Flexizyme-aminoacylated shortened tRNAs demonstrate that only the aminoacylated acceptor arms of the two tRNA substrates are required for cyclodipeptide synthase activity.

Cyclodipeptide synthases (CDPSs) use two aminoacyl-tRNAs (AA-tRNAs) to catalyse cyclodipeptide formation in a ping-pong mechanism. Despite intense studies of these enzymes in past years, the tRNA regions of the two substrates required for CDPS activity are poorly documented, mainly because of two limitations. First, previously studied CDPSs use two identical AA-tRNAs to produce homocyclodipeptides, thus preventing the discriminative study of the binding of the two substrates. Second, the range of tRNA analogues that can be aminoacylated by aminoacyl-tRNA synthetases is limited. To overcome the limitations, we studied a new model CDPS that uses two different AA-tRNAs to produce an heterocyclodipeptide. We also developed a production pipeline for the production of purified shortened AA-tRNA analogues (AA-minitRNAs). This method combines the use of flexizymes to aminoacylate a diversity of minitRNAs and their subsequent purifications by anion-exchange chromatography. Finally, we were able to show that aminoacylated molecules mimicking the entire acceptor arms of tRNAs were as effective a substrate as entire AA-tRNAs, thereby demonstrating that the acceptor arms of the two substrates are the only parts of the tRNAs required for CDPS activity. The method developed in this study should greatly facilitate future investigations of the specificity of CDPSs and of other AA-tRNAs-utilizing enzymes.

INTRODUCTION

Cyclodipeptide synthases (CDPSs) constitute a family of tRNA-dependent enzymes involved in the biosynthesis of 2,5-diketopiperazines (2,5-DKPs) (1), a class of natural microbial products with therapeutically promising bioactivities, such as antibacterial, antifungal or antitumoral activity (2,3). CDPSs use two aminoacyl-tRNAs (AA-tRNAs), diverted from their canonical role in ribosomal protein synthesis, to catalyse the formation of various cyclodipeptides, the simplest members of the 2,5-DKP family (4).

CDPSs can be split phylogenetically into two subfamilies, named the NYH and XYP subfamilies according to the occurrence of two sets of conserved catalytic residues (5,6). Both subfamilies have been extensively studied (for reviews, see (1,7–9)). The catalytic mechanism used by AlbC from Streptomyces noursei, the first NYH CDPS to be identified, has been elucidated (10–12). It begins with the binding of the first substrate, Phe-tRNAPhe, followed by the transfer and covalent attachment of its phenylalanyl moiety to a conserved serine residue, resulting in the formation of a phenylalanyl-enzyme intermediate. The second substrate, Leu-tRNALeu, interacts with the phenylalanyl-enzyme, to which its leucyl moiety is transferred, generating a dipeptidyl-enzyme intermediate. Finally, the dipeptidyl moiety undergoes cyclization, with a conserved tyrosine residue acting as a proton relay, to yield the final cyclodipeptide cyclo(l-Phe-l-Leu) (cFL). The crystallographic structures of AlbC alone and in complex with a dipeptidyl intermediate analogue, and those of six other NYH or XYP CDPSs have been solved (10–11,13–15). Despite the very low levels of amino-acid sequence identity between CDPSs (below 10% in some cases), they share a common architecture, with a monomer built around a Rossmann fold domain that is very similar structurally to the catalytic domain of homodimeric class-Ic aminoacyl-tRNA synthetases (AARSs). The catalytic residues conserved throughout the CDPS family are located in the same position in all structures, strongly suggesting that all CDPSs, regardless of the subfamily to which they belong, act via similar mechanisms.

The two AA-tRNAs used successively as substrates by CDPSs are differently accommodated in the enzyme (11). The first substrate interacts with the enzyme through its aminoacyl moiety accommodated in a surface-accessible pocket named P1 and its tRNA moiety interacting either with basic residues at the surface of a conserved helix for NYH CDPSs (10,13–16) or basic residues present in the C- and N-terminal regions for XYP enzymes (17). The second substrate binds to the enzyme with its aminoacyl moiety accommodated in another pocket named P2 and the binding of its tRNA moiety is not documented (14,16). For each of the two substrates, the regions of the tRNA moieties essential for catalysis remains to be determined. In class-Ic AARSs, which have a conserved homodimeric quaternary structure, each monomer contains an anticodon-binding domain involved in recognizing the anticodon loop of a tRNA that is then acylated by the catalytic domain of the other monomer (18). CDPSs do not contain these anticodon-binding domains, almost all these enzymes are monomeric (1) and they use amino acids attached to the acceptor arm of tRNAs. They may not, therefore, be able to interact with the anticodon loop or another distal part of the tRNAs. Several studies had convergent results suggesting that the acceptors arms of the substrates were involved in the interaction with the enzyme (16,17). In order to further investigate the tRNA regions required for CDPS activity, an attractive strategy would involve the use of shortened tRNA analogues (minitRNAs), which have already proved valuable in the study of AARSs (19,20). Flexizymes, short AARS-like ribozymes, are a promising tool for the aminoacylation of diverse minitRNAs (21) but they have never been used for the study of AA-tRNA-utilizing enzymes. Flexizymes recognize their RNA acceptors by base-pairing with only three bases of their conserved 3′ termini. They can, therefore, accept a wide range of RNA acceptors as substrates: entire tRNAs, shortened tRNAs, such as microhelices (miHxs), and even tetramer oligonucleotides mimicking the 3′ NCCA tail of tRNA (21–23).

Here, we demonstrate using various flexizyme-aminoacylated minitRNAs (AA-minitRNAs) that the acceptor arms of the two substrates are the only parts of the tRNAs required for CDPS activity. To perform the study leading to this conclusion, we selected a new model CDPS that uses two different AA-tRNAs as substrates thus allowing to discriminate their bindings to the enzyme. We also produced relevant AA-minitRNAs thanks to the development of a new method combining the use of flexizymes with the separation of the AA-minitRNAs by anion-exchange chromatography. Finally, we set up an in vitro assay for the model CDPS, which allows to determine the behaviour of the CDPS towards its two AA-tRNA substrates and the produced AA-minitRNAs.

MATERIALS AND METHODS

Protein production

The His-tagged versions of T7 RNA polymerase, Escherichia coli AlaRS, E. coli GluRS and Nbra-CDPs were produced and purified according to published protocols ((24–26) and (4), respectively), modified slightly, as described in the detailed protocols in section I.A. of the Supplementary Data. The purified proteins were quantified by UV spectrophotometry with a DS-11 spectrophotometer (Denovix). Theoretical extinction coefficients were used for T7 RNA polymerase, E. coli AlaRS and E. coli GluRS. The experimental extinction coefficient of Nbra-CDPS was determined by acidic hydrolysis and amino-acid analysis (27).

Production of Escherichia coli tRNAs

Plasmids for the constitutive expression of tRNAAlaUGC, tRNAAlaGGC and tRNAGlu (see Supplementary Table S1 for a summary of the sequences of all the RNAs used in this study) were constructed by inserting the corresponding tRNA sequences into pBSTNAV2, which was kindly provided by Pr. Yves Mechulam (section I.B.ii. of Supplementary Data and Supplementary Table S2). We generated the tRNAs according to a protocol provided by the team of Pr. Yves Mechulam, as detailed in (28) (see section I.B.ii of Supplementary Data). We quantified the tRNAs and all other RNAs used in this study by UV spectrophotometry with a DS-11 spectrophotometer (Denovix), assuming that 1 OD260nm unit represented 40 ng/μl RNA.

Aminoacylation of tRNAs

The tRNAs were aminoacylated under the following conditions: 50 mM HEPES-KOH pH7.5, 20 mM KCl (only for tRNAAla), 15 mM MgCl2, 1 mM or 10 mM dithiothreitol (DTT) for tRNAGlu and tRNAAla, respectively, 4 μM adenosine triphosphate (ATP), 15 μM tRNAs, 2 μM of the corresponding E. coli AARS and 250 μM of the corresponding l-amino acid. Reactions were incubated at 37°C for 20 min and stopped by adding sodium acetate pH 5.0 to a final concentration of 100 mM. The AA-tRNAs and the remaining non-acylated tRNAs were purified on a 5 ml HiTrap Q HP anion-exchange chromatography column (GE Healthcare), with a linear gradient of 0 to 1 M NaCl in 10 mM sodium acetate pH 5.0. Fractions containing tRNAs were subjected to ethanol precipitation and the precipitate obtained was stored as a dried pellet at −20°C. The proportion of aminoacylated tRNAs was determined by enzymatic assays, with high concentrations (500 nM) of Nbra-CDPS (section II.B.iii. of Supplementary Data).

Production of flexizymes (dFx) and minitRNAs

Microhelices (miHxs) mimicking the acceptor arms of E. coli tRNAs are abbreviated as miHxYyy_X. Yyy refers to the amino acid charged on the tRNA mimicked, and X is the number of base pairs retained in the stem (so, for example, miHxAla_7 reproduces the entire acceptor arms of both tRNAAla isoacceptors, which are identical). MiHxAla_4, miHxGlu_7, miHxGlu_3 and tetramer oligos ACCA and GCCA were synthesized chemically by Eurogentec. MiHxAla_7 and dFx were synthesized by in vitro transcription and then purified by anion-exchange chromatography. In vitro transcription and treatment of the reactions were carried out using a protocol from the team of Prof. Suga (29), modified slightly to prevent the formation of template-independent transcripts (30) (see detailed protocol in section I.B.iv. of Supplementary Data). DNA matrices were obtained by mixing T7_F primer and the corresponding reverse primer (Supplementary Figure S1 and Table S3). After phenol extraction and removal of nucleotides and aborted transcripts using centrifugal filter, transcripts were renatured by heating at 80°C for 5 min and then rapidly decreasing the temperature to 4°C, before purification by HPLC on an Elite LaChrom HPLC system (VWR). Samples were loaded onto a DNAPAC-PA100 semi-preparative column (9 × 250 mm, 13.5 μm, Thermo Fisher Scienfitic). Separations were performed under non-denaturing conditions with an increasing linear gradient of buffer B (2.5 M ammonium acetate pH 5.2, 0.5% acetonitrile) in buffer A (25 mM ammonium acetate pH 5.2, 0.5% acetonitrile), at a flow rate of 5 ml/min (31). The gradient was optimized for each transcript, so a number of different gradients were used (Supplementary Table S4). Fractions of interest were desalted to 25 mM ammonium acetate pH 5.2 with a Hi-Trap 26-10 desalting column (GE Healthcare), freeze-dried and stored at −20°C. Purified transcripts were heat-renatured with the protocol described above and analysed on a DNAPAC-PA100 analytical column (4 × 250 mm, 13.5 μm, Thermo Fisher Scientific) with the same gradients, at a flow rate of 1 ml/min.

Production of AA-minitRNAs with dFx

MinitRNAs were aminoacylated by dFx with 3,5-dinitrobenzyl esters (DBE) of l-alanine or l-glutamate (Supplementary Figure S2). These chemically activated amino acids were synthesized according to standard procedures (22,32). Standard aminoacylation conditions were used (29): 25 μM minitRNA and 37.5 μM of the corresponding dFx were mixed in 50 mM HEPES-KOH pH 7.5 buffer, heated at 95°C for 2 min, and then allowed to cool at room temperature for 5 min. We added 600 mM MgCl2 and incubated the mixture at room temperature for 5 min and then on ice for 3 min. The reaction was initiated by adding the corresponding DBE-activated amino acid at 5 mM and incubating on ice for 1.75 h for DBE-activated alanine and 3 h for DBE-activated glutamate. Reactions were quenched by adding sodium acetate pH 5.2, subjected to ethanol precipitation and the RNA pellets were stored at −80°C. Pellets were dissolved in 20 mM sodium acetate pH 5.2 and the RNA was renatured by applying a gentle heating/quick cooling protocol: heating to 60°C for 30 s, followed by a rapid decrease in temperature to 5°C. All samples were purified and analysed on DNAPAC-PA100 columns as described above, with the gradients described in Supplementary Table S4.

Enzymatic assays

Dried pellets of the substrates to be tested, AA-tRNAs and AA-minitRNAs, were solubilized in 20 mM sodium acetate pH 5.2 shortly before use and kept on ice during all experiments, to limit deacylation. Enzymatic assays were performed in 100 mM phosphate-sodium buffer pH 7.5, 50 mM KCl, 15 mM MgCl2 and 0.1 mM β-mercaptoethanol. Substrates were added to the solution, which was then incubated for 1 min at 20°C, and reactions were initiated by adding 5 nM Nbra-CDPS. Aliquots were withdrawn after different times (see below), acidified with 2% trifluoroacetic acid (TFA) to stop the reaction and mixed with known concentrations of stable isotope-labelled internal standards, for quantification of the cAE and cAA produced in LC-MS/MS analyses, as previously described (16) (see section I.C. of Supplementary Data). The stable isotope-labelled standards, cAE and cAA, each containing one 13C3-15N-L-alanine residue, were chemically synthesized (see section I.C.i. of Supplementary Data, Supplementary Schemes S1 and 2, Figures S3 and 4). LC-MS analyses were performed on an Elute SP HPLC chain (Bruker Daltonik GmbH) equipped with a DAD detector coupled via a split system to an Amazon SL ion trap mass spectrometer (Bruker Daltonik GmbH) set in positive mode. Samples were injected onto a Hypercarb column (2 × 150 mm, 3 μm, 100 Å, ACE), at a flow rate of 0.2 ml/min, with a linear gradient from 2 to 42% solvent B at 0.2 ml/min over 20 min (solvent A: 0.1% (v/v) formic acid in H2O; solvent B: 0.1% (v/v) formic acid in acetonitrile/H2O (90/10)). Cyclodipeptide concentrations were calculated from the ratio of the EIC peak areas of the analyte and the standard compounds (section I.C.ii. of Supplementary Data and Supplementary Figure S5). For the determination of initial velocities, we collected three aliquots during the first 2 min of incubation and plotted the concentrations obtained at the three different sampling times (Supplementary Figure S10). For endpoint assays, the cyclodipeptides were quantified after incubation for 30 min with the appropriate substrates.

RESULTS

Selection of the model CDPS and its substrates

Choice of the CDPS

The study requires to discriminate the binding of the two substrates. Therefore, it should be carried out on a CDPS that uses two different substrates to produce an heterocyclodipeptide. This is not the case for the two CDPSs that have already been submitted to enzymological studies: Rv2275 uses Tyr-tRNATyr as its unique substrate to synthesize cyclo(l-Tyr-l-Tyr) (cYY) (13) and AlbC uses Phe-tRNAPhe more efficiently than any of the Leu-tRNALeu isoacceptors as the second substrate to mostly produces cFF (16). An attractive candidate for overcoming this issue is the recently identified CDPS from Nocardia brasiliensis, Nbra-CDPS. Upon its heterologous expression in E. coli, Nbra-CDPS uses Ala-tRNAAla and Glu-tRNAGlu as substrates to produce mostly cAE, only about 2% cAA and no cEE (5). Protein sequence analyses have established that the aminoacyl moiety of Ala-tRNAAla is accommodated in P1, whereas that of Glu-tRNAGlu is accommodated in the second pocket P2 (5,6). Ala-tRNAAla can also be accommodated in P2 but it is a very poor substrate, as demonstrated by the very small amounts of cAA produced. Nbra-CDPS therefore appears to be an appropriate candidate for use in studies attempting to elucidate the use of the two substrates.

Choice of the substrates

Since Nbra-CDPS is active in E. coli, we based our study on the use of E. coli tRNAs as substrates, as previously described for other characterized CDPSs (4,13,16). The E. coli genome encodes one tRNAGlu and two tRNAAla isoacceptors, tRNAAlaUGC and tRNAAlaGGC. Only six nucleotide differences, none of them in the acceptor arms, are found between the two alanine isoacceptors (Figure 1A). All three full-size tRNAs were obtained by overexpression in E. coli.

Figure 1.

Sequences of substrates and substrates analogues used in the study. (A) Secondary structures of Escherichia coli aminoacylated tRNAsAla and tRNAGlu. The differences between the two tRNAAla isoacceptors are shown in red. (B) Secondary structures of the minitRNAs used. Post-transcriptional modifications in full-size RNAs were taken from the MODOMICS database (52) and are abbreviated as follows: D: dihydrouridine; G7: 7-methylguanosine; V: uridine 5-oxyacetic acid; T: 5-methyluridine; ψ: pseudouridine; S: 5-methylaminomethyl-2-thiouridine; A2: 2-methyladenosine.

In order to investigate the regions of tRNAAla and tRNAGlu required for Nbra-CDPS activity, we designed a set of minitRNAs based on the sequences of E. coli tRNAs (Figure 1B). This set of minitRNAs includes tRNA miHxs, stem-loop RNAs in which the stems corresponded to the acceptor arms of the tRNAs closed by the tetramer loop 5′-UUCG-3′, which is known to favour the stability of short stems (33,34). The two tRNAAla isoacceptors of E. coli have similar acceptor arms and were, therefore, mimicked by the same miHxAla. We included miHxs with 7 bp stems that mimicked the entire acceptor arm, and the miHxs with the shortest stems predicted to be stable, miHxGlu_3 and miHxAla_4 (miHxAla_3 was excluded from the study because the corresponding hairpin conformation was not predicted to be stable) (Supplementary Table S5). Finally, we also included RNA tetramers mimicking the 3′ NCCA tail of the tRNAs. All minitRNAs were obtained by in vitro transcription or chemical synthesis (Supplementary Table S1).

Production of purified flexizyme-aminoacylated minitRNAs

The full procedure for producing AA-minitRNAs was developed with Ala-miHxAla_7. MiHxAla_7 was synthesized by in vitro transcription, purified by anion-exchange chromatography on a semi-preparative DNAPAC column, desalted and lyophilized (see ‘Materials and Methods’ section). Anion-exchange chromatography was performed under non-denaturing conditions, and miHxAla_7, like many small RNA hairpins, can fold into different intra- and intermolecular conformations (Supplementary Table S5). We ensured that the most stable intramolecular conformation was favoured by heating and rapidly cooling miHxAla_7 samples to ensure their renaturation before purification and analysis (35). Such renaturation protocol had a dramatic effect on HPLC profiles. For example, renaturation decreased the number of peaks observed after the injection of purified miHxAla_7 from two to one (Supplementary Figure S13). We alanylated miHxAla_7 with dFx under standard conditions. A rate of about 60% alanylation was achieved, as demonstrated by acid-PAGE (polyacrylamide gel electrophoresis) analysis (Supplementary Figure S14A). We developed a new procedure for purifying Ala-miHxAla_7 and separating it from residual miHxAla_7, based on the use of the DNAPAC column. We assumed that the change in charge following RNA aminoacylation (the primary α-amine function of amino acids is positively charged in neutral or slightly acidic conditions) would induce a different pattern of behaviour during anion exchange chromatography. A comparison of the chromatogram obtained for the alanylation miHxAla_7 with that obtained for purified miHxAla_7 revealed a new peak, with a shorter retention time (Supplementary Figure S14B). The corresponding eluted compound was purified and its characterization by acid-PAGE analysis and MALDI-TOF-MS showed it to be Ala-miHxAla_7 (Supplementary Figure S14A and C). Figure 2 sums up the various steps in the purification process, from the crude in vitro transcription reaction products to purified Ala-miHxAla_7. Ala-miHxAla_7 deacylation occurred during purification, but often at rates of less than 20%, as shown by the ratio of the areas of the peaks corresponding to Ala-miHxAla_7 and miHxAla_7 in purified Ala-miHxAla_7 (Figure 2D). The aminoacylation reaction and the purification process could easily be scaled up for the production of several tens of micrograms of Ala-miHxAla_7. Interestingly, the purification protocol also effectively separated and purified the dFx from the reaction mixture (Supplementary Figure S14B). It was therefore possible to purify the dFx and to reuse them several times, with no loss of efficiency.

Figure 2.

Purification of Ala-miHxAla_7. Chromatograms corresponding to the injection onto the DNAPAC column of (A) 20 μg crude in vitro transcription mixture for miHxAla_7, (B) 1 μg purified miHxAla_7, (C) the alanylation of 1 μg miHxAla_7 with a dFx and (D) 0.25 μg purified Ala-miHxAla_7.

We performed a similar experiment for the production of Glu-miHxGlu_7. Chemically synthesized miHxGlu_7 was renatured and then glutamylated with dFx. Aminoacylation was analysed on the DNAPAC column. The chromatogram obtained was similar to that for Ala-miHxAla_7, with the presence of an additional peak relative to the chromatogram of miHxGlu_7, with a shorter retention time than the peak corresponding to miHxGlu_7 (Supplementary Figure S15A). MALDI-TOF confirmed that the eluted compound was Glu-miHxGlu_7 (Supplementary Figure S15B). Despite the separation being performed with similar gradients, the glutamylation of miHxGlu_7 resulted in a lower shift in retention time (about 2.8 min) than the alanylation of miHxAla_7 (about 5 min). This may be due to the negative charge on the side chain of glutamate at pH 5.5, partly compensating for the effect of the additional positive charge on the backbone amine group.

Similar experiments were performed with smaller minitRNAs, which were also retained on the DNAPAC column. The length of the minitRNA conditioned its retention time, with shorter minitRNAs having shorter retention times (Supplementary Figure S16). Anion-exchange chromatography was used to analyse the flexizyme aminoacylation reactions for minitRNAs. Additional peaks corresponding to AA-minitRNAs were observed. Aminoacylation yields were found to be about 60% for minitRNAsAla and 45% for minitRNAsGlu, consistent with the results of Prof. Suga and coworkers (22,23). By scaling up the reactions, we were able to obtain several nmoles of aminoacylated minitRNAs. The identity of the aminoacylated products was confirmed by MALDI-TOF (Supplementary Figure S17). Deacylation during the purification process did not exceed 20% for alanylated minitRNAs and 10% for glutamylated minitRNAs, as shown by analytical anion-exchange chromatography.

The anion-exchange chromatography method presented here was also used for time-course analyses of aminoacylation by dFx and to determine optimal reaction times. These optimal reaction times were found to be 1.75 h for minitRNAAla and 3 h for minitRNAGlu(Supplementary Figure S18), which is coherent with previously reported data.

Principle and optimization of the enzymatic assay

Previous enzymological studies of CDPSs used coupled CDPS/AARS assays, in which E. coli AA-tRNAs are generated in situ by E. coli AARSs (10,16). This approach is unsuitable for studies of minitRNAs, most of which are not AARS substrates. We therefore developed a direct assay, in which tRNAs and their minitRNA analogues were aminoacylated and purified before their testing as substrates for cyclodipeptide production by NBra-CDPS. Standard procedures were used for the AARS-based aminoacylation of full-size tRNAs and their purification. For minitRNAs, the new method of production described above was used.

Careful optimization of the various components and parameters of the assay was required to maximize the cyclodipeptide-synthesizing activity of Nbra-CDPS and to ensure that the spontaneous hydrolysis of the aminoacyl-linkage of the substrates to tRNAs during the assays could be neglected. Preliminary results suggested that the activity of Nbra-CDPS was similar for the two tRNAAla isoacceptors. We therefore optimized the in vitro assay with Ala-tRNAAlaUGC before performing accurate comparison between both isoacceptors. The nature of buffer greatly affected Nbra-CDPS activity. The two buffers previously used in enzymatic assays, Tris–HCl and HEPES-KOH (10,16), strongly inhibit Nbra-CDPS activity (Supplementary Figure S6). The observed inhibition may be due to the binding of HEPES in the P1 pocket of Nbra-CDPS, as previously observed for CHES and CAPSO in the crystal structure of the CDPS YvmC (14). Sodium phosphate was selected as a more appropriate buffer in this context. Nbra-CDPS activity significantly increased as pH increased from 7.00 to 7.75 (Supplementary Figure S7A). However,the spontaneous deacylation of the substrates is also favoured by increasing pH (36). We limited this reaction by performing the assay at pH 7.50. AA-tRNA deacylation is also favoured by increasing temperature (37), and the activity of Nbra-CDPS is similar at 20°C and 30°C (Supplementary Figure S7B). We therefore performed reactions at 20°C. Finally, 5 nM Nbra-CDPS was shown to be a convenient concentration for the measurement of initial reaction rates (Supplementary Figure S8). In these conditions, the maximal duration of the assay required for the determination of initial reaction rates was 2 min. In order to confirm that the deacylation occurring over such time ranges is negligible, we used the chromatographic method described above to study the spontaneous deacylation of Ala-miHxAla_7 and Glu-miHxGlu_7 in the conditions of the cyclodipeptide production assay. We found the half-life of these products to be 60 and 145 min, respectively (Supplementary Figure S19 and Table S6), consistent with previous reports showing alanylated-tRNAs to be the least stable AA-tRNAs (32). Based on these findings, we estimated that spontaneous deacylation rates were limited to 2.3% for alanylated substrates and 1.0% for glutamylated substrates after 2 min (maximal duration of the assays used to calculate initial reaction rates) (Supplementary Table S6).

Enzymatic characterization of Nbra-CDPS using its two AA-tRNA substrates

We began by characterizing the behaviour of the first substrate of Nbra-CDPS, Ala-tRNAAla, by measuring the initial rate of reaction for a fixed concentration of Glu-tRNAGlu and a variable concentration of Ala-tRNAAla. We obtained similar results for the two isoacceptors (Supplementary Figure S9), suggesting that the six nucleotides in the TψC stem and the anticodon stem-loop that differ between these isoacceptors (Figure 1A) are not involved in substrate recognition. We therefore performed all subsequent experiments with Ala-tRNAAlaUGC. We measured cyclodipeptide production with 600 nM Glu-tRNAGlu and various concentrations of Ala-tRNAAlaUGC, ranging from 300 nM to 12 μM (Figure 3A and Supplementary Figure S10). For concentrations of Ala-tRNAAlaUGC up to about 4.8 μM, the initial rate of cAE production increased with increasing substrate concentrations, and almost no cAA production was detected. At higher Ala-tRNAAlaUGC concentrations, saturable rates of cAE production appeared to be achieved, followed by a decrease in rate with increasing substrate concentrations. The rate of concomitant cAA production was low, and increased with increasing substrate concentration.

Figure 3.

Kinetic studies of the interaction of Nbra-CDPS with its AA-tRNA substrates. Initial rates of cAE (dark blue) and cAA (light blue) production were measured with (A) 600 nM Glu-tRNAGlu and various concentrations of Ala-tRNAAlaUGC and (B) 1200 nM Ala-tRNAAlaUGC and various concentrations of Glu-tRNAGlu. Error bars indicate the standard error between triplicates. The cyclodipeptide titration curves used to determine initial rates are shown in Supplementary Figure S10.

Similar experiments were performed to characterize the second substrate, with Ala-tRNAAlaUGC at a constant concentration of 600 nM and various concentrations of Glu-tRNAGlu. However, saturable rates of cAE production were observed regardless of the concentration of Glu-tRNAGlu used. We therefore tested lower concentrations of Glu-tRNAGlu with a higher fixed concentration of Ala-tRNAAlaUGC (1200 nM) to increase cyclodipeptide production and facilitate detection (Figure 3B and Supplementary Figure S10). Up until a concentration of about 600 nM Glu-tRNAGlu, the initial rate of cAE production increased and the rate of cAA production decreased. At higher Glu-tRNAGlu concentrations, saturable rates were observed, followed by a decrease in the rate of production for cAE and cAA production was not detected.

Taken together, the results clearly demonstrate that competition and/or inhibition occur at the two binding sites. When Ala-tRNAAla is present in large excess, it replaces Glu-tRNAGlu as the second substrate. However, cAA production requires a large molar excess of Ala-tRNAAla relative to Glu-tRNAGlu (at least 4-fold) and cAA levels remain much lower than cAE levels (Figure 3), confirming the strong preference of the enzyme for Glu-tRNAGlu as the second substrate. When Glu-tRNAGlu is present in large excess, a decrease in cAE production is observed, suggesting that Glu-tRNAGlu may act as an inhibitor at the first or another binding site (Figure 3B). These results reflect the complexity of the reaction catalysed by Nbra-CDPS, making it difficult to determine kinetic parameters accurately. However, the results seem to indicate a large difference in the affinity of Nbra-CDPS for its two substrates, with maximal initial rates reached at a concentration of about 5000 nM for Ala-tRNAAlaUGC and about 600 nM for Glu-tRNAGlu. In both cases, the ‘pre-saturation’ increase in initial rates could be modelled with a Michaelis–Menten equation, yielding apparent KM values of 930 ± 114 nM for Ala-tRNAAlaUGC and 93 ± 8 nM for Glu-tRNAGlu (Supplementary Figure S11). This suggests that the affinity of the second binding site for Glu-tRNAGlu is greater than that of the first binding site for Ala-tRNAAlaUGC. Apparent kcat values for Nbra-CDPS were estimated at 0.33 s−1 for Ala-tRNAAlaUGC and 0.25 s−1 for Glu-tRNAGlu (Supplementary Figure S11), although it is clear that there is a risk of these values being substantially underestimated due to the observed competition and/or inhibition. Apparent catalytic efficiencies were estimated by calculating kcatapp/KMapp ratios. We obtained values of 0.4 × 106 M-1.s−1 for Ala-tRNAAla as the first substrate and 2.7 × 106 M-1.s−1 for Glu-tRNAGlu as the second substrate.

This study using entire AA-tRNAs confirmed that Nbra-CDPS was a relevant model for discriminative studies of the use of two substrates and set reference values for the comparative study of shortened substrates.

Enzymatic characterization of Nbra-CDPS using AA-mini-tRNAs as substrates

We first investigated the use, by Nbra-CDPS, of Ala-miHxAla_7 and Glu-miHxGlu_7, which mimic the entire acceptor stems of tRNAAla and tRNAGlu, with an approach similar to that used for AA-tRNA substrates (i.e. by measuring the initial rate of cyclodipeptide production for a range of concentrations of Ala-miHxAla_7 in the presence of 600 nM Glu-tRNAGlu or for a range of concentrations of Glu-miHxGlu_7 in the presence of 1200 nM Ala-tRNAAlaUGC). The results were compared with those obtained with the corresponding AA-tRNAs (Figure 4A and B). The initial rates of cAE production obtained with Ala-miHxAla_7 were very similar to those obtained with Ala-tRNAAlaUGC (Figure 4A), indicating that Ala-miHxAla_7 was as good a substrate as Ala-tRNAAla at the first binding site. As for Ala-tRNAAlaUGC, no cAA was produced at the concentrations tested (Supplementary Figure S12). With Glu-miHxGlu_7, the initial rates of reaction was also close to that for Glu-tRNAGlu. When using low and increasing concentrations of Glu-miHxGlu_7, the increase in cAE production (Figure 4B) and the decrease in cAA production (Supplementary Figure S12) was slightly faster than with Glu-tRNAGlu. When used at high concentrations, Glu-miHxGlu_7 inhibited cAE production even more strongly than Glu-tRNAGlu. This suggests that Glu-miHxGlu_7 is at least as good a substrate as Glu-tRNAGlu and that it competes more efficiently than Glu-tRNAGlu with Ala-tRNAAla for the second binding site.

Figure 4.

Kinetic study of the interaction of Nbra-CDPS with AA-minitRNA substrates. Initial rates of cAE production were measured with (A) 600 nM Glu-tRNAGlu and various concentrations of Ala-minitRNAAla and (B) 1200 nM Ala-tRNAAlaUGC and various concentrations of Glu-minitRNAGlu. End-point assays were performed with (C) 600 nM Glu-tRNAGlu and 600 nM Ala-minitRNAAla and (D) 1200 nM Ala-tRNAAlaUGC and 600 nM Glu-minitRNAGlu. Error bars indicate the standard errors between duplicates. For the sake of clarity, cAA production is not indicated on Figure 4. Data for cAA production are presented in Supplementary Figure S12.

We then performed similar experiments using aminoacylated tetramer oligonucleotides mimicking the 3′ NCCA of tRNAs (Figure 1B). No cAE production was detected with Ala-ACCA and Glu-tRNAGlu or Glu-GCCA and Ala-tRNAAlaUGC (Figure 4A and B). The 3′ tails of the tRNAs are, therefore, clearly not sufficient for use as substrates by Nbra-CDPS in the conditions of our assay. Moreover, as the low amount of cAA remains constant when Ala-tRNAAla is used with increasing concentrations of Glu-GCCA (Figure 4B and Supplementary Figure S12B), this analogue is not an efficient competitor for the second binding site of the enzyme.

The in vitro assay do not appear to be suitable for the characterization of very poor substrates, because of the detection threshold for cyclodipeptide production (section I.C.ii of Supplementary Data). We therefore used an endpoint assay, in which cyclodipeptide production was quantified after incubation for 30 min with the substrates (it should be noted that after 30 min incubation in the conditions of the assays deacylation of the substrate could reach up to 29.3 and 13.3%, for alanylated and glutamylated substrates, respectively (Supplementary Table S6)). Upon incubation of 600 nM Ala-ACCA with 600 nM Glu-tRNAGlu or of 600 nM Glu-GCCA with 1200 nM Ala-tRNAAlaUGC, no cAE production was detected, confirming that these aminoacylated 4-mer cannot act efficiently as substrates of Nbra-CDPS (Figure 4C and D). Upon incubation of Glu-GCCA, cAA production was detected (Supplementary Figure S12D), as might be expected for an inefficient competitor of Ala-tRNAAla at the second biding site.

Finally, we used the same endpoint assay to evaluate the activity of Nbra-CDPS with the miHxs with the shortest stable stems. We found that Ala-miHxAla_4 could act as a substrate for Nbra-CDPS, but that it was a much poorer substrate than Ala-miHxAla_7 (Figure 4C). Similarly, Glu-miHxGlu_3 was found to act as a substrate of Nbra-CDPS, but much less cAE was produced than with Glu-miHxGlu_7 (Figure 4D).

DISCUSSION

We show here that Nbra-CDPS is a relevant model for studies aiming to discriminate between the substrates used at the two binding sites of CDPSs. It was not possible to determine kinetic parameters accurately, due to the apparent cross-inhibition or cross-competition of substrates and probable changes in the conformation of the CDPS between its free, aminoacyl and dipeptidyl states. However, our data suggest that Nbra-CDPS has significantly different affinities for its two substrates: the affinity of the first binding site for Ala-tRNAAla appears to be lower than that of the second binding site for Glu-tRNAGlu. Only two kinetic studies on CDPSs have been published, both focused on the use of Tyr-tRNATyr by the cYY-producing CDPS Rv2275 (13,38). However, the authors of both studies used a one-substrate model to estimate these kinetic parameters, which is incompatible with a ping-pong mechanism involving two substrates at two different binding sites, making it impossible to discriminate between these two sites for the purposes of characterization.

The feasibility of direct in vitro CDPS assays combined with AA-minitRNA production by flexizymes made it possible to investigate the use of various substrates modified in the RNA part at each of the two binding sites. We used this approach to improve our understanding of the regions of the two tRNAs required for CDPS activity. We were primarily interested in determining the role of the acceptor arms of tRNAs, as we previously demonstrated the involvement of this region in the interaction with CDPSs (16,17). We demonstrated that AA-miHxs, which mimic the entire 7 bp stem of tRNAAla and tRNAGlu acceptor arms, are at least as good as the corresponding AA-tRNAs in terms of their ability to act as substrates for Nbra-CDPS. By contrast, no activity was observed when the substrates tested were limited to the aminoacylated 3′ tails. The tRNA acceptor arms are, therefore, required at the two binding sites for Nbra-CDPS activity, whereas the three stem-loops of tRNAs are fully dispensable. We explored the minimal substrates of CDPSs further, by focusing on the smallest stable microhelices, Ala-miHxAla_4 and Glu-miHxGlu_3. We found that these microhelices were poorer substrates than the corresponding entire AA-miHxs, suggesting that Nbra-CDPS may require distal nucleotides from both acceptor arms. However, one major concern with these shorter miHxs is that the tetramer loops introduced to close the miHxs may hamper the interaction with the binding sites of the enzymes by causing artificial steric hindrance. The promising results obtained in this study suggest that flexizyme technology will pave the way for further biochemical and structural investigation of CDPS specificity and for elucidating the specificity determinants of CDPS and other enzymes using AA-tRNAs as substrates.

Other families of enzymes have evolved the ability to hijack AA-tRNAs from their canonical role for various types of reaction (1). In particular, three families of enzymes and their mode of interaction with their AA-tRNA substrates have been investigated in detail: class I lanthipeptide dehydratases (LanBs), which use Glu-tRNAGlu for the successive glutamylation and dehydrogenation of serine and threonine residues in ribosomally synthesized and post-translationally modified peptides (39), Fem transferases X (FemX), which transfer the alanyl moiety of Ala-tRNAAla to the UDP-MurNAc-pentapeptide during peptidoglycan biosynthesis (40), and L/F transferases, which transfer the leucyl moiety of Leu-tRNALeu (and, to a lesser extent, the phenylalanyl moiety of Phe-tRNAPhe) to the N-terminus of proteins as a label for targeted degradation (41). Interestingly, despite their differences in substrate specificity and function, LanBs, FemX and L/F transferases have all been shown, via various experimental strategies, to interact with the acceptor arm of the tRNA moieties of their substrates.

A first approach to investigating the tRNA regioselectivity of AA-tRNA-utilizing enzymes is based on the use of mutant tRNAs that are acylated by AARSs. Mutations are introduced into the tRNAs, which are then assayed as substrates to assess the effect of the mutations on their activity. An effect of the mutation is interpreted as indicating that the mutated nucleotide is involved in recognition by the enzyme. This approach has been used to show that L/F transferases and FemX interact with up to the first five and four base pairs, respectively, of the acceptor arms of their substrates (42,43). Similar studies have also demonstrated that the LanB-like MibB interacts specifically with the discriminator base A73 and the nucleotide U72 within the Glu-tRNAGlu (44), and that the CDPS AlbC interacts with the first base pair of its second AA-tRNA substrate (16). However, methods based on the use of AARSs for aminoacylation are intrinsically limited in terms of the range of RNA acceptors that can be tested, because the specificity determinants required for aminoacylation must be conserved (45). One interesting way to get around this limitation is to synthesize AA-tRNA analogues by chemoenzymatic methods, through the enzymatic ligation of aminoacylated dinucleotide to tRNA analogues lacking two nucleotides at their 3′ end (46,47). This strategy makes it possible to aminoacylate a much broader range of tRNA-like molecules than AARS-based methods, including shortened tRNA analogues, such as miHx or tRNAs, with mutations of AARS specificity determinants. This approach was used to generate various analogues of Ala-tRNAAla for the study of FemX, including shortened analogues, such as Ala-miHxAla, which was shown to be a good substrate of FemX, and various mutated miHxs, which were used to elucidate the role of the nucleotide sequence in determining specificity (31). It would be particularly interesting to apply this strategy to other AA-tRNA-utilizing enzymes, including CDPSs in particular. However, it necessitates the use of 2′-deoxycytidine as the penultimate nucleotide, and preliminary results with such molecules (kindly prepared by our collaborators M. Fonvielle, M. Ethève-Quelquejeu and M. Arthur) and the CDPS AlbC have suggested that this modification induces a decrease CDPS activity. The flexizyme-based production pipeline described here thus constitutes a valuable alternative. Given the infinite diversity of RNA acceptors that can be aminoacylated with flexizymes and to the ability of these ribozymes to use any amino acid, it should be possible to produce AA-minitRNAs with any sequence desired. This would make it possible to dissect out the sequence-dependent recognition of tRNAs at the two binding sites of Nbra-CDPS or of other CDPSs.

Three-dimensional structures have been determined for various members of the three families of AA-tRNA-utilizing enzymes (15,39,48–49). Only CDPSs present a Rossmann-fold domain similar to that of class I AARSs (15). Thus, several different folds have evolved the ability to recognize AA-tRNAs. Structures of enzymes in complex with AA-tRNA analogues have also been reported, but the size and intrinsic flexibility of tRNAs have made it impossible to obtain high-resolution structures with full-size substrates. Minimal substrate analogues have, therefore, been used. We previously reported the structure of AlbC in complex with a dipeptidyl intermediate, shedding light on aminoacyl binding by CDPSs (11). In the case of LanBs and L/F transferases, structures have been reported with synthetic analogues of puromycin, which mimic the 3′-terminal aminoacylated adenosines of AA-tRNAs and in which the labile ester bond between the nucleotide and the amino acid is replaced by a stable amide bond (48,50). Using peptidyl-RNA conjugates produced by a solid-phase synthesis/click chemistry approach, Fonvielle et al. reported the structure of FemX in complex with a peptidyl-RNA miHx via two base pairs (43). The flexizyme-based production of various aminoacylated RNAs appears to be of great interest for determining the structure of CDPSs (or other AA-tRNA-utilizing enzymes) in complex with such analogues, particularly since the recent report from Prof. Suga and coworkers indicating that flexizymes are active on RNA molecules with a final adenosine residue bearing the 3′-deoxy-3′-amino modification, making it possible to produce analogues in which the ester bond of the aminoacyl linkage is replaced by the more stable amide bond (51).

The diverse methods for obtaining AA-tRNA analogues have different advantages and drawbacks. Methods based on AARSs are simple and efficient, but the spectrum of amino acids and RNA acceptors that can be used is restricted by the natural specificity of these enzymes. The chemoenzymatic synthesis of aminoacylated or peptidylated tRNA analogues provides access to a broad range of AA-tRNA analogues bearing diverse modifications. However, this approach requires the use of 2′-deoxycytidine (dC) as the penultimate nucleotide, which may be problematic if the 2′ hydroxyl of the C75 nucleotide of AA-tRNA is involved in the interaction with the enzyme. One of the major limitations of such approaches is that multi-step synthesis routes are not easily reproducible in a typical biological laboratory. Flexizymes constitute a convenient tool for aminoacylation, because they provide good aminoacylation yields in a single step in mild conditions with an almost infinite versatility of amino acids and RNA acceptors. The combination of flexizymes with anion-exchange chromatography presented here constitutes a valuable tool for studies of AA-tRNAs and associated mechanisms, as it can be used to generate a very broad range of aminoacylated RNAs. The application of anion-exchange chromatography to the study and manipulation of small aminoacylated RNAs makes it possible to characterize the proportion of an RNA sample that is aminoacylated, in a precise manner, by HPLC. This approach constitutes an interesting addition to the experimental toolbox used for these molecules, and its accuracy appears to exceed that of the conventional method based on acid-PAGE densitometry. We believe that this method will be of great value to the community of scientists studying the growing number of AA-tRNA-utilizing enzymes.

Click here for additional data file.