Multi-omic dissection of RNA control reveals convergent cis-regulatory programs during glucose starvation in yeast

Abstract

Cells reprogram gene expression at multiple, often independently studied, regulatory layers during nutrient stress. Here we use direct RNA sequencing of polysome-fractionated transcripts that simultaneously measures transcript abundance, mRNA degradation, poly(A) tail length, and three RNA modifications (m5C, m6A, pseudouridine) to build an integrated, multi-omic view of the Saccharomyces cerevisiae response to acute glucose starvation. Lasso-penalised mixture-of-regressions clustering of stochastic translation efficiency ({delta}STE) against this combined feature set partitions 5,033 transcripts into five translational programs with distinct functional identities, from a strongly upregulated translation-machinery cluster to a strongly repressed glycosylation cluster. CDS sequence composition discriminates these extremes, while 5' and 3'UTR composition contributes only marginal, largely non-significant effects after length correction. Extending this analysis, we show that the repressed glycosylation program is distinguished by a convergent stack of independent cis-regulatory signatures: weak Kozak initiation context, high upstream open reading frame burden, excess RNA secondary structure specifically over the start codon, and depletion for the targets of major stabilising and export-associated RNA-binding proteins. In other programmes, translation efficiency tunes with initiation-context strength indicating that downstream regulatory layers coregulate translation. These results provide an integrated, feature-resolved map of post-transcriptional regulation during acute nutrient stress.