RNA recognition by designed peptide fusion creates "artificial" tRNA synthetase

Abstract

The genetic code was established through aminoacylations of RNA substrates that emerged as tRNAs. The 20 aminoacyl-tRNA synthetases (one for each amino acid) are ancient proteins, the active-site domain of which catalyzes formation of an aminoacyl adenylate that subsequently reacts with the 3′ end of bound tRNA. Binding of tRNA depends on idiosyncratic (to the particular synthetase) domains and motifs that are fused to or inserted into the conserved active-site domain. Here we take the domain for synthesis of alanyl adenylate and fuse it to ``artificial'' peptide sequences (28 aa) that were shown previously to bind to the acceptor arm of tRNA^Ala. Certain fusions confer aminoacylation activity on tRNA^Ala and on hairpin microhelices modeled after its acceptor stem. Aminoacylation was sensitive to the presence of a specific G:U base pair known to be a major determinant of tRNA^Ala identity. Aminoacylation efficiency and specificity also depended on the specific peptide sequence. The results demonstrate that barriers to RNA-specific aminoacylations are low and can be achieved by relatively simple peptide fusions. They also suggest a paradigm for rationally designed specific aminoacylations based on peptide fusions.

The development of the genetic code was essential for the transition from the putative RNA world to the theatre of proteins (1, 2). The rules of the modern code are determined in aminoacylation reactions, where each amino acid is linked to the tRNAs bearing the anticodon triplets specific for that amino acid (3). The tRNA itself is comprised (typically) of 76 nucleotides ending at the 3′ end with the universal trinucleotide CCA₇₆, where A₇₆ is the amino acid attachment site (Fig. 1A). In three dimensions, the cloverleaf secondary structure is rearranged into two distinct domains fixed at right angles to form an L shape. One domain contains the 12-bp acceptor-TψC minihelix with the amino acid attachment site at the 3′ end. The other domain, at right angles, contains the anticodon at a distance of ≈75 Å from the amino acid attachment site (4–6).

Fig. 1.

Domain organizations of E. coli tRNA^Ala and AlaRS. (A) Sequence and cloverleaf structure of E. coli tRNA^Ala (Left) and hairpin structure of alanine microhelix (Right). The G3:U70 base pair is highlighted in bold and the acceptor stem sequence is designated by a shaded box (48). (B) Schematic diagram of the full-length E. coli AlaRS where each functional domain is specified (36, 49). The two functional modules used as negative controls (fragment 368N) and positive controls (fragment 461N) are also delineated.

The two domains of tRNA are thought to have arisen independently, with the minihelix being an outgrowth of the earliest substrates for aminoacylation (7–10). Significantly, over half of the synthetases have been shown to charge RNA oligonucleotide substrates based on the minihelix or smaller pieces such as a 7-bp microhelix hairpin that is designed after the tRNA acceptor stem (see below; e.g., refs. 11–16). RNA determinants for specific aminoacylation of microhelices have been referred to as a second genetic code or an operational RNA code (1, 17). The second genetic code may have been the precursor to the modern genetic code (18, 19).

Alanyl-tRNA synthetase (AlaRS) is a particularly striking example of an enzyme that catalyzes robust aminoacylation of RNA oligonucleotide substrates based on the acceptor stem of tRNA^Ala (Fig. 1 A; refs. 11 and 20). A single G:U wobble base pair at position 3:70 in the acceptor stem is a critical determinant for aminoacylation (21, 22). Transfer of G3:U70 into nonalanine tRNA frameworks confers alanine acceptance on them (11, 21, 22). Minor identity elements in the acceptor stem include the A73 ``discriminator'' nucleotide and a G2:C71 base pair (23–25). In contrast, the anticodon makes no contribution to the specificity or efficiency of recognition of tRNA^Ala. Indeed, the enzyme makes no physical contact with the anticodon (26). Thus, the determinants that specify alanine acceptance are completely segregated from the triplets that encode alanine. In terms of RNA recognition, AlaRS may be representative of an early synthetase.

The catalytic domains of aminoacyl tRNA synthetases are divided into two classes of 10 enzymes each, based on the two distinct architectures of the active sites (27, 28). These domains are ancient and, for virtually all synthetases, basically fixed throughout evolution. Thus, these catalytic domains were established at the time of the last common ancestor to the tree of life (2, 29). The domains contain the site for amino acid activation (aminoacyl adenylate synthesis) and for interaction with the 3′ end of tRNA. Other motifs and domains are fused to or inserted into the active-site domains. These elements are essential for specific RNA recognition and, in general, are idiosyncratic to the synthetase and not conserved among enzymes of the same class (30).

From the foregoing perspective, the development of RNA-specific aminoacylations came from idiosyncratic acquisitions of RNA-binding elements by preexisting primordial domains for adenylate synthesis. To test the feasibility of this scenario, we exploited previous work in which a short peptide was selected by phage display methods (31) to discriminate a G3:U70 base pair from other possibilities (G:C, U:G, and I:U) in microhelix^Ala (32). The selected peptide, MFβ2, efficiently bound microhelix^Ala and discriminated G3:U70 with high selectivity (Fig. 2A). Recognition depended on the presence of the unpaired 2-amino group of G3 that projects into the minor groove. This same amino group is critical for recognition by native AlaRS.

Fig. 2.

Sequence and structure of the selected peptide MFβ2 and assorted variants. (A) The peptide contains two 9-aa RGG motifs (black; ref. 42) that flank the10-aa sequence selected (by phage display methods) to recognize specifically a G:U base pair (red; ref. 32). The standard one-letter abbreviations for amino acids are used. (B) Helical wheel representation of the predicted helical structure corresponding to the parental (``wild-type'')MFβ2 sequence (Left) and mutated sequences used in this study (Right, mutations are circled in red). In the wild-type sequence on the left, red residues were shown important for specific recognition of the unpaired 2-amino group of G3 that projects into the minor groove.

In this work we fused versions of the MFβ2 peptide to the catalytic domain of the 875-aa Escherichia coli AlaRS (33). A fragment of the N-terminal 368 aa (fragment 368N) encodes the domain for adenylate synthesis and has determinants for contacting the 3′ end of tRNA^Ala (34, 35; Fig. 1B). Fragment 368N lacks polypeptide determinants needed for recognition of the G3:U70 base pair and consequently for binding tRNA^Ala or microhelix^Ala (36, 37). (These determinants are contained in a 93-aa segment that extends from T369 to D461.) Peptides based on MFβ2 were joined through a linker to 368N and tested for their capacity for aminoacylation and for specificity of base-pair recognition at the 3:70 position. The results show that base pair-sensitive aminoacylation systems based on peptide fusions are accessible and suggest a paradigm for the rational design of specific aminoacylation systems.

Materials and Methods

Preparation of Artificial AlaRSs. AlaRS fragments 461N and 368N and the different fragment-MFβ2 fusions were overexpressed and purified in E. coli. Genes were cloned into expression vectors pQE70 and pQE30 (Qiagen, Courtaboeuf, France) fused to a C- or N-terminal His₆ tag, respectively. Induction (1 mM isopropyl β-d-thiogalactoside) was performed for 3 h at 37°C under moderate shaking (180 rpm). AlaRS variants were purified on a nickel-nitrilotriacetic acid column according to the Qiagen protocol. Proteins were quantified by UV measurement at 280 nm by using a calculated extinction coefficient ε = 48,220 M·cm^–1 for fragment 368N and for the various fusion proteins (which have no chromophores that absorb at 280 nm). Proteins were stored at –20°C in 50 mM phosphate buffer (pH 7.2)/150 mM KCl/50% glycerol/10 mM 2-mercaptoethanol. All AlaRS constructs yielded one band on SDS/PAGE and from an overloading of material the purity was estimated as >99%.

Fragment 461N with an N-terminal His₆ tag had an activity reduced compared with the C-terminal His₆-tagged protein. However, all N-terminal-tagged 368N-MFβ2 fusions were inactive. Thus, C-terminal His₆-tagged 461N was used as a positive control to compare with N-terminal-tagged 368N and fusions of 368N with peptides.

The DNA sequence coding for 368N was cloned into pQE30, between SpHI and HindIII sites. The linker and the different peptide sequences were subsequently fused by using overlapping oligonucleotides with specific restriction sites. The restriction sites led to minor sequence variations in the fusion proteins: substitution of R³⁶⁸ in 368N with the sequence K³⁶⁸L in the fusion proteins and insertion of the dipeptide sequence Glu-Phe between the linker and the MFβ2-encoding peptides (Fig. 3B). Finally, all proteins contain a C-terminal sequence AAAKLN encoded by the plasmid pQE30. In a test experiment, this hexapeptide extension did not have an effect on the activity.

Fig. 3.

Schematic diagrams and sequences of fusion proteins. (A) Functional domains are indicated with different colors: 368N is shown in gray, linker is shown in green, RNA-binding motifs (RGG boxes) are shown in red, and the 10-aa recognition sequence from MFβ2 is shown in blue. (B) Corresponding primary sequences of fusion peptides of 368N-MFβ21 and 368N-MFβ25.

Preparation and Aminoacylation of RNAs. RNA microhelices were chemically synthesized on a Pharmacia (Piscataway, NJ) gene assembler synthesizer as described and subsequently purified on a 16% polyacrylamide gel. Concentrations of RNAs were determined by absorbance at 260 nm using calculated extinction coefficients based on base compositions.

After optimization, aminoacylation of alanine microhelices variants was performed at 37°C in 25 mM Tris·HCl (pH 7.5)/10 mM MgCl₂/5 mM ATP/75 mM NaCl/1 mM dithioerythritol/0.5 μg/μl BSA/0.01 units/μl inorganic pyrophosphatase (Sigma–Aldrich)/50 μM l-[³H]alanine (Amersham Pharmacia, Orsay, France) and appropriate amounts of RNA microhelices. Before aminoacylation, transcripts were renatured in H₂O by heating at 85°C for 90 s and fast cooling on ice prior to the addition of MgCl₂ (to a final concentration of 50 μM). Aminoacylation reactions were initiated by the addition of appropriate amounts of enzyme diluted in 100 mM Hepes-KOH (pH 7.4)/10% glycerol/1 mM dithioerythritol/5 μg/μl BSA. Reactions were stopped (after 2.5-, 5-, 10-, and 20-min incubations at 37°C) by the addition of 5% trichloroacetic acid and treated in the conventional way by filtration of precipitates on Whatman 3 MM paper (38). Kinetic constants (k_cat and K_m) were derived from Lineweaver–Burk plots. RNA concentrations extended from 20 to 100 μM, and protein concentrations ranged from 30 nM for fragment 461N to 5 μM for 368N and for the fusion proteins 368N-MFβ21 and 368N-MFβ25. Kinetic plots and plateaus represent an average of at least two independent experiments. Values of k_cat/K_m for replicate experiments varied by at most 10%.

Comments on Optimization of Aminoacylation Conditions. In the first set of experiments, 368N-MFβ25 was chosen to optimize the aminoacylation conditions. Under starting conditions (25 mM Tris·HCl, pH 7.5/75 mM NaCl/10 mM MgCl₂/1 mM dithioerythritol) this fusion charged up to 1.2% of the RNA substrate present in the assay. (This level was low compared with the plateaus obtained in the positive control with fragment 461N that reach 80%.) The low activity was explicitly associated with the presence of the peptide MFβ2 in the fusion, because the negative control (fragment 368N) did not show any charging activity (data not shown).

Four parameters then were varied to optimize the activity of 368N-MFβ25: the presence of inorganic pyrophosphatase and BSA, concentration of alanine (25–100 μM), and temperature. Three of these parameters significantly affected the overall rate. Thus, the presence of 0.01 units/μl inorganic pyrophosphatase together with 0.5 μg/μl BSA increased the rate by 3-fold, whereas raising the temperature from 25 to 37°C increased the rate by 5-fold. Thus, by combining the two sets of variables, a gain of 15-fold in aminoacylation rate and ≈10-fold in yield was achieved over that realized with the starting conditions.

Measurements of Affinities for Microhelix^Ala. Affinities for microhelix substrates were measured by polyacrylamide coelectorphoresis using the protocol described in ref. 39. Samples of radiolabeled microhelix (0.1 pmol of 5′ ³²P-labeled microhelix were mixed with 20 pmol of E. coli total tRNA in a final volume of 20 μl), 10% glycerol, 50 mM Tris·HCl (pH 7.5), and 10 mM MgCl₂ were electrophoresed for 90 min (70 V at 4°C) through a 10% acrylamide/bisacrylamide gel (37.5:1; 200 × 100 × 1.5 mm³) containing increasing concentrations of fragment 461N or fusion proteins 368N-MFβ21 and 368-MFβ25 (25–50 μM) in Tris buffer (100 mM Tris base/100 mM boric acid). Gels were dried and analyzed on a FUJIX BAS 2000 bioimaging analyzer (Raytest, Courbevoie, France) with work station 1.1 software.

Results

Rational Design of Fusion Proteins. The overall architecture of E. coli AlaRS is based on four modular units responsible for alanyl-adenylate synthesis, tRNA recognition, tRNA-dependent editing, and assembly of the tetrameric structure (Fig. 1B). Although the native enzyme is tetrameric, fragment 368N is a monomer that catalyzes synthesis of alanyl adenylate with the same efficiency as that of the native enzyme (34, 40). Fragment 368N has no activity for aminoacylation or tRNA binding. Aminoacylation of tRNA^Ala requires the segment extending from T369 to D461 to give fragment 461N (36). All contacts with the acceptor stem are encoded by 461N. As a consequence, 461N aminoacylates microhelix^Ala with the same catalytic efficiency as does the native enzyme. [The native enzyme has a smaller K_m for charging tRNA^Ala than does 461N because of contacts with tRNA^Ala that lie outside the acceptor arm that are contributed by portions of AlaRS that lie on the C-terminal side of D461 and therefore are missing from 461N (40).] Thus, in our work we sought to replace the 93-aa segment from T369 to D461, which is needed for acceptor helix contacts. This replacement was a short, ``artificial'' peptide that was selected (from a phage display library) for its ability to recognize the acceptor arm of tRNA^Ala in a G3:U70-dependent way (32).

Peptide MFβ2 consists of a central decapeptide flanked with two RGG-containing motifs (KRGGKRGGK) that enhance general RNA binding (refs. 41 and 42; Fig. 2 A). The peptide forms a specific complex with microhelix^Ala with a K_d of 300 nM (32). The decapeptide central region was strongly predicted to form an α-helix. By mutational analysis 4 of the 10 residues were demonstrated important for specific recognition of the G3:U70 base pair (Fig. 2B). These residues are S2, A4, E6, and N9. Three of these (S2, E6, and N9) are on the same side of the helix.

Using a linker (four repeats of a GS dipeptide), we fused 368N to MFβ2-containing peptides. Altogether, seven fusion proteins (368N-MFβ21 to 368N-MFβ27) were constructed. All contained fragment 368N, the linker, and the G:U-specific decapeptide MFβ2. The number and localization of the RGG boxes neighboring the selected decapeptide were varied to generate the seven different constructs (Fig. 3).

Aminoacylation with Fusion Constructs. By using optimized conditions, the seven fusion constructs were tested for aminoacylation of microhelix^Ala and compared (Fig. 4). The first construct, 368N-MFβ25, had one RNA-binding motif located on the N-terminal side of the G-U-recognizing peptide and had significant activity for charging of microhelix^Ala. In contrast, under the same conditions no aminoacylation was observed with fragment 368N (as expected). Each of the six other constructs also showed activity, with the different combinations of RGG motifs leading to more or less efficient enzymes. The most efficient fusion constructs were 368N-MFβ21 and 368N-MFβ22, whereas the least efficient were 368N-MFβ26 and 368N-MFβ27. The latter fusion was virtually inactive. The most efficient constructs corresponded to fusions with four (368N-MFβ21) or three (368N-MFβ22) RGG boxes, whereas low activities were associated with 368N-MFβ26 and 368N-MFβ27, which had one or no RGG box, respectively. However, 368N-MFβ25 had only one RGG box and showed a relatively strong activity. This construct charged microhelix^Ala more effectively than did either 368N-MFβ23 or 368N-MFβ24 that each hold two RGG motifs.

Fig. 4.

Aminoacylation of microhelix^Ala by fusion proteins. Sequences of the different fusion constructs are shown in Fig. 3. The activity of one representative of each group is shown: 368N-MFβ21, most efficient; 368N-MFβ25, moderately active; 368N-MFβ27, low activity; and 368N, no activity. (Inset) Activity of all other constructs (368N-MFβ22, 368N-MFβ23, 368N-MFβ24, and 368N-MFβ26) compared with 368N-MFβ21. Aminoacylation reactions were run at pH 7.5 and 37°C with 30 μM microhelix^Ala and 5 μM of each fusion protein.

Thus, the presence of only one RGG motif in the artificial AlaRS is enough to form an active complex with microhelix^Ala. To act efficiently as an anchor of the RNA, it seems that an RGG motif has to be localized at the N-terminal side of the critical decapeptide (compare 368N-MFβ25 with 368N-MFβ26). Interestingly, the most active construct, 368N-MFβ21, encoded the original 28-mer selected in the earlier study (32).

Binding and Catalytic Parameters. To put these results on a quantitative footing, two of the most active fusion constructs were chosen for further investigation. Affinity coelectorphoresis of microhelix^Ala binding and the microhelix^Ala concentration dependence of aminoacylation were investigated for 368N-MFβ21, 368N-MFβ25, and the control fragment 461N (Table 1). Apparent kinetic parameters for the two fusion proteins yielded k_cat/K_m values that were reduced 1/600 (368N-MFβ21) and 1/1,500 (368N-MFβ25) compared with fragment 461N. These values correspond to differences in the apparent free energy of activation of 4–5 kcal·mol^–1 compared with the free energy of activation for 461N. Interestingly, most of the difference from fragment 461N lies in the k_cat parameter, unsurprisingly suggesting that positioning of the 3′ end of microhelix^Ala is not as optimal when bound to either of the fusion proteins as when bound to the natural active site. In contrast, K_m values for the fusion proteins are comparable to that for 461N. In addition, microhelix^Ala binds (K_d values) more tightly to the two fusion proteins than to 461N. The lower K_d values for the artificial enzymes are possibly related to the nonspecific RNA-binding components (RGG motifs) that are built into the MFβ2 peptides that were fused to 368N.

Table 1. Binding and aminoacylation parameters of microhelix^Ala with 461N and fusion proteins 368N-MFβ21 and 368N-MFβ25.

Construct	Kd, μM	Km, μM	kcat, min-1	kcat/Km, 10-3	Loss in efficiency
461N (C-terminal 6-His)	3.100	100	100	1,000	1
Fusion proteins
368N-MFβ21	0.075	150	0.25	1.66	1/600
368N-MFβ25	1.500	150	0.1	0.66	1/1,500

Kinetic parameters were determined at pH 7.5, 37°C. K_d values were determined at 4°C.

Specificity of Microhelix Aminoacylation. Three constructs (368NMFβ21, 368N-MFβ22, and 368N-MFβ25) were tested for their capacity to discriminate between a G3:U70 base pair versus G:C, U:G, and I:U pairs (Fig. 5A). Thus, RNA microhelices differing only at the 3:70 position were used as substrates. Of the three constructs, 368N-MFβ25 was the most efficient for discriminating between G:U and the other three base pairs. Indeed, all three alternatives, G:C, U:G, and I:U, had the same reduction in activity with 368N-MFβ25. The other two constructs, 368N-MFβ21 and 368N-MFβ22, were more efficient at charging the G3:U70-containing substrate but also had higher activities with substrates harboring any of the other three base pairs. Still, these fusion constructs were most efficient with native microhelix^Ala.

Fig. 5.

Specificity of aminoacylation of microhelix^Ala with various substitutions at the critical 3:70 position, which is a G:U pair in most tRNA^Alas throughout evolution (50). Aminoacylation efficiencies were calculated as pmol·min^–1 of aminoacylated microhelix synthesized during a 2.5-min incubation. (B) Discrimination at the 3:70 position by mutant forms of 368N-MFβ25, with mutations placed at specific positions in the 10-aa recognition motif of the free MFβ2 peptide. Mutations correspond to those shown previously to diminish specificity of binding of MFβ2 to microhelix^Ala (32).

Mutational Analysis of Fusion Peptide. To test whether the efficiency of aminoacylation depended on the central decapeptide sequence that contained the putative determinants for binding microhelix^Ala, substitutions were introduced at three of the four positions previously shown to be important for recognition of G3:U70. These residues are S2, E6, and N9 (Fig. 2B). S2 was replaced by leucine. (Introduction of an alanine at this position led to the production of a nonsoluble fusion protein.) E6 and N9 were replaced with alanines.

Different combinations were tested in the context of the sequence of 368N-MFβ25: single substitutions at position 2 (S2L) or 9 (N9A), double substitutions at positions 2 and 9 (S2L/N9A), and the triple substitution (S2L/E6A/N9A). Aminoacylation activity progressively dropped as more substitutions were included (Fig. 5B). For example, charging efficiency dropped almost 2-fold when either the S2L or N9A substitution was introduced and 5-fold when both substitutions were combined (S2L/N9A). With the triple substitution, aminoacylation efficiency was reduced 10-fold. Both single mutants lost specificity, showing an increase in efficiency of charging of G3:C70-, U3:G70-, or I3:U70-containing microhelices relative to charging of native microhelix^Ala. This loss of specificity was seen also with the double and triple substitutions, each of which had a further decline in aminoacylation activity. Thus, the charging activity and specificity of the fusion 368N-MFβ25 construct strongly depends on the sequence of the previously selected central decapeptide sequence.

Discussion

``Fusing'' to 368N a 93-aa natural AlaRS peptide segment from T369 to D461 confers RNA-binding activity sufficient for specific aminoacylation of microhelix^Ala or tRNA^Ala (36, 37). In previous work, large nonspecific RNA-binding domains of 175 and 228 aa (C-terminal domain of Saccharomyces cerevisiae Arc1p and N-terminal appended domain of S. cerevisiae glutaminyl-tRNA synthetase, respectively) were fused to fragment 368N (43). Significantly, the resulting chimeric proteins catalyzed aminoacylation of microhelix^Ala. Unsurprisingly, no preference for the critical G3:U70 pair was seen (indeed, A3:U70 was somewhat preferred over G3:U70). These large protein fusions gave early support to the idea that a simple fusion to a domain for adenylate synthesis could result in the acquisition of aminoacylation function. Not clear was whether much smaller motifs (<75–100 aa) could also confer aminoacylation and, if so, whether even a modest specificity could be achieved. Here peptide motifs of just 28 or 15 aa, fused via a linker of 8 aa to 368N, generated an enzyme that catalyzed aminoacylation with a clear preference for a G3:U70 pair in a microhelix substrate (368N-MFβ21 and 368N-MFβ25, respectively).

Importantly, the two components, adenylate synthesis and specific RNA binding, were generated independently. In particular, the MFβ2 peptide framework used in this work was generated through reiterative selections imposed on a phage display library (31). Selection was specifically to obtain a sequence that bound with strong preference to G3:U70 in the context of microhelix^Ala (32). Thus, the results are consistent with the idea that early tRNA synthetases arose from small, idiosyncratic RNA-binding elements being fused to domains for adenylate synthesis. These RNA-binding elements might have developed originally to bind and protect ribozymes (to give early ribonucleopeptides or ribonucleoproteins; refs. 44–47). The fusions of RNA-binding peptides to domains for adenylate synthesis may have been the first step in developing protein-based synthetases that overcame the ribozyme-based system of aminoacylation.

The 28-aa MFβ2 peptide obtained in the earlier selections bound to microhelix^Ala in a 1:1 complex with a K_d of 300 nM at pH 7.5, 4°C, by using the same affinity coelectorphoresis procedure adopted for the present work. For the 368N-MFβ21 peptide fusion, the resulting K_d was of a similar order (75 nM; see Table 1). Thus, most of the RNA-binding energy for the fusion construct almost certainly came from the 28-aa peptide element. This result suggests that the cost of a fusion, in terms of lost RNA-binding activity compared with the free peptide, is inconsequential. More significant, however, is the reduction in specificity. In a competition assay, U3:G70-substituted microhelix^Ala was the best competitor for binding of the MFβ2 free peptide to microhelix^Ala. In particular, the peptide bound microhelix^Ala by 20- to 25-fold over U3:G70-containing microhelix^Ala. The specificity for G:U versus U:G at the 3:70 position of microhelix^Ala was reduced ≈5-fold for 368N-MFβ25 relative to the free MFβ2 peptide. This reduction probably reflects the serendipitous weak nonspecific RNA-binding determinants in fragment 368N that affect the context for RNA binding by the fusion peptide (35). Further selections could be placed on the fusion proteins to ameliorate these effects. For example, the disposition of the fusion peptide relative to the active site within 368N might be most affected by variations in the length and character of the linker sequence.