The process of variable translation of mRNAs through alternative start codon selection produces multiple protein variants with distinct cellular destinations and has been associated with mitochondrial function and rare diseases. Investigators at the Whitehead Institute for Biomedical Research, Massachusetts Institute of Technology, and Boston Children’s Hospital identified that alternative start codons can determine a protein’s localization to the mitochondria or its retention in the cytosol and nucleus. This mechanism, detailed in a Molecular Cell research article, has medical implications, as rare disease variants can selectively impair one isoform while leaving another intact, resulting in divergent molecular and clinical outcomes.
Ancient origins of alternative start codons
Many core biochemical activities—DNA replication, translation, and tRNA maturation—must happen both inside mitochondria and elsewhere in the cell. Several evolutionary solutions to localize these essential biochemical activities have emerged, such as gene duplication and organelle‑specific paralogs. Another solution, the production of protein isoforms, is a codified part of every Biology 101 class.
Protein isoform production is primarily attributed to different arrangements of mRNA generated through splicing and alternative promoters. This study delves deeply into a strategy of complementary isoform diversification, where a single mRNA transcript sequence can produce multiple functionally distinct proteins by initiating translation at different start codons. In fact, more than 50% of mRNAs produce more than one protein product. This approach is essential for generating mitochondrial targeting signals encoded at protein N-termini through alternative upstream or downstream initiation, which can alter subcellular localization by adding, removing, or masking these signals.
Using ribosome profiling in human cells, the investigators mapped translation initiation genome‑wide and identified thousands of alternative start sites that generate N‑terminal extensions or truncations in over two thousand genes. Computational localization prediction and direct fluorescent tagging experiments showed that a substantial fraction of these alternative isoforms localize differently from the annotated proteins. Many of these dual-localized isoforms are distinguished by the fact that some are truncated and remain cytosolic, while others are extended and gain targeting signals and enter mitochondria. Additionally, some extensions “mask” an existing mitochondrial targeting peptide, causing the protein to be sent to another location, such as the nucleus or nucleolus.
The study documents several evolutionary solutions for producing alternative isoforms: leaky ribosome scanning, alternative transcription start sites that remove upstream starts, and—in some lineages—gene duplication yielding paralogs specialized for different compartments. Phylogenetic analyses revealed the conservation of many dual-localized isoforms across eukaryotes, likely originating around the time of mitochondrial endosymbiosis. The conserved example of TRNT1 (tRNA nucleotidyltransferase) exemplifies this: one isoform (initiating at Met1) is mitochondrial, whereas an internal Met30 generates a nuclear/cytosolic isoform; this configuration is maintained across numerous species that preserve mitochondrial tRNAs.
Clinical implications of isoform-selective alleles
To probe biological function, the authors made isoform‑selective CRISPR knockouts. For Aurora kinase A-interacting protein 1 and mitochondrial ribosomal protein S38 (AURKAIP1/MRPS38), a single protein with multiple functions, an extended CUG‑initiated isoform localizes to the nucleolus and promotes cytosolic ribosome biogenesis and translation, while the shorter mitochondrial isoform supports mitochondrial ribosome function. Selective loss of the extended AURKAIP1/MRPS38 isoform impaired cytosolic translation and cell growth, demonstrating that the isoforms carry distinct, essential roles.
The researchers discovered a clinically significant finding that numerous pathogenic variants recorded in ClinVar, which archives and aggregates information about relationships among variation and human health, do not completely eliminate all protein products from a gene but rather selectively remove one isoform—designated by the authors as isoform-selective alleles (ISAs). For example, nonsense or frameshift mutations that occur between two in‑frame start codons can destroy the longer mitochondrial isoform while leaving a downstream cytosolic isoform intact. The authors identified dozens of such cases, including TRNT1, and validated that patient alleles often preserve downstream truncated proteins.
Importantly, the clinical presentation can differ depending on which isoform is lost. In TRNT1‑related SIFD (sideroblastic anemia, immunodeficiency, fevers, developmental delay), patients with mutations that disrupt both isoforms present with classic severe disease and early mortality; by contrast, patients carrying TRNT1 mutations that selectively eliminate the mitochondrial isoform—or selectively impair initiation at the downstream Met30—exhibit milder or atypical phenotypes (for example, retinal dystrophy without severe anemia, or periodic fevers with preserved hematologic status). Similarly, missense changes that alter an internal start codon (rather than protein structure) can be pathogenic by altering isoform ratios, despite being scored as benign by standard missense predictors.
This work highlights alternative translation initiation as a pervasive mechanism shaping organelle proteomes, with deep evolutionary roots and direct relevance to human disease. It argues that variant interpretation must account for alternative start codons and possible isoform‑specific effects—areas not routinely considered in clinical genetics pipelines. Clinically, recognizing ISAs could explain why patients with variants in the same gene have divergent symptoms and could reshape diagnosis, prognosis, and potentially therapeutics (for example, by restoring initiation at an alternative start site or modulating isoform ratios). The study calls for broader integration of translation‑level annotations into variant databases and clinical reporting to uncover missed diagnoses and refine genotype–phenotype links in rare disease.
