Evolutionary inference across eukaryotes identifies specific pressures favoring mitochondrial gene retention

Since their endosymbiotic origin, mitochondria have lost most of their genes. Despite the central cellular importance of mtDNA, selective mechanisms underlying the evolution of its content across eukaryotes remain contentious, with no data-driven exploration of different hypotheses. We developed HyperTraPS, a powerful methodology coupling stochastic modelling with Bayesian inference to identify the ordering and causes of evolutionary events. Using a large dataset from over 2000 complete mitochondrial genomes, we inferred evolutionary trajectories of mtDNA gene loss across the eukaryotic tree of life. We find that proteins fulfilling central energetic roles in the assembly of mitochondrial complexes are prefentially retained in mitochondria across eukary-otes, biophysically supporting colocalization for redox regulation. We also provide the first quantitative evidence that a combination of GC content and protein hydrophobicity is required to explain mtDNA gene retention, and predicts the success of artificial gene transfer experiments. Our results demonstrate that a combination of these three characteristics explains most mtDNA gene variability, addressing the long-debated question of why particular genes are retained in mitochondrial genomes.