Deciphering the reading of the genetic code by near-cognate tRNA

Significance

Protein translation is a key cellular process in which each codon of mRNAs has to be accurately and efficiently recognized by cognate tRNAs of a large repertoire of noncognate tRNAs. A successful decoding process is largely dependent on the presence of modified nucleotides within the anticodon loop, especially of tRNAs having to read A/U-rich codons. In this latter case, their roles appear to stabilize the codon–anticodon interaction, allowing them to reach an optimal energetic value close to that of other interacting tRNAs involving G/C-rich anticodons. In this work we demonstrate that, while helping an efficient translation of A/U-rich codons, modified nucleotides also allow certain unconventional base pairing to occur, as evidenced in the case of stop codon suppression.

Abstract

Some codons of the genetic code can be read not only by cognate, but also by near-cognate tRNAs. This flexibility is thought to be conferred mainly by a mismatch between the third base of the codon and the first of the anticodon (the so-called “wobble” position). However, this simplistic explanation underestimates the importance of nucleotide modifications in the decoding process. Using a system in which only near-cognate tRNAs can decode a specific codon, we investigated the role of six modifications of the anticodon, or adjacent nucleotides, of the tRNAs specific for Tyr, Gln, Lys, Trp, Cys, and Arg in Saccharomyces cerevisiae. Modifications almost systematically rendered these tRNAs able to act as near-cognate tRNAs at stop codons, even though they involve noncanonical base pairs, without markedly affecting their ability to decode cognate or near-cognate sense codons. These findings reveal an important effect of modifications to tRNA decoding with implications for understanding the flexibility of the genetic code.

The ribosome uses transfer RNAs (tRNAs) to decode codons of mRNAs into a sequence of amino acids in a polypeptide. The specificity of this process depends on two main functions of the tRNA: its recognition by cognate aminoacyl-tRNA synthetases (1) and its anticodon pairing with the mRNA codons (2). In all living organisms, correct reading of the genetic code by tRNAs is essential, to prevent the misincorporation of amino acids and premature termination. However, translation fidelity is not maximal, and misreading rates can vary from 10⁻³ to 10⁻⁶, depending on the tRNA and the organism considered (3, 4). Accurate decoding depends on the ability of the ribosome to discriminate between correct (cognate) and incorrect (near-cognate or noncognate) codon–tRNA interactions, while allowing the tRNA to decode synonymous codons. It was long thought that this ability was conferred by monitoring of exclusively the stability of the codon–anticodon complex, mainly based on the number of hydrogen bonds. However, recent data suggest that the ribosome plays an active role in decoding and thereafter accommodates both cognate and near-cognate tRNAs in a similar manner by forcing mismatched base pairs to adopt a Watson–Crick geometry as normal A-U/U-A or C-G/G-C base pairs (5).

More than a 100 base and ribosome modifications in tRNAs of different organisms have been shown to contribute to various aspects of gene expression (6). They can be classified into two groups: one group of modifications that serves to stabilize the core of the tRNA and its overall L-shaped structure, and a second group of modifications that are crucial for correct mRNA decoding and fine-tuning of the translation process (7). The latter modifications are mainly present within the anticodon loop, in particular at positions 34, the first nucleotide of the anticodon, and 37, located immediately 3′ to the anticodon. These modifications have a direct impact on the capacity of the tRNA to read synonymous codons and to prevent decoding of near-cognate codons (8), although this view has recently been challenged by studies revealing that, in some cases, these modifications increase the misreading of codons (3). For example, it has been suggested that the combination of 5-methylene derivatives and 2-thiolation modifications of U₃₄ restrict the decoding of codons ending with A (9, 10); moreover, 2-thiolation increases affinity of binding to the cognate codon and reduces tRNA rejection (11). These modifications have also been implicated in protein homeostasis (12), in reading-frame maintenance (13), and in the enhancement of recognition by aminoacyl-tRNA synthetases (14).

However, data are not easily transposable from one organism to another because notable differences exist in the decoding strategies of distantly evolutionary related organisms (15). The impact of nucleotide modifications in the anticodon loop on the likelihood of near-cognate tRNA being used by the ribosome has yet to be analyzed in eukaryotes. Here, we used a system in which cytoplasmic near-cognate tRNAs do not compete with cognate tRNAs to study the impact of these modifications on the recognition of naturally occurring stop codons (Fig. 1). We constructed yeast mutant strains with deletions of genes encoding specific modification enzymes and used LC-MS/MS and ribosome profiling techniques to investigate the impact of these deletions on the selection of natural suppressor tRNAs, which are, by definition, near-cognate tRNAs. We found that modifications of the tRNA anticodon loop are required for decoding a near-cognate stop codon, but that they have no marked impact on the decoding of cognate codons. We also showed that overall stop codon suppression efficiency appears as a poor indicator of the ability of individual tRNAs to decode stop codons, due to adjustments between the various competing near-cognate cellular tRNAs at reading the same stop codon. Taken together, our results highlight the flexibility of the genetic code and reveal an unexpected capacity of the Saccharomyces cerevisiae translation machinery to discriminate between sense and nonsense near-cognate codons. Although one cannot rule out that increased readthrough due to Ψ₃₅ is a direct consequence of the increased stability of the codon:anticodon pairing, one appealing hypothesis is that genes coding for enzyme catalyzing Ψ₃₅ have been conserved through evolution to regulate stop codon readthrough according to the cellular needs.

Fig. 1.

(A) A schematic diagram of the anticodon loops from the three near-cognate tRNAs inserted at the various stop codons. Modifications studied in this work are shown in red, with the name of the corresponding gene (*= mcm⁵s²). The position of the mismatch (or nonconventional base pairing) between the anticodon and the stop codon is highlighted in blue (or in purple if this position is also modified). Lysine tRNA_CUU and arginine tRNA_UCU also carry t⁶A₃₇ modification that is not studied in this work because of the strong growth defect link to the absence of this modification. (B) The quantification of the various fractions of the polysome profiles obtained for each mutant. Three to five independent experiments were performed for each mutant.

Results

Deletion of Genes Coding Enzymes Responsible for tRNA Modifications Have No Major Impact on Cell Growth and Polysomes.

To study the impact of several yeast tRNA modifications on the decoding of near-cognate codons, we used a reporter system based on the stop codon readthrough-dependent expression of a gene encoding a GST protein. This system provides a precise quantitative analysis of the incorporation of natural suppressor tRNAs at stop codons. In a previous study, we identified tyrosine, glutamine, and lysine tRNAs as the cytoplasmic tRNAs incorporated at UAA and UAG codons, and tryptophan, cysteine, and arginine tRNAs as the tRNAs incorporated at UGA codons (Table S1) (16).

Here, we systematically deleted six genes encoding enzymes responsible for posttranscriptionnal modification of bases or hydroxyl-riboses in the anticodon loop of the above-identified suppressor tRNAs (Fig. 1A). Three of these genes (Elp3p, Trm9p, and Tuc1p) are involved in conversion of U₃₄ residue in the glutamine tRNA_UUG, lysine tRNA_UUU, and arginine tRNA_UCU into a hypermodified 5-methoxycarbonylmethyl-2-thiouridine residue (mcm⁵s²U) (6). The deletion of ELP3 or TRM9 results in glutamine and lysine tRNAs carrying only s²U₃₄ or cm⁵s²U₃₄, respectively (U₃₄ and cm⁵s²U₃₄ for arginine tRNA), whereas the deletion of TUC1 results in glutamine and lysine tRNAs with mcm⁵U₃₄ (17, 18). The anticodons of the unique tyrosine and tryptophan suppressor tRNAs are also modified by two other enzymes: Pus7p, which catalyses the pseudouridylation of U₃₅ in the tyrosine tRNA and in U2 snRNA (19); and Trm7p, which methylates C₃₄ and C₃₂ in the tryptophan tRNA (19, 20). Another enzyme of interest is Mod5p, which converts the nucleotide at position 37, 3′-adjacent to the anticodon, to i⁶A in the tyrosine and cysteine tRNAs (21). We first confirmed that deletions of these genes have no severe effect on cell viability (Fig. S1). Doubling times were similar for the parental and deletion strains, except for the ΔTRM9 and ∆TRM7 strains, which grew more slowly, as reported in previous studies (20). Polysome profile analysis revealed that the proportions of 40S, 60S, and 80S ribosomes were unaffected in each strain carrying the above-selected deletion of modification enzyme genes (Fig. 1B).

U₃₄ Modification Affects Stop Codon Recognition by the Gln, Lys, and Arg tRNAs.

In yeast, most of the uridine residues in position 34 of tRNAs are hypermodified. We investigated the influence of different elements of mcm⁵s²U modification on stop codon decoding by natural suppressor tRNAs by deleting the ELP3, TRM9, and TUC1 genes separately in the parental strain. Because these enzymes modify U₃₄ only in glutamine tRNA_UUG, lysine tRNA_UUU (reading UAA), and arginine tRNA_UCU (reading UGA) (Fig. 1A), we assessed their impact on these two stop codons. We first used a dual reporter system (22) to analyze the effect of each deletion on the levels of stop codon suppression. We found no significant effect of TRM9 deletion, but a slight decrease in readthrough efficiency in ∆elp3 mutants, for the UAA and UGA codons only (Fig. 2A).

Fig. 2.

Consequences on stop codon suppression of deletion genes coding for U34 modification in various tRNAs. (A) Quantification of readthrough efficiency at UAA and UGA codons in three mutants in which mcm⁵s²U₃₄ formation was abolished. At least five independent experiments were performed for each value. The central lines show the medians; the box limits indicate the 25th and 75th percentiles. (B) Corrected intensities of readthrough peptides normalized according to the amount of GST were quantified by determining the mean value for three experiments. Quantification was performed as described in Materials and Methods.

Next, we assessed the contributions of the glutamine tRNA_UUG, lysine tRNA_UUU, and arginine tRNA_UCU to the levels of stop codon expression observed, by transforming the Δelp3, Δtrm9, and Δtuc1 strains with the GST reporter system (16) and performing mass spectrometry on the GST proteins generated by stop codon suppression, to quantify the frequency of incorporation of the amino acids corresponding to the three tRNAs, as previously described (Fig. S2B) (16). Quantification of the proportions of amino acid present at UAA codon revealed that incorporation of the glutamine tRNA_UUG occurred less frequently in the three deletion strains than in the WT, this tRNA being replaced by the tyrosine tRNA, the incorporation of which was not affected by the deletions (Fig. 2B). We also found that deletions of ELP3 (resulting in mcm⁵-lacking s²U₃₄) and TUC1 (resulting in the absence of the 2-thio group of uracil at position 34 bearing the 5-mcm group) decreased glutamine incorporation more strongly than the deletion of TRM9 (resulting in m-lacking cm⁵s²U₃₄). Quantitative analyses for the UGA codon showed that deletion of ELP3 had no effect on the ratios of the various amino acids, whereas the deletion of TRM9 strongly stimulated incorporation of arginine and also slightly increased cysteine incorporation, even though the corresponding cysteine tRNA is not thought to be modified by TRM9 (Fig. 2B). Clearly, 5-mcm modification does not have the same effect on the ability of glutamine, lysine, and arginine tRNAs to read UAA and UGA stop codons.

Modification of the Nucleotides Adjacent to the Anticodon Plays an Important Role in Near-Cognate Decoding.

Beside the arginine tRNA_UCU, the two main other suppressor tRNA of the UGA codon are the tryptophan and cysteine tRNAs, both of which carry nucleotide modifications. The tryptophan tRNA anticodon loop is modified at C₃₂ and C₃₄ into 2′-methyl cytosine (Cm) by Trm7p, while the Cys tRNA_GCA is modified at A₃₇ into i⁶A₃₇ by Mod5p (Fig. 1A). They both belong to the same four-codon family boxes. Suppression involved a mismatch base pair at position 34 of the anticodon and the third base (A) of the codon (respectively, Cm₃₄-A₃ for tryptophan tRNA and G₃₄-A₃ for cysteine tRNA). As shown in Fig. S3A, efficiency of UGA readthrough is not affected by TRM7 deletion (Fig. S3A), which also had no effect on the proportions of tryptophan, cysteine, and arginine incorporated (Fig. S3B). In contrast, the absence of Mod5p affected only the relative proportions of suppressor tRNAs not the level of the UGA suppression. Indeed, cysteine was not incorporated anymore at the UGA codon in the ΔMOD5 strain (Fig. S3B).

Like the cysteine tRNA, the tyrosine tRNA is modified by Mod5p, resulting in an i⁶A residue at position 37. We analyzed the effect of the elimination of this modification from the tyrosine tRNA on the level of stop codon suppression. Cysteine tRNA harbors a G₃₄-A₃ mismatch when reading UAA and tyrosine tRNA harbors a G₃₄-G₃ mismatch when reading UGA). Quantitative analyses for the ΔMOD5 strain showed that stop codon suppression efficiency at both the UAA and UAG codons was much lower than in the parental strain (Fig. 3A). Contribution of the tyrosine tRNA to the global suppression observed in the ΔMOD5 strain was next investigated by quantifying the incorporation of suppressor tRNAs at UAA and UAG codons. Levels of tyrosine incorporation were slightly lower at the UAG codon and much lower at the UAA codon (Fig. 3B). Thus, the effect of the i⁶A modification of the tyrosine tRNA depends on the codon decoded.

Fig. 3.

Consequences on stop codon suppression of deletion genes coding for enzymes catalyzing formation of i⁶A₃₇ and Ψ₃₅ in various tRNAs. (A) Quantification of readthrough efficiency at UAA and UAG codons in the two mutants lacking Ψ₃₅ and i⁶A₃₇ formation. At least five independent experiments were performed for each value. The box plot is done as in Fig. 2A. (B) Quantification of the relative proportions of the amino acids identified at the two stop codons as in Fig. 2B.

The Presence of a Pseudouridine Residue at Position 35 of the Tyr tRNA Is Important for Stop Codon Suppression.

In S. cerevisiae, there is only one cytosolic tyrosine tRNA, which acts as the major suppressor at the UAA and UAG stop codons, despite the creation of a G₃₄-A₃ and a G₃₄-G₃ mismatch, respectively. This tyrosine tRNA is unique because, in addition to being modified at position 37 (i⁶A), it is the only a suppressor in yeast that also has a pseudouridine (Ψ) instead of a usual uridine at the center (position 35) of the anticodon. This modification is posttranscriptionnally catalyzed by Pus7p (Fig. 1A).

The lack of this pseudouridine residue on suppression of the UAA and UAG stop codons was investigated as above for the other tRNA modifications. PUS7 deletion did not change tyrosine tRNA stability as measured by four-leaf clover (FL)–qRT-PCR (23) (Fig. S5), but resulted in lowering the level of stop codon suppression at UAG with almost no effect on stop codon suppression at the UAA codon (Fig. 3A). However, a quantification of amino acids incorporated at UAA and UAG codons revealed that the frequency of tyrosine tRNA incorporation at UAG was strongly decreased at the benefit of the noncognate glutamine or lysine tRNAs (Fig. S2A), even more significantly than in the ΔMOD5 strain (resulting in i⁶-lacking A₃₇). While in the case of UAA suppression, the incorporation of tyrosine tRNA is now almost exclusively replaced by the noncognate glutamine tRNA_UUG (Fig. 3B). We next checked whether the observed phenotype was solely due to the absence of Pus7p, by reexpressing either the WT protein or a catalytic mutant of Pus7p, Pus7-D256A [which has been reported to be inactive (19)] in the ΔPUS7 strain. Results indicate that reexpression of the WT protein restored parental amino acid levels, whereas expression of the single mutant did not (Fig. S4). Thus, both modifications of the tyrosine tRNA (i⁶A₃₇ and Ψ₃₅) are important for stop codon suppression, the i⁶A₃₇ modification being important only for UAA readthrough, whereas the Ψ₃₅ modification appears crucial mainly for the recognition of UAG codon. We also tested the absence of both modifications by constructing a strain in which the PUS7 and MOD5 genes were deleted. Subsequent quantification of stop codon readthrough efficiency (Fig. 3A) and amino acid insertion at the UAA and UAG codons (Fig. 3B) demonstrated a significant additive effect on stop codon readthrough, especially on the UAG codon.

Absence of the Pseudouridine Residue at Position 35 in the Tyr tRNA Does Not Prevent the Decoding of Tyrosine Codons.

In S. cerevisiae, synonymous tyrosine codons UAU and UAC are almost equally used in cytoplasmic mRNAs. Only one tyrosine tRNA isoacceptor containing a GΨA anticodon exists for reading each of these two codons. Because, tyrosine tRNA is dependent on Ψ₃₅ for decoding the two near-cognate stop codons (UAA and UAG, see above), it was of interest to determine how tyrosine tRNA lacking of Ψ₃₅ behaves for decoding the two tyrosine sense codons (UAU/C). We used a ribosome profiling technique that allows in vivo quantification of translation speed through calculation of the genome-wide ribosome residence time at each codon (24). We first checked the periodicity of the signal to demonstrate that the signals obtained corresponded to translating ribosomes (Fig. 4A). Then, in both the parental and ∆PUS7 strains, we compared the normalized ribosome density at each tyrosine and stop codons in the A site for each gene (Materials and Methods). We found that UAU and UAC were decoded with similar efficiency in both ∆pus7 and parental strains (P = 0.4 and 0.6, respectively) (Fig. 4B) and that this efficiency was not different from that for all of the other codons. We concluded that Ψ₃₅ in Tyr tRNA does not allow a better decoding of the near-cognate UAU codon, despite it playing an important role for decoding near-cognate UAA and UAG codons. Interestingly, we did not observe any significant difference in ribosome occupancy at stop codons (Fig. 4C), whereas a reduction of in-frame ribosome footprints downstream the stop codon is evident (P = 1.5 × 10⁻⁴) (Fig. 4D). This confirms that stop codon readthrough is impaired in the ∆pus7 strain without accumulation of ribosomes at natural stop codons when tyrosine tRNA is lacking Ψ₃₅.

Fig. 4.

Ribosome profiling and codon occupancy in the WT and ∆PUS7 strains. (A) Metagene analysis of 28-mers in the A site of the ribosome, between −20 and +100 nt from the start codon of each CDS. Phasing is indicated by a color scheme: phase 1 (blue), 2 (red), 3 (green). Periodicity is clearly visible for both strains, confirming that the footprints correspond to active ribosomes. (B and C) Quantification of the ribosome occupancy at tyrosine and stop codons in the A site. The lines in the center correspond to medians. “Other” corresponds to all codons other than UAU, UAC, and stop codons. n = 1,557 (UAU), 2,143 (UAC), 209 (UAA), 75 (UAG), 344 (UGA), 169,210 (Other) sample points. (D) Normalized ribosome-protected mRNA fragments (RPF) counts found downstream of the CDS stop codon and upstream the next in-frame stop codon (3′ extension). Box plot is done as previously indicated. Differences are significant according to a student test (P = 1.5 × 10⁻⁴). n = 593 WT and ∆pus7.

Discussion

Decoding of mRNAs by tRNAs into polypeptides on the ribosome was considered for a long time as a relatively simple process, where formation of a set of H-bonds between three complementary base pairs of codons and anticodons within a minihelix configuration plays a major role. However, it is now clear that other elements of tRNA molecules, as well as of rRNA of the ribosome, are also important for accurate and efficient mRNA decoding (5, 8, 25, 26). The biochemical properties of tRNAs depend strongly on posttranscriptional modifications of nucleotides within the anticodon (positions 34 and 35) and the so-called “proximal anticodon loop” of tRNA (positions 37, 38, and/or 32) (Fig. S6) (6, 15). The main role of these nucleotide modifications is generally argued to notably fine-tune the accuracy of decoding: that is, to restrict the decoding of split box codons (2:2 or 3:1 decoding boxes) or to extend the decoding of nonsplit four-codon boxes. However, it is now clear that, together with other bases of the anticodon loop, their roles are to mainly achieve a uniform ribosome binding by stabilizing the codon–anticodon interaction of tRNAs (27). This is especially important for tRNA harboring an A/U-rich anticodon triplet that have to reach about the same optimal interaction energies than tRNAs harboring G/C-rich anticodon. Base or ribose modifications lead to various chemical and physical consequences, such as keto-enol tautomerism, base protonation, uridine isomerization to pseudouridine (Ψ), anti to syn base transconformation, freezing a 2′-O-glycosidyl bond into its 3′-endo configuration, improvement of base stacking with neighboring bases, or additional interactions with ribosomal elements. One corollary of these multiple stabilizing effects is to allow certain noncanonical base pairs, which are isosteric with standard Watson–Crick pairs, to occur during decoding (28, 29).

In a preceding work, we and others identified in yeast eight naturally occurring tRNAs that misread stop codons as sense codons under in vivo conditions (Fig. 1) (16, 30). Although we are using a [PSI⁺] strain, it has been previously shown that this does not impact tRNA ratios found at the stop codon during readthrough (30). Here, we investigated the consequences of the absence of a given modification normally present in the anticodon loops (at positions 34, 35, and/or 37) on their relative efficacy to readthrough stop codons. This is a different situation from studies of missense errors at sense codons during the elongation process (3), because the near-cognate tRNAs of a stop codon will never be in competition with cognate tRNA that normally does not exist. All tRNAs in Fig. S6 belong to the so-called “intrinsically weak, A/U-rich” interacting tRNAs (Lys, Tyr) or “intermediate” interacting tRNAs (Gln, Arg, Trp, Cys), which are supposed to be among the most prone for miscoding (for details, see ref. 25). First, in the glutamine and lysine tRNAs, uridine-34 is modified to mcm⁵s²U (symbolized as sU*) and to mcm⁵U only (U*) in the arginine tRNA_UCU. The chemical adduct at C⁵ of U₃₄ has been shown to favor binding to G₃ over A₃, via enol-keto tautomerism, while the thiol (s²) group, beside reinforcing base pairing mostly with A₃, also reinforces stacking, and thus stability and conformational rigidity of both base 34 and the neighboring base pair N₃₅-N₂ (Fig. S6) (9, 10). Deletion (one by one) of each of the three genes responsible for U₃₄ modification into mcm⁵s²U₃₄ in S. cerevisiae (ELP3, TRM9, and TUC1) shows a slight growth defect for only the deletion of TRM9 (affecting only the last methylation step of cm⁵U₃₄) (Fig. S1). Deletion of these genes had only a slight impact on UAA readthrough efficiency, the most pronounced effect being observed for the ∆elp3 mutant corresponding to the total lack of 5-mcm group on U₃₄ (Fig. 2A). The interesting observation is that, while the efficacy is not much affected in both the ΔELP3 (lacking 5-mcm but not 2-thio) and ∆tuc1 (lacking 2-thio but not 5-mcm) strains, glutamine tRNA_UUG was no longer able to compete with tyrosine tRNA, now becoming almost the only suppressor tRNA for UAA readthrough, whereas in the ΔTRM9 strain (resulting in the lack of one methyl group on cm⁵s²U) a weak competition persists between glutamine and tyrosine tRNAs (Fig. 2B). These results demonstrate that both the fully methylated 5-mcm and 2-thio adducts of U₃₄ greatly improve the stability of the codon–anticodon interaction so that a noncanonical G₃₆:U₁ can occur, allowing glutamine tRNA_UUG to be stably incorporated at UAA. Because all of the base pairs of the minihelix have to adopt a strict Watson–Crick geometry (5, 26), this G₃₆:U₁ pair probably fulfill this requirement through a keto-enol tautomerization of either base (31). The data obtained with lysine tRNA reading the same UAA are going in the same direction. Only when U₃₄ is fully modified to mcm⁵s²U₃₄, a U₃₆:U₁ opposition in this case can be accommodated, yet obviously less efficiently than for a G₃₆:U₁. Such a U:U opposition, with possibly only one single H-bond within a Watson–Crick helix geometry, seems to require a water molecule (28).

Slightly different results were obtained for reading UGA stop codon, involving a strong middle C-G pair, by the minor arginine tRNA_UCU harboring also a 5-mcm residue but not a 2-thio group at U₃₄ (U*). ELP3 deletion (absence of 5-mcm on U₃₄) did not alter significantly the ratio of amino acids inserted at UGA (Fig. 2B), whereas both arginine and cysteine incorporation levels were a bit higher in the ∆trm9 strain (lacking of a methyl group on cm⁵U₃₄). The better incorporation of arginine tRNA at UGA may originate from the generation of a free carboxyl group due to the lack of the methyl on cm⁵U₃₄. However, this conclusion has to be taken with caution because a slight increase in cysteine incorporation has also been observed, while the cysteine tRNA is not normally modified by Trm9p. We believe instead that the strong C₃₅-G₂ pair is the main stabilization element and the adduct on U34 only slightly contributes to the ability of arginine tRNA to play a role of UGA suppressor. These results point out the importance of considering the global energetic and stability of a codon–anticodon interacting system, including the proximal extended anticodon (25), before generalizing any observation.

The opal UGA codon is decoded mostly by tryptophan tRNA, followed by cysteine tRNA, both belonging to the same four-codon box. In tryptophan tRNA, A₃₇ is unmodified while in cysteine tRNA it is modified by Mod5p to i⁶A₃₇. Conversely, in tryptophan tRNA, C₃₄ is modified to Cm₃₄ (as well as Cm₃₂) by Trm7p, while G₃₄ in cysteine tRNA is not modified. The presence of a hydrophobic isopentenyl group on A₃₇ is known to reinforced the stacking power of the purine adenine on the adjacent base pair A₃₆-U₁ (both by lateral and interstrand stacking, thus with the first codon base) (32). A 2′-O-metyl group on the hydroxyl of C₃₄ is known to favor a more rigid 3′-endo conformation of the ribose, confines the H-bound formation, and increases hydrophobic surface, hence also stacking with neighboring bases (33). No significant effect of TRM7 deletion (lacking Cm₃₄) is found on either global stop codon readthrough efficiency (Fig. S3A) or the nature of the amino acids incorporated, attesting again that the strong central C₃₅-G₂ pair rather than the 2′-O-methylation of C₃₄ (and C₃₂) is probably the main stabilizing element (Fig. S3B). An isosteric C₃₄ or Cm:A₃ opposition probably occurs through an amino ionization of either C/Cm or A; alternatively, Cm:A₃ can also occur in the absence of a hydrogen bond (28, 29).

In contrast, deletion of the isopentenyl group on A₃₇ of cysteine tRNA in the ΔMOD5 strain prevents incorporation of cysteine at UGA stop codon (Fig. S3B). The same trend was observed with i⁶A₃₇ containing tyrosine tRNA, the suppression efficiency being more severely affected in the case of UAA than of UAG (Fig. 3B). Thus, lowering the stability of a A₃₆-U₁ Watson–Crick pair by the lack of an isopentenyl group on A₃₇ in cysteine or tyrosine tRNA affects the efficacy of a triplet pairing involving a G₃₄ mispair with A₃ of stop codon UGA and UAA, respectively, or with G₃ in the case of stop codon UAG. Formation of an isosteric base pair involving two purines requires that one of the two purines switches conformation from anti to syn, probably the one of the codon, to form a Watson–Crick/Hoogsteen base pair (28, 29). Notice that codon–anticodon involving noncanonical G₃₄:A₃ or G₃₄:G₃ have been demonstrated for several tRNAs involving G/C-rich anticodons and reading codons of the unsplit four-decoding boxes. For a long time, this type of decoding process has been designated as a “two-out-of-three” decoding rule (34).

Tyrosine tRNA carries an unusual Ψ at the central position of the anticodon that is catalyzed by Pus7p. Pseudouridine enables engaging a strong base pairing (almost as strong as C-G), rigidifies the sugar-phosphate backbone, and improves stacking with a neighboring base pair, forcing a Ψ-A pair to adopt an A-form conformation, as in a genuine Watson–Crick helix (35, 36). In agreement with an earlier similar observation in a plant translation system (37), we found that tyrosine tRNA in a ∆Pus7 strain (lacking Ψ₃₅) is a less-efficient suppressor at both UAA and UAG stop codons, but mainly at UAA (Fig. 3B), although the tyrosine tRNA is equally stable in WT and ∆Pus7 strains (Fig. S5). The effect of Ψ₃₅ appears independent of A₃₇ isopentenylation, as shown by the additive effect of the PUS7 and MOD5 double deletion on UAG (Fig. 3B). The level of UAA readthrough was not affected by PUS7 deletion, even though this deletion prevented the incorporation of the tyrosine tRNA at UAA (Fig. 3). These results illustrate that a strong stabilization of codon–anticodon interaction allows a noncanonical G₃₄:A₃ base opposition to occur. Interestingly, a ribosome profiling (RiboSeq) experiment performed in parental WT and ∆PUS7 strains allows us to demonstrate that a tyrosine tRNA isoacceptor (with or without Ψ₃₅) reads equally well both synonymous tyrosine codons, UAU and UAC (Fig. 4). This indicates the lack of involvement of Ψ₃₅ in discrimination between these two tyrosine codons in a situation where there is no real mismatch between the anticodon and the codon (whereas a G₃₄:A₃ or G₃₄:G₃ base opposition systematically occurs when tyrosine tRNA reads stop codons). A similar situation exists with echinoderm asparagine tRNA_GUU harboring a Ψ₃₅ in reading a lysine near-cognate codon AAA, while not affecting the normal reading of asparagine codons AAC/U (38).

In conclusion, our findings reveal the importance of certain modifications for the suppression of stop codons. Consistent with the findings of another recent study (3), we found that tRNA modifications did not solely serve to restrict the decoding capacity of the tRNA to its cognate codon, but also allow the decoding of near-cognate stop codons. This involves an expected mismatch at the wobble position and also noncanonical base pairs at the third anticodon position and the first codon position. The allowed base mispairs or base oppositions are, however, only those that would mold within a minihelix of the Watson–Crick type of geometry that is mandatory for the aminoacyl-tRNA to be accepted and finally captured by the mRNA-ribosome machinery (26). We observed that mismatch with a stop codon never occurred at the middle position of the anticodon, while such possibilities have been demonstrated at sense codons (6, 7). The reason is that there is no naturally occurring tRNA that could sustain such base opposition (like U:G) within a Watson–Crick minihelix in the whole tRNA repertoire of S. cerevisiae.

Only a small subset of tRNAs have been analyzed and it is likely that studies of other tRNAs in other translation systems, as well as organisms, will turn up other surprises. In this work, we propose that the activities of certain tRNA modification enzymes can be a regulatory device for the production of certain functional “readthrough” proteins. This work should help at elaborating synthetic or mutated tRNAs able to introduce nonproteinous amino acids at specific locations of a mRNA where a sense codon has been appropriately mutated into a stop codon (39).

Materials and Methods

Detailed information on materials and methods used in this study is provided in SI Materials and Methods.

Strains and Plasmids.

All of the strains used in this study were derived from 74-D694 (mata ade1-14 [UGA] ura3-52 trp1-289 [UAG] his3∆200 leu2-3,112 [PSI⁺]).

Mass Spectrometry.

LC-MS/MS analyses.

Proteolytic peptides were identified and quantified with a Triple-TOF 4600 mass spectrometer (ABSciex) coupled to the nanoRSLC system (Thermo Fisher Scientific) equipped with a trap column (Acclaim PepMap100C18, 75 μm i.d. × 2 cm, 3 μm) and an analytical column (Acclaim PepMapRSLCC18, 75 μm i.d. × 25 cm, 2 μm, 100 Å).

Relative quantification of readthrough peptides.

The intensity of each chromatographic peak was corrected by a factor taking the ionization and digestion efficiencies of each readthrough peptide into account (16). The means of three technical replicates are reported.

RNA-Seq and Ribosome Profiling.

Cells were grown to an OD₆₀₀ of 0.6 in 1 L of liquid glucose-YNB supplemented with CSM and 2× adenine and flash-frozen. Total RNA and polysomes were extracted as previously described (40).