The on-again, off-again relationship between mitochondrial genomes and species boundaries

Abstract

The study of reproductive isolation and species barriers frequently focuses on mitochondrial genomes and has produced two alternative and almost diametrically opposed narratives. On one hand, mtDNA may be at the forefront of speciation events, with co-evolved mitonuclear interactions responsible for some of the earliest genetic incompatibilities arising among isolated populations. On the other hand, there are numerous cases of introgression of mtDNA across species boundaries even when nuclear gene flow is restricted. We argue that these seemingly contradictory patterns can result from a single underlying cause. Specifically, the accumulation of deleterious mutations in mtDNA creates a problem with two alternative evolutionary solutions. In some cases, compensatory or epistatic changes in the nuclear genome may ameliorate the effects of mitochondrial mutations, thereby establishing coadapted mitonuclear genotypes within populations and forming the basis of reproductive incompatibilities between populations. Alternatively, populations with high mitochondrial mutation loads may be rescued by replacement with a more fit, foreign mitochondrial haplotype. Coupled with many nonadaptive mechanisms of introgression that can preferentially affect cytoplasmic genomes, this form of adaptive introgression may contribute to the widespread discordance between mitochondrial and nuclear genealogies. Here, we review recent advances related to mitochondrial introgression and mitonuclear incompatibilities, including the potential for cointrogression of mtDNA and interacting nuclear genes. We also address an emerging controversy over the classic assumption that selection on mitochondrial genomes is inefficient and discuss the mechanisms that lead lineages down alternative evolutionary paths in response to mitochondrial mutation accumulation.

Keywords: co-evolution; cytonuclear interactions; introgressive hybridization; mitochondrial introgression; mutation accumulation; reproductive isolation.

Figure 1.

Bateson-Dobzhansky-Muller incompatibilities and mitonuclear epistasis. A) The BDMI model applied to a mitonuclear interaction. N₁ and N₂ are the ancestral and derived nuclear alleles, respectively. M₁ and M₂ are the ancestral and derived mitochondrial alleles, respectively. For simplicity, nuclear genotypes are written as haploid. The curved, dotted line indicates hybridization and admixture, which produces all possible combination of mitonuclear genotypes. The genotype that had not previously been “tested” by selection is boxed in red. B) Although BDMIs are conventionally described as resulting from substitutions that occur in parallel in two different populations (see panel A), the same logic can apply to two sequential substitutions within the same population (Orr 1995). This distinction is important because much of the theory related to mitonuclear interactions involves coevolutionary pressures that are expected to produce mitochondrial and nuclear changes in the same lineage. Once again, the curved, dotted line indicates hybridization/admixture, and the previously “untested” mitonuclear genotype is boxed in red. C) Examples of mitonuclear epistasis in which an initial mitochondrial change (from M₁ to M₂) can make a nuclear change (from N₁ to N₂) that previously would have been deleterious now be beneficial or at least neutral. The ancestral and derived nuclear alleles are shown in gray and blue, respectively. Black arrows indicate the direction of evolutionary change. The examples differ based on the fitness effects of the initial mitochondrial change (deleterious, neutral, or beneficial) and of the subsequent nuclear changes. The examples do not represent a comprehensive set of all possible arrangements, but any of these paths (and others) would arrive at coadapted mitonuclear genotypes that can be disrupted by hybridization.

Figure 2.

An example of intimate interactions between mitochondrial and N-mt proteins: 3D structure of the OXPHOS complex IV (cytochrome c oxidase) isolated from bovine heart tissue (PDB accession 1OCC). Three mitochondrially encoded subunits (COX1, COX2, and COX3) are shown in yellow, while 10 nuclear-encoded subunits (COX4, COX5A, COX5B, COX6A2, COX6B1, COX6C, COX7A1, COX7B, COX7C, COX8B) are shown in green. COX5 subunits, which have been implicated in cointrogression with an mtDNA haplotype that has moved between Drosophila species (Beck et al. 2015), are drawn with green spheres. The remaining subunits are drawn with ribbons.

Figure 3.

Barriers to cointrogression for unlinked loci. A deterministic (i.e., infinite population size) and random- mating model for the introgression of mitochondrial (orange) and nuclear (blue) alleles after a one-time admixture event that introduces foreign alleles at a frequency of 5%. The two plots show the same results on two different timescales (note the differences in scales for both axes). The introduced mitochondrial haplotype is favored by selection and undergoes adaptive introgression. The selection coefficient for the introduced mtDNA (s_m) is 0.05. The solid lines depict a scenario in which the nuclear locus is neutral. Under this scenario, the introduced nuclear allele increases in frequency over the first few generations by hitchhiking with the introgressing mtDNA. But its frequency plateaus as mitonuclear LD is rapidly eliminated by random mating. The dotted lines depict a scenario in which there is a symmetrical mitonuclear incompatibility between the native and introduced genotypes. The epistatic selection coefficient (s_i) acting against “mismatched” mitonuclear genotypes is −0.02. These incompatibilities are assumed to exhibit incomplete dominance, so heterozygotes experience half of this cost (0.01). Under this scenario, the rate of spread of the introduced mtDNA is slower because of its negative interactions with native nuclear alleles, but it nevertheless introgresses rapidly because of the strong selection on it (s_m is still 0.05). After the short initial hitchhiking phase, the introduced nuclear allele that is “matched” to the introgressing mtDNA actually starts to decline in frequency because of its harmful interactions with the native mtDNA, which remains far more abundant in early generations. The introduced nuclear allele does not begin to increase in frequency until the foreign mtDNA spreads to >50% in the population, creating a very long lag. Under a more realistic model with a finite population size, the introduced nuclear allele would be at risk of stochastic loss from the population because of its early decline in frequency and the long lag before there is any selection in favor of it. The code used to generate these models is provided as supplementary material.

Figure 4.

Accelerated and variable rates of protein sequence evolution in mitochondria. Branch lengths are scaled based on the proportion of amino acid substitutions in translated cox1 genes sequences from select bacterial and mitochondrial genomes (gray and black branches, respectively). Sequences were aligned with MAFFT (E-INS-i), and divergent N- and C-terminal portions of the alignment were manually trimmed. Branch length estimates were then generated in PAML v4.9a, using an LG+G model and a topology that was constrained based on published sources (Burki 2014; Lavrov & Pett 2016).

Box 2 Figure.

An example of introgression that is consistent with the possibility of mitochondrial mutation load selecting for adaptive replacement. The recipient species (Herichthys minckleyi) exhibits significantly higher nonsynonymous divergence than the donor species (H. cyanoguttatus) based on Tajima’s relative rate test (p = 0.009). Herichthys minckleyi also has a lower N_e and shows evidence of relaxed selection on the mitochondrial genome (Hulsey et al. 2016). Nonsynonymous branch lengths and w values (i.e., d_N/d_S) were estimated in PAML using a concatenation of all mitochondrial protein coding genes. The arrow indicates the direction of mitochondrial introgression. The cichlid image is from PhyloPic (M. Tan).