Two weeks ago, an article was published online in the inaugural issue of the new journal Nature Ecology & Evolution entitled “Experimental test and refutation of a classic case of molecular adaptation in Drosophila melanogaster“. As the title suggests, the article soundly refutes a classic example of molecular adaption that was first established in a seminal paper by McDonald and Kreitman in 1991.
As expected, creationists and ID proponents have rushed to the scene, crowing about this “failure” of the theory of evolution. Instead of doing novel research themselves, they’re content to sit on the sidelines and wait patiently for scientists to publish something that they can wave around as though it vaguely supports their ideas or present problems for evolution. Let’s take a look at what was actually refuted, and see just how desperate the science-deniers really are in promoting this article.
The case of molecular adaptation in question is the adaptation of the Adh gene in a species of fruit fly (D. melanogaster) to improve the catalytic power of the ADH (Alcohol Dehydrogenase) enzyme and therefore caused the flies to have an increased alcohol tolerance. D. melanogaster resides in an ethanol-rich niche, feeding on fermenting (rotting) fruit, while closely related species do not. In 1991, John McDonald and Martin Kreitman of Princeton University published a paper testing the hypothesis that the Ahd gene had undergone adaptive evolution in the D. melanogaster lineage relative to its sister species. The test they implemented has since become known as the McDonald-Kreitman (MK) test, and forms the basis of a lot of studies inferring adaptive molecular evolution.
Without going into all the details, the test requires sequence information from multiple individuals from least two species, in order to gather data about the number of nucleotide substitutions that are polymorphic (vary within the species) and that are fixed (only vary between species, not within them), and how many of these are synonymous (don’t change the encoded amino acid) and non-synonymous (do change the amino acid). The test compares the ratio of non-synonymous and synonymous fixed nucleotide differences (dN/dS) to the ratio of non-synonymous and synonymous polymorphic nucleotide differences (pN/pS) . If the dN/dS ratio is significantly larger than the pN/pS ratio, this represents a departure from neutrality and therefore can represent the signature of adaptive molecular evolution.
The reasoning here is that if the sequence had evolved neutrally, we would expect that the polymorphic ratio should be the same as the fixed ratio, because there would be no selective pressure to cause the fixation of a higher number of non-synonymous (potentially function-altering) substitutions. In simple terms then, the MK test calculates whether there have been a higher proportion of potentially function-altering mutations in a lineage than would be expected by neutral evolution. In their paper, McDonald and Kreitman found that the substitution frequency table for the Adh gene in 3 closely-related species of fruit fly looked like this:
From the table, it’s easily to calculate dN/dS = 7/17 = 0.41, and pN/pS = 2/42 = 0.05. 0.41 is significantly higher than 0.05, so the null hypothesis of neutral evolution can be rejected. Easy. It’s important to note, however, that departure from the expected neutral results isn’t “proof” that the sequence in question has evolved adaptively, it is just a good indication. Alternate explanations for the departure from neutrality include relaxed constraints on the sequence, or simply a series of unlikely events. Another important point is that this test doesn’t tell us which of the species in question underwent the adaptive evolution, it just tells us whether or not there is a signature of selection somewhere in the group or not.
Fast-forward 25 years, and the technology is now available to infer and synthesise ancestral gene sequences to see how their products actually behaved. The team of researchers in Joe Thornton’s lab at the University of Chicago did exactly that, reconstructing the Adh gene sequence that was present in the last common ancestor of D. melanogaster and its closely related species, and raising transgenic flies with this ancestral gene sequence in their genome instead of the modern one. The expectation was that the flies with the ancestral gene should be less tolerant to alcohol, and separate biochemical tests were run to directly assess the ADH enzyme’s activity. To their surprise, they found that the ancestral enzyme performed no differently to the modern enzyme! Short of errors reconstructing the ancestral sequence (although the method itself is known to be reliable), this is clear evidence that the protein-coding gene sequence of the ADH enzyme has not experienced adaptive evolution to provide increased alcohol tolerance.
As good scientists, the researchers didn’t stop there – they wanted to find out why exactly there results were at odds with the results of the seminal MK test. First they tested whether or not the original MK test had been done with insufficient polymorphism data, so they re-did the original MK test with an expanded dataset, and although the signature of selection was much less significant (P = 0.049 as opposed to P = 0.0007 in the original 1991 test), it remained just below the P = 0.05 threshold, as was therefore still technically statistically significant.
Next, they tested to see if the signature of selection was coming from lineages other than D. melanogaster, because as I mentioned earlier, the general test alone can’t give us information about a specific lineage. This including performing the MK test on just the sister species of D. melanogaster alone: D. yakuba and D. simulans. They found that the signature of selection was in fact coming from D. yakuba, not D. melanogaster. This signature is also very weak (P = 0.047), and is unlikely to represent adaptive evolution toward alcohol tolerance as the species is less tolerant to alcohol than its sister species.
So, the takeaway message here is that in this specific application of the MK test, McDonald and Kreitman were primarily limited by the dataset and methodologies they had available to them, although they probably should have also run the test on subsets their 3 species, and Thornton’s team in for the 2017 paper, because they would have found that the subset which excluded D. melanogaster also gave a statistically significant signature of selection! This is a good lesson in hypothesis testing: the size of your dataset can have a big impact on the statistical significance of your results, and it’s important to employ rigorous tests. Just because there is a good correlation between the data, doesn’t mean the narrative itself is necessarily correct. In this case, the narrative was that the Adh gene itself underwent adaptive evolution in the lineage that led to D. melanogaster, as this fit with the data that D. melanogaster had a higher alcohol tolerance.
Does the fact that this narrative turned out to be wrong change the fact that the D. melanogaster lineage evolved a higher alcohol tolerance? Not at all, multiple papers have been published previously that suggested that this trait is actually due to increased expression of Adh through changes in regulatory regions (e.g. this one and this one) controlling the gene, or even other genes altogether (e.g. this one and this one). As you can see in these papers, the idea that differences in the Adh gene were responsible for D. melanogaster‘s heightened alcohol tolerance has been waning for years, this 2017 paper was just the final conclusive nail in the coffin.
Does the 2017 paper cast doubt on the validity of the MK test as a conceptual basis for inferring adaptive evolution? Not in the slightest. As I explained earlier, the authors actually demonstrated that the Adh gene in D. melanogaster actually fails the MK test, which is consistent with their experimental results that show that it didn’t undergo adaptive evolution. The MK test and more advanced methods based on the same principle have been used in thousands of rigorous studies over the last 25 years, and they have demonstrated unequivocally that they’re extremely useful and reliable.
In summary then, the 2017 study from Thornton’s lab is certainly an important one, and lessons can be learned from it, but it does nothing to cast doubt on the theory of evolution as creationists and ID proponents would have you believe.
Comments and queries are welcome.