Temporal dynamics of mRNA translation dysregulation and codon decoding during murine stroke evolution.

Abstract

Ischemic stroke imposes acute oxidative and metabolic stress, yet how this reshapes mRNA translation remains poorly defined. We generated a time-resolved RNA-seq and ribosome-profiling atlas of mouse ischemic cortex at 1, 6 and 24 hours. Stroke produced early discordance between mRNA abundance and ribosome occupancy, revealing translational efficiency changes not captured by transcriptomics. The hyperacute phase was marked by a transient shift toward G/C-ending codon usage and a burst of 3' untranslated region ribosome footprints consistent with stop-codon readthrough-like termination stress. Machine-learning analyses linked this response to stop-proximal ribosome queuing, 3' untranslated region frame usage, local RNA structure and transcript architecture rather. From 6 to 24 hours, ischemic cortex developed progressive A-site pausing and altered coding-frame occupancy. Hidden Markov modeling localized a subset of frame-disruption events to discrete coding-sequence windows and machine learning revealed reproducible codon, amino-acid, nucleotide-composition and RNA-structure features associated with frameshifting. ORF-level analyses showed that many frame changes coincided with differential open reading frame usage. Late injury was characterized by increased upstream open reading frame engagement and 5' untranslated region ribosome occupancy, inversely correlating with coding-sequence translation. These data identify staged translational remodeling after stroke that could not be captured via traditional mRNA or protein expression analysis.