Human trans-editing enzyme displays tRNA acceptor stem specificity and relaxed amino acid selectivity

Accurate translation of genetic information into proteins is vital for cell sustainability. ProXp-ala prevents proteome-wide Pro-to-Ala mutations by hydrolyzing misacylated Ala-tRNAPro, which is synthesized by prolyl-tRNA synthetase (ProRS). Bacterial ProXp-ala was previously shown to combine a size-based exclusion mechanism with conformational and chemical selection for the recognition of the alanyl moiety, while tRNAPro is selected via recognition of tRNA acceptor stem elements G72 and A73. The identity of these critical bases changed during evolution with eukaryotic cytosolic tRNAPro possessing a cytosine at the corresponding positions. The mechanism by which eukaryotic ProXp-ala adapted to these changes remains unknown. In this work, recognition of the aminoacyl moiety and tRNA acceptor stem by human (Hs) ProXp-ala was examined. Enzymatic assays revealed that Hs ProXp-ala requires C72 and C73 in the context of Hs cytosolic tRNAPro for efficient deacylation of mischarged Ala-tRNAPro. The strong dependence on these bases prevents cross-species deacylation of bacterial Ala-tRNAPro or of Hs mitochondrial Ala-tRNAPro by the human enzyme. Similar to the bacterial enzyme, Hs ProXp-ala showed strong tRNA acceptor-stem recognition but differed in its amino acid specificity profile relative to bacterial ProXp-ala. Changes at conserved residues in both the Hs and bacterial ProXp-ala substrate binding pockets modulated this specificity. These results illustrate how the mechanism of substrate selection diverged during the evolution of the ProXp-ala family and provides the first example of a trans-editing domain whose specificity evolved to adapt to changes in its tRNA substrate.


Introduction
Correct pairing of amino acids with their corresponding tRNAs by aminoacyl-tRNA synthetases (ARSs) is fundamental for faithful translation of genetic information into proteins. These enzymes are divided into two classes based on structural differences within their aminoacylation active sites. The chemical and structural similarities of certain classes of amino acids challenge the ability of the synthetic active site, wherein aminoacyl-adenylate formation occurs, to properly select only cognate amino acids at this first step. Consequently, ARSs frequently ligate tRNAs with the wrong amino acid during the second step of aminoacylation (1). If left uncorrected, misacylation may lead to the incorporation of amino acids in response to the wrong codon (i.e., mistranslation); accumulation of these errors has been linked to cell defects in diverse organisms (1)(2)(3)(4)(5). To prevent mistranslation, many ARSs have acquired quality control mechanisms that enable deacylation of tRNAs immediately after their synthesis in a reaction known as post-transfer editing. This reaction may occur in cis in a domain distinct from the aminoacylation active site and seven members of both classes of ARSs possess such an editing domain (1). Class II synthetases have additionally been shown to edit aminoacyl-tRNAs in trans, that is, following substrate dissociation and rebinding (6)(7)(8). In addition, free-standing editing domains that are evolutionally related to the editing domains of three class II ARSs have been identified in all Domains of life; these enzymes lack aminoacylation activity and catalyze deacylation exclusively in trans (8).
The critical role of aminoacyl-tRNA editing is highlighted by the different phenotypes associated with editing deficiencies (9)(10)(11). For example, murine cells harboring an editing-deficient transgene for valyl-tRNA synthetase transition to an apoptotic-like state due to the incorporation of Thr at Val codons (2), while a mild defect in the editing activity of AlaRS provokes ataxia in mice (3). In E. coli, severe oxidative conditions inactivate the editing activity of threonyl-tRNA synthetase, which impairs cell growth (12), whereas heritable genome mutations have been identified in aging E. coli cells resulting from accumulation of translational errors from an editing defective isoleucyl-tRNA synthetase (13).
Prolyl-tRNA synthetases (ProRSs) commonly misactivate Ala and Cys, which are smaller or similar in size to cognate Pro (14)(15)(16)(17). In most bacteria, mistranslation of Pro codons with Ala is prevented by the ProRS editing domain, which is known as the insertion (INS) domain due to its location between the class II consensus Motifs 2 and 3 (18). This domain specifically hydrolyzes Ala-tRNA Pro but rejects Pro-and Cys-tRNA Pro (14). The later error is corrected by a single-domain trans-editing enzyme known as YbaK, which is homologous to the INS domain but possesses unique specificity for Cys-tRNAs (19). While tRNA Pro misacylation with Ala is an inherent characteristic of ProRSs, the INS domain is not present in some bacteria nor is it found in most eukaryotes (15). Instead, many organisms that lack INS encode ProXp-ala, another homolog of the INS domain that hydrolyzes Ala-tRNA Pro in trans (20)(21)(22). Structural studies revealed that the INS domain from Enterococcus faecium and ProXp-ala from Caulobacter crescentus (Cc) share a conserved tertiary structure and key catalytic residues (23). Biochemical studies revealed that Cc ProXp-Ala and Escherichia coli (Ec) ProRS share similar sizeexclusion and chemical-based mechanisms for recognition of the aminoacyl moiety, limiting their ability to hydrolyze tRNA Pro charged with cognate Pro as well as other amino acids that are larger than Ala (24,25). In contrast to their shared mechanism of amino acid recognition, INS and bacterial ProXp-ala developed distinct strategies for tRNA Pro selection. The INS domain lacks intrinsic tRNA specificity and relies on the recognition of the tRNA anticodon by the anticodon binding domain of ProRS, whereas Cc ProXp-ala displays strong recognition of the acceptor stem elements of bacterial tRNA Pro (26). The tRNA specificity of both enzymes is crucial for preventing deacylation of cognate Ala-tRNA Ala .
Bacterial tRNA Pro sequences show high conservation of the A73 discriminator base and the C1:G72 base pair, which is unique to the tRNA Pro acceptor stem (27). Cc ProXp-ala recognizes tRNA Pro through specific interactions with G72 and A73 (26). These bases are also critical for aminoacylation of tRNA Pro by bacterial ProRS (28); thus, they play a dual role in ensuring the correct aminoacylation of tRNA Pro with Pro and in editing of misacylated Ala-tRNA Pro . The identity of these bases at the top of the tRNA Pro acceptor stem changed during evolution; eukaryotic cytosolic (cyto) tRNA Pro encodes C73 and G1:C72 (Fig. 1A) (27). These differences in the acceptor stem prevent aminoacylation of human (Hs) cyto tRNA Pro by bacterial ProRS (29) as well as deacylation of Hs cyto Ala-tRNA Pro (hereafter referred to as Hs Ala-tRNA Pro ) by bacterial ProXp-ala (26). Even though bacterial tRNA Pro is a poor substrate for Hs ProRS, the human enzyme only weakly relies on recognition of the tRNA Pro acceptor-stem bases, while strongly relying on interactions with the anticodon, as well as other structural features of Hs tRNA Pro (30). A previous study using a variant of tRNA Pro that could readily be misacylated with Ala by alanyl-tRNA synthetase (AlaRS) concluded that Hs ProXp-ala has amino acid specificity but lacks tRNA specificity (31). Here, we prepared native aminoacyl-tRNA Pro substrates to investigate the amino acid and tRNA specificity of Hs ProXp-ala. Our findings suggest that the substrate binding pocket of Hs ProXp-ala displays different amino acid selectivity relative to that of the bacterial enzyme, but the human enzyme maintains strong acceptor stem recognition. We also identified residues in the substrate binding pocket of both bacterial and Hs ProXp-ala that modulate their amino acid specificity. The results reported here provide insights into the evolutionary relationship between class II aminoacyl-tRNA synthetases and editing domains.

Hs ProXp-ala robustly deacylates Ala-tRNA Pro
In this work, we used in vitro transcribed Hs cyto tRNA Pro (Fig. 1A) and full-length recombinant Hs ProXp-ala expressed and purified in E. coli. To obtain sufficient yield of misacylated Hs Ala-tRNA Pro substrates without the need for tRNA acceptor stem mutation, we used flexizyme technology, which allows tRNAs to be charged with virtually any amino acid (proteinogenic or non-proteinogenic) (Fig. S1) (32). Our initial experiments showed that Hs ProXp-ala robustly deacylated Ala-tRNA Pro in vitro. Based on an analysis of Hs Ala-tRNA Pro deacylation by varying concentrations of Hs ProXp-ala under singleturnover conditions, we determined an apparent dissociation constant (Kd) of 10.2 µM and a maximum rate constant (kmax) of 3.8 min -1 (Fig. 2). Relative to the cyto tRNA, much weaker deacylation of Hs mitochondrial (mito) Ala-tRNA Pro by Hs ProXp-ala was observed (Fig. 1B). Conversely, Cc ProXp-Ala deacylates mito Ala-tRNA Pro with similar efficiency as bacterial Ala-tRNA Pro (Fig. 1C). We hypothesize that differences in the acceptor stem sequences of these two tRNAs (Fig. 1A) may account for these differences in deacylation activity.

Human ProXp-ala specificity is determined by tRNA acceptor stem elements
We previously showed that bacterial ProXp-ala cannot efficiently deacylate Hs cyto Ala-tRNA Pro because of the differences in the identities of the bases at position 72 and 73 in this eukaryotic tRNA Pro (Fig. 1C) (26). To test whether Hs ProXpala exhibits cross-species deacylation capability, assays with Ec Ala-tRNA Pro were performed under single-turnover conditions. We found that Hs ProXp-ala only weakly deacylated Ec Ala-tRNA Pro with a 66-fold decrease in kobs ( Fig. 3 and Table 1). This is consistent with the observation that Hs mito Ala-tRNA Pro (which shares the C1:G72 and A73 acceptor stem elements of Ec tRNA Pro ) is a poor substrate for Hs ProXp-ala (Fig. 1B). To identify the tRNA features that determine the specificity of Hs ProXp-ala, we prepared three Hs cyto tRNA Pro acceptor-stem mutants (C73A, double mutant G1C:C72G, and triple mutant G1C:C72G/C73A) that mimic the acceptor stem of Ec tRNA Pro (Fig.  1A). We found that the C73A and G1C:C72G tRNA Pro mutants were hydrolyzed 13-and 16-fold slower than WT Ala-tRNA Pro , respectively, whereas the triple mutant G1C:C72G/C73A tRNA Pro was deacylated at an ~100-fold reduced rate, similar to the decrease observed with Ec WT tRNA Pro (Fig. 3 and Table 1). These data indicate that the first base pair and discriminator base of tRNA Pro are important elements for Hs ProXp-ala recognition and deacylation activity. To further validate the role of these bases, we prepared a series of Ec tRNA Pro acceptor-stem variants containing Hs tRNA Pro elements (Fig. 1A). We found that Ec C73 and G1:C72 tRNA Pro mutants were significantly better substrate for the human enzyme than Ec WT tRNA Pro , while Ec G1:C72/A73 tRNA Pro was deacylated slightly more efficiently than Hs WT tRNA Pro (Fig. 3 and Table 1). Together, these results show that the deacylation activity of Hs ProXp-ala depends strongly on the identity of the nucleotides at the top of the acceptor stem of Hs cyto tRNA Pro .
To further study the effect of different substitutions at the 73 position, we charged or mischarged three Ec tRNAs with Ala: tRNA Ala , tRNA Ser and tRNA Cys , which contain A73, G73 and U73, respectively (all have G1:C72). We found that the efficiency of Ala-tRNA Ala deacylation is very similar to other G1:C73 tRNA Pro variants with A73 (Table 1). Ala-tRNA Cys (U73) was a slightly better substrate and Ala-tRNA Ser (G73) displayed the lowest kobs value of the three. These results emphasize a key role of the discriminator base for tRNA deacylation by Hs ProXp-ala.

Aminoacyl substrate specificity of Hs ProXp-ala
We next investigated the aminoacyl specificity of Hs ProXp-ala using Hs tRNA Pro charged with amino acids of different classes including 2aminobutyric acid (Abu), Ser, Thr, Pro, and azetidine-2-carboxylic acid (Aze) (Fig. 4). The deacylation rate of tRNA Pro charged with the larger non-polar Abu amino acid is 20-fold slower than Ala-tRNA Pro (Table 2). Likewise, for tRNA Pro charged with polar amino acids, the larger Thr-tRNA Pro is a poorer substrate than Ser-tRNA Pro . Thus, for these amino acids, the deacylation efficiency correlates with size of the aminoacyl moiety, with a preference for the smaller substrate. In contrast, Aze, a 4-membered ring analog of Pro, with a molecular volume slightly larger than Ser, was deacylated 6-fold more efficiently than Ser and only 3.4-fold less efficiently than Ala. Aze-tRNA Pro deacylation is also slightly greater (~2-fold) than the rate of Pro-tRNA Pro deacylation by the human enzyme, which surprisingly is only ~5-fold lower than that of Ala-tRNA Pro ( Table 2).

The amino acid specificity of Hs ProXp-ala is tunable
The aminoacyl specificity of the Ec INS editing domain is modulated by conserved residues in the substrate binding pocket that include L266 and H369 (18,25). Whether the corresponding residues play a role in determining the overall specificity of Hs ProXp-ala is unknown. L266 and H369 of Ec INS align with M35 and H137 of Hs ProXp-ala, respectively (Fig. 5). We previously showed that L266 plays a role in the specificity of the INS towards Ser-tRNA; when this residue in Ec INS is mutated to Ala, Ala-tRNA Pro is deacylated less efficiently, while the mutant enzyme hydrolyzes Ser-tRNA Pro more robustly than WT INS (25). We now show that replacing M35 of Hs ProXp-ala with smaller Gly increases the rate of Ser-tRNA Pro deacylation by 2.3-fold while decreasing the rate of Ala-tRNA Pro deacylation by 4.7-fold ( Fig. 6 and Table 3). The highly conserved His residue in the substrate binding pocket of Ec INS has previously been shown to function to prevent cognate substrate deacylation; mutation of this residue to Ala significantly increased Pro-tRNA Pro deacylation (18). We now show that a homologous mutation (H137A) in Hs ProXp-ala also improves its activity towards Pro-tRNA Pro , which is now deacylated only 2-fold weaker than Ala-tRNA Pro by the same mutant ( Fig. 7A and Table 3). Similarly, a Cc H130A ProXp-ala variant showed strong deacylation of Pro-tRNA Pro (Fig. 7B). Interestingly, the Ala-tRNA Pro deacylation activities of Cc H130A and Hs H137A ProXp-ala were not significantly different than those of the WT enzymes.

Phylogeny and distribution of the ProXp-ala family
Previous bioinformatic analyses showed that bacterial ProXp-ala is primarily encoded in organisms belonging to the Alphaproteobacteria, Lactobacillales, Clostridia, and Negativicutes classes (20,22), whereas eukaryotic ProXp-ala have only been reported for Hs, Mus musculus, Arabidopsis thaliana, Trypanosoma brucei and Plasmodium falciparum (20). To update our understanding of the overall phylogenetic distribution of ProXp-ala in Eukarya, we searched all available 540 eukaryotic sequenced genomes and identified ProXp-ala genes in approximately 54% of the organisms analyzed (Table S1, Fig. 8). ProXp-ala is widely distributed in mammals, birds, fish, reptiles and amphibians, while it is absent in other analyzed Animalia groups such as insects, arthropods, and worms. ProXp-ala is also encoded by almost all sequenced plants. The predicted plant ProXp-ala gene sequence is almost twice the size (~320 amino acids) of that found in bacteria and the Animalia kingdom. The novel C-terminal extension of plant ProXp-ala shares no primary sequence homology with any known protein and its function is unknown. In contrast to plants and animals, ProXp-ala is fused to the N-terminus of ProRS in a group of animal parasites belonging to the Stramenopila, Aveolates and Rhizaria (SAR) supergroup, which includes P. falciparum and Toxoplasma gondii. This ProXp-ala-ProRS fusion is also observed in Trypanosoma brucei ProRS and some other species from Kinetoplasida (e.g., Leishmania). We also identified putative ProXp-ala genes in several archaeal species including members of the Asgard archaea superphylum, Candidatus Prometheoarchaeum syntrophicum MK-D1 and Lokiarchaeum sp. GC14_75. Both organisms have been associated with eukaryogenesis and may be provide a link to the evolution of ProXp-ala (33,34).
A phylogenetic tree of the ProXp-ala family revealed the formation of distinct clades corresponding to eukaryotic and bacterial ProXpala with only two instances of possible horizontal gene transfer events in the eukaryotes Entamoeba dispar and Ostreococcus tauri (Fig. 9). The phylogenic relationship of the ProXp-ala family underlines their divergence during evolution, which is reflected in distinct substrate features.

Discussion
In contrast to a previous report, which concluded that Hs cyto ProXp-ala lacks tRNA specificity (31), we found a strong dependence on elements at the top of the acceptor stem for efficient deacylation by the Hs enzyme (Table 1). In the earlier study, the reported activity was weak even at an elevated enzyme concentration of 5 µM (31). It is possible that the relatively weak deacylation activity was due to the use of a tRNA Pro mutant with three substitutions (C3G, G70U, and C73A) relative to WT tRNA Pro . These mutations were introduced to facilitate the preparation of Ala-tRNA Pro using AlaRS (31). While we did not test the third base pair in our work directly, we found that C73 is a critical recognition element for Hs ProXp-ala and G3:U70containing Ec Ala-tRNA Ala is a relatively poor substrate despite the fact that it contains G1:C72 (Table 1). In addition, differences in the N-terminus of the recombinant Hs ProXp-ala used previously relative to the genome-encoded sequence available in UniProt (35) may have impacted enzymatic function (Fig. S2) (31).
The changes in the sequence of the acceptor stem of tRNA Pro during evolution led to the development of distinct modes of tRNA recognition by ProRSs across the three domains of life (28,30,36). Here we found that these same changes also exerted pressure during the evolution of the trans-editing enzyme ProXp-ala from Bacteria to Eukarya, enabling its adaptation to cytosolic tRNA Pro . However, Hs ProRS and ProXp-ala altered their substrate selection mechanisms in response to the tRNA Pro acceptor stem changes differently. Hs ProRS relaxed its requirement for specific acceptor stem nucleotides and instead gained a stronger affinity for the anticodon, as well as a specific D-stem loop requirement (30). In contrast, Hs ProXp-ala specificity relies on the G1:C72/C73 nucleotides of tRNA Pro (Fig. 3, Table 1), which are the same acceptor stem positions recognized by bacterial ProXp-ala, albeit with different base identities (i.e. C1:G72/A73) (26). Consequently, Hs ProXp-ala only weakly deacylated bacterial and mitochondrial Ala-tRNA Pro (Fig. 1B), and its strong tRNA specificity likely prevents efficient deacylation of Ala-tRNA Ala (Table 1). These results, and the fact that C73, which is rarely found at this position and is a universal feature of all eukaryotic cyto tRNA Pro , indicate that ProXp-ala is likely to act exclusively in the cytosol of higher eukaryotes. Studies to determine the cellular localization of Hs ProXp-ala are required to confirm this hypothesis. Thus, whereas acceptor stem elements of bacterial tRNA Pro play a dual role in aminoacylation and editing, the acceptor stem elements that only weakly mark cytosolic tRNA Pro for aminoacylation with Pro by Hs ProRS, play a more important role in the proofreading of aminoacyl-tRNA Pro by Hs ProXp-ala.
Biochemical characterization also showed that Hs ProXp-ala displays an aminoacyl specificity profile that does not follow a simple relationship between the rate of hydrolysis and the size (molecular volume) of the amino acid. For example, Hs ProXpala deacylated Abu-tRNA Pro with similar efficiency as Ser-tRNA Pro , despite Ser being significantly smaller than Abu. In contrast, tRNA Pro charged with Aze (which resembles the volume of Ser) was deacylated several fold better than Ser-tRNA Pro and with similar efficiency as Pro-tRNA Pro (Fig. 4, Table 2). Thus, in comparison to the bacterial system (24), the Hs ProXp-ala active site better accommodates the pyrrolidine and azetidine side chains of Pro and Aze and shows greater discrimination against polar OH-containing amino acids. Consequently, Cc and Hs ProXp-ala appear to have evolved not only different tRNA acceptor stem recognition specificities but also different preferences in amino acid selection.
Our study uncovered key residues that define the amino acid specificity of the ProXp-ala family. In Hs ProXp-ala, mutations at M35 and H137 modulated amino acid recognition, similar to the corresponding Leu and His residues in the substrate binding pocket of the Ec INS domain (25). Substitution of M35 with G enlarges the Hs enzyme's binding pocket and allowed better accommodation of Ser-tRNA Pro . The corresponding position in Cc ProXp-ala corresponds to G32, and WT Cc ProXp-ala has a binding pocket that enables it to efficiently accommodate and robustly hydrolyze Ser-tRNA Pro (24).
Our results also confirmed that a highly conserved His residue in the aminoacyl-binding pocket of INS/ProXp-ala family members functions as a universal gatekeeper to prevent or reduce cognate Pro-tRNA Pro deacylation. In the case of INS, the H369A mutation led to a switch of specificity from Ala-to Pro-tRNA Pro (18), whereas the analogous mutation in Cc and Hs ProXp-ala resulted in dual activity for Ala-and Pro-tRNA Pro . Notably, H137 of Hs ProXp-ala is less effective at preventing Pro-tRNA Pro deacylation relative to the bacterial system (Fig. 7). The rather efficient deacylation of cognate Pro-tRNA Pro observed for the Hs trans-editing enzyme in vitro is surprising and suggests that other factors in the cell likely prevent this undesired reaction. Based on the ability of bacterial elongation factor Tu (EF-Tu) to protect correctly charged tRNAs from free-standing editing domains, including bacterial ProXp-ala (22,37), we propose that Hs EF-1A may outcompete Hs ProXpala for binding to Pro-tRNA Pro , allowing only deacylation of the mischarged Ala-tRNA Pro species in cells. If editing occurred in cis (i.e., by a ProRS editing domain prior to substrate dissociation) rather than in trans, this competition would be less likely to be successful and may explain why eukaryotic ProRSs rely on a free-standing transediting enzyme for proofreading. Whether Aze-tRNA Pro is protected by elongation factors is unknown. Aze is a proline analog synthesized in many plants; it is also misactivated by Hs ProRS (15). High levels of Aze administered to zebrafish embryos lead to Aze-tRNA Pro accumulation and cell toxicity (38). Thus, it is possible that the dual activity towards Ala and Aze, albeit weaker for the latter, displayed by Hs ProXp-ala serves to prevent mistranslation of Pro codons with either amino acid when present at low levels.
The results reported here underscore the complex evolutionary relationship between class II aminoacyl-tRNA synthetases and editing domains. The tRNA Pro -specific INS/ProXp-ala family of editing domains likely arose from the biological pressure to prevent mistranslation of Pro codons due to the critical nature of this amino acid for protein structure and function in all Domains of life. The ProRS INS domain, found only in bacteria (14,39), and the ProXp-ala family, found primarily in bacteria and eukarya (Fig. 9), have the same specificity toward editing of Ala-tRNA Pro ; yet, ProXp-ala is usually encoded as a free-standing domain (with a few exceptions, see Fig. 9) and INS is always found appended to ProRS (14,20,22,39). The latter observation may be due to the lack of strong tRNA acceptor stem specificity of this domain (26), which would result in promiscuous trans-editing of Ala-tRNA Ala . Each organism has likely evolved to use the type of Ala-tRNA Pro editing domain (INS vs. ProXp-ala) and domain structure (free-standing or synthetase-appended) to ensure optimal levels of fidelity required for cell survival.

Materials
All amino acids, nucleotides, enzymes and other materials were purchased from MilliporeSigma unless otherwise noted.

Protein preparation
All WT and mutant proteins used in this study were prepared as previously described: Hs ProXp-Ala

Preparation of tRNAs and aminoacyl-tRNA substrates
All tRNA variants used in these studies were prepared by in vitro transcription as previously described (14). Briefly, for each tRNA, a linear DNA template for T7 RNA polymerase was obtained by BstN1 (NEB) digestion of a plasmid carrying the corresponding gene. Transcribed tRNA was purified using denaturing 10% polyacrylamide gel electrophoresis. To ensure high yields of tRNA variants with cytidine at position 1, a hammerhead ribozyme was inserted at 5′ of the tRNA gene (42). A plasmid encoding Hs tRNA Pro C1:72G variant was ordered from Genewiz; all other acceptor stem tRNA Pro mutants were made by SLIM (41) and confirmed by DNA sequencing (The Genomics Shared Resource at The Ohio State University Comprehensive Cancer Center). All tRNAs used in deacylation assays were labeled with 32 P at the 3´-A76 using tRNA nucleotidyltransferase (43). Preparation of aminoacylated Hs tRNA Pro was carried out in 50 mM HEPES, pH 7.5, 20 mM KCl, 4 mM ATP, 10 mM MgCl2, 0.1 mg/mL BSA, and 1 mM DTT by incubating 10 µM Hs ProRS, 10 µM tRNA Pro , and 0.03 mg/mL pyrophosphatase (Roche) with the following amino acids: 900 mM Ala, 30 mM Pro, or 300 mM Aze for 5 min at 37 °C. Ec Pro-and Ala-tRNA Pro were prepared under similar conditions with editing deficient Ec K279A ProRS (25). Aminoacyl-tRNAs were phenol chloroformextracted followed by ethanol precipitation. Acceptor-stem tRNA Pro variants as well as Ec tRNA Ser , Ec tRNA Ala , and Ec tRNA Cys were charged with Ala using dinitro-flexizyme (dFx) and Ala-3,5-dinitrobenzyl ester (Ala-DBE) as described (44). Briefly, 32 P-labeled tRNA and dFx were refolded by heating at 90 °C for 1 min followed by addition of MgCl2 and cooling to room temperature for 3 min. The aminoacylation reaction was carried in 100 mM HEPES-KOH (pH 7.5), 20 mM MgCl2, 25 µM tRNA, 25 µM dFx, 5 mM Ala-DBE, Ser-DBE or Abu-DBE for 20 h on ice. Thr-tRNA Pro was prepared using enhanced-flexizyme (eFx) and Thr-4-chlorobenzyl thioester (44). Substrates for deacylation assays were ethanol precipitated and dissolved in 3 mM sodium acetate, pH 5.2, and stored at -80 °C prior to use.

Deacylation assays
Single-turnover aminoacyl-tRNA Pro deacylation reactions were performed as previously described (24). Hs ProXp-Ala (WT and mutants) reactions typically contained 0.1 μM 32 P-labeled aminoacyl-tRNA Pro and 1.5 μM ProXp-ala in deacylation buffer A (50 mM HEPES pH 7.5, 20 mM KCl, 5 mM MgCl2, 0.1 mg/ml BSA, 1 mM DTT) and were performed at 30 °C. Cc ProXp-Ala reactions contained 0.1 μM 32 P-labeled aminoacyl-tRNA Pro and 0.75 μM ProXp-ala in deacylation buffer B (150 mM potassium phosphate pH 7.0, 5 mM MgCl2, 0.1 mg/mL BSA) and were performed at 18 ºC. Reactions were initiated by mixing equal volumes of aminoacyl-tRNA Pro and enzyme. At the chosen time points, 2-µL aliquots were quenched into a solution containing 20 units S1 nuclease, 1 mM zinc acetate, and 200 mM NaOAc mM sodium acetate, pH 5.0. Product formation was monitored by separating aminoacyl-A76 from A76 on PEIcellulose plates (EMD Millipore) using a mobile phase of 0.05% ammonium chloride/5% acetic acid. Radioactive products were detected by autoradiography using a Typhoon FLA 9500 (GE Healthcare) and quantified using ImageQuant TL 8 software (GE Healthcare). Comparative analysis of numerous tRNA and enzyme variants was performed using single-turnover rate conditions with the concentration of Hs ProXp-ala several fold smaller than the apparent Kd.
Under these conditions, the observed rate constant (kobs) reflects kcat/Km and allows comparison of the enzyme efficiency for various substrates. Observed rate constants were obtained by fitting the time course for aminoacyl-tRNA deacylation with a single-exponential equation using SigmaPlot (Systat Software). Each reported rate constant is an average of three independent assays. The graphs show representative time courses. All data points were corrected for non-enzymatic buffer hydrolysis.

Amino Acid Volume Calculations
The structure of each free amino acid (Ala, Abu, Aze, Pro, Ser, and Thr) was obtained from the Protein Data Bank. Molecular volumes for each amino acid were calculated by loading each .pdb file into the Volume Assessor module of the 3V webserver (http://3vee.molmovdb.org/volumeCalc.php) using a 0.1 Å probe radius and high grid resolution (0.5 Å voxel size) (45).

Phylogenetic analysis of ProXp-ala
A total of 542 eukaryotic sequenced genomes (retrieved from the KEGG database, www.genome.jp, and the Joint Genome Institute Genomes Online Database (46)) were individually searched for ProXp-ala genes using the National Center for Biotechnology Information BLASTp service (Table S1). ProXp-ala reference sequences from Hs, Cc, Clostridium sticklandii, Agrobacterium tumefaciens, and Arabidopsis thaliana were used as queries. A global search of ProXp-ala genes in Archaea was also carried out. The resulting putative genes were designated as ProXp-ala only if they had the signature residues of ProXp-ala (i.e. K45, H130, GXXXP/A, numbering based on C. crescentus ProXp-ala). The eukaryotic ProXp-ala phylogenetic tree was built via multiple sequence alignment using Clustal Omega (47) followed by a Maximum Likelihood analysis with 100 bootstraps using MEGA X with default settings (48). The ProXp-ala family tree was built using the NGPhylogeny.fr platform (49). A workflow with Clustal Omega sequence alignment followed by tree inference using PhyML with 100 bootstraps was used. The bacterial, eukaryotic, and archaeal ProXp-ala sequences from the organisms listed in Table S1 were used for the tree. The iTOL online tool was used to display both trees (50).           The branches corresponding to bacterial, eukaryotic, and archaeal ProXp-ala are colored in red, blue, and green, respectively. The phylum, clade, or kingdom to which each organism belongs are indicated. ProXp-ala is fused to ProRS in organisms from the Discoba, Alveolata, and Stramenopiles clades. Branches with bootstrap values higher than 70 are indicated by black dots.