Deciphering the logic of mammalian mitochondrially-localised translation

Localised translation broadly enables spatiotemporal control of gene expression yet currently there are limited tools for monitoring subcellular translation in mammalian cells. Here we present LOCL-TL (LOV-domain Controlled Ligase for Translation Localisation) an optogenetic approach for labelling ribosomes at any defined subcellular location and quantitatively determining their translation with codon precision through ribosome profiling. Application of LOCL-TL to mitochondrially localised translation revealed that ~20% of human nuclear-encoded mitochondrial genes are robustly translated on the outer mitochondrial membrane (OMM). Mitochondrially-translated messages form two classes distinguished by the length of proteins they encode, and the mechanism, and function of their recruitment. An evolutionarily ancient mechanism allows nascent chains to drive the cotranslational recruitment of long proteins via a bipartite targeting signal, which acts on mostly mitochondrial matrix proteins of bacterial origin. Conversely, mRNAs of short proteins, especially eukaryotic origin electron transport chain (ETC) components, are directly recruited to the OMM, in a manner that depends on the 3’UTR, mRNA splicing and the OMM protein A-Kinase Anchoring Protein 1 (AKAP1) and disruption of localised translation through the loss of AKAP1 lowers ETC production. Our studies thus reveal a hierarchical strategy used to enable preferential translation of a subset of proteins on the OMM.