Selection for biparental inheritance of mitochondria under hybridization and mitonuclear fitness interactions

Abstract

Uniparental inheritance (UPI) of mitochondria predominates over biparental inheritance (BPI) in most eukaryotes. However, examples of BPI of mitochondria, or paternal leakage, are becoming increasingly prevalent. Most reported cases of BPI occur in hybrids of distantly related sub-populations. It is thought that BPI in these cases is maladaptive; caused by a failure of female or zygotic autophagy machinery to recognize divergent male-mitochondrial DNA 'tags'. Yet recent theory has put forward examples in which BPI can evolve under adaptive selection, and empirical studies across numerous metazoan taxa have demonstrated outbreeding depression in hybrids attributable to disruption of population-specific mitochondrial and nuclear genotypes (mitonuclear mismatch). Based on these developments, we hypothesize that BPI may be favoured by selection in hybridizing populations when fitness is shaped by mitonuclear interactions. We test this idea using a deterministic, simulation-based population genetic model and demonstrate that BPI is favoured over strict UPI under moderate levels of gene flow typical of hybridizing populations. Our model suggests that BPI may be stable, rather than a transient phenomenon, in hybridizing populations.

Keywords: bet-hedging; biparental inheritance; heteroplasmy; hybrid; mitonuclear; rare-allele.

Figure 1.

A schematic of the population under examination in the model. Movement of diploid cells occurs between demes (with a maximum migration distance of 1 deme per generation) and mating occurs within demes. As in this schematic, figures throughout the text display the ‘AA’ home range on the left. Reference to the ‘cline’ in the text refers to the zone where both genotypes are represented (at least at a meaningful frequency), rather than the entire 50-deme stretch. (Online version in colour.)

Figure 2.

Mean frequency of the u allele throughout the entire cline at equilibrium or after 75 000 generations; whichever came first. The u allele is injected into a metapopulation of hybridizing cells at a frequency of 0.01 after migration/selection equilibrium has been reached along the cline. It is then able to evolve freely. For the parameter space explored, u invades more readily under the heteroplasmic-advantage model than the risk-avoidance model. Yellow represents a high equilibrium frequency of u and blue represents little or no invasion, as denoted by the scale on the right-hand side of the figure. In each case, the conditions where u reaches a high equilibrium frequency correspond to areas of low selection and higher migration, where gene flow is high. Parameter values: diploid mitochondrial number M = 100, number of demes C = 50. (Online version in colour.)

Figure 3.

Comparison of the fitness of U and u individuals aggregated over specific N-mt genotypes and over the entire population. (a) In regions of the cline where the A allele is more common, u carries a cost to AA individuals and in regions of the cline where the A allele is less common, u provides an advantage. (b) u individuals have lower variance in fitness in the cline centre since they have a much lower chance of producing either a mismatched ‘low fitness’ genotype or a fully matched ‘high fitness’ genotype. (c,d) Mean normalized fitness of the U and u alleles averaged across all genotypes under low frequency of u (c) and high frequency of u (d). u offers a net advantage in the middle two demes (25 and 26) and a disadvantage in the demes either side of the centre. This trend is exaggerated as the frequency of u increases (from c,d). While the difference in the cline centre increases roughly linearly with increasing frequency of u (red line, (e)), the difference outside the cline centre increases quadratically (blue line, (e)). Parameter values: M = 100, C = 50, migration rate h = 0.1, selection strength s = 0.2 (moderate gene flow). Fitness curves: ‘risk-avoidance’. (Online version in colour.)

Figure 4.

A schematic of how, through recombination, an AU gamete in its ‘home range’ risks eventually producing a totally mismatched aaU zygote when paired with a rare a allele. This is contrasted with the lower variance pathway of an Au gamete following the same path but producing descendants all of intermediate fitness. This scenario is played out in the ‘A side’ of the cline, where A and 1 haplotypes (represented by blue circles) are the most common and a and 0 haplotypes (red circles) are relatively rare. (Online version in colour.)