In one of the most important papers in population genetics, Begun and Aquadro showed that levels of DNA sequence polymorphism are positively correlated with recombination rate. There are three ways of interpreting this result:

  1. Recombination is mutagenic, and polymorphism is higher with increased mutation.

  2. Positive selection on beneficial mutations decreases levels of linked neutral polymorphism. This effect is greater in regions of decreased recombination and is known as hitchhiking.

  3. Selection against deleterious mutations also decreases linked neutral polymorphism. Like hitchhiking, this effect is greatest in regions of low recombination. This is known as background selection.

Each of these scenarios makes certain predictions regarding patterns of divergence and polymorphism. We can determine whether mutations due to elevated recombination rate are responsible for variation in polymorphism by comparing polymorphism with divergence. This is shown in the figure below the fold:


If the mutagenic effects of recombination were responsible for elevated amounts of polymorphism in regions of high recombination, then we would also expect to see a similar correlation when we look at levels of divergence (the top two graphs). In contrast, if natural selection were responsible for the correlation of recombination rate and polymorphism, divergence should be uncorrelated with recombination rate (the bottom two graphs).

In looking at the data from Drosophila, Begun and Aquadro so no correlation between divergence and recombination rate (the bottom scenario). This led them to conclude that genetic hitchhiking is responsible for the correlation between polymorphism and recombination rate. The Charlesworths argued that background selection could also explain the observed correlation. But neither group claimed that mutagenic effects of recombination led to a correlation between polymorphism and recombination rate.

I wouldn’t have told you all of that if there wasn’t a twist, and there is one. Gil McVean and colleagues have performed a rigorous analysis of recombination rate and polymorphism in humans (summary here). They were able to examine recombination rate on a more fine scale than previously possible. Based on their analysis, they conclude that the mutagenic effects of recombination are responsible for the correlation between recombination rate and polymorphism.

But if recombination increases polymorphism by inducing mutations, we would also expect that divergence would be correlated with recombination rate. Why don’t we see such a correlation? McVean’s group argues that fine scale recombination hotspots in mammals are ephemeral. That means the same variations in mutation rate that increase levels polymorphism do not persist long enough to affect rates of divergence.

So what does this mean for models that connect recombination rate and polymorphism through natural selection (rather than mutation rate)? Well, it matters which system one is studying. In mammals, recombination hot-spots and cold-spots occur along the length of a chromosome. Begun and Aquadro’s initial study was performed on Drosophila melanogaster, where the recombination landscape is much simpler: recombination is suppressed at the centromeres and telomeres, and elevated in the center of the chromosome arms. That suggests that areas of elevated recombination should persist longer in Drosophila than in humans because they are tied to more enduring features — centromeres and telomeres.

Another thing to take into consideration is effective population size. Drosophila have larger populations than humans, so the effects of hitchhiking and background selection will be different in the two species. Larger population sizes mean the population wide recombination rate will be higher, which means the effect of selection events will be more localized. If the population is small enough, the effect of selection events may be spread out over the chromosome so much that they are no longer associated with a recombination hotspots. Or maybe I’m just grasping at straws.

Anyway, I think it’s important to take two things into consideration regarding McVean’s result. First, variation in recombination rate is not constant between taxa. And, second, the effects of selection depend on population size, which also varies between taxa. An association between polymorphism and mutation rate may explain the correlation between polymorphism and recombination rate in humans, but natural selection may still be responsible for the correlation in Drosophila.

Begun DJ and Aquadro CF. 1992. Levels of naturally occurring DNA polymorphism correlate with recombination rates in D. melanogaster. Nature 356: 519-520. doi: 10.1038/356519a0

Spencer CCA, Deloukas P, Hunt S, Mullikin J, Myers S, et al. 2006. The Influence of Recombination on Human Genetic Diversity. PLoS Genet 2: e148. doi: 10.1371/journal.pgen.0020148


  1. #1 Peter Ellis
    November 10, 2006

    I can see a fourth possibility:

    4) Polymorphism is recombinogenic.

    Naively, it doesn’t seem impossible that when homologous chromosomes synapse during meiosis, recombination foci are preferentially targeted to regions where there’s greater divergence between the homologs. Up to the point where divergence is sufficiently large that you can’t achieve synapsis, anyway.

    How would this hypothesis play out in the above analysis?

    Thinking about it, from this hypothesis you could make a couple of other predictions:

    1)Highly inbred lab strains will show reduced recombination relative to an outbred population

    2)You should see increased recombination in F1 crosses between different inbred strains relative to either parent strain.

    However, if there are *global* constraints on the number of recombination foci per cell, or per chromosome, you wouldn’t see that effect.

    To truly test it, you’d have to look at partial consomics. Take one mouse inbred mouse strain, cross it with another mouse strain that’s exactly the same except that one chromosome arm is derived from a different strain – then check if that specific chromosome arm shows an increased level of recombination relative to either parent strain.

  2. #2 RPM
    November 10, 2006

    The problem is that in order to detect recombination you need genetic variation. It’s a catch 22. What you could do, however, is create a genome with equal amounts of variation across the entire chromosome.

  3. #3 Peter Ellis
    November 13, 2006

    The problem is that in order to detect recombination you need genetic variation. It’s a catch 22. What you could do, however, is create a genome with equal amounts of variation across the entire chromosome.

    Yes, e.g. a Balb/C mouse where one of its copies of chromosome 2 is brought in from C57, or similar. However, as I said, if there’s some kind of constraint on the *total* number of crossovers per chromosome, that might obscure any effect you’re looking for.

    That’s why I went for the partial version – e.g. a Balb/C mouse, but where only *half* of one of its copies of chromosome 2 is C57-derived. You can then see if that half of the chromosome gets more crossovers than expected.

    As for detecting the recombination, it depends what resolution you want. If you were looking at large enough region, you could do it just by staining for recombination foci at pachytene and counting a number of cells.

New comments have been temporarily disabled. Please check back soon.