Natural Homolog of tRNA Synthetase Editing Domain Rescues Conditional Lethality Caused by Mistranslation*

AlaXp is a widely distributed (from bacteria to humans) genome-encoded homolog of the editing domain of alanyl-tRNA synthetases. Editing repairs the confusion of serine and glycine for alanine through clearance of mischarged (with Ser or Gly) tRNAAla. Because genome-encoded fragments of editing domains of other synthetases are scarce, the AlaXp redundancy of the editing domain of alanyl-tRNA synthetase is thought to reflect an unusual sensitivity of cells to mistranslation at codons for Ala. Indeed, a small defect in the editing activity of alanyl-tRNA synthetase is causally linked to neurodegeneration in the mouse. Although limited earlier studies demonstrated that AlaXp deacylated mischarged tRNAAla in vitro, the significance of this activity in vivo has not been clear. Here we describe a bacterial system specifically designed to investigate activity of AlaXp in vivo. Serine toxicity, experienced by a strain harboring an editing-defective alanyl-tRNA synthetase, was rescued by an AlaXp-encoding transgene. Rescue was dependent on amino acid residues in AlaXp that are needed for its in vitro catalytic activity. Thus, the editing activity per se of AlaXp was essential for suppressing mistranslation. The results support the idea that the unique widespread distribution of AlaXp arises from the singular difficulties, for translation, poised by alanine.

tRNA synthetase (AlaRS) 2 (14,15). Normally, the mischarged tRNAs are cleared by a distinct editing activity, specific to each synthetase. This activity is able to distinguish the correct from the incorrect amino acid fused to its cognate tRNA. In addition to the editing activity that is part of the synthetase and encoded as a separate domain (16 -18), there are a few instances of freestanding editing domain homologs encoded by various genomes (12, 19 -22). The most widespread (through all three kingdoms of life) of these is AlaXp, a small protein that is a homolog of the editing domain of AlaRS and has been shown to edit Ser-tRNA Ala and Gly-tRNA Ala in vitro (21)(22)(23)(24). Not clear, however, is whether AlaXp plays a role in vivo in guarding against mistranslation.
The importance of editing for the (conditional) survival of bacteria is well established (7,25,26). The connection of editing defects to disease in mammalian systems has also been demonstrated from investigations in vivo of the editing activities of valyl-tRNA synthetase (ValRS) and AlaRS. For example, in the case of ValRS, inducing the expression of a transgene encoding an editing-defective ValRS in mammalian cells creates a transdominant, pathological phenotype (27). In the mouse, a mild editing defect in AlaRS leads to neurodegeneration characterized by ataxia and disintegration of Purkinje cells in the cerebellum (28). The editing-defective murine AlaRS synthesizes small amounts of Ser-tRNA Ala , and the tiny amount of incorporation of Ser at codons for Ala eventually triggers the unfolded protein response and cellular markers for apoptosis.
Thus, the need for editing to guard against mistranslation is now recognized, but the contribution of components like AlaXp remains unanswered because activity in vivo has never been demonstrated. Because Escherichia coli is an exception to the usual case where AlaXp is present, it offers an opportunity to investigate an AlaXp in vivo in circumstances where the bacterium has an editing-defective AlaRS. Previous work showed that serine was toxic to E. coli when it harbored an editing-defective AlaRS (14), thus giving an example, in bacteria, of the significance of mistranslation at codons for alanine. Given these circumstances, the question we posed was whether a transgene encoding an AlaXp could rescue an editing-defective strain of E. coli and, if so, whether that rescue depended on determinants in AlaXp that were essential for its editing function.

EXPERIMENTAL PROCEDURES
Preparation of tRNA-E. coli tRNA Ala(UGC) was prepared in vitro as described previously (14), except the transcription was done using a MEGAshortscript kit (Ambion, Austin, TX). Misacylated [ 3 H]Ser-tRNA Ala was produced in vitro using E. coli AlaRS C666A/Q584H as described previously (14).
Protein Expression and Purification-E. coli AlaRS and Methanosarcina mazei AlaXp proteins used in activity assays were prepared by expression from plasmid pET21b (Novagen, Madison, WI) as described previously (24).
Deacylation Assay-[ 3 H]Ser-tRNA Ala (ϳ5 M) was incubated at room temperature with AlaRS (50 nM) or AlaXp (10 nM) in assay buffer before being quenched into 96-well Multiscreen filter plates as described previously (29). Liberated [ 3 H]Ser was collected by centrifugation into 96-well flexible plates, and radioactive counts were measured on a MicroBeta plate reader.
Halo Assay-Media for the halo assay was made with 15 g/liter agar, and each plate contained 25 ml. 10 l of an overnight culture was diluted into 1 ml of sterile water and spread on a M9-Kan/Amp/Cm plate. 100 l of 1 M L-serine was added to a well cut into the center of the plate and allowed to diffuse into the agar, creating a radial gradient. Plates were incubated at 37°C and imaged after 48 h.
Direct Growth Assay-To confirm the serine sensitivity of AlaRS C666A/Q584H, M9-Kan/Amp plates were poured with 0 or 5 mM L-serine. 10 l of an overnight culture was streaked on the plates and incubated at 37°C for 40 h. For testing AlaXp, M9-Kan/Amp/Cm plates were poured with 0, 5, 7.5, or 10 mM L-serine. Overnight cultures were diluted 1:100, and 5 l was streaked on the plate. Plates were incubated at 37°C and imaged after 72 and 96 h.
Nitrocellulose Filter Binding Assay-In vitro transcribed tRNA Ala was end-labeled with 32 P using a KinaseMax kit (Ambion). The filter binding assays were carried out as described previously (33). Briefly, [ 32 P]tRNA Ala (20 nM) was incubated with AlaXp (0 -3.75 M) in 60 mM KH 2 PO 4 (pH 5), 10 mM MgCl 2 , 0.1 mg/ml bovine serum albumin for 10 min at 25°C. Reactions were spotted onto nitrocellulose filters (Millipore) soaked in 5% trichloroacetic acid and washed over a vacuum with assay buffer. Filters were dried, and captured [ 32 P]tRNA Ala was measured by scintillation counting.

RESULTS
Overall Experimental Strategy-The challenge was to find a way to test in vivo for the activity of AlaXp in isolation from any contribution of editing activity from endogenous AlaRS. For this purpose, we needed a strain of bacteria that harbored an AlaRS with completely ablated editing activity. Because the sites for aminoacylation and editing are discrete, the activity for editing can in principle be abolished without disrupting the essential aminoacylation function. To abolish the editing function, two mutations were placed in the editing domain of AlaRS. The purpose of this mutant was to have total elimination of the editing function while maintaining aminoacylation function. The double mutant was introduced into a bacterial strain carrying a deletion of the endogenous AlaRS gene. With the endogenous editing activity from AlaRS completely abolished, we then set out to test AlaXp function by introducing an AlaXp-encoding transgene into the editingdefective strain.
Construction and Testing Editing-defective C666A/Q584H AlaRS-Cys-666 in the editing domain is conserved through evolution for all AlaRSs. It is important for editing, because substitution of Cys-666 with Ala severely diminishes the ability of AlaRS to edit mischarged Ser-tRNA Ala or Gly-tRNA Ala (14). As a consequence, C666A AlaRS misacylates tRNA Ala with Ser or Gly. (In work reported below, we focused on Ser-tRNA Ala to study editing activities of mutant and wild-type proteins.) Separately, the Q584H mutation was serendipitously discovered to further enhance mischarging by C666A AlaRS in vitro (14). To further investigate the properties of C666A/Q584H AlaRS, we first established that its aminoacylation activity with alanine was unchanged from that of wild-type AlaRS (Fig. 1A). However, in contrast to the wild-type enzyme, C666A/Q584H AlaRS mischarged Ser onto tRNA Ala (Fig. 1B). Consistent with this mischarging, deacylation of Ser-tRNA Ala by C666A/ Q584H AlaRS was undetectable (at the level of background) (Fig. 1C).
Next, we tested the ability of C666A/Q584H AlaRS to sustain cell growth. For this purpose, we used E. coli strain W3110 (lacI q recA ⌬1 Kan r alaS⌬2) that harbors a chromosomal deletion of alaS (32). This ⌬alaS strain is maintained by a plasmid bearing wild-type alaS and a temperature-sensitive replicon. The maintenance plasmid is lost at 42°C, and growth can then be sustained only by introduction of a second plasmid that expresses a functional AlaRS and is not lost at the higher temperature. Thus, the question was whether a plasmid encoding C666A/Q584H AlaRS (under the control of the arabinose operon pBAD promoter) could rescue growth of the null strain after ejection of the maintenance plasmid.
We found that W3110 (lacI q recA ⌬1 Kan r alaS⌬2) harboring the gene for editing-defective C666A/Q584H AlaRS was able to grow on minimal media containing no additional amino acids but failed to grow on plates containing 5 mM serine (Fig. 1D). By contrast, a strain containing a plasmid encoding wild-type AlaRS, under the control of the same promoter, was able to grow on plates containing serine. This sensitivity of C666A/ Q584H AlaRS to high concentrations of serine is consistent with its inability to edit mischarged Ser-tRNA Ala and to prevent misincorporation of serine at codons for alanine.
In Vitro Editing Activity of M. mazei AlaXp and Its Dependence on Two Conserved Lysines-M. mazei AlaXp is a 270amino acid polypeptide that is homologous to the editing domain of AlaRS ( Fig. 2A). Like AlaXps from other sources (21,23), M. mazei AlaXp edits tRNA Ala mischarged with Ser or Gly in vitro (24). 3 We selected M. mazei AlaXp (a "type I" AlaXp) because it is small and, from earlier work, was shown to be soluble when expressed in E. coli (24). In eukaryotes, including mammals, AlaXp is extended by inclusion of the C-terminal domain of AlaRS (to give a "type II" AlaXp) that contributes nonspecific RNA binding elements (24). For the shorter AlaXps, such as the one from M. mazei, the sequence divergence (between AlaXp and the core editing domain of AlaRS) in a loop between two ␤ strands (␤6 and ␤7 in Fig. 2B) is thought to compensate for the determinants in the C-terminal domain of AlaRS that are missing in the type I AlaXps.
Key among those in the abovementioned loop is Lys-251 and Lys-253 (Fig. 2B). The double mutant K251A/K253A M. mazei AlaXp was shown previously to be inactive for editing in vitro (24). Our interest in this mutant was based on the idea that, if AlaXp was shown to be active in vivo by rescue of an editing-defective host strain, then finding that expression of an inactive (in vitro) mutant version of AlaXp did not rescue growth (of the editing-defective strain) would confirm that it was the editing activity per se of AlaXp that was needed for cell survival.
Pursuant to this line of investigation, we wanted to learn more about the basis for the lack of activity of K251A/K253A M. mazei AlaXp. From molecular modeling based on the structures of Pyrococcus horikoshii AlaXp and the threonyl-tRNA synthetase-tRNA Thr co-crystal (the editing domains of ThrRS and AlaRS are homologous) (22), these lysines (Lys-251 and Lys-253) within the basic loop come into close contact with the acceptor stem of tRNA Ala . Individual K251A and K253A mutant AlaXps were constructed, and each single mutant, the double mutant, and the wild-type protein was tested for activity toward Ser-tRNA Ala (Fig. 2C). The K253A AlaXp had a modest decrease in editing activity, whereas K251A AlaXp had a more severe editing defect. Importantly, K251A/K253A AlaXp had an even further reduction in editing activity, to nearly background levels. Thus, most of the previously observed defect in K251A/K253A AlaXp is due to the mutation of Lys-251. Consistent with this finding, of the two lysines, Lys-251 is the most conserved (Fig. 2B).
We postulated that the basic loop of M. mazei AlaXp contributed to tRNA binding affinity. However, when wild-type and mutant AlaXps were assayed for tRNA Ala binding by the nitrocellulose filter binding method, the dissociation constants were similar for all the AlaXp proteins. The K d values were 77 nM (wild-type), 71 nM (K253A), 48 nM (K251A), and 45 nM (K251A/K253A). Because the filter binding assays were done at pH 5.0, we wondered whether the lack of difference in the K d values was a consequence of the low pH. For that reason, we did 3 Y. E. Chong, unpublished observation. a kinetic analysis at pH 7.5 to determine relative values of K m for the wild-type and mutant proteins. The relative K m values were 1 (wild-type), 5.4 (K253A), 11.0 (K251A), and 11.4 (K251A/ K253A). Thus, under these conditions, the wild-type protein associates (as defined by K m ) 5-10-fold more strongly. Still, most of the difference (over 500-fold) in catalytic activity (k cat / K m ) of the wild-type versus the K251A/K253A mutant protein is in the k cat parameter. Thus, we conclude that the basic loop containing Lys-251 and Lys-253 may have an interaction with tRNA Ala that includes a component affecting k cat , such as anchoring the acceptor stem to facilitate positioning of the CCA-3Ј end in the catalytic site for editing.

M. mazei AlaXp Rescues Serine Toxicity of Editing-defective Strain-
To investigate the biological function of AlaXp in vivo, we transformed a plasmid carrying M. mazei AlaXp into the editing-defective C666A/Q584H AlaRS-bearing strain. The flow of these experiments is shown in Fig. 3A. First, the editing-defective strain was established by introducing a maintenance plasmid encoding C666A/ Q584H AlaRS into a ⌬alaS strain. Next the plasmid encoding M. mazei AlaXp was introduced. Both plasmids were under the control of pBAD arabinose-inducible promoters (31), with distinct markers and origins of replication to allow for co-transformation.
To determine whether AlaXp expressed from the transgene was able to rescue the serine toxicity experienced by the editing-defective AlaRS strain, cells were subjected to high concentrations of serine. For this purpose, a radial gradient was generated by diffusion of serine from a central well cut into the agar media. Sensitivity to serine was seen as a "halo" of cell toxicity around the central well. The zone of growth inhibition in the strain containing wild-type AlaXp was markedly smaller in size compared with that of a strain carrying the empty vector (Fig. 3B). The reduction in halo size (Fig. 3C) showed that the presence of AlaXp allows editingdefective AlaRS to sustain growth at much higher concentrations of serine.
To confirm that the editing activity per se of AlaXp is responsible for this phenomenon, another strain was created with the K251A/K253A AlaXp double mutant that, as noted previously, is defective in its editing activity in vitro (Fig. 2C). Growth of this strain in the presence of the radial serine gradient was only marginally better than that of the empty vector control (Fig. 3, B and C). No toxicity was observed in any of the strains when water or alanine was added to the central well in place of serine. Thus, the editing activity of AlaXp is able to rescue serine toxicity in the editing-defective AlaRS strain by preventing mistranslation at alanine positions. To determine whether results seen in the halo assays were due to disparate levels of one or more of the proteins (C666A/ Q584H AlaRS, AlaXp, K251A/K253A AlaXp) expressed from the introduced plasmids, cultures were inoculated from the plates at the conclusion of the halo assays. Cell lysates were separated on SDS-PAGE and immunoblotted with an ␣-His tag antibody. (The introduced plasmids encoding C666A/Q584H AlaRS and the AlaXps also encode a C-terminal His tag for each.) Expression of C666A/Q584H AlaRS was similar for all three strains (with slightly stronger expression in the strain bearing an empty vector in place of the plasmid encoding one of the AlaXps) (Fig. 3D). This result shows that rescue by AlaXp of the growth defect of the strain harboring C666A/Q584H AlaRS was not due to higher levels of C666A/Q584H AlaRS expression that, in turn, could have conferred a higher aminoacylation activity and faster growth rate. In addition, expression of wildtype and K251A/K253A AlaXps were comparable, with slightly stronger expression of the mutant AlaXp (Fig. 3D). Thus, the rescue seen with wild-type AlaXp was due to its catalytic activity rather than to a difference in expression levels.
To further confirm the ability of AlaXp to rescue the serine toxicity of the editing-defective strain, plates with various amounts of serine were prepared, and instead of a halo assay, colony growth was observed directly. The C666A/ Q584H AlaRS strain, which failed to grow in the presence of 5 mM serine (Fig. 1D), was able to grow when transformed with wild-type AlaXp (Fig. 3E). Growth of the strain encoding wild-type AlaXp was also observed at 7.5 and 10 mM serine (although the growth rate was diminished at these higher concentrations). In contrast, strains carrying either the empty vector or K251A/K253A AlaXp failed to grow at the same serine concentrations (Fig. 3E). The data in Fig. 3 strongly support the conclusion that the free-standing editing domain homolog, AlaXp, is functional for clearance of mischarged tRNA Ala in vivo.

DISCUSSION
The genetic complementation assay here was able to provide a clear demonstration of the activity in vivo of AlaXp. When challenged with serine, cells harboring an editing-defective AlaRS could only survive by expression of AlaXp in trans. The results demonstrate the significance of "post-transfer" editing; that is, clearance of mischarged tRNA after the aminoacyl group has been transferred from the adenylate (as aminoacyl-AMP) to the 3Ј-end of the tRNA. (In pre-transfer editing, the adenylate is hydrolyzed before the amino acid is transferred to the tRNA.) Interestingly, AlaXp was able to capture and hydrolyze Ser-tRNA Ala before the mischarged tRNA was bound and protected by EF-Tu and carried to the ribosome. It is possible that AlaXp weakly associates with AlaRS to improve the efficiency of clearance. On that point, the Ybak protein binds to ProRS to clear Cys-tRNA Pro (19).
It is also of interest to note that M. mazei AlaXp deacylates E. coli Ser-tRNA Ala in vivo, just as it does in vitro (24). This cross-species ability to recognize tRNA Ala in vivo is most likely because the domain for editing, like that for aminoacylation (34), recognizes the G3:U70 base pair that is common to almost all tRNA Ala sequences (24).
An unanswered question from earlier work was the rationale for the need for genome-encoded, free-standing homologs of C, relative areas of growth inhibition in the presence of a radial gradient of serine. The bars represent two independent experiments. D, expression of C666A/Q584H AlaRS and AlaXps in E. coli. Cultures were inoculated from the halo assay plates, and cell lysates were probed for expression using an ␣-His tag antibody. E, editingdefective AlaRS strain containing empty vector, wild-type AlaXp, or K251A/K253A AlaXp was streaked onto minimal media containing no amino acid or varying amounts of serine. The no serine and 5 mM serine plates were imaged after incubation at 37°C for 72 h. The 10 mM serine plate was imaged after 96 h of incubation. editing domains. The work here provides a direct demonstration that, indeed, one example of these free-standing domains can provide the editing function in a way that ensures cell survival. That survival is dependent on AlaXp, in this particular experimental system, places further emphasis on the sensitivity of cells to mistranslation. In particular, the results stimulate further interest in the question of why a small, 2-fold defect in the editing activity of murine AlaRS should be sufficient to cause neurodegeneration in the sti mouse (28). In addition to neurodegeneration, these mice have other pathologies, such as substantially reduced body mass and altered fur. These pathologies are present even though the sti mouse encodes wild-type AlaXp. Thus, mistranslation appears to be near a "tipping point" presented by the singular difficulties of discriminating, in this instance, glycine and serine from alanine.
We can also raise the question of why AlaXp is the most widespread of these genome-encoded editing domains. No examples are known of separate, genome-encoded CP1 editing domains of IleRS, LeuRS, or ValRS, and only a limited evolutionary distribution of the ThrXp domain (that is closely related to AlaXp) is seen (20,21). ProXp (and the related Ybak) is also widely distributed and has been shown to clear Ala-tRNA Pro (12,21). Thus, the precise insertion of alanine into proteins (not confusing Gly or Ser for Ala, and not confusing Ala for Pro) seems to present a major problem for many species through evolution. The "alanine problem" may be more difficult to overcome than other examples of amino acid misrecognition, such as the confusion of valine for isoleucine, threonine for valine, or serine for threonine. Whether the singular effects associated with mistranslation of alanine codons is due to the greater impact on protein folding of serine (or glycine) substitutions for alanine, or alanine substitutions for proline, is not clear but remains a formal possibility.