Secondary Structure and Hybridization Accessibility of Hepatitis C Virus 3′-Terminal Sequences

Abstract

The 3′-terminal sequences of hepatitis C virus (HCV) positive- and negative-strand RNAs contribute cis-acting functions essential for viral replication. The secondary structure and protein-binding properties of these highly conserved regions are of interest not only for the further elucidation of HCV molecular biology, but also for the design of antisense therapeutic constructs. The RNA structure of the positive-strand 3′ untranslated region has been shown previously to influence binding by various host and viral proteins and is thus thought to promote HCV RNA synthesis and genome stability. Recent studies have attributed analogous functions to the negative-strand 3′ terminus. We evaluated the HCV negative-strand secondary structure by enzymatic probing with single-strand-specific RNases and thermodynamic modeling of RNA folding. The accessibility of both 3′-terminal sequences to hybridization by antisense constructs was evaluated by RNase H cleavage mapping in the presence of combinatorial oligodeoxynucleotide libraries. The mapping results facilitated identification of antisense oligodeoxynucleotides and a 10-23 deoxyribozyme active against the positive-strand 3′-X region RNA in vitro.

The untranslated regions (UTRs) of the hepatitis C virus (HCV) positive-strand genome and negative-strand intermediate contain cis-acting sequences essential for viral translation and RNA replication. The 341-nucleotide (nt) 5′-UTR of the positive strand acts as an internal ribosome entry site (IRES) to direct cap-independent translation of the single ∼3,000-codon-long HCV open reading frame (38, 40). The tripartite 3′-UTR consists of a short upstream variable region, a central poly(U)-polypyrimidine stretch of variable length, and a terminal highly conserved 98-nt sequence (38). With the exception of the variable region, all sequence elements in the 3′-UTR are necessary for intracellular replication of HCV RNA (16, 29, 56).

A number of in vitro studies (26, 35, 36, 39, 58) have established that the terminal 98-nt X region sequence can serve as a minimal template for de novo initiation of negative-strand synthesis by the viral RNA-dependent RNA polymerase, NS5B. The 3′ terminus of the HCV negative strand reportedly plays an analogous role in positive-strand RNA synthesis (25, 36, 39). The binding of HCV 3′ termini to various host proteins may exert subtle effects on IRES-mediated translation (16, 23, 34, 55) or protect viral transcripts from degradation by cytoplasmic RNases (15, 48, 49).

The replicative and protein-binding functions of heteropolymeric regions in the HCV 3′ termini are, in many instances, dependent on the ability of the primary sequence to fold into higher-order RNA structure. In vitro, the X region sequence is capable of adopting a three-stem-loop structure, with the 3′-terminal 46 nt forming a thermodynamically stable stem-loop, SL I (6). Mutational analysis indicates that the duplex structure forming the base of SL I influences both the site and efficiency of de novo initiation by NS5B (26, 35). Preservation of the interior stem-loops, SL II and SL III, is critical for binding of the host factor polypyrimidine tract-binding protein (PTB) (11, 22), whereas disruption of SL I or SL II structure inhibits binding by several ribosomal proteins (55). Secondary-structure mapping of the negative-strand 3′ terminus has yet to be reported, but the sequence has been predicted to form various stem-loop structures (3, 31, 36). Mutational analysis suggests that binding of the HCV NS3 helicase to this region is sensitive to both the primary and secondary structure of a proposed terminal stem-loop (3).

The essential replicative functions and strict sequence conservation of HCV UTRs make these elements and their negative-strand complements promising targets for the development of antisense nucleic acid inhibitors (12, 51). Numerous studies (reviewed in reference 24) have explored the development of oligodeoxynucleotide (ODN) and ribozyme constructs targeting the HCV 5′-UTR and adjacent core protein N-terminal coding sequence, as antisense-mediated binding or cleavage of this region can inhibit IRES-directed gene expression. Recently, we characterized deoxyribozymes (DNA-based sequence-specific RNases) active against the 5′-UTR/core region (37). In contrast to the IRES domain, the 3′-UTR and negative-strand sequences lack readily assayable functions, and until recently, their contribution to RNA replication could not be directly assessed in cell culture models. In this regard, one study (31) has evaluated the utility of ribozymes active against the negative-strand 3′ terminus. Although ribozymes targeting the X region have been developed (30), antisense-mediated cleavage within this sequence has not yet been reported.

A major impediment to application of the antisense strategy is its sensitivity to secondary structure of the target RNA, which can interfere with hybridization of the construct (47). Identification of optimal antisense inhibitors has traditionally been a laborious empirical process, requiring individual testing of constructs. Prediction of accessible sites by thermodynamic modeling, chemical probing, and/or enzymatic mapping of RNA secondary structure has permitted a more rational and modestly reliable approach to antisense design (46, 57). Recently, combinatorial methods with sequence-randomized oligonucleotide libraries have been used to exhaustively screen all target sites within a given RNA. Hybridization of a fully or partially randomized ODN pool to structured RNA, followed by digestion with RNase H, can be used to determine simultaneously the relative accessibility of every potential hybridization site (a method referred to as hybridization accessibility mapping or oligonucleotide scanning). Target sequences thus identified have proven optimal for ODN-mediated (19, 20, 43) or ribozyme-mediated (1, 44, 45) inhibition of various eukaryotic genes in cell culture-based assays. Oligonucleotide scanning of the HCV 5′-UTR facilitated the design of antisense ODNs which inhibited IRES-mediated translation in vitro and in cell culture (32).

In the present study, we have used secondary-structure prediction, partial RNase digestion, and hybridization accessibility mapping to investigate the secondary structure of the HCV negative-strand 3′ terminus and to identify potential antisense target sites within the sequence. Additionally, we evaluated oligonucleotide scanning as a method of target site selection for the positive-strand X region RNA, whose secondary structure was characterized previously (6, 11, 22). Sequence-specific hybridization by individual anti-X deoxyribozymes and ODNs was assessed by measurement of deoxyribozyme- and RNase H-mediated cleavage within structured 3′-UTR transcripts in vitro.

MATERIALS AND METHODS

Plasmid construction.

Viral sequences were derived from p90/HCV FL-long pU (generously provided by Charles Rice, Rockefeller University), a consensus cDNA clone of HCV genotype 1a, isolate H77 (28). The XbaI-BstBI fragment encompassing the full-length viral sequence was subcloned into the vector pSP64 Poly(A) (Promega) in the antisense orientation with respect to the SP6 promoter (p64FL). The XbaI-KpnI fragment, including the 5′-UTR and core N-terminal sequence (nt 1 to 585), was subcloned into the vector pGEM-7Zf(+) (Promega) in the antisense orientation with respect to the SP6 promoter (pG5). The complete 3′-UTR sequence (nt 9375 to 9648) was PCR amplified with primers 3AP and 3AN or 3BP and 3BN (Table 1) and cloned into the vector pGEM-T (Promega). The resulting plasmids (pG3A and pG3B, respectively) were sequenced across the insert region. The HindIII-BamHI 3′-UTR fragment of pG3A was subcloned into pcDNA3.1(+) (Invitrogen) in the sense orientation with respect to the T7 promoter (pC3A).

TABLE 1.

Sequences of oligonucleotides used in this study

Type	Designation	Sequence
Primers for PCR and reverse transcription	3AP	5′-GCCAAGCTTAGGTTGGGGTAAACACTCCGGCCT-3′
	3AN	5′-GCCGCTAGCGGATCCTGTACATGATCTGCAGAGAGGCCAGTATC-3′
	3BP	5′-GCCGGATCCAGGTTGGGGTAAACACTCCGGCCT-3′
	3BN	5′-GCCGAATTCAAGCTTGTACATGATCTGCAGAGAGGCCAGTATC-3′
	3CN	5′-GGAATTCCACCACACTGGACTAG-3′
	P2	5′-CTAGATAATACGACTCACTATAG-3′
	P118	5′-GAGTGTCGTGCAGCCTCCAG-3′
	P237	5′-CCTGGAGATTTGGGCGTGC-3′
Antisense oligodeoxynucleotides and deoxyribozymesa	X10	5′-GCTAAGA(T)GGAGCCA-3′
	X16	5′-ACTAGGG(C)TAAGATG-3′
	X28	5′-CAGCTAG(C)CGTGACT-3′
	X42	5′-TCACGGA(C)CTTTCAC-3′
	X46	5′-CGGCTCA(C)GGACCTT-3′
	X61	5′-CTCTCTG(C)AGTCATG-3′
	X74	5′-GCCAGTA(T)CAGCACT-3′

^aFor deoxyribozyme constructs, the target site residue (in parentheses) was replaced by the wild-type 10-23 sequence, 5′-GGCTAGCTACAACGA-3′ (41), or, in the case of the X46 mutant, by the 10-23 G14C sequence, 5′-GGCTAGCTACAACCA-3′ (9).

In vitro transcription.

Template plasmids were linearized by digestion with XbaI (p64FL and pG5), EcoRI (pC3A), or RsaI (pG3B) and purified by phenol-chloroform extraction and ethanol precipitation. Vector pC3A was transcribed in vitro with T7 RNA polymerase (Gibco) in a 100-μl reaction mixture containing 40 mM Tris-HCl (pH 8.0), 8 mM MgCl₂, 2 mM spermidine-(HCl)₃, 25 mM NaCl, 5 mM dithiothreitol, a 1 mM concentration of each of the four nucleoside triphosphates (Promega), 60 U of ANTI-RNase (Ambion), 2 μg of template, and 100 U of enzyme. Templates p64FL and pG5 were transcribed with SP6 RNA polymerase in reaction buffer containing 40 mM Tris-HCl (pH 7.9), 6 mM MgCl₂, 2 mM spermidine-(HCl)₃, and 1 mM dithiothreitol. Unlabeled RNAs were purified by digestion with DNase I (Boehringer Mannheim), phenol-chloroform extraction, and ethanol precipitation. Transcript yields were calculated based on absorbance at 260 nm (Perkin Elmer Lambda 2 UV/VIS Spectrometer) and verified by agarose-formaldehyde gel electrophoresis.

T7-mediated transcription of pG3B was carried out in a 50-μl reaction mixture containing 500 μM ATP, 500 μM GTP, 2 mM UTP, 12.5 μM CTP, 25 μCi of [α-³²P]CTP (800 Ci/mmol, SP6/T7 grade; Amersham Pharmacia), 500 ng of template, and 50 U of polymerase. A ³²P-labeled RNA ladder was generated by T7-mediated transcription of Century Marker Template Plus (Ambion). Radiolabeled transcripts were purified by DNase I digestion and elution from NucAway spin columns (Ambion). Eluate RNA concentration was calculated based on specific activity determination (Packard TriCarb 4530 liquid scintillation counter), and transcript purity was verified by polyacrylamide-urea gel electrophoresis. Immediately prior to digestion experiments, RNAs were diluted to the desired concentration in 1× TMK buffer (30 mM Tris-HCl [pH 7.5], 10 mM MgCl₂, 270 mM KCl), incubated at 75°C for 2 min, renatured by cooling slowly to 45°C, and stored on ice, as described previously (22).

Partial RNase digestion.

Enzymatic probing of HCV negative-strand RNA structure was performed as reported previously for the positive-strand X RNA (22). RNA samples (1 μg) were treated for 5 min at 37°C with 0.5, 2.5, or 12.5 U of nuclease S1 (Promega); 0.1, 0.5, or 2.5 ng of RNase A (Sigma); 0.1, 0.5, or 2.5 U of RNase T₁ (BRL); and 0.02, 0.1, or 0.5 U of RNase T₂ (Gibco) in a total reaction volume of 5 μl. Control reactions lacking RNA or enzyme were carried out concurrently. Digestion buffer for S1 nuclease contained 50 mM sodium acetate (pH 4.5), 280 mM NaCl, and 4.5 mM ZnSO₄ (plus 0.2× TMK derived from RNA buffer); all other digestion reactions were carried out in 1× TMK buffer. Reaction products were purified by phenol-chloroform extraction and ethanol-sodium acetate precipitation with yeast tRNA carrier. Pellets were washed in 80% ethanol and dissolved in 15 μl of H₂O or annealing buffer (40 mM Tris-HCl [pH 7.5], 25 mM MgCl₂, 50 mM NaCl).

Oligodeoxynucleotide library and RNase H digestion.

Hybridization accessibility mapping was carried out with a protocol similar to that of Ho et al. (19). Chimeric 2′-O-methyl-RNA-DNA 11-mer oligonucleotide libraries (Howard Hughes Medical Institute-Keck, Yale University) of the form 5′-N_mN_mN_mXNNNN_mN_mN_mN_m-3′, where X is A, C, G, or T, N is a randomized nucleotide, and the subscript m denotes a 2′-O-methylated linkage, were hybridized to renatured HCV target RNAs. Digestion of such hybrids with Escherichia coli RNase H results in hydrolysis at a specified position 3′ of the ribonucleotide base-paired to X (13). One fully randomized (X = any nucleotide) and four partially randomized (X = A, C, G, or T) libraries were used to generate independent cleavage patterns for each RNA. Following a 5-min, 37°C hybridization of 300 pmol of library ODN to 1 pmol of RNA, digestion was initiated by the addition of 2 or 0.2 U of RNase H (Gibco) or 1 μl of H₂O (mock digest). The resulting 10-μl reaction mixtures, containing 50 mM Tris-HCl (pH 7.5), 10 mM MgCl₂, 150 mM KCl, and 6 U of ANTI-RNase, were incubated for 10 min at 37°C. Reactions were quenched by phenol-chloroform extraction and precipitated in 70% ethanol with 150 mM sodium acetate and 5 μg of yeast tRNA. Pellets were washed in 80% ethanol and redissolved in 9 μl of diethylpyrocarbonate-treated water.

Primer extension.

Oligodeoxynucleotide primers (Gibco) P2, P118, P237, and 3CN (Table 1) were used for reverse transcription of HCV RNA digests. Each primer was end labeled with 10 U of T4 polynucleotide kinase (Gibco) at 37°C for 1 h in a 25-μl reaction containing 1× T4 exchange buffer (50 mM imidazole-HCl [pH 6.4], 12 mM MgCl₂, 1 mM 2-mercaptoethanol, 70 μM ADP) and 50 μCi of [γ-³²P]ATP (3,000 Ci/mmol; Amersham Pharmacia Biotech). Primer was eluted in deionized water from a Sephadex G-50 (Pharmacia) column, and 1 μl of the peak elution fraction (∼10 to 50 nCi) was added to 3 μl of each digestion product (primer in molar excess relative to RNA). Primer-substrate annealing was achieved by incubation at 75°C for 3 min, followed by slow cooling to 45°C.

Extension was performed with 100 U of Moloney murine leukemia virus (MMLV) reverse transcriptase (Gibco) at 42°C for 50 min in a 10-μl reaction mix containing 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl₂, 1 mM dithiothreitol, and a 2 mM concentration of each of the four deoxynucleoside triphosphates (plus 5.6 U of ANTI-RNase for RNase H digestion products). cDNA products were diluted in an equal volume of loading buffer (90% formamide, 1× Tris-borate-EDTA [TBE], 0.5% bromophenol blue) and heated for 3 min at 90°C immediately prior to electrophoresis. Dideoxy sequencing reactions were carried out on unmodified RNA and run in parallel with the primer extension products on 8% polyacrylamide-7 M urea sequencing gels.

Deoxyribozyme and targeted RNase H cleavage assays.

³²P-labeled HCV 3′-UTR RNA samples (0.01 pmol) were digested for 1 h at 37°C in a 10-μl reaction containing 10 mM Tris-HCl (pH 7.5), 10 mM NaCl, 140 mM KCl, 10 mM MgCl₂, 8 U of ANTI-RNase, and 20 pmol of 15-mer antisense ODN or deoxyribozyme (Invitrogen). All deoxyribozymes were designed with 7-mer hybridizing arms (7 × 7) flanking the 10-23 catalytic domain (Table 1). ODN digestions were carried out in the presence of 1 U of E. coli RNase H (Ambion). Where indicated, deoxyribozyme digestions were supplemented with 2 U of E. coli RNase H (Gibco). Digestion reactions were mixed in a 1:2 volume ratio with Loading Buffer II (Ambion), denatured at 80°C for 5 min, and resolved on 5% polyacrylamide-8 M urea gels. Quantitation of RNA cleavage products was carried out with the Storm 860 PhosphorImager (Molecular Dynamics).

RNA structure prediction.

HCV negative-strand RNA structure was analyzed with mfold, version 3.1 (59), under default folding conditions (37°C, 1 M NaCl, and no limit on distance between paired bases). Secondary-structure plots were generated with RnaViz, version 2.0 (14).

RESULTS

Enzymatic probing and thermodynamic modeling of HCV negative-strand RNA secondary structure.

We used endoribonuclease digestion monitored by primer extension analysis (50) to characterize folding of the HCV negative-strand 3′-terminal RNA sequence in vitro. Substrate RNAs containing partial (3′-terminal 585 nt) or full-length HCV negative-strand sequences were generated by SP6-mediated in vitro transcription of vectors pG5 and p64FL. Transcripts were thermally renatured and probed at 37°C with serial dilutions of single-strand-specific endoribonucleases, S1 nuclease at low pH and RNases A, T₁, and T₂ at physiological pH. For subsequent primer extension by MMLV reverse transcriptase, RNA was annealed to ³²P-labeled oligodeoxynucleotide primers P2, P118, and P237 in Tris-Mg²⁺-Na⁺ buffer, as described previously (22), or hybridized in the absence of additional buffer-derived salts.

Gel electrophoresis of primer extension products yielded cleavage data for nt 5 to 356 of the negative strand (nucleotide positions are numbered in ascending order from 3′ to 5′ to facilitate comparison with the complementary positive strand). Representative digestion results, obtained with the partial-length negative-strand template, are shown in Fig. 1. The observed cleavage products were consistent with known base preferences for RNases T₁ and A (3′ of guanosine and pyrimidine residues, respectively) and with the sequence-independent nuclease activity of S1 and T₂. Qualitatively identical patterns of nuclease cleavage were observed with the full-length negative-strand RNA substrate (data not shown), suggesting the absence of long-range intramolecular base-pairing interactions between this region and upstream sequences.

FIG. 1.

Secondary-structure mapping of the HCV negative-strand 3′ terminus. cDNAs derived from MMLV-mediated extension of primers P2 (A) and P237 (B) on nuclease- and mock-digested HCV RNA templates are shown, with dideoxy sequencing products as molecular size standards. SP6 in vitro transcripts from XbaI-linearized template pG5 were renatured in TMK buffer, and 1-μg samples were digested for 5 min at 37°C. Lanes: 1, negative control, template RNA omitted; 2, TMK buffer mock digest; 3, RNase T₁ (0.5 U); 4, RNase A (0.1 ng); 5, RNase T₂ (0.1 U); 6 to 9, dideoxy sequencing reactions; 10, S1 nuclease (2.5 U); 11, S1 buffer mock digest. Cleavage products were redissolved in H₂O and analyzed by primer extension as described in Materials and Methods.

Analysis of single-strand-specific RNase cleavage within the negative-strand 3′ terminus revealed patterns of nuclease sensitivity characteristic of stable stem-loop structures. To further analyze the RNA folding pattern, secondary-structure prediction for the terminal 365-nt sequence was carried out with mfold, version 3.1 (59). Inspection of the mfold energy dot plot and p-num values (not shown) indicates that the 3′-terminal 217 nt are well determined, whereas upstream sequences can adopt a variety of structurally diverse yet thermodynamically equivalent structures. Analysis of the optimal structure (Fig. 2A) revealed that the terminal 220 nt fold into five stem-loops (I′, IIz′, IIy′, IIIa′, and IIIb′; see Discussion for domain nomenclature). The 5′ aspect of the largest stem structure, IIz′, is interrupted by 5-nt and 7-nt single-stranded bulges. Nucleotides C320 to G363 also appear to form a relatively well-defined stem-loop (IV′). The five terminal stem-loops and, to a lesser degree, the internal stem-loop are retained among optimal structures even when upstream HCV sequences are included in the computation.

FIG. 2.

Predicted secondary structure and observed nuclease cleavage of the HCV negative-strand 3′ terminus. Secondary-structure energy minimization of the 3′-terminal 365-nucleotide region was performed with the mfold program, version 3.1. (A) The thermodynamically optimal folding configuration is shown, with sites exhibiting reproducible, dose-dependent endonuclease cleavage indicated on the nucleotide 5′ of the hydrolyzed bond. Cleavage sites for RNases T₂ (squares), A (triangles), and T₁ (circles) are reported as preferred (solid symbols) or minor (open symbols). Preferred and minor sites for nuclease S1 are denoted by bold and italic type, respectively. (B) Alternative base-pairing configuration for nt 1 to 20. (C) Positive-strand 5′-UTR/core region (nt 1 to 365) secondary-structure domains (40). An RNA pseudoknot structure, psk, is formed by base-pairing of subdomain IIIf and sequences upstream of domain IV.

The poorly structured domain (IIIcdef′), comprising nt C222 to C320, is capable of forming a multilobed structure, as in Fig. 2A, but, in many foldings, participates in long-range base-pairing interactions with various upstream or downstream sequences. This promiscuity of base-pairing results in notable structural variation among the suboptimal mfold structures. For example, the terminal stem-loop (I′, nt 5 to 20) is absent from several structures (Fig. 2B); in those cases, nt 1 to 20 base-pair extensively with nt 235 to 254. The IV′ element is disrupted in various foldings via interaction with downstream sequences. In another prominent variant (5′-UTR-like), the basal stem regions of domains IIy′ and IIIcdef′ are reorganized into a single branched stem structure, as in the positive-strand 5′-UTR (Fig. 2C).

In Fig. 2A, the cleavage sites exhibiting conclusive nuclease dose-response are superimposed on the thermodynamically optimal RNA structure predicted by the mfold program. The cleavage data are consistent with the predicted optimal structure and correlate closely with mfold ss-count values, i.e., the number of computed foldings (out of the total predicted by mfold) in which the residue is unpaired (not shown). The most highly preferred cleavage sites correspond to regions predicted to lie within exposed single-stranded regions, such as the 4-nt apical loop (A56 to A59) and 7-nt bulge (A91 to A97) of IIz′, the 10-nt loop of IIIb′ (A194 to A203), and the 7-nt loop of IV′ (U339 to U345). Extensive nuclease cleavage was also found to occur in predicted hinge regions (e.g., nt C152 to U155), whereas cleavage was largely absent from predicted stem regions.

mfold analysis was repeated on the sequence under constrained folding conditions. Through B, D, H, V mfold annotation, the 29 nucleotides which were reproducibly susceptible to cleavage by two or more nucleases (Fig. 2A) were prohibited from base-pairing (the prohibition is exempted if the 3′-neighboring nucleotide is single stranded). When the folding was thus constrained to reflect our enzymatic probing data, the optimal structure shown in Fig. 2A was retained. In contrast, suboptimal structures exhibiting disruption of domain IIz′, IIy′, or IV′, including 5′-UTR-like foldings, were associated with a relatively great increase in ΔG and were therefore less strongly supported by experimental results.

Oligonucleotide scanning of the HCV negative-strand 3′ terminus.

Single-stranded regions of the HCV 3′-terminal sequence were further mapped by measurement of RNA accessibility to hybridization by antisense oligonucleotides. Hybrid formation is thought to stabilize transient structural conformations, thereby revealing accessible sites which are underrepresented in traditional single-strand-specific cleavage maps (4). As with the above mapping protocol, a partial digestion-primer extension method was employed. Renatured RNA transcripts bearing the 3′-terminal 585-nucleotide HCV sequence were incubated at 37°C in the presence of a 300-fold molar excess of a fully or partially randomized 11-mer oligonucleotide library. (Given the complexity of each library population, this stoichiometry corresponds to an oligonucleotide:RNA ratio of ∼1:14,000 for each member of a fully randomized library and ∼1:3,500 for partially randomized libraries.) Hybridized RNA was digested with E. coli RNase H, and cleavage products were reverse transcribed with ³²P-labeled ODN primers P2, P118, and P237. Representative cDNA products, shown in Fig. 3A, reveal a heterogeneous distribution of sites within the target transcript that are susceptible to hybridization and dose-dependent RNase H cleavage.

FIG. 3.

Hybridization accessibility mapping of the HCV negative-strand 3′ terminus. SP6 in vitro transcripts from XbaI-digested template pG5 were renatured in TMK buffer and digested with 0.2 or 2 U of RNase H at 37°C in the presence of a 300-fold molar excess of fully randomized (N) or nucleotide-specific (C, T, A, and G) chimeric 11-mer oligonucleotide libraries. (A) Overview of RNase H cleavage pattern for the entire 3′-terminal region. cDNAs were derived from MMLV-mediated extension of primer P2. Lanes: 1, negative control, HCV RNA omitted; 2, RNase H (2 U) mock digest, library omitted; 3 to 14, nucleotide-specific library digests and dideoxy sequencing products; 15 to 17, N library digests. (B) Detail of the cleavage pattern for RNA digested with 2 U of RNase H. cDNAs were derived from extension of primers P2 (lower left), P118 (upper left), and P237 (right). Lanes: 1, mock digest, library omitted; 2, mock digest, RNase H omitted; 3, N library digest; 4, negative control, HCV RNA omitted; 5 to 12, nucleotide-specific library digests and dideoxy sequencing products.

To permit mapping of the observed hybridization zones to 1-nucleotide resolution, the experiments were carried out with chimeric 2′-O-methyl RNA-DNA libraries (19, 20). Specifically, libraries were of the form 5′-(N_m)₃X(N)₃(N_m)₄-3′, where X is A, C, G, T, or randomized nucleotide N and the subscript m denotes a 2′-O-methylated linkage. The 2′ backbone modification permits Watson-Crick base-pairing with the RNA but prohibits RNase H cleavage at specific ribose-phosphate linkages opposite the modified region (Fig. 4A). Scission of the RNA takes place at only a single position, 3′ of the nucleotide base paired to the library residue X, as directed by the interior stretch of 4 unmodified deoxyribose linkages (13). Therefore, each observed cleavage site can be unambiguously attributed to hybridization at a particular 11-nt range in the target RNA. Such precision of mapping is not possible with unmodified libraries, in which case cleavage can occur at any of several positions within a given hybrid (Fig. 4B).

FIG. 4.

Determination of sites accessible to hybridization within the HCV negative-strand 3′ terminus. (A) Schematic representation of a 2′-O-methyl RNA-DNA chimeric 11-mer oligonucleotide hybridized to target RNA. The unique site of RNase H cleavage within the target sequence is indicated. The cDNA generated by primer extension of the cleaved RNA differs by 1 nt in length from a dideoxy sequencing reaction terminated at nucleotide X. (B) RNase H-mediated cleavage directed by an unmodified oligodeoxynucleotide. Possible endonuclease cleavage sites are indicated. Exonuclease activity degrades processively toward nucleotide Y. The 3′ product of digestion can vary in length because its 5′ extent depends on the site of endonuclease cleavage. (C) Sites within the 3′-terminal 356 nt of the HCV negative strand exhibiting sequence-dependent hybridization and dose-dependent RNase H cleavage. The pattern of preferred and minor cleavage sites (solid and open arrows, respectively) is a composite derived from primer extension analysis with primers P2, P118, and P237 for two independent sets of RNase H digests. Regions of high and universal sequence conservation (53) are indicated by underlined and boxed nucleotide positions, respectively.

The cDNA products generated by accessibility mapping of HCV negative-strand RNA were consistent with sequence-specific hybridization and 2′-O-methyl-restricted RNase H cleavage. Most of the digestion products observed in the A, C, G, and T library digest lanes migrated at a molecular size 1 nt less than that of a corresponding dideoxy sequencing product (Fig. 3). Such a cleavage pattern results from the formation of a complementary base pair between the constrained library deoxynucleotide X and the ribonucleotide Y adjacent to the cleavage site (Fig. 4A). In contrast to previous reports (19, 20), we obtained satisfactory data with both fully and partially randomized libraries. Each of the base-specific libraries yielded cleavage products greater in intensity than the corresponding products from fully randomized library digests (Fig. 3B). A fourfold difference in intensity was anticipated based on the relative complexity of the oligonucleotide pools used (4¹⁰ sequences for the A-, C-, T-, and G-specific libraries versus 4¹¹ for the N library).

The HCV negative-strand RNase H cleavage pattern was highly reproducible, and the results of two independent sets of cleavage reactions are summarized in Fig. 4C. Regions of RNA accessible to antisense hybridization are evidenced by strong cleavage at consecutive nucleotides: A101 to A103 and G111 to G115, flanking the IIz′-IIy′ junction hinge; C146 to G156, on the 5′ stem face and hinge upstream of domain IIy′; A196 and G197, within the apical loop of IIIb′; A227 to A234, on the 3′ stem face and hinge downstream of IIIcdef′; and A248 to A305, the entire multilobed apical structure of IIIcdef′. With respect to the proposed HCV negative-strand structure (Fig. 2A), antisense constructs appeared to hybridize most efficiently to putative interdomain hinge regions and large apical loops. The poor hybridization observed for nt 18 to 75, 81 to 95, 118 to 141, 159 to 174, 184 to 192, 201 to 223, and 326 to 356 suggests that proposed apical duplex stems, especially those terminating in relatively small loops, are resistant to RNase H cleavage. These results indicate that hybridization of library oligonucleotides occurs principally within or on the margin of proposed single-stranded regions.

Hybridization accessibility mapping of the HCV positive-strand X region.

To evaluate the above RNase H mapping method as a tool for the design of antisense nucleic acid constructs, the procedure was repeated with the HCV positive-strand 3′ terminus as a substrate RNA. Plasmid pC3A was used to generate T7 transcripts encompassing the complete HCV 3′-UTR sequence and bearing a downstream vector-derived tag, which served as the annealing site for subsequent primer extension. cDNAs derived from MMLV-mediated extension of primer 3CN are shown in Fig. 5A. In contrast to the negative-strand results, the full-length cDNAs from all 3′-UTR digests, including the mock digest, were heterogeneous in length. This was most likely due to abortive synthesis and slippage by reverse transcriptase within the poly(U)-polypyrimidine tract of the template RNA. Such phenomena have been reported previously for the HCV RNA-dependent polymerase NS5B as it traverses 3′-UTR templates (26, 36, 39).

FIG. 5.

Hybridization accessibility mapping of the HCV positive-strand 3′-UTR RNA. (A) cDNAs derived from MMLV-mediated extension of primer 3CN on RNase H- and mock-digested HCV RNA substrates. T7 in vitro transcripts from EcoRI-linearized template pC3A were renatured in TMK buffer and subjected to digestion with 2 U of RNase H. Lanes: 1, negative control, HCV RNA omitted; 2, mock digest, library omitted; 3, mock digest, RNase H omitted; 4, N library digest; 5 to 12, nucleotide-specific library digests and dideoxy sequencing products. (B) Sites within the X region sequence exhibiting sequence-dependent hybridization and dose-dependent RNase H cleavage. The pattern, presented as in Fig. 4C, is derived from primer extension analysis for two independent sets of RNase H digests.

The cDNA products resulting from RNase H-directed cleavage within the 3′-X sequence were catalogued, and a pattern of reproducible cleavage sites from two independent sets of digestion reactions is presented in Fig. 5B. Four zones of efficient hybridization are evident: U3 to U7, U13 to C17, A40 to U43, and G46 to A49. The two upstream regions reside in SL III of the X region structure, whereas the latter two constitute the 3′ stem face of SL II (Fig. 6A). The remainder of SL II as well as the entirety of SL I exhibited dramatically lower accessibility to hybridization.

FIG. 6.

Antisense-mediated cleavage of the HCV positive-strand 3′-X region. (A) Deoxyribozyme (Dz) target sites selected for cleavage analysis, superimposed on the X region secondary structure (6). See Table 1 for antisense construct sequences. The deoxyribozyme target nucleotide and hybridization zone for X46 constructs are indicated by italic and bold type, respectively. Digestions were carried out for 1 h under simulated physiological conditions (37°C, pH 7.5, 10 mM Mg²⁺) with 1 nM ³²P-labeled 3′-UTR target RNA and 2 μM construct. (B) In vitro RNase H-mediated cleavage of 3′-UTR RNA, directed by 15-mer antisense oligodeoxynucleotides. Lanes: 1, negative control X10 digest, RNase H omitted; 2 to 8, ODN plus 1 U of E. coli RNase H; 9, mock RNase H digest, ODN omitted. (C) Intrinsic cleavage activity and hybridization of 10-23 deoxyribozymes on 3′-UTR RNA in vitro. Lanes: 1, mock digest, deoxyribozyme omitted; 2 to 9, deoxyribozyme digests; 10 to 18, digests supplemented with 2 U of RNase H. The 3′ products of cleavage are shown.

In vitro hybridization and cleavage activity of individual antisense constructs targeting the HCV X region sequence.

To assess the hybridization mapping results as a predictor of in vitro antisense activity, we designed a series of 15-mer oligodeoxynucleotides and corresponding 10-23 deoxyribozymes directed against the X region RNA (Table 1). As with conventional RNA-based ribozymes, the modular structure of deoxyribozymes comprises a conserved catalytic domain sequence (e.g., 10-23, which cleaves within a purine-pyrimidine target site dinucleotide), and variable substrate-binding arms, which can be specified to interact with a given substrate RNA by Watson-Crick base-pairing (41). We selected deoxyribozyme hybridizing arms and antisense ODNs to target 15-mer sequences within the X RNA exhibiting diverse hybridization properties. Based on the RNase H cleavage map (Fig. 5B), efficient hybridization was predicted for ODN and deoxyribozyme constructs X10, X16, X42, and X46, and weak or no activity was predicted for X28, X61, and X74 (construct numbers refer to the nucleotide position 5′ of the deoxyribozyme cleavage site [Fig. 6A]).

The accessibility of each target site to hybridization was assayed under simulated physiological conditions in vitro (see Materials and Methods) by measurement of antisense-mediated 3′-UTR RNA cleavage. For this assay, ³²P body-labeled target transcripts bearing genuine HCV 3′ termini (28) were generated by T7-mediated transcription of RsaI-digested plasmid pG3B. When electrophoresed on denaturing polyacrylamide-urea gels, these 3′-UTR transcripts exhibited a remarkably large range of apparent molecular sizes (Fig. 6B, lanes 1 and 9). The heterogeneity in length likely resulted from polymerase slippage within the poly(U) tract, as described above. Transcripts generated by SP6 transcription from other vectors (e.g., pG3A) were unsuitable for use because they yielded less-than-full-length transcription products, consistent with premature termination within the poly(U) tract (data not shown).

Individual ODN digests generated 3′ and 5′ products in the correct molecular size range, consistent with E. coli RNase H-directed cleavage of the body-labeled RNA transcript at multiple sites within the hybrid (Fig. 6B). The 5′ products, owing to heterogeneity of the poly(U) tract, could not in all cases be completely resolved from the full-length uncleaved transcript. Sequence-dependent cleavage was not observed if either ODN or RNase H was omitted from the reaction (Fig. 6B, lanes 1 and 9). Quantitation of low-molecular-weight cleavage products for each ODN (corrected to account for background intensity observed in mock digests) was carried out by PhosphorImager analysis. Calculation of transcript cleavage efficiency was based on the assumption that RNase H exhibits, to the maximum extent possible, processive 3′-to-5′ exonuclease activity within the hybrid (Fig. 4B), as reported previously (13, 42).

A comparison of cleavage activity for the various ODN constructs is presented in Table 2. At a construct:RNA ratio of 2,000:1, ODNs X10, X16, X42, and X46 each achieved ≈60 to 80% cleavage of the 3′-UTR transcript. In contrast, only ≈30% cleavage was obtained with constructs X28 and X74, and cleavage products were not detected with ODN X61. The differences observed are greater than could be accounted for by any possible disparity in RNase H processivity. For the panel of seven constructs, cleavage activity correlates strongly with accessibility as determined by oligonucleotide scanning. The close agreement between the mapping results and the individual ODN cleavage assays suggests that the RNase H map is not patently biased by positive or negative cooperativity of binding resulting from the presence of multiple ODN sequences in the library. The 15-mer ODN-RNase H digestion results thus provide validation for the 11-mer ODN-RNase H hybridization accessibility mapping method.

TABLE 2.

In vitro cleavage activity of 10-23 deoxyribozymes and corresponding 15-mer oligodeoxynucleotides targeting the HCV X region RNA

Target site	Hybridization range	GC contenta	Mapping resultb	3′-UTR transcript cleavagec (%)
				ODN + RNase Hd	Deoxy- ribozyme
A10	U3-C17	8/15	+	76 ± 4	ND
G16	C9-U23	7/15	+	63 ± 15	ND
G28	A21-G35	9/15	−	29 ± 3	ND
G42	G35-A49	8/15	+	60 ± 4	ND
G46	A39-G53	10/15	+	62 ± 8	17.0 ± 1.5
G61	C54-G68	8/15	−	ND	ND
A74	A67-C81	8/15	−	30 ± 19	ND

^aIncludes target site nucleotide, which does not participate in hybridization by deoxyribozyme constructs.

^bSee Fig. 4C.

^cSee Materials and Methods for reaction conditions; values are reported as mean and standard deviation for three or four independent digestion experiments. ND, not detected. ODN, oligodeoxynucleotide.

^dAssumes complete 3′-to-5′ exonuclease processivity by RNase H.

The predictive capacity of the hybridization map did not extend, however, to selection of optimal deoxyribozyme target sites. Of the seven deoxyribozymes tested, only one, X46, exhibited detectable cleavage activity against 3′-UTR RNA, as evidenced by the appearance of a 3′ product 52 nt in length (Fig. 6C). In contrast to the ODN-RNase H digests, the deoxyribozyme cleavage product migrated as a single, sharp band, consistent with cleavage at a unique site in the target transcript. The X46 deoxyribozyme-mediated cleavage exhibited dose-response behavior and was sensitive to the magnesium ion concentration in the reaction. An approximately twofold reduction in activity was observed upon depletion of the Mg²⁺ concentration from 10 mM to 2 mM (data not shown). A G14C point mutation (9) in the 10-23 catalytic core rendered the X46 deoxyribozyme construct enzymatically inactive (Fig. 6C, lane 7), demonstrating that the observed cleavage activity was indeed due to deoxyribozyme activity and not RNase H contamination.

It was of interest to determine why three out of four deoxyribozymes (X10, X16, and X42), which were predicted to anneal to the target sequence based on accessibility mapping and 15-mer ODN-RNase H digestion results, failed to cleave the target RNA. In order to detect annealing of deoxyribozyme constructs to the target RNA, we repeated the deoxyribozyme digestion in the presence of E. coli RNase H. This enzyme is able to cleave within either of two 7-mer DNA-RNA hybrids formed by annealing of the deoxyribozyme arms to the target RNA. As shown in Fig. 6C, deoxyribozymes X10, X16, and X46 and the X46-G14C mutant deoxyribozyme were each able to direct RNase H cleavage, consistent with hybridization of one or both construct arms to their respective target sequences. No RNase H cleavage product was observed with deoxyribozyme X42. These results suggest that the X10 and X16 deoxyribozymes are able to anneal to their respective target sequences but that substrate binding forces the 10-23 core catalytic domain to adopt a conformation incompatible with deoxyribozyme enzymatic activity. In contrast, the X42 deoxyribozyme is apparently unable to anneal, even though the corresponding 15-mer ODN hybridizes efficiently. The difference might be attributed to the relatively low GC content of the X42 target site, in that hybridization may require formation of the G42-dC base pair, which is present in the 15-mer ODN hybrid but not in the 7 × 7-mer deoxyribozyme hybrid.

DISCUSSION

The HCV negative-strand 3′-terminal structure is thought to play an essential role in the initiation of positive-strand RNA synthesis by direct interaction with the NS5B polymerase (25, 36, 39) and/or NS3 helicase (3). The RNA structure itself may also render the negative strand resistant to cellular RNases in the absence of viral protein binding (49) and carry out other functions by direct interaction with host factors (48, 52). Characterization of the HCV 3′-terminal RNA fold may therefore be of value for the identification of structural features essential for RNA replication (17, 27) and protein binding (3, 48), whose location in the negative-strand sequence has been mapped by deletion mutagenesis.

The partial nuclease digestion and thermodynamic prediction presented in this study indicate that the 3′-terminal 365 nt of the HCV negative strand fold into a seven-stem-loop structure (Fig. 2A). This result differs greatly from the folding reported previously by Lieber et al. (31). We were unable to identify a structure resembling their result among the mfold output from a variety of folding conditions. The structure determined by Lieber et al. is also inconsistent with our enzymatic probing data, in that nucleotides residing in proposed duplex stem regions (e.g., nt 56 to 59, 93 to 96, and 152 to 154) were observed to be highly sensitive to single-strand-specific nucleases. Notably, our nuclease mapping analysis could not rule out foldings in which the terminal stem-loop (nt 5 to 20) is obliterated by base-pairing to nt 235 to 250 (Fig. 2B). The mutational analysis presented by Banerjee and Dasgupta (3), which demonstrated binding of HCV NS3 protein to the terminal stem-loop, was carried out with RNA transcripts lacking the upstream region.

The proposed HCV negative-strand structure (Fig. 2A) does not simply constitute a mirror image of its positive-strand complement (Fig. 2C). The solution structure of the 5′-UTR sequence was investigated by Brown et al. (8) and has been subsequently refined through other nuclease mapping studies and mutational analysis (reviewed in reference 40). Several stem-loop structures of the 5′-UTR, including domains I (nt 5 to 20) and IV (nt 331 to 354) and subdomains IIIa (nt 156 to 171) and IIIb (nt 177 to 222), are retained in the proposed negative-strand structure (designated I′, IV′, IIIa′, and IIIb′, respectively). However, domain II and the surrounding interdomain regions are reorganized into two large stem-loops, which we designated IIz′ (nt 21 to 104) and IIy′ (nt 107 to 151). The complement of subdomains IIIc to IIIf adopts a thermodynamically unstable multilobed structure (IIIcdef′, nt 329 to 317). Even in 5′-UTR-like suboptimal foldings, the region of subdomains IIId to IIIf is reorganized. The retention of domains I(a), IIIa, IIIb, and IV, with concomitant reorganization of domains II and IIId to IIIf, was a result common to mfold analyses of both HCV and pestivirus negative-strand RNAs (bovine diarrhea virus type I and hog cholera virus; not shown).

In light of the perceived structural differences between the 5′-UTR and its complement, we find it intriguing that the negative-strand hybridization accessibility map (Fig. 4C) bears little resemblance to its positive-strand counterpart presented by Lima et al. (32). In that study, RNase H mapping with an unmodified 10-mer oligodeoxynucleotide library identified a preferred hybridization site at nt 29 to 38 and other, lower-affinity sites. Under our assay conditions, we observed a more robust RNase H cleavage activity overall, yet nt 29 to 38 of the negative strand were found to be inaccessible to hybridization. Notably, these nucleotide positions lie within a large single-stranded region between domains I and II of the 5′-UTR, whereas in the proposed negative-strand structure, they take part in extensive base-pairing to form the central portion of the stem structure in domain IIz′. Therefore, our hybridization accessibility mapping results provide further evidence that the secondary structure of domain II differs between the positive and negative strands.

Although the proposed structure was generated by using only the HCV 1a genome sequence, without the aid of phylogenetic comparison, it appears to be applicable to other HCV strains and genotypes. The negative-strand folding is compatible with observed patterns of nucleotide covariance that were used previously to define the positive-strand 5′-UTR structure (40). Four consecutive covariant base pairs (178 to 181:218 to 221) within subdomain IIIb are retained. Other covariant pairings within domains II and III are not retained; however, many of these contain GU wobble pairs in the positive strand (56:107, 59 to 60:104 to 105, and 78:98) or reside in unpaired regions of the negative strand. The proposed folding is also modestly supported by the findings of Vizmanos et al. (53), who reported sequence variability for nt 49 to 287 of the HCV genome, based on a comprehensive comparison of GenBank sequences. Upon evaluation of the variability data with regard to negative-strand RNA sequence, we found that most but not all of the nonunique nucleotide substitutions occur in unpaired positions or are otherwise compatible with the structure. Mutational analysis of secondary structure, such as that carried out on the positive-strand 5′-UTR (40), is necessary to confirm the existence of the putative negative-strand domains.

Recently, using a cell culture-based chimeric HCV RNA replicon system in which viral protein synthesis was uncoupled from IRES function, Friebe et al. (17) and Kim et al. (27) showed that sequences within the 5′-terminal ∼120 nt of the HCV genome are essential for RNA replication. Constructs bearing deletions of nt 5 to 20, 24 to 40, or 72 to 96 did not replicate to detectable levels in transiently transfected cells (17). The results suggest that domains I and II of the positive-strand structure are necessary and sufficient for RNA synthesis (27). Alternatively, the minimal replication signal may reside in the 3′-terminal ∼120 nt of the negative strand. This sequence encompasses domains I′ and IIz′ of the proposed structure (Fig. 2A). Oh et al. (36) demonstrated that a minimal sequence element for in vitro NS5B template utilization is contained within the 3′-terminal 122 to 239 nt (domains I′, IIz′, and possibly IIy′-IIIb′) of the negative strand. Deletion of the 3′-terminal 45 nt severely impairs utilization by NS5B (39). The region mapped by Friebe et al. and Kim et al. may therefore delimit the cis-acting negative-strand structural element which acts as a promoter for NS5B-mediated initiation of positive-strand RNA synthesis.

Interestingly, an interior section of the negative-strand 3′ terminus (nt 107 to 222) bears secondary structure similarity to the positive-strand 3′-X terminus (Fig. 7). As has been noted previously (36), negative-strand nt 122 to 239 exhibit a high degree of primary sequence identity to the 98-nt X RNA. Our structural analysis indicates that, in the context of the surrounding negative-strand sequence, this region likely folds into three stem-loops: IIy′, IIIa′, and IIIb′. The 3′-most of these, IIy′, is a G-rich 45-nt stem-loop reminiscent of SL I of the X region. The hinge region upstream of IIy′ includes the pentanucleotide CCGCA (nt 149 to 153). This sequence occupies the SL I-II hinge in the X RNA of most HCV genotypes. Furthermore, the 3′ border of the 10-nt loop in IIIb′ bears a GAAAGG (nt 192 to 197) motif, a feature shared by SL II of the X region.

FIG. 7.

Region of X RNA structural homology within the HCV negative-strand 3′ terminus. Domains IIIb′ to IIy′ of the proposed negative-strand RNA structure (see Fig. 2) are shown for comparison with the X RNA structure (inset) reported previously (6). Short stretches of primary sequence identity are indicated by bold type. The solid bar delineates a region of the X RNA previously reported to bind recombinant HCV NS5B polymerase in vitro (35).

Notably, the primary sequence of nt 190 to 197 is universally conserved among HCV isolates (53). Although conservation of this region may be attributable to its proposed role in positive-strand IRES function (8), it is also possible that the strict sequence conservation preserves the interaction of domain IIIb′ with a viral or host factor. The secondary-structure context in which the pentanucleotide and polypurine sequences are presented in both 3′ termini is strikingly similar, and it will be of interest to determine whether the three-stem-loop domains contribute similar functions in cis to both HCV RNA strands.

Given that very few antisense constructs targeting the HCV negative strand have been reported to date, the utility of our RNase H map in predicting optimal target sites remains to be assessed. Welch et al. (54) claimed success with a hairpin ribozyme, CNR3, against the negative strand, but the data have not yet been published. Lieber et al. (31) identified a hammerhead ribozyme, Rz2, active against HCV negative-strand RNA in cultured hepatocytes. Although we found the region surrounding this ribozyme cleavage site (nt 342, within the apical loop of domain IV′) to be highly sensitive to single-strand-specific nucleases S1, T₁, T₂, and A (Fig. 2A), the site was inaccessible to hybridization by 11-mer ODNs and/or resistant to cleavage by RNase H (Fig. 4C).

The incongruity of the result of Lieber et al. with our RNase H map might be attributed to the different substrate RNA used (full-length hepatocyte-derived virus versus partial-length in vitro transcript). It may also reflect genuine differences in the annealing properties of RNA and 2′-O-methyl-RNA-DNA constructs or substrate conformational constraints unique to the hammerhead catalytic core and RNase H. That result notwithstanding, the mapping data presented in Fig. 4C are potentially of great value for the development of novel antisense-based diagnostic and therapeutic constructs, targeting the most highly conserved hybridization-accessible regions of the HCV negative strand.

Based on the results summarized in Table 2, it is evident that the oligonucleotide scanning method presented in this study is valid for prediction of accessible hybridization sites within the HCV 3′-X region. The RNase H map revealed positional variations in accessibility which could not be predicted solely from knowledge of secondary structure, single-strand nuclease sensitivity, or other theoretical considerations, such as the GC content of the target site. Our findings dispute the assertion that 2′-O-methyl modification alters the annealing properties of library ODNs so as to appreciably bias the mapping results (4, 19). Therefore, 11-mer 2′-O-methyl RNA-DNA libraries may be useful for the design of therapeutic antisense ODN constructs, including those whose length and backbone modification state differ from that of the library. However, the RNase H mapping data were of limited utility for prediction of susceptible 10-23 deoxyribozyme target sites because these constructs do not necessarily anneal to the same extent as corresponding ODNs and because deoxyribozyme activity is influenced by factors other than the extent of hybridization (Fig. 6).

Under the simulated physiological cleavage assay conditions tested, deoxyribozyme X46 achieved less than 20% cleavage of 3′-UTR RNA even at a 2,000-fold molar excess over target RNA. Cleavage was undetectable below a 20:1 ratio of construct to target. This activity is considerably lower than that reported for other 10-23 deoxyribozymes targeting hepatitis B and C virus transcripts (2, 18, 37). However, it should be noted that the method of our cleavage assay was unique in that it included an RNA renaturing step. Depending on the protocol used to purify the in vitro transcript, the target RNAs used in the previous studies may have sustained a substantial disruption of secondary structure prior to digestion, rendering them more accessible to hybridization by the deoxyribozyme. Also, in several cases (2, 37), the deoxyribozyme arm length was considerably longer than 7 × 7-mer, which may have contributed to the enhanced deoxyribozyme activity. We are presently optimizing the arm length of deoxyribozyme X46 to maximize in vitro cleavage activity and evaluating backbone modifications to enhance stability and pharmacokinetics for intracellular application.

The precedent of other hybridization accessibility mapping studies indicates that efficiency of RNA cleavage in vitro is often (19, 20, 31, 32) but not always (1) a reliable predictor of antisense construct efficacy in cell culture. The unexpectedly low intracellular activity observed with some in vitro-optimized constructs has commonly been attributed to masking of the target site by RNA-binding proteins or complementary RNAs. Although HCV RNA duplexes are reportedly absent from infected hepatocytes (7), various proteins have been shown by deletion mapping or site-directed mutagenesis to bind within the HCV 3′-X sequence (3, 11, 21, 22, 52, 55). The viral RNA-dependent RNA polymerase NS5B has been reported to bind the X region with low affinity (10), though in one instance, binding of this enzyme to the SL I-II junction (Fig. 7) was sufficient to inhibit RNase cleavage in a footprinting assay (35). It will be of interest to determine what effect, if any, the presence of HCV nonstructural proteins has on deoxyribozyme X46 activity. A recently developed oligonucleotide scanning method utilizing endogenous RNA and human RNase H derived from cell extracts (43, 44) may also permit more comprehensive modeling of in vivo antisense hybridization to HCV RNA.

To our knowledge, this is the first report to characterize antisense-directed cleavage of the HCV X region RNA. Further study is needed to evaluate possible inhibitory effects in the context of hepatocyte infection. Deoxyribozyme- or RNase H-mediated cleavage occurring upstream of SL I will ostensibly render the cleaved strand replication defective (16, 29, 56). However, the potential for cross-reactivity with cellular RNAs (30) and the emergence of viral escape mutants bearing nucleotide changes in the target sequence remain to be assessed. In future studies, antisense constructs targeting HCV 3′ termini should be evaluated using cell culture-based models of NS5B-mediated RNA replication (5, 17, 27, 33).