 and he's going to tell us about something in Pseudomonas aeruginosa, I know. Yes, I'd like to thank Petorio and the other organizers for the invite. It's a pleasure to be here. I am a microbiologist who tries to learn about quantitative approaches that I'm trying to incorporate into my talk, but I would have never thought that as a biologist I would set foot into a center of theoretical physics, but here we are. So first a quick overview of what my lab has been up to at Oregon State for the last 10, 11 years. We are basically interested in how and why bacteria, bacterial populations communicate and cooperate. We basically have two model systems. One that has dominated in my lab is acetylhomocerein lactone quorum sensing Pseudomonas aeruginosa and you've already heard about that from Pete and others. The second project that was inspired by my sabbatical a couple of years ago and that I've worked on more recently is iron acquisition via pioverdin sederophores in Pseudomonas. And I guess where my lab finds the niche is in sort of integrating questions about the how and the why, the proximate and the ultimate answers to understanding cooperation. And in particular we're interested in identifying mechanisms that help stabilize cooperation. So I'm going to talk about iron acquisition in Pseudomonas today. Iron is often acquired from the environment via sederophores that are secreted under iron limiting conditions. They're secreted into the extracellular medium where they bind iron with extremely high affinity, 10 to the minus 30 molar or so. Then they bind to a cognate receptor where they are internalized. The iron is unloaded and then often the sederophore is recycled. These are small secreted high affinity key letters and they are necessary in the environment because extracellular iron is often low and the host that is bound by, chelated by host proteins and in aerobic environments iron is often in its ferric insoluble form. Pseudomonas secreted pioverdins among other iron chelaters among other sederophores. These are small peptides that are non-ribosomally made and it's a convenient system that is amenable to experimental manipulation and investigation because pioverdin is yellow so you can easily identify a pioverdin producer from a non-producer by its color and pioverdin is also fluorescent so you can easily quantify this trait in growing cultures. There's a great diversity, tremendous diversity of pioverdin in pseudomonas species. Over 70 different structures have been identified and that's also an interesting question that we pursue is what drives this diversity although that's not a question I will deal with today. So pioverdin is a secreted product. It is a public good that is useful for cooperation in bacterial populations because a secreted product not only benefits the focal producer but it benefits other cells in the environment and it's commonly assumed that secretion is costly. These non-ribosomal peptides cost a lot of ATP to produce and non-producers that we call cheaters should spread. One basic question that we want to answer is that this model system is useful for some basic physiological insight into the cost of secretion. How costly is secretion actually? Is it costly at all times? And how does the environment and especially the nutritional environment impact costs and hence the stability of cooperation? Why do we care about nutrition? Well I guess for bacteria almost everything is about nutrition and it's about nutrient limitation because essentially every microbial growth environment is nutrient limited be it the host, be it marine, terrestrial, aquatic environments. As we all know nutrient availability, nutrient limitation has a tremendous impact on gene expression on the phenotype of the microbe. So to get a first pass at this question how costly is pioverdin production? We looked at the enzymology. This is a graphic from a review and it basically outlines the major players here at the center of the pioverdin biosynthesis pathway or machinery is the non-ribosomal peptide synthetases. These are huge enzymes. They have a modular build. They have three domains, an adenylation domain, a thylation domain and a condensation domain. Just like on a conveyor belt you have amino acids that are being incorporated into this peptide one at a time. This is costly. Each individual amino acid costs two high energy phosphates. Then if you do the whole calculation including precursor reactions, maturation and experts we estimate that about 26 high energy phosphates per pioverdin molecule are required to make it. Considering how much pioverdin is made in cultures of pseudomonas we estimate that about 15% of the total ATP that the cell needs is devoted to secretion of pioverdin. At first pass that's a huge metabolic burden. 15, 15. Next we did some in silico modeling of how the nutritional environment affects these fitness costs. We used a whole genome metabolic model that Jason Papens group at the University of Virginia had built several years ago. It did not contain the pioverdin reaction so we added them to this model. Then we did flux balance analysis basically asking how an objective function, in this case growth, that's the objective I guess most bacteria have, could grow as fast as possible, how an objective function growth rate changes with increasing pioverdin concentration. When different nutrients are limiting so we chose carbon, nitrogen, phosphorus, iron and sulfur. And you can see with carbon and nitrogen there is an immediate trade-off in resource allocation between cellular biomass accumulation i.e. growth rate or pioverdin production. As pioverdin production increases growth rate decreases. With these other nutrients phosphorus, iron and sulfur it's like the bacterium doesn't care. It grows at the same rate until the threshold is crossed when carbon again becomes limiting which in this case is set to the maximally experimentally observed uptake rate. And then there's so much pioverdin produced that carbon becomes limiting and then there is this trade-off just as we see with carbon limitation in the first place. So we can reconcile these observations when we consider the structure and particularly the molecular composition of pioverdin. So fitness costs then that is the trade-off between growth and secretion or the reduction in growth rate is a function of nutrient and public good composition. So this public good the pioverdin is as a peptide as I said so it is carbon and nitrogen rich. And when carbon and nitrogen are limiting and you have this trade-off then there's a fitness cost but not when other nutrients are limiting. So we can, yes? So this does not depend on the concentration of iron in the environment? That's right. We always said in these cases we always said one nutrient to be limiting is we limit the uptake rate and we set the other uptake rates to either be open-ended to either be just as high as can be or to experimentally determine maximal values. Yes, in this case there is no regulation built into the system. Yes, that's a good point. We're working on a dynamic model where the regulation is actually part of the whole story. Very good point. So based on these fitness cost considerations then we can make inferences about the evolutionary stability and cooperate or cheat the dynamics because if there is a fitness cost then cheaters should invade a cooperating population if there's no fitness cost then cheaters should not invade. So we tested this idea with a chemostat system and it's been a couple of talks that used the chemostat or explained the chemostat. Just go through it briefly. So it's basically a bioreactor, a vessel that allows you to grow bacteria at a constant cell density and at a constant growth rate. So they're basically at steady state. The growth rate is determined by the dilution rate of the inflowing nutrient and then there's also waste and spent media and bacteria that flow out. The system is well aerated and well mixed. So what we use here as the growth medium is a synthetic minimal medium that contains chelated iron so it requires Siderophore production for growth. A mutant could not grow, a Siderophore production mutant could not grow in the system by itself and we have either glucose or phosphate as either carbon or phosphorus as the growth rate limiting nutrient in the system. And we initiate about one-to-one co-cultures of a pyroverdent producer that is the wild type and the pyroverdent mutant that in this case is a sigma factor mutant PVDS that sigma factor directs the biosynthesis of pyroverdent. I guess I should say some of you may be surprised that the sort of poor man's built poor man's chemostat that is in fact intentional. We tried first high-tech stainless steel bioreactor but we could never get iron limited conditions obviously. So this is our rigged design that I actually got from Pete years ago. I guess we modified it a little bit. So as I said we had one-to-one competitions between a producer and a non-producer and we assessed relative fitness by measuring the frequency of the population at the beginning and at the end of the growth period that is after six days of continuous culture we calculated a relative fitness of the mutant versus the wild type and the relative fitness is basically the ratio of Malthusian parameters that is the ratio of the average growth rates and as predicted under carbon limitation the mutant invades. Its fitness is significantly higher than one means no invasion whereas under phosphorus limitation the mutant does not invade. It does just as well as the wild type. Now this effect is not due to simply different expression levels of pyroverdin. Under carbon and under phosphorus limitation both expression levels are indistinguishable. So to take a closer look at the dynamics under carbon limited conditions we indeed see that in Oculum was not quite 50-50 at the beginning but almost the pyroverdin biosynthesis mutant readily invades the population at the expense of the total pyroverdin level. Of course there's fewer and fewer wild type in the system that contribute to pyroverdin production and despite of this invasion there's no decrease in the total culture density so that means that the wild type whatever pyroverdin it produces is far more than is necessary. The pyroverdin is clearly not limiting in the system. Yes please. Spatial structure is also modeled in the pyroverdin system and it clearly has a role but we just were interested in the theme of being reductionists. We only wanted to model nutritional effects not spatial structure. Spatial structure has a tremendous effect. That is true and that's been done by just looking at bacteria on colonies basically where you introduce this. Well viscosity of the medium has been altered. That's been done too, obviously. The molecule could be modified in a way and there's also the idea that a partial privatization of products that Jeff Gore has worked on, yeast and vertase results in a partial private, actually large, no. That's the key question that I was one of the key questions here that I was going to get at. So there seems to be an equilibrium and we have additional data where we start with higher proportions of cheaters where we find that there's never extinction, that there's this apparent equilibrium. Now we don't know currently what the reason for that is but our favorite explanation is that there is an automatic switch from carbon limited to de facto iron limited because as I said pyrovertein levels decrease here dramatically and so through pyrovertein you indirectly get iron limitation in the medium and so this switch to iron limitation could mean that now there is no cost to cooperation anymore and therefore it doesn't pay to cheat. So you basically end up with this equilibrium. So then in summary what we've shown that if the building block of the public good is limiting indicated in this case by these orange dots here then you have this tradeoff between resource allocation to either growth or secretion. Growth is slowed and cheaters can invade. If another nutrient is limiting now the blue dots that are not part of the public good that are not incorporated into the CDERA4 or other secretions growth rate is as high as it can be, as high as cheaters and so there's no invasion. So in principle cooperation is stable. Now what about iron limitation? After all CDERA4s are produced when iron is limiting. So there are two cases that we haven't talked about yet. Now when iron is abundant this of course is the trivial case then you don't need CDERA4 so you don't have a social situation. When iron is actually growth rate limiting then as I already alluded to you have a situation where you have no cost, the iron is scavenged and when you have no cost then two cheaters cannot invade and this sort of regulation scheme is akin to what's been described as metabolically prudent regulation in Pseudomonas and specifically Ramnolipids that's a bios effect that Pseudomonas produced in a very nice paper by Jao Javier and Kevin Foster. And these Ramnolipids too are only made when nitrogen but not carbon is limiting. Ramnolipids are carbon rich so they're only made when it doesn't cost anything. So then in conclusion what we've shown is that the costs of cooperation are highly context dependent and particularly the chemical composition of the public good and the nutrient source determine the evolutionary stability of cooperation and this provides we think a nice framework to consider the presence of secretion negative isolates and infections. Why are they there and can we make predictions of when they should arise in certain environments that are nutrient limited one way or another. So now I'd like to come back to the situation that Jeff alluded to namely the fact that we never saw a population collapse or a tragedy of the commons in this chemostat which would manifest in just wash out of the population. From cheaters emerging we certainly think we have a reasonable explanation it's a switch in nutrient limitation although there are other potential explanations for why there's a stable equilibrium. In this case nevertheless conventional wisdom tells us or what we are told in various publications cheaters exploit cooperators to the point of population collapse. The tragedy of the commons should be widespread but actually when you look in the literature it isn't. So a tragedy of the commons is not very common in the microbial systems that have been investigated. Of course the first case that demonstrated cheating and Greg Velliser's group was in fruiting bodies of myxocococcentus analogous to what we've heard about Dicty yesterday. So you have these fruiting bodies that consist of spurangium and stock and these cheaters when they emerge they preferentially form spores they don't like to be altruistic stock cells and so they enrich in the spurangium until the whole system collapses because there isn't enough cells left to build the stock. Now a similar situation in pseudomonas fluorescence where you have pellicles that form at the air-liquid interface as you see here and Paul Rainey's lab now at New Zealand has worked on this and shown that cheating too can lead to extinction can lead to collapse of this pellicle because pellicle formation requires the production of an exopolysaccharide but cheaters that don't produce this exopolysaccharide put so much weight on the pellicle that it eventually sinks to the bottom of the ocean floor or actually the test tube. So in search of evidence in the literature does a tragedy of the commons population collapse happen in a chemostat you couldn't find any there was nothing maybe some of you guys know and in general we couldn't find any mathematical proof that a tragedy of the commons really happens now there is game theoretical literature that suggests that cheaters should always win but typically that doesn't feed back on the population productivity Am I correct? Okay, so again this is sort of new territory and I've collaborated with a mathematician at Oregon State and two other mathematicians at Arizona State to actually prove the tragedy of the commons that in principle it should occur in the chemostat when the public good is always costly it is fully shared, it's not partially privatized and there is a well-mixed environment. So we have a cooperator and we have a cheater here and in this case they actually use an enzyme therefore they secrete an enzyme as the public good that converts the substrate into a product and then that product is the growth rate limiting nutrient at the same time both are taken or cooperators and cheaters take up this product at the same rate the cheater incorporates it all into biomass but then in the in the cooperator there is this trade-off between either growth or secretion. Now when we start with the cooperator only so only this guy in the chemostat not this one we have an initial phase of growth the red guy is the is the cooperator from a low inoculum and so you have a commentant increase in enzyme production in product formation and the depletion of substrate of course but slowly thereafter you have an equilibrium forming and this equilibrium of course depends on initial conditions and parameter values but now when we introduce the cheater here we have the following situation we have the cheater at very low frequency it enriches at the expense of the cooperator but only to a certain degree at some point there is not enough cooperator in the system to sustain the cheater and both go extinct and actually this is very robust this always happens no matter what the initial conditions or what the parameter values are whether you have a citrofoam or an enzyme and what not so the only thing is that the cooperator pays a cost that the cheater does not so this seems biologically intuitive certainly to many of us but it was apparently not mathematically trivial as Patrick tells me and even though it may be biologically intuitive we did actually a little poll non-representative poll with colleagues some were mathematicians and some were engineers to ask them so what do you think should be the outcome of the system would you always expect a population collapse and not a single one said yes that's what you would expect okay so anyhow the next step is a little bit of deviation from the within species cheating now we go to between between species this is an interaction that you may call cheating but it's commonly called crossfeeding in the literature so following on the theme that we've heard this morning excuse me so this is the bacterium that we've already introduced to you pseudomonas ergenosa that produces a cedarophore that it can bind, it can utilize itself but of course there are other bacteria other species like this pseudomonas prodigens that cross feeds because it produces a matching receptor oftentimes these cross feeders also produce their own cedarophore another question that we wanted to ask here how is pyrovertein production and reception regulated when this species is presented with the opportunity to cross feed would it then forego producing its own cedarophore because it has the opportunity to cheat basically and how does this regulation affect relative fitness so we chose a model system here the consummate cross feeder I would say, pseudomonas prodigens that a colleague at Oregon State at Joyce Loper has worked on and it's done beautiful experimental work to identify the ferre pyrovertein receptors so there's five heterologous receptors out of 45 ton B dependent auto membrane proteins so all these auto membrane proteins could potentially bind cedarophores cross feed they may bind other small molecules as ton B dependent proteins have demonstrated to do but we have these five heterologous ferre pyrovertein receptors one, two, three, four, five and the matching bacterial cedarophore and the producing species have also been identified and of course the six cedarophore here is the one that binds the homologous or the self-produced cedarophore so with the system we first asked is the receptor up-regulated only when presented with the matching cedarophore we did this by real-time PCR we looked at the suite of six receptors and the corresponding cedarophore there's always the receptor and the cedarophore in a matching color and as you can clearly see in most cases the receptor is only expressed when the cedarophore is present in the supernatant and I should add we have not used purified pyrovertein here we've used supernatant but in previous studies we used purified pyrovertein equivalent so even when you add two cedarophores two matching receptors are expressed there are some exceptions in this case BN7 even though we know that the brown receptor is required for the use of this brown cedarophore it is not induced we surmise therefore that expression is already constitutive and in the case of B10 here the green receptor is indeed induced in the presence of its matching cedarophore but so are many others and this is probably a case of structural promiscuity where receptors can bind multiple cedarophores in other words this receptor, this cedarophore can bind to multiple receptors so next we turned our attention to the expression of the endogenous pyrovertein produced by PF5 so does PF5 still produce its cedarophore its pyrovertein when presented with the opportunity to cross feed so we looked at three different strains here, the wild type and the two functionally equivalent pyrovertein mutants we looked at growth in culture we looked at fluorescence fluorescence excitation basically as a measurement of pyrovertein production so in the absence or in the presence of these various pyroverteins from the different species so when there is no pyrovertein provided then you find that only the wild type can grow the two mutants don't and you find that of course the producing strain produces cedarophore because it is required in this growth medium when you add the respective cedarophores you find that all strains can grow at the wild type level so the mutants are restored to the wild type level so the amount of cedarophore that we've added is fully sufficient and nevertheless as the fluorescence spectra show they still make their own cedarophore so basically here shaded in gray is the contribution of the fluorescence from the cedarophore that's added that doesn't change over time it's always constant but then above that you see the contribution of the endogenously produced cedarophore so even when given the opportunity to cross feed this strain still produces its own cedarophore so to look into this to sort of identify the reason why they might be doing this we used a co-culture system of pseudomonas prodigens with pseudomonas aeruginosa so pseudomonas aeruginosa produces pioverdin that it itself can use but that pseudomonas prodigens pf5 can also use but then in turn prodigens only produces cedarophore that it itself can use there's no cross feeding so there is in this model you have a unidirectional cross feeding and we verified that this is indeed the case with synthesis mutants so the pioverdin synthesis mutant requires its own cedarophore whereas pf5 can grow on its own cedarophore or that provided by pio1 the two the red and the green are virtually overlapping here so then we did co-culturing experiments to determine the fitness of pf5 relative to pio1 again 50-50 co-culturing experiments and you find when you put the two together without any additional pioverdin or supernatant just endogenously produced by both species you find that pf5 does a little bit better than pio1 when you take away the ability to produce pioverdin by pio1 then it's relative fitness drops greatly it can be chemically complemented by adding more than that by adding a supernatant now to further elucidate what's going on we looked at how pioverdin amendment affects the growth rate of pio1 and to pio1 pure cultures we either added no supernatant or supernatant containing pioverdin from pf5 or supernatant devoid of the pf5 pioverdin and we find that indeed pf5 greatly reduces the growth rate of pio1 in the presence or through probably sequestration of iron so in conclusion then we can say that heterologous pioverdins induce expression of the cognate receptors made to order so to speak and self-produced pioverdin is expressed during cross feeding and provides a competitive advantage so finally I'd like to acknowledge people that did the work so what I've presented today was Incidero4 cooperation so there's a graduate student who just graduated Joe Sexton together with an undergrad Amanda Singh and also the people that I've collaborated with during my sabbatical Katja Bettenbrock and Stefan Klumpt at the Max Planck Institute Markterburg they helped basically get the metabolic model rolling and helped a little bit with chemostat stuff and Radhakrishnan Mahadevan from the University of Toronto is also a metabolic modeler who helped with this and Joyce Loper helped with expertise on pf5 and provided reagents and we also have another project in the lab that revolves around quorum sensing in there I have a student that also just graduated Kyle Asfall and a new graduate student Tana Robinson I think funding agencies and thank you for your attention