A mitochondrion carries its own DNA, and the arrangement of that genome explains why RNA-level control matters there more than almost anywhere else. Human mitochondrial DNA spans 16,569 base pairs and encodes 37 genes: 22 transfer RNAs, 2 ribosomal RNAs and 13 subunits of the oxidative phosphorylation machinery, distributed around a circular molecule with a control region that carries the promoters for both strands. Nearly every transcript derives from a small number of long polycistronic precursors that must be processed, modified, translated and degraded with precision, and the machinery for that processing is largely nuclear-encoded. Transcriptional initiation is therefore not the main lever. Post-transcriptional regulation carries the weight, and RNA modification sits at the centre of it.
Where the modifications come from, and what they do, is the substance of the review. Chemical marks once associated only with nuclear transcripts appear on mitochondrial RNA, including N6-methyladenosine, 5-methylcytosine, N1-methyladenosine, pseudouridine and formylation of the initiator methionine, and candidate writer, reader and eraser enzymes have been proposed for each. Some of those enzymes are shared with the cytosol, some have mitochondrial isoforms produced by alternative translation initiation or splicing, and assigning function is complicated by the fact that mitochondrial transcripts are short, heavily structured and present in high copy number. Functional consequences are still being mapped, but the pattern is consistent: marks on transfer RNA influence decoding fidelity, aminoacylation and fragmentation; marks on ribosomal RNA influence subunit assembly and translation efficiency; marks on messenger RNA influence stability and, in some cases, localisation to the mitochondrial surface. Because oxidative phosphorylation supplies most cellular ATP, small shifts in translation efficiency propagate into metabolic phenotype. Assigning function requires methods that quantify a mark on a specific transcript rather than across a pool of fragments, and current mapping techniques are only beginning to reach that resolution.
Cancer is where the implications concentrate. Mitochondrial metabolism supports biosynthetic growth, redox balance and survival under stress, and tumour cells rewire it extensively, often becoming dependent on specific electron transport chain complexes or on mitochondrial one-carbon metabolism. Modifications that tune oxidative phosphorylation therefore influence tumourigenesis directly, and they connect to metabolic reprogramming in transformed cells and to the immune consequences of mitochondrial dysfunction, including effects on antigen presentation and on the metabolic fitness of tumour-infiltrating lymphocytes. Treating these marks as actionable targets follows naturally, since mitochondrial RNA and its modifying enzymes are chemically and genetically distinct from the corresponding nuclear machinery, so modulation can in principle be selective rather than global.
Selectivity is exactly where the difficulty lies. Mitochondria are essential in every cell, so broad inhibition of mitochondrial gene expression carries obvious toxicity, and the therapeutic window narrows accordingly. Delivery adds a second constraint: oligonucleotide and small-molecule cargos must reach the organelle, which has prompted mitochondria-targeted peptides, ligand- or antibody-conjugated nanoparticles and oligonucleotides engineered for mitochondrial import. Responsive activation, triggered by tumour or microenvironment signals, is a further route to reducing off-target effect, as is engineering dependence on a vulnerability that tumour cells carry and normal cells do not. The fundamental questions are equally pressing. Which marks are stoichiometric and functional rather than incidental? Which enzymes act in the organelle rather than elsewhere? How do changes in mitochondrial marks reshape nuclear transcriptional programmes, given that mito-nuclear communication runs in both directions and the field has so far watched mainly one? Conditional and inducible perturbation systems, combined with bulk and single-cell sequencing and chromatin profiling, are proposed as the means to answer the last question, and integrating those mechanistic results with better detection technology is what would move mitochondrial epitranscriptomics from description toward therapy. Sequencing mitochondrial transcripts directly, rather than inferring marks from nuclear assays, is the technical step that would make such integration possible.
DOI
10.1007/s11684-026-1233-z