Cells alter their tRNA abundance to selectively regulate protein synthesis during stress conditions

Abstract

Decoding the information in mRNA during protein synthesis relies on tRNA adaptors, the abundance of which can affect the decoding rate and translation efficiency. To determine whether cells alter tRNA abundance to selectively regulate protein expression, we quantified changes in the abundance of individual tRNAs at different time points in response to diverse stress conditions in Saccharomyces cerevisiae We found that the tRNA pool was dynamic and rearranged in a manner that facilitated selective translation of stress-related transcripts. Through genomic analysis of multiple data sets, stochastic simulations, and experiments with designed sequences of proteins with identical amino acids but altered codon usage, we showed that changes in tRNA abundance affected protein expression independently of factors such as mRNA abundance. We suggest that cells alter their tRNA abundance to selectively affect the translation rates of specific transcripts to increase the amounts of required proteins under diverse stress conditions.

Fig. 1

Quantification of change in abundance of yeast tRNAs during stress. (A) Yeast cells were grown in YPD (yeast extract, peptone, and dextrose) at 30°C and then challenged with four different stress conditions: (i) temperature stress by increasing incubation temperature from 30° to 37°C, (ii) osmotic stress by increasing sorbitol concentration from 0 to 1 M, (iii) oxidative stress by increasing H2O2 concentration from 0 to 0.5 mM, and (iv) diauxic shift by changing the carbon source of the media from 2% glucose to 2% ethanol. For each condition, change in abundance of individual tRNAs was measured by qPCR with respect to normal, nonstressed conditions at three different time points: 20, 60, and 120 min (all in biological triplicates). (B) Hierarchical clustering of tRNAs based on their relative changes in abundance over time (average fold change of n = 3 biological replicates). (C) Pie charts of the proportion of up- and down-regulated tRNAs under different stress conditions. DNase I, deoxyribonuclease I; cDNA, complementary DNA; RNase A, ribonuclease A.

Fig. 2

Analysis of tRNA expression patterns. (A) Multivariate analysis of the tRNA expression patterns using t-SNE and K-means clustering. Three groups of tRNAs (C1, C2, and C3) are highlighted. All tRNAs except tRNA^Leu(CAA), tRNA^Leu(UAG) (which were outliers, assignable to C1), and the initiator tRNA^Met(CAU) (assignable to C3) were included for the visualization. (B) Fold change distribution of tRNA abundance in each group (dot plots). Horizontal dashed lines indicate threshold values for up-regulation (+50%) and down-regulation (−50%). The effect sizes (rank-biserial correlation using Wendt’s criterion) of comparisons between C1-C2 and C2-C3 are indicated for each stress condition. P values are computed using Mann-Whitney U test, *P < 0.05, **P < 0.01, ***P < 0.001. (C) Pairwise correlation matrices showing the similarity of changes in tRNA abundance across the four stress conditions after 20, 60, and 120 min. Each cell shows the Pearson’s r correlation coefficient, with negative and positive values colored in shades of red and blue, respectively. μ_r represents global r values for each time point.

Fig. 3

Changes in the tAI during stress. (A) n-tAI (x axis) and s-tAI (y axis) were both normalized to range between 0 and 1. Dots indicate individual genes. Orange genes have increased their z score of shift in rank more than 1.28 times in terms of tRNA adaptation (10% of genes that are likely to be most efficiently translated), and red genes have decreased their z score of shift in rank more than 1.28 times (10% of genes that are likely to be least efficiently translated) during stress compared to normal conditions. The other genes are shown in gray. Some exemplars of significant gene ontology (GO) slim enrichment are indicated in the figure (data file S4). (B) Ribosome profiling data during oxidative stress for a different group of genes (n = 4157 genes). Numbers within the box plots denote effect size. (C) Increase in protein abundance during oxidative stress for the different group of genes. To control for mRNA abundance, three plots are shown for low (n = 1386 genes), medium (n = 1386 genes), and high (n = 1385 genes) mRNA abundance (tertile classification). (D) Experimentally measured protein production rates of the four variants of mEGFPs calculated as an increase in fluorescence per time (ΔA.U./s) under different stress conditions (average of n = 3 biological replicates). Amino acid similarity (%AA), nucleotide similarity (%nt), n-tAI, and s-tAI values are shown. mEGFP protein production rates for the four variants under each of the four stress conditions (16 measurements) plotted against s-tAI. Individual r values for each stress condition are: 0.66 (diauxic shift), 0.77 (oxidative stress), 0.98 (osmotic stress), and 0.77 (temperature stress). A.U., arbitrary units; Nt, N terminus; Ct, C terminus; TCA, tricarboxylic acid.

Fig. 4

Simulation of protein translation based on the experimentally measured changes in tRNA. (A) Simultaneous simulation of protein production rates for 3727 yeast genes. (B) Distribution of s-tAI values and fold change in protein production rates for the ESR up-regulated genes (blue) compared to the other genes (gray). Fold change in protein abundance (right) in oxidative stress based on confocal microscopy. (C) Bayesian network inference. The explanatory variables used to study protein production rate fold change include the following: variation in mRNA abundance, tAI, initiation frequency, and elongation speed. In total, 29,816 observations were pooled from simulations of the four stress conditions. The numbers on the edges in the network represent the magnitude of the link strength. P values are computed using Mann Whitney U test, *P < 0.05, **P < 0.01, ***P < 0.001. Numbers within the box plots denote effect size.

Fig. 5

A model for how changes in the tRNA abundance could selectively influence protein production rates of specific transcripts. (A) Under prolonged stress, the production of proteins is reshaped as a result of changes in the abundance of mRNAs and tRNAs. In this scenario, there is a balance between codon demand and tRNA supply, whereby adaptation to the tRNA pool may result in higher levels of protein production. (B) Under optimal growth conditions, the transcriptome consists of highly abundant mRNAs coding for growth-related genes whose codon usage is adapted to tRNA abundance under normal conditions and whose proteins are produced at a high rate and abundance (gray). Another part of the transcriptome consists of lower abundance mRNAs for stress-responsive genes whose codons are less adapted to tRNA abundance under normal conditions and whose proteins are produced at basal or low levels (blue). After prolonged stress, the tRNA pool is significantly altered. Growth-related genes tend to have fewer transcripts and show relatively slower elongation due to reduced codon adaptation to the new tRNA pool, resulting in decreased protein production. Stress-responsive genes tend to have more transcripts whose elongation is also globally slower than in normal conditions but relatively faster compared to the other genes because of better codon adaptation to the new tRNA pool, resulting in an overall increase in protein production.