Lifestyle modifications: coordinating the tRNA epitranscriptome with codon bias to adapt translation during stress responses

Abstract

Cells adapt to stress by altering gene expression at multiple levels. Here, we propose a new mechanism regulating stress-dependent gene expression at the level of translation, with coordinated interplay between the tRNA epitranscriptome and biased codon usage in families of stress-response genes. In this model, auxiliary genetic information contained in synonymous codon usage enables regulation of codon-biased and functionally related transcripts by dynamic changes in the tRNA epitranscriptome. This model partly explains the association between synchronous stress-dependent epitranscriptomic marks and significant multi-codon usage skewing in families of translationally regulated transcripts. The model also predicts translational adaptation during viral infections.

Fig. 1

Prototypical RNA species and their modifications. All types of RNA are modified; structures for three of more than 120 modifications are shown in the lower panel. RNA secondary structures were adapted from the following sources: XIST long non-coding RNA (lncRNA) [82]; 5S and 5.8S rRNA, tRNA, microRNA (miRNA), small nuclear RNA (snRNA), transfer-messenger RNA (tmRNA), 16-18S, 23-28S rRNA [83]. Abbreviations: ac⁴C N4-acetylcytidine, ac⁴Cm N4-acetyl-2’-O-methylcytidine, Am 2’-O-methyladenosine, Cm 2’-O-methylcytidine, Gm 2’-O-methylguanosine, I inosine, i⁶A N6-isopentenyladenosine, mcm⁵U 5-methoxycarbonylmethyluridine, m⁵C 5-methylcytidine, m⁵U 5-methyluridine, m⁶A N⁶-methyladenosine, nt nucleotide, 3′-U/A 3′-uridylation/adenylation, 2’-O-Me 2’-O-methylation, Um 2’-O-methyluridine, Y pseudouridine

Fig. 2

tRNA reprogramming and codon-biased translation of stress-response proteins. a Model illustrating the effects on Mycobacterium bovis BCG of hypoxia encountered during infection. Hypoxia induces expression of 48 proteins in the Dos regulon that causes the cell to become dormant. b The tRNA epitranscriptome consists of over 120 post-transcriptionally inserted modified ribonucleosides. During hypoxia, the relative quantities of 40 tRNA modifications (rows in heat map) change as a function of time during the response to the stress (days 0–21; columns in heat map) and again during O₂ resuscitation (days 22–24). In early hypoxia (day 9), the wobble position of tRNAThr^UGU, which reads the codon ACG, switches from 5-methoxyuridine (mo⁵U) to 5-oxyacetic acid uridine (cmo⁵U; structure shown). c RNase/LC-MS maps cmo⁵U to the wobble of tRNAThr^UGU. d Families of response genes are organized by biased use of synonymous codons. The heat map shows over-use (purple) and under-use (yellow) of 62 codons (columns) across all genes (rows) in BCG. The gene for DosR, the master regulator of the 48-gene Dos regulon, over-uses the ACG codon and under-uses ACC, the most common Thr codon. e Codon analysis of proteomics data shows that > 80% of proteins upregulated in early hypoxia use ACG to code for Thr, whereas downregulated proteins are enriched in the so-called ‘optimal’ codon for Thr, ACC. Evidence that the auxiliary information in the genetic code is utilized for regulatory purposes is supported by examining codons associated with highly upregulated and downregulated proteins across all time-points of hypoxia in BCG. Pairs of synonymous codons are differentially enriched in upregulated and downregulated proteins, with the codon enrichments defining functional gene families. Alk phos alkaline phosphatase, PLS partial least squares, Thr threonine

Fig. 3

Patterns of synonymous codon usage define families of stress-response genes and might predict epitranscriptomic responses to viral infections. a The idea that translation regulation uses auxiliary genetic information in the form of codon bias arose by linking systems-level analyses of stress-induced proteomic upregulation and downregulation [21, 28, 34] with codon analytics [28, 38]. The heat maps shown here are examples of genome-level application of a codon-counting algorithm and visualization approaches [28, 34] to the genomes of Mycobacterium bovis BCG and Saccharomyces cerevisiae. The maps show over-use (yellow) and under-use (purple) of 62 codons (columns) across all genes (rows). For each organism, clusters represent groups of genes that have distinctly different codon-usage patterns compared with genome averages, with two opposing groups of genes identified in humans. As shown in Fig. 2d, the smaller of two clusters of codon-biased genes in M. bovis BCG consists of the DosR regulon of 48 genes that control the response to hypoxic stress [34]. b Widely differing codon-usage patterns in the human genome and a representative dengue serotype 2 genome (DENV2) could predict reprogramming of the host cell tRNA epitranscriptome both to accommodate the codon mismatch in the viral RNA genome and to respond to the stress of viral infection. Codon frequency data were generated using the web-analysis interface on the published Codon Utilization Tool (CUT) [56] and the publicly available human and dengue sequence information (Human refseq_hg38 and dengue virus 2, complete genome NCBI Reference Sequence: KM204118.1), together with human and dengue frequency data found in Additional file 1. The analysis shows that the DENV2 genome is biased toward A-ending codons, whereas the human genome is biased toward G- and C-ending codons. This leads to the testable hypothesis that DENV2 infections will cause changes in the host cell tRNA pool, both modifications and copy numbers, to simultaneously accommodate translation of the viral mRNA and facilitate translation of codon-biased host stress-response genes