What inversion lengths can tell us about their evolution

Abstract

Genome structural differences, such as inversions, are extremely common between species and within populations. Theoretical models of how and why such inversions evolve have affirmed that they are able to evolve under both adaptive and non-adaptive scenarios (reviewed in Kirkpatrick (2010)). What has remained difficult, however, is distinguishing these scenarios from each other. In this issue of Molecular Ecology, Connallon and Olito (2022) present a model that examines how adaptive and non-adaptive scenarios lead to different distributions of inversion sizes. The authors present several important predictions including an expectation that larger inversions should evolve under local adaptation scenarios and much smaller inversions should evolve when they are either underdominant or directly beneficial. Finally, the authors ask how the presence of deleterious mutations within populations affects the probability of fixing inversions of different types. The paper is therefore an important step in synthesizing decades of inversion theory.

It is widely agreed that inversions contribute to both speciation and local adaptation (Faria et al., 2019). An inversion can have both direct effects (by changing the coding sequence or expression of genes at or near the inversion’s breakpoints) or indirect effects by suppressing recombination between the alleles it captures. In turn, there has been great interest in understanding the selective forces that act on inversions. Four broad categories of inversion models have emerged. Inversions may simply be neutral, underdominant (with lower fitness in heterozygotes than either homozygous standard or homozygous inversion genotypes), directly beneficial due to effects of their breakpoints, or indirectly beneficial by capturing a favorable set of alleles (Figure 1B). In the last scenario, a case that has received a lot of attention is when an inversion captures a set of locally adapted alleles. Because inversions reduce recombination rates in heterozygotes, this allows a sets of co-adapted alleles to be maintained together.

Figure 1:

A) The four types of inversions considered by Connallon and Olito. B) How inversion fitness is predicted to change with inversion size for each of the models. C) A toy example of how combining the probability that an inversion of some length occurs with the probability that it fixes allows Connalon and Olito to describe the distribution of inversion sizes that should be observed under each of the processes.

While many potential explanations for the evolution of inversions now exist, it has so far been difficult to distinguish between these processes. Inversions that are underdominant, directly beneficial, or that carry locally adapted alleles have all been identified (see Wellenreuther and Bernatchez (2018) for a review), and sometimes all three processes are found in the same system, such as the inversions of Drosophila pseudoobscura (Schaeffer, 2008), but what processes are responsible for the overall distribution of inversions across a genome? Connallon and Olito turn to the distribution of inversion lengths to distinguish between the processes that have fixed them. They note that expected inversion fitness varies with inversion size (where size is the fraction of the chromosome the inversion captures) depending on the evolutionary scenario (Figure 1B). By combining analytical models of the probability that an inversion will be fixed with the probability that an inversion of a particular length will occur, Connallon and Olito are able to predict the distribution of inversion lengths across the genome that arises from each of the four processes (Figure 1C). Underdominant inversions are likely to be the smallest, as larger inversions have increasingly deleterious effects. Neutral and beneficial inversions are also likely to be small, primarily since smaller inversions are simply more likely to occur, and the effect of the inversion does not depend on its size in either model. On the other hand, small inversions are unlikely to capture multiple locally adapted alleles, and so local adaptation drives the evolution of much larger inversions.

Connallon and Olito additionally explore a previously under-studied factor influencing the fate of inversions – the capture of deleterious mutations. Many slightly deleterious mutations remain polymorphic across the genome due to both drift and new deleterious mutations being constantly introduced. A large inversion that is free of rare deleterious mutations will have a fitness advantage over the average haplotype in the population, but larger inversions are also expected to capture more deleterious mutations. By incorporating this complexity into models of inversion fixation, Connallon and Olito demonstrate that deleterious mutations decrease the overall size of inversions expected to fix under most scenarios, but might actually lead to larger, strongly underdominant inversions fixing in populations with a high level of genetic load. A key result is that deleterious mutations can cause a lack of large inversions even when inversions are directly beneficial. In turn, populations or chromosomes with a high level of genetic load should show fewer large inversions. Since this result depends on the density of deleterious mutations per chromosome arm, if it is true, species which have smaller chromosome arms should have relatively larger inversions (proportional to the total size of the chromosome). The distribution of inversion lengths is therefore highly indicative of the processes that shaped their evolution.

Connallon and Olito synthesize a very broad set of theoretical results to present both predictions for empirical data and directions for future theoretical work. The manuscript does an excellent job of summarizing existing data-sets of inversion lengths and outlining what theoretical threads remain untied. But it also opens the door to testing which processes are more common in inversion evolution. In nature, species are likely to harbor inversions of all four types, as well as additional ones that capture hybrid incompatibility loci (Navarro & Barton, 2003), loci associated with mating preference (Trickett & Butlin, 1994) or both (Dagilis & Kirkpatrick, 2016). Previous studies to test the distribution of fixed inversion size (for example, Cheng and Kirkpatrick (2019)) have used simple assumptions about the relationship between inversion size and fitness coupled with simulations to produce probability densities of various inversion sizes. Incorporating the results of Connallon and Olito into a framework like Cheng and Kirkpatrick would potentially allow researchers to test the relative frequency of each of the four processes. The paper has therefore opened the gates to determining what evolutionary forces have driven the evolution of inversions.