Mistranslation suppresses mistranscription in eukaryotes

Phenotypic mutations are non-heritable changes in biomolecule sequences, such as synthetic errors resulting from mistranscription or mistranslation. Even though both types of errors contribute to a wide variety of biological processes, such as protein aggregation in neurodegenerative diseases and novel trait emergence, little is known about their relationship. In this study, we conducted a genome-wide assessment of mistranslations and mistranscriptions in five model organisms using mass spectrometry and Circ-Seq data, respectively. Transcription shows a lower per-site probability of error than translation, but the difference is much narrower when it comes to error probability per gene. In all five species, genes with frequent mistranslations tend to exhibit a lower mistranscription rate, a pattern we hypothesized as created by negative epistasis between the two types of errors. We tested our hypothesis through systematic experimental measurements of within-gene epistasis. It was found that such epistasis is predominantly negative, making proteins affected simultaneously by both types of errors significantly more deleterious than expected. More importantly, an in silico simulation of molecular evolution suggests that the extra deleterious effects caused by negative epistasis, when scaled by the error rates, are sufficiently large to facilitate selection against mistranscription for genes with frequent mistranslations. Finally, our hypothesis is further supported by the observation that genes with more frequent mistranslation purge nonsynonymous mistranscription more efficiently, and transcripts that are more frequently translated tend to have a lower mistranscription rate. Combined, our results reveal a previously unknown interaction between mistranscription and mistranslation.