Queuine links translational control in eukaryotes to a micronutrient from bacteria

Abstract

In eukaryotes, the wobble position of tRNA with a GUN anticodon is modified to the 7-deaza-guanosine derivative queuosine (Q34), but the original source of Q is bacterial, since Q is synthesized by eubacteria and salvaged by eukaryotes for incorporation into tRNA. Q34 modification stimulates Dnmt2/Pmt1-dependent C38 methylation (m5C38) in the tRNAAsp anticodon loop in Schizosaccharomyces pombe. Here, we show by ribosome profiling in S. pombe that Q modification enhances the translational speed of the C-ending codons for aspartate (GAC) and histidine (CAC) and reduces that of U-ending codons for asparagine (AAU) and tyrosine (UAU), thus equilibrating the genome-wide translation of synonymous Q codons. Furthermore, Q prevents translation errors by suppressing second-position misreading of the glycine codon GGC, but not of wobble misreading. The absence of Q causes reduced translation of mRNAs involved in mitochondrial functions, and accordingly, lack of Q modification causes a mitochondrial defect in S. pombe. We also show that Q-dependent stimulation of Dnmt2 is conserved in mice. Our findings reveal a direct mechanism for the regulation of translational speed and fidelity in eukaryotes by a nutrient originating from bacteria.

Figure 1.

Queuosine modification of tRNAs increases the translational speed of the C-ending codons for His and Asp and decreases the speed of the U-ending Asn and Tyr codons. (A) Overview of the experimental conditions for ribosome profiling to determine the effect of Q modification on translation. m⁵C38 indicates the level (%) of C38 methylation of tRNA^Asp in a given sample. Q designates samples with (+) or without (–) Q34-modification of tRNAs. (B) Codon-specific changes in ribosomal A-site occupancy in the absence of Q modification in comparison to with Q (mean ± SEM, n = 3), as measured by ribosome profiling. C-ending Q codons are boxed in red, U-ending codons in yellow. Thin gray lines give the SEM.

Figure 2.

Translational speed of C-ending Q codons relative to U-ending codons was decreased in the absence of Q modification (mean ± SEM (thin gray lines), n = 3). Experimental conditions used are as in Figure 1.

Figure 3.

Global effect of Dnmt2/ Pmt1-dependent tRNA^Asp methylation on codon occupancy. Codon occupancy in samples without m⁵C38 tRNA^Asp methylation was compared to those with m⁵C38. (A) Overview of sample comparisons regarding the presence or absence of m⁵C38 and Q. Representation as in Figure 1A. Sample comparisons were pmt1Δ cells carrying a vector control (+v) compared to control pmt1⁺ OE (overexpression) cells (cells cultivated in minimal medium, gray); pmt1Δ + Q compared to wt +Q (green); and pmt1Δ compared to wt + Q (blue, same data as in Figure 1B). Values are mean ± SEM (representation as in Figure 1B). (B) Codon-specific changes in ribosomal A-site occupancy in the absence of m⁵C38 (pmt1Δ) compared to with m⁵C38 by pmt1⁺ overexpression (OE) or by the addition of Q. Representation as in Figure 1B.

Figure 4.

Misreading errors by tRNA^Asp of the Gly codon GGC, but not GGU, decrease in the presence of Q34 modification. The activity of wt Escherichia coli β-galactosidase expressed in S. pombe wt or pmt1Δ strains was measured in the absence (white bars) or presence of queuine (black bars, mean ± SD). The codon–anticodon pairing required for a (mis-)reading event is diagrammed above the graph. The upper line represents the codon, the lower line the anticodon (AC). Vertical lines represent Watson–Crick pairs, filled circles canonical wobble pairs, and open circles pairs requiring a tautomeric shift to form. (A) Activity of wt β-gal (wt, n = 5; pmt1Δ, n = 4). (B) β-gal activity of mutant version in which the Asp codon 201 is replaced by the Gly codon GGC (wt, n = 5; pmt1Δ, n = 6). (C) as in B, but Asp 201 replaced by the GGU Gly codon (n = 3). (D, E) Activity of β-gal with Asp 201 replaced by the Glu codon GAA or GAG. (F, G) Activity of β-gal with Tyr 503 replaced by the Cys codon UGC or UGU (n = 4). **P < 0.005; *P < 0.01. All other comparisons of –/+Q were statistically not significant.

Figure 5.

Effect of Q and m⁵C38 modification on translational efficiency. (A) Effects of Q on translational efficiency. Plots of the log₂-fold change in translational efficiency of two compared conditions are shown relative to the log₂ of the median expression of transcripts from the two conditions. Left, TE of pmt1Δ +Q/pmt1Δ; middle, TE of wt +Q/ pmt1Δ; right, TE of wt +Q/wt. Colours indicate transcripts with significantly different translational efficiency. Blue dots, P < 0.05, red dots, Benjamini–Hochberg adjusted (adj) P < 0.1. The number of genes in the respective P-value categories that are up- or down-rgeulated are given in the upper right and lower right corner of the plots, respectively. The triangle represents a data point that is outside the plotted frame and corresponds to pmt1⁺. (B) Effect of C38 methylation by Dnmt2/ Pmt1 on translational efficiency. The log₂-fold change in TE of pmt1⁺ overexpression (OE)/pmt1Δ + vector (v) (left) and wt +Q/ pmt1Δ+Q (right) is plotted against the log₂ median expression of transcripts of the two conditions. Representation as in (A). (C) Levels of Cox2 protein were decreased in pmt1Δ + Q. Protein extracts were separated by SDS-PAGE and analysed by Western blotting using an antibody against Cox2 (top) and, as a control, with an α-Tubulin antibody (Tub1). Below, quantification of Cox2 levels relative to Tub1 (mean ± SEM, n = 3, normalized to pmt1Δ value. *P = 0.11; **P = 0.002). (D) Overlap of differentially translated mRNAs between pmt1⁺ OE/ pmt1Δ +v (see Figure 4B, left) and wt +Q/ pmt1Δ +Q (Figure 4B, right) (adj. P < 0.1). (E) Overlap of mRNAs differentially translated in wt +Q/ pmt1Δ (Figure 4A, middle) and wt +Q/ pmt1Δ +Q (Figure 4B, right). (F) Translational efficiency of genes with stretches of Q codons was decreased in the absence of Q. The translational efficiency (TE) of genes with at least one stretch of 2 consecutive Asp, His, Asn or Tyr codons or three or more consecutive Q codons (size of the group of genes is given in parenthesis) was compared to that of all other genes not fulfilling this condition. The difference between the median translational efficiency of genes with the respective stretch compared to that of genes without the stretch is given. Only values that are statistically significantly different (Benjamini–Hochberg adjusted P < 0.05) are shown.

Figure 6.

Q modification caused a respiratory defect in S. pombe cells and a growth defect when pmt1⁺ was overexpressed. (A) Serial dilutions of wt, pmt1Δ and qtr2Δ S. pombe strains were spotted on full medium (YES) with glucose and on medium with 3% glycerol in the presence or absence of Q. (B) Strains as in (A) were grown on medium with 300 mM CaCl₂ with our without Q. (C) Overexpression of pmt1⁺ caused a growth defect in the presence of Q in wt and pmt1Δ, but not qtr2Δ. (D) No growth defect of the strains was observed with the vector control.

Figure 7.

Stimulation of Dnmt2-dependent m⁵C38 methylation by Q is conserved in mice. (A) Mouse embryonic stem cells were cultivated in 2i medium and treated for 1, 2 or 3 days with 0.05 or 0.5 μM Q. RNA was extracted, and m⁵C38 levels in RNA^Asp were determined by high-throughput bisulfite sequencing. (B) m⁵C38 levels in RNA^Asp were determined as in (A) in RNA extracted from brain or liver tissues from wt or Qtrd1−/− mice (tgt −/−). Mean ± SD is reported (n = 3). (C) Graphical summary of the effect of tRNA Q modification on translational speed in S. pombe. The top row shows the tRNAs carrying either Q34 or G34. They translate the respective C- or U-ending codons. Q modification results in faster translation of the C-ending codons for Asp and His (bold arrow; ‘intermediate-strength’ codons according to (4)), but has no effect on the respective U-ending codons. Q modification slows down translational speed of the U-ending codons for Asn and Tyr (dashed arrow, ‘weak’ codons according to (4)). (D) Graphical summary of the effect of Q modification on suppressing translational errors. Q modification of tRNA^Asp reduces misreading of the C-ending, but not the U-ending Gly codon.