 Okay welcome everyone again it's a real pleasure to be here I will add so I'm probably one of the few here that was born in Trieste and grew up in Trieste and my love for the city actually will be apparent throughout my talk a little bit I was trained as a physicist I left Trieste as a particle physicist and gradually transformed into a biologist or some mixture so I'm gonna tell you as the first lecture of this workshop and it's really just amazing to see the diversity of geography and scientific background here and a lot of colleagues that inspire me throughout the years are here so this is really wonderful I'm gonna give you an overview of some of the concepts that I think are interesting in thinking about microbial communities with examples from my own work and touching upon things that we'll hear about in the next few days so I'll start by showing this slide that to me represents some of the reasons I'm interested in I was attracted by microbial communities and two examples of how microbial communities affect human life and we are all very familiar with this this represents let's see oops the effect of microbes on our own health and I like to think of the little people around Gulliver as being helpers and I think that's how we think now about most of the human microbiome the other side which is perhaps less well-known is the fact that microbes affect humans at the planetary scale and this is phytoplankton bloom of the coast of France and what is just amazing to me is that this little entities can have a huge impact on the metabolism and microbial composition of the biosphere and in fact microbes like this ones are the ones that gave rise to the oxygen that made multicellular life possible about two billion years ago so I think it's still very interesting to to understand how microbes affect planetary scale metabolism now I as you all know metabolism and microbial metabolism is a multi-scale problem so we can think of microbes as from the perspective of what I'm gonna tell you about as bags of enzymes where each organism has an internal network and and it's not clear whether in trying to understand how microbes interact with each other for example through cross feeding or competition or antibiotics fighting whether we can understand what happens between different microbes by understanding what happens inside them I think that's one of the big challenges but one could imagine looking at microbes as entities that are represented by one single variable looking at lot of altera type equations that describe each microbes as as an entity without looking at what happens inside and at the other extreme and I'll mention this briefly later one could think of whether it matters which enzymes which molecules and networks are present in each organism or whether one can study a whole community as a big soup of enzymes so does compartmentalization matter so I like to think of these different scales and different problems as really spanning multiple hierarchies and multiple level of descriptions and I think this is one of the exciting challenges of studying microbial communities so let's start with the challenge of trying to figure out whether the network of interactions between microbes can be somehow inferred or understood in terms of the interactions and networks that are present inside each individual organism why specifically what we want to do this there are multiple reasons these are some of the specific questions were interested in my group at Boston University we want to try and and see whether it's possible to really understand the dynamics of these communities based on the metabolic networks present inside each organism can metabolize help explain the stability or instability of microbial communities the unculture ability of many strains perhaps through metabolic interdependencies and the diversity of microbial ecosystems the other side of this is an engineering side many people here and in the world are interested in synthetic microbial communities can we engineer and can we understand microbial communities well enough to be able to engineer them for specific functions for example shifting the balance between a disease associated microbiome and a healthy one or making communities that can produce something useful like such as biofuel molecules so when you start thinking about metabolism and modeling metabolism this is the problem you face this is the network that you see often the chart that you see hanging in many labs of all known metabolic reactions across all organisms and this seems and is indeed a difficult task and this is where an famous Italian writer from Trieste helps us so this is a quote from Italo Svevo one of one of the most famous writers from Trieste this is a quote from Zeno's conscience a piece of literature that is famous in Italy and interested in particular as one of the key writings in the early 1900 and if you walk around Trieste you'll see the statue of Svevo this is in the heart of the old city next to the library and not far from the old port so it's a beautiful place to visit so this quote says I'll read just a English version one cannot expect a chemist to know the world by heart and I can't go into the context of why Zeno was saying this in the in novel but why this is important to us because indeed we can't hope to keep in mind all the metabolic network and we can represent this especially as we try to do quantitative modeling we can represent this network as a matrix and and build pipelines from databases such as keg that includes all known reactions in metabolism through the information content of genome so we can see which of these reactions are present in the genome of each in which of organism and we can build a network that is specific for a specific microbe in this case this is the metabolic network of E. coli where each link represent a reaction each node represent a metabolite and this counts of the order of 1500 reactions approximately 1000 metabolites and one can represent this with what is called a stoichiometric matrix S ij representing the number of moles of molecules of molecule I that participate in reaction j so this makes it possible to do to represent this complex network into a mathematically tractable form and gives rise to a lot of methods some of which I'll mention today so if anybody that has not familiar with this and it is interested there are there are a lot of these networks available for example through the department of energy k-base a database of reactions or on this website from Bernard Paulson at UCSD that has tens and now increasingly more hundreds of reconstructed networks for specific organisms for which one can do trying to detail modeling so this was a little bit mistranslator from Mac to PC but this shows a little bit of the challenge that we have so once you have a network like this for a specific organism this is not the end of the story it's the beginning of the story what we would like to do is be able to predict what are the fluxes what are the rates of each of these reactions and for example given an environmental condition nutrients available in the environment how will this organ is grow what is how there are resources allocated in this network so that the organ is can grow and what could be potentially secreted by the by the organism affecting other species now if you try and do modeling of a whole network of a microbe using standard kinetic modeling this is a very difficult problem you'd have to write differential equations for these 1500 reactions and have a lot of kinetic parameters this traditional Michaelis Menten parameters for all of these complex reactions that are typically unknown so this is a very very difficult problem and one of the oops okay one of the ways around this which is what is the method called flux balance analysis allows us to do models of a whole cell metabolic network in a more efficient less precise but more efficient and very valuable way especially as you will see from the perspective of modeling communities so this is again the same network but now instead of thinking of this as a dynamical system we think of this as a resource allocation problem really where there are nutrients coming in the cell needs to produce amino acids nucleotides cofactors and so on in a very precise set of proportions shown here and these proportions are known from experimental measurements so we know how much of each amino acid is needed to build a new cell and now you can think of metabolism as this big factory where out of glucose and nitrogen coming in and so on the cell has to decide how much ATP to produce how much building blocks to produce so that construction of the biomass can be made in an efficient way and this becomes now a more tractable problem and I will give you a very quick idea of how this can be done this is the in one my one slight version of flux balance analysis some of you may be familiar with this for many with this may be the first time I will just I want to give you a flavor of what this is all about we focus here on one individual reaction one key simplification that makes it possible to model metabolism at this large scale is to take a steady-state approximations we assume that this the steady-state that the amount of each metabolite is is constant so the cell is at steady-state not in a certain equilibrium a dynamical steady-state where the molecules the reactions producing this metabolite in this case glucose 6-phosphate are balanced exactly by the reactions consuming this metabolite so there is a net zero flux out of the system and now what is interesting about this type of approach is that instead of dealing with the concentration of metabolites as a function of time you deal with the rates of the reaction the fluxes and they are related to each other by the simple linear relationship that just imply the conservation of the metabolite or what is called mass balance so we have simple linear equations in the fluxes and you have one such equation of conservation for each metabolite in the network it turns out that this this simplifies the problem mathematically so much that you can really use very efficient tools to end up predicting each of these fluxes for a large number of reactions in in a metabolic network of a microbe now there are additional constraints here for example availability of nutrients from the external environment so you can model the specific environment under which an organism is grown for example how much glucose is available from the environment you can model known irreversibility of certain reactions and it turns out that all these constraints are linear constraints that allow one to solve this problem through linear optimization and I'm not gonna go into any more detail but you can for example ask what is the set of rates of reactions given the environmental constraints that will allow the cell to grow in the most efficient way maximizing the rate of production of its biomass and you can in a fraction of a second find both the growth rate of a cell under these conditions and all the reaction rates under these conditions yes molecular the composition yeah so yeah so here are not considering the molecular structure except for this stoichiometry so you do consider the stoichiometry the specific reactions that take place in E. coli so in to some extent obviously for each chemical reaction you conserve atomic composition and so on but you don't look at you don't take into account the molecular structure in detail in in the network that's a very interesting question I think I can point to some literature people started looking at motifs in metabolism but I don't think that that it's so well studied partly I think because they're well known pathway structure so we yeah there are a lot of linear pathways there are some typical sets of reactions but I think it's an interesting question I don't know that anybody has studied let's say motifs in the same way as we've been studied for transcriptional networks for metabolic networks and yes yeah so there here you really just study the steady state you cannot look at the transient I'll mention a little bit some pseudo dynamics shortly and how one can study partially the dynamics of the system as well that's right so why well when you do there the reason this method is so efficient is because you do the steady-state approximation so you really get rid of the concentrations completely and you study only the steady states and your variables become just the fluxes there are no more concentrations okay I won't go into the detail here just this is just meant to show that their flux balance has some good aspects to it it can predict under certain conditions match the predictions can match experimental fluxes but not always and it's an ongoing process so you shouldn't think of flux balance analysis is a perfect tool for predicting microbial metabolism but as a way of predict making hypotheses that are testable it has certain advantages fast computation it's scalable there is no need for kinetic parameters and most importantly we care about fluxes we're interested in the rates of reactions and the cell probably cares about fluxes so this is why it's important to think about it fluxes in this way but it has many limitations for example you cannot look at metabolite concentrations there is no dynamics you look at population time average and there are many still many gene functions that are unknown which limits our capacity to do wholesale modeling in general but let me get back to communities and think of how can we use these models of metabolism in single-organism to understand exchange in communities and so here we have another adopted Trieste rider in helping us so some of you may not know James Joyce lived for 10 years in Trieste and if you walk across one of the channels walking towards the sea you'll see his statue and I'm putting here a sentence to live to her to fall to triumph to recreate life out of life from a portrait of the artist as a young man what is interesting about this it kind of hints to synthetic biology and in fact this sentence was encoded in Craig's Ventures first mycoplasma synthetic genome and it sparked a lot of interesting discussions later on both because there were some copyright issues but also at the same time the copyright issues were interestingly kind of made irrelevant by evolution because their mutations were quickly changing the sentence so it's an interesting issue but for us what is important here is to mention that in the same way as you can think of synthetic biology as a way of reconstructing and tinkering with microbes you can also think of a synthetic biology as a way of tinkering with communities and generating communities and understanding their function so this is an example of work when we need Shu who is here today and we'll talk in a few days and she and collaborators a few years ago created one of the first examples of a synchrophic pairs of organisms so these are two yeast that you were transformed into oxotrophic we're not able each to produce a certain molecule so they were only able to grow in presence of each other so by tinkering the internal network somehow you could induce a synchrophic interaction and obligate cross-feeding between these two microbes now as we were starting thinking of how to use models to study communities we realized that one interesting twist to this idea was to engineer the environment instead of engineering the microbes so with a former student in my lab Nils Clitcord we thought how about taking two organisms two microbes as they are without changing their internal network but asking whether by providing the right mixture of nutrients we could induce an obligate interaction between the two organisms so is it possible to engineer the environment so that two organisms are necessary have to rely on each other in order to survive so the way this is done in flux balance analysis is by building multi-compartment models so you can imagine two organisms being sub-compartments in a larger compartment that is the environment and now two organisms can exchange nutrients and as shown in this toy model these organisms this toy organism that need A, B and C as molecules in order to produce biomass they wouldn't be able to grow on nutrient A unless they're in presence of each other where they can exchange B and C respectively now the way this is done in flux balance modeling is by having labels for specific metabolites in each compartment the details don't matter but this was the first instance from David Stahl's lab of a microbial community model based on flux balance analysis now one can use this to ask the following very simple question if you take two organisms and find a set of environments that can support the two organisms together so you take these two organisms and you ask what is a set of nutrients for example glucose ammonia and sulfur source and so on and so forth that will enable the organism one and two to grow in co-culture and you could find it turns out typically millions of different environments for a number of pairs you can choose and all of this is computational work and now you can ask given that these two organisms can grow together in this environment you can ask whether each organism can grow on its own and you can have one of these four situations one possibility is that each organism is also able to grow on a given environment on a given set of nutrients in this case this is kind of a neutral interaction but it's possible that on a given environment where the two species can grow one species is able to grow also on its own the other cannot so this would mean that that environment induces a one-directional interaction where one provides an essential molecule to two and the most interesting case perhaps is the case where the two organisms can grow together but none of them can grow on their own and in this case the interaction would be this type of cross-feeding interaction so one could computation into this very easily you can screen many many different environments and try to find environments that would induce this type of interaction so what Niels in the lab did a few years ago was building a matrix of this possible interaction between different species you see here the name of the organisms are a variety of microbes partly from human-associated partly environmental associated but what is crucial here is that for each pair of organ you say E. coli and Salmonella you can see the pie chart represents the number of metabolites that are the number of environments that could support growth of E. coli and Salmonella together and the portion of this the green portion of these environments is the one that would induce neutral interaction that is this would be a minimal medium that is common to E. coli and Salmonella and as you could imagine E. coli and Salmonella being very similar there are many such environments but what is most interesting the yellow portion of this pie chart represents environments that induce an obligate syntrophic interaction between the two organisms so anytime you see a yellow portion here this means for example B. Sautilis B. Sautilis and Salmonella there are several of the order of hundreds of thousands of environments such that on those environments the prediction prediction would be that B. Sautilis and Salmonella would need each other in order to survive so what was striking about this was that somehow based on the stoichiometry of networks of metabolic networks across organisms there seems to be a lot of opportunities for cross feeding in the microbial world so there is a lot of possibilities for metabolites that could be secreted by one organism and used by another for growth and the question is does this represent what happens in reality how to probe this further both computationally in more detail and experimentally I'm not keeping track of time is someone or you I don't know how much so I will tell you a little bit about how we approach this in a more detailed way through another method that is an expansion of flux balance analysis and it's called dynamic flux balance analysis and gave rise to what we call comets computation of microbial ecosystem in time and space which is a way of incorporating genome scale models into a spatial framework to model more realistically microbial metabolism in communities and this is an effort sparked by Bill Reel a former student in my lab and I developed them together with Will Harcombe and Chris Marx were here and Pankaj Mehta who's also here so this is still an ongoing process but it was very exciting to be able to come together with microbiologists to be able to test some of these predictions but the main concept here I want to highlight is that it's possible to make a little bit less assumption in particular you don't need to make any assumption about the existence of an interaction in order to see one emerging from simulations and the idea is the following you can model in a certain region in space you can model flux balance for a specific organism as shown before and you can transform the concentration of metabolites in the medium into an uptake rate very similar using something that is basically a Michaelis Menten equations you can determine the rate of uptake of metabolites based on concentrations and solve different time steps this flux balance problem so you can get an instantaneous rate of growth for a given organism at a given time and measure the rate of consumption of metabolites and by iterating this process in different time step you can obtain an approximation of the growth curve in an approximation of the abundance of metabolites in the medium so this is still not bypassing the problem of intracellar metabolite concentration so you monitor only extracellar metabolite concentration environmental concentration metabolites and the growth of individual species so nowadays you'll find a lot of people are using this dynamic flux balance methods to model microbial communities because you can imagine modeling a specific organism certain region of space that organisms can grow based in on its own rules produce a certain metabolite that metabolite could diffuse away another organism potentially in the same region or in a different region could detect the presence of that metabolite and use it for growth maximizing its own capacity to to produce biomass but there is no a priori assumption of these two organism having to interact so if there is an interaction across feeding this will be the outcome of each organ is trying to do what it's best for for itself and now by embedding this into a spatial temporal the spatial temporal setting you can make models of microbial communities that look a little bit like this so there we like to think of these are as virtual petri dishes where you can inoculate different species let them grow and see what happens and there are a number of applications of this which I'll mention soon I will just quick quickly go through some of the validation of this will my talk about this more in the next few days but this is really work that will Harkom perform testing predictions for mutualistic pair of an E. coli and a Salmonella that depend on each other and again I imagine you'll hear more for will on this but this was the first test of comets showing that you can really make a reasonable predictions of growth for this organism I'll skip this this is an example again from Chris Mars and will Harkom's work on a three species artificial community again showing that these models do a reasonable job and actually quite surprisingly good job at predicting the abundance of species at different time points given their metabolic network model so I will just conclude the part on comets just by saying that there are more things to add to this for example one can model growth as a convection diffusion problem and make diffusion dependent on the materials so for example if you're modeling the growth of a colony diffusion of oxygen through the colony itself will affect the way the colony grows and this is something that can be taken into account and will have important consequences so for anybody that is interested someone already asked me earlier this morning comments in open source and I'll be happy to discuss with people throughout the week so I want to I'm skipping a few things because I want to get to some other final concepts so one of the key questions right we we can use these models to try and understand simple toy communities but one of the big challenges is whether we can model large national community so this was one first example from prior work on trying to model a gut microbiome toy community where you can see each organ is represented is it in its own network but what is interesting here is I can you see the emergence of an ecological network whose nodes are the organism themselves and the molecules that are being exchanged between different species so that would be one of the challenges and interesting questions was can we predict this ecology and the molecular exchange between different microbes in a natural community and here there is another poet actually a poet from Tresa that helps us think about this this is Umberto Saba whom you'll find next to Corso Italia this is another beautiful area of Tresa next to the ancient theater Roman theater and Saba was a poet and one of his verses that I borrowed from this has how beautiful my city must have been back then a big open market and I think this is a good way I think of thinking about microbial communities as a big open market of molecules where organism secret molecules that can then be used by other molecules and I think as I'm approaching the time I have I will just really quickly mention two ways of thinking about this open market one is recent work by a postdoc in my lab Aliso Morotti largely inspired by Jeff Gore's work so one can think of microbes as potentially leaking metabolites and there is a lot of interesting work on what has been called the Black Queen hypothesis of how these exchanges could happen given that production of the molecules is often costly so one can study the chances of emergence of an interactions given the cost of these metabolites and the fact that producing a certain metabolite will reduce the fitness of the producer to the advantage of a cheater that will give up producing that metabolite because it's available in the environment and one can study this type of dynamics also for pairs of organisms for example leakiness of two amino acids in E. coli and try and understand under what conditions the emergence of a mutual cross feeding could emerge so and what is interesting is that one can use game theory and compute Nash equilibrium for the system to estimate under what conditions what are the leakiness levels that would support the emergence of these communities so again I won't have able to go into this now but I'll be happy to chat about this and I know some of the talks in the next few days will touch upon game theory or similar concepts the last thing about this open market is I just want to mention quickly something about this possibility that perhaps in these complex communities maybe compartmentalization matters only to some extent so can we think of communities as soups of enzymes does it make sense so one hint to this is that one can look at the whole metabolic network of across all organism and irrespective of who does what function we can ask what is the capability of an ecosystem you can think of the metabolism of the ecosystem as a whole without taking into account what reactions are present in what organism and one example of how this can be useful is using what is called the network expansion algorithm that was developed by Oliver Ebenhagen collaborators a few years ago where the idea is very simple you can have in this toy network you can start with a seed of metabolites and ask what molecules are producible by this network in this case for example you can only produce these two final molecules and you're stuck you cannot do anything more so this is the scope of the network if you have this additional metabolite in the seed you can produce these two reactions you can take this example to the to the large network and ask given a set of initial molecules for example environmentally available molecules what is producible in principle by this network and this is a very simple algorithm we actually apply this algorithm to study early life recently in my lab Josh Gold for a student my work in my lab asked the question of how far can you go in a puritive primordial metabolism where phosphate may not been available yet and this is a very interesting question from the origin of life perspective because phosphate is poorly available in and is very difficult to get out of rocks so the question is how easy it is to obtain or can you obtain any network from this initial molecule and again making a long story short the result of this was that you can obtain a fairly large unexpectedly large network of metabolites from this initial seed showing that somehow there is phosphate independent subnetwork hidden in current metabolism which has again implication for early metabolism but also potential implication for natural communities nowadays in fact we know that many communities can use sulfolipids instead of phospholipids under phosphorous limited conditions and I will stop here and just actually with this last thought that this idea of a soup of enzymes as a way of modeling communities has to I think temptation so there is the temptation that is practical given that we have a lot of metagenomic sequencing data can we use this data directly to model metabolism in communities without going through the construction of individual organism and the other more interesting perhaps it an intellectual temptation and that is the question of whether really you know does does it or doesn't it matter that enzymes are enclosed in specific organism or could you think really as the main driving force of ecosystem metabolism as really collective phenomenon of just reactions together without taking into account where they belong to and I you know in my thinking I go back and forth between thinking that yes compartmentization matter or no maybe not but I think this is an interesting topic for future discussion and I will stop here and thank my students and postdocs and collaborators and funding sources and all of you for listening