De novo synthesis and development of an RNA enzyme

Abstract

Arbitrary manipulation of molecular recognition at the atomic level has many applications. However, systematic design and de novo synthesis of an artificial enzyme based on such manipulation has been a long-standing challenge in the field of chemistry and biotechnology. In this report, we developed an artificial RNA ligase by implementing a synthetic strategy that fuses a series of 3D molecular modelings based on naturally occurring RNA–RNA recognition motifs with a small-scale combinatorial synthesis of a modular catalytic unit. The resulting ligase produces a 3′–5′ linkage in a template-directed manner for any combinations of two nucleotides at the reaction site. The reaction rate is 10⁶-fold over that of the uncatalyzed reaction with a yield higher than those of previously reported ligase ribozymes. The strategy may be applicable to the synthesis and development of a variety of nonnatural functional RNAs with defined 3D structures.

Many RNA receptors and enzymes have been selected in vitro from combinatorial libraries that consist of a long random sequence (1). Anatomies of the selected RNAs are difficult to predict because their higher-order structures were not specified before the selection. However, if a precisely designed assemblage of well established modular units is used as a scaffold for preparing such functional RNAs, redesign of the resulting selected RNAs will be facilitated. They will thus enable a variety of artificial evolutionary pathways, starting from the selected molecule.

In general, molecular design of biopolymers such as RNAs or proteins is difficult at the 3D level because of their highly complicated folding process. In the 1990s, biochemical and structural analyses revealed that many functional noncoding natural RNAs are organized into modules and fold into defined 3D structures (2–5). Moreover, several commonly used RNA–RNA binding motifs in these RNAs were identified by phylogenetic comparison (6) and high-resolution structural analyses (7–10). Consequently, it has become possible to design self-folding RNAs precisely by employing such motifs and mimicking the modular organization of natural RNAs (11–13). As one such example, we have previously reported the design of a self-folding RNA consisting of standard double-stranded helices connected by the two motifs: a tetraloop–receptor interaction and consecutive base-triples (Fig. 1 A and B) (13). Results indicated that the constructed RNA folds compactly into the designed 3D structure.

Fig. 1.

Synthetic scheme and structures of DSL ligases. (A) From left to right, structural motifs, GAAA loop (red), 11-nt receptor (green), and triple-helical scaffold (blue), used in the molecular design of the scaffold (far left), simplified secondary structures of the scaffold (second left), cis-DSL (third left), trans-DSL-1 (fourth left), and trans-DSL–2 (far right). A putative secondary structure of the selected catalytic module inserted at the R region in B is shown in purple. GAAA loop specifically interacts with the 11-nt receptor. Nicks in the DSLs correspond to the reaction site. (B) Secondary (Left) and 3D structure (Right) of the scaffold RNA. (Left) The putative reaction site is indicated with an arrow. Two sets of substrate oligonucleotides (yellow) and their complementary sequences (positions 19–23) (Left, S2/Type 2; Right, S1/Type 1) are shown. L and R indicate the regions replaced with 30 random nucleotides for constructing the libraries. Bio indicates biotin. (Right) A model 3D structure. Colored regions correspond to those in the secondary structure. Yellow sphere indicates the reaction site. To show the relative position between the reaction site and L or R region without obstacles, a view from the side opposite to that shown at Left is depicted. (C) Secondary structures of the selected catalytic module. (Top) Proposed secondary structure of the catalytic module of cis-DSL-01 which consists of the selected 30 nucleotides (red) and its adjacent regions. (Middle) The consensus sequence deduced from the second selection. In the library for the second selection, positions 51–55 and 75–105 of cis-DSL-01 were replaced with randomized (25% each) and degenerate (61% original + 3 × 13% other three) bases, respectively. X-Y, Py, and n indicate Watson–Crick base pairing, pyrimidine base, and nonconserved base, respectively. Within the box, capital and small letters indicate highly and modestly conserved nucleotides. Invariant nucleotides are shown as red. The nucleotides whose identify seem responsible for improving the activity are marked with green (compare Top and Bottom). (Bottom) The sequence of cis-DSL-1S. The nucleotides responsible for improving the activity are marked with green. Italic letters correspond to the nucleotides different from those in cis-DSL-01 (Top).

In this paper, we report the synthesis and development of an artificial RNA ligase as shown in the scheme (Fig. 1 A). First, a reaction site for RNA–RNA ligation was installed into the designed RNA scaffold, and a different region of the RNA was subject in vitro selection from a small combinatorial library to provide a catalytic center. The ligation reaction was chosen as a convenient target because several RNA ligases had already been obtained from large-scale pools by in vitro selection (ref. 14 and references cited therein). Biochemical characterization and second-selection of the catalytic unit were performed for the resulting RNA, termed cis-DSL. For further development of the molecule, the organization of cis-DSL, which conducts the reaction in an intramolecular manner, was redesigned and converted into two different RNA enzymes (trans-DSL-1 and -2), each of which recognizes and ligates a different RNA substrate.

Materials and Methods

Sequences. The sequences of the synthetic oligonucleotides used in the experiments are shown in Data Set 1, which is published as supporting information on the PNAS web site.

Construction of the RNA Library. The DNA templates for a pool containing 30 random nucleotides in the L or R region in the P3a helix of the self-folding RNA (Fig. 1B) were constructed by PCR with three synthetic oligonucleotides. The three oligonucleotides for constructing the pool L containing 30 random nucleotides in the L region were pool L-a, pool N30L-b, and pool L-c. The oligonucleotides for the pool R were pool R-a, pool N30R-b, and pool R-c. In 6 ml of reaction mixture, 15 cycles of PCR were performed with 120 pmol of the pool N30L-b or pool N30R-b to produce ≈200 pmol PCR products. By using these PCR products as templates for in vitro transcription, ≈1,000 pmol (7 × 10¹³ molecules, eight copies for each sequence) of the RNA was obtained. The two pools were mixed to prepare 2,000 pmol of RNA for the selection. A doped cis-DSL-01 library was constructed by primer extension with two synthetic oligonucleotides, KTL-1 dope sense and KTL-1 dope antisense. By using 1,000 pmol of each oligonucleotide, primer extension was performed with Ex Taq polymerase (Takara Shuzo, Tokyo).

The resulting mixture was used for transcription without further purification to obtain 800 pmol of RNA. The DNA templates were digested with RQ I DNase (Promega) after the transcription, and the resulting RNAs were purified by 5% denaturing PAGE followed by precipitation with ethanol.

In Vitro Selection. The purified RNAs were dissolved in H₂O, and then denatured at 80°C for 3 min, followed by preincubation at 37°C for 5 min. The RNA folding was initiated by adding 5-fold concentrated reaction buffer at 37°C. After 10 min, a 10-fold concentrated substrate RNA was added to start the ligation reaction. The final concentrations of the pool RNA, the substrate RNA, and Tris·Cl (pH 7.7) were 1 μM, 2 μM, and 30 mM, respectively. As described in Results and Discussion, two sets of 5′-biotinylated substrate RNAs (S-1 and S-2) and their guide sequences (types 1 and 2) were used alternatively. The S-1/type 1 set was used in rounds 1, 3, and 5, and the S-2/type 2 set was used in rounds 2 and 4. The conditions are summarized in Table 1, which is published as supporting information on the PNAS web site. The reaction was stopped by ethanol precipitation. The ligated products were captured on streptavidin paramagnetic particles (Promega) and hybridized with a DNA primer complementary to their 3′ region. Reverse transcription was performed with ReverTra Ace (Toyobo, Osaka) with three primers, Rv-S in rounds 1 and 4, Rv-M in rounds 2 and 5, or Rv-L in round 3. The resulting cDNAs were eluted with 150 mM KOH, followed by neutralization with 150 mM HCl. The cDNAs were selectively amplified by PCR using the primer described above together with the selective primers complementary to the sequences of the substrate RNAs. PCR for regeneration and type-switching was carried out with the primer containing the sequence of the T7 promoter together with the primer containing the template for the guide sequence of type 1 or 2. The resulting DNAs were used as templates for the next round.

The ligated RNAs from the fifth round pool were isolated by using 5% denaturing PAGE, followed by reverse transcription and PCR, and cloned into pGEM-T vectors (Promega). Individual clones were sequenced by using the BcaBEST Dideoxy Sequencing kit (Takara Shuzo) with an automated DNA sequencer (ALF express II, Amersham Pharmacia). For in vitro selection from the doped cis-DSL-01 library performed in a similar manner, the RNAs after the fourth-round selection were isolated and sequenced. The conditions are summarized in Table 2, which is published as supporting information on the PNAS web site.

Ligation Assays. Uniformly ³²P-labeled DSL ligases were dissolved in 35 μl of H₂O, then denatured by incubating at 80°C for 3 min, followed by preincubation at 37°C for 5 min. RNA folding was initiated by adding 5-fold concentrated reaction buffer (10 μl) at 37°C. After 10 min, a 10-fold concentrated substrate RNA (5 μl) was added to start the ligation reaction. The standard conditions were as follows. Final ligase and substrate concentrations were 25 nM and 1 μM, respectively. The reaction was performed with MgCl₂, KCl, and Tris·Cl (pH 7.7) (50, 200, and 30 mM, respectively) at 37°C. At each time point, an aliquot was treated with an equal volume of the stop solution consisting of 85% formamide, 100 mM EDTA, and 0.1% xylene cyanol (XC). Samples were separated on 5% polyacrylamide denaturing gels and quantified with a Bio-Imaging Analyzer (BAS2500; Fuji). The data were fitted to single exponential curves that give the observed rate constant (k_obs) and calculated final yield. All experiments were performed at least in duplicate, and those of cis-DSL-01 and cis-DSL-1 were performed five and four times, respectively.

Mutant Ribozymes and Substrate RNAs. Template DNAs for transcription of cis-DSLs and their derivatives for determining specificity at the ligation site (see Fig. 3) were prepared by PCR with plasmid bearing cis-DSL-01, cis-DSL-1, or its variants together with the primers within which the T7 promoter sequence and respective mutations were introduced. Sequences of substrate RNAs used for assay in Fig. 3 are listed in Data Set 1. For trans-DSLs (see Fig. 4), the templates for synthesizing the catalytic unvit were prepared by PCR, and those for substrate RNAs were prepared by primer extension with two synthetic oligonucleotides. The DSLs were prepared by in vitro transcription with T7 RNA polymerase and purified by electrophoresis on 5% or 10% denaturing polyacrylamide gels.

Fig. 3.

Effect of helical contexts in P1 region of DSL ligases. (A) (Upper) Structures and sequences of five types of substrate–P1 combinations. Nucleotides different from the original type 1 (with original substrate–P1) are indicated with red. For substrate RNAs, last five nucleotides are shown (their intact sequences are shown in Data Set 1). (Lower) Autoradiograms depicting the activities of variant cis-DSL-1 ligases that possess four different substrate–P1 combinations. Sub, substrate; IGS, internal guide sequence; Rib, ligase ribozyme; SubRib, ligated product. (B) (Upper) Reaction scheme of successive nucleotidyl-addition reactions. pppX, a nucleoside triphosphate; Y, a complementary nucleotide to X. pp–, inorganic pyrophosphate released by the reaction. (Lower Left) An autoradiogram of nucleotidyl addition reaction of cis-DSL-1S's variants. The reaction mixtures were run on 15% denaturing polyacrylamide gel. Asterisks indicate the components labeled with ³²P. (C) Substrate specificity of the reaction with nucleoside triphospahte (pppX) and substrate-1 (sub1). Relative intensities of the products containing a variety of sub1 and ³²P-labeled mononucleotide are indicated at the bottom of each autoradiogram.

Fig. 4.

Structure and reactivity of trans-DSL-1 and -2. (A) Secondary structure of the substrate unit comprising the corresponding substrates 1 and 2 components (Upper Left) and trans-DSL-1 (Upper Right). Tetraloop, receptor, and catalytic module are indicated with red, green, and purple, respectively. Ligation site is indicated with an arrow. (Lower) Time course of the reaction. Filled circle, with designed original substrate unit; filled triangle, with control substrate unit A [the 11-nt receptor in the original substrate unit is substituted with the B7.8 receptor (20), which presumably interacts weakly with the ligand]; filled square, with control substrate unit B (two consecutive 5′C–G3′ pairs in the B7.8 receptor were deleted from the control substrate unit A). Conditions were as follows: 0.1, 1, and 1 μM trans-DSL-1, substrate 1, and substrate 2, respectively; and 25, 25, 5, 1, and 40 mM MgCl₂, NaCl, DTT, spermidine, and Tris·Cl (pH 7.5), respectively, at 37°C. (B) Secondary structure of the substrate unit, comprising the corresponding substrate 1 and 2 units (Upper Left) and trans-DSL-2 (Upper Right). Tetraloop, receptor, and catalytic module are indicated with red, green, and purple, respectively. In the substrate unit, ligation site is indicated with an arrow. (Lower) Autoradiogram of the reaction of the designed RNAs and controls. Substrate unit: wt (designed substrate unit), Δ (a boxed A-U pair in the designed substrate unit is deleted), and insert (a 5′U-A3′ pair is inserted between the boxed A-U pair and the receptor, Upper). Catalytic unit: wt (designed trans-DSL-2) and insert (a 5′C–G3′ pair is inserted at a site of trans-DSL-2 indicated by an arrow). The lower and upper bands correspond to the substrate 1 and the ligated product, respectively. The bands are broad because of the 3′ heterogeneity of the transcription products. The product yield is the percentage fraction of the reacted substrate. The conditions were: 6, 1.5, and 1.5 μM trans-DSL-2, substrate 1, and substrate 2, respectively; 25, 50, and 30 mM MgCl₂, KCl, and Tris·Cl (pH 7.5), respectively, at 37°C for 2 h.

Assay of Base Specificity at 5′ End of Ligation Site. A cis-DSL-1S derivative containing a modified P1 (12 μM), S-1 RNA (10 μM), and an appropriate [α-³²P]nucleoside triphosphate (≈5 × 10⁷ cpm) were mixed in H₂O. The mixture was heated at 80°C for 5 min, then allowed to cool to 37°C. A 1/4 volume of the 5-fold concentrated reaction buffer for kinetic analyses was added to the mixture, and the mixtures were incubated at 27°C for 6 h, 32°C for 6 h, and then 37°C for 6 h. The reaction mixtures, stopped by adding an equal volume of the stop solution, were analyzed on 5% or 15% polyacrylamide denaturing gels with a Bio-Imaging Analyzer (BAS2500; Fuji).

Results and Discussion

To construct cis-DSL, a putative reaction site and a combinatorial library for selecting a catalytic modular unit were incorporated into a scaffold RNA (Fig. 1B). The reaction site comprised one RNA with a 5′ triphosphate and a second terminating with 2′,3′-OH groups; these were designed to form part of the P1 helix, so that the substrate RNA oligonucleotide (marked with yellow in Fig. 1B) could form five base pairs with an internal guide sequence (Fig. 1B) (14). A pool of 30 nucleotides with random sequence was inserted at the 5′ and 3′ strands in the P3a region in place of the original four nucleotides of scaffold RNA (L and R marked with purple in Fig. 1B) because it appeared from 3D modeling that the reaction site could be fully surrounded with a portion of 30 nucleotides attached to L or R region (Fig. 1B Right).

Two RNA pools consisting of random nucleotides at the 5′ and 3′ strands of the P3a region with 1.4 × 10¹⁴ different sequences were mixed and subjected to in vitro selection by employing two sets of substrate RNA and respective P1 region for the selection (Fig. 1B) (15). After four rounds of selection, the RNA pool with type 1 P1 and its cognate substrate (S-1) exhibited significant ligation activity (k_obs = 9.2 × 10^–4 per min). Fourteen ligated RNAs were subsequently cloned after the fifth round of selection. They had identical sequences at the 3′ strand of P3a with no mutations in the rest of the RNA (Fig. 1C Top). The resulting RNA was termed cis-DSL-01.

Cis-DSL-01 was converted to a shortened form (cis-DSL-1) by truncating its 3′ terminal tag region used for the selection (positions 116–140, which are shown as a gray bar in Fig. 1B). The truncation somewhat influenced catalytic efficiency of the ligase (Fig. 2A; see also Table 3, which is published as supporting information on the PNAS web site). The combination of cis-DSL-1 possessing type 1 P1 region and its corresponding substrate (S-1) exhibited highest activity (k_obs = 5.3 × 10^–3 per min) compared with the rest of the catalyst-substrate combinations (Fig. 2 A and Table 3), and specifically produced a 3′–5′ linkage at the ligated junction (Fig. 5, which is published as supporting information on the PNAS web site).

Fig. 2.

Structural requirement for cis-DSL ligases. (A) Self-ligation reactions of cis-DSL-01 and cis-DSL-1 with four different substrate–P1 combinations (see Fig. 1B). The reactions were carried out for 2 h under the standard conditions (see Materials and Methods). Rib and SubRib indicate bands corresponding to the ribozyme and the ligated product, respectively. (B) Structures of the receptor motifs in P3 (two at left) and the sequences at P2 region (three at right) in the derivatives of cis-DSL-1. The motifs and the sequences were substituted for the P2 region (boxed) and the 11-nt receptor (green), respectively, in Fig. 1B. (C) The activities of the cis-DSL-1 variants possessing mutation(s) in P1 loop and/or P2 region. The reactions were carried out under the standard conditions for 0, 2, and 18 h (see Materials and Methods). (D) The activities of the cis-DSL-1 variants possessing different of P1 loops and GNRA receptors. The reactions were carried out under the standard conditions for 0 and 18 h (see Materials and Methods).

One remarkable feature of cis-DSL-1 is the final yield of its ligated product (81%, Table 3), indicating that the majority of cis-DSL-1 folds into a catalytically active form. The yield is higher than those of previously reported ligase ribozymes (65–70%; refs. 16 and 17) and comparable to that of a self-splicing intron RNA (18), indicating that the designed scaffold allows little misfolding (13). Presumably, a catalytic module that least interferes with the folding process was selected. Thus, in this aspect, a self-folding RNA seems to be advantageous as a scaffold for constructing an enzyme.

To determine whether cis-DSL-1 folds into the designed structure, we constructed derivatives predicted to have defective tertiary interactions (Fig. 2B) (7, 13, 19–21). One or two mutations at the loop–receptor interaction and/or consecutive base-triples significantly lowered or abolished the activity, respectively, indicating that the loss of the interactions destabilized the folded 3D structure (Fig. 2C). The compensatory mutations that restored the P2 helix also restored the activity (Fig. 2C). The GAAA loop–11-nt-receptor interaction between the P1 and P3 regions was also replaced with other loop–receptor combinations (Fig. 2D). The variant cis-DSL-1 possessing B7.8 or C7.34 receptor motifs that prefer GUAA or GGAA to the rest of GNRA tetraloops, respectively, reacted most effectively with the corresponding loop (Fig. 2D) (20).

To improve the activity of cis-DSL-1, a second selection was performed from a newly constructed pool in which a mixture of nucleotides was substituted at each position of the catalytic unit, together with five completely randomized nucleotides on the opposite side (Fig. 6, which is published as supporting information on the PNAS web site; see also Fig. 1C Middle). After four rounds of selection, the cloned RNAs exhibited a consensus sequence that indicates 17 invariant positions in the catalytic unit (Fig. 1C Middle). Among the base substitutions, a simultaneous C-to-U change at positions 80 and 101 (green in Fig. 1C) was responsible for an ≈10-fold improvement in the activity (Fig. 6). Furthermore, a G-to-A substitution at position 54, which pairs with U76 (green in Fig. 1C), improved the activity ≈2-fold (Fig. 6). All variants containing the three substitutions exhibited activity one order of magnitude higher than cis-DSL-1: k_obs of the most active variant (cis-DSL-1S) (Fig. 1C), which was 0.12 per min with the optimized P1-substrate combination (S-19, see Table 4, which is published as supporting information on the PNAS web site), was 23-fold higher than that of cis-DSL-1 with the S-1 substrate. Cis-DSL-1S ligase accelerates the ligation reaction ≈10⁶-fold over the uncatalyzed reaction. This rate acceleration is comparable to those of previously selected RNA ligases producing 3′–5′ linkage, except class I ligase, which is 10³-fold faster than cis-DSL-1S (Table 5, which is published as supporting information on the PNAS web site; refs. 17 and 22–25).

Fig. 3A and Table 4 show that cis-DSL-1 and -1S are able to use A, C, G, or U at the 3′ end of the substrate RNA, as long as the template nucleotide is complementary. A trimolecular reaction system consisting of a substrate, a variant ribozyme lacking the first nucleotide of cis-DSL-1S, and a nucleoside triphosphate serving as an alternative for the first nucleotide (step 1 of Fig. 3B Upper) demonstrated that the ligase can use all four nucleotides at the 5′ end as well as at the 3′ end (Fig. 3 B and C). These results indicate that the selection provided a versatile catalytic unit. [Note that Sub1-X-ribozyme products were also detected, indicating that cis-DSL-1S can carry out two successive nucleotidyl addition reactions (step 2 of Fig. 3B, see also Fig. 7, which is published as supporting information on the PNAS web site)]. In contrast, the previously reported ligases with the exception of class I can use only limited combinations of the two substrate nucleotides at the reaction site (14, 24–26).

The ligase was transformed into two molecules, a trans-acting catalytic RNA (trans-DSL) and a substrate unit consisting of two RNAs, on the basis of its structural organization (Fig. 1 A). DSL-1S was dissected into a catalytic RNA consisting of the P3 region (trans-DSL-1) and a substrate unit consisting of the P1 region by replacing the triple-helical scaffold with a GAAA loop–11-nt-receptor interaction (12) (Fig. 1 A). The substrate RNAs were designed to bind specifically to the designed enzyme by adjusting the relative positions of the reaction site and the catalytic modular unit. The resulting enzyme ligated the substrates efficiently and specifically, whereas two mismatched substrates reacted very weakly or not at all (Fig. 4A). Accordingly, another set, trans-DSL-2, was designed and constructed successfully by swapping one set of loop–receptor units in trans-DSL-1 (Fig. 4B and Fig. 8, which is published as supporting information on the PNAS web site). This finding indicates that the activity of trans-DSLs accurately reflects the designed conformations. The two sets of constructions demonstrate that the catalytic modular unit serves as an installable central device for building a variety of RNA machinery.

The modular engineering of DSL indicates that its further evolution is feasible (Fig. 4). For example, it may be possible to engineer trans-DSL to conduct the reaction more efficiently by attempting further in vitro selection and/or by adding extra components for enhancement of the activity. Trans-DSL-2 may also be converted to a versatile ligase that performs the reaction on a canonical double-stranded RNA by replacing its two GAAA loops with receptors that nonspecifically bind to double-stranded RNA. In principle, new enzymes for conducting a variety of reactions may also be synthesized and developed by employing the present or a newly designed scaffold as a mother RNA. Likewise, artificial receptor RNAs that specifically recognize structural motifs of DNA, RNA, or protein might also be constructed. In the field of nano-bioscience and engineering, these constructs may serve as useful components for constructing novel RNA machines.