Site-and-branch-heterogeneous analyses of an expanded dataset favour mitochondria as sister to known Alphaproteobacteria.

Determining the phylogenetic origin of mitochondria is key to understanding the ancestral mitochondrial symbiosis and its role in eukaryogenesis. However, the precise evolutionary relationship between mitochondria and their closest bacterial relatives remains hotly debated. The reasons include pervasive phylogenetic artefacts as well as limited protein and taxon sampling. Here we developed a new model of protein evolution that accommodates both across-site and across-branch compositional heterogeneity. We applied this site-and-branch-heterogeneous model (MAM60 + GFmix) to a considerably expanded dataset that comprises 108 mitochondrial proteins of alphaproteobacterial origin, and novel metagenome-assembled genomes from microbial mats, microbialites and sediments. The MAM60 + GFmix model fits the data much better and agrees with analyses of compositionally homogenized datasets with conventional site-heterogenous models. The consilience of evidence thus suggests that mitochondria are sister to the Alphaproteobacteria to the exclusion of MarineProteo1 and Magnetococcia. We also show that the ancestral presence of the crista-developing mitochondrial contact site and cristae organizing system (a mitofilin-domain-containing Mic60 protein) in mitochondria and the Alphaproteobacteria only supports their close relationship.