Trade balances in microbial communities

So this week seems to be turning out to be cooperation week.  We talked about cooperative behavior of the proteins that make up the cytoskeleton on Monday, cooperation in breeding behavior in birds on Wednesday, and now it’s time to talk about cooperative behavior in bacteria.  If only I’d planned it in advance.  Ah well; can’t have everything.

You know, of course, that in the wild bacteria do not typically live in monocultures: different varieties of bacteria both compete and collaborate, and the complex interactions that result are not easy to study.  A number of labs have been working on developing well-defined synthetic communities to ask questions about how communities evolve.  The Silver lab has now taken a rather different approach (Wintermute and Silver 2010, Emergent cooperation in microbial metabolism. Molecular Systems Biology 6: 407 PMID: 20823845), by exploring the interactions among 46 different metabolically impaired strains of E. coli and rationalizing the results in terms of a flux-balance-analysis model of interacting strains.

What interested Wintermute and Silver was the fact that bacterial communities can perform all kinds of important metabolic tricks that individual species can’t manage.  This is not hard to understand, or at least to imagine that you understand: one species finds an efficient way to produce rare metabolite A, another species develops an efficient way to produce rare metabolite B, when you mix the two together they both have the advantages of a supply of both A and B, but they’ve effectively halved the cost of production (making all those enzymes) by sharing.  But is that really how it works?

Wintermute and Silver [by the way, it must be wonderful to share a name with the manipulative superintelligent computer from William Gibson’s Neuromancer — a brilliant book — especially when your advisor is named after a robot from another vaguely cyberpunk novel] acquired 46 conditionally lethal mutants that block pathways involved in producing essential metabolites (such as amino acids and nucleotides).  They labeled each mutant with either a red fluorescent protein or a yellow fluorescent protein, co-cultured pairs of mutants, and measure the growth of each strain.  The idea here is that if mutant 1 makes a metabolite that mutant 2 needs, mutant 2 will grow better.  But if mutant 1 is not growing very well, the amount of help it can give mutant 2 will be limited.  The best pairings will be those where both mutants help each other, so both can grow much better than they would if they were growing alone.

When the authors tried co-culturing each of the 46 strains with all 45 other strains, they found that 17% of the pairwise co-cultures grew dramatically better than the rest. This was clearly due to cooperation between the strains.  Using an ODE model, Wintermute and Silver were able to extract from the growth curves a “cooperation level” for each strain. More cooperation is better, up to a point: if strain A cooperates well, it allows its partner (B) to grow and produce the metabolite A lacks.  But if A cooperates too well, B will grow so fast that there are no nutrients left to support the growth of A.

The interaction between two (mutant) strains that grow better together than either one does on its own is reminiscent of (though different from) the interaction between two mutated genes in a single organism.  In a single organism you might think about synthetic lethality: two mutations that individually aren’t lethal become lethal when they are put together.  In this case, you’re looking at synthetic mutualism — rescue — between different strains, which the authors call synthetic mutualism in trans (SMIT). And you can analyze it in much the same way as you can analyze synthetic lethality.  Mutations in the same pathway are unlikely to show synthetic lethality; similarly, they’re unlikely to show synthetic mutualism.  If two mutations each block the production of the same metabolite, they’re unlikely to rescue each other.  And the pattern of mutant combinations that give rise to synthetic rescues is likely to be informative, just like the pattern of combinations that produce synthetic lethalities.  By clustering rescue profiles, Wintermute and Silver show that related mutants show related profiles — the enzymes responsible for making tryptophan, for example, all cluster together; they all have a similar deficit, so they can all be rescued by the same set of cooperators.

Given all of these results, I think we’re justified in beginning to consider a combination of two strains with different metabolic impairments as a single community organism, a bit like a community of differentiated cells in a multicellular species — although, Wintermute and Silver’s pairs of mutually dependent organisms have not yet had a chance to evolve and become truly interdependent.  This is just the raw material from which interdependence could evolve.  With that caveat, let’s look at whether the behavior of the co-cultures can be predicted based on what we already know about the E. coli metabolic network and the way metabolites flow through the network.  This is an approach pioneered by Bernhard Palsson’s group and, locally, intelligently pursued by Daniel Segrè.  Wintermute and Silver create a joint flux model that combines the flux in both mutants and asks how well the combined superorganism can grow.

The results are mixed.  While the joint flux model does a reasonably good job of predicting which pairs can grow, it generally over-estimates the amount of growth.  This seems to be a consequence of treating the pair of organisms as one: in determining the optimal joint flux, the model acted as if benefit to strain A is exactly equivalent to benefit to strain B.  But the control mechanisms acting on strain A are all internal to that strain; strain A has no idea whether strain B needs more glucose.  So the model is making too strong an assumption of optimality.  Wintermute and Silver therefore switched to a different modeling approach that effectively puts a “price” on each exchange of metabolites — metabolites that are easy for strain A to produce (based on the flux of metabolites in strain A) are cheap for strain A, though they may be extremely valuable to strain B.  The ideal cooperation is one in which both strains are getting valuable goods from the other in exchange for goods they have made cheaply.  [Rather as the US exchanges cheaply-produced education for material goods from China; though this particular analogy may not last much longer.]  Using this approach, the authors came out with predictions that matched their data well.

So it appears that it’s relatively easy for bacteria to trade the results of metabolic innovations, creating communities that can perform transformations that no single species can accomplish alone; and for bacteria, as for people or countries, the best trading partners are the ones that want what you have too much of.  And this can be rationalized in the context of a model that determines the cost to one species of producing a benefit to the other. A nice step forward in the analysis of community behavior.

Wintermute EH, & Silver PA (2010). Emergent cooperation in microbial metabolism. Molecular Systems Biology, 6, 407 PMID: 20823845