Exclusive Use of trans-Editing Domains Prevents Proline Mistranslation*

Background: Editing domains of aminoacyl-tRNA synthetases ensure high fidelity in codon translation, but the extent of single-domain trans-editing is unclear. Results: A bioinformatics analysis reveals distribution of five distinct homologs of the bacterial prolyl-tRNA synthetase editing domain. Conclusion: A triple-sieve mechanism of editing in Caulobacter crescentus relies exclusively on trans-editing. Significance: A diversity of approaches is used by bacteria to prevent proline mistranslation. Aminoacyl-tRNA synthetases (ARSs) catalyze the attachment of specific amino acids to cognate tRNAs. Although the accuracy of this process is critical for overall translational fidelity, similar sizes of many amino acids provide a challenge to ARSs. For example, prolyl-tRNA synthetases (ProRSs) mischarge alanine and cysteine onto tRNAPro. Many bacterial ProRSs possess an alanine-specific proofreading domain (INS) but lack the capability to edit Cys-tRNAPro. Instead, Cys-tRNAPro is cleared by a single-domain homolog of INS, the trans-editing YbaK protein. A global bioinformatics analysis revealed that there are six types of “INS-like” proteins. In addition to INS and YbaK, four additional single-domain homologs are widely distributed throughout bacteria: ProXp-ala (formerly named PrdX), ProXp-x (annotated as ProX), ProXp-y (annotated as YeaK), and ProXp-z (annotated as PA2301). The last three are domains of unknown function. Whereas many bacteria encode a ProRS containing an INS domain in addition to YbaK, many other combinations of INS-like proteins exist throughout the bacterial kingdom. Here, we focus on Caulobacter crescentus, which encodes a ProRS with a truncated INS domain that lacks catalytic activity, as well as YbaK and ProXp-ala. We show that C. crescentus ProRS can readily form Cys- and Ala-tRNAPro, and deacylation studies confirmed that these species are cleared by C. crescentus YbaK and ProXp-ala, respectively. Substrate specificity of C. crescentus ProXp-ala is determined, in part, by elements in the acceptor stem of tRNAPro and further ensured through collaboration with elongation factor Tu. These results highlight the diversity of approaches used to prevent proline mistranslation and reveal a novel triple-sieve mechanism of editing that relies exclusively on trans-editing factors.

adenylate and possesses a cis-editing domain (INS) responsible for Ala-tRNA Pro clearance (7,8,11). E. coli also encodes a single-domain INS homolog known as YbaK, which deacylates Cys-tRNA Pro in trans (10,12). Therefore, a "triple-sieve" editing mechanism operates in E. coli, wherein the aminoacylation site of ProRS acts as the coarse sieve to reject amino acids larger than proline allowing the synthesis of Ala-and Cys-tRNA Pro , in addition to cognate Pro-tRNA Pro . The cis-editing activity of INS is responsible for clearing Ala-tRNA Pro in the second or fine sieve. Finally, in the third sieve, YbaK edits Cys-tRNA Pro in trans. The latter has been shown to occur via chemical discrimination using the unique sulfhydryl-specific chemistry of the cysteine substrate side chain (13). This mechanism explains how E. coli and many bacteria likely achieve high fidelity in proline codon translation. Although the INS domain is found widely distributed throughout bacteria, a significant number of bacteria encode a ProRS lacking a full-length INS. Thus, the triple-sieve mechanism that is operational in E. coli is not applicable in these cases.
INS and YbaK belong to a family of homologous INS-like proteins that appears to include four other single-domain proteins previously named PrdX, ProX, YeaK, and PA2301 (9,10,(12)(13)(14)(15). In vitro activity studies have been reported for Clostridium sticklandii PrdX, which was demonstrated to be a robust Ala-tRNA Pro deacylase (9,12). Based on this activity we now rename this protein as ProXp-ala. Because C. sticklandii ProRS lacks an INS domain, it was proposed that some organisms use trans-editing factors to compensate for the lack of a cis-editing domain (9). The function of the other INS-like domains is unknown, and their phylogenetic distribution has not been comprehensively investigated. Here, we perform a global phylogenetic analysis to better understand the distribution of INSlike proteins in bacteria and to gain potential insights into their functional roles. This analysis revealed that a significant number of bacterial species lack INS in the context of ProRS, but contain a wide variety of alternative combinations of INS-like proteins. Caulobacter crescentus was chosen for experimental analysis as this organism encodes a ProRS with a severely truncated INS, in addition to YbaK and ProXp-ala. We show that a novel triple-sieve editing mechanism that makes exclusive use of trans-editing factors for the deacylation of Ala-and Cys-tRNA Pro exists in this species. Moreover, C. crescentus ProXpala recognizes elements in the acceptor stem of tRNA Pro and collaborates with elongation factor Tu (EF-Tu) to avoid hydrolysis of correctly charged Ala-tRNA Ala . Enzyme Preparation-C. crescentus proS, ybaK, and prdX (NCBI GenBank accession numbers: NP420738, NP_419779, and NP_418930, respectively) genes encoding ProRS, YbaK, and ProXp-ala, respectively, were PCR-amplified from C. crescentus CB15 genomic DNA (provided by Dr. Kenneth Keiler, Pennsylvania State University) using primers that included flanking restriction sites for BamHI and NdeI endonucleases (New England Biolabs). Each gene was cloned into pCR2.1-TOPO (Invitrogen) and later subcloned into pET15b (Novagen). Protein expression was carried out in E. coli BL21 (DE3) RIL cells. C. crescentus ProRS and ProXp-ala overexpression was induced with 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside for ϳ12 h at room temperature. C. crescentus YbaK overexpression was induced with 1 mM ␤-D-1-thiogalactopyranoside for 4 h at 37°C. The histidine-tagged proteins were purified using HIS-select nickel resin (Sigma-Aldrich). Wild-type (WT) E. coli alanyl-tRNA synthetase (AlaRS) (16), WT ProRS (17), E. coli K279A ProRS (17), E. coli EF-Tu (18), and E. coli tRNA nucleotidyltransferase (19) were prepared as described previously. Concentrations of E. coli AlaRS, E. coli K279A ProRS, C. crescentus YbaK, C. crescentus ProXp-ala, E. coli EF-Tu, and E. coli tRNA nucleotidyltransferase were determined by the Bradford assay (20). The concentrations of C. crescentus and E. coli ProRS were determined by active site titration (21).
Aminoacylation Assays-Aminoacylation reactions to determine kinetic parameters (for the amino acid) were performed in buffer A at 37°C using trace amounts of 3Ј-32 P-labeled tRNA, 20 M unlabeled tRNA, and variable concentrations of alanine (10 -300 mM) or proline (0.01-1 mM). Reactions were initiated by addition of C. crescentus ProRS to a final concentration of 0.1 M for proline charging or 0.5 M for alanine charging. Alanine mischarging of tRNA Pro by C. crescentus ProRS was carried out in buffer A using 1 M ProRS, 5 M E. coli tRNA Pro , and 250 mM alanine. E. coli AlaRS charging of tRNA Ala with alanine was accomplished with 50 nM AlaRS, 5 M E. coli tRNA Ala , and 50 M alanine. Reactions were initiated by addition of 4 mM ATP and incubated at 37°C. In some reactions, 1 M C. crescentus PrdX and 10 M E. coli EF-Tu were included. Prior to use in aminoacylation assays EF-Tu was activated as reported previously (24). Briefly, purified E. coli EF-Tu-GDP (10 M) was incubated for 1 h at 37°C in buffer containing 50 mM HEPES, pH 7.5, 1 mM dithiothreitol, 68 mM KCl, 6.7 mM MgCl 2 , 0.5 mM GTP, 2.5 mM phosphoenolpyruvate, and 30 g/ml pyruvate kinase. Graphs for all assays were prepared using SigmaPlot (Systat Software, San Jose, CA) with error bars representing the S.D. of triplicate data.
ATP:PP i Exchange Assays-ATP:PP i exchange assays were carried out using published conditions (8) with the following amino acid concentrations: 0.025-5 mM proline, 0.1-15 mM cysteine, and 50 -1000 mM alanine. C. crescentus ProRS (10 nM) was used for proline activation, and 250 nM enzyme was used for cysteine and alanine activation.
Pre-transfer Editing Assays-ATP hydrolysis assays were performed as described previously (8). The C. crescentus ProRS concentration was 0.5 M, and the amino acid concentrations were as follows: 3 mM proline, 3 mM cysteine, and 500 mM alanine. Assays were performed both in the absence and presence of 10 M E. coli tRNA Pro .
Bioinformatics-Bacterial organisms with partial or complete sequenced genomes (Ͼ2500) were used to identify ProRS genes (NCBI GenBank). Sequence alignments using ClustalW2 (25) were carried out to categorize the 1657 ProRS genes identified into one of three classes depending on whether they (i) contain a full-length INS, (ii) lack INS but instead contain a distinct C-terminal extension (C-term), or (iii) contain a truncated INS domain (mini-INS) (supplemental Fig. 1). In cases where multiple strains existed in the database, redundant sequences were removed. These 1657 organisms were then used to identify INS-like genes using a cut-off of 25% sequence identity for gene classification. The candidate INS-like genes were aligned using ClustalW2 with sequences representing the previously reported INS homologs: Haemophilus influenzae, Rhodopseudomonas palustris, C. crescentus, and E. coli YbaK; Aeropyrum pernix, R. palustris, C. sticklandii, and Thermus thermophilus ProX; C. sticklandii, Homo sapiens, C. crescentus, and Agrobacterium tumefaciens PrdX; E. coli, Bordetella avium, and Streptomyces griseus YeaK; and Bordetella parapertussis, Streptomyces albus, and Frankia alni PA2301 (9, 13, 26 -29). Based on homology to ProRS and known function, we have renamed PrdX as ProXp-ala, as mentioned above. We also renamed ProX, YeaK, and PA2301, homologs of unknown function, as ProXp-x, ProXp-y, and ProXp-z, respectively.
The resulting phylogenetic tree was constructed on the phylogeny.fr platform (30) by aligning 120 sequences (20 sequences from each of the six subfamilies) using MUSCLE 3.7. MultiSeq 2.0 included in the VMD software was used to isolate the INS domain sequences from full-length of ProRS (31). The tree was generated using PhyML (32) using default settings. Branch sup-port was provided by bootstraps of 100 replicates. The tree was displayed and edited using the "interactive Tree Of Life" (iTOL) tool (33).

RESULTS
Bacterial ProRSs and the INS Superfamily-ProRSs are class II synthetases that all contain the consensus motifs 1, 2, and 3 that constitute the core catalytic domain. However, several architectures for "extra" domains have been identified for ProRSs throughout the three domains of life (34,35). In bacteria, three distinct ProRS species have been previously reported. The predominant form of bacterial ProRS contains a full-length INS editing domain inserted between motifs 2 and 3. A subset of bacteria contain a severely truncated INS (mini-INS) domain that lacks catalytic activity (supplemental Fig. 1) (6). Finally, some bacterial species lack an INS domain altogether, and instead contain a smaller C-terminal extension (C-term) with a distinct architecture. The function of the C-terminal domain is believed to be structural (36,37). To determine the distribution of the different architectures in bacteria, we gathered ProRS sequence information from 1657 bacteria with partially or complete sequenced genomes and used multiple sequence alignments to classify a particular ProRS into one of the three groups (INS, mini-INS, or C-term). Based on this analysis, 64% of bacteria encode an INS-containing ProRS, 10% have a mini-INS, and 22% possess a C-term domain (supplemental Table 1). Interestingly, mini-INS ProRSs are mostly found in organisms from the ␣-proteobacteria phylum (and some ⑀and ␤-proteobacteria). In contrast, the C-term and INS ProRSs are widely distributed in the rest of bacterial phyla.
Our phylogenetic analysis strongly supports the existence of six distinct INS-like protein domains (Fig. 1). INS and ProXpala sequences cluster together in accordance with their alaninespecific editing activities. In contrast, ProXp-ala is more distantly related to ProXp-x, ProXp-y, and ProXp-z, which is in agreement with the lack of robust Ala-tRNA Pro editing by the latter enzymes from R. palustris, E. coli, and B. parapertussis, respectively. 3 We found that 64% of bacteria that lack INS but instead encode a mini-INS-or C-term-containing ProRS, also encode ProXp-ala. Therefore, although the correlation is high, the occurrence of ProXp-ala does not strictly correlate with the absence of INS, suggesting the possibility of functional redundancy. The well studied Cys-tRNA deacylase, YbaK, is the most abundant INS-like domain, present in 57% of bacteria. The most common combination of domains is the INS/YbaK pairing, which occurs in 25% of bacterial genomes. However, a wide variety of other combinations are also observed, as summarized in supplemental Table 2. Interestingly, 55 organisms encode two ProRS genes, one with an INS and one lacking INS. Fifteen percent lack an INS or INS-like domain altogether, suggesting that they do not require tRNA Pro proofreading or have evolved an alternative pathway for that function. Sixty-seven organisms that lack INS encode both YbaK and ProXp-ala (and in some cases ProXp-x as well), suggesting that they may rely exclusively on trans-editing factors to avoid proline mistranslation. To test this hypothesis, we chose to study the ␣-proteobacterium C. crescentus, which encodes a mini-INS, YbaK, and ProXp-ala.
Amino Acid Specificity and Misacylation by C. crescentus ProRS-ATP-PP i exchange experiments were carried out to probe the amino acid specificity of C. crescentus ProRS. A ϳ3-fold higher K m was measured for cysteine relative to cognate proline (Table 1). This observation is similar to previous reports using E. coli and other bacterial ProRSs (6,8). However, the turnover number (k cat ) for cysteine activation is ϳ240-fold lower relative to cognate proline. Thus, overall, the relative efficiency of cysteine activation by C. crescentus ProRS is ϳ780fold reduced relative to proline. The K m for alanine is substantially higher than that of proline (Ͼ1000 mM), and overall alanine activation by C. crescentus ProRS is at least ϳ10 4 -fold less efficient relative to proline.
The activation levels measured for cysteine are above the threshold of 1/3000 where editing is predicted to be required (38). Despite the relatively low alanine activation by C. crescentus ProRS, the parameters reported here are similar to those previously reported for E. coli ProRS (7). Although to our knowledge the relative amino acid levels in C. crescentus have not been reported, when taking into account the relative concentration of proline and alanine in E. coli cells (39), the "effective discrimination factor" in E. coli is reduced to 1200 (7), and this value is expected to be similarly lowered in C. crescentus. In vitro, C. crescentus ProRS mischarges alanine onto tRNA Pro to levels comparable with the post-transfer editing-deficient E. coli K279A ProRS ( Fig. 2A). In contrast, the robust posttransfer editing activity of WT E. coli ProRS prevents accumulation of Ala-tRNA Pro (Fig. 2A).
To further understand the mischarging capabilities of C. crescentus ProRS, we determined the steady-state parameters for the amino acids in the overall aminoacylation reaction. 32 P-Labeled tRNA Pro was used in these assays, which allowed the use of saturating amino acid concentrations (23). In vitro transcribed E. coli tRNA Pro/UGG , which is 82% identical to C. crescentus tRNA Pro/UGG (40) and contains all of the elements essential for aminoacylation by ProRS (41), was used in these assays. Aminoacylation data indicate that C. crescentus ProRS has an elevated K m for alanine compared with proline (ϳ5000-fold difference), but similar k cat values (only ϳ2-fold reduced). Thus, the overall in vitro discrimination factor for alanine compared with proline in this assay is ϳ10 3 ( Table 1). Overall, our results indicate that C. crescentus ProRS can mischarge tRNA Pro with both cysteine and alanine to levels where proofreading activity is likely to be necessary to prevent mistranslation. Pre-and Post-transfer Editing by C. crescentus ProRS-We next investigated the pre-and post-transfer editing activities of C. crescentus ProRS. Accumulation of AMP in the presence of ATP and amino acid reflects the presence of pre-transfer editing activity. In the case of E. coli ProRS, this activity is independent of the presence of tRNA (8). Therefore, we monitored the formation of AMP in the absence of tRNA in reactions containing proline, cysteine, or alanine. As expected, no accumulation of AMP is observed in reactions containing cognate proline (Fig. 2B). Similarly, cysteine does not stimulate significant levels of ATP hydrolysis. In contrast, substantial ATP hydrolysis is observed when alanine is present. As previously observed for E. coli ProRS (8), this activity was not stimulated by the addition of tRNA Pro (data not shown). These results demonstrate that C. crescentus ProRS possesses pre-transfer editing activity toward alanine but not cysteine.
Direct measurement of post-transfer editing activity is possible using deacylation assays. Cys-and Ala-tRNA Pro were prepared using post-transfer editing-deficient E. coli K279A ProRS. Deacylation assays demonstrate that C. crescentus ProRS does not possess post-transfer editing activity, consistent with the presence of a mini-INS domain that lacks catalytic function (Fig. 3). Interestingly, even though C. crescentus ProRS can hydrolyze alanyl-AMP via pre-transfer editing, this proofreading mechanism is insufficient to prevent formation of Ala-tRNA Pro (Fig. 2A). This result is consistent with the fact that high levels of alanine mischarging are observed for E. coli K279A ProRS, despite the fact that this enzyme still carries out pre-transfer editing (8). Taken together, our data indicate that C. crescentus ProRS cannot prevent formation of mispaired Ala-and Cys-tRNA Pro and is likely to rely on alternative mechanisms to clear these mischarged species.
C. crescentus YbaK and ProXp-ala Deacylation Activities-We next cloned, expressed, and purified C. crescentus YbaK and ProXp-ala and characterized their post-transfer editing activities in vitro. These experiments confirmed robust deacylation activity of C. crescentus YbaK and ProXp-ala for Cys-tRNA Pro and Ala-tRNA Pro , respectively (Fig. 3). Moreover, deacylation by ProXp-ala is not affected by the presence of C. crescentus ProRS (data not shown). These results support the formation of an alternative triple-sieve editing mechanism that operates in C. crescentus, which relies on the trans-editing activities of YbaK and ProXp-ala.
tRNA Specificity of C. crescentus ProXp-ala-A previous in vitro study indicated that C. sticklandii ProXp-ala deacylates correctly charged Ala-tRNA Ala , albeit with a reduced hydrolysis rate (12). We also observed weak Ala-tRNA Ala deacylation by C. crescentus ProXp-ala, with the rate of hydrolysis ϳ28fold reduced relative to Ala-tRNA Pro deacylation (Fig. 4A). Sequence comparison of bacterial tRNA Ala and tRNA Pro  reveals differences at the top of the acceptor stem domain (Fig.  4B) (40). Although both tRNAs share an A73 "discriminator" base, bacterial tRNA Pro has a unique C1:G72 base pair, which is important for ProRS recognition (41). In contrast, tRNA Ala contains a G1:C72 base pair at the end of the acceptor stem, and a unique G3:U70 wobble base pair, which is essential for AlaRS recognition (42,43). The G3:U70 base pair is not only critical for AlaRS aminoacylation, but also for editing by AlaRS, as well as by the trans-editing homolog, AlaXp (44). To establish whether C. crescentus ProXp-ala recognizes acceptor stem sequences, we transplanted the G1:C72 and U70 elements of tRNA Ala into the framework of tRNA Pro (Fig. 4B). The incorporation of these elements into tRNA Pro reduced the rate of ProXp-ala deacylation by 12-fold, a level similar to that of tRNA Ala (Fig. 4A). When only the G3:U70 base pair was introduced into tRNA Pro the rate of hydrolysis was reduced 4-fold.
These results indicate that ProXp-ala is sensitive to the identity of the first and third base pairs of tRNA Pro .

EF-Tu Discrimination against Mischarged
Ala-tRNA Pro -Despite a significant preference for the tRNA Pro acceptor stem, tRNA specificity of ProXp-ala may not be sufficient to prevent hydrolysis of Ala-tRNA Ala in vivo, as weak in vitro deacylation of the correctly charged tRNA Ala is observed (Fig. 4A). EF-Tu binds with a wide range of affinities to different aa-tRNAs (45). Dale and Ulhenbeck have shown that the operational binding affinity of EF-Tu depends on the individual thermodynamic contributions of the amino acid moiety and the tRNA. Consequently, a binding compensation occurs, and EF-Tu binds to correctly charged tRNAs with affinities that allow efficient delivery of aa-tRNAs to the ribosome. Because alanine and tRNA Pro are considered weak EF-Tu binders (46), we hypothesized that EF-Tu is likely to protect Ala-tRNA Ala , but not Ala-tRNA Pro from ProXp-ala hydrolysis. To test this hypothesis, we performed in vitro charging and mischarging assays in the presence of both ProXp-ala and E. coli EF-Tu. Addition of EF-Tu to a C. crescentus ProRS misacylation assay does not affect the overall synthesis of Ala-tRNA Pro (Fig. 5A). The  addition of C. crescentus ProXp-ala (equimolar with C. crescentus ProRS) decreases the formation of Ala-tRNA Pro by at least 20-fold both in the absence and presence of EF-Tu (Fig. 5A). In contrast, when cognate charging assays are performed with E. coli AlaRS, the concentration of Ala-tRNA Ala only decreases ϳ3-fold in the presence of a 20:1 ratio of ProXp-ala to AlaRS and E. coli EF-Tu promotes accumulation of Ala-tRNA Ala (by ϳ1.5-fold) (Fig. 5B). In good agreement with the thermodynamic compensation model, EF-Tu binds to Ala-tRNA Ala with higher affinity than Ala-tRNA Pro and can therefore protect a significant portion of Ala-tRNA Ala but not Ala-tRNA Pro from ProXp-ala hydrolysis (Fig. 5).

DISCUSSION
The wide distribution of INS-like proteins in bacteria highlights the importance of tRNA Pro proofreading in most bacteria. The overlapping distribution at the phylum level also suggests that the three INS-like domains of unknown function may have distinct substrate specificities. Indeed, the INS-like trans-editing factors studied experimentally to date, which include YbaK from H. influenza (10), E. coli (12,13), and C. crescentus (present work), and ProXp-ala from C. sticklandii (9,12), H. sapiens (29), and C. crescentus (present work), possess distinct substrate specificities and unique mechanistic features. For example, the mechanism of catalysis of YbaK involves thiol-specific chemistry, and thus, YbaK substrate specificity is not tunable (13). Although the detailed mechanism of ProXp-ala has not yet been explored, it has the same substrate specificity as INS, which uses a steric exclusion mechanism and water-mediated hydrolysis. Changes in substrate specificity can therefore be engineered through mutagenesis of specific residues within a tunable hydrophobic pocket (11). Whereas mutation of a highly conserved histidine residue in E. coli INS to alanine led to the hydrolysis of cognate Pro-tRNA Pro , mutation of the same histidine residue in human ProX (a member of the ProXp-ala subfamily defined here) did not alter the substrate specificity (29). These results suggest that although ProXp-ala and INS share Ala-tRNA specificity, the details of substrate binding and recognition may be different. Finally, the lack of robust Ala-or Cys-tRNA Pro deacylation by ProXp-x, ProXp-y, and ProXp-z 3 supports the presence of orthologous groups within the INS superfamily that correlate with well defined clades in the phylogenetic tree (Fig. 1).
In C. crescentus, the activities of YbaK and ProXp-ala establish a novel triple-sieve mechanism that exclusively relies on trans-editing domains for post-transfer editing. In this new model (Fig. 6), the aminoacylation domain of C. crescentus ProRS acts as a fine sieve to reject amino acids larger than proline allowing the formation of Pro-tRNA Pro , as well as noncognate Ala-and Cys-tRNA Pro (similar to the E. coli model (10)). In the second and third sieves, YbaK and ProXp-ala hydrolyze Cys-tRNA Pro and Ala-tRNA Pro in trans, respectively. The order in which aa-tRNA Pro species are proofread by YbaK and ProXpala is still unknown.
Because ProXp-ala and YbaK deacylate cognate Ala-tRNA Ala and Cys-tRNA Cys and both lack the tRNA binding domain of ProRS, these enzymes were proposed to act as general deacylases (12,29). However, we now show that the unique C1:G72 base pair of bacterial tRNA Pro contributes to efficient ProXp-ala hydrolysis. Nevertheless, ProXp-ala cannot effectively discriminate against Ala-tRNA Ala , and because the cellular concentration of Ala-tRNA Ala is likely to be higher than Ala-tRNA Pro a mechanism to prevent hydrolysis of Ala-tRNA Ala by ProXp-ala must exist. Our experiments show that the translational factor EF-Tu is responsible for preventing Ala-tRNA Ala deacylation by ProXp-ala. Based on these results, we propose that upon synthesis of Ala-tRNA Ala by AlaRS, EF-Tu can sequester Ala-tRNA Ala and prevent trans-editing by ProXp-ala, which cannot effectively compete against EF-Tu for binding to this substrate. In contrast, the tRNA specificity of ProXp-ala together with the weak affinity of EF-Tu for Ala-tRNA Pro prevents accumulation of Ala-tRNA Pro . Taken together, our data suggest that EF-Tu and ProXp-ala collaborate to prevent hydrolysis of Ala-tRNA Ala and ultimately mistranslation of proline.
The fact that ProXp-ala and INS coexist in some organisms suggests that the trans-editing function may serve as a redundant mechanism to prevent alanine-for-proline mistranslation. A similar scenario exists with AlaRS and AlaXp, which both prevent serine-for-alanine mistranslation (9,47). Alternatively, ProXp-ala may function to deacylate other Ala-tRNA species. Although we have shown that this enzyme has a preference for C1:G72 over G1:C72 in the context of tRNA Pro , it may tolerate 1:72 substitutions in the context of other tRNA frameworks.
Some ARSs such as human mitochondrial LeuRS (48), yeast and human mitochondrial PheRSs (49,50), mitochondrial and cytosolic eukaryotic ProRSs, archaeal ProRSs, and some bacterial ProRSs such as C. crescentus ProRS investigated here (51,52), have lost their editing function and are in many cases more selective at the initial amino acid activation step (48,(52)(53)(54). Nevertheless, transediting factors such as ProXp-ala are encoded in some genomes. The scattered distributions of the six INS-like domains in bacteria suggests lateral gene transfer as a mechanism for introducing editing activities into organisms that would benefit from the function of trans-editing domains, possibly in response to particular environmental conditions (14). As shown here, these domains combine in a species-specific manner and based on their overlapping distribution are likely to prevent accumulation of distinct mispaired aa-tRNA molecules. Characterization of the substrate specificities of ProXp-x, ProXp-y, and ProXp-z will be necessary to gain a complete understanding of the role of the INS superfamily in bacteria.