Is artificial selection a useful analogy for natural selection?
You will recall that Darwin presented examples of artificial selection as evidence that natural selection could produce the same effect on a species: a shift in average characteristics within a population over time. Darwin, of course, had no idea about how heredity worked. Now that we have access to genome sequences, however, we have the opportunity to test the strength of Darwin’s analogy at the molecular level. To do that, however, we need to compare the molecular details of both types of selection events. Interestingly, the dog genome also shows signs of natural selection during the early domestication process, which we can compare to examples of artificial selection.
Meat or potatoes? Natural selection during early dog domestication
The very same study (which we discussed yesterday) that demonstrated significant selection for variation in nervous system genes in dogs during the early domestication process also identified selection on a class of genes involved in starch metabolism. Starch is a long chain of sugar (glucose) molecules strung together that plants use as an energy storage mechanism. Wolves do not eat a diet with high amounts of starch, but do ingest some in the wild fruit they sometimes consume (e.g. berries). In order to benefit from starch, mammals use a class of enzymes called amylases that break the starch chain down into separate glucose molecules. In wolves, amylase is produced in the pancreas. Like wolves, dogs also have a pancreatic amylase enzyme—but instead of having just one gene, all modern dogs have between 2 and 15 copies of this gene, whereas all wolves have only one. The dog copies sit side-by-side in the dog genome, right next to where the original amylase gene is found in wolves, indicating that the gene copies were duplicated during chromosome replication. These extra amylase gene copies greatly increase the amount of amylase enzyme that dogs make compared to wolves, and allow dogs to benefit from a high-starch diet in a way that wolves never could.
This diet, of course, is a distinctly human diet: large-scale use of starchy plants is a feature of human agriculture, which comes on the scene about 10,000 years ago, give or take. The association of dogs with humans at this time thus seems to have provided a selective advantage to dogs that could derive more benefit from the food they were receiving (or scavenging) from human sources. In other words, the environment the dogs were in (access to high starch foods associated with proximity to human settlements) provided the selection: dogs that could derive increased nutritional benefit from starch would be able to reproduce at a greater frequency than dogs that could not. Over time, the average ability of dogs to metabolize starch would improve in the population, as more and more dogs would have duplications. Since no conscious human choice was made to identify dogs with duplicated amylase genes and select them for breeding (since the characteristic would not be readily observable) this is an example of environmental, or natural selection. Even today humans do not have the ability to select dogs for increased starch metabolism, despite the fact that modern dogs remain variable for the number of amylase gene copies they have.
From dog to dachshund: artificial selection for a new gene
Having domesticated wolves into dogs, our selective efforts didn’t stop there, of course: humans used artificial selection to create 400-plus dog breeds, and we can see the effects of this selection on the dog genome when we compare different breeds to one another. One striking example of a difference between breeds is leg length. Some breeds are defined by an unusually short leg without a proportional reduction in body size, such as in basset hounds and dachshunds. This trait, known by the technical term “chondrodysplasia,” is one that was selected for based on its utility for certain hunting roles, such as the pursuit of burrowing animals. The genetic basis of this trait turns out to be another duplication event that happened once within domestic dogs: all short leg breeds share the same genetic innovation.
Note: the molecular details of this duplication event are somewhat more complex than for the simple side-by-side duplications of the amylase gene that we have just discussed. I’ve included the details below for those readers who are interested in digging into the finer points. The take-home message, however, is simple: in this case, a new trait (shortened legs) came into being when a gene was duplicated and the copy acquired new properties during the duplication event. The shortened leg trait was noticed and intentionally selected for by human breeders, making this an example of artificial selection leading to breeds with a specific characteristic. If the details of the duplication event itself are not of interest, feel free to skip down and pick up the story under the “Comparing artificial and natural selection in dogs” heading.
The molecular basis of how this new gene responsible for chondrodysplasia in dogs came to be is very similar to something we have discussed in detail before: the new gene is in effect a processed pseudogene that acquired a function at the time of its duplication. Since the new copy was functional from the beginning, it is not best described as a pseudogene, but rather as a retrogene: the mRNA copy of a gene that was (a) reverse transcribed from mRNA back into DNA, (b) inserted into the genome next to sequences that could direct its expression, and (c) came under selection to maintain it:
The new retrogene turns out to be a copy of the fgf4 gene, an important regulator of growth and development. For reasons that are not yet clear, the fgf4 retrogene interferes with normal bone growth and causes the short leg trait in chondrodysplastic dogs:
Another interesting feature of the new fgf4 retrogene is that its regulatory sequence (shown in yellow in the figure above) is derived from a transposon. As we have discussed before, transposons are autonomous, self-replicating DNA parasites that can, on occasion, become part of the genes of their hosts – and this is another example. The new fgf4 retrogene is “cobbled together” from transposon sequences and the original fgf4 coding sequence and has a new function (attenuating leg growth) not seen in either the original fgf4 gene or the transposon sequence.
Comparing artificial and natural selection in dogs
While the mutation that led to shortened legs in some dog breeds is a particularly dramatic example of a new variation arising (since it involves the birth of what is effectively a new gene), there were many other genomic regions selected during the creation of dog breeds. Other, more mundane examples abound: the small body size common to the “toy dog” group is determined by selected variation in the Insulin-like growth factor 1 (IGF1) gene; variation in three key genes have been identified as responsible for coat color variation; and even variation in a gene that is responsible for the characteristic skin wrinkling seen in the Shar-Pei has been described. In these cases, it is not the production and selection of duplicated or new genes that is responsible for these traits, but rather small mutations in existing genes that alter the function of those genes compared to the ancestral state in wolves or early dogs. Again, the main theme is clear: small changes in DNA, combined with artificial selection, can add up to large changes in form within a population in a short amount of time.
So, how comparable are natural selection and artificial selection? Both forms of selection have the ability to shift average characteristics of a population over time. Additionally, the molecular underpinnings of the mutation events for both types of selection events are comparable: the trait must arise through a mutation that produces a new heritable variant in the population. Note well: in the case of artificial selection, human intelligence and agency cannot produce the variation needed, but only select for it once it arises. In both cases, we have examples of small sequence mutations, duplications, and even new genes arising. Indeed, the only difference is the selection step—the filter that allows some variants to reproduce preferentially over others.
Human agency might be an efficient form of selection, but so, too, is nature. As such, Darwin’s use of artificial selection as evidence for selection in nature remains a valid approach, even at the molecular level.
In the next post in this series, we’ll explore how natural selection has shaped amylase function in the human lineage as well as in dogs.
For further reading:
Akey, J.M., et al. (2010). Tracking the footprints of artificial selection in the dog genome. PNAS 107; 1160 – 1165. (link)
Axelsson, E., et al. (2013). The genomic signature of dog domestication reveals adaptation to a starch-rich diet. Nature 495; 360 – 364. (link)
Parker, H.G., et al. (2009). An Expressed Fgf4 Retrogene Is Associated with Breed-Defining Chondrodysplasia in Domestic Dogs. Science 325: 995 – 998. (link)