Let's say I want to evolve a bacterium that is resistant to an antibiotic. I want to do this by growing initially clonal populations of bacteria in presence of this antibiotic for a long time.

I have two options:

  • Make one 50 ml flask culture
  • Make ten 5 ml tube cultures

What are the comparative advantages and disadvantages of these two? Clearly, the ten culture version is better at probing the space of all possible resistance mechanisms. However, what about the rate of evolution? If I just want to get a resistant strain as fast as possible, which one should I use?

Note that "antibiotic" is just a proxy for any selective pressure, and "bacteria" is proxy for any organism. If possible, I think it will be interesting to address the effects of sexual vs. asexual reproduction in the organism that is being evolved.

There is already a question, How does rate of evolution/innovation scale with population size? that is very similar. I think mine is different, because it is not one small population vs. one big population, but one big population vs. many small populations. In other words, I am attempting to control for number of selection/mutation events, and focus only on the stratification of the total population.

  • $\begingroup$ Here are my thoughts. I might be wrong. Such process are mainly limited by the availability of mutations. In consequence, as long as the product of the selection differential and the population size is high enough, it does not really matter whether you split your populations or not. $\endgroup$ – Remi.b Dec 19 '14 at 4:32
  • $\begingroup$ If it's the same number of individuals in either set up then the rate of mutation will be the same, smaller populations are prone to drift so this could reduce the speed/quality of response in the smaller populations, but replicate populations will be requisite if you want to publish something, and interesting comparisons could be made between replicates (e.g. do they evolve different solutions to the same problem?) - my preference would therefore fall to many replicates $\endgroup$ – rg255 Dec 19 '14 at 10:07

Note: This is not an area where I know the litterature well

Where are many counteracting processes to consider for this question. For instance, the rate of evolution will be affected by the rate of mutation, the distribution of positive and deleterious mutations, strength of selection, whether the fitness effects are small or large, if fitness effects depend on interactions between multiple mutations or single mutations with large effect, asexual vs. sexual populations, and processes such as genetic drift.

A recent review that is useful for your question is Lanfear et al (2014). They focus on how substitution rate is affected by effective population size, and go through both theoretical and empirical evidence. They conclude that in most cases, most mutations are deleterious, and that drift and selection are the most important effects in determining rates of evolution. They also state the effect of effective population size (Ne) is likely to depend if mutations are dominated by deleterious or advantagious mutations, and they expect the rate of evolution to be negatively correlated with Ne when deletious mutations dominate evolution and a positive correlation to Ne when advantagious mutations dominate.

A paper by Rozen et al (2008) indicates that while large populations are more effective in fixing mutations with large positive effects, small populations can be more effective in complex "rugged" adaptive landscapes.

While such determinism speeds adaptation on the smooth adaptive landscape represented by the simple environment, it can limit the ability of large populations from effectively exploring the underlying topography of rugged adaptive landscapes characterized by complex environments.

However, this does not deal directly with the rate of evolution, but mostly about attaining the highest possible fitness in a particular fitness landscape. Another interesting paper is Desai et al. (2007), but I haven't looked through this yet. However, they find (for asexual bacteria) that the speed of adaptation scales less than linearly with population size, and similar results have also been found in studies from Lenski's group. Their interpretation is that evolution is driven by multiple concurrent mutations of small effect and not one-by-one fixation of single mutations (which would favour large populations more).

For your case of evolution of resistance, I think one crucial thing is if you expect resistance to be driven by a few single mutations or combinations of many mutations that interact (related to whether resistance is "easy" or "hard" to evolve). From what I understand, the first case should favour a single large population (since they will fix mutations for resistance more quickly), while the second case might favour many smaller populations (since you might get more effective exploration of combinations of mutations)

These are just a few starting points in a very large field, and I hope that you find them useful.

  • $\begingroup$ Very helpful! I especially appreciate the links to literature, thanks. $\endgroup$ – Superbest Dec 19 '14 at 13:33
  • 1
    $\begingroup$ Nice answer (+1). As I said in my comment (and I might be wrong) I would tend to think that the difference between several small or one big population is totally negligible in this case because 1) $2\cdot N\cdot s$ will much large than $1$ anyway and 2) adaptation is mutations limited $\endgroup$ – Remi.b Dec 19 '14 at 17:01

With very large population sizes like this the effect of genetic drift goes to zero, which is greatly simplifying. I'm also assuming selection is very strong, so that fixation times are small, and that mutation rates and population size more or less cancels out, so there are a finite number of mutations and the selection space is relatively unexplored. If selection is weak everything takes longer, but I'm pretty sure the results are the same. There's greater exploration of alternatives, but that doesn't affect our thought experiment comparing population size.

In asexual populations: 10 cultures will give you ten answers to your selection criteria, one of which will be 'best'. One culture will give you the best solution a little bit later (larger populations take longer to reach fixation).

In sexual populations: One culture is better. The space of alleles is larger, and positive or conditionally positive mutations will be tested against more possible synergistic partners.

Consider that there are ten positive mutations split across your population. A sexual population in ten sub-populations has a tiny chance of getting all ten mutations into the same gene pool, but all ten sub-populations combined make it near-certainty that all ten mutations will make it into the same organism.

There's a limit to sexual population size above which evolution doesn't accelerate with additional individuals, because any additional mutations they bring to the table are eventually guaranteed to already exist in the gene pool. That limit is related to the genome size and mutation rate and probably other things, but for biological species it is enormous.


Both have its benefits and drawbacks.

The single big culture would have a bigger risk of contamination, but a bigger chance that the resistant mutated individuals are present, so in this case you would have a lower risk of complete extermination of your culture.

The smaller several cultures on the other hand will let you try several dosages of the antibiotic without fearing exterminating your culture. On the other hand each individual culture probably will have lower diversification, so lowers the chance of new strains.

  • $\begingroup$ RE the first segment of your answer, @superbest is using clonal populations as a base so this is not true in this case. There would be no standing genetic variance in the population(s) regardless of the population size, all resistance would have to come via new mutation. If the meta populations are of equal size (e.g. one population of $N$ individuals or 10 populations of $1/10N$) then there is no difference between the expected number of mutations between the two alternatives. $\endgroup$ – rg255 Dec 19 '14 at 13:02
  • $\begingroup$ While you are correct and make valid points, in retrospect I realize that I failed to adequately specify that I am interested in the biology, not the methodology. As such I neglect contamination risk, and apply the same conditions to all cultures. (this is more realistic, if you think of natural populations) $\endgroup$ – Superbest Dec 19 '14 at 13:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.