Ancestral AlaX Editing Enzymes for Control of Genetic Code Fidelity Are Not tRNA-specific*

Background: Both AlaRS editing domains and their free-standing editing forms, AlaX, are thought to exclusively deacylate mischarged tRNAAla. Results: Contrary to this belief, we find that AlaX-S deacylates Ser-tRNAAla and Ser-tRNAThr. Conclusion: AlaX-S is likely to be the ancestor of AlaX, AlaRS editing, and bacterial-like ThrRS editing domains. Significance: The biochemical activities of AlaX-S link, for the first time, the activities of the AlaRS and ThrRS editing domains. Accurate protein synthesis requires the hydrolytic editing of tRNAs incorrectly aminoacylated by aminoacyl-tRNA synthetases (ARSs). Recognition of cognate tRNAs by ARS is less error-prone than amino acid recognition, and, consequently, editing domains are generally believed to act only on the tRNAs cognate to their related ARSs. For example, the AlaX family of editing domains, including the editing domain of alanyl-tRNA synthetase and the related free-standing trans-editing AlaX enzymes, are thought to specifically act on tRNAAla, whereas the editing domains of threonyl-tRNA synthetases are specific for tRNAThr. Here we show that, contrary to this belief, AlaX-S, the smallest of the extant AlaX enzymes, deacylates Ser-tRNAThr in addition to Ser-tRNAAla and that a single residue is important to determine this behavior. Our data indicate that promiscuous forms of AlaX are ancestral to tRNA-specific AlaXs. We propose that former AlaX domains were used to maintain translational fidelity in earlier stages of genetic code evolution when mis-serylation of several tRNAs was possible.

Mistranslation is a biological phenomenon caused by the misacylation of tRNAs by aminoacyl-tRNA synthetases (ARSs). 3 Bona fide translation errors are often deleterious, but mistranslation can also be an adaptive strategy to correct other kinds of insults (1)(2)(3). The reported cases of toxic mistranslation are invariably due to errors in amino acid recognition. In contrast, adaptive mistranslation is mostly due to tRNAs that are charged promiscuously or by ARSs that acylate different tRNAs (4 -8).
Toxic mistranslation is prevented by cells through proofreading mechanisms (9). Error-prone ARSs are toxic to both bacterial and mammalian cells, and, to avoid mischarging, they have evolved editing domains to recognize mischarged tRNAs (10 -15). In some instances, editing domains display tRNA selectivity, and, in other cases, the domains per se are not tRNAspecific and rely on the specificity of other domains for their activity (16 -19). For example, YbaK is not tRNA-specific in vitro and will deacylate any tRNA charged with Cys (20,21). In vivo, YbaK tRNA specificity requires interactions with ProRS (20). In contrast, AlaXs and ProXp-ala (a YbaK homolog) are tRNA-specific and recognize the same conserved tRNA acceptor stem identity elements (16,22,23).
The deacylation activities of the AlaXs vary. Although AlaX-M, AlaX-L, and AlaRS hydrolyze both Ser-tRNA Ala and Gly-tRNA Ala (22,25,27,32), AlaX-S only hydrolyzes mischarged Ser-tRNA Ala (29), possibly because of the lack of the Gly-rich domain ( Fig. 2A), which results in an enlarged amino acid binding pocket and a much lower binding affinity toward Gly (29). In contrast, bacterial ThrRS editing domains hydrolyze Ser-tRNA Thr (26,34). AlaX freestanding domains are generally thought to be specific for tRNA Ala , and their specificity has been linked to a highly conserved arginine residue (22) that is believed to interact with the unique G3:U70 pair in the tRNA Ala acceptor stem. A single R693K mutation in the editing domain of Escherichia coli AlaRS allows this enzyme to hydrolyze Ser-tRNA Thr (22).
We observed that, in archaeal AlaX-S, this arginine is substituted by glutamic acid (Fig. 2B and 2C). Intrigued by the lack of conservation of such a critical residue, we investigated the tRNA specificity of AlaX-S. We found that, although AlaX-S retains specificity for serylated tRNAs, it is not tRNA-specific and deacylates Ser mischarged onto tRNA Ala , tRNA Thr , and tRNA Pro . These features set AlaX-S apart from the rest of the AlaX family of enzymes. A phylogenetic analysis of AlaX sequences shows that AlaX-S constitutes an ancient clade that separates from other AlaXs and that is more closely related to bacterial ThrRS editing domains than any other AlaX member. Therefore, on the basis of our biochemical and phylogenetic analyses, we propose a novel evolutionary model for the AlaX family in which the ancestral form of the AlaX family of editing enzymes likely was a tRNApromiscuous enzyme whose function was to correct widespread mis-serylation of tRNAs by ancient primitive ARSs. This ancestral form would afterward have evolved to deacylate either tRNA Ala (AlaX-M, AlaX-L, AlaRS editing domain) or tRNA Thr (ThrRS editing domain), fulfilling the need for higher levels of discrimination before being incorporated as the editing domains now characteristic of contemporary synthetases.

EXPERIMENTAL PROCEDURES
Phylogenetic Analysis of AlaX Domains-Hidden Markov Model profiles were built for each protein containing an AlaX domain (AlaX-S, AlaX-M, AlaX-L, and AlaRS editing domain and ThrRS editing domain) using the HMMER package (35), which were used as query to find homologues in the Uniprot database. The resulting sequences were aligned using T-Coffee (36) with the structure-guided alignment mode, and phylogeny was built using the maximum likelihood method with bootstrapping using PHYLIP 3.68 (37). The final tree was visualized with TreeView (38). To build figures that illustrate the structure-guided multiple-sequence alignment (MSA) results, a representative subset of species was selected for each AlaX clade from the full multiple-sequence alignment. Structure-guided multiple-sequence alignment results were visualized with Seaview (39).
Protein Expression and Purification-Pyrococcus horikoshii AlaX-S and AlaRS were a gift from Prof. Sokabe (University of California, Davis). The Pyrococcus abyssi ThrRS N-terminal editing domain (PaThrRS-NTD) was a gift from Prof. Sankaranarayanan (Centre for Cellular and Molecular Biology, India). Expression was induced in E. coli Rosetta TM BL21 (Novagen) at A 600 ϭ 0.6 with 1 mM isopropyl ␤-D-thiogalactoside for 6 h at 37°C. Enzymes were prepared as described previously (29,40).
In Vitro tRNA Transcription and Purification-Sequences for P. horikoshii tRNA Ala and tRNA Thr were taken from gtRNAdb (41) (Fig. 3). In vitro-transcribed P. horikoshii tRNA Ala and tRNA Thr were prepared as described previously (42). E. coli tRNA Pro was prepared following protocols published previously (43). Briefly, each tRNA was transcribed from an NspI-digested pUC19 plasmid template using T7 RNA polymerase. The sample was then loaded into a denaturating 10% polyacrylamide gel, and the resulting tRNA band was cut out, electroeluted, and ethanol-precipitated.

Ancestral AlaX Enzymes Have Broad tRNA Specificity
Preparation of Aminoacyl-tRNA Substrates-In vitro-transcribed E. coli tRNA Pro , P. horikoshii tRNA Ala and tRNA Thr were 3Ј 32 P-labeled using E. coli tRNA nucleotidyltransferase and [␣-32 P]ATP (PerkinElmer Life Sciences) as described previously (44). Aminoacylation of tRNAs was carried out using biotinylated dFx (Thermo Scientific) using published conditions with minor modifications (45). Briefly, tRNA and dFx (42 M each, 18-l volume) and trace amounts of 3Ј 32 P-labeled tRNA were heated to 95°C for 1 min, followed by addition of 25 mM MgCl 2 and 5 mM of serine-3,5-dinitrobenzyl ester. dFxcatalyzed aminoacylation was carried out for 6 h on ice. Reactions were quenched with 120 l of 0.3 M NaOAc (pH 5). Biotinylated dFx was removed by incubating the reaction mixture with 150 l of streptavidin-agarose resin (Novagen) for ϳ15 min at room temperature, followed by centrifugation in a tabletop centrifuge at 4°C for 2 min at 2000 ϫ g. The supernatant containing the aminoacylated tRNA was removed, phenol-chloroform-extracted, and ethanol-precipitated. The pellet was dissolved in diethylpyrocarbonate-treated water and stored at Ϫ80°C.

RESULTS
Sequence and Structural Analysis of the AlaX Domain Superfamily-Previous work showed that the G3:U70 base pair is critical for aminoacylation of tRNA Ala by AlaRS (46,47). This base pair interacts with a conserved Arg in the AlaRS editing domain (Arg-693 in E. coli), and the substitution of this residue with Lys results in loss of tRNA specificity (22). Although this residue was initially predicted to be universally conserved, our analysis here shows that AlaX-S enzymes contain a Glu residue in the equivalent position (Glu-145 in P. horikoshii AlaX-S) (Fig. 2B). Structural superimposition of E. coli AlaRS, P. horikoshii AlaX-M, and P. horikoshii AlaX-S crystal structures demonstrates that Glu-145 is in a similar orientation as Arg-693 and Arg-210 of E. coli AlaRS and P. horikoshii AlaX-M, respectively (Fig. 2C). On the basis of these observations, we hypothesized that the presence of Glu-145 in AlaX-S alters its substrate specificity, allowing hydrolysis of Ser from different tRNAs.
AlaX-S Is a Nonspecific Ser-tRNA Deacylase-We next examined the tRNA specificity of AlaX-S in vitro. We purified recombinant P. horikoshii AlaX-S, P. horikoshii AlaRS-N752 (PhAlaRS lacking the C-terminal C-Ala domain), and the isolated N-terminal editing domain of P. abyssi ThrRS (ThrRS-NTD). We also prepared P. horikoshii Ser-tRNA Ala and Ser-tRNA Thr and E. coli Ser-tRNA Pro using the Flexizyme (dFx) system (48). It should be noted that the P. abyssi ThrRS editing domain is not related to either the bacterial/eukaryotic ThrRS editing domains or the AlaX family (49 -51).
As expected, deacylation assays confirmed robust editing activity of Ser-tRNA Ala by AlaX-S and PhAlaRS. In contrast, ThrRS-NTD does not possess significant deacylation activity for this substrate (Fig. 4A) but efficiently hydrolyzes Ser-tRNA Thr (50). PhAlaRS exhibited only weak deacylation activity for Ser-tRNA Thr , whereas AlaX-S robustly edited this substrate, in support of our hypothesis (Fig. 4B). The lack of tRNA specificity of AlaX-S was further confirmed by showing that Ser-tRNA Pro , which is not known to occur in nature, is also a substrate for the enzyme (Fig. 4C).
We then investigated whether the lack of tRNA specificity observed in AlaX-S is also accompanied by a lack of amino acid specificity. Gly-tRNA Ala had already been shown to not be a substrate of AlaX-S (29). We also tested the deacylation activity of AlaX-S upon Pro-tRNA Pro , finding that AlaX-S is not capable of deacylating this substrate either (data not shown).
Taken together, these results support the idea that AlaX-S retains characteristics of an ancestral freestanding editing protein, able to deacylate multiple mischarged Ser-tRNAs. In contrast, the editing domain of archaeal P. abyssi ThrRS is capable of discriminating Ser-tRNAs and appears to be highly specific for Ser-tRNA Thr . Phylogenetic Analysis of the AlaX Superfamily-Several phylogenetic trees that include bacterial ThrRS, AlaRS, and freestanding AlaXs have been reported (10,27). These studies did not distinguish AlaX-S from AlaX-M, which were previously classified together as AlaXp-I (27). For this reason, we performed a new phylogenetic analysis to try to resolve all three types of freestanding AlaX sequences (Fig. 5). We found that, although AlaX sequences are represented in the three domains of life, AlaX-S sequences are only found in Archaea. Previous studies have included several bacterial and eukaryal sequences within the set of AlaX-S sequences (27). However, we believe that these sequences may reflect misannotations because of the shorter sequence length of some bacterial and eukaryal AlaX-M proteins. To confirm the archaeal distribution of AlaX-S, we verified that shorter AlaX sequences from bacteria and eukarya had an Arg residue in the homologous position to Arg-693 of E. coli AlaRS and were evolutionary closer to AlaX-M sequences, indicating that AlaX-S members are only found in Archaea.
Our phylogenetic analysis shows that the editing domain of ThrRS is closest to AlaX-S, which positions AlaX-S as an intermediate evolutionary step between ThrRS and the group of AlaRS, AlaX-M, and AlaX-L, suggesting an early separation that split the original enzyme into two different specificities, one for tRNA Thr and the other for tRNA Ala . Therefore, our results suggest that all forms of AlaX evolved from a common ancestor and shows that both AlaX-M and AlaX-L are derived from AlaX-S.
Surprisingly, a structural superimposition of all the homologous AlaX-type editing domains shows that an Arg residue is also found in bacterial ThrRS editing domains at the equivalent position to E. coli AlaRS Arg-693 (Fig. 6). This implies that the Arg residue by itself does not confer tRNA Ala specificity, given that bacterial ThrRS has been shown to specifically deacylate Ser-tRNA Thr (26). Additional sequence and structural elements are likely to contribute to specific G3:U70 recognition (52).

DISCUSSION
Editing domains are universal features of some ARSs (10,26,53). Originally, however, their function might have been unrelated to protein synthesis quality control and linked to other roles in the RNA world (54). Structural data indicate that the incorporation of editing domains to the ARS ancient catalytic core happened first in class I enzymes (55). However, the existence of free forms of these domains raises the question of whether they originated as freestanding hydrolytic enzymes or as ARS-attached domains.
It is generally accepted that early genetic codes were prone to errors that were gradually minimized as organisms required more faithful genetic translation (56). Here we studied the tRNA specificity of a putative ancestral editing domain, AlaX-S. Our goal was to gain a better understanding of the original function of AlaX-type editing enzymes and the origin of tRNA specificity within AlaX-type editing enzymes. Our data show that AlaX-S is capable of deacylating Ser from multiple tRNA acceptors (Fig. 4), whereas AlaRS, other AlaXs, and the archaeal ThrRS editing domain (ThrRS-NTD) are tRNA-specific.
Our phylogenetic analysis shows that AlaX-S forms a distinct clade that separates AlaXs and AlaRS from ThrRS sequences (Fig. 5). One possibility is that an editing domain with relaxed tRNA specificity (AlaX-S) would have evolved in modern archaeal cells. However, this possibility would not explain the phylogenetic position of AlaX-S between ThrRS and other AlaX-type domains. The correct clustering of the different editing enzymes on the basis of its biochemical activities, and the robust bootstrap support, reinforce the possibility that AlaX-S is the closest relative of the ancestor of all AlaX-type editing domains. This ancestral nature of AlaX-S is further supported by the distribution of the Gly-rich motif responsible for the recognition of Gly in addition to Ser- (29), which is found in AlaX-M, AlaX-L, and AlaRS editing domain but not in the AlaX-S or ThrRS editing domains ( Fig. 2A). Therefore, AlaX-S is likely an evolutionary intermediate between the AlaRS/AlaX-M/AlaX-L editing domains and bacterial ThrRS editing domains.
On the basis of both our biochemical and phylogenetic results, we propose that AlaX-S is a direct descendant of an ancestral deacylation enzyme that lacked tRNA specificity and is the functional ancestor of the AlaRS and ThrRS editing domains, including their freestanding forms (Fig. 7). This ancestral enzyme might have been in charge of deacylating misserylated tRNAs generated by synthetases with poor amino acid selectivity or by seryl-tRNA synthetases with poor tRNA specificity. This activity would be required in an earlier stage of genetic code development where mis-serylation of tRNAs was particularly frequent. Future work reconstructing ancestral sequences and testing their tRNA specificities will provide additional evidence on the evolution of this family.
The precise order by which amino acids entered the genetic code remains unknown (55,57,58), but it is generally accepted that similar amino acids were added through the reassignment of similar codons. The extant codons for Ala, Thr, Pro, and Ser occupy a common space in the genetic code and, possibly, initially coded for a single amino acid. Their diversification during the growth of the code might have generated a fidelity problem that was solved by AlaX-S, which can deacylate serylated tRNA Ala , tRNA Thr , and tRNA Pro . By extension, it is plausible that other editing proteins or domains might have retained their ancestral tRNA specificities and, therefore, may provide additional information on the evolution and growth of the genetic code.