We often think of evolving traits as phenotypic variables, such as wingspan, spitting velocity, and pupil geometry (holy cow, check out cuttlefish eyes!). One of the challenges of understanding evolution is determining what makes a trait meaningful for a research question. Quantitative genetics literature in general gives limited guidance on this point (Lynch and Walsh, 1998), as any quantitative trait simply needs to be a quantifiable quality of an organism, and not a discrete class; it needs to have a continuous distribution and be measurable on a scale. For a trait to be applicable to evolutionary quantitative inquiry, it needs to be heritable. Beyond these requirements, the selection of which trait to use depends on what the researcher is trying to understand about the response of traits to natural selection and/or population history.
What often is missed is that the degree to which such traits are able respond to selective pressures is itself an evolvable trait. This measure is (appropriately) referred to as the “evolvability” of the trait, and is generally described as the expected generational response of a trait to directional selection (Hansen 2003, Hansen & Houle 2008, Bolstad et al. 2014). That is, a trait could be subject to natural selection, but it may not be able to respond to that selection. As I have blogged about previously, genetic covariance between traits can function to alter the response of those traits to directional selection. Though measuring evolvability is fairly straightforward for a single trait, interactions between traits make the calculation of multivariate evolvability considerably more complex.
Hansen and Houle (2008) devised several measures of multivariate evolvability, four of which are explained by Bolstad and colleagues (2014): Evolvability is the ability of a trait to change over generations in the direction of selection. A related concept is respondability, which is the freedom of a trait to change in response to selection. Some traits may change over time, but because of genetic covariance with other traits, cannot change in the direction of selection. Other traits may affect fitness, but are unable to change for a variety of reasons. Stabilizing selection on a trait or covarying traits might be one reason. We can measure conditional evolvability to determine the ability of a trait to change in the direction of selection when stabilizing selection restricts change in other directions. We may further ascertain if constraints imposed by other traits are restricting evolvability through measuring autonomy, which is the proportion of evolvability that remains after conditioning on other traits.
As Hansen and Houle (2008) described them, these measures of evolvability are calculated based on the relationship between a genotypic or phenotypic variance-covariance matrix (which represents the correlations between traits) and a given selection gradient (the partial regression of a trait on fitness) (see the vector diagram above). Because these measures provide insight into how correlated traits respond to selection, they are often calculated for and compared between different species to examine long-term evolutionary patterns (for example, Marriog et al. 2009, Young et al. 2010, Grabowski et al. 2011). Since we have evidence that the correlations between measures of body form have strongly affected human evolution across ecogeographic regions (Savell et al. 2016), I decided to see if the measures of evolvability between these human populations would notably differ.
Perhaps unsurprisingly, they don’t. In humans, the covariation among traditional measures of body form (limb segment lengths, bi-iliac breadth, and femoral head diameter) was consistent enough for the measures of evolvability to stay centered around a common average. We might conclude, then, that any human group migrating to a different ecogeographic region (latitude), if subjected to the same selective pressures humans encountered before, would demonstrate the same ability to respond to selection as other groups. There are some fairly extreme outliers within each measure of evolvability calculated, but these do not appear to follow an obvious pattern, and more importantly are not the same across estimated evolvability, respondability, conditional evolvability, and autonomy. It would appear that for some groups, there is a difference in how traits are able to respond to selection, to but understand why will require further exploration, which will be the subject of a future post. Stay tuned!
Bolstad GH, Hansen TF, Pélabon C, Falahati-Anbaran M, Pérez-Barrales R, Armbruster WS. 2014. Genetic constraints predict evolutionary divergence in Dalechampia blossoms. Phil Trans R Soc B. 369: 20130255.