Selection of RNA Aptamers That Are Specific and High-Affinity Ligands of the Hepatitis C Virus RNA-Dependent RNA Polymerase

Abstract

In order to find small RNA molecules that are specific and high-affinity ligands of nonstructural 5B (NS5B) polymerase, we screened by SELEX (systematic evolution of ligands by exponential amplification) a structurally constrained RNA library with an NS5BΔC55 enzyme carrying a C-terminal biotinylation sequence. Among the selected clones, two aptamers appeared to be high-affinity ligands of NS5B, with apparent dissociation constants in the low nanomolar range. They share a sequence that can assume a stem-loop structure. By mutation analysis, this structure has been shown to correspond to the RNA motif responsible for the tight interaction with NS5B. The aptamers appeared to be highly specific for the hepatitis C virus (HCV) polymerase since interaction with the GB virus B (GBV-B) NS5B protein cannot be observed. This is consistent with the observation that the activity of the HCV NS5B polymerase is efficiently inhibited by the selected aptamers, while neither GBV-B nor poliovirus 3D polymerases are affected. The mechanism of inhibition of the NS5B activity turned out to be noncompetitive with respect to template RNA, suggesting that aptamers and template RNA do not bind to the same site. As a matter of fact, mutations introduced in a basic exposed surface of the thumb domain severely impaired both the binding of and activity inhibition by the RNA aptamers.

Hepatitis C virus (HCV) is a positive-strand RNA virus of the Flaviviridae family, which affects more than 3% of the world population. About 80% of the infected patients develop liver cirrhosis and, in some cases, hepatocarcinoma (13). Besides interferon-based treatments, effective therapies to counteract this important public health problem are still lacking.

The HCV positive-strand RNA viral genome contains a single open reading frame flanked by 5′- and 3′-untranslated regions. The open reading frame encodes a polyprotein of ca. 3,010 amino acids which is processed into at least 10 mature proteins (C, E1, E2, p7, NS2, NS3, NS4A, NS4B, NS5A, and NS5B) by both host signal peptidases and viral proteases (10, 31). In analogy with other positive-strand RNA viruses, HCV replication is supposed to proceed through the synthesis of negative-strand RNA, which is in turn used as a template for the production of genomic RNA molecules. A virally encoded RNA-dependent RNA polymerase (RdRp) is considered one of the key enzymes involved in both steps of HCV replication and is, therefore, a primary target for the development of antiviral drugs. The HCV RdRp activity has been localized in the 66-kDa nonstructural 5B (NS5B) protein (2). In vitro, purified NS5B has been shown to be a processive enzyme (39) capable of transcribing the full-length HCV genome (28) essentially via a snap-back mechanism. Recent studies indicate that NS5B does direct de novo replication, requiring neither an exogenous primer nor a snap-back priming event on a variety of RNA templates (22, 30, 37, 43, 44). The lack of specificity toward the HCV RNA genome suggests that NS5B corresponds to the elongation factor of the HCV replication complex: interaction with viral and/or cellular partners may be required in order to achieve specific recognition of RNA sequences or structures functioning as cis-acting signals and for the initiation of RNA synthesis. Alternatively, RNA structures directly recognized by NS5B might only be formed upon binding of other factors to the HCV genome and may, therefore, not exist in the in vitro-produced RNA templates.

The full-length purified enzyme has a very poor catalytic activity (27, 28, 39). Deletions of the C-terminal membrane localization signal (42) have allowed the production of proteins with enhanced solubility and activity (14, 39; L. Tomei et al., unpublished results), but the polymerase efficiency in vitro still remains manyfold lower than expected compared to other polymerases. Carroll et al. (8) demonstrated that only a small fraction of the purified protein is engaged in productive polymerase-RNA complexes. Either nonspecific binding to the RNA or secondary structures in the RNA template might account for the low amounts of active complexes.

As an approach to the dissection of the RNA-enzyme interaction, we have used the SELEX (systematic evolution of ligands by exponential enrichment) procedure (12, 41) to screen for specific and high-affinity RNA ligands of the HCV RdRp. This methodology has been successfully used to select RNA binders directed against various proteins that are (23, 24, 38) or are not (3, 15, 34) nucleic acid-binding proteins. In most cases the selected aptamers turned out to be potent and selective inhibitors of the target proteins, making this procedure a valuable tool to develop inhibitors of protein functions (23) or to dissect protein functions either in vitro (7, 20, 25) or in vivo (4, 9, 35).

In this study, we describe the characterization of RNA aptamers that bind with high affinity and specificity the HCV RNA-dependent RNA polymerase and that are potent inhibitor of its in vitro activity. The selected RNA molecules interact with the enzyme at a site distinct from the template/primer-binding site that is located on a solvent exposed surface of the thumb domain.

MATERIALS AND METHODS

Recombinant plasmids and protein purification.

The plasmid pT7-NS5BΔC55 contains the HCV-BK sequence from nucleotides (nt) 7600 to 9207 coding for an NS5B protein lacking of the C-terminal 55 amino acids into the pT7-7 expression vector. A single point mutation was introduced by PCR in the NS5B-ΔC55 sequence to construct the plasmid pT7-NS5BΔC55-R498E. pT7-NS5BΔC55bio codes for an NS5B-ΔC55 protein carrying the biotinylation sequence GGGLNDIFEAQKIEWH (33) at its C terminus. pT7-GB/NS5BΔC23 encodes a GBV-B NS5B protein lacking the C-terminal 23 amino acids (residues 1 to 567).

Expression in Escherichia coli BL21(DE3) (36) and purification of the HCV NS5B and the GBV-B NS5B proteins were carried out essentially as already described (39). Depending on the specific protein, buffers at different pH values were used (NS5BΔC55 and ΔC55bio, pH 7.5; ΔC55-R498E, pH 7.0; GBNS5BΔC23, pH 8.0). For expression of NS5BΔC55bio, BL21(DE3) cells contained the plasmid pACYC184 (33). Growth and induction were performed in the presence of 50 μM biotin in standard Luria-Bertani medium. For the purification of NS5BΔC55bio, the lysis buffer and the chromatographic buffer used for the heparin-Sepharose column contained 200 and 50 μM biotin, respectively. The full-length NS5B protein was purified from Sf9 cells infected with a recombinant baculovirus as already described (39).

In vitro selection of RNA aptamers.

The SELEX procedure was carried out essentially as described previously (17). The initial template was generated by 10 cycles of amplification of 1 nmol of single-stranded DNA library SSL25/10 with 10 nmol each of primer T7P2 (5′-GGGAAGCTTAATACGACTCACTATAGGGATGCTTCGGCATCCC-3′) and primer 3′RSP.2 (5′-CCCAAGCTTACGTACCGCCGAAGCGGTAC-3′). The amplified DNA was purified from a 3% agarose gel and transcribed with T7 RNA polymerase in the presence of 20 μCi of [α-³²P]GTP as tracer nucleotide. Contaminant DNA was digested with 10 U of RNase-free DNase I (Boehringer), and the RNA was purified on an 8% denaturing gel. Denaturation-renaturation of the purified RNA was performed in a buffer containing 10 mM Tris-Cl (pH 7.5), 100 mM KCl, and 1 mM MgCl_2. After 10 min of incubation at 65°C, the RNA solution was allowed to reach 23°C and adjusted to 50 mM Tris-Cl (pH 7.5)-100 mM KCl-50 mM NaCl-5 mM MgCl_2.

In all of the selection cycles, the RNA was always in molar excess over the protein. M-280 streptavidin Dynabeads (Dynal) were coated with 100 pmol of purified NS5BΔC55bio by 20 min of incubation at 23°C in 50 mM Tris-Cl (pH 7.5)-100 mM KCl-50 mM NaCl-5 mM MgCl₂ (5B-selection buffer), followed by four washes in 5B-selection buffer. RNA (0.8 to 1.5 nmol) was incubated in 250 μl of 5B-selection buffer with NS5B-ΔC55bio-coated beads for 15 min at 23°C on a rotating wheel. The selected RNA was eluted with a buffer containing 50 mM Tris-Cl (pH 7.5), 5 mM EDTA, 1.5% sodium dodecyl sulfate, 300 mM NaCl, and 1.5 mg of proteinase K/ml and then reverse transcribed with 500 pmol of primer 3′RSP.2 and 25 U of AMV-RT (20 min, 56°C). After phenol extraction and ethanol precipitation, the cDNA was used for the PCR with 500 pmol of T7P2 and 3′RSP.2 primers (30 cycles, annealing temperature of 60°C). The DNA template was purified and transcribed as described above, and the resulting RNA pool was used for the successive rounds of selection. Before the incubation with NS5BΔC55bio-coated beads, the RNA was incubated with M280 beads to eliminate molecules nonspecifically retained on the beads. PCR-amplified DNA from the fourth round of selection was cloned into pUC19 and sequenced.

Gel retardation and filter-binding assays.

Binding reactions were carried out with 20 μl of buffer containing 20 mM Tris-HCl (pH 7.0), 1 mM dithiothreitol (DTT), 0.25 U of RNasin/μl, 100 ng of bovine serum albumin/μl, 250 mM NaCl, 0.03% n-octyl-β-d-glucopyranoside, 5 mM MgCl₂, and 2% glycerol. In the gel shift experiments, the amounts of purified NS5B specified in the figures were incubated with 0.12 nM labeled RNA aptamers (20,000 Cerenkov counts) for 15 min at 23°C. At the end, 5 μl of 20% Ficoll was added, and the complexes were analyzed by 6% polyacrylamide gel electrophoresis-0.25× Tris-borate-EDTA run at 4°C.

Apparent K_d values were measured by filter-binding assays where increasing amounts of polymerase were incubated with 0.12 nM labeled RNA aptamers (20,000 Cerenkov counts), as described above. The binding reaction mixtures were filtered through MultiScreen-HA 96-well filter plates (Millipore). After extensive washes with binding buffer, scintillation counting quantitated the radioactivity retained on the filters. Apparent K_d values were calculated through parametric fitting of the experimental data with a theoretical curve. Before each binding experiment, the labeled RNA molecules were denatured and renatured as described above.

For mutational analysis, RNA oligonucleotides carrying mutated sequences of the B.2 aptamer were synthesized by EuroGenetec.

Structure-specific enzymatic probing.

Enzymatic probing of the SLII molecule was performed by partial digestion with RNase T₁ (0°C and 37°C, 0.1 U; Boehringer) and RNase A (0 and 37°C, 12.5 to 25 pg/μl). Digestion of 5′-end-labeled RNA (30,000 Cerenkov counts) was carried out in a 5-μl reaction volume containing RNase buffer (30 mM Tris-HCl, [pH 7.5], 20 mM MgCl₂, 300 mM KCl, 1 mM DTT) and 4 or 15 μg of tRNA for RNase T₁ or RNase A, respectively. After RNase digestion, an equal volume of urea-dye mix (10 M urea, 2 mM EDTA, 0.06% each bromophenol blue and xylene cyanol) was added, and the digested products were analyzed on a 20% denaturing polyacrylamide gel. The RNA ladder was generated by alkaline hydrolysis of 5′-end-labeled RNA (10 min at 95°C in 50 mM NaHCO₃ [pH 9.5]).

Polymerase assay.

Polymerase assays were performed with poly(rA)/oligo(rU₁₈) as a template/primer. Reactions were carried out with 20 μl of buffer containing 20 mM Tris-HCl (pH 7.0), 1 mM DTT, 0.25 U of RNasin/μl, 100 ng of bovine serum albumin/μl, 50 mM NaCl, 0.03% n-octyl-β-d-glucopyranoside, 5 mM MgCl₂, and 2% glycerol as already described (39). The template/primer concentrations reported in the figure legends refer to oligo(rU₁₈).

The 50% inhibitory concentration (IC₅₀) values were calculated by using a three-parameter logistic equation, and inhibition data were fitted by using Kaleidagraph software.

Inhibition mechanisms were determined by performing substrate titration experiments. Kinetic parameters were calculated from a least-square fit of initial rates as a function of substrate concentration assuming Michaelis-Menten kinetics.

RESULTS

Selection of RNA aptamers with HCV NS5BΔC55bio.

To isolate high-affinity RNA ligands for the HCV NS5B polymerase, we screened a structurally constrained combinatorial RNA library (SSL25/10; Fig. 1A) by using the SELEX procedure. The library contains 35-nt random sequences in two segments of 25 and 10 nt, divided by a constant core sequence of 10 nt. This is in turn flanked by 18-nt constant regions at the 5′ and 3′ ends, respectively. The design of this library was inspired by the structure of an aptamer (SSL2.5) selected previously from a library called SSL30 (17). Each of the three segments of fixed sequence has the potential to fold into an independent stem-loop structure, although the central stem-loop has only the predisposition of a 3-bp stem which, on its own, will not be particularly stable. This structure might not form in the context of most aptamer sequences, as in fact was observed for the aptamer selected in this work. A single-stranded 3′-end tail of 11 bases was added to increase the efficiency of reverse transcriptase and the PCRs.

FIG. 1.

Properties of the combinatorial RNA library SSL25/10 and of the selected aptamers. (A) 5′-CSL, the core sequence, and 3′-CSL correspond to fixed sequences that can assume the stem-loop structure. The core sequence separates two random sequences 25 (N25) and 10 (N10) nt long. The single-stranded 3′ tail is a fixed sequence that anneals to the oligonucleotide used for reverse transcriptase and PCRs. (B) Sequences of N25 and N10 fragments (capital letters) in the selected aptamers. Lowercase letters refer to the fixed core sequence. Sequences common to aptamers of the same class are underlined (classes A and B) or indicated in bold characters (class C). (C) Predicted structures of selected aptamers belonging to classes A and B. The sequences shared by aptamers of the same class are shown in bold. The stem-loop-bulge structure common to aptamers of B class is indicated with SLII. The K_d values measured for each reported aptamer are shown.

Initially, the randomized RNA pool was challenged with streptavidin beads coated with biotinylated NS5BΔC55 protein (NS5BΔC55bio), and the complexed RNA was eluted, amplified (reverse transcription and PCR), and in vitro transcribed by using T7 RNA polymerase in the presence of a labeled nucleotide tracer. The products of transcription were used as an RNA pool for the next cycle of selection. At the fourth selection cycle, the RNA pool was roughly 140-fold enriched in the molecules with high binding affinity for NS5BΔC55bio, as judged by the radioactivity retained on the beads (not shown). As already noticed, in the case of structurally constrained RNA libraries (17) additional selection cycles failed to increase the amount of the selected RNA. Therefore, the PCR DNA from the fourth cycle was cloned, the nucleotide sequences of 15 individual clones were determined, and their structural motifs were analyzed. Based on the presence of common sequence motifs, the clones were grouped in three families (Fig. 1B). Analysis of the secondary structure of each sequence by the MFOLD program (21) revealed that the overall structure of the individual aptamers belonging to each class was conserved, with the common sequence motifs as part of analogous structural domains. As an example, the structures predicted for selected aptamers of A and B classes are reported in Fig. 1C.

Binding of NS5B polymerase by selected aptamers.

In order to determine the binding affinity of the selected aptamers and analyze the RNA-protein interaction, gel mobility retardation and filter-binding assays were performed. Initially, binding reactions were carried out by challenging a constant concentration of the labeled RNA aptamers (0.12 nM) with increasing concentrations (0.19 to 48 nM) of the NS5BΔC55bio protein used for the selection (not shown). However, the same results were obtained when the NS5BΔC55 protein was used instead of NS5BΔC55bio, implying that the biotinylation sequence added at the C terminus of the protein is not involved in aptamer selection. Therefore, all of the data reported below reflect results obtained with the nonbiotinylated NS5BΔC55.

As shown in Fig. 2, the RNA aptamer B.2 turned out to be a tight binder of the NS5BΔC55 polymerase, with a K_d value of 1.5 ± 0.2 nM, derived from filter-binding experiments (Fig. 2B). A similar dissociation constant was measured regardless of the salt concentration of the buffer (50 to 250 mM; not shown) and of the addition of unspecific RNA as competitor (tRNA, 200 nM [data not shown]), suggesting that binding of the RNA aptamer is highly specific. However, even though the labeled RNA was always denatured and renatured before each binding experiment, only 80% of the probe was complexed with saturating protein concentrations. This suggests that the RNA molecules were not homogeneous in structure and that only a specific conformation could be bound by the polymerase. The formation of a unique RNA/ΔC55 shifted band in gel retardation experiments, even at high protein concentrations (Fig. 2A), indicates that the protein/RNA ratio in the complex was always 1 to 1, suggesting that the RNA binds to a specific site of the protein. Binding of the B.2 aptamer was highly specific for the HCV NS5B, since interaction with the GBV-B NS5BΔC23 protein was not observed (Fig. 3, lane 2). As shown in Fig. 3, both full-length NS5B (lane 5) and NS5BΔC21 (lane 4) efficiently interacted with the B.2 RNA, thus eliminating the possibility that the selected aptamer was specific for NS5B proteins lacking the 55 C-terminal residues.

FIG. 2.

Gel retardation (A) and filter-binding (B) assays for the interaction of the B.2 aptamer and the NS5BΔC55 protein. Assays were performed as described in Materials and Methods with a constant amount of labeled B.2 RNA (0.12 nM) and increasing concentrations of purified protein between 0.19 and 48 nM.

FIG. 3.

Gel retardation experiments to assess the specificity of the interaction with HCV NS5B. Labeled B.2 RNA (0.12 nM) was challenged with a constant amount (6 nM) of GBV-B NS5BΔC23 (lane 2), HCV NS5B full-length (lane 4), or HCV NS5BΔC21 (lane 5) purified proteins. Lanes 1 and 3, unbound probe.

As reported in Fig. 1C, the aptamer A.2, which has a completely different sequence and predicted structure, showed a dissociation constant 10-fold higher than that measured for the B.2 aptamer (K_d = 16 ± 1.6 nM). In class B aptamers, B.1 also appeared to efficiently interact with NS5B polymerase (K_d = 3 nM; data not shown), while B.3 was less efficient (K_d = 10 ± 1.6 nM; Fig. 1C). Both B.1 and B.2 contain the sequence UAUGGACCAGUGGC, with GGACCA predicted to constitute the loop domain of a stem-loop structure which involves the 25-nt variable sequence and part of the fixed core sequence (Fig. 1C). Only part of that sequence (UAUGGACCA) is present in the B.3 aptamer and, even though the overall stem-loop structure could still be predicted, the loop is 1 nt shorter and an additional base pair separates the loop from the bulge.

Aptamer structural requirements for binding to NS5B.

The results reported above suggest that the structural determinants responsible for the high binding affinity of the B.2 RNA to the NS5B polymerase reside in the stem-loop domain indicated as SLII in Fig. 1C. In order to verify this hypothesis and to identify the minimal RNA domain required for binding with NS5BΔC55, a mutational analysis of the B.2 RNA aptamer was undertaken.

As the first step, a number of oligonucleotides were synthesized (Fig. 4A) in which both the 5′- and 3′-constrained stem-loop (5′- and 3′-CSL in Fig. 1C) (mutΔ1), the N10 variable sequence (mutΔ2), and the constant 10-nt core sequence (mutΔ3) were sequentially lacking. While mutΔ1 displayed a binding affinity comparable to that of the full-length B.2 sequence (K_d = 1.8 ± 0.4 nM), both mutΔ2 and mutΔ3 were not able to efficiently interact with the polymerase. In the case of mutΔ2, the destruction of the small stem-loop downstream of SLII could have destabilized SLII structure, while favoring an alternative structure that could be predicted by the MFOLD program (not shown), thus explaining its lower binding activity (K_d = 10 ± 0.7 nM). A further three bases deletion from the 3′ end of mutΔ2 produced a sequence exactly corresponding to the SLII stem-loop domain which retained the almost full binding capability of the wild-type aptamer (Fig. 4A) (SLII, K_d = 3 ± 0.5 nM). These results indicated that SLII does contain all of the structural features needed for the high-affinity binding of the B.2 RNA aptamer to the NS5B polymerase. Further mutational analysis of the SLII sequence (Fig. 4B) showed that neither a 2-bp shortening of the 9-bp long stem preceding the GC bulge [SLII(−2), K_d = 2 ± 0.4 nM] nor a further 2-bp deletion [SLII(−4), K_d = 4 ± 0.3 nM] affected binding efficiency. On the contrary, binding capability was abolished either by a 1-nt deletion of the loop sequence [SLII(ΔG)] or by a 1-bp insertion into the stem between the loop and the GC bulge [SLII(GC)]. In addition, the GC bulge appeared to be an important feature of the functional structure as demonstrated by the loss of binding activity by the B.2 mutant SLII(S13). SLII(S13) should not form an internal bulge due to the insertion into the opposite strand of the stem of 2 nt complementary to the unpaired GC residues.

FIG. 4.

Mutational analysis of the B.2 aptamer. (A) Deletion mutants of the B.2 aptamer. mutΔ1 contains the sequences corresponding to the N25 and N10 fragments and the intervening constant core sequence (boldface characters); in mutΔ2 the N10 sequence has been also deleted; mutΔ3 also lacks the fixed core sequence; and SLII has the B.2 sequence adopting the stem-loop structure common to class B aptamers. Arrows point to the nucleotide at the 3′ end of the mutant sequence. (B) Mutants of B.2 SLII. Mutants SLII(−2) and SLII(−4) lack 2 and 4 bp, respectively, at the base of the stem. Arrows indicate nucleotides at the 5′ and 3′ ends. A G (encircled) was deleted from the loop in mutant SLII(ΔG). In mutant SLII(GC) a GC base pair has been inserted at the base of the loop (arrow). (C) A GC dinucleotide (boxed) was inserted in mutant SLII(13) to allow base pairing of the GC bulge. The K_d values measured for each mutant are reported.

In order to provide experimental evidence supporting the predicted secondary structure, the 5′-end-labeled B.2 SLII RNA was subjected to enzymatic probing by using RNases T₁ and A, which cleave at the 3′ end of G and the pyrimidine nucleotides, respectively. Partial digestion with each enzyme produced the expected cleavage pattern (Fig. 5A). The nuclease-sensitive sites mainly corresponded to the unpaired nucleotides in the loop and bulge regions (Fig. 5B). However, additional cleavage sites were also noticed at the level of the predicted U-G base pairs at the base of the loop and at the top of the bulge, suggesting that U and G base pairing in these regions is unstable.

FIG. 5.

Enzymatic probing of the B.2 SLII structure. (A) Partial digestion of 5′-end-labeled SLII. RNase A was used at 12.5 (lanes 2 and 3) or 25 (lanes 4 and 5) ng/ml, and digestions were performed at 0°C (lanes 2 and 4) or 37°C (lanes 3 and 5). Digestions with 20 U of RNase T₁/ml were carried out at 0°C (lane 6) or 37°C (lane 7). Lanes 1 and 8, RNA ladder. Digestion products were separated on 20% urea-acrylamide gels. (B) Scheme of the RNase A (gray arrows)- and RNase T₁ (black arrows)-sensitive sites on the B.2 SLII RNA as deduced by partial digestion experiments.

Inhibition of NS5B polymerase activity.

The selected B.2 aptamer and its mutants were tested for their ability to inhibit the polymerase activity of NS5B. Polymerase assays were performed by preincubating the NS5BΔC55 protein with poly(rA)/oligo(rU₁₈) and increasing concentrations of the RNA aptamer before the addition of UTP. Efficient inhibition was observed by both the full-length B.2 RNA (IC₅₀ = 10 ± 0.5 nM; Fig. 6A) and mutants mutΔ1 or SLII (not shown), but not by SLII(S13) mutant (Fig. 6A), which was not able to interact with the NS5B enzyme. As expected, neither GBV-B NS5B nor poliovirus 3Dpol activities were affected by the addition of B.2 (not shown), demonstrating that the inhibition effect of the selected aptamer does depend on its specific interaction with the HCV polymerase.

FIG. 6.

Inhibition of NS5B polymerase activity by B.2 RNA. (A) Purified NS5BΔC55 (2 nM) was incubated with poly(rA)/oligo(rU₁₈) at 0.5 μM (referred to as the primer concentration) and increasing amounts of B.2 RNA (•) or mutant SLII(13) (▪). Aptamer concentrations varied between 2.5 and 160 nM with a twofold increase for each point. After 15 min of preincubation, 5 μM UTP was added and the elongation reaction was allowed to proceed for 20 min at 37°C. Polymerase efficiency was expressed as the percentage of the activity measured in the absence of B.2 RNA. (B) Inhibition mechanism of B.2 RNA. Polymerase reactions were performed as described above at 0 nM (•), 5 nM (▪), and 20 nM (⧫) B.2 RNA and increasing amounts of poly(rA)/oligo(rU₁₈). The template/primer concentration was varied with twofold increases between 30 nM and 2 μM and is referred to as oligo(rU₁₈). Kinetic parameters were calculated as described in Materials and Methods. (C) Binding competition experiments. Filter-binding assays were performed under polymerase activity conditions (50 mM NaCl). 5′-Labeled B.2 SLII (0.12 nM) was incubated with 1.5 nM NS5BΔC55 and without or with increasing amounts of cold poly(rA)/oligo(rU₁₈) at concentrations between 125 nM and 2 μM (▪). In the reverse experiment, 30 nM polymerase was incubated with a 5 nM concentration of labeled poly(rA)/5′-[³²P]oligo(U₁₈) in the absence or presence of cold SLII RNA at concentrations between 5 and 160 nM (•).

The mechanism of inhibition by the B.2 aptamer was determined with respect to the poly(rA)/oligo(rU₁₈) template/primer RNA. To this aim, V_max of the reaction and the K_m for poly(rA)/oligo(rU₁₈) were calculated at increasing concentrations of B.2 RNA. The data, reported in Fig. 6B, suggested a noncompetitive type of inhibition, since both V_max and K_m values were affected by the presence of the aptamer. The higher K_m values indicated that the interaction with B.2 RNA decreased the affinity of the polymerase for the poly(rA)/oligo(rU₁₈). This observation was also supported by the results of binding competition experiments, performed by filter-binding assays (Fig. 6C), where cold B.2 RNA appeared to inhibit polymerase binding to poly(rA)/oligo(5′-³²P)U₁₈, while, in contrast, increasing amounts of cold poly(rA)/oligo(rU₁₈) were not able to displace B.2 RNA binding. These results suggest that the RNA aptamer and the template RNA bind to different sites of the protein, which do not reciprocally influence each other.

Mapping of aptamer-binding site on NS5B.

We reasoned that, since the RNA aptamer does not interact with the template site, another site must be involved which could be presumably located on an exposed surface of the NS5B protein. The recent crystal structure of NS5B revealed the existence of a basic patch of residues on the solvent-exposed surface of the thumb subdomain, toward the very C terminus of the protein. We thought, therefore, that a candidate site for the interaction with the B.2 RNA could be located in this region. Since five arginine residues are contained in a single α-helix residing in this region (helix T, residues 497 to 513 [5]), we decided to mutagenize one of them and introduce the mutation in the context of the ΔC55 protein. Therefore, the mutant protein NS5BΔC55-R498E was constructed, and its ability to interact with the selected aptamer was verified by gel shift experiments. As shown in Fig. 7A, the mutant protein had a dramatically reduced affinity for the SLII RNA, and protein-RNA complexes were observed only at 50 to 100 times higher protein concentrations than the wild-type polymerase. Moreover, as expected, the polymerase activity of the R498E mutant was only slightly affected by the addition of the SLII RNA (Fig. 7B). It should be noted that mutation R498E did not affect NS5BΔC55 polymerase activity, thus eliminating the possibility that the inefficiency of the aptamer was due to an altered protein conformation. The kinetic parameters of NS5BΔC55 and its R498E mutant (NS5BΔC55-R498E) on poly(rA)/oligo(rU₁₈) were, respectively, as follows: K_m (UTP), 3.6 ± 0.6 and 4.6 ± 0.5 μM; k_cat, 5 ± 0.34 and 5.2 ± 0.30 min⁻¹. These results implicate R498 in the interaction with the selected RNA aptamer, suggesting a localization of the aptamer-binding site in the vicinity of the basic patch on the surface of the thumb domain. Very recently, a specific GTP-binding site was identified by X-ray crystallography on the surface of the thumb domain, very far from the polymerase catalytic site (6). This site does not directly involve R498, which is, however, very close to it. In order to test the possibility that the aptamer-binding site overlaps with the surface GTP-binding site, we verified whether GTP was specifically able to interfere with SLII binding in gel shift experiments. As shown in Fig. 8, while CTP was unable to displace SLII binding, even at very high concentrations, GTP was found to efficiently compete for aptamer binding. The GTP concentration that produced 50% of aptamer displacement was, however, very high (ca. 4 mM), in agreement with the low-affinity nature of the surface GTP-binding site (6).

FIG. 7.

Interaction with the R498E mutant enzyme. (A) Band-shift experiments. 5′-End-labeled B.2 SLII2 (0.12 nM) was incubated with increasing amounts of purified NS5BΔC55-R498E mutant enzyme at concentrations between 0.19 and 48 nM, and the resulting complex was resolved by native gel electrophoresis. (B) Inhibition of polymerase activity. Mutant R498E polymerase (2 nM) was incubated with 0.5 μM poly(rA)/oligo(rU₁₈) and increasing amounts of B.2 SLII RNA before the addition of UTP. Aptamer concentrations were varied with twofold increases of from 2.5 to 160 nM. Polymerase efficiency was expressed as the percentage of activity measured in the absence of SLII RNA.

FIG. 8.

Competition of SLII binding by GTP and CTP. The NS5BΔC55 protein (1 nM) was incubated with 5′-end-labeled B.2 SLII2 (0.12 nM) in the absence (0) or in the presence of increasing amounts (1.25 to 10 mM) of GTP or CTP, and the resulting complex was resolved by native gel electrophoresis.

DISCUSSION

In this study we describe the characterization of an RNA aptamer that binds to the HCV NS5B polymerase with high affinity and selectivity and is a potent inhibitor of its activity in vitro. We chose to use, as an initial pool of RNA, a combinatorial RNA library in which the randomized sequence is contained in a constant background that imposes structural constraints on the molecule. This kind of library has been successfully used for selection of aptamers binding to various proteins (16, 17, 18, 40). It has been observed that the selection process requires only four to five cycles of enrichment when constrained libraries are used, possibly by increasing the specific enrichment of aptamers obtainable per cycle of selection. This could be due to a reduced number of structural isomeric forms that can be formed by an individual aptamer containing a constraint, but this question has not been addressed systematically. After only four rounds of selection we obtained, in fact, a number of individual clones which shared structural and sequence homologies and were efficient ligands of the NS5B polymerase (Fig. 1). Among the selected RNA molecules, aptamers belonging to the B class displayed the highest affinity for NS5B and, among them, B.2 was the most potent ligand. Class B molecules share a similar stem-loop domain (SLII, Fig. 1C) that we demonstrated to correspond to the structural motif responsible for NS5B interaction. Both hairpin loops and bulges are often found as part of protein recognition sites (11). Bulges can create unique recognition sites either directly or indirectly by distorting the RNA backbone and allowing access to base pairs in a widened deep groove (19). Even though we cannot distinguish between these possibilities, the results of mutational analysis (Fig. 4) indicated the GC bulge as one of the features of the SLII domain strictly required for efficient binding. This is further supported by the observation that the GC bulge is contained in all class B aptamers: in ligand B.3, which contains only part of the sequence that in B.1 and B.2 forms the hairpin-loop structure, the GC bulge derives from the fixed core-sequence contained in the library, indicating that the 3-bp stem imposed as a potential constraint is not dominant and also that its presence did not impair the selection of a specific aptamer. Both the loop size and the length of the stem at its base also appeared to be fundamental features of the active molecule since shortening of the loop to 5 nt or insertion of 1 bp in the stem completely abolished binding of the isolated B.2 SLII domain. It should be noted that ligand B.3 contains both a 5-nt loop and a 4-bp stem between the loop and the bulge but was still able to interact with the NS5B protein with a relatively high affinity (K_d = 10 nM; Fig. 1C). This apparent discrepancy could be explained assuming that the simultaneous presence of a 5-nt loop and a 4-bp stem in the same molecule allowed a feature of the RNA three-dimensional structure needed for efficient interaction with NS5B that was otherwise disturbed.

The B.2 ligand and the isolated SLII domain are potent and selective inhibitors of the HCV polymerase activity. The high specificity of the phenomenon is suggested by the inability of B.2 to affect the activity of the poliovirus RNA-dependent RNA polymerase and, more impressively, that of NS5B from the HCV-related GBV-B, despite the high homology of the two NS5B proteins. This latter result reflects the absence of binding of B.2 to GBV-B NS5B observed in gel shift experiments (Fig. 3).

Kinetic analyses pointed to a noncompetitive mechanism of inhibition with respect to template/primer. This conclusion was supported by the fact that B.2 binding prevented the template/primer-NS5B complex formation, but a large excess of poly(rA)/oligo(rU₁₈) did not interfere with the B.2-NS5B complex formation (Fig. 6C). Altogether, these data suggest that B.2 interferes with template/primer binding on NS5B by interacting with the enzyme at a site distinct from the template/primer-binding site.

The crystallographic structure of the NS5B polymerase, in the absence of nucleotides and template RNA, has recently been determined (1, 5, 26). The protein folds into the characteristic fingers, palm, and thumb subdomains. As a peculiarity, two extended loops span the fingers and thumb domains at the top of the active site cavity accounting for the compact shape of the HCV polymerase. Moreover, a long β-hairpin formed by residues 442 to 456 in the thumb domain protrudes into the active site cavity and may thus restrict this site precluding binding of the template/primer. In the NS5BΔC21 polymerase (1, 26) the C-terminal residues 547 to 556 are buried in the putative template/primer-binding cleft, further occluding it. The absence of this sequence in the NS5BΔC55 truncation could be reflected in a more accessible RNA-binding cleft, thus explaining the 20- to 50-fold increase of activity of this enzyme form with respect to the ΔC21 truncation (14; Tomei et al., data not shown). However, in conditions that do not allow polymerase reaction, the stability of the interaction of the RNA in the template/primer-binding site could still be insufficient to allow the selection of aptamers specifically interacting in that region of the protein. Moreover, to facilitate the selection process, we have added at the C terminus of NS5BΔC55 a 16-amino-acid sequence in order to introduce the biotinylation modification. The resulting modified protein was three- to fourfold less active than the parental ΔC55 protein (data not shown). Even though we do not know which step of the polymerase reaction was affected, the possibility exists that the terminal peptide and/or the introduced biotinylation could have further limited the accessibility of the template/primer-binding cleft. It is therefore conceivable that a more exposed region of the protein was the primary site targeting the interaction with RNA ligands. As a matter of fact, introduction of the R498E mutation into NS5B abolished aptamer binding and, consequently, inhibition of the polymerase activity of the mutant protein by B.2. Arg498 lies in α-helix T (residues 497 to 513 [5]) located toward the C terminus of the protein in a basic solvent-exposed surface of the thumb domain. Despite the large change in charge, a Arg 498-to-Glu mutation does not affect polymerase activity (see above), suggesting that the mutation does not alter protein conformation. However, the relevance of helix T in polymerase function is indicated by the fact that Ala substitutions of all of the residues from 500 to 505 abolished enzyme activity (32, 42). The binding of the RNA aptamer on the surface of the thumb might freeze the molecule in a closed conformation, impeding binding of the RNA substrate in the active site and movement of the thumb during elongation. Interestingly, Arg 498 lies very close to a shallow pocket at the molecular surface of the enzyme that has been recently identified, by X-ray crystallography, as a specific GTP-binding site (6). The observation that GTP, but not CTP, was able to interfere with B.2 binding on NS5B (Fig. 8) suggests that the aptamer-binding site comprises the noncatalytic GTP-binding site. This GTP surface site might have a role in the regulation of the initiation step of HCV replication, since, in vitro, GTP is required for de novo initiation of RNA synthesis (22, 30, 43). This possibility is currently being verified and the RNA aptamer could be a helpful tool in the clarification of the role played by this site.

Since B.2 was a strong binder of NS5B, we searched the HCV genome for sequences similar to the 15 nt constituting the hairpin loop-bulge structure without success. Obviously, one cannot exclude that RNA structures similar to that selected, but adopted by a different sequence, might exist in the HCV genomic RNA. On the other hand, the strong inhibition of polymerase activity makes unlikely that the interaction of NS5B with such a structure could have a physiological role in HCV replication.

Experiments are in progress to test whether the B.2 RNA is able to affect NS5B activity in cells supporting subgenomic replication (29). This possibility would open up a new experimental avenue for identifying vital functional domains of the HCV polymerase by genetic approaches and could also provide an alternative approach for HCV therapy.