Research Evolution of multicellularity, bet hedging and microbial cooperation Spatial dynamics of microbial cooperation and conflict Why killing your neighbors can structure microbial communities and lead to the evolution of cooperation About 25% of gram negative bacteria kill each other with poison-tipped harpoons (the Type 6 secretory system). It turns out that when two well-mixed strains of bacteria kill each other, they form neat physical patterns by driving a type of phase separation never before seen in biology (a 'Model A order-disorder transition). Now, this is pretty cool in its own right, but it gets better. This phase separation (which works even in dense surface-attached populations with no net growth!) creates precisely the conditions conductive to the evolution of public-goods cooperation. So, does T6-mediated killing explain cooperation in the real-world? We looked for this bioinformatically, creating a phylogeny of >400 bacteria that contain at least one T6 system. We make a few simple assumptions: 1) having more ways of killing neighbors should create more structured populations-that is, if a strain only has one toxin/antitoxin, many competitors could cohabit with it, but a strain that has 7 ways toxin/antitoxins is probably going to be pretty clonal. 2) That secreted proteins are potentially exploitable by competitors, so we can get a rough measure of 'cooperativity' by looking at the fraction of a bacterium's genome that codes for secreted products. After correcting for phylogeny, the number of Type 6 secretion systems and effector proteins explained 90% of the variation in secretome size! That's an astonishingly large effect. One of the things I love about this work is its scope: we link the physics neighbor killing to the emergence of novel ecological patterning, and then show that this patterning can (and actually does) change evolutionary trajectories, favoring cooperative extracellular metabolism. Read the paper here. We’re currently pursuing a number of follow-up projects. Bonus: Listen to Moselio Schaechter read the first paragraph from our paper. He’s like the David Attenborough of microbiology- this made my week! Horizontal gene transfer of T6 weaponry One of the neat things about cholera is that the same ecological conditions that trigger T6-mediated killing also turn on natural competence, causing them to take up the DNA of slain competitors and incorporate some if it into their own genome. In a paper with Brian Hammer’s group, we show that choera readily swap out their T6SS effectors (antibiotics) and immunity genes. We modeled the spatial dynamics of this horizontal gene transfer, and two things became clear: HGT is very costly, mostly because transformant cells come into conflict with former clonemates and are usually killed. But, if their competitor has superior T6 weaponry, it pays to acquire it through horizontal gene transfer. Finally, we show that acquiring the weapons of fallen competitors can act as a bet-hedging strategy, maximizing long-run fitness when the future competitors are unpredictable and variable. Bet hedging We describe the first individual-level bet-hedging behavior in a microbe The unpredictable nature of life has significant evolutionary consequences. Organisms that have large swings in their fitness through time pay a price for this variation (for more detail on the ‘geometric mean principle’ of fitness that demonstrates this, read this blog post). One solution to the problem of unpredictability is to “put your eggs in multiple baskets” and produce offspring well-suited to several possible future environments, improving the chances that some of them succeed. Historically, this process of bet hedging has been studied from two different perspectives. Those studying multicellular organisms focus on offspring diversity at the level of a single individual organism. For example, individual plants can produce seeds that germinate immediately, or after several years of dormancy. In contrast, microbiologists have focused on how a single microbial genotype can produce multiple phenotypes at the population level.  Individual microbes typically do not produce offspring that are different form their parents.  Instead, they have a very low chance of producing a different phenotype (often less than once per thousand divisions), but this starts a new lineage with the new phenotype.  At the population level, a single genotype can thus exhibit multiple phenotypes, but individuals rarely diversify. We found a novel bet hedging mechanism in the rhizobium Sinorhizobium meliloti. These bacteria symbiotically fix nitrogen inside legume root nodules. To reinfect new legumes, they need to enter the soil, take up a free-living existence, and find a new legume host. To help with soil survival, these bacteria can store carbon and energy in the polyester poly-3- hydroxybutyrate (PHB). PHB can be used either to keep rhizobia alive during starvation, a common problem in soil, or to fuel the production of new offspring. Here’s the challenge: if the duration of starvation is unpredictable, how should these rhizobia use their PHB? It turns out that they hedge their bets against short- and long-term starvation. Starved S. meliloti divide asymetrically, producing one low-PHB ‘grower’ and one high-PHB ‘persister’. The low-PHB phenotype has a fitness advantage when resources (like a new legume root or exogenous food) are encountered quickly, while the high-PHB rhizobia has a survival advantage during prolonged starvation and antibiotic stress. This novel bet-hedging behavior allows even a single cell to compete for both short-term resources and survive long-term starvation. The structure of risk dictates optimal diversification strategies The following is in collaboration with Eric Libby. Environmental risks (like drought) can vary in both their spatial and temporal characteristics. The best summary of this is from our 2015 Evolution paper: “ A single unpredictable event may vary in scale from population-wide (e.g., a landscape-level process like unpredictable season length) to local (e.g., chance of nest discovery by a predator). Similarly, risk may affect populations randomly in time or it may occur in correlated series. Using the above examples, season length is largely uncorrelated from year to year, but a predator that discovers a nest site may revisit it frequently.” In this paper we demonstrate that risks that are not correlated in time favor rapid diversification (the faster the better), while risks that are autocorrelated favor slow diversification. This is because slow diversification allows for adaptive tracking of risks that tend to occur in series. Differences in the ecology of plants and animals and microorganisms may explain some of the differences in observed diversification rates (microbes tend to be much slower diversifiers, but also may experience more autocorrelated risk). We see some interesting time-dependent effects, resulting in layered bet hedging strategies. In a follow-up paper (preprint here), we examine the dynamics of stochastic phenotype switching when risk is correlated in time in more detail. Over the short-term, slow rates of diversification are favored (allowing for adaptive tracking of risk, as described above), but over long time periods this is costly, leaving a lineage much more susceptible to extinction when the environment changes quickly. As a result, over very long time periods (tens of thousands of generations), environmental risk favors intermediate rates of diversification. In the paper we explore how this selective tug-of-war, which acts over quite different time scales, presents a challenge for adaptation. Programmed cell death can increase the efficacy of microbial bet hedging One of the surprising things microbes do is kill themselves. While there are a number of hypotheses out there to explain this apparently self-destructive behavior (see this great review by Nedelcu et al.), these arguments tend to be pretty straight- forward ‘for the good of the clone/group’ altruism, and few (if any) have been rigorously modeled. We have a new, fairly general idea: programmed cell death (PCD) in microbes can drive population cycling, even when carrying capacity has been reached, which gives microbes more opportunities to diversify through stochastic phenotype switching. It works surprisingly well in our model- take a look at the preprint here to see the nitty gritty of our analytical results, or our spatial simulation (which we did to see if we can get enough clonal assortment arising from the demographics of growth and death to allow PCD+ strains to resist cheating by a PCD- competitor. And yes, we totally do.). Legume-rhizobium symbiosis A bacterial storage resource plays a key role in the evolution of cheating rhizobia Rhizobia are symbiotic bacteria that live inside legume root nodules, where they fix nitrogen using reduced carbon supplied by the host. Nitrogen fixation by rhizobia is largely altruistic: they use only a tiny fraction for their own growth, the rest is supplied to the legume. Multiple strains of rhizobia typically infect a single host plant, so any benefit they provide to the plant is shared with other rhizobial strains inside the same plant. Rhizobia that invest less in N2 fixation have more carbon available for reproduction, so why haven’t non-fixing rhizobia displaced high-N2 fixing rhizobia entirely? It turns out that plants punish rhizobia that fix little N2, limiting their reproduction. My thesis focused on the following question: Why, in the face of such effective punishment, are poor mutualists globally persistent? Key findings. In addition to using reduced carbon for N2 fixation and reproduction, rhizobia in root nodules can synthesize the storage lipid poly-3-hydroxybutyrate (PHB), often accumulating PHB to more than 50% cell dry weight. Rhizobia are horizontally transmitted symbionts and must survive in the soil between hosts. I found that rhizobia escaping nodules can use stored PHB to survive starvation and reproduce up to 3-fold (Ratcliff et al., 2008), but PHB synthesis is energetically expensive and trades-off with N2 fixation.  As a result, PHB synthesis increases rhizobium fitness while reducing legume fitness, and is a central (but previously unappreciated) mechanism in the evolution of conflict between rhizobia and legumes. I found that some rhizobia have evolved sophisticated mechanisms that increase PHB accumulation. Bradyrhizobium elkanii produces rhizobitoxine, a chemical that blocks the legume’s ability to synthesize the hormone ethylene. Rhizobitoxine production reduces the host’s growth, decreasing rhizobia per nodule for all strains on a plant, but substantially increases PHB accumulation for rhizobitoxine-producing rhizobia alone (Ratcliff and Denison, 2009). Further, I demonstrated that accurate estimates of rhizobial fitness should include the reproductive capacity of stored PHB (Ratcliff et al. 2012 and Kiers et al. 2012). Microbial social evolution Cryptic indirect effects in the evolution of microbial dormancy. Since Hamilton’s (1964) insights into inclusive fitness, evolutionary biologists have attempted to disentangle the role of direct and indirect fitness effects in behavioral evolution. This has mostly been limited to traits where direct and indirect effects act in opposite directions (i.e. altruistic cooperation). In a recent paper in The American Naturalist, we describe a new approach for quantitatively separating direct and indirect fitness effects of a trait that affects both: microbial dormancy. Dormant individuals are resistant to antibiotics (a direct benefit), but spare external resources, increasing the growth of nondormant kin (an indirect benefit) and competitors (an indirect cost). We find that dormancy provides an indirect fitness benefit only when the population is sufficiently structured, but that these cryptic indirect effects can have a large effect on fitness. Antibiotics: weapons, signals, cues, or manipulation? There have been many explanations for the ecological role of antibiotics in nature, but no organized approach for how microbiologists should think about these various explanations. In 2011 Ford Denison and I co-authored a perspective in Science using social evolutionary theory to examine antibiotic production. We classified antibiotics into the four categories above, as determined by the fitness consequences to the antibiotic producer and recipient. Read the paper. A model of two bacterial strains (red and blue) growing on a surface, engaging in Type-6 secretory system conflict. They start out well-mixed, but quickly phase separate. Simulations show that HGT of T6 weaponry is adaptive only when competitors have superior T6SS effectors.