Minihelix-loop RNAs: minimal structures for aminoacylation catalysts

Abstract

We report here an in vitro selected ribozyme, KL17, which is active in charging amino acids on its own 5′-OH group. The ribozyme consists of two catalytic domains, one of which (consisting of P5/P6/L6) recognizes amino acid substrates based on the steric environment of the side chain, whereas the other recognizes an aminoacylated oligonucleotide. The secondary structure of this ambidextrous ribozyme arranges into a pseudoknot, where L6 docks onto the 3′-terminal single-stranded region. The formation of this pseudoknot structure brings the P6 region, in which the essential catalytic core is most likely embedded, into the proximity of the 5′-OH group. Our studies show that the P6–L6 domain can be separated from the main body of KL17 and the derived P6–L6 minihelix-loop RNA can act as a trans-aminoacylation catalyst. In this report, we also compare this ribozyme with an analogous aminoacylation system previously characterized in our laboratory and illuminate the similarities and differences between these catalytic systems.

INTRODUCTION

Aminoacyl-tRNAs are involved in decoding the genetic information stored in mRNA during translation, thereby making the tRNA aminoacylation system an essential component of the translation apparatus (1). The modern translation system utilizes a set of protein enzymes, called aminoacyl-tRNA synthetases (ARSs) to carry out tRNA aminoacylation (2). Since ARSs are sophisticated enzymes, they could not have existed before the advent of the modern day protein-based translation system. Hence, in the prototypical translation system, tRNA aminoacylation may have been catalyzed by some other primordial macromolecule. One such macromolecule is RNA and it has been postulated that RNA enzymes might have catalyzed the necessary reactions for translation, including aminoacylation of tRNAs (or tRNA-like molecules) (3–8) as well as peptidyl transfer (9–11).

The role of RNA enzymes in the aminoacylation of tRNA has been the focus of ongoing investigations in our laboratory. As a first step, a self-aminoacylating ribozyme was evolved and characterized (12–15). This ribozyme, named ATRib (acyltransferase ribozyme), transfers an N-biotinyl-l-aminoacyl group from the 3′-terminus of a donor oligonucleotide to its 5′-OH group and then it transfers the aminoacyl group to the 3′-terminus of a tRNA (6). Since ATRib-catalyzed tRNA aminoacylation is solely dependent on sequence complementarity between the ATRib internal guide sequence (IGS) and the substrate oligonucleotide, instead of the amino acid, ATRib transfers various amino acids with nearly equal efficiency. For ATRib to behave like a true ARS, however, it needs to recognize the amino acid specifically and then carry out tRNA aminoacylation. In order to confer ATRib with the amino acid recognition ability, a random sequence-containing domain (N70) was appended to ATRib at its 3′-end (Fig. 1A). This construct was then subjected to in vitro selection and evolution using a N-biotinyl-l-glutaminyl-cyanomethyl ester (biotin-Gln-CME) substrate (6). The selection yielded a GlnRS-like ribozyme, referred to as AD02, which exhibits a preference towards glutamine and discriminates against other amino acids by a factor of 20–1000-fold (16,17). AD02 also transfers the glutamine substrate to a tRNA and, thus, it has earned the distinction of being the first RNA-based catalyst that can aminoacylate a tRNA. In doing so, it emulates two essential functions (amino acid recognition and charging it on tRNA) of ARSs.

Figure 1.

In vitro selection of ribozymes that react with the Leu substrate. (A) Schematic illustration of the comparison of the selection for AD02 and KL17. (B) Sequence alignment of clones obtained during the selection using the Leu substrate. Three classes, I–III, are shown. The ATRib domain is not shown except for the IGS. The U75 insertion is highlighted in bold. The L6 and 3′-IGS^L6 sequences involved in base pairing in the class III clones are highlighted by green boxes. (C) Self-aminoacylation of KL17: a, KL17; b, biotin-aminoacyl-KL17 complexed with SAv (streptavidin). Reactions were carried out with 1 µM KL17 and 1 mM biotin-Leu-CME (Leu, lanes 1–4) or 2.5 µM biotin-l-Met-ACCAAC-5′ (Oligo, lanes 6 and 7) at 25°C. Two aliquots of the reaction were taken at 15 and 30 min. Control experiments (30 min time point) were performed with no SAv (lane 3), 5′-PPP-KL17 (lane 4) and no substrate (lane 5).

The success with AD02 encouraged us to increase the repertoire of the ARS-like ribozymes. Gln was the amino acid of choice in evolving the amino acid recognition domain for AD02 because its amide side chain might interact readily with the RNA. Although in vitro selections have yielded ribozymes that interact with aromatic amino acids such as Phe/Tyr (4,18) or RNA aptamers that interact with aliphatic amino acids such as Val/Ile (19,20), so far there has been no report of a ribozyme that reacts with an aliphatic amino acid. In order to examine the ability of a ribozyme to interact with a hydrophobic amino acid and carry out catalysis, we decided to use Leu as the substrate for selection. In this study we discuss the in vitro selection and characteristics of a ribozyme that reacts with biotin-Leu-CME. We then compare and contrast this new ribozyme with the previously characterized AD02. A striking similarity between AD02 and our ribozyme is that their minimal catalytic core, consisting of a minihelix-loop RNA motif, is able to aminoacylate ATRib or its derivative in trans (16). Although the two ribozymes use different strategies to place the catalytic core of the minihelix-loop RNAs in close proximity to the 5′-OH of ATRib domain, the nature of their catalytic role seems quite similar. These two systems add credibility to the view that a ‘minihelix-loop RNA world’ may have aided in establishing the primitive aminoacylation system in the RNA world (5,21).

MATERIALS AND METHODS

In vitro selection

The synthesis of the RNA pool used in this study has already been reported in our earlier work (6) and the general protocols for negative/positive selection, reverse transcription–PCR and transcription are the same as those in the previous report. The significant differences in the protocols in this study from the previous study are the alternation of substrates used in each round and the inclusion of negative/positive selection after round 8. This was done in order to enrich sequences that maintain the ambidexterity and 5′-OH-specific aminoacylation activities. Thus, biotin-Leu-CME was used in rounds 1–3, 8, 9, 11, 12 and 14, and biotin-l-Met-3′-ACCAAC-5′ was used in rounds 4–7, 10, 13 and 15. The selection conditions were as follows: 1 µM (10 µM for the first round) RNA was incubated in EK buffer (50 mM EPPS, 500 mM KCl, pH 7.5) at 95°C for 5 min, cooled to 25°C for 5 min, after which 50 mM MgCl₂ was added. The reaction was then initiated by addition of substrate and carried out for 3 h (rounds 1–11) or 30 min (rounds 12–15). The remaining procedures were carried out as in the previous selection. The RNA pool after round 15 was cloned and 22 clones were used for sequencing and further analyses.

RNA mapping assays

In vitro transcribed cold KL17 RNA was treated with calf intestine phosphatase (CIP) and purified by 6% denaturing PAGE. The resulting ribozyme was 5′-end-labeled with [γ-³²P]ATP by T4 polynucleotide kinase, followed by repurification of the labeled KL17 by 6% denaturing PAGE. Approximately 0.5 µM 5′-³²P-labeled KL17 was incubated with the EK buffer at 95°C for 5 min and then cooled to 25°C for 5 min. For the cleavage reactions performed in the presence of Mg²⁺, 50 mM MgCl₂ was added and equilibrated for 5 min. Cleavage reactions were initiated by addition of Pb(OAc)₂ (0.25 mM in the absence of Mg²⁺ and 1 mM in the presence of Mg²⁺) or the RNase S1 (1 U), T1 (5 U) or T2 (0.05 U) and reactions were incubated for 5–120 min (see Fig. 2B and D for the specific times). Upon completion, reactions were quenched by addition of 80 mM EDTA and ethanol precipitated. The cleavage products were resolved by 12 and 8% denaturing PAGE and quantified using a Molecular Imager FX (Bio-Rad).

Figure 2.

Secondary structure of KL17. (A) Secondary structure of the ATRib domain. The LR domain is shown in the purple box and the IGS is highlighted by a thick line. U75 is shown in bold. (B) RNase and Pb²⁺ mapping of the ATRib domain. Controls: no probe (lane 1); alkaline hydrolysis for 5 min (lanes 2 and 15); T1 digestion for 10 min (lanes 3 and 14); RNase T2 digestion for 5 min (lanes 4 and 13). Mapping: S1 nuclease digestion in the, absence (lanes 5 and 6) for 5 and 10 min, respectively, and presence (lanes 7 and 8), for 5 and 15 min, respectively, of 50 mM Mg²⁺; Pb²⁺-induced RNA cleavage in the absence (lanes 9 and 10) for 5 and 10 min, respectively, and presence (lanes 11 and 12) for 1 and 2 h, respectively, of 50 mM Mg²⁺. (C) Proposed secondary structure of the LR domain. The ATRib domain is shown in the blue box, the IGS is highlighted by a thick line and U75 is shown in bold. The L6–3′-IGS^L6 interaction is shown by a shaded arrow. The bases in the rectangular box were mutated to confirm the secondary structure. The five bases A165–C169 (5′-AUCAC-3′) at the 3′-end of the LR domain have been substituted with a line for simplicity. (D) RNase and Pb²⁺ mapping of the LR domain. All the lanes are as in (B).

Self-aminoacylation assay

One micromolar internally labeled [³²P]KL17 was incubated in EK buffer at 95°C for 5 min, cooled to 25°C for 5 min and equilibrated with 50 mM MgCl₂ for 5 min. The reaction was initiated by adding 1 mM biotin-Leu-CME (or 1 mM biotin-Gln-CME) or 2.5 µM biotin-l-Met-3′-ACCAAC-5′. Aliquots were taken at intervals of 15–60 min and quenched with 55 mM EDTA. After two ethanol precipitations, pellets were dissolved in 3 µl water, mixed with 5 µl MEUS buffer (25 mM MOPS, 5 mM EDTA, 8 M urea, 10 mM streptavidin, pH 6.5) and the products were resolved by 8% denaturing PAGE at 4°C and quantified using Molecular Imager FX (Bio-Rad). For the mutations or deletions in the L6 and 3′-IGS^L6 regions, the KL17 DNA template was amplified using the original 5′-primer and 3′-primers containing the mutations or deletions at the designated positions. For the mutation at U75, the ATRib domain was amplified using the original 5′-primer and primers that are specific for the 3′-end of the ATRib domain (C55–U90) containing the mutation(s) at the designated position(s). Similarly, the LR domain was amplified with a 5′-primer (which is antisense to the above 3′-primer of the ATRib domain) containing the mutation(s) in the designated position(s) and the original 3′-primer of the LR domain. These two DNAs were then mixed and subjected to PCR extension without primers. The extended DNA was amplified using the 5′- and 3′-primers corresponding to the ATRib and LR domains, respectively, to get the desired full-length mutants. These DNA templates were transcribed to the corresponding mutant ribozymes and the self-aminoacylation assays were performed in the same manner as above. The inhibition experiment was performed in the same manner as that previously reported elsewhere (16).

Trans-aminoacylation assay

A synthetic DNA template containing the P6–L6 minihelix-loop sequence (G99–G134) was amplified using a 5′-primer containing the T7 promoter sequence (5′-CATATGTAATACGACTCACTATAGGGCACC-3′) and the 3′-primer (5′-TTTTTTTTTTCACCCCTGTAAACTTGG-3′). This DNA was then in vitro transcribed by T7 RNA polymerase and the RNA was isolated by a standard protocol. Nested PCR was done to generate the two ATRib derivatives, ATRib^P5-3′-IGS and ATRib^3′-IGS, using the original 5′-primer and the 3′-primer (5′-CGTCCAACGGCCTCTCTTCAGACTAGATACTAAAGAGAACCTAACCAAAAAACAAAAAGC-3′ for ATRib^P5-3′-IGS; 5′-CGTCCAACGCTAACCAAAAAACAAAAA-3′ for ATRib^3′-IGS). These DNA templates were transcribed in the presence of [α-³²P]UTP and then the labeled ribozymes were dephosphorylated by CIP. The independently folded ATRib derivative and the minihelix-loop RNA (10 equivalents of the ATRib derivative) were mixed in EK buffer and the remaining procedure was carried out in a manner similar to that of the self-aminoacylation assay except that the reaction was carried out for 30 and 60 min.

Kinetic assays

Kinetic assays were performed in a manner similar to those described in our previous study (16). In brief, 1 µM RNA was folded in EK buffer, equilibrated with 50 mM Mg²⁺ and mixed with various concentrations of biotin-aminoacyl-CME at 25°C. Aliquots were removed at various time points, quenched with 55 mM EDTA and ethanol precipitated twice. The remaining procedure was the same as in the self-aminoacylation assay. Velocities were determined by taking at least six points from the linear regions of the time course. The data obtained from six different substrate concentrations were fitted to a non-linear regression curve according to the Michaelis–Menten equation using KaleidaGraph (Abelbeck Software).

RESULTS

Selection strategy and sequence analysis

We used an RNA pool containing a random sequence appended to the 3′-end of ATRib. This pool was previously used to select the AD02 ribozyme, specific to Gln, but this time it was applied to select RNA populations active towards Leu. In order to evolve a domain within the ribozyme that would interact with Leu, we used biotin-Leu-CME as a substrate for selection. Biotin is a selectable tag that allows us to separate active RNAs from inactive ones using immobilized streptavidin. The CME group is a leaving group that has a good balance of activation and hydrolytic stability, and its poor hydrogen bonding property makes it suitable for selecting ribozymes that recognize the amino acid moiety rather than the leaving group. We also used an oligonucleotide-based substrate, biotin-l-Met-3′-ACCAAC-5′, to retain the aminoacyl transfer activity of the ATRib domain. Note that the ATRib activity is dependent on the sequence complementarity between the IGS in ATRib and the oligonucleotide, but not dependent on the type of aminoacyl group (6). These two substrates were used during selection in order to isolate ribozymes capable of using both new and pre-existing catalytic domains, thus exhibiting ambidexterity. Therefore, the ribozymes can potentially execute 5′-aminoacylation and then charge the amino acid onto the 3′-OH of tRNA.

Our initial thrust was to select active sequences that recognize Leu using the random sequence region. Hence, we first planned to begin our selection using biotin-Leu-CME and subsequently switch the substrate to biotin-l-Met-3′-ACCAAC-5′. This protocol had already been used successfully to select the Gln-specific ambidextrous ribozyme AD02 (6). However, in the selection towards the Leu substrate, we encountered two setbacks. First, the active sequences that emerged after round 4 exhibited more aminoacylation at the internal 2′-OH or terminal 3′-OH rather than at the 5′-OH, based on the observation that the 5′-PPP-RNA pool exhibited self-aminoacylation activity. Second, the pool lacked self-aminoacylation activity in the presence of biotin-l-Met-3′-ACCAAC-5′, suggesting that the selected sequences in the new domain interfered with the ATRib activity. These results suggested that the internal or 3′-terminal activity in the pool readily overrides the 5′-terminal activity in this selection and the exclusive use of biotin-Leu-CME in earlier rounds biases the activity of the pool towards the Leu substrate completely.

To overcome the first setback, we used a combination of negative and positive selection procedures. For the negative selections, 5′-PPP-RNA was used instead of the 5′-OH-RNA to remove 5′-OH-independent active sequences, which should be trapped on the streptavidin column by virtue of aminoacylation. The unbound pool population was then collected, dephosphorylated to expose the 5′-OH group and used for the positive selection to isolate 5′-OH-dependent active sequences. To overcome the second setback, we used an alternation of the two substrates in each round, thereby preserving the ambidexterity of the ribozyme. Finally, after doing 15 rounds of selection according to the above protocols (see Materials and Methods for details), we successfully evolved an RNA population showing 5′-OH-dependent activities with both substrates.

The round 15 RNA pool was cloned and 22 clones were sequenced. Alignment of the sequences using CLUSTALW yielded three classes (I–III), with class I being the major class containing 14 members, while the others contain four members each (Fig. 1B). The ATRib domain in all the clones contained the salient features that have been observed in the case of AD02 and ATRib alone, such as the GAAA loop, a G:U wobble base pair in the P3 stem and the tandem G:U wobble base pairs in the P1 stem (13,15). Thus, these features suggest that the ATRib domain structure and activity is most likely retained in all the clones. In fact, representative clones from each class show acyl-transferase activity using biotin-l-Met-3′-ACCAAC-5′. It should be noted, however, that the ribozymes from class III contain an additional U in the IGS of the ATRib domain (Fig. 1B). On the other hand, no sequence homology is observed in the new domain between the different classes, indicating that each class evolved from an independent progenitor sequence. Furthermore, the new domain in all the classes has no sequence similarities to the glutamine recognition (QR) domain in AD02. This new domain is, therefore, referred to as the leucine recognition (LR) domain. One representative clone chosen from each class exhibited 5′-self-aminoacylation for both Leu and oligo substrates, indicating that all classes have the ambidextrous feature (data not shown).

Our preliminary assays of representative clones from the individual classes revealed that a clone from class III, KL17, has the best balance of self-aminoacylation activities for both biotin-Leu-CME and biotin-l-Met-3′-ACCAAC-5′ substrates, among other clones tested. In fact, 44% of KL17 self-aminoacylated in the presence of the Leu substrate over 30 min (Fig. 1C, lanes 1 and 2), and the controls agreed well with its 5′-OH-dependent activity (lanes 3–5). Furthermore, KL17 self-aminoacylated in the presence of the oligo substrate with a 24% yield over 30 min (lanes 6 and 7). It should be noted that KL17 has a unique U75 insertion between U74 and the IGS (Fig. 1B), which was also attractive for further investigation. We therefore chose KL17 for structural and kinetic characterization in this work.

Secondary structure of KL17

We used MFOLD, an RNA folding program, to predict the secondary structure of KL17 (22). The secondary structure of the ATRib domain in KL17 maintains the identity of ATRib, with four helices (P1–P4) and three loops (L2–L4) (Fig. 2A). Most importantly, its essential catalytic motifs, the tandem G:U wobble base pairs in P1, the G:U wobble base pair in P3 and the GAAA loop in L3, were completely conserved. On the other hand, the LR domain consists of two long helices (P5 and P6), a 10 nt loop (L6) and an asymmetrical internal loop (J5/6) (Fig. 2C). In order to verify the above secondary structure, a set of endoribonucleases, RNases T1 (as a sequencing marker), T2 and S1, was used to map the enzyme-accessible single-stranded regions. In addition to this mapping method, we also used lead (Pb²⁺) to probe the RNA structure. Pb²⁺ can also behave as a nuclease reagent similar to endoribonucleases, but because of its small size and an ability to localize in high affinity metal-binding sites, it can probe RNA structure in aspects different from that of the endoribonuclease method.

For the ATRib domain, the S1 nuclease cleavage profile in the absence of Mg²⁺ (Fig. 2B, lanes 5 and 6) showed cleavage sites at the three major loops, L2, L3 and L4 and the single-stranded IGS. Among these regions, L3 was considerably protected when S1 digestion was performed in the presence of 50 mM Mg²⁺ (lanes 7 and 8). Pb²⁺ cleaved not only the loops and IGS region but also the junction regions of the four stems (U28–U31 and U46–A47) in the absence of Mg²⁺ (lanes 9 and 10). However, the addition of Mg²⁺ significantly protected all these regions (lanes 11 and 12). We previously performed extensive mapping of ATRib alone (as an independent ribozyme) and its mutants using Pb²⁺ and Tb³⁺, and site-directed mutations in the G:U wobble base pair in P3 and the tandem G:U wobble base pairs significantly altered the cleavage signatures (15). These results showed that the localization of metal ions at the high affinity metal-binding site consisting of these G:U wobble base pairs plays a critical role in catalysis. The Pb²⁺ cleavage profile observed for the ATRib domain in KL17 was identical to that observed for ATRib alone, implying that the tertiary structure of this parental domain is conserved in KL17. Therefore, we believe that the mechanistic nature of the ATRib-dependent acyl-transferase function of KL17 is preserved from the original ATRib, but with the exception of the single U insertion in the IGS.

In the case of the LR domain, S1 nuclease digestion without Mg²⁺ produced a fingerprint with cleavage at L6 G116–C118, J5/6 U89–U96 and U138–A141 (Fig. 2D, lanes 5 and 6), whereas the addition of Mg²⁺ resulted in protection of these regions from cleavage (lanes 7 and 8). Pb²⁺ also cleaved the same regions as above, but more intense and specific cleavages were observed at U94–U96 and U117–C119 (lanes 9 and 10). Again, these regions were strongly protected from cleavage in the presence of Mg²⁺ (lanes 11 and 12). The Mg²⁺-dependent protection observed for the L6 and J5/6 regions from cleavage by the two distinct hydrolytic reagents suggests that when Mg²⁺ binds to the ribozyme, these regions may tightly fold into compact structures or be involved in contact with other regions of the ribozyme in a Mg²⁺-dependent manner (see below).

The above nuclease and Pb²⁺ cleavage experiments on KL17 provided support for the proposed secondary structure (Fig. 2A and C). To further confirm this, we introduced mutations at C102–G107 and C127–G132 in the P6 stem (highlighted by a rectangular box in Fig. 2C) to disrupt the base pairing and then restored the base pairing by generating the compensatory mutant. As expected, the mutants with the mismatch displayed a complete loss of self-aminoacylation, whereas the compensatory mutant with the restored base pairing completely recovered the activity (data not shown), thus lending strong support to the proposed secondary structure.

Intramolecular cross-talk in the LR domain and an intermolecular aminoacylation system

Despite the fact that L6 is a 10 nt loop, during the enzymatic and Pb²⁺ cleavage assays it was substantially protected (Fig. 2D). We speculated that the loop sequence might be involved in a long-range interaction with another part of the ribozyme. The Gln-specific AD02 ribozyme also has a 9 nt loop (called L6b), with the sequence 5′-UAACCA-3′, which forms six base pairs with the IGS (5′-UGGUUG-3′), resulting in a unique pseudoknotted structure (16,17). We therefore considered a similar possibility in KL17, hypothesizing the formation of base pairs between L6 and IGS. In fact, we found a possible base paring interaction between U75–U80 and G116–A121, albeit containing double wobble base pairs (U80:G116 and U79:U117). Unfortunately, our attempt at introducing compensatory mutations in these regions was unsuccessful in restoring the activity, indicating that a similar interaction does not exist in KL17.

Hence, we turned our attention to another single-stranded region, the 3′-terminus, for potential base pairing. The 3′-terminal sequence has 13 bases, of which seven, 5′-UUGGACG-3′, can form Watson–Crick bases pairs with 5′-UGUCCAA-3′ in L6 (Fig. 2C). To test the importance of these seven bases at the 3′-end, referred to as 3′-IGS^L6, we generated two deletion mutants. In one mutant, the entire single-stranded region (13 bases) at the 3′-end was deleted, while in the other only five bases (5′-₁₆₅AUCAC₁₆₉-3′) beyond the 3′-IGS^L6 were deleted (these five bases are shown by a line in Fig. 2C). The result was striking: the former mutant showed no self-aminoacylation activity (Fig. 3A, lane 2), whereas the latter mutant virtually retained the wild-type activity (Fig. 3A, lane 1 versus 3). This indicates the necessity of the seven bases in 3′-IGS^L6 for catalysis. The base pairing interaction between L6 and 3′-IGS^L6 was confirmed when mutants that disrupted this interaction exhibited significantly diminished self-aminoacylation activity (Fig. 3B, lanes 3–6). Conversely, the compensatory mutant that restored the base pairing displayed significant recovery of self-aminoacylation activity (Fig. 3B, lanes 7 and 8 for the mutant versus lanes 1 and 2 for the wild-type). These results clearly indicate that this base pair interaction is indeed present and essential for functioning of the ribozyme.

Figure 3.

L6–3′-IGS^L6 interaction in KL17: a, KL17 or its various mutants; b, biotin-Leu-KL17 or its mutants complexed with SAv. Reactions were carried out with 1 µM ribozyme and 1 mM biotin-Leu-CME at 25°C. (A) Deletion of 3′-IGS^L6. Reactions are incubated for 30 min. WT, wild-type KL17 (lanes 1 and 4); 3′-Δ13, a mutant in which 13 bases at the 3′-end were deleted (lanes 2 and 5); 3′-Δ5, a mutant in which five bases from the 3′-end were deleted (lanes 3 and 6). Control experiments in lanes 4–6 have no SAv. (B) Complementarity between L6 and 3′-IGS^L6: wild-type KL17 (lanes 1 and 2) and mutants that disrupt the complementarity (lanes 3–6) and compensatory mutants (lanes 7 and 8). The three bases in L6 and 3′-IGS^L6 studied are shown along with the numbers. Two time points for the reaction were taken at 30 and 60 min.

For L6 to interact with the 3′-IGS^L6, P6 very likely bends at J5/6. This led us to hypothesize that the asymmetric internal loop in J5/6 might play an important role in governing the above long-range interaction. We therefore substituted C139–U143 with a single adenosine (C139–U143→A), which gives a long stem–loop structure. This should prevent the necessary bending in the LR domain, thereby disallowing the long-range interaction. This mutation resulted in an ∼6–7-fold reduction in activity of the wild-type (Fig. 4A, lanes 1–3 versus 4–6). However, we considered that this degree of reduction is unexpectedly small, because if bending of the stem governs the L6–3′-IGS^L6 long-range interaction a nearly complete loss of activity should occur. As an alternative explanation, the observed activity can be attributed to an intermolecular interaction between L6 and 3′-IGS^L6, thereby aminoacylating the 5′-OH of ATRib in trans.

Figure 4.

Deletion of the asymmetric loop, trans-aminoacylation and mutation/deletion of U75. (A) Deletion of the asymmetric loop of the LR domain: a, KL17 mutant; b, biotin-Leu-KL17 mutant complexed with SAv; c, KL17; d, biotin-Leu-KL17 complexed with SAv. C139–U143→A, a mutant in which the bases C139–U143 were replaced by a single adenosine (lanes 1 and 2); SAv-deficient control of lane 2 (lane 3); WT, wild-type KL17 (lanes 4 and 5); SAv-deficient control of lane 5 (lane 6). Reactions were carried out with 1 µM ribozyme and 1 mM biotin-Leu-CME at 25°C. (B) Trans-aminoacylation systems. A P6–L6 minihelix-loop RNA paired with ATRib^P5-3′-IGS (left) or with ATRib^3′-IGS (right). The base numbering of these constructs was kept the same as the full-length KL17. The interaction between L6 and 3′-IGS^L6 is indicated by the shaded arrow. Five bases in the unpaired 3′-end sequence of the LR domain were deleted in this series of studies. (C) Trans-aminoacylation of ATRib derivatives by the P6–L6 minihelix RNA: a, ATRib^P5-3′-IGS; b, biotin-Leu-ATRib^P5-3′-IGS complexed with SAv; c, ATRib^3′-IGS; d, KL17; e, biotin-Leu-KL17 complexed with SAv. ATRib^P5-3′-IGS (lanes 1–4), ATRib^3′-IGS (lanes 5–8) and KL17 (lane 9). Controls: no minihelix-loop RNA (lanes 3 and 7) and no SAv (lanes 4 and 8). (D) Mutational studies of U75: a, KL17 or U75 mutants; b, biotin-aminoacyl-KL17 or mutants complexed with SAv. Reactions were carried out with 1 µM ribozyme and 1 mM biotin-Leu-CME (lanes 1 and 3–5) or 2.5 µM biotin-Met-3′-ACCAAC-5′ (oligo, lane 2) at 25°C for 30 min. Abbreviations for mutants are described in the text.

The above observation led us to design experiments testing the 5′-aminoacylation of ATRib derivatives by a minihelix-loop RNA consisting of P6–L6. We constructed two systems for the intermolecular aminoacylation. For both systems, the trans-acting minihelix-loop RNA consists of P6–L6 (G99–G134), with incorporation of two guanosines and 10 adenosines at its 5′- and 3′-ends, respectively, for efficient in vitro transcription. For the first construct, we generated an ATRib derivative, ATRib^P5-3′-IGS (Fig. 4B, left), which contains the entire P5 stem and 3′-IGS^L6. The second construct of ATRib, ATRib^3′-IGS (Fig. 4B, right), has only 3′-IGS^L6 and the remaining regions were deleted. We generated these two systems based on the consideration that proper juxtaposition of 3′-IGS^L6 governed by P5 is probably necessary to form the active complex with the minihelix-loop RNA. As expected, the first system showed the 5′-aminoacylation of ATRib^P5-3′-IGS (Fig. 4C, lanes 1 and 2), whereas the second system exhibited no aminoacylation of ATRib^3′-IGS (lanes 5 and 6). A control experiment for the first system (lane 3) demonstrated that the observed activity is minihelix-loop RNA-dependent, and its activity was only 2-fold reduced compared to the cis-acting system (lane 9) under the optimized conditions. Thus, the minihelix-loop RNA can act as a trans-aminoacylation catalyst upon docking onto the 3′-IGS^L6 of ATRib^P5-3′-IGS.

U insertion in the IGS

The necessity of the proper juxtapositioning of 3′-IGS^L6 for minihelix-loop RNA-dependent trans-aminoacylation suggested that the L6–3′-IGS^L6 interaction lies close to the IGS in the ATRib domain. Since the 5′-OH in P1 is the aminoacylation site, the entities L6–3′-IGS^L6 and P1–IGS might form the catalytic center. Interestingly, the IGS in KL17 contains a unique U insertion at position 75, which was not recognized in either ATRib or AD02. To verify the importance of this U75, we constructed a series of mutants (Fig. 4D).

The U deletion mutant (U75Δ) showed a 4-fold decrease in LR-dependent 5′-self-aminoacylation activity (Fig. 4D, lane 1 versus 5), while it showed an ∼2-fold increase in ATRib-dependent activity (Fig. 4D, lane 2; for comparison see Fig. 1C, lane 7). These results suggest that the U insertion increases the LR-dependent activity, but mildly reduces the ATRib-dependent activity. The U75C mutant also displayed a 5-fold drop in activity (lane 3 versus 5), indicating that the role of U75 is not simply spacing between P1 and the IGS, but more likely the specific U base is necessary for activity. We further confirmed the importance of U insertion by testing a mutant with a U75 deletion and a single A insertion between A81 and G82 (U75Δ-A), in which 3′-IGS^L6 might be placed at the same distance from the 5′-OH as in the wild-type. This mutant again showed a 4-fold reduction in activity (lane 4 versus 5). These mutational studies on U75 indicate its important role in the LR-dependent activity, presumably in interacting with the amino acid substrate or in formation of the catalytic core in the vicinity of L6–3′-IGS^L6 and P1–IGS.

Amino acid recognition

Our selection procedures in this study as well as the earlier study did not include a negative selection process against other amino acid substrates. Despite this, the AD02 ribozyme isolated in the earlier study exhibits remarkable amino acid selectivity towards the cognate Gln substrate (16). We therefore wondered whether KL17 exhibits selectivity towards the Leu substrate over others. It was also of interest to directly compare the catalytic efficiencies of these two distinct ribozymes towards the same substrates. Thus, we first compared the activity of KL17 and AD02 with three substrates, biotin-l-Met-3′-ACCAAC-5′, biotin-Leu-CME and biotin-Gln-CME.

The activity of KL17 towards the Leu substrate is ∼50-fold higher than the activity of AD02 towards the Leu substrate (Fig. 5, lane 1 versus 4), clearly indicating that KL17 is able to self-leucinylate much more efficiently than AD02. On the other hand, in the case of the ATRib activity using biotin-l-Met-3′-ACCAAC-5′, KL17 shows a 2-fold lower activity than AD02 (lane 2 versus 5) due to the U75 insertion. To our surprise, when KL17 was reacted with the Gln substrate, we observed a tremendous increase in the yield of self-aminoacylated ribozyme compared with Leu, giving near completion of self-glutaminylation in 15 min (lane 3). In comparison, AD02 self-glutaminylated with 18% yield in 15 min (lane 6), while its activity towards the Leu substrate was negligible (lane 4), thereby reiterating its high specificity towards Gln. Since these experiments used a sub-end-point assay for KL17, they do not necessarily represent its catalytic rate for the substrate. Therefore, we extended our studies to kinetic experiments and determined the Michaelis constants for not only these two substrates but also other biotin-l-aminoacyl-CMEs available in our laboratory.

Figure 5.

Comparison of self-aminoacylation of KL17 and AD02: a, KL17 or AD02; b, biotin-aminoacyl-KL17 or AD02 complexed with SAv. Reactions were carried out with 1 µM ribozyme and 1 mM biotin-Leu-CME (Leu) or 1 mM biotin-Gln-CME (Gln) or 2.5 µM biotin-l-Met-3′-ACCAAC-5′ (Oligo). Aliquots were taken after 15 min. Self-aminoacylation of KL17 (lanes 1–3) and AD02 (lanes 4–6) with biotin-Leu-CME (lanes 1 and 4), biotin-l-Met-3′-ACCAAC-5′ (lanes 2 and 5) and biotin-Gln-CME (lanes 3 and 6).

The specificity constant (the second order rate constant) of KL17 for each substrate was generated by the Michaelis–Menten parameters and these are listed in Table 1. Consistent with the preliminary assay, the specificity constants revealed that KL17 is 36-fold more active towards l-Gln than l-Leu. Even more interestingly, the activity towards l-Gly and l-Met was 10-fold higher than l-Leu, and that for l-Phe was 2-fold higher. l-Val is the only exception, having 10-fold less activity than l-Leu. Thus, the activity of KL17 seems to depend upon the degree of steric hindrance of the side chain on the amino acid. The side chains of l-Gln, l-Met and l-Gly, all of which have no γ-branch, are sterically less challenged compared with γ-branched l-Leu, which may explain the greater reactivity of these substrates. The similar activity observed for the γ-branched l-Phe as well as the poor activity observed for β-branched l-Val also support this idea. The 3-fold enhancement for the reactivity of l-Gln over l-Met and l-Gly can be attributed to preferential interactions of the amide side chain which resides within or near the active site in the ribozyme.

Table 1. Michaelis–Menten parameters of KL17 for various amino acid substrates.

Amino acid	kcat × 103 (s–1)	Km × 10–3 (M)	kcat/Km (s–1 M–1)	Relative specificity constant
Leu	0.42	0.71	0.6	1 (0.03)
Gln	6.10	0.28	21.2	36 (1)
Gly	6.25	1.00	6.3	11 (0.30)
Met	4.86	0.81	6.0	10 (0.28)
Phe	0.40	0.31	1.3	2.1 (0.06)
Val	0.087	1.24	0.07	0.1 (0.003)

Relative specificity constants were determined by the observed second order rate constant of each substrate divided by that of the Leu substrate or divided by that of the Gln substrate (parentheses).

To determine whether the CME group is an important recognition element of the substrate, we performed an inhibition experiment using biotin-Gln-O^– (free carboxylate). Dixon analysis revealed that the K_i for biotin-Gln-O^– was ∼0.75 mM, which is only 2-fold higher than the K_m value determined for biotin-Gln-CME. This indicates that the CME group is a minor recognition element of the substrate. On the other hand, when biotin or glutamine was used as an inhibitor to compete with biotin-Gln-CME, neither inhibitor exhibited a detectable degree of inhibition. The biotinyl group is commonly present in all substrates tested in Table 1, but yet the ribozyme is able to preferentially react with substrates that have a sterically less hindered side chain. In the light of all the above observations, it is very likely that both the steric environment of the amino acid side chain and the N-acyl moiety of biotin (the linking amide group between the amino acid and biotin) cooperatively contribute to binding to the ribozyme.

Another important result to emerge from these experiments is that the KL17 self-aminoacylation rates observed for l-Leu and l-Gln were 40- and 13-fold, respectively, faster than those of AD02 (16). These results indicate that KL17 has greater aminoacylation activity than AD02, but has gained a faster rate at the expense of a loss of specificity for the amino acid.

DISCUSSION

Selection for KL17 versus AD02

In the current study and in our previous study we used the same RNA pool containing a random domain appended to the ATRib domain and carried out selection towards two distinct amino acids. In both the approaches, in order to obtain active sequences capable of carrying out self-aminoacylation, the pool was subjected to two specific selective pressures: (i) the ability of the RNA to aminoacylate only the 5′-OH of the ATRib domain (as against the internal 2′-OH or the terminal 3′-OH); (ii) its ability to retain the ATRib activity. Therefore, our selection, with such strict and specific pressures, is more demanding and difficult compared with other examples of selection that use a completely random RNA pool and enrich for sequences that modify any sites of the RNA (23). Despite such restrictions, we succeeded in selecting ribozymes that are able to 5′-self-aminoacylate and also conserve the activity of the parental ATRib domain.

However, the profile of emerging active sequences in the two selections turned out to be quite different from each other. In the selection using the Gln substrate, we found that the majority of active clones in the pool showed 5′-OH-specific activity rather than any other site-specific activity, so we performed only one round of negative selection to suppress the latter activity (6). The ATRib activity in the enriched pool was low but readily enhanced by only three rounds of continuous selection using the oligo substrate (biotin-l-Met-3′-ACCAAC-5′) while maintaining the activity toward the Gln substrate. Cloning of the final pool yielded one major class of the desired ambidextrous ribozymes (AD02 is a representative clone from this class) and five independent ambidextrous ribozyme clones. On the other hand, in the case of selection towards the Leu substrate, we encountered difficulties in enriching the 5′-specific activity as well as preserving the ATRib activity, and therefore it was necessary to perform a series of negative/positive selections for the 5′-OH-specific activity, alternating between the Leu and oligo substrates. This might indicate that the active sequences capable of reacting towards the Leu substrate while maintaining the activity of the ATRib domain were rarer in the pool than those reacting towards Gln substrate, thereby making it more difficult to enrich such ambidextrous sequences.

Interestingly, the KL17 ambidextrous ribozyme selected towards the Leu substrate exhibited greater activity towards Gln over Leu. Our kinetic studies using five different amino acid substrates suggest that the ribozyme recognizes the steric environment on the γ-branched carbon center, thus charging Gln, Met and Gly better than Leu (Table 1). This is in stark contrast to the AD02 ribozyme, where the carboxyamide moiety of the Gln side chain is critical for substrate activity, which results in discrimination against other amino acids by a factor of 20–1000-fold (16). Although KL17 has broader amino acid selectivity than AD02, it is important to note that its k_cat/K_m value for Gln is 36-fold greater than that of AD02. This raises a question as to why our previous selection using the Gln substrate did not yield the KL17 class, i.e. only yielded the AD02 class. Our mutational studies on KL17 have shown that U75, which is newly inserted between U74 and the IGS in the constant region of the original pool, plays a critical role in catalysis. This ‘fortuitous’ insertion of U (note that since error-prone PCR was not performed in the selection from the random pool, this insertion is truly fortuitous) results in an increase in the self-leucinylation activity of the KL17 class, in contrast to clones of the AD02 family found in the previous selection for self-glutaminylation, where the constant region of ATRib was completely conserved (6). Therefore, the reason for the lack of isolation of the KL17 class in the pool that yielded the AD02 class is simply the lack of this position-specific U insertion during the selection. This in turn allowed us to isolate the AD02 class (without enriching the KL17 class in the pool), which has a better balance of QR and ATRib activities as well as high specificity towards Gln. Comparison of the activity of KL17 (Table 1) and AD02 (6) towards the Gln substrate revealed that the higher activity of KL17 can be attributed to improvements in both k_cat and K_m compared with those observed for AD02 (k_cat = 0.49 × 10^–3 s^–1 and K_m = 0.80 × 10^–3 M^–1). However, the enhancement in k_cat appeared to be more significant than that in K_m (12- versus 3-fold). Thus, the difference in the catalytic motifs between KL17 and AD02 (see below) appears to have a greater impact on k_cat.

Minihelix-loop RNA as a trans-aminoacylation catalyst

Despite the fact that the primary sequence and the secondary structure of the amino acid recognition domain of KL17 are very different from those of AD02, an interesting structural feature is common: both ribozymes form a pseudoknotted secondary structure between a loop and an IGS and this loop–IGS interaction is essential for activity. However, formation of this pseudoknot is fairly different. In KL17, L6 interacts with the single-stranded region, named 3′-IGS^L6, located at the 3′-end of the LR domain (Fig. 2C), whereas in AD02, L6b interacts with the IGS located between the ATRib and QR domains (16). Additionally, the appropriately designed ATRib constructs, ATRib^P5-3′-IGS and ATRib, can be aminoacylated by the cognate minihelix-loop RNAs, P6 and P6a–P6b, respectively (Fig. 6). Thus, the minihelix-loop RNAs can act as trans-aminoacylation catalysts in both systems.

Figure 6.

Comparison of the trans-aminoacylation system derived from KL17 (A) and AD02 (B). L6 and 3′-IGS^L6 in KL17 and L6b and the IGS in AD02 are highlighted by green boxes and the interactions of these regions are shown by shaded arrows. The base numbering in both systems are kept the same as the cis-acting full-length ribozymes. Base pairing of the triple U residues with GAA (P6, KL17) and AGA (P6b, AD02) are shown in the red boxes.

Our kinetic studies on KL17 showed that this ribozyme functions well towards the Gln substrate, which brings our attention to finding the sequence similarity between the two catalytic minihelix-loop RNAs. As a matter of fact, both RNAs have triple U residues paired with GAA (P6, KL17) or AGA (P6b, AD02) and purine-rich unpaired sequences in the L6 loops (Fig. 6). Our extensive structural studies on AD02 using chemical modifications and nucleotide analog interference mapping (NAIM) revealed that an 11 nt cluster embedded in P6b (C110–A113/U123–G126) and the unpaired bases in L6b (A114–A116) constitute the Gln-binding site (Fig. 6B) (16). Although we still await more extensive biochemical mapping of KL17 to define its amino acid-binding site, the similarity of the P6 motif of the KL17 minihelix-loop RNA to that of AD02 certainly suggests a critical role of the P6 motif in binding to the amino acid. More interesting questions remaining to be answered include, how do these residues form the amino acid binding site and what is the difference in the tertiary structure between these two catalytic minihelix-loop RNAs (e.g. how is U75 in KL17 involved in formation of the catalytic core)?

Our findings in these ribozyme systems derived from KL17 and AD02 show a remarkable ability of minihelix-loop RNAs for aminoacylation in trans. These two systems share similar secondary structural motifs embedded in the stem, but yet show different degrees of catalytic activity and specificity towards amino acids, presumably due to the different tertiary contacts of the catalytic residues with the substrate. The catalytic ability as well as amino acid recognition ability of the minihelix-loop RNAs suggest that these simple RNA molecules could have played critical roles in not only accepting amino acids (5,21,24) but also charging amino acids in the ancient RNA world.