Mitonuclear conflict and cooperation govern the integration of genotypes, phenotypes and environments

Abstract

The mitonuclear genome is the most successful co-evolved mutualism in the history of life on Earth. The cross-talk between the mitochondrial and nuclear genomes has been shaped by conflict and cooperation for more than 1.5 billion years, yet this system has adapted to countless genomic reorganizations by each partner, and done so under changing environments that have placed dramatic biochemical and physiological pressures on evolving lineages. From putative anaerobic origins, mitochondria emerged as the defining aerobic organelle. During this transition, the two genomes resolved rules for sex determination and transmission that made uniparental inheritance the dominant, but not a universal pattern. Mitochondria are much more than energy-producing organelles and play crucial roles in nutrient and stress signalling that can alter how nuclear genes are expressed as phenotypes. All of these interactions are examples of genotype-by-environment (GxE) interactions, gene-by-gene (GxG) interactions (epistasis) or more generally context-dependent effects on the link between genotype and phenotype. We provide evidence from our own studies in Drosophila, and from those of other systems, that mitonuclear interactions-either conflicting or cooperative-are common features of GxE and GxG. We argue that mitonuclear interactions are an important model for how to better understand the pervasive context-dependent effects underlying the architecture of complex phenotypes. Future research in this area should focus on the quantitative genetic concept of effect size to place mitochondrial links to phenotype in a proper context. This article is part of the theme issue 'Linking the mitochondrial genotype to phenotype: a complex endeavour'.

Keywords: GxE; GxG; conflict; cooperation; epistasis; mitonuclear.

Figure 1.

Mitochondrial effects are dominated by epistatic and environmental interactions. (a) illustrates a mitonuclear epistatic (GxG) interaction for a phenotype. The three nuclear genotypes at an autosomal locus (NN, Nn, nn) might show different norms of reaction across alternative mtDNA backgrounds (M and m). (b) Different mitonuclear genotypes (denoted as mtDNA; nuclear genotype) may have different phenotypes in different sexes. Here, the sex of the organisms provides a distinct environment for the genotype. (c) A typical genotype-by-environment (GxE) interaction where alternative genotypes have different phenotypes across a range of environments. The relative contributions of mtDNA and nuclear genes to these phenotypes are context dependent. Importantly, (d) demonstrates that the effect size for a given mtDNA or genotype might be large in any one environment but may be very small when averaged across all genetic and environmental backgrounds encountered in nature (i.e. marginal effect sizes are small). (Online version in colour.)

Figure 2.

Mitonuclear phenotypic landscapes. (a) A number of genetically tractable model organisms should be introgressed to test the phenotypic effects of three axes of variation: (i) nuclear DNA (ΔnDNA), (ii) mitochondrial DNA (ΔmtDNA) and (iii) environment (ΔEnv.). A combination of genome DNA sequencing, RNA-seq(uencing) and whole-organism phenome measurements can help construct a comprehensive map of the phenotypic landscape. (b) shows five simulated datasets of 6400 mitonuclear genotypes (80 mtDNAs × 80 nDNAs), responding randomly to diet (alternative colours). Regions of these landscapes that are modified by environment (b) are associated with GxGxE loci. Comprehensive sequencing of these lines can help identify associated modifying (or sensitive) loci. Likewise, gene–gene or protein–protein interaction networks (c) with overlapping across-taxa hotspots of mitonuclear dysregulation can pinpoint mtDNA-sensitive genes or sub-networks. This is ‘guilt-by-association’ across evolutionarily divergent species, to map evolutionary conserved networks and pathways [73]. These identified regions (red hotspots in (c)) can then be genetically manipulated to narrow down the search space for mitonuclear epistasis and/or GxGxE for forward genetics confirmation. Shown is a subsection of the Drosophila protein–protein interaction network [79]. (Online version in colour.)

Figure 3.

‘Candidate’ mitonuclear epistasis mapping. The YARS2 gene (the human homologue of Drosophila Aatm) is shown in its dimeric crystal form in (a). We have highlighted the causative amino acid polymorphism associated with a large-effect negative mitonuclear epistasis in Drosophila. The SNP is highly conserved across metazoa yet mutated in the negative epistatic OregonR nuclear background. The tRNA^Tyr molecule docks the YARS2 protein and is presumably modified by the polymorphic locus. Different coloured regions represent different protein domains (labelled in (c)). Given this knowledge, mitochondrial aminoacyl-tRNA synthetases (mt-aaRSs) may be rewarding proteins to use for targeted (candidate) perturbation. There are many ways to conduct this, and we show two methods. (b) First, highly conserved mt-aaRS gene sequences across taxa can be screened in genetic panels of fully sequenced nuclear backgrounds. Any genetic lines with non-synonymous amino acid polymorphisms at conserved loci would be good candidates to introgress with a panel of mtDNA haplotypes demonstrating variation in the cognate mtDNA-encoded tRNA gene. (c) Alternatively, targeted knock-in can be used at any regions of the gene of interest, e.g. YARS2 (shown in (a,c)). To accomplish this, in silico knock-ins could be made to select the highest likelihood negative perturbation [114]. Red stars in (c) show example targets for mutation via CRISPR. In figure 3d, the mitonuclear panel of genotypes is represented by different mtDNA haplotypes (columns) and nuclear backgrounds with specific mt-aaRS polymorphisms (rows). The panel can be scaled-up to include as many mt-aaRSs and mtDNA haplotypes as required. These constructs may produce lethal offspring, but both approaches would identify mitonuclear interacting loci with effects on the phenome. MTS: mitochondrial targeted sequence (pink), catalytic: catalytic domain (red), insertion: insertion domain (orange), ACB: alpha-helical domain (green), S4-like: anticodon binding domain (blue). (Online version in colour.)

Figure 4.

Empirical and simulated phenotypic landscapes. Dietary composition of altered protein (P): carbohydrate (C) geometry under isocaloric conditions alters development time in mitonuclear Drosophila strains (a–d) (data from [26]). Each figure contains a heat component corresponding with the magnitude of the phenotype (development time). The largest main effect is diet type (high P:C (a), equal P:C (b), low P:C (c) and laboratory food (d)) are shown. The second most important explanatory variable is the nuclear effect followed by mtDNA variation. Mitonuclear epistases are shown as mtDNA modification of nuclear effects. Mitonuclear epistasis can be both positive and negative. A simulated phenotypic landscape of mitonuclear variation (80 × nDNAs by 80 × mtDNAs = 6400 mitonuclear genotypes) in a single environment (e). The heat component corresponds with the magnitude of a phenotype. The major axis shows nuclear variation (nDNA), which consistently explains more of the variance in measured phenotypes than mtDNA (mtDNA). Mitonuclear epistases are shown as sharp peaks and troughs in the landscape. (Online version in colour.)