The random walk of evolution
March 11, 2011 § 2 Comments
One of the reasons that some people have a hard time accepting evolution is that the organisms we can study today are so darn complicated that it’s hard to see how they could have arisen from many small steps. Take chemotaxis in bacteria, for example. The motor for movement, the flagellum, has frequently been cited as an example of so-called “irreducible complexity“; proponents of intelligent design claim that it could not have evolved because every one of its 40 or so parts is needed for the motor to function at all. To me, this argument has a strange sort of circularity about it: the underlying assumption seems to be that the flagellum needed to be exactly the way it is, and so the problem that evolution faced was how to build it in such a way that each step improved motor activity. If you can’t imagine a path in which motor activity smoothly improved from zero to current-day, evolution must be wrong; and because you started out by imagining what is in effect an engineering design process, you end up concluding that the only way it could have succeeded is through the actions of a designer. But what is wrong instead is the idea that evolution takes linear paths that lead ever upward, and that have always been focused on the improvement of the function we see today. It’s so much easier to imagine evolution as a march of progress than to try to get your head around over 3 billion years of a meandering multibranched process of accident, failure and occasional success. In particular, it’s hard to keep in mind that the advantage that leads to increased evolutionary success can be very very small; often so small that it’s extremely hard to measure.
A recent paper (Wei et al. 2011. The population dynamics of bacteria in physically structured habitats and the adaptive virtue of random motility. PNAS doi/10.1073/pnas.1013499108) set out to disentangle a different question in the evolution of chemotaxis, namely the question of what drove the evolution of motility in the first place. This is a less trivial question than it may appear at first sight. The genes that allow the bacterium to sense and respond to chemical cues in the environment are separate from those that allow movement, so either this is another example of irreducible complexity or (my choice) one of these two functions is useful without the other. Does it do you any good to sense food if you can’t move towards it? Maybe not, though I wouldn’t rule out the possibility that at some distant time in the past it might have been advantageous for bacteria to be able to grow selectively in the direction of food. (If nobody else can move either, why not?) But it seems more likely, from several lines of evidence, that being able to move is the function that evolved first, and that chemotaxis evolved as a way to bias the motion towards useful directions.
So what good is movement without direction? Under what conditions does a randomly-moving bacterium have an advantage over one that can’t move, or can’t move as fast? This question requires us to think about the structure of the environment the bacteria live in, in a way that we normally don’t. Experiments on bacterial growth typically make great efforts to minimize lack of homogeneity in the environment, ranging from shakers to chemostats; and similarly modeling approaches generally consider the bacteria to be, as Wei et al. have it, “cavorting about in an environment that, from the perspective of an individual bacterium, is [spatially] dimensionless”.
In the case of experiments, it’s easy to change to an inhomogeneous environment — mostly, you just have to stop trying hard to make the environment homogeneous. For example, you can grow the bacteria on a solid support such as agar, and measure the density at different points by taking samples from the agar and suspending the bacteria from the samples in saline for counting. It’s slightly trickier to work out the theory for an inhomogeneous environment, but Wei et al. put together a system of partial differential equations (PDEs) to model resource-limited growth in an environment where food is unevenly distributed. For both theory and experiment, when you compete (in the case of the theory, “compete”) a motile strain with a non-motile strain, the motile strain wins. The model shows that the advantage of the motile strain is larger for environments with higher viscosity and more food. There is also support for the idea that more is better — faster-moving cells have a larger advantage, other things being equal.
In the experiments, the authors tried two ways of inoculating petri dishes containing soft agar. In the first, they placed a mixture of motile and non-motile bacteria in the center of the dish; in the second, they introduced the same mixture into the agar while it was still liquid, then allowed the agar to set. In the first condition, it’s kind of obvious why the motile strain has a major advantage: the non-motile strain is stuck in the middle, and when the food in the middle runs out those cells can’t grow any more. Only the motile strain has access to the resources in the rest of the dish. In the second condition, the reason behind the result is less obvious: the motile strain might just as easily be using its mobility to move into a region depleted of food as to move into a richer region. Nevertheless, in both cases the motile strain has a noticeable advantage; the ratio of motile to non-motile cells is in the range of 10-100 in most conditions. Any competition for resources (whether between the two strains or between individual cells of a strain) can be evaded, with some probability, by the motile cells but not by the non-motile cells. (The PDE model is based on the first condition, in which the inoculation is all in one spot, and the match between model and experiment is rather good for this condition.)
The experimental work I’ve described so far used bacteria that can not only move, but also bias their motion in the direction of food. Wei et al. repeated their experiments using a motile strain that lacks the chemotaxis functions, and found that the motile strain is still better off than the non-motile strain in agar, although the competitive advantage is smaller. So here’s an evocative view of evolution as a random walk made up out of small steps: motility alone can give a bacterium a slight evolutionary advantage, and each incremental step towards faster motility gives a larger advantage; a bacterium that finds a way to bias its motility in useful ways gains a yet larger advantage. One can’t know for sure that motility came first (it was all a very long time ago…), but at least there seems to be no reason why motility could not have evolved independently of chemotaxis.
Wei Y, Wang X, Liu J, Nememan I, Singh AH, Weiss H, & Levin BR (2011). The population dynamics of bacteria in physically structured habitats and the adaptive virtue of random motility. Proceedings of the National Academy of Sciences of the United States of America PMID: 21325053
I have a question regarding the “3 billion years of a meandering multibranched process”. Do you know of any work to model such a meandering multibranched process? While this is not my own area of expertise, it seems like population study mathematics might be useful in this regard (like the popular sim games).
I ask the question because 3 billion just doesn’t seem that big a number to me. Particularly when you consider the period of a branch increases with the complexity of the organism.
As an example consider that we have exhausted the available IPv4 address space (a value of that order) in less than 50 years. Consider that the process by which the addresses were consumed was meandering and multibranched (if not truly random). Consider also that regardless of all “possible” meandering multibranched processes, we are only concerned with the small subset of paths that actually yielded out current state (the single IPv4 address space if you will). I this light 3 billion seems almost puny in terms of scale, to me at least.
Thank you for your consideration.
hi, and thanks for the comment. There are lots of efforts to model specific parts of the evolutionary process, but I’m not aware of an effort to model all the way from the beginning to the present day. I am no expert, but I think there are several reasons for this, including the fact that we don’t know where to start (what did life look like on the molecular level when it began?), and we don’t know what to optimize (what conditions were early organisms facing, and how did conditions change over time?). If I implied that it is only the amount of time that makes the problem complicated, I didn’t mean to… my apologies!