Empty DNA

Mitochondrial DNA (mtDNA) is one of the most used markers in molecular ecology¹. A good molecular marker for population level studies should be neutral, so that researchers can use it to infer things like:

• Population size and changes in population size (expansions and bottlenecks);
• Population structure;
• And phylogeographic relationships.

A recent meta-analysis of mtDNA polymorphism revealed that there is no relationship between levels of diversity and population size. The authors of that study explain the lack of a relationship on recurrent selection on mtDNA genes, which affects the entire molecule because it lacks recombination. This is not a good property for a neutral marker. Now that we know mtDNA is less than ideal, will ecologists continue to use it?

The meta-analysis was published in April of this year, so I don't expect to see its effects any time soon. I wonder, however, whether mtDNA will remain a popular marker despite the fact that it violates many of the assumptions required for inferring demographic properties of populations. At the Evolution Meeting, I saw a talk in which the presenter inferred a population expansion in her study organism (an arctic bird) using only mtDNA. Now, she was greatly limited in marker selection -- she was working with museum specimens, so nuclear microsatellites were out -- but I was curious whether she was aware of the limitations of using mtDNA. When I asked her, as politely as I could (acknowledging her technical limitations), whether she was concerned about drawing demographic inferences from a non-neutral locus, she merely said she had no other choice.

I felt quite guilty for pointing out such a major flaw in her analysis. I attempted to relieve some of my guilt by asking a fellow grad-student whether my actions were warranted. The grad-student pointed out that the extremely negative Tajima's D calculated from the mtDNA sequences could not be due to a population expansion alone (as the presenter suggested), but would also require a wicked selective sweep.

I'm of a far more outspoken persuasion than my fellow grad-student, but I wish she would have asked the presenter whether any coalescent simulations were performed to estimate the expansion parameter for this population. This question -- rhetorical as it may be -- would have been far superior to mine. I merely suggested that mtDNA ain't so hot. Trying to estimate the rate of expansion may have revealed that a Tajima's D that negative could only result from a population expanding from four individuals to four billion (or something equally absurd). This would provide further evidence that any demographic inferences made from mtDNA should be taken with quite a few grains of salt.

If mtDNA is unacceptable for inferring demographic parameters, what are we left with? Not everyone works in model organisms, so direct sequencing of multiple nuclear markers is not feasible in most cases. Even nuclear microsatellites can be difficult to identify. There are quick and dirty markers such as RAPD, AFLP, and RFLP. I used to be skeptical about using these markers, but the majority of my skepticism is now directed at mtDNA. For instance, only having read the abstracts of two papers in which the authors study populations of pigs (one using mtDNA, the other using AFLP) I would trust the conclusions in the AFLP paper much more than the mtDNA paper.

1- I draw a distinction between population genetics and molecular ecology in that molecular ecology tends to be interested in the whole organism, whereas population geneticists as much more reductionist. Molecular ecologists use molecular markers to infer demography in their population(s) of interest. Population geneticists are interested in how the actual sequences evolve (ie, how they respond to mutation, selection, demography, recombination, etc). The two groups are using similar data, but moving in opposite directions.