Structural Analysis of the Activation of Ribavirin Analogs by NDP Kinase: Comparison with Other Ribavirin Targets

Materials and Methods

Synthesis of Ribavirin Triphosphate Analogs.

Analytically pure ribavirin was either a kind gift from Pr. Chris Meier (Hamburg, Germany) or was purchased from ICN Pharmaceuticals. 2′,3′-Dideoxy-2′,3′-didehydroribavirin (d4R), 2′,3′-dideoxyribavirin (ddR), and 2′,3′-anhydro-ribavirin (epoxyR) (Fig.1) were synthesized from ribavirin according to the procedures described by Upadhya et al. (1990). 2′-Deoxyribavirin (2′dR) (Pochet and Dugué, 1998) and 3′-deoxyribavirin (3′dR) (Upadhya et al., 1990) were obtained in four steps by radical deoxygenation of appropriately bis-silylated ribavirin at position 2′ and 3′, respectively, as described for the preparation of the 3′-deoxyadenosine (Meier and Huynh-Dinh, 1991). Conversion of ribavirin and analogs into their corresponding triphosphate was performed according to the method of Ludwig and Eckstein (1989). Spectra from electrospray ionization mass spectrometry were recorded in the negative-ion mode. HPLC analyses were performed on a PerkinElmer series 200 unit using a reverse-phase analytical column (Uptisphere UPS HDO, C18, 120 Å, 4.6 × 250 mm, 5 μm; Interchim, Montluçon, France) with detection at 230 nm. The eluants used were acetonitrile/10 mM triethylammonium acetate buffer, pH 6.6, 1/19 (A) and acetonitrile/10 mM triethylammonium acetate buffer, pH 6.6, 3/17 (B). A linear gradient from A to B over 20 min was used with a flow rate of 1 ml/min. Characteristic data for the target compounds are as follows: ribavirin 5′-triphosphate: HPLC, 4.91 min; C₈H₁₅N₄O₁₄P₃(483.98) m/z 483 (M-H); 2′-deoxyribavirin 5′-triphosphate: HPLC, 5.56 min; C₈H₁₅N₄O₁₃P₃(467.98) m/z 467 (M-H); 3′-deoxyribavirin 5′-triphosphate: HPLC, 4.87 min; C₈H₁₅N₄O₁₃P₃(467.98) m/z 467 (M-H); 2′,3′-dideoxy-2′,3′-didehydroribavirin 5′-triphosphate: HPLC, 5.99 min; C₈H₁₁N₄O₁₂P₃(449.97) m/z 449 (M-H); 2′,3′-dideoxyribavirin 5′-triphosphate: HPLC, 6.01 min, C₈H₁₅N₄O₁₂P₃(451.99) m/z 451 (M-H); and 2′,3′-anhydroribavirin 5′-triphosphate: HPLC, 5.79 min; C₈H₁₃N₄O₁₃P₃(465.97) m/z 465 (M-H).

Expression and Purification of NDP Kinases.

Human NDP kinase A was purified as a recombinant protein overexpressed inEsherichia coli (BL21-DE3) using plasmid pJC20, a kind gift from M. Konrad (Max Planck Institute, Göttingen, Germany), according to the procedure described by Schneider et al. (2000). The protein, stored at −20°C in Tris-HCl, pH 7.5, buffer containing 1 mM DTT, 20 mM KCl, and 50% glycerol was >97% pure as determined by SDS-polyacrylamide gel electrophoresis. The H122G single-mutant and the F64W-H122G double-mutant NDP kinases from Dictyostelium discoideum were overexpressed in E. coli (XL1-Blue) and purified as described previously for F64W NDP kinase (Schneider et al., 1998). Neither mutant had measurable NDP kinase activity. Protein concentration is expressed as 17 kDa subunits. It was determined either colorimetrically or using (for a 1 mg/ml solution) an absorbance coefficient A₂₈₀ = 0.55 for H122G andA₂₈₀ = 0.85 for F64W-H122G.

Crystallization of the Complex Ribavirin Triphosphate with NDP Kinase and Data Collection.

H122G D. discoideum NDP kinase in complex with RTP was crystallized by transferring tiny crystals obtained with a higher concentration of PEG, to a drop containing 13% PEG 550 monomethyl ester, 100 mM Tris-HCl, pH 7.5, 30 mM MgCl₂, 10 mg/ml protein, and 20 mM RTP, over wells containing 26% PEG 550 monomethyl ester in the same buffer. The crystals belong to the trigonal space group P3₁21 with unit cell a = b = 71.7 Å, c = 153.8 Å. The asymmetric unit contains a trimer.

X-ray diffraction data from a single crystal were collected at λ = 1.542 Å and 18°C on a Rigaku generator with a MAResearch image plate detector. Diffracted intensities were evaluated with the programs DENZO and SCALEPACK (Otwinowski and Minor, 1997) and further processed using the CCP4 program suite (CCP4, 1994). Overall statistics are given in Table 1. Because the crystals were isomorphous to those of H122G NDP kinase-d4T triphosphate, the 1.85-Å model (Meyer et al., 2000) could be used to calculate phases. Electron density maps were examined using Turbo-FRODO (Roussel and Cambillau, 1991). The first 2Fo-Fc electron density showed clear density for both the protein and the ligand, needing only minor modifications of the protein structure. An RTP molecule and an Mg²⁺ion are bound to each monomer. Water molecules were gradually added during further conjugate gradient refinement with CNS (Brünger et al., 1998). Residues 2 to 5 are missing in the final model for each monomer.

Fluorometric Binding Studies on NDP Kinase.

The affinity of RTP analogs for NDP kinase was determined by following the variation of the intrinsic fluorescence upon nucleotide binding as described previously (Schneider et al., 1998). The fluorescence of the F64W-H122G mutant in T₁ buffer (50 mM Tris-HCl, 75 mM KCl, and 5 mM MgCl₂, pH 7.5) was measured at 330 nm with excitation at 310 nm (2-nm excitation slit and 4-nm emission slit) (PTI, New Brunswick, NJ). Successive aliquots of nucleotide were added to a 1 μM enzyme solution. The inner filter effect was negligible. Experimental binding curves were fitted to a quadratic equation after correction for dilution.

Stopped-Flow Kinetic Experiments and Analysis of Kinetic Results.

Stopped-flow kinetic experiments were performed with a Hi-Tech DX2 (Salisbury, UK) microvolume stopped-flow reaction analyzer equipped with a high-intensity xenon lamp as described previously (Schneider et al., 1998). The excitation wavelength was 304 nm, with a 2-nm excitation slit and a 320-nm cutoff filter at the emission. Mixing was achieved in less than 2 ms. After mixing NDPK (1 μM) and NTP (10–500 μM), the intrinsic protein fluorescence was recorded for 10–200 s. In each experiment, 400 pairs of data were recorded, and the data from three or four identical experiments were averaged and fitted to a number of nonlinear analytical equations using the software provided by Hi-Tech. All curves fitted to single exponentials.

As described previously (Schneider et al., 1998), the data were analyzed using the reaction scheme:The observed single step could be attributed to the phosphotransfer between the nucleotide and the enzyme. Because the product concentration remains very low, the product binding can be neglected. The binding steps are expected to be fast and not detectable in the time range of the stopped-flow experiments. Saturation could not be obtained with the concentrations of nucleotide triphosphate used. Nevertheless, the data are sufficient to determinek₁k₂/k₋₁ork₂/K_S, where K_S is the dissociation constant of the E·NTP complex. This represents the catalytic efficiency of the enzyme phosphorylation step, equivalent to a second-order rate constant and allowing a reliable comparison of different NDP kinase substrates.

Expression and Purification of HCV 1b Polymerase.

TheNS5B gene encoding HCV 1b polymerase was PCR-amplified using the primers HCV-1b-BamHI forward (5′-GGCCGGATCCTCAATGTCCTACACATGGACA-3′) and HCV-1b-SalI reverse (5′-GGCCGTCGACCTAGAGTTTAAGTTTGGATTTTAC-3′) from a clinical HCV isolate serotype 1b to yield Δ55 NS5B 1b cDNA, which specifies an NS5B protein truncated at its C terminus by 55 amino acids. The amplified fragment was digested by BamHI andSalI, cloned into PQE-30 (Amersham Biosciences, Orsay, France), which specifies an N-terminal His₆ tag, and the resulting vector was used to transform BL21pDNAy E. coli cells. Bacteria were cultured at 37°C in Luria broth medium supplemented with 100 μg/ml ampicillin and 25 μg/ml kanamycin untilA_{600 nm} reached 0.6. The temperature of the culture was then switched to 27°C, and expression was induced by the addition of 0.3 mM isopropyl β-d-thiogalactoside for 2 h. Bacteria were harvested by centrifugation, washed twice in phosphate-buffered saline, and the pellet was stored at −80°C until use. Bacteria were lysed for 30 min on ice in 3 ml of buffer 1 (50 mM sodium phosphate, pH 7.5, and 20% glycerol) per gram of bacteria supplemented with 1 mg/ml lysozyme, 1 mM phenylmethylsulfonyl fluoride, and 1 mM benzamidine. Lysates were diluted with 3 ml of buffer 2 (50 mM sodium phosphate, pH 7.5, 20% glycerol, 0.6 M NaCl, 10 mM β-mercaptoethanol, 1.6% Igepal, 20 mM imidazole, and 7 μg of DNase) per gram of bacteria, left on ice for 30 min, and sonicated on ice. Lysates were centrifuged for 30 min at 75,000g, and the cleared lysates were loaded on a 1-ml Hi-Trap heparin column (Amersham Biosciences). After extensive washes in buffer 3 (50 mM NaPO₄, pH 7.5, 10% glycerol, 0.3 M NaCl, 5 mM β-mercaptoethanol, and 10 mM imidazole), bound proteins were eluted with 3 ml of buffer 3 adjusted to 900 mM NaCl. Eluted proteins were then diluted with buffer 3 without NaCl to an NaCl concentration of 0.3 mM and applied to 1.5 ml of a nickel-nitrilotriacetic acid agarose column (QIAGEN, Hilden, Germany). The column was washed extensively before elution of the His₆-tagged HCV polymerase with buffer 3 containing 350 mM imidazole. The fractions containing the HCV polymerase were dialyzed against buffer 3 without imidazole and stored at −20°C.

HIV reverse transcriptase (HIV RT) and phage T7 RNA polymerase (T7 RNAP) were purified as described in by Boretto et al. (2001) and Sousa (2000), respectively.

Polymerase Assays.

Polymerase activity was assayed by monitoring the formation of radiolabeled nucleic acid product absorbed onto DE-81 ion exchange paper discs. All radioactive nucleotides come from Amersham Biosciences. For HIV reverse transcriptase experiments, reactions were performed in RT buffer (50 mM Tris-HCl, pH 8.0, 50 mM KCl, 10 mM MgCl₂, and 0.05% Triton X-100) at 37°C with a bacteriophage MS2 template (Roche, Meylan, France) annealed to the primer (5′-CTCGGTCAGCTACCGAGGAGA-3′; 20 nM primer/template), dATP, dCTP, and dGTP (10 μM each), decreasing amount of ribavirin or analogs (4 mM, 0.8 mM, 160 μM, 32 μM, 6.4 μM, and 1.3 μM), and 5 μM [³H]dTTP (0.12 μCi of [³H]TTP at 1 μCi/μl). Reactions were initiated by the addition of recombinant RT (50 nM). After 30 min, aliquots were withdrawn and spotted onto DE-81 paper discs. Filter paper discs were washed three times for 10 min in 0.3 M ammonium formate, pH 8.0, washed two times in ethanol, and dried. The radioactivity bound to the filter was determined using liquid scintillation counting.

For phage T7 polymerase experiments, reactions were performed in T7 buffer (40 mM Tris-HCl, pH 7.5, 6 mM MgCl₂, 2 mM spermidine, and 10 mM DTT) at 37°C, using 10 ng of DNA template, 1 or 0.5 mM ATP, CTP, UTP, and [α-³²P]GTP (0.1 μCi of [³²P]GTP at 10 μCi/μl). Reactions were initiated by the addition of 1 μg of recombinant T7 RNAP in presence of 0.5 or 1.5 mM RTP or analogs. Aliquots were withdrawn after 30 min and 1 h and spotted onto DE-81 paper discs. Filter papers were processed as described above.

HCV polymerase experiments were performed in RdRp buffer (50 mM HEPES, pH 8.0, 10 mM KCl, 1 mM MnCl₂, 5 mM MgCl₂, and 5 mM DTT) containing 125 ng/μl of homopolymeric cytosine RNA template (Amersham Biosciences) annealed to a guanosine dinucleotide primer (ESGS/Cybergene, Evry, France), and 0.5 mM [α-³²P]GTP (0.1 μCi of [³²P]GTP at 10 μCi/μl). Reactions were initiated by the addition of 300 ng of purified recombinant HCV 1b polymerase and incubated at 30°C. Aliquots were withdrawn overtime and spotted onto DE-81 paper discs. For elongation inhibition, the reaction was incubated without RTP or analogs at 30°C. After 15 min, 0.5 mM RTP or RTP derivatives were added to the samples, and the reaction was allowed to continue for 80 min. Aliquots were withdrawn overtime and spotted onto DE-81 paper discs. The latter were washed and processed as described above. Results shown are representative of three different experiments.

Results

X-Ray Structure of Ribavirin Triphosphate Bound to NDP Kinase.

D. discoideum NDP kinase has 60 and 58% sequence identity, respectively, with the major A and B isoforms of the human enzyme, also called Nm23-H1 and Nm23-H2 (Lacombe et al., 2000). All eukaryotic NDP kinases are hexamers, including the D. discoideum and human proteins. Their subunit folds are very similar, and their active sites are essentially identical (Janin et al., 2000). Structural data show the same mode of binding for GDP to human NDP kinase B (Morera et al., 1995) and ADP or TDP to wild-type D. discoideum NDP kinase (Cherfils et al., 1994; Morera et al., 1994). Thus, the D. discoideum enzyme is a good model of human NDP kinase for functional studies. In addition, the H122G mutant is useful when studying nucleoside triphosphate binding, because the substitution of the active site histidine prevents transfer of the γ-phosphate to the protein. We previously used the mutant in the analysis of a complex with d4T triphosphate at 1.85-Å resolution (Meyer et al., 2000). With the wild-type protein, bound ATP can be mimicked by a complex of ADP and either aluminum or beryllium fluoride (Xu et al., 1997). The present X-ray structure of H122G in complex with ribavirin triphosphate has an R factor of 20.7% (R_free = 24.9%) at 2.9 Å resolution and good stereochemistry (Table 1). The protein has essentially the same conformation as in the d4T triphosphate complex, a superposition yielding a RMS distance of 0.26 Å for all equivalent Cα positions. The superposition with GDP-bound human NDP kinase B is almost as good, with an RMS distance of 0.41 Å, excluding nine C-terminal residues that differ in conformation between the D. discoideum and human proteins (Lacombe et al., 2000).

RTP binds in the same site and orientation as other nucleotide substrates or nucleotide analogs (Fig.2). The substituted triazole that replaces the base is in a slit opening on the protein surface and stacks between the side chains of Val-116 and Phe-64. The 2.9-Å electron density map uniquely defines the plane of the five-membered ring, but it is compatible with a 180° rotation of the ring about the glycosidic bond. Thus, the complex may contain either thesyn and anti conformers of the nucleoside analog or a mixture of the two. In addition, the carboxamide group in C3 of the triazole can also flip by 180° relative to the ring. None of these rotations enables the substituted triazole to make polar interactions with the protein. In the GDP complex, guanine makes a hydrogen bond with the side chain carboxylate of the C-terminal glutamate of an adjacent subunit in the hexamer. The distance of the carboxylate to the carboxamide group in C3 of the triazole is more than 5 Å, too long for a hydrogen bond.

Unlike d4T, which has a double bond and no hydroxyl in C2′-C3′, ribavirin carries an unmodified ribose. The sugar in bound RTP has the same 3′-endo ring pucker as in ADP bound to D. discoideumNDP kinase (Morera et al., 1994) or GDP in the human enzyme (Morera et al., 1995). The 2′- and 3′-hydroxyl groups make the same interactions as in these two complexes (i.e., with the amide group of Asn-119 and the amino group of Lys-16). All polar interactions that involve the phosphate groups are also conserved (Table2). The γ-phosphate of RTP, like that of d4T triphosphate (Meyer et al., 2000) interacts with Lys-16, Tyr-56, and the main chain NH of Gly-123. In the wild-type protein, fluoride ions make the same interactions in the ADP-beryllium and aluminum fluoride complexes (Xu et al., 1997).

All the complexes cited above contain a Mg²⁺ ion. The metal interacts with oxygens of two phosphates if the ligand is a diphosphate and oxygens of three phosphates if it is a triphosphate, which is the case of RTP. In either case, water molecules complete the octahedral coordination. The metal takes two positions approximately 2 Å apart in the complexes with di- and triphosphates, illustrated in Fig. 2, respectively, by the human NDP kinase B-GDP and the H122G-RTP structures. The metal-oxygen distances remain in the range 2.0 to 2.4 Å and the phosphate groups themselves do not move: the α- and β-phosphates of bound RTP and GDP superimpose to within 0.5 Å (that is, within the error in the coordinates of a X-ray structure at 2.9 Å resolution).

Catalytic Efficiencies of NDP Kinase for Ribavirin Triphosphate and Derivatives.

We followed the time course of the reaction of RTP with human NDP kinase A in fast kinetic experiments by monitoring the intrinsic fluorescence of the protein. The fluorescence is quenched upon phosphorylation of the active site histidine by a NTP substrate and, reciprocally, enhanced upon its dephosphorylation by NDP (Deville-Bonne et al., 1996). The reaction with RTP yields a monoexponential fluorescence decay without a lag, as previously reported for other nucleoside triphosphates (Schaertl et al., 1998;Schneider et al., 1998). The pseudo–first-order constantk_obs of the exponential decay increased linearly with RTP concentrations in the range investigated (5–50 μM) (Fig. 3). The slope of the linear fit, which measures the catalytic efficiency of RTP as a substrate for NDP kinase, wask₂/K_S= 3 × 10⁵M⁻¹s⁻¹. This makes RTP a good substrate, comparable with the natural substrate CTP, although inferior to GTP, which is the best known substrate of NDP kinase (Table3).

The phosphorylation kinetics were also studied with RTP analogs modified on the ribose. The absence of a 2′-hydroxyl group in 2′dRTP led to a 5-fold loss in catalytic efficiency, comparable with the 2-fold change observed between GTP and dGTP (Fig. 3A and Table 2). Removal of the 3′-hydroxyl group had a much larger effect, causing a drop in activity by a factor of 10⁴ with both 3′dRTP and d4RTP. The factor reached 10⁵ in the case of the 2′,3′-dideoxy analog (ddRTP) (Fig. 3B and Table 2). Very similar observations have been made with thymidine analogs: the d4T derivative was 10-fold better that the dideoxy derivative when tested as a substrate of NDP kinase (Schneider et al., 2000). The 2′3′-epoxy analog of RTP (epoxyRTP) is also better than the dideoxy compound. Acyclovir triphosphate was also assayed for reaction with NDP kinase. Acyclovir, which is used in herpes virus therapy, is a guanosine analog with a linear connection between the base and 5′-hydroxyl group. The phosphorylation reaction followed in the fluorescence test was 10⁵ times slower than for GTP and 10⁴ times slower than for RTP. This confirms the dominant role of the sugar hydroxyl groups and the minor importance of the base in the reaction with NDP kinase.

Binding Affinities of Ribavirin Triphosphate and Derivatives for NDP Kinase.

Although the equilibrium dissociation constantK_S of the NDP kinase-NTP complex is not accessible in these kinetic experiments, it can be estimated by using a mutant NDP kinase, which lacks the catalytic histidine and where the phenylalanine (Phe-64 in D. discoideum NDP kinase) stacking on the base is replaced by a tryptophan. In the H122G-F64W double mutant, the protein fluorescence changes upon nucleotide binding although no phosphorylation takes place. Titration by RTP yieldedK_S = 24 ± 5 μM, a 3-fold decrease in affinity relative to CTP and 160-fold relative to GTP (Table 3). RTP analogs with a modified sugar showed a further decrease in affinity by a factor of 2 for 2′dRTP, 10 for 3′dRTP and d4RTP, and 100 for ddRTP. Although smaller than for the catalytic efficiencyk₂/K_S, these ratios indicate that all analogs bind NDP kinase with lower affinities than the natural substrates.

Inhibition of Viral Polymerase by RTP and Analogs.

Ribavirin has been reported to exhibit some anti-HIV activity, but its combination with other nucleotide analogs leads to serious adverse effects (Sim et al., 1998). Indeed, although ribavirin is a ribonucleoside analog, RTP may also inhibit the DNA polymerase activity of HIV RT. Because 2′,3′-dideoxynucleotides are efficient against HIV RT, it was of interest to test whether 2′- or 3′-deoxy-modified RTP is efficient against HIV RT and retains inhibitory power against HCV polymerase.

We first analyzed the activity of RTP and derivatives against DNA polymerization by HIV RT. Because RMP can be incorporated opposite either cytidine or uridine, we used templates containing all four natural ribonucleotides. HIV RT was not affected by any of the compounds at concentration up to 4 mM (data not shown), suggesting that they are either not incorporated or not chain terminators. We next tested whether ribavirin derivatives inhibit RNA polymerases. Phage T7 RNAP is the best known RNA polymerase and constitutes an interesting model for inhibition. Polymerization was assayed using a DNA template containing the T7 promotor, the four natural nucleotides, and increasing amounts of RTP or derivatives. No significant polymerization inhibition was observed, suggesting that RTP and derivatives are not inhibitors of T7 RNA polymerase (data not shown).

RTP and derivatives were then tested against HCV polymerase. HCV polymerase activity is characterized by a nonprocessive initiation step, followed after 10 to 15 min by linear incorporation (elongation step). We measured the nucleotide incorporation over time in presence of RTP or RTP analogs. Polymerase activity was determined by calculating the rate of nucleotide incorporation after the initiation step. RTP, 2′dRTP, and epoxyRTP inhibited HCV polymerase; 2′dRTP was the most efficient (Fig. 5, 85% of relative inhibition). The other RTP analogs inhibited HCV polymerase weakly, if at all.

We wanted next to test whether the lower nucleotide incorporation was caused by the initiation or elongation step. During the first 15 min, we observed a decrease of nucleotide incorporation induced by RTP, epoxyRTP, and 2′dRTP, whereas 3′dRTP, ddRTP, or d4RTP had no effect or a slight activating effect (Fig. 6A). To measure inhibition of the elongation step, RTP analogs were added after 15 min at the end of the initiation step, and the reaction continued for 80 min. Under these conditions, elongation was inhibited only by RTP, with little effect of the other analogs (Fig. 6B).

In summary, RTP inhibits both initiation and elongation, 2′dRTP and epoxy-RTP act only on initiation, and 3′dRTP, ddRTP, and d4RTP have no effect on nucleotide incorporation. This suggests that both the 3′ and 2′ hydroxyl groups are required for inhibiting elongation but that the presence of an oxygen at the 3′ position is sufficient for inhibiting initiation.

Discussion

Unlike nucleoside analogs directed against HIV RT, which lack a 3′OH, ribavirin contains a normal ribose moiety and cannot act as a chain terminator. The derivatives tested here combine the base modification of ribavirin and some of the sugar modifications found in RT inhibitors. We tested the capacity of NDP kinase to produce the triphosphate form of these derivatives and the effect of the triphosphate form on several target viral polymerases.

Correlation between Binding Affinities and Catalytic Efficiencies of Nucleotides for NDP Kinase.

RTP and its analogs were found to be excellent substrates for NDP kinase on the condition that the 3′hydroxyl group is present. This is in line with previous studies of natural substrates and analogs. Figure 7 shows a correlation between the catalytic efficiencies and the equilibrium association constantK_A for natural nucleotides and analogs. Substrates are found to belong to two classes: those carrying a 3′-OH and those lacking this hydroxyl group. The two lines obtained for the two classes indicate a change in catalytic efficiency by a factor 500 to 1000, illustrating the well-established contribution of the 3′-OH to substrate-assisted catalysis by NDP kinase. The lines themselves show that, within each class, the catalytic efficiency correlates linearly to the binding affinity.

Tighter substrate binding makes for more efficient phosphotransfer. Thus, GTP has both the highest reactivity and the best affinity (K_D = 0.15 μM). Affinity and reactivity are less for RTP. The structures of GDP bound to the active enzyme and RTP bound to the H122G variant (Fig. 1) suggest a simple reason for the difference: the ribavirin base, which is smaller than guanine or other natural bases, buries less surface upon binding the enzyme. Thus, the nonpolar contribution to the binding energy is less, whereas other interactions are conserved. All complexes of NDP kinase with nucleotides (at least 10 X-ray structures are known) indicate a similar mode of binding for all, except for the presence or absence of the 2′-OH, and the size of the base surface in nonpolar contact with the protein. Both the polar interactions made by the 3′OH and the nonpolar contact contribute to hold the substrate in place.

The Use of Ribavirin Nucleotide as RNA Polymerization Inhibitors.

Ribavirin is currently used as a therapeutical agent against HCV in combination with interferon (Lauer and Walker, 2001). Although it is a weak antiviral drug when used alone, ribavirin is phosphorylated by cellular kinases up to the triphosphate level, and ribavirin monophosphate is incorporated into the viral RNA by the recombinant HCV polymerase (Maag et al., 2001). Templates containing RMP can also block RNA elongation. This makes ribavirin an interesting candidate for chemical modification aiming to increase its antiviral effect.

The RNA replication of the HCV genome goes through two steps relevant to our discussion: the initiation step, comprising the incorporation of the very first nucleotide, up to the 4th or 5th nucleotides, and the elongation step, which consists of processive extension of the neosynthesized RNA. The initiation step, and especially the synthesis of the first phosphodiester bond between the two first nucleotides, is nonprocessive and rate-limiting (Kao et al., 2001). Moreover, the switch from initiation to elongation requires a conformational change. Therefore, polymerization can be inhibited during the initiation step, by preventing the switch from initiation to elongation, or during processive synthesis. We have observed that 2′dRTP and epoxyRTP are HCV polymerase inhibitors if they are present at the beginning of the reaction, but not if they are added after 15 min of reaction. This suggests that 2′dRTP and epoxyRTP are initiation inhibitors only. 3′dRTP, ddRTP, and d4RTP, which were thought to be chain terminators, do not have the expected inhibitory effect, during either initiation or elongation, suggesting that they are not incorporated by the HCV enzyme. In contrast, RTP inhibits both initiation and elongation.

What is the detailed mechanism of this inhibitory effect? HCV polymerase activity is stimulated in the presence of high concentrations of GTP (Lohmann et al., 1998), and kinetic studies revealed two K_M values for GTP, one similar to that of other nucleotides, and a higher one (Luo et al., 2000). Because ribavirin is a guanosine analog, it can compete either for the incorporation of GTP in the nascent RNA or for the stimulating effect of GTP. Recently, an allosteric GTP-binding site has been characterized far away from the catalytic site and is not involved in the polymerization reaction (Bressanelli et al., 2002). This allosteric GTP-binding site might also represent a target for GTP analogs. Indeed, because RTP, epoxyRTP, and 2′dRTP inhibit the initiation step of the polymerization, they may compete for this GTP-binding site and may inhibit the allosteric activation or the switch from initiation to elongation, normally induced by the binding of GTP to this site.

Epoxy-RTP, which does not carry a 3′-OH, is a putative chain terminator, but it does not show the expected effect, at least during the elongation step of polymerization. RTP and 2′dRTP carry a 3′-OH and should not be chain terminators. Nevertheless, they, like epoxy-RTP, inhibit the initiation of synthesis by HCV polymerase. This suggests that a 3′-hydroxyl group, or at least the presence of an oxygen at the 3′ position, is required to inhibit this step. Inhibition of elongation requires both hydroxyls in position 2′ and 3′, making RTP the only efficient inhibitor of this step. RTP may act at either step by competing for the NTP binding site or by blocking incorporation per se, as already observed by Maag et al. (2001).

In conclusion, the experiments suggest that 2′-deoxy ribavirin may be an alternative to ribavirin in therapies against C hepatitis. It is efficiently activated to the triphosphate form by NDP kinase and is a potent specific inhibitor of HCV polymerase initiation. The absence of the 2′OH on 2′deoxyribavirin is likely to result in a weaker interaction with IMP dehydrogenase, the cytotoxic cellular target, because the active site side chains interact with both 2′OH and 3′OH of the nucleotide (Colby et al., 1999). Cellular studies are now needed to validate this in vitro study.