|Home | About | Journals | Submit | Contact Us | Français|
Accurate transfer RNA (tRNA) aminoacylation by aminoacyl-tRNA synthetases controls translational fidelity. Although tRNA synthetases are generally highly accurate, recent results show that the methionyl-tRNA synthetase (MetRS) is an exception. MetRS readily misacylates non-methionyl tRNAs at frequencies of up to 10% in mammalian cells; such mismethionylation may serve a beneficial role for cells to protect their own proteins against oxidative damage. The Escherichia coli MetRS mismethionylates two E. coli tRNA species in vitro, and these two tRNAs contain identity elements for mismethionylation. Here we investigate tRNA mismethionylation in Saccharomyces cerevisiae. tRNA mismethionylation occurs at a similar extent in vivo as in mammalian cells. Both cognate and mismethionylated tRNAs have similar turnover kinetics upon cycloheximide treatment. We identify specific arginine/lysine to methionine-substituted peptides in proteomic mass spectrometry, indicating that mismethionylated tRNAs are used in translation. The yeast MetRS is part of a complex containing the anchoring protein Arc1p and the glutamyl-tRNA synthetase (GluRS). The recombinant Arc1p–MetRS–GluRS complex binds and mismethionylates many tRNA species in vitro. Our results indicate that the yeast MetRS is responsible for extensive misacylation of non-methionyl tRNAs, and mismethionylation also occurs in this evolutionary branch.
Translational fidelity, which is critical for cell survival, depends on the incorporation of the correct amino acid to its transfer RNA (tRNA). Aminoacyl-tRNA synthetases (aaRSs) aminoacylate tRNAs with their cognate amino acids. Despite the high fidelity of catalytic and editing domains in aaRSs, misacylation can occur when the aaRS catalyzes the aminoacylation of a cognate tRNA with a non-cognate amino acid (1). aaRS can also misacylate tRNAs through catalyzing the aminoacylation of a non-cognate tRNA with a cognate amino acid. Misacylated tRNAs that are used in translation produce mutant proteins. However, mistranslation at low levels is not always detrimental to cells and in some cases is tolerated or may even be beneficial in stress response (2,3).
A recent study has provided evidence that tRNA misacylation with methionine is actively regulated in mammalian cells and tRNA mismethionylation may provide a benefit to cells under oxidative stress (4). Previous findings show that genetically encoded methionine residues can protect proteins against reactive oxygen species by oxidation of methionine residues on the surface or near active sites of proteins (5,6). Mismethionylated tRNAs can extend this protective function by substituting certain non-methionine residues in proteins at strategic locations. For this function, methionine substitution through mismethionylated tRNAs is likely to occur at solvent-exposed residues on the surface or near the active site of target proteins; such residues more commonly have charged or polar side chains.
tRNA mismethionylation occurs in both prokaryotes and eukaryotes. The recombinant, purified E. coli methionyl-tRNA synthetase (MetRS) mismethionylates two E. coli tRNAs in vitro, one coding for a charged amino acid, tRNAArgCCU, and the other for a polar amino acid, tRNAThrCGU (7). In mammalian cells, MetRS is part of an 11 protein complex, eight of which are aaRSs (8,9). The affinity purified human complex containing MetRS can mismethionylate two human tRNAs coding for a charged amino acid, tRNALysCUU and tRNALysUUU (4). In mammalian cells, mismethionylation occurs at a basal level of ~1%, which includes tRNAs coding for several charged amino acids. This level of mismethionylation increases to up to 10% under innate immune activation and chemically triggered oxidative stress (4).
Here we investigate tRNA misacylation with methionine in baker’s yeast, Saccharomyces cerevisiae, both in vivo and in vitro. Saccharomyces cerevisiae is evolutionary distant from mammals and bacteria, so studies of yeast broaden the evolutionary reach of this unusual behavior in altering the aminoacylation and translational fidelity. Using tRNA microarrays and classical filter retention assays, we identified many mismethionylated tRNAs in S. cerevisiae. The extent of misacylated tRNAs is dependent on yeast growth conditions. Mismethionylated tRNAs show similar in vivo utilization kinetics as correctly charged tRNAMets, and several misacylated peptides are detected by mass spectrometry, indicating that they are used in translation. The yeast MetRS is part of a three protein complex that includes a general tRNA-binding protein, Arc1p and the glutamyl-tRNA synthetase [GluRS, (10)]. Previous reports have shown that the anticodon is not sufficient for tRNA binding to the yeast MetRS; efficient aminoacylation with methionine requires primary, secondary and tertiary determinants in the tRNA (11–13). The Arc1p protein binds to the MetRS and GluRS by N-terminal interactions and to tRNA by C-terminal interactions, which facilitate tRNA binding to the MetRS and GluRS for aminoacylation with methionine and glutamic acid, respectively (10). However, since Arc1p binding to tRNA is non-specific, tRNA binding to the MetRS determines the specificity of aminoacylation (10,14). We examine tRNA binding to the recombinant Arc1p-GluRS-MetRS (AME) complex using tRNA microarrays and show that AME can bind almost all yeast tRNAs, consistent with the previous reports. We also show that AME extensively mismethionylates many tRNAs in vitro. Our results suggest that MetRS being a part of a multi-protein complex provides eukaryotes with another mechanism of tRNA mismethionylation by allowing many tRNAs to bind to the MetRS for misacylation with methionine.
Yeast strains used in this study were WY798 (MATα URA3 LEU2 TRP1) (15), BY4742 (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) and ΔArc1 (MATa his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0 arc1Δ). BY4742 and ΔArc1 were purchased from Open Biosystems. Yeast strains were grown on YPDA medium at 30°C. Single colonies were grown overnight in synthetic complete (SC) media. For stationary phase experiments, these cultures were grown to OD600 4–8. For mid-log phase experiments, the overnight cultures were diluted to OD600 0.1 and grown to OD600 0.4–0.5.
35S-Met pulse labeling of yeast cells was adapted from established procedures with minor modifications. Briefly, yeast cells were starved for methionine by spinning down and resuspending in an equal volume of SC–Met media 1 h prior to pulse labeling to maximize 35S-signal. After pelleting, yeast cells were resuspended in 300 μl of pulse labeling media consisting of 0.02 mCi/OD600 35S-Met (Perkin-Elmer, Boston, MA) in SC–Met. OD600 12.5 cells were typically sufficient to yield ~100 μg RNA. Pulse labeling proceeded at 30°C for 1 or 8 min. For chase experiments, 300 μl of SC–Met supplemented with 1 mg/mL fresh methionine, and 200 μg/mL cycloheximide, where appropriate, was added after the pulse period and incubated at 30°C for 1 min. Reactions were stopped by addition of 300 μl ice cold 0.3 M sodium acetate/acetic acid buffer with 10 mM EDTA, pH 4.8, and submersion in ice, after which cells were further rinsed twice with the same buffer. For cycloheximide treatments, 200 μg/mL cycloheximide was maintained in the acetate/EDTA buffer solution throughout washes and lysis to maintain translational arrest.
For in vitro experiments, total RNA was isolated from yeast grown to stationary phase overnight in YPDA medium, pelleted, resuspended in 300 µl 0.3 M KCl, 50 mM KOAc, and transferred to a tube containing 300 µl acetate-saturated phenol-CHCl3, pH 4.8 and 0.5 mm acid-treated glass beads. The sample was vortexed three times by alternating vortexing for 1 min and incubating on ice. The sample was then spun at 14 000 rpm for 15 min at 4°C, transferred to a new tube containing 300 µl acetate-saturated phenol-CHCl3, pH 4.8, and vortexed for an additional 1 min. The sample was spun at 14 000 rpm for 10 min at 4°C, and the aqueous layer was transferred to a clean tube, ethanol precipitated twice, and resuspended in 10 mM Tris, pH 7.5, 1 mM EDTA.
Following pulse labeling in vivo, total RNA was isolated from yeast by transferring the sample to a clean tube containing 300 µl acetate-saturated phenol-CHCl3, pH 4.8, and 1 Retsch 7 mm stainless steel ball and vortexing at room temperature for 30 min. The sample was then spun at 14 000 rpm for 15 min at 4°C and followed the remaining procedure as in the in vitro experiments. Once resuspended in 10 mM Tris, pH 7.5, 1 mM EDTA, the RNA was again spun at 14 000 rpm for 15 min at 4°C and transferred to a clean tube.
A plasmid overexpressing the AME complex under IPTG control was transformed into BL21 DE3 E. coli cells. The cells were grown in LB with 100 mg/L ampicillin until OD600 0.6, and then overexpression was induced with 0.2 mM IPTG at 37°C. Expression continued for 4 h, and the cells were then harvested. Cells were lysed in lysing buffer (50 mM K-HEPES, pH 7.6, 30 mM NaCl, 5 mM β-mercaptoethanol) in the presence of protease inhibitors and 2000 U DNase per 50 mL extract. Following centrifugation, the complex was purified by FPLC by elution from a Ni-NTA column using an imidazole gradient. The purification buffers contained 50 mM K-HEPES, pH 7.6, 150 mM NaCl, 5% glycerol, 10 mM BME, and 20 mM or 500 mM imidazole. The complex eluted around 300 mM imidazole.
The affinity purified AME complex was passed through a Superdex 200 column at 4°C to analyze by gel filtration using the buffer containing 20 mM Tris, pH 7.4, 30% glycerol, 2 mM DTT and 1 M NaCl.
Saccharomyces cerevisiae tRNAMetCAU, tRNAGluCUC, tRNAGluUUC(12) and tRNAGluUUC(1) sequences were obtained from the genomic tRNA database (16). Mutants 1-3 were created by swapping nucleotides from the tRNAGluCUC and tRNAGluUUC(12). For tRNAMetCAU, the transcript was made by in vitro transcription of overlapping oligonucleotides and purified as described previously (17,18).
All mature tRNAGlu start with U at the 5′-position. Since T7 RNA polymerase transcription works poorly with U-starting RNA, the three tRNAGlu transcripts and mutants 1-3 were first transcribed similarly as the tRNAMet transcript but with a 5′-leader sequence of 5′-gggacaaata-tRNAGlu. These transcripts were then cleaved with Bacillus subtilis RNase P holoenzyme to obtain the appropriate tRNA sequences. The cleavage was performed by first reconstituting the B. subtilis RNase P holoenzyme (19). The final buffer concentration of the reconstituted holoenzyme was 50 mM Tris-HCl, pH 8, 18 mM MgCl2, 0.2 M NH4Cl. The holoenzyme was reconstituted by first mixing P RNA with water and Tris–HCl, pH 8, and heating at 85°C for 2 min, then at room temperature for 3 min, followed by adding MgCl2 and incubating at 50°C for 5 min, and finally adding equal moles of P protein and NH4Cl and incubating at 37°C for 5 min. The transcription mixture was then incubated with the reconstituted B. subtilis RNase P at 37°C for 5 min. Cleavage was stopped with 15 mM EDTA, and the cleaved transcription mixture was ethanol precipitated and purified by denaturing PAGE.
Filter-based aminoacylation reactions with methionine were performed at 30°C in 20 mM K-HEPES (pH 7.2), 100 μM methionine, 10 mM MgCl2, 5 mM DTT, 4 mM ATP, 150 mM NH4Cl, 0.1 mM EDTA, 0.5 μCi/mL l-[35S] methionine, 0 or 2.5 μM tRNA transcripts and 0.1 μM AME enzyme. Filter-based aminoacylation reactions with glutamic acid were performed at 30°C in 100 mM K-HEPES (pH 7.2), 100 μM glutamic acid, 10 mM MgCl2, 10 mM DTT, 2 mM ATP, 30 mM KCl, 50 μM l-[3H] glutamic acid, 0 or 2.5 μM tRNA transcripts and 0.1 μM AME. Filters were counted on a Perkins–Elmer scintillation counter. For kinetics experiments, aminoacylation reactions were performed with up to 15 µM transcripts.
In vitro aminoacylation of tRNA for microarray analysis was performed at 30°C for 6 or 20 min in 20 mM K-HEPES (pH 7.2), 100 mM NH4Cl, 0.1 mM Na-EDTA, 2 mM ATP, 1.5 mM MgCl2, 2.5 mM DTT, 1 mCi/mL l-[35S]methionine (Perkins–Elmer, Boston, MA), 0.4 mg/mL gel-purified total yeast tRNA and 0.5 µM purified AME enzyme.
Total tRNA was gel-purified from total RNA isolated from yeast at pH 4.8. The purified tRNA was dephosphorylated with calf intestinal phosphatase in 50 mM Tris, pH 8, 0.1 mM EDTA, extracted from phenol/CHCl3 and ethanol precipitated. The tRNA was 5′-32P-labeled with T4 PNK and renatured. Binding experiments contained 10 pmol AME with excess (40 pmol total) renatured tRNA doped with 5′-32P-labeled tRNA in 0.1 M K-HEPES, pH 7.2, 1.5 mM Mg2+ and 0.1–0.4 M KCl. The binding mixture was incubated at 30°C for 10 min and then added to Genscript Ni2+-MagBeads following the Genscript protocol 2.1.2 for purification of polyhistidine-tagged proteins under native conditions. The wash and elution buffers were the same as used in the Ni-NTA purification of the AME complex. The eluted tRNA was ethanol precipitated and analyzed by tRNA microarray.
Hybridization to microarrays and controls using radioactive detection on a Genomic Solutions Hyb4 station has been described previously (4). Mismethionylation and tRNA binding to AME for yeast was assessed with manually printed arrays containing 40 nuclear and 24 mitochondrial probes for S. cerevisiae and 31 probes for E. coli as controls. The arrays contained eight replicates for each probe. Experiments with 35S-Met detection used 20 µg total RNA per array. Signals were quantified using Fuji Imager software.
Mass spectrometry data and FASTA sequences for nine abundant yeast proteins (ADH1, CDC19, ENO1, ENO2, FBA1, PDC1, PGK1, TDH2 and TDH3) were obtained from Geiler-Samerotte et al. (20) and were analyzed by MaxQuant (21). Additional FASTA sequences were created for each protein with one methionine substitution at each lysine and arginine residue. As trypsin-digested peptides are cleaved at Lys and Arg residues, the MaxQuant data were analyzed for longer peptides representing a methionine misincorporation at the lysine and arginine residues.
To determine if tRNAs are misacylated in S. cerevisiae, we chose to work first with the S. cerevisiae strain 798, a fully prototrophic strain whose tRNA abundance and charging characteristics have been characterized previously (15). We detected tRNAs that are either correctly or incorrectly aminoacylated with methionine after pulse labeling with 35S-Met using arrays containing probes for all cytosolic and mitochondrial tRNAs of S. cerevisiae. The array includes eight repeats each of 40 cytosolic and 24 mitochondrial S. cerevisiae tRNA probes. In addition, the array includes eight repeats each of 1 blank control and 31 E. coli tRNA probes, which serve as negative controls. The probe sequences used were identical to those described previously to measure tRNA charging in yeast and E. coli (22,23).
We observed 35S-signals from numerous yeast tRNA probes for both methionyl and non-methionyl-tRNAs (Figure 1A). The most intense signals were derived from both cytosolic tRNAMets and both mitochondrial tRNAMets, as expected. The strongest signal was derived from cytosolic elongator tRNAMete. Signals from mitochondrial tRNAs were weaker than tRNAMete, presumably due to the significantly lower abundance of these tRNAs. Unexpectedly, signal from cytosolic initiator tRNAMeti was much weaker than that of tRNAMete, even though the abundance and charging levels of tRNAMeti and tRNAMete should be similar according to previous studies (23,24). Varying pulse labeling time from 1 to 8 min did not markedly change this behavior (data not shown). This result is significantly different from mammalian studies where similar levels for 35S-charging signals were observed for tRNAMeti and tRNAMete (4). At this time, we do not understand the reasons for the low 35S-detected charging levels of tRNAMeti in pulse labeling. One possible explanation is that yeast cells may use distinct intracellular methionine pools to charge tRNAMeti and tRNAMete; our result would be consistent with methionine used for tRNAMete coming from immediate Met uptake and methionine used for tRNAMeti coming from Met obtained or de novo synthesized at earlier times.
We performed a series of controls to ensure that the majority of the 35S-signals present in non-methionyl-tRNAs are derived from mismethionylated tRNAs (Figure 1B) as was done previously for the mammalian misacylation study (4). To rule out signals due to cross-hybridization, we added excess oligonucleotides complementary to all cytosolic and mitochondrial tRNAMet to the hybridization mixture. All signals from Met-tRNAs could be eliminated with little change in the signal intensity from non-Met-tRNAs. To rule out signals due to peptidyl-tRNAs with an N-terminal 35S-methionine, we treated the RNA sample with aminopeptidase-M before array hybridization. Most signals from non-Met-tRNAs remain after aminopeptidase treatment (Figure 1C). Post-transcriptional thio-modifications of tRNA may be radio-labeled via catabolism of 35S-Met. To distinguish signals due to post-transcriptional thio-modifications of tRNA, we hydrolyzed the labile aminoacyl bond of the charged tRNA sample at pH 9 prior to array hybridization to remove all 35S-signal due to methionine charging. Yeast tRNAs known to contain thio-modifications at the wobble position of the anticodon include tRNALysUUU, tRNAGluUUC, tRNAGlnUUG (5-methoxycarbonylmethyl-2-thiouridine or mcm5s2U), tRNAArgUCU and tRNAThrIGU (2-thio-U or s2U (25)). Signals from these five tRNAs were readily detected after deacylation. Surprisingly, signals from several other tRNAs were also present after deacylation, including tRNALysCUU, tRNAIleIAG, and tRNAThrUGU; these tRNAs do not contain known thio-modifications (Figure 1C). It remains to be determined whether these deacylation resistant 35S-signals are also derived from thio-modifications in these tRNA.
We performed one additional control to ensure that misacylation observed in yeast is not caused by methionine starvation in the standard pulse-labeling protocol (Figure 1D). In the standard pulse-labeling experiments, yeast cells were first starved for methionine for 1 h before the addition of 35S-Met. This step decreases the intracellular pools of cold Met, resulting in increased specific activity of Met in labeled protein or RNA. This step was not necessary for mammalian cells because they are intrinsically auxotrophic for Met, and the intracellular pools of cold Met is much lower. Misacylation still occurred when yeast had not been starved prior to pulse labeling, although signals for both Met and non-Met-tRNAs were substantially reduced.
A summary of the misacylation result is shown in Figure 1C. Approximately two-thirds of all non-Met-tRNA probes show detectable 35S-signals. Alkaline deacylation prior to array hybridization removed signals from ~70% of these probes, suggesting that the majority of these signals are derived from mismethionylation. No mismethionylation was detected for all mitochondrial tRNAs. At the stationary phase, the cumulative extent of misacylation is over 10% relative to all Met-tRNAMet signals. Since the 35S-signal for tRNAMeti is significantly lower in yeast than in mammalian cells and the misacylation extent is normalized to 35S-signals of all Met-tRNAs, this level of mismethionylation in yeast is comparable to the highest level observed in mammalian cells.
We determined whether tRNA mismethionylation depends on cell growth conditions in S. cerevisiae (Figure 2A and B). The growth state of yeast is known to impact numerous cellular factors, including gene expression, metabolic rate and oxidative stress load (26–28). In addition, the stationary phase is thought to impose an oxidative stress relative to the mid-log phase (29). Yeast 798 strain was grown to either mid-log (OD600~0.5) or stationary (OD600~4–8) phase and pulse-labeled with 35S-Met. Mismethionylation was observed under both conditions; however, a greater extent of misacylation occurred in the stationary phase than in the mid-log phase (Figure 2B). This result indicates that the extent of misacylation depends on the yeast growth phase. In mammals, mismethionylation is increased upon innate immune or chemically triggered oxidative stress.
In vitro experiments using HeLa multi-synthetase complexes containing MetRS suggest that the higher order structure of MetRS may play a role in regulating misacylation (4). Yeast MetRS is part of a three protein complex, including the GluRS and a general tRNA binding protein Arc1p. We tested a potential role of Arc1p in tRNA mismethionylation by performing pulse-labeling experiments with an Arc1 null yeast strain. For this purpose, we switched to yeast cells in the BY4742 background, where the wild-type and the isogenic ΔArc1 strains are readily available. No difference in misacylation relative to the wild-type BY4742 strain was detected (Figure 2C). This result suggests that the MetRS alone may be able to methionylate non-methionyl tRNAs in vivo.
Aminoacylation of tRNA does not guarantee its use in translation. At least in bacteria, the utilization of an aminoacyl-tRNA depends upon a compromise of specific tRNA and their charged amino acid interactions with the elongation factor and on the ribosome (30,31). Very little is known how yeast elongation factor (eIF1α) and yeast ribosome choose how misacylated tRNA is utilized in translation. We performed one experiment using cold chase and cycloheximide to test whether mismethionylated tRNAs are likely used in translation in yeast (Figure 3). Cycloheximide inhibits ribosome elongation and is widely used in cellular studies of protein synthesis. To examine the kinetics of turnover of mismethionylated tRNAs, cells were first pulse-labeled with 35S-Met, followed by a rapid cold chase of a large excess of non-radioactive Met in the absence and presence of cycloheximide. The resulting 35S-labeled tRNAs were then examined by microarrays. In the absence of cycloheximide, signals from both Met-tRNAs and non-Met-tRNAs were reduced by ~40-fold, suggesting that both types of charged tRNAs are turned over with similar kinetics in cells. In the presence of cycloheximide, the amount of 35S-labeled tRNAs for the 798 strain is reduced by <1.8-fold for Met-tRNAs and non-Met-tRNAs (Figure 3C). This result shows that inhibition of translation also inhibits the turnover kinetics of mismethionylated tRNAs to a similar extent as the turnover of correctly charged Met-tRNAs, consistent with mismethionylated tRNAs being used in translation in yeast cells.
To determine that misacylated tRNAs are indeed used in translation in yeast cells, we analyzed mass spectrometry data from Geiler-Samorette et al. (20) using MaxQuant (21). We chose to analyze the peptides for some of the most abundant yeast proteins, including nine proteins involved in glycolysis and fermentation (Figure 4A). The proteins had been trypsin-digested, which cleaves the peptides at Lys and Arg residues, prior to mass spec analysis. Since lysyl- and argininyl-tRNAs are two of the misacylated tRNA species, we chose to look for longer peptides representing methionine misincorporation at Lys and Arg residues. We found low abundant peptides representing misincorporation of methionine at both Lys and Arg residues in seven of these nine proteins at a frequency of 0.66% of all observed peptides for these proteins. Example spectra are shown in Figure 4B for a wild-type and its mistranslated peptide from pyruvate kinase, CDC19. The Arg codon at this misincorporated position in CDC19 is AGA. AGA is read by tRNAArgUCU, which is the only tRNAArg isoacceptor that shows high levels of misacylation (Figure1C). These results indicate that the misacylated tRNA species are used in translation.
We used recombinant, purified yeast AME complex from E. coli to demonstrate that this complex is sufficient to mismethionylate yeast tRNAs (Figure 5). In vitro aminoacylation with S. cerevisiae tRNAMet previously used yeast MetRS alone in the presence or absence of Arc1p, not the full AME complex (10,32). To better recapitulate cellular conditions, we used the AME complex for all of our in vitro aminoacylation studies. The reaction scheme involves incubating the purified AME complex with total yeast tRNA with 35S-Met, followed by hybridization of the reaction mixture on the microarray (Figure 5A). Our affinity purified AME complex was derived from an overexpression plasmid in E. coli; it contains an amino-terminal His6 tag on the Arc1p protein. Since the His6 tag is only on the Arc1p protein, excess Arc1p appears to be purified with the complex during Ni-NTA affinity purification, as seen by SDS-PAGE (Figure 5B). As expected, essentially all synthetase molecules are associated with Arc1p, and the majority of both synthetases form a single peak in size exclusion chromatography (Figure 5C).
We charged total RNA isolated from stationary phase yeast with 35S-Met using the purified AME complex (Figure 5D). Many tRNA species were misacylated and mis-methionylation increases over time (Figure 5E). All the misacylated tRNA species in vitro were also misacylated in vivo. The greater extent of misacylation in vivo may be due to the AME complex association with polysomes, although this remains to be determined. The mammalian multi-synthetase complex is associated with polysome (33), and the polysome-associated multi-synthetase complex misacylates more tRNA species in vitro compared to the multi-synthetase complex alone (4). When the in vitro charging sample was first deacylated at pH 9 followed by array hybridization, no deacylation-resistant tRNA was observed, as expected due to the lack of post-translational modification enzymes in the in vitro charging mixture (Figure 5E). Quantification of the in vitro misacylated tRNA species showed high levels for tRNA species mostly coding for charged and polar amino acid side chains (Figure 5D). Although a very extensive pattern of in vitro misacylation is present, still more mismethionylated tRNA species are observed in vivo than in vitro. Similar results have been seen for tRNA mismethionylation in mammalian systems (4).
The Arc1p in the AME complex is a generic tRNA-binding protein and assists tRNA binding to the MetRS and GluRS for aminoacylation (10,14). The extensive level of misacylated tRNA species in S. cerevisiae suggests that Arc1p may also shuttle non-Met-tRNAs to the MetRS to be aminoacylated with methionine. We performed an experiment to compare tRNA binding versus mismethionylation by the AME complex (Figure 6). We first incubated 5′- 32P-labeled total yeast tRNA with the purified AME complex at varying concentrations of KCl followed by affinity pull down of the AME complex. Bound 32P-labeled tRNAs were then examined by microarray (Figure 6A and B). Almost all detectable tRNA species are bound by AME at 0.1 M KCl; increasing KCl to 0.4 M reduced the number of tRNA species bound to the complex as expected (Figure 6C). This result is consistent with the ability of AME to bind essentially any tRNA (32). Among the bound tRNAs at 0.1 M KCl, only a subset is mismethionylated, suggesting that mis-methionylation has additional requirement beyond simple binding.
Finally, we applied the classical filter-based aminoacylation assays to confirm that the AME complex is capable of misacylating unmodified tRNA transcripts at high efficiency (Figure 7). We chose to work with transcripts of tRNAMete (Figure 7A, left) and variants of tRNAGlu (Figure 7A, middle) because the AME contains both MetRS and GluRS. The reference yeast genome contains two tRNAGlu isoacceptors: 2 copies of tRNAGluCUC and 13 copies of tRNAGluUUC. The tRNAGluUUC isoacceptor family has two isodecoders, a major form with 12 copies and a minor form with a single copy (16,34). The tRNAGluUUC(12) and tRNAGluUUC(1) isodecoders differ by one nucleotide in their acceptor stems, and the isoacceptor tRNAGluCUC differs by five nucleotides from tRNAGluUUC(12). These three tRNAGlus were not distinguishable on our microarrays due to their sequence similarities. We also made three tRNAGlu mutants (Figure 7A, right) in the acceptor stems and anticodon loop to probe possible sequence identity elements of misacylation.
The purified AME complex charged all the tRNA transcripts with their cognate amino acids, tRNAMete with methionine or all tRNAGlus with glutamic acid (Figure 7B). All three tRNAGlu variants as well as all three tRNAGlu mutants were mismethionylated at similar levels under this reaction condition. As a positive control, very little mischarging of tRNAMete with Glu is present, showing that misacylation occurs exclusively by the MetRS (Figure 7B). This result suggests that all three tRNAGlu variants could have contributed to the total 35S-signal for the tRNAGlu on our microarray. We also measured the charging kinetics of tRNAMete and the major form of tRNAGlu with Met or Glu (Figure 7C). A two-phase charging kinetics was observed in all cases. Both phases have essentially the same Km values, but they differ by over 100-fold in kcat values. The reason for such two-phase charging behavior by the purified AME complex is unclear. However, both the Km and the kcat values for the fast phase is within the same order of magnitude observed previously, although all previous experiments were performed with just the MetRS or GluRS protein not the AME complex. The AME charged tRNAMete with Met at kcat/Km of 0.39 μM−1s−1 and tRNAGluUUC(12) with Glu at kcat/Km of 1.1 μM−1s−1, indicating that our recombinant AME is highly active. The AME mischarged tRNAGluUUC(12) with Met at kcat/Km of 0.2 μM−1s−1, which is only 2-fold lower than cognate charging. This result is significantly different from in vitro mischarging of the AME complex using total yeast tRNA (Figure 5D), which suggests that certain modifications in tRNAGlu may significantly reduce mismethionylation.
We have demonstrated here that tRNA misacylation with methionine occurs in S. cerevisiae. This represents the second example of tRNA mismethionylation in cells in addition to mouse and humans. Since yeast is evolutionarily distant from mammals, this result suggests that tRNA mismethionylation is conserved from fungi to mammalian lineages. We also show that tRNA mismethionylation is derived from the activity of the MetRS. Mismethionylation occurs in vitro when the yeast MetRS is associated with the other two proteins in the AME complex. This association in the complex, however, is not required for mismethionylation in vivo as an Arc1 deletion yeast strain also shows similar level of mismethionylation. We further show by in vivo utilization kinetics and proteomic mass spec analysis that misacylated tRNAs are used in translation in yeast, as in mammals (4).
Our results lead to two wide-open biological questions. First, how do ribosome choose mismethionylated tRNAs in translation? Misacylated tRNA may or may not be used in translation depending on elongation factor selection and ribosome utilization. In E. coli, misacylated tRNAs can bind to the elongation factor EF-Tu at different affinities compared to correctly charged tRNAs (30,35). This differential binding has been shown to result in the exclusion of some misacylated tRNAs to EF-Tu binding while some other misacylated tRNAs cannot be properly delivered into the A-site of the ribosome (31). How mismethionylated tRNAs are selected by EF-Tu is, however, unclear as the EF-Tu selection of misacylated tRNAs depends on the tRNA and the amino acid identity; mismethionylated tRNAs were not used in previous studies. In the fungal CTG clade species, the tRNA with anticodon CAG is charged with either serine or leucine, and both Ser and Leu charged tRNAs are used in translation (36). This is one example of experimental evidence for the utilization of mischarged tRNAs in eukaryotes, suggesting that EF-1α does not vigorously discriminate misacylated tRNAs. Many studies have been conducted on how ribosome discriminates codon-anticodon mismatched tRNAs in the A and the P sites. Mismethionylated tRNAs, however, can enter the A site while maintaining perfect codon–anticodon matches. For instance, a mismethionylated tRNALysCUU is expected to enter the A site containing the cognate AAG codon. We do not know how mismethionylated tRNA in the A site might perturb peptide bond formation or translocation, due to a lack of previous studies that specifically considers mismethionylated tRNAs.
Although EF-1α may not discriminate misacylated tRNAs, our proteomic analysis here shows that misacylated lysyl- and argininyl-tRNAs are used in translation. To identify the rules of ribosome utilization of mismethionylated tRNAs in the future, it should be possible to conduct proteomic studies to specifically identify Met substitutions in proteins that can be considered to derive from all the mismethionylated tRNA species. Each mismethionylated tRNA species is only present at an average level of ~0.5% of tRNAMets. If we consider that ribosome uses all such tRNAs, it will still represent a sub-1% presence of Met at individual non-Met positions in proteins. Met-containing peptides are also prone to oxidation in mass spectrometry analysis, which pose an additional challenge for proteome-wide identification. Unlike Cys-containing peptides, which can be specifically enriched from the proteomic mixture, no reliable chemical method is yet available to enrich Met-containing peptides.
The second biological question deals with the potential function of mis-translation using mismethionylated tRNAs. We have proposed previously for mammalian cells that low-frequency substitution of non-Met residues with Met in stress response proteins can enhance the known, protective function of genetically encoded Met residues against ROS inactivation of cells’ own proteins. ROS refers to a collection of highly reactive radicals or peroxides, byproducts of the electron transport chain, and as such used as signaling molecules for cell health and stress (37). In mammals, ROS is also used as chemical weapons against invading microbes or undesired molecules. Their high chemical reactivity easily leads to damages of a cell’s own molecules, including proteins. To protect their own proteins against ROS inactivation, certain Met residues in an endogenous protein are positioned at strategic places to react first with diffusing ROS molecules before they can oxidize sensitive amino acid side chains in, e.g. an active site of an enzyme to result in permanent inactivation (5,38). Our previous proposal suggests that substituting certain non-Met residues with Met during translation can enhance this effect, in particular, during oxidative stress response. Our yeast result here is consistent with this idea in that many tRNA encoding charged and polar amino acids are mismethionylated so that a Met substitution at these residues would enable such a protective function but less likely produce misfolded proteins. In order to test this functional proposal for mis-translation using mismethionylated tRNAs, one would need MetRS mutants that have diminished ability to mismethionylate but still maintain a wild-type level activity to charge tRNAMet. Such MetRS mutants are available for the E. coli enzyme (7), and we are actively identifying such mutants using the recombinant AME as the starting point.
To rule out misacylation caused by other enzymes in yeast cells, we purified the recombinant AME complex from E. coli and performed in vitro aminoacylation reactions with purified total yeast tRNA and tRNA transcripts. Mismethionylation was prevalent for many tRNAs and for all tRNAGlu transcripts tested. All the misacylated tRNAs in vitro are also misacylated in vivo, suggesting that the AME complex binds many tRNAs and facilitates interaction of many tRNAs with the MetRS for methionylation. It appears that the Arc1p protein in the AME complex facilitates interactions of MetRS with tRNAMet and a large number of other tRNAs. However, Arc1p is not required for methionylation in vivo.
The E. coli MetRS misacylates only two E. coli tRNAs, tRNAArgCCU and tRNAThrCGU in vitro (7). These two tRNAs have anticodons that differ from the anticodon for methionine, CAU, by one nucleotide. The anticodon and several regions of the E. coli MetRS that interact with the anticodon loop are responsible for misacylation. In addition, several studies have shown that the CAU anticodon plays a key role in yeast tRNA methionylation and discrimination (39). As there are a large number of different misacylated tRNAs in S. cerevisiae, the misacylated tRNAs have little similarity among their anticodons. This raised the question of how so many different tRNAs are misacylated with methionine in S. cerevisiae. We found here that many tRNA species bind to the AME complex, including all the tRNAs that are misacylated. Since Arc1p is a generic tRNA-binding protein, Arc1p can bind many tRNAs and likely transfer them to the MetRS for aminoacylation with methionine, potentially explaining why the AME complex misacylates many tRNA species in vitro.
In summary, tRNA mismethionylation occurs in mammals and as shown here in yeast. As in mammals, a large number of tRNAs can be mismethionylated in yeast, and these misacylated tRNAs are used in translation. The misacylated tRNAs code for charged or polar amino acids, corroborating a role for Met substituting more solvent exposed positions in proteins. This work provides another role of multisynthetase complexes in eukaryotes and expands our knowledge on the mystery of molecular and biological roles of mis-translation.
National Institutes of Health (NIH) [DP1GM105386 to T.P.]. An NIH pre-doctoral training grant in Chemistry and Biology [T32-GM008720 to E.W.]. Funding for open access charge: NIH [DP1GM 105386].
Conflict of interest statement. None declared.
We thank Dr Chuan He for his advisory role on EW and his insightful comments on this project. We also thank Dr Hubert Becker for comments on the manuscript and providing the AME complex plasmid, Dr Allan Drummond for providing the mass spectrometry dataset, Dr Marc Parisien for his computational assistance in the mass spectrometry analysis, the members of Dr Joseph Piccirilli’s lab for use of their scintillation counter and Dr Thomas Jones for useful discussions.