Cross-editing by a tRNA synthetase allows vertebrates to abundantly express mischargeable tRNA without causing mistranslation

Abstract The accuracy in pairing tRNAs with correct amino acids by aminoacyl-tRNA synthetases (aaRSs) dictates the fidelity of translation. To ensure fidelity, multiple aaRSs developed editing functions that remove a wrong amino acid from tRNA before it reaches the ribosome. However, no specific mechanism within an aaRS is known to handle the scenario where a cognate amino acid is mischarged onto a wrong tRNA, as exemplified by AlaRS mischarging alanine to G4:U69-containing tRNAThr. Here, we report that the mischargeable G4:U69-containing tRNAThr are strictly conserved in vertebrates and are ubiquitously and abundantly expressed in mammalian cells and tissues. Although these tRNAs are efficiently mischarged, no corresponding Thr-to-Ala mistranslation is detectable. Mistranslation is prevented by a robust proofreading activity of ThrRS towards Ala-tRNAThr. Therefore, while wrong amino acids are corrected within an aaRS, a wrong tRNA is handled in trans by an aaRS cognate to the mischarged tRNA species. Interestingly, although Ala-tRNAThr mischarging is not known to occur in bacteria, Escherichia coli ThrRS also possesses robust cross-editing ability. We propose that the cross-editing activity of ThrRS is evolutionarily conserved and that this intrinsic activity allows G4:U69-containing tRNAThr to emerge and be preserved in vertebrates to have alternative functions without compromising translational fidelity.


INTRODUCTION
Aminoacyl-tRNA synthetases (aaRSs) establish the rules for genetic code expression by matching each of the 20 pro-teinogenic amino acids to their cognate transfer RNAs (tR-NAs), which harbor anticodon trinucleotides to allow the 'translation' of mRNA into proteins within the ribosome (1). Faithful translation of the genetic information is of central importance in biology (2). Because the accuracy of the aaRSs in pairing tRNAs with their cognate amino acids is greater than that of subsequent steps of ribosomal protein synthesis (3), the fidelity of translation is predominately dictated by aaRSs.
The aaRS-catalyzed tRNA aminoacylation is a two-step reaction: first, the amino acid is activated with ATP to form an enzyme-bound aminoacyl-adenylate; second, the aminoacyl moiety of the adenylate is transferred onto its cognate tRNA to generate the aminoacyl-tRNA product (4). To ensure the accuracy in aminoacylation of tRNAs, elaborate mechanisms of recognition for both the correct amino acid and the cognate tRNA by an aaRS have been evolved. The amino acid binding pocket at the active site of an aaRS plays the major role in identifying the correct amino acid. However, for certain aaRSs, the active site is not sufficient in selecting out the cognate amino acid due to high similarity with some noncognate amino acids in size and/or chemical properties. For example, serine can be misactivated by both AlaRS and ThrRS (5,6). Therefore, an editing domain has been incorporated into each synthetase to selectively hydrolyze the noncognate aminoacyladenylate (pre-transfer editing) or remove the noncognate amino acid from tRNA (post-transfer editing) (7)(8)(9). The importance of editing has been extensively demonstrated, as even mild editing defects will cause severe diseases (10).
As for the cognate tRNA recognition, it often involves the anticodon and the acceptor stem of the tRNA to be specifically identified by the anticodon binding domain and the catalytic domain, respectively, of the corresponding aaRS. Mischarging a cognate amino acid onto a noncognate tRNA is less frequently reported (11)(12)(13)(14)(15). In this sce-nario, because the amino acid is cognate to the synthetase, neither pre-nor post-transfer editing is effective to remove the error. A recent study found that, under stress conditions, MetRS could misacylate methionine onto various noncognate tRNAs. The lack of editing of the mischarged noncognate tRNAs leads to mis-incorporation of methionine into proteins, which could protect cells against oxidative damage (11). Although mistranslation may provide beneficial effects for a short term as in this case, long-lasting mistranslation is likely to be detrimental for cells.
Interestingly, certain aaRSs are prone to mischarging of noncognate tRNAs. For example, AlaRS, which lacks an anticodon binding domain, recognizes its cognate tRNA based on a single G3:U70 base pair in the acceptor stem (16), and thus is prone to potential perturbation in pairing accuracy (14,17). Indeed, using a tRNA microarray system, we detected that human AlaRS can mischarge alanine onto noncognate tRNAs with a G4:U69 base pair, including tRNA Cys and tRNA Thr (14). Although AlaRS can mischarge both tRNA Cys and tRNA Thr , we only detected a cysteine-to-alanine, but not threonine-to-alanine, substitution in a reporter protein expressed in human cells (14), suggesting the existence of a trans-editing mechanism to specifically remove the mischarged alanine from tRNA Thr but not tRNA Cys , among other possible explanations.
In this work, we extensively studied the mischargeable G4:U69-containing tRNA Thr to understand its apparent lack of mistranslation in human cells. We found that the mischargeable tRNA Thr species are ubiquitously and highly expressed among various mammalian cell lines and tissues. Upon rigorous analysis, we again failed to detect the corresponding Thr-to-Ala mistranslation in the human proteome. We identified a robust cross-editing mechanism that removes the mischarged alanine from tRNA Thr . While AlaRS itself is unable to correct this mistake, ThrRS efficiently deacylates the mischarged Ala-tRNA Thr at its editing site. Therefore, while wrong amino acids are corrected within an aaRS, a wrong tRNA is handled in trans by an aaRS cognate to the mischarged tRNA species. AlaRS and ThrRS thus constitute a mischarging-editing cycle which protects the cell from noncognate tRNA charging and its detrimental effects. We outline a process by which organisms can evolve novel translation-independent functions of specialized tRNA species without compromising translational fidelity.

In vitro transcription of tRNA
Human tRNA Thr and tRNA Ala was in vitro transcribed as described previously with modification (18). Template DNA for tRNA transcription was amplified by PCR. After 5-h transcription reaction at 37 • C, the reaction was terminated and the product loaded onto a DEAE column (GE healthcare, USA) equilibrated in low salt buffer containing 50 mM MES pH 6.5, 150 mM NaCl, 0.2 mM EDTA to remove proteins and NTPs. The tRNA was eluted with high salt buffer containing 400 mM NaCl. tRNA fractions were pooled and precipitated by ethanol. The purity of tRNA was checked by urea-PAGE. Prior to use in assays, tRNA was incubated at 80 • C for 5 min followed by refolding at 60 • C in 5 mM MgCl 2 and gradually cooling down to room temperature.

Northern blot
RNA samples, either transcribed tRNAs or total RNA from HEK293 cells and mouse tissues, were separated by 15% urea-PAGE, followed by electrophoretic transfer onto a nylon membrane. The crosslinked membrane was hybridized with DNA probes ([5 -32 P] labeled using T4 PNK) at a specific temperature for 20 h. The washed membrane was incubated with phosphor imaging plate to develop blots. For detection of different RNAs, the membrane was stripped of the previous probe each time before the next one was applied. After stripping, the absence of residual activity from the previous probe was checked on a phosphorimager. A pre-stained marker for small RNA series was added in a separate lane and was cut from the membrane before hybridization. The two membrane fragments were subsequently aligned to indicate the size of RNAs. In vitro transcribed and purified tRNAs were used to precisely indicate the size of the tRNAs purified from cells and tissues.

Expression and purification of recombinant proteins from E. coli
Coding sequences for AlaRS and ThrRS were cloned into plasmid pET-21a, respectively. ThrRS mutants (H155A/H159A, D259A) were constructed by the Quickchange method. The recombinant proteins with his-tag were first purified by affinity chromatography, and then the eluted proteins went through ion exchange column, HiTrap Heparin (GE Healthcare) for ThrRS and its mutants while HiTrap Q for AlaRS. Finally, proteins were loaded onto HiLoad 200 16/60 and eluted proteins were pooled and concentrated. Proteins at each step of purification were monitored by SDS-PAGE.

Protein and tRNA overexpression in mammalian cells
Genes encoding human AlaRS, GAPDH and tRNAs were cloned into the pCDNA6-V5/His-C vector (Life Technologies, Grand Island, NY, USA), respectively. The plasmids were transfected into HEK293 cells using Lipofectamine 3000 (Life Technologies, Grand Island, NY, USA), which were further cultured for 48 h in DMEM supplemented with 10% fetal bovine serum (FBS). The cells were harvested and washed with cold PBS and stored for further use.

Aminoacylation and deacylation assays
The aminoacylation assays were performed as described previously (14). The reactions were incubated with 50 mM HEPES pH 7.  Figure 3D). For alanylation of tRNA Thr (G4:U69) by AlaRS in the presence of ThrRS, the reaction was carried out with 1 M AlaRS and 20 M tRNA Thr (G4:U69), with or without 500 nM ThrRS/ThrRS H155A/H159A ( Figure 3E). At varying time intervals, 5 l aliquots were applied to MultiScreen 96-well filter plate pre-wetted with quench solution containing 0.5 mg/ml DNA and 100 mM EDTA in 300 mM NaOAc (pH 3.0), followed by the same procedures described in previous paper (14). For deacylation assay, 2 M [ 3 H]Ala-tRNA Thr was incubated with 200 nM ThrRS, ThrRS mutants, or ATD in the buffer containing 50 mM HEPES pH 7.5, 20 mM KCl, 5 mM MgCl 2 , 0.2 mg/ml BSA. Reactions were quenched at various time points, and the deacylation rate was calculated based on the reduced [ 3 H]Ala-tRNA Thr signals. All raw data from the aminoacylation and deacylation assays are included as Supplementary Table S5.

Electrophoretic mobility shift assay
Each tRNA (400 nM) was incubated with increasing amount of AlaRS for 30 min at room temperature. The tRNA-protein complex was resolved on a Native PAGE containing 10% acrylamide in Tris-glycine buffer.

Identification of Thr-to-Ala misincorporation by mass spectrometry
For mass spectrometry (MS) analysis of HEK293 cell lysate, the proteins were extracted by methanol/chloroform, which were digested by trypsin and further desalted before loading onto MS. For MS analysis of recombinant GAPDH, GAPDH was isolated and enriched by immunoprecipitation using mouse V5 antibody, which was digested and desalted before MS analysis. For MS analysis of luciferase protein, luciferase gene with an amber stop codon at Thr348 position was inserted into pCDNA6 vector, which was cotransformed into HEK293 with pCDNA6 vector containing tRNA Thr (G4:U69) gene or the chimeric positive control tRNA Thr (G3:U70) with the anticodon changed into that corresponds to amber stop codon. The full-length luciferase protein with the read-through of amber stop codon was enriched by immunoprecipitation using mouse V5 antibody, and applied for MS analysis as above.
The digested samples were analyzed on a Q Exactive HFX mass spectrometer (Thermo). Samples were injected directly onto a 25 cm, 100 m ID column packed with BEH 1.7 m C18 resin (Waters). Samples were separated at a flow rate of 300 nl/min on a nLC 1200 (Thermo). Buffer A and B were 0.1% formic acid in water and 90% acetonitrile, respectively. A gradient of 1-25% B over 180 min, an increase to 40% B over 40 min, an increase to 90% B over another 10 min and held at 90% B for a 10 min was used for a 240 min total run time. Column was re-equilibrated with 15 l of buffer A prior to the injection of sample. Peptides were eluted directly from the tip of the column and nanosprayed directly into the mass spectrometer by application of 2.8 kV voltage at the back of the column. The Q Exactive was operated in a data dependent mode. Full MS scans were collected in the Orbitrap at 120K resolution with a mass range of 400-2000 m/z. The 10 most abundant ions per cycle were selected for MS/MS and dynamic exclusion was used with exclusion duration of 10 sec.
Protein and peptide identification were done with Integrated Proteomics Pipeline -IP2 (Integrated Proteomics Applications). Tandem mass spectra were extracted from raw files using RawConverter (19) and searched with Pro-LuCID (20) against human UniProt database. The database was appended with the sequence of Luciferase where each of the 20 amino acids are possible at the amber stop codon site. The search space included all fully-tryptic and half-tryptic peptide candidates. Carbamidomethylation on cysteine was considered as a static modification. Data was searched with 50 ppm precursor ion tolerance and 600 ppm fragment ion tolerance. Identified proteins were filtered using DTASelect (21) and utilizing a target-decoy database search strategy to control the false discovery rate to 5% at the spectrum level. The spectra covering the amber codon stop site were manually validated (22).

G4:U69-containing mischargeable tRNA Thr are ubiquitously and abundantly detected in mammalian cells and tissues
According to our previous tRNA microarray analysis, both human cytoplasmic tRNA Cys and tRNA Thr can be mischarged with alanine by AlaRS ( Figure 1A), with significantly higher levels of mischarged Ala-tRNA Thr than for Ala-tRNA Cys (14). Analysis of the tRNA genome showed that the number of G4:U69-containing tRNA Thr genes (hereafter denoted as tRNA Thr (G4:U69)) is greater than that of G4:U69-containing tRNA Cys genes (hereafter tRNA Cys (G4:U69)) in human and mouse cells (Supplementary Tables S1-S4). Moreover, in vertebrates including mammals, tRNA Thr (G4:U69) genes, but not tRNA Cys (G4:U69) genes, are absolutely conserved in all genetically annotated organisms ( Figure 1B), suggesting the existence of strong selective pressure to retain the G4:U69 wobble base pair within tRNA Thr genes in vertebrates.
The effect of mischarging by AlaRS on translation would greatly depend on the expression levels of tRNA Thr (G4:U69) genes. To obtain insight into the expression of the six tRNA Thr (G4:U69) genes, we analyzed the Chromatin Immunoprecipitation (ChIP) data of RNA Polymerase III (Pol III) in human cells (23), and human and mouse liver tissues (24), where Pol IIIbinding indicates active transcription of tRNA genes. According to this analysis, tRNA Thr (G4:U69) genes show high Pol III occupancies in all four available cell lines--HEK293 ( Figure 1C), HeLa, Jurkat, and HFF (Supplementary Figure S2A), as well as human and mouse liver tissues (Supplementary Figure S2B). Most of the tRNA Thr (G4:U69) genes showed higher Pol III binding than the average for all tRNA Thr genes ( Figure 1C and Supplementary Figures S2A, B), suggesting relatively high transcription levels of tRNA Thr (G4:U69) genes in these cell lines and liver tissues. Consistently, deep sequencing data for tRNA transcripts (tRNA-Seq) (25) directly show that tRNA Thr (G4:U69) isodecoders, in particular from the tRNA Thr (G4:U69)-AGU isoacceptor family, are the most abundant tRNA Thr species expressed in HEK293T cells ( Figure 1D and Supplementary Figure S2C). By contrast, low Pol III occupancy (Supplementary Figure S3A) and low tRNA transcript levels (Supplementary Figure S3B) were detected for the single tRNA Cys (G4:U69) gene in HEK293 cells. The latter may be related to its less stable secondary clover-leaf structure, as indicated by a low tRNAScan score (Supplementary Table S3).

Mischarging of alanine onto tRNA Thr (G4:U69) does not yield mistranslation
The conservation and abundant expression of the mischargeable tRNA Thr (G4:U69) species pose the question whether it can lead to mistranslation. Although beneficial effects of deliberate modifications of translation fidelity have been documented (26), widespread mistranslation would be detrimental. In order to investigate the tRNA Thr (G4:U69)-mediated mistranslation, we exhaustively explored this possibility in HEK293 cells by mass spectrometry analysis as detailed below.
To generate a positive control, we created the chimeric tRNA Thr (G3:U70) by moving up the GU base pair in a natural tRNA Thr (G4:U69) (i.e. Thr-AGU-7) along the acceptor stem of the tRNA (Supplementary Figure S5) to enhance its capacity to be mischarged by AlaRS ( Figure  2B). To ensure that we can detect potential threonine-toalanine substitution by mass spectrometry analysis, we incorporated a translation readthrough strategy (Supplemen- Fraction of all tRNA All tRNA average   tary Figure S6A). The gene of the reporter protein (luciferase) contains an amber stop codon TAG in the middle and the DNA sequence encoding a V5-tag at the Cterminus. Only when the stop codon is readthrough by tR-NAs with an CUA anticodon, the full-length protein is expressed with a V5-tag for capture. We replaced the anticodon to CUA in the chimeric tRNA Thr (G3:U70) and the natural tRNA Thr (G4:U69) to allow the translation readthrough. The replacement should not affect mischarging by AlaRS because the AlaRS-tRNA recognition does not involve the anticodon (16). Indeed, when the amber tRNA Thr (G3:U70) is expressed as the suppressor tRNA, out of the 21 readthrough peptides we detected, 19 contain an Ala, while only 2 have a Thr, at the amber stop codon site (Supplementary Figure S6B), suggesting that the amber tRNA Thr (G3:U70) is mostly charged with alanine by AlaRS as expected and that threonine-to-alanine substitution in proteins can be successfully detected. In contrast, when the natural mischargeable tRNA Thr (G4:U69) with the engineered anticodon is expressed as the suppressor tRNA, out of the 14 readthrough peptides we detected, all of them had the cognate Thr at the amber stop codon position. Therefore, although the anticodon of tRNA Thr is an identity element, tRNA Thr containing the amber anticodon can still be recognized and aminoacylated by ThrRS, consistent with previous reports (27,28). To further confirm the lack of Thr-to-Ala mistranslation in HEK293 cells, we used another reporter protein (GAPDH) without the readthrough design. In this case, we can directly overexpress the mischargeable tRNA Thr (G4:U69) with its natural anticodon. Intended as a positive control, the chimeric tRNA Thr (G3:U70) was separately overexpressed. However, no Thr-to-Ala substitution was found in the reporter protein with either tRNA overexpressed and when AlaRS co-overexpressed with each tRNA to further enhance the alanine mischarging ( Figure  2C). We also attempted to detect a Thr-to-Ala substitution in the HEK293 whole cell lysate. Again, no mistranslation was detected even with the co-expression of chimeric tRNA Thr (G3:U70) and AlaRS ( Figure 2C). Considering the sensitivity limitations of mass spectrometry, especially when the amount of the target peptide is low, we cannot rule out the possibility that mistranslation may happen below the detectable level. No obvious cellular toxicity was observed upon overexpression of the natural or the engineered tRNA Thr , consistent with the lack of detectable mistranslation.

ThrRS is the main factor for editing Ala-tRNA Thr to prevent mistranslation
The lack of detection of threonine-to-alanine substitutions in the reporter proteins as well as in the proteome of HEK293 cells suggests the existence of an editing activity that removes the mischarged alanine from tRNA Thr . Recently, a trans-editing factor ATD (Animaliaspecific tRNA deacylase), also known as DTD2 (Daminoacyl tRNA deacylase-2), was reported to deacylate Ala-tRNA Thr (G4:U69) (29). However, based on the HUMAN PROTEOME MAP database (http://www. humanproteomemap.org/) and the Human Protein Atlas (https://www.proteinatlas.org/), expression of ATD is barely detected in many tissues and cell types ( Supplementary Figure S7A). Given the ubiquitous and abundant expression of tRNA Thr (G4:U69) ( Figure 1E and F), it is therefore unclear whether ATD is sufficient or at all available for correcting Ala-tRNA Thr in all tissues.
An alternative candidate for hydrolyzing the mischarged Ala-tRNA Thr is ThrRS. ThrRS possesses an editing domain with known cis-acting post-transfer editing activity to remove noncognate amino acid from tRNA Thr after it is being mischarged in the synthetic site (6). Importantly, ThrRS is ubiquitously and highly expressed in all cell and tissue types (Supplementary Figure S7A). We detected and quantified ATD and ThrRS levels in various cells including HEK293 cells. Indeed, the expression level of ThrRS is substantially higher than that of ATD in all cells ( Supplementary Figures S7B and C). Particularly, in HEK293 cells, the concentration of ThrRS is about 5-fold higher than that of ATD ( Supplementary Figures S7B and S7C).
Next, we investigated the potential deacylation activity of ThrRS against Ala-tRNA Thr and compared it with that of ATD. Ala-tRNA Thr (G4:U69) can be rapidly hydrolyzed by ThrRS (Figure 3A), and the editing activity of ThrRS is more efficient than that of ATD (Supplementary Figure S8). Combining the activity and the expression analyses, ThrRS is more likely to be the main factor in vivo that edits Ala-tRNA Thr to prevent mistranslation. Moreover, as expected, although AlaRS also have an editing domain, it cannot edit Ala-tRNA Thr ( Figure 3B).

The editing domain active site of ThrRS is responsible for hydrolyzing Ala-tRNA Thr
Previous studies of the post-transfer editing activity of ThrRS focused on deacylating the mischarged Ser-tRNA Thr (6). It was shown that, while the cognate threonine moiety is sterically excluded from the editing site, the absence of the methyl group in the side chain of serine allows it to fit into the editing domain active site pocket ( Figure 3C) (30). This led us to assume that alanine, with an even smaller side chain, would fit into the editing pocket of ThrRS as well ( Figure 3C). To confirm that hydrolysis of Ala-tRNA Thr occurs in the editing site of ThrRS, we introduced editing site mutations, including a double mutant H155A/H159A and a single mutant D259A, (Figure 3C), each has been shown to impair editing of Ser-tRNA Thr by ThrRS (6). Indeed, these mutations completely abolished the editing of Ala-tRNA Thr ( Figure 3C).

ThrRS cannot mischarge alanine onto tRNA Thr but can cross-edit the mischarged Ala-tRNA Thr
The erroneous capacity of ThrRS to activate serine and to generate Ser-tRNA Thr necessitates the evolutionary conservation of its post-transfer editing activity (6,31). Our observation that ThrRS also edits Ala-tRNA Thr raises the question whether ThrRS itself is able to mischarge alanine onto tRNA Thr . To address this point, firstly, we show that no Ala-tRNA Thr is formed with ThrRS ( Figure 2A). However, the lack of mischarging could result from the above demonstrated editing activity of ThrRS. Indeed, no Ser-tRNA Thr can be detected with ThrRS either ( Figure 3D). Yet, when the editing activity of ThrRS is abolished by the H155A/H159A mutation, only Ser-tRNA Thr , but not Ala-tRNA Thr , can be formed ( Figure 3D), demonstrating that human ThrRS cannot mischarge alanine onto tRNA Thr independent of its editing activity. Therefore, ThrRS only possesses cross-editing activity for Ala-tRNA Thr .
To further investigate the mischarging-editing-cycle of tRNA Thr (G4:U69) catalyzed by two different aaRSs, we monitored the mischarging of tRNA Thr (G4:U69) by AlaRS in the presence of ThrRS. In these experiments, although AlaRS was in excess of ThrRS (2-fold), the presence of ThrRS markedly reduced the accumulation of misacylated Ala-tRNA Thr ( Figure 3E). Importantly, this effect of ThrRS was abolished by the H155A/H159A mutation in the editing site. Moreover, the addition of ThrRS H155A/H159A slightly enhanced the mischarging of tRNA Thr (Figure 3E), presumably by competing with AlaRS for binding to Ala-tRNA Thr , thereby facilitating the turnover and the release of the mischarged tRNA from AlaRS.
Nucleic Acids Research, 2020, Vol. 48, No. 12 6453 Prokaryotic ThrRS also possesses cross-editing ability to deacylate Ala-tRNA Thr Our previous study found that eukaryotic, but not prokaryotic, AlaRS can mischarge tRNA with a G4:U69 base pair (14). The mischarging capacity is determined by several key residues in the tRNA binding domain of AlaRS, which are divergent between eukaryotes and prokaryotes (14). Interestingly, although the mischarged Ala-tRNA Thr is unlikely to exist in prokaryotes, E. coli ThrRS, like human ThrRS, can cross-edit the mischarged Ala-tRNA Thr (G4:U69) with an efficiency similar to that of the human enzyme ( Figure  3F). Therefore, ThrRS seems to have an inherent capacity to edit Ala-tRNA Thr , regardless of the existence of the mischarged species within the biological system.

DISCUSSION
Aminoacylation is a strictly controlled process with multiple proofreading mechanisms ensuring high accuracy. While numerous mechanisms have been described for how aaRSs prevent misacylation of noncognate amino acids onto their cognate tRNA, no editing mechanisms have been described so far for cases in which an aaRS mischarges its cognate amino acid onto a noncognate tRNA. In fact, MetRS could mischarge methionine onto various noncognate tRNAs, which would then, due to the absence of any proofreading mechanism to clear the mistake, be used in translation (11,12). Using a microarray assay, we previously found that alanylation of tRNA Thr accounts for 58% of all misalanylated tRNA. In contrast, tRNA Cys accounts for only 5%, and yet misincorporation of alanine in lieu of cysteine could be detected by mass spectrometry. This suggests that the levels of Ala-tRNA Thr should be high enough to cause mistranslation if it was not cleared by editing factors (14). However, we show here that mischarging of tRNA Thr by AlaRS unlikely results in mistranslation presumably due to the novel cross-editing activity of ThrRS. Although other factors such as ATD may contribute to removing mischarged alanine from tRNA Thr , the higher in vitro activity and its ubiquitous expression as a housekeeping protein suggest that ThrRS is the main factor responsible for editing Ala-tRNA Thr in vivo.
It would be ideal if we can demonstrate the significance of the cross-editing activity of ThrRS in vivo. In theory this may be achieved by creating cell lines that are defective in ThrRS editing to reveal the otherwise corrected alanine mischarging of tRNA Thr and Thr-to-Ala mistranslation. However, we have shown that the same editing site used to correct the mischarged alanine-tRNA Thr in trans is also used to correct commonly mischarged serine-tRNA Thr in cis, indicating that the editing activity of ThrRS is likely to be indispensable for cell viability. Indeed, it has been demonstrated that a similar editing activity from AlaRS is essential. Mouse homozygous in expressing a severe editingdeficient AlaRS (AlaRS-C723A) died at early stage of embryonic development (32). It is worth noting that a ThrRSlike protein (TARSL2) has been identified in higher eukaryotes and was demonstrated to have both tRNA aminoacylation and editing activities (33). The editing domain of TARSL2 is highly homologues to that of the canonical ThrRS, including strict conservations of the key residues for editing, indicating TARSL2 would also be capable to transedit the mischarged alanine-tRNA Thr . The existence of both ThrRS and a ThrRS-like protein would further explain why vertebrates can abundantly express mischargeable tRNA Thr without causing mistranslation.
Editing of a cognate amino acid from a noncognate tRNA poses a unique challenge to synthetases. Given that the correct amino acid is being used, neither classical prenor post-transfer editing mechanisms would recognize the mistake. In addition, most editing mechanisms employed by aaRSs also rely on specific identity elements in the cognate tRNA (34), exacerbating the difficulties for cis-acting editing sites to clear a cognate amino acid from a noncognate tRNA. This concept is demonstrated here by the fact that human AlaRS can produce Ala-tRNA Thr (Figure 2A), due to its relaxed specificity toward the relocated G:U identity determinant, but is unable to correct the mistake in its own editing site ( Figure 3B). Instead, the mischarged Ala-tRNA Thr is edited by ThrRS, even though ThrRS itself does not create this mischarged tRNA. Thus, the function of ThrRS in this context is independent from its synthetic role but relies on its hydrolytic activity as a cross-editing factor for an error introduced by another synthetase ( Figure  4). Although several free-standing trans-editing enzymes (e.g. AlaX, ProX, YbaK, DTD, ATD) have been described to clean up mischarged tRNAs that escaped from aaRSs (29,(35)(36)(37)(38)(39)(40), these free-standing factors often do not have clear tRNA specificity or have to rely on association with an aaRS (e.g. YbaK) to obtain tRNA specificity (35,41). In contrast, the cross-editing activity of ThrRS possesses an intrinsic specificity toward its cognate tRNA. Historically, E. coli PheRS was shown to be able to cross-edit Ile-tRNA Phe mischarged by IleRS (42). However, the mischarging activity of IleRS toward tRNA Phe appears to be extremely weak under normal conditions (15), thus the mischarging and the editing may not actually happen in cells. Nevertheless, the potential cross-editing activity of PheRS may serve as another example of how a wrong tRNA is handled in trans by an aaRS cognate to the mischarged tRNA species to provide intrinsic specificity.
We show here that G4:U69-containing tRNA Thr are strictly conserved in vertebrates ( Figure 1B), and that these mischargeable tRNA Thr (G4:U69) form the most prevalent and highly expressed isodecoders of tRNA Thr (Figure 1C-F). However, no Thr-to-Ala mistranslation was detected in mammalian cells, presumably due to the ability of ThrRS to clear mischarged Ala-tRNA Thr before its delivery to the translating ribosome. Interestingly, no mistranslation was detected even with the expression of the engineered tRNA Thr (G3:U70) with the G:U base pair relocated to enhance its mischarging by AlaRS (Supplementary Figure S5A). We have confirmed that the mischarged Ala-tRNA Thr (G3:U70) can still be efficiently deacylated by ThrRS (Supplementary Figure S5B), thus explaining this lack of mistranslation. In contrast, we successfully detected mistranslation with the amber tRNA Thr (G3:U70) in the luciferase reporter. In this case, the tRNA Thr (G3:U70) was further engineered in the anticodon in order to decode amber stop codon for translation readthrough (Supplementary Figure S6A). Presumably, the engineered anticodon would affect the cognate tRNA recognition by ThrRS, thus re- In vertebrates, not only can tRNA Thr (G4:U69) be correctly charged with Thr by ThrRS, it can also be mischarged with Ala by AlaRS to some extent, while AlaRS with editing domain cannot hydrolyze the mischarged Ala-tRNA Thr . However, the mistake can be corrected by ThrRS through a crossediting mechanism, whereas ThrRS itself is unable to mischarge Ala onto tRNA Thr . The cross-editing function of ThrRS can prevent deleterious effects caused by mistranslation in organisms, and thus allow the conservation of mischargeable tRNA Thr , probably for alternative functions beyond translation, such as stress sensing or other regulatory functions through tRNA fragmentations.
ducing its editing efficiency and allowing mistranslation to occur (Supplementary Figure S6B). However, this artificial scenario is not present in natural cells.
It is interesting to note that the ability of ThrRS to clear mischarged Ala-tRNA Thr appears to be evolutionarily conserved. Although the mischarged Ala-tRNA Thr is unlikely to exist in prokaryotes, we show that E. coli ThrRS is able to deacylate human Ala-tRNA Thr with similar efficiency as human ThrRS ( Figure 3F). Thus, by co-opting an intrinsic activity of ThrRS, vertebrates were free to relax the specificity constraints on AlaRS and allow tRNA Thr to access an extended sequence space, without compromising translation accuracy. The evolution of the exquisite functional network between AlaRS, ThrRS, and tRNA Thr (G4:U69) may reflect a general principle by which higher eukaryotes balance the need to preserve translational fidelity, while at the same time allowing functional expansion to support increasing complexity (Figure 4).
Considering the evolutionary conservation and the high expression level of tRNA Thr (G4:U69), we speculate that tRNA Thr (G4:U69) may have a biological function not directly related to translation. In recent years, it has become increasingly evident that there are many alternative functions of tRNA beyond translation (43,44), such as amino acid addition through tRNA for antibiotic biosynthesis and cell envelope remodeling (45)(46)(47), uncharged tR-NAs regulating gene expression in response to the dynamics in amino acid availability (48)(49)(50), and formation of tRNA fragments for translation regulation and gene silencing (51)(52)(53)(54). G4:U69 serves no obvious beneficial purpose in the canonical function of tRNA Thr . Yet, once established tRNA Thr (G4:U69) was fixed in the vertebrate lineage, suggesting strong selective pressure to retain this sequence element.
One possible function of the mischargeable tRNA Thr (G4:U69) may be related to stress sensing and response. It has been reported that oxidative stress induces tRNA mischarging by Salmonella PheRS to enhance translational quality control by increasing the rate of editing (55). Interestingly, a high-throughput sequencing method to determine the level of charged tRNA revealed that while most cytosolic tRNAs are charged at levels over 80% in HEK293T cells, tRNA Ser and tRNA Thr are constitutively charged at lower levels (56). The reason and purpose for lower charging levels of these two tRNA species were not clear, however, uncharged tRNA is a well-known activator of the GCN2 pathway in response to amino acid starvation and other stresses (57). The AlaRS/ThrRS-mediated mischarging and deacylation cycle described here may contribute to the relatively low charging levels of tRNA Thr and may serve as a sensitized system for stress sensing (Figure 4).
Additionally or alternatively, tRNA Thr (G4:U69) may serve regulatory functions through its fragmentation (Figure 4). Increasing evidence suggests that tRNA halves or tRNA-derived fragments (tRFs) could regulate translation and gene expression. From our northern blot analysis, we could clearly detect the differential occurrence of tRFs (or tRNA halves) derived from tRNA Thr (G4:U69) in tissues ( Figure 1E). Especially in small intestine, lung, and pancreas, the level of tRNA Thr (G4:U69)-CGU fragments is comparable to the full length tRNA Thr (G4:U69)-CGU. In-Nucleic Acids Research, 2020, Vol. 48 The fragments (5and 3 -halves) are listed according to their predicted minimum free energy from low to high, putting the most stable tRNA fragments on the top, which are from G4:U69-containing tRNA Thr . The secondary structure of potential tRNA Thr fragments and their minimum free energy are predicted by RNAfold server. The secondary structure of representative fragments is shown aside. The sizes of tRNA 5 -and 3 -halves are approximated based on the size ranges of tRNA fragments reported so far.
terestingly, codon usage corresponding to tRNA Thr -CGU is much lower than other threonine codons ( Figure 5A). Based on the pairing rules, tRNA Thr -CGU can only decode ACG codon, which has a low usage, while other isoacceptors with AGU and UGU anticodons can be used for decoding all or almost all four codons and can compensate for the absence of tRNA Thr -CGU ( Figure 5A). Therefore, the necessity of tRNA Thr -CGU for translation is relatively low, suggesting tRNA Thr -CGU, especially G4:U69-containing tRNA Thr -CGU (see below), may be required for alternative functions beyond translation. It was reported that a stable stem-loop secondary structure is crucial for tRNA halves to avoid degradation by RNases (58). Prediction of the secondary structure of tRNA Thr -derived tRNA halves by the RNAfold server shows that both 5 -and 3 -halves of tRNA Thr (G4:U69)-AGU and the 5 half of tRNA Thr (G4:U69)-CGU form stable stem-loop structures ( Figure 5B). In fact, tRNA Thr (G4:U69)-AGU halves rank on the top in stability among all tRNA Thr halves. Interestingly, although there is only one other nucleotide difference between the 5 half of tRNA Thr (G4:U69)-AGU4 and that of tRNA Thr (U4:G69)-AGU6 (in addition to the G4/U4 position), the former exhibits a much more stable secondary structure than the latter ( Figure 5B). Moreover, prediction by RNAstructure bifold showed a high probability for homo-or heterodimer formation by tRNA Thr (G4:U69)-AGU 5 halves (Supplementary Figure S9A), which can further protect the tRNA fragments from degradation (58). In addition, based on the database of tRNA fragments (Mintbase) in human tissues, fragments of tRNA Thr (G4:U69), especially tRNA Thr (G4:U69)-CGU, are the most abundant populations among all sequences derived from tRNA Thr (Supplementary Figure S9B). Therefore, the G4:U69 base pair may play a role in stabilizing tRNA Thr fragments.
Finally, mischarging tRNA Thr with alanine may in itself serve as a signal for production of tRFs ( Figure 4). Honda et al. reported a novel type of tRNA-derived small RNA, named hormone-dependent tRNA-derived RNAs (SHOT-RNAs), which enhance cell proliferation in breast and prostate cancers. Unlike other tRFs, SHOT-RNAs are exclusively produced from charged tRNAs, where the 3 amino acid may play a role in tRNA selectivity by ANG (59). In light of our finding, it is possible that mischarging of tRNA Thr (G4:U69) with Ala may be important for fragmentation of tRNA Thr (G4:U69), as mischarged tRNA Thr may have a lower binding affinity to EF1A and therefore is more accessible to be captured and cleaved by RNases.
In summary, the properties of tRNA Thr (G4:U69) discussed above suggest that it serves a function that would explain its vertebrate-specific conservation. Further studies, such as vertebrate-related phenotype characterization under manipulation of tRNA Thr (G4:U69), are required to elucidate the roles of these specific tRNAs. Importantly, the fact that these specific, mischargeable tRNAs can exist in evolution is likely due to the efficient and intrinsic crossediting function of ThrRS demonstrated in this study.