Loss of a conserved tRNA anticodon modification perturbs cellular signaling

Abstract

Transfer RNA (tRNA) modifications enhance the efficiency, specificity and fidelity of translation in all organisms. The anticodon modification mcm(5)s(2)U(34) is required for normal growth and stress resistance in yeast; mutants lacking this modification have numerous phenotypes. Mutations in the homologous human genes are linked to neurological disease. The yeast phenotypes can be ameliorated by overexpression of specific tRNAs, suggesting that the modifications are necessary for efficient translation of specific codons. We determined the in vivo ribosome distributions at single codon resolution in yeast strains lacking mcm(5)s(2)U. We found accumulations at AAA, CAA, and GAA codons, suggesting that translation is slow when these codons are in the ribosomal A site, but these changes appeared too small to affect protein output. Instead, we observed activation of the GCN4-mediated stress response by a non-canonical pathway. Thus, loss of mcm(5)s(2)U causes global effects on gene expression due to perturbation of cellular signaling.

Figure 1. Genetic ablation of mcm⁵ or s² leads to ribosome accumulation at specific codons.

(A) (left) mcm⁵s²U is found at the 5′ nucleotide of the anticodon in three yeast tRNAs. (right) The structure of mcm⁵s²U, and the subset of modification genes whose mutants were profiled in this study are indicated. (B) (top) Anatomy of a ribosome footprint, with P-site offset for 28 mer reads indicated. (bottom) Metaplot of WT ribosome footprint reads summed across all start codons. The peak of upstream reads corresponds to ribosomes with start codons in their P site. The location of this peak is used to determine the location of A, P and E sites for each read length. (C) (left) Explanation of metacodon plots. Similar to panel B, all in-frame instances of a given codon in the genome are aligned, and the reads mapping around those positions are summed. The resulting plot is then offset by the P-site distance, and normalized to the average peak height of the outer sites (±1, ±2). The peak heights for each site are the bulk codon occupancies, a proxy for the amount of time the ribosome spends with a given codon in each site, compared to its neighbors. (right) ATG codons and stop codons display the expected distributions with this metric. All plots are from WT yeast. (D and E) Changes in bulk codon occupancy in MSUM mutants. Both plots are the same, with different codons highlighted. Independent biological replicates were done for ncs6Δ and uba4Δ. All mutants are compared to a WT sample prepared and processed simultaneously.

Figure 2. A single-codon occupancy metric shows that ribosome footprint accumulations at AAA, CAA, and GAA are statistically significant.

(A) Description of the single codon occupancy metric. The occupancy for a given codon in a given site is the number of in-frame reads for that codon in that site, compared to the average in-frame read density for the parent gene. (B) Cumulative distributions of single-codon occupancy for select codons in ncs6Δ and uba4Δ. (C) Heatmap of K-S test p-values for all sense codons in all mutants. For ncs6Δ and uba4Δ, mutant and WT replicates were pooled to improve the accuracy of the metric.

Figure 3. Single codon occupancy changes may be insufficient to affect protein output.

(A) Fold changes for all single codons in uba4Δ are plotted against their read density in grey. Colored lines are the mean fold changes for the specified codons over read-coverage bins of width 0.2 (log₂ scaled). “Other” is a pool of all non-VAA codons. (B) Metaplot of ribosome footprint density around all AAA and CAA codons with ≥2-fold change in uba4Δ, and ≥32 reads in both datasets. Reads at each position were normalized by the total number of reads for the parent gene, and averaged across all host genes that overlap that position. The plot is offset such that 0 corresponds to having the codon in the A site. The expected location of a ribosome queuing event is indicated, and a diagram of such an event is shown below. The dip in ribosome footprint density at −10 is a computational artifact, due to an inability to determine read lengths of poly-adenylated fragments when they end in one or more adenosines.

Figure 4. MSUM strains show the gene-expression signatures of GCN4 activation.

(A) Comparison of RNA-seq and Ribo-seq RPKM changes in uba4Δ. GCN4 targets and statistically significant Ribo-seq changes are indicated. Values are the means of 2 biological replicates. (B) Venn diagram of overlap between GCN4 functional targets (blue) and significant Ribo-seq RPKM changes in uba4Δ (pink) and ncs6Δ (green). The significance of the overlap was computed using the hypergeometric distribution. (C) Cumulative distribution plots of fold Ribo-seq changes for GCN4 targets (solid lines) compared to all other genes (dashed lines) in uba4Δ (top) and ncs6Δ. P values are from a KS test of GCN4 targets against the rest of the genome. (D) Mean±SEM of predicted Gcn4p occupancy for groups of genes from panel B and figure S5, as determined by high-throughput in vitro binding assays . Bars are colored to match groups in panel B. P values are from t-tests comparing the indicated gene set against all genes in the genome.

Figure 5. GCN4 is induced independently of GCN2 in MSUM strains.

(A) Ribo-seq and RNA-seq RPKMs for the GCN4 open reading frame. Standard deviations are indicated for strains with replicate data. (B) The indicated strains were transformed with a reporter containing the promoter and transcript leader of GCN4 fused to lacZ. LacZ activity and mRNA levels were measured in log phase after overnight growth in YPD. (C) LacZ assays were performed as in panel B, with the addition of double mutant strains. P values are for t-test against WT unless otherwise indicated.

Figure 6. Disruption of the GCN pathway partially suppresses the stress sensitivity of MSUM strains, independently of tRNA overexpression.

(A) Yeast was grown to saturation in selective media. 5-fold serial dilutions were spotted onto YPD containing the indicated drug, and grown at the indicated temperature. (B) The independent rescue of MSUM phenotypes by gcnΔ and hc-tRNA suggests that two independent pathways contribute to the mutant phenotypes.