2′-Deoxy-4′-azido Nucleoside Analogs Are Highly Potent Inhibitors of Hepatitis C Virus Replication Despite the Lack of 2′-α-Hydroxyl Groups*

RNA polymerases effectively discriminate against deoxyribonucleotides and specifically recognize ribonucleotide substrates most likely through direct hydrogen bonding interaction with the 2′-α-hydroxy moieties of ribonucleosides. Therefore, ribonucleoside analogs as inhibitors of viral RNA polymerases have mostly been designed to retain hydrogen bonding potential at this site for optimal inhibitory potency. Here, two novel nucleoside triphosphate analogs are described, which are efficiently incorporated into nascent RNA by the RNA-dependent RNA polymerase NS5B of hepatitis C virus (HCV), causing chain termination, despite the lack of α-hydroxy moieties. 2′-Deoxy-2′-β-fluoro-4′-azidocytidine (RO-0622) and 2′-deoxy-2′-β-hydroxy-4′-azidocytidine (RO-9187) were excellent substrates for deoxycytidine kinase and were phosphorylated with efficiencies up to 3-fold higher than deoxycytidine. As compared with previous reports on ribonucleosides, higher levels of triphosphate were formed from RO-9187 in primary human hepatocytes, and both compounds were potent inhibitors of HCV virus replication in the replicon system (IC50 = 171 ± 12 nm and 24 ± 3 nm for RO-9187 and RO-0622, respectively; CC50 >1 mm for both). Both compounds inhibited RNA synthesis by HCV polymerases from either HCV genotypes 1a and 1b or containing S96T or S282T point mutations with similar potencies, suggesting no cross-resistance with either R1479 (4′-azidocytidine) or 2′-C-methyl nucleosides. Pharmacokinetic studies with RO-9187 in rats and dogs showed that plasma concentrations exceeding HCV replicon IC50 values 8-150-fold could be achieved by low dose (10 mg/kg) oral administration. Therefore, 2′-α-deoxy-4′-azido nucleosides are a new class of antiviral nucleosides with promising preclinical properties as potential medicines for the treatment of HCV infection.

RNA polymerases effectively discriminate against deoxyribonucleotides and specifically recognize ribonucleotide substrates most likely through direct hydrogen bonding interaction with the 2-␣-hydroxy moieties of ribonucleosides. Therefore, ribonucleoside analogs as inhibitors of viral RNA polymerases have mostly been designed to retain hydrogen bonding potential at this site for optimal inhibitory potency. Here, two novel nucleoside triphosphate analogs are described, which are efficiently incorporated into nascent RNA by the RNA-dependent RNA polymerase NS5B of hepatitis C virus (HCV), causing chain termination, despite the lack of ␣-hydroxy moieties. 2-Deoxy-2-␤-fluoro-4-azidocytidine (RO-0622) and 2-deoxy-2-␤-hydroxy-4-azidocytidine (RO-9187) were excellent substrates for deoxycytidine kinase and were phosphorylated with efficiencies up to 3-fold higher than deoxycytidine. As compared with previous reports on ribonucleosides, higher levels of triphosphate were formed from RO-9187 in primary human hepatocytes, and both compounds were potent inhibitors of HCV virus replication in the replicon system (IC 50 ‫؍‬ 171 ؎ 12 nM and 24 ؎ 3 nM for RO-9187 and RO-0622, respectively; CC 50 >1 mM for both). Both compounds inhibited RNA synthesis by HCV polymerases from either HCV genotypes 1a and 1b or containing S96T or S282T point mutations with similar potencies, suggesting no cross-resistance with either R1479 (4-azidocytidine) or 2-C-methyl nucleosides. Pharmacokinetic studies with RO-9187 in rats and dogs showed that plasma concentrations exceeding HCV replicon IC 50 values 8 -150-fold could be achieved by low dose (10 mg/kg) oral administration. Therefore, 2-␣-deoxy-4-azido nucleosides are a new class of antiviral nucleosides with promising preclinical properties as potential medicines for the treatment of HCV infection. 3 infection is a major cause of chronic liver disease, cirrhosis, and hepatocellular carcinoma and is the leading cause of liver transplantation. Current treatment options available to HCV-infected persons have limitations with regard to efficacy and tolerability. Only about 50% of individuals infected with HCV genotype 1 achieve sustained virological response when treated with a combination of pegylated interferon ␣ and ribavirin (1,2). In addition, high viral load, age, body weight, co-infection with human immunodeficiency virus, and cirrhosis negatively affect the probability of achieving sustained virological response (3,4). Therefore, there is an urgent need to develop new and more effective therapies for the treatment of HCV infection. A number of new antiviral candidates are currently being evaluated in clinical studies, the majority targeting either the HCV protease or HCV polymerase enzymes, which are essential for viral replication (5). The HCV RNA-dependent RNA polymerase, NS5B, contains the active site responsible for viral RNA synthesis and functions as part of a membrane-associated replicase complex. Nucleoside and non-nucleoside inhibitors of HCV polymerase have successfully progressed into clinical development. Most antiviral nucleoside analogs are further metabolized to the corresponding nucleoside 5Ј-triphosphate analogs by cellular enzymes. Nucleoside 5Ј-triphosphate analogs then function as alternative substrates for the viral polymerase, competitively inhibit viral nucleic acid synthesis, and can terminate nucleic acid synthesis after incorporation.

Hepatitis C virus (HCV)
Previously, R1479 was identified as a potent and selective inhibitor of HCV replication in cell culture (6,7). R1626, a prodrug of 4Ј-azidocytidine, is presently in phase II clinical development. In a multiple ascending dose phase I study, R1626 showed dose proportional pharmacokinetics and robust antiviral effects in HCV-infected persons. When administered for 14 days, R1626 achieved mean HCV viral load reductions of 1.2, 2.6, and 3.7 log IU/ml at dose levels of 1500, 3000, and 4500 mg, twice a day, respectively (8). R1626 therefore provided clinical proof-of-concept for 4Ј-modified ribonucleoside analogs as inhibitors of HCV replication. Further studies were aimed at increasing the antiviral potency of 4Ј-modified nucleosides as potential second generation inhibitors with improved efficacy for the treatment of HCV infection. Antiviral potency of nucleosides is often limited by phosphorylation efficiency, as modified analogs can be poor substrates for human nucleoside kinases (9). To identify novel nucleosides with increased antiviral potency, the focus of optimization was therefore on increasing phosphorylation efficiency by human kinases as well as on increasing intrinsic incorporation efficiency by the recombinant NS5B enzyme. RNA polymerases are believed to discriminate against 2Ј-deoxyribonucleotide incorporation by forming a hydrogen bond interaction with the 2Ј-␣-hydroxyl group that differentiates ribonucleotides from deoxyribonucleotides (10 -12). Structural analysis of HCV polymerase in complex with ribonucleotides suggested that the conserved Asp-225 of HCV NS5B functions in this manner to confer ribonucleotide specificity (13). We tested the hypothesis that the addition of a 4Ј-substituent could provide novel interactions in the HCV polymerase active site to compensate for the absence of a 2Ј-␣-hydroxy group. 2Ј-Deoxy-2Ј-␤-fluoro-4Јazidocytidine (RO-0622) and 2Ј-deoxy-2Ј-␤-hydroxy-4Ј-azidocytidine (RO-9187) were found to be surprisingly potent inhibitors of HCV replication with antiviral potencies up to 50-fold higher than that of R1479, despite the lack of hydrogen bonding potential with the conserved Asp-225 of NS5B.

EXPERIMENTAL PROCEDURES
Human Nucleoside Kinase Assays-Recombinant human deoxycytidine kinase (dCK), thymidine kinases 1 and 2 (TK1 and TK2), and uridine-cytidine kinases 1 and 2 (UCK-1 and UCK-2) were expressed and purified using a bacterial vector system Escherichia coli BL21 (DE3) pLysS (14 -18). The proteins were purified by metal chelate affinity chromatography using a Ni 2ϩ -nitrilotriacetic acid-agarose resin. The His tag was removed from the purified proteins with thrombin. Kinase activity was measured in an adenosine 5Ј-triphosphate transfer assay performed with 0.05 M [␥-32 P]ATP (10 Ci/l), 100 M ATP, 50 mM Tris-HCl, pH 7.6, 5 mM MgCl 2 , 100 mM KCl, 0.5 mg/ml bovine serum albumin, 10 mM dithiothreitol, and different concentrations of nucleoside analogs. The reaction was initiated by adding 50 ng of the enzyme followed by incubation at 37°C for 25 min and terminated by boiling. 2 ml of the supernatant was applied to polyethyleneimine cellulose F TLC plates. Chromatography was performed for 8 -12 h with isobutyric acid/NH 4 OH/H 2 O (66:1:33) (v/v) as the mobile phase. The products of the kinase reaction were detected by autoradiography. The spots were excised and eluted with 0.5 ml of 0.2 M KCl, 0.1 M HCl (1:1, v/v) and quantified by liquid scintillation counting or by the Fuji BAS 2500/LAS 1000 PhosphorImaging system, with an Image Reader version 1.7E. Kinetic parameters were determined using nonlinear regression analysis with SigmaPlot 2001 software and Michaelis-Menten equation.
HCV Replicon Assays-HCV replicon assays were performed using either the 2209-23 cell line or the transient replicon sys-tem in cured Huh-7 cells as described (6,19). Nucleoside analogs were synthesized at Medivir, dissolved in Me 2 SO, and then diluted in Dulbecco's modified Eagle's medium with 5% (v/v) fetal bovine serum before addition to cells. The final concentration of Me 2 SO was 1% in all experiments. Quantification of Renilla and firefly luciferase activities was performed using luciferase kits from Promega according to the manufacturer's instructions. The WST-1 cell proliferation assay (Roche, Palo Alto, CA) was used to measure cell viability. The effect of compounds on the incorporation of tritiated thymidine into cellular DNA was measured using the [ 3 H]thymidine incorporation scintillation proximity assay system from Amersham Biosciences.
HCV Polymerase Assay-The HCV polymerase activity was measured as the incorporation of radiolabeled nucleotide monophosphates into acid -insoluble RNA products using NS5B570-BK protein and cIRES RNA template as described previously (6)  Gel-based Nucleotide Incorporation Assay-The RNA template-directed incorporation and extension of the nucleotide and nucleotide analogs by HCV NS5B were performed with a 19-nucleotide RNA oligonucleotide (5Ј-AUGUAUAAUUAU-UGUAGCC-3Ј) and a 5Ј-end-labeled GG primer (Dharmacon, Inc.). The RNA template was predicted to form a single stemloop with an unpaired 3-nucleotide sequence at the 3Ј-end. The nucleotide incorporation reactions and RNA product separation on denaturing polyacrylamide gel were conducted as described (6), except that the reactions were carried out with 0.25 M GG primer and nucleotide triphosphates at the indicated concentrations for 30 min. The kinetics of the single nucleotide incorporation of CTP and CTP analogs were determined with the same RNA oligonucleotide template and primer pair for 25 min at the following assay conditions: 7.5 M 19-nucleotide RNA template, 10 M unlabeled GG primer, 0.075 M 5Ј-end-labeled GG primer, 3 M NS5B, and serial dilutions of CTP or each CTP analog. The GG primer input and the GGC RNA product were quantified using a Typhoon9400 PhosphorImager scanner and ImageQuant software (GE Healthcare). The reaction velocity was plotted as a function of the nucleotide concentration, and the single nucleotide incorporation kinetic parameters were derived by fitting the data to the Michaelis-Menten equation.
Analysis of Phosphorylation in Primary Human Hepatocytes-Nucleoside phosphorylation efficiency and the intracellular half-life of the 5Ј-triphosphate of RO-9187 were determined by incubating fresh primary human hepatocytes with tritium-labeled RO-9187 as described (20) with the following modifications of the cell extraction procedure. The soluble content of the primary hepatocytes was incubated with 500 l of 1.5 M perchloric acid for 30 min on ice, followed by neutralization with 300 l of 2.5 N potassium hydroxide. The supernatant was collected after centrifugation at 10,000 ϫ g for 10 min to remove the cell debris and precipitated macromolecules. The pH of the supernatants was adjusted by the addition of 200 mM Tris-HCl, pH 7.5. The phosphates of RO-9187 were separated by ion exchange high performance liquid chromatography as described (20).
Compound Crystallography-Suitable crystals of NM107 were obtained by evaporation of a saturated solution with methanol as solvent. Crystals were mounted in loops and cooled to 100 K in a nitrogen stream. Diffraction data were collected at the Swiss light source beamline X10SA using a MAR CCD225 detector with synchrotron radiation (0.70 Å), and data were processed with the program XDS. The crystal structures were solved and refined with SheIXTL (Bruker AXS, Karlsruhe, Germany). Crystals of RO-9187 were also obtained from methanol evaporation, mounted in loops, and cooled to 89 K. Data from one crystal were collected on a STOE imaging plate diffraction system (STOE & Cie GmbH, Darmstadt, Germany) with molybdenum radiation (0.71 Å), and data were processed with STOE imaging plate diffraction system software.
Pharmacokinetic Studies-Pharmacokinetic studies were performed in Hanover-Wistar rats (Charles River Laboratories, Inc., Hollister, CA), beagle dogs, and cynomolgus monkeys. Three animals per dose regime were administered either single 10 mg/kg intravenous bolus doses (50% cyclodextran/water) or single 10, 200, or 2000 mg/kg oral solution using an aqueous vehicle containing 0.5% hydroxypropylmethylcellulose, 0.4% polysorbate 80, and 0.9% benzyl alcohol. Plasma concentrations of test compounds were determined by liquid chromatography and tandem mass spectroscopy with a lower limit of quantification of 5 ng/ml.

Deoxyribonucleosides and Dideoxyribonucleosides Are Poor
Substrates and Inhibitors of HCV Polymerase-RNA polymerases synthesize RNA molecules by polymerization of ribonucleotide triphosphate substrates through a divalent metal ion-catalyzed reaction mechanism. Discrimination against 2Ј-␣-deoxyribonucleotide triphosphates is believed to be achieved through conserved amino acid residues in the active site of RNA polymerases, which form hydrogen bonds with the 2Ј-␣-hydroxy group of ribonucleotides (10,11). To determine the substrate specificity of nucleotide incorporation by HCV polymerase, RNA synthesis reactions were performed in the presence of either CTP, 2Ј-dCTP, 3Ј-dCTP or 2Ј,3Ј-ddCTP substrates, and the CTP analogs were also characterized as inhibitors of RNA synthesis by HCV polymerase. The chemical structures of nucleosides used in this study are shown in Fig. 1. CTP and 3Ј-dCTP were efficiently incorporated into nascent RNA. 3Ј-dCTP prevented further chain elongation as an obligatory chain terminator. As compared with the natural substrate CTP, the incorporation efficiency of 3Ј-dCTP was 14-fold lower, mostly driven by an increased K m value for 3Ј-dCTP (Table 1 and Fig. 2). 3Ј-dCTP was also a potent inhibitor of RNA synthesis by HCV polymerase in vitro ( Table 2). In contrast, 2Ј-dCTP was not efficiently incorporated and did not inhibit RNA synthesis from cIRES RNA template (Table 2 and Fig. 2). Similarly, 2Ј,3Ј-ddCTP did not inhibit RNA synthesis by HCV polymerase ( Table 2). The comparative results between CTP and 3Ј-deoxy-CTP on one hand and 2Ј-deoxy-CTP and 2Ј,3Ј-deoxy-CTP on the other hand indicated a significant contribution of the 2Ј-␣hydroxy group on binding and incorporation by HCV polymerase. These results are consistent with the model of 2Ј-deoxyribonucleo-  side discrimination through the lack of hydrogen bonding interaction in the RNA polymerase active site. 4Ј-Substitution Can Increase the Inhibitory Potency of 2Ј-␣-Deoxyribonucleotides-Interestingly, 2Ј-␤-hydroxy-CTP (ara-CTP) showed a moderate inhibitory effect on RNA synthesis despite the lack of a 2Ј-␣-hydroxy group, suggesting that substituents in the ␤-configuration may allow additional binding interactions in the NS5B active site (Table 2). However, the incorporation efficiency of ara-CTP was 140-fold lower as compared with CTP, in line with an important interaction of the 2Ј-␣-hydroxy group. Earlier, we discovered that 4Ј-substituents on ribonucleoside analogs may pick up additional binding interactions in the HCV polymerase active site, with 4Ј-azido being the most effective to allow productive binding to both HCV polymerase and human nucleoside kinases (6,7,9). The addition of a 4Ј-azido moiety to the ara-CTP molecule resulted in an unexpected 15-30-fold increase in inhibitory potency of 2Ј-␣-deoxy-2Ј-␤-hydroxy-4Ј-azido-CTP (RO-9187-TP) as compared with 2Ј-␣-deoxy-ara-CTP. With this, the overall inhibitory potency of RO-9187-TP was similar to that of the ribonucleoside analog 4Јazido-CTP (R1479-TP) in the cIRES RNA-dependent RNA polymerase assay ( Table 2). The ␤-hydroxy moiety could be replaced with a ␤-fluoro moiety with similar inhibitory potency in the RNA synthesis assay (2Ј-␣deoxy-2Ј-␤-fluoro-4Ј-azido-CTP; RO-0622-TP). Both RO-9187-TP and RO-0622-TP were used as substrates for incorporation into nascent RNA by HCV NS5B and prevented further chain elongation, similar to the obligatory chain terminator 3Ј-dCTP (Fig. 2). 4Ј-Substitution improved incorporation efficiencies of RO-9187-TP and RO-0622-TP ϳ8and 18-fold as compared with ara-CTP, respectively (Table 1).
RO-9187 and RO-0622 Are Phosphorylated by Human Deoxycytidine Kinase-To achieve antiviral potency as competitive inhibitors of HCV polymerase, nucleoside analogs need to be phosphorylated to their 5Ј-triphosphate analogs by human nucleoside kinases. RO-9187 and RO-0622 together with control compounds 4Ј-azidocytidine (R1479), 2Ј-␤-methylcytidine (NM107), ara-C, dC, dT, and C were tested as substrates of recombinant human dCK, thymidine kinases 1 and 2 (TK1 and TK2), and uridine-cytidine kinases 1 and 2 (UCK-1 and UCK-2). In an initial screen, nucleoside analogs were incubated with human kinases at a single concentration of 100 M, and the phosphorylation products were quantified by thin layer chromatography. Kinetic parameters were then determined for the kinase, which provided sufficient product formation for each nucleoside. None of the compounds tested were quantifiable substrates of TK1 or UCK-2 under the reaction conditions. Nucleoside analogs R1479, RO-9187, RO-0622, and ara-C were substrates for human dCK. Ara-C and RO-0622 were also substrates for human TK2. R1479 and RO-9187 were phosphorylated with lower efficiency as compared with the natural substrate 2Ј-deoxycytidine (dC), whereas RO-0622 and ara-C were phosphorylated with 3-and 2-fold increased efficiency relative to deoxycytidine, respectively (Table 3). With this, RO-0622 was phosphorylated with 11-fold increased efficiency as compared with R1479. Interestingly, 2Ј-␤-methylcytidine (NM107) was not phosphorylated by dCK but was a substrate of UCK-1.
RO-9187 and RO-0622 Are Potent Inhibitors of HCV Replication-The antiviral potencies of the novel nucleoside analogs RO-9187 and RO-0622 were determined in the HCV genotype 1b (Con1) subgenomic replicon cell line 2209-23 and in the transient replicon system as described previously (6,19). Both compounds were potent inhibitors of HCV replication in these systems (Table 4). RO-9187 had similar or 5-10-fold higher antiviral potency as compared with R1479 and NM107. RO-0622 was 50 -100-fold more potent as compared with R1479 and NM107 (Table 4). RO-9187, RO-0622, as well as R1479 and NM107 inhibited HCV genotype 1a-and genotype  1b-driven replication with similar potency, as expected for nucleoside analogs (Table 4). Interestingly, RO-9187 showed 4 -5-fold higher antiviral potency in the stable 2209-23 cell line as compared with the transient replicon system, whereas all other compounds showed similar potencies between the stable and transient systems (Table 4). RO-9187 and RO-0622 were not cytotoxic and did not inhibit tritiated thymidine incorporation in 2209-23 cells, i.e. these compounds did not inhibit cell proliferation (Table 4). In contrast, ara-C was potently cytostatic and inhibited 2209-23 cell proliferation with an IC 50 of 29 nM, which prevented the assessment of its potential antiviral effect in cell culture. Similarly, 2Ј-␤-fluorocytidine (RO-8013) was potently cytostatic in 2209-23 cells. These results demonstrate that the introduction of a 4Ј-substituent effectively abolished the inhibitory effect on cell proliferation and increased antiviral selectivity as compared with ara-C and RO-8013.
Interestingly, RO-0622 and RO-9187 were also potent inhibitors of replicon variants carrying S96T or S282T point mutations in NS5B. These mutations were previously shown to confer resistance to R1479 and NM107, respectively (19). Therefore, the combination of 4Ј-and 2Ј-␤-substitution in the absence of a 2Ј-␣-substituent resulted in two compounds, which show no apparent cross-resistance with 4Ј-azidocytidine and 2Ј-␤-methylcytidine, respectively.

RO-9187 Is Efficiently Phosphorylated in Primary Human
Hepatocytes-The kinetics of RO-9187-triphosphate (RO-9187-TP) formation in human hepatocytes was determined using tritium-labeled RO-9187. As shown in Fig. 3, the formation of RO-9187-TP increased in a time-and dose-dependent manner. The maximal triphosphate concentration at 24 h was 87 pmol/10 6 cells with half-maximal triphosphate formation achieved at 12 M RO-9187. 15.6 pmol/10 6 cells of RO-9187-TP were formed at 24 h from incubation of primary hepatocytes with 2 M RO-9187. In the absence of extracellular RO-9187, intracellular RO-9187-TP disappeared in a biphasic decay pattern with an effective half-life (50% reduction of initial TP concentration) of 5.4 h and a terminal half-life of 28 h (Fig. 3C). At 24 h after removal of extracellular RO-9187, the intracellular RO-9187-TP concentrations were still well above 4 pmol/10 6 (1.3 M) cells, and thus above the antiviral IC 50 , assuming a 3-l volume of normal human liver parenchymal cells (20,21), reflective of very robust phosphorylation efficiency of RO-9187 in primary human hepatocytes.

RO-9187 and NM107 Crystal Structures Show Different Ribose Conformations-
The high potency of inhibition of the HCV RNA-dependent RNA polymerase by 2Ј-deoxynucleoside analogs RO-9187 and RO-0622 was unexpected. The impact of the 4Ј-azido moiety on both antiviral potency and increased selectivity and reduced toxicity could be related to HCV-specific interaction of the 4Ј-substituent. However, a conformational effect on the ribose configuration was also a possibility. Crystal structures of RO-9187 and NM107 were solved to compare the structures of two potent HCV replication inhibitors with 2Ј-␣-deoxy (RO-9187), 2Ј-␣-hydroxy (NM107), and 2Ј-␤substituted (RO-9187, NM107) configurations, respectively. Interestingly, RO-9187 crystallized in the 2Ј-endo conformation, similar to other 2Ј-deoxyribonucleosides and other aranucleosides and typical of B-form DNA, whereas NM107 crystallized in the 3Ј-endo conformation, typical of A-form RNA (Fig. 4). The addition of the 4Ј-azido substitution to ara-C to

TABLE 4 Inhibition of HCV replication, cytotoxicity, and inhibition of cell proliferation of nucleoside analogs in the HCV replicon system
Mean values and standard deviations were determined from at least three experiments. form RO-9187 therefore did not affect the 2Ј-endo ribose conformation preference of ara-C typical of B-form DNA (22,23).    anatomic pathology in rats dosed with RO-9187, and the NOEL was 2000 mg/kg/day. In contrast, clinical observations in ribavirin-treated animals included pale appearance, labored respiration, and reduced body weight gain. Hematology findings in ribavirin-treated animals included granulocytopenia, erythrocytopenia, reduced mean hemoglobin, and an increase in alanine transferase, and microscopic changes were noted in multiple organs, including bone marrow, liver, and lymphoid organs. The plasma exposures of ribavirin increased ϳ2-fold between day 0 and day 13 of the study period, were similar to those achieved with RO-9187 at 200 mg/kg, and significantly lower than those obtained with RO-9187 at 2000 mg/kg. Therefore, RO-9187 showed an improved safety profile in rats as compared with ribavirin, when administered orally for 2 weeks.

DISCUSSION
The development of novel antiviral agents targeting HCV replication is paramount to achieve improved response rates to therapy among HCV-infected persons, especially those infected with difficult to treat genotype 1 virus. Nucleoside analogs have provided key medicines to treat viral diseases, including treatment of infections by herpesviruses, human immunodeficiency virus, and hepatitis B virus. The majority of marketed antiviral nucleoside analogs target viral DNA polymerases. In contrast, no specific nucleoside inhibitors of RNA polymerases have been developed to date, although novel ribonucleoside analogs targeting the RNA-dependent RNA polymerase of HCV have recently achieved promising results in clinical studies, encouraging further exploration within this chemical class. NM283, the 3Ј-valinate prodrug of NM107 (2Ј-C-methylcytidine), has demonstrated clinically significant antiviral potency in HCV-infected patients (24), and R1626, the triisobutyrate prodrug of R1479 (4Ј-azidocytidine), has shown robust antiviral potency in HCV-infected patients with dosedependent mean viral load reductions up to 3.7 log after 14 days of dosing in monotherapy (8), the largest antiviral effect seen to date with a polymerase inhibitor in monotherapy (25). Options to further increase the antiviral potency of nucleoside analogs as inhibitors of HCV replication include optimization of nucleotide analog incorporation efficiency by HCV NS5B, or optimization of phosphorylation efficiency to increase the intracellular concentration of the nucleoside triphosphate analog. In particular, the first step in the pathway to the triphosphate, the formation of nucleoside monophosphate, appears to be a limiting step for a majority of nucleoside analogs (9,26,27).
HCV is a positive strand RNA virus encoding an RNA-dependent RNA polymerase essential for the replication of viral RNA in the cytoplasm of infected cells. RNA polymerases are generally highly selective for incorporation of ribonucleotides into nascent RNA molecules and are believed to discriminate against 2Ј-deoxynucleotide triphosphates substrates by direct interaction with the 2Ј-␣-hydroxy moiety of ribonucleotides (10,11). Therefore, efforts into the design of nucleoside inhibitors against RNA polymerase targets have generally focused on satisfying this interaction by providing either ␣-hydroxy or ␣-fluoro substituents. The role of hydrogen bonding potential of fluorine on ribose molecules has been questioned, however, and a conformational impact of 2Ј-␣-substituents may also be of biological relevance. In particular, 2Ј-␣-substitution is conducive to the formation of the RNA-specific 2Ј-endo conformation, whereas 2Ј-␣-deoxyribonucleosides form preferentially the 3Ј-endo conformation typical of B-type DNA (22, 23, 28 -30). Previous studies with recombinant HCV polymerase (NS5B) suggested high selectivity of HCV NS5B for RNA templates and ribonucleotide triphosphate substrates, although oligodeoxyribonucleotides could be used as primers with homopolymeric RNA templates (31,32). The crystal structure of an NS5B-UTP complex suggested that the conserved amino acid Asp-225 in the NS5B active site forms a hydrogen bond with the 2Ј-␣-hydroxyl group of UTP and could therefore function as the key residue conferring NTP substrate selectivity over dNTPs by HCV NS5B (13).
Consistent with the high ribonucleotide substrate binding selectivity of HCV NS5B, we detected only low level incorporation of 2Ј-dCTP into nascent RNA in a gel-based single nucleotide incorporation assay and no inhibition in an in vitro RNA synthesis assay using recombinant NS5B and HCV cIRES RNA template. These results demonstrate that CTP competitive binding of 2Ј-dCTP to HCV NS5B was negligible under the assay conditions, consistent with high level active site binding discrimination against 2Ј-deoxy-CTP.
Interestingly, moderate inhibition of RNA synthesis by NS5B was observed with ara-CTP, suggesting that new interactions of ␤-substituents in the HCV polymerase active site may partially compensate for the loss of 2Ј-␣-hydroxy hydrogen bonding. Similarly, 2Ј-␤-hydroxy-GTP was described as a moderate inhibitor of HCV NS5B, but 2Ј-␤-hydroxy-ATP was inactive (33). The increased base pairing stability of G:C base pairs may therefore be required to allow sufficient binding affinity of these compounds in the NS5B active site to confer measurable inhibitory potency. Based on the discovery that 4Ј-substitutions on ribonucleosides could provide potent and selective inhibitors of HCV RNA synthesis (6,7) and assuming that additional interactions with the 4Ј-substituent could further compensate for the loss of ␣-hydroxy interaction, the effect of such substitutions on deoxyribonucleosides was assessed. The results were unexpected in their magnitude.
The synthesis of 2Ј-␣-deoxy-2Ј-␤-hydroxy-4Ј-azido-CTP (4Ј-azido-ara-CTP, RO-9187-TP) provided a novel inhibitor of HCV NS5B with inhibitory potency in the cIRES RNA-dependent RNA polymerase assay similar or slightly improved to that of 4Ј-azido-CTP (R1479-TP). RO-9187-TP was incorporated into nascent RNA by NS5B and prevented further RNA elongation as an effective chain terminator in the HCV transcription initiation assay. Interestingly, the incorporation efficiency measured in a gel-based RNA synthesis initiation assay was 7.8-fold higher than that of ara-CTP but 4-fold lower than that of R1479-TP. Effects on both K m and V max values contributed to this difference in incorporation efficiency. The inhibitory potency of RO-9187-TP was similar to or slightly higher in the cIRES-based RNA synthesis assay as compared with R1479-TP, despite the lower incorporation efficiency during RNA synthesis initiation. It is possible that RO-9187-TP and R1479-TP incorporation efficiencies are different during initiation and elongation phases of RNA synthesis by NS5B and that differences in incorporation efficiency during RNA elongation may therefore contribute to the overall inhibitory potency of both compounds in the cIRES-RNA-dependent NS5B RNA synthesis assay. Studies to quantify incorporation during elongation are currently in progress and will help to clarify this potential difference between these two nucleoside analogs.
The ␤-hydroxy group of RO-9187-TP could be replaced by a ␤-fluoro moiety with only minor effects of reduced inhibitory potency in the RNA synthesis assay. RO-0622-TP was also efficiently incorporated into nascent RNA by NS5B and was an effective chain terminator at transcription initiation. Fluoro substitution may provide hydrogen bond acceptor properties as a replacement of hydroxyl moieties, but the resulting hydrogen bond strength is expected to be substantially lower as compared with a hydroxyl group (34). However, it has been shown that under certain circumstances the 2Ј-hydroxy group may be replaced by a fluoro moiety without major reduction of inhibitory potency against certain RNA polymerases. 2Ј-␣-Fluoro-NTPs showed antiviral potencies against influenza virus and were substrates for the RNA-dependent RNA polymerase of influenza virus (35,36). 2Ј-␣-Fluorocytidine was also assessed as an inhibitor of HCV replication but was a potent inhibitor of cell proliferation in the HCV replicon system (7,37). The addition of a 2Ј-␤-methyl substituent resulted in the discovery of PSI-6130 (2Ј-␣-fluoro-2Ј-␤-methylcytidine), a potent inhibitor of HCV replication (38). Interestingly, PSI-6130 showed high selectivity against the closely related positive strand RNA virus bovine viral diarrhea virus, whereas the 2Ј-␣-hydroxy analog 2Ј-C-methylcytidine (NM107) inhibited both HCV and bovine viral diarrhea virus replication in cell culture (38). Fluoro-nucleosides may therefore provide considerable hydrogen bond acceptor functionality in the RNA polymerase active site (34), although such substitutions may also function by facilitating the formation of an RNA-like 3Ј-endo conformation in the RNA polymerase active site (39) or by accelerating the chemical reaction through electronic stabilization of the nucleophilic substitution reaction in the polymerase active site (40). Consistent with the conformational effect of the 2Ј-␣-fluoro substituent, PSI-6130 crystallized in the RNA typical 3Ј-endo conformation (38) similar to NM107 (Fig. 4) and other ribonucleosides and 2Ј-␣-fluoronucleosides (22,23). In contrast, the 2Ј-␣-deoxyribonucleoside RO-9187 crystallized in the DNA-like 2Ј-endo-conformation, similar to other 2Ј-deoxyribonucleosides.
Unexpectedly, the antiviral potency of RO-0622 increased more than 50-fold as compared with R1479 in HCV replicon cells, whereas RO-9187 was 7.5-fold more potent than R1479, even though all three compound triphosphates showed similar inhibitory potencies in the NS5B RNA synthesis assay in vitro, and even though R1479-TP was the most efficient substrate for incorporation by NS5B during transcription initiation. This increase in potency was most likely related to a substantially increased phosphorylation efficiency of the deoxynucleoside analogs in human cells. All three compounds (RO-0622, RO-9187, and R1479) were substrates of human dCK, but RO-0622 was phosphorylated with an unprecedented efficiency more than 13-fold higher than that of R1479 and 3-fold higher than that of 2Ј-deoxycytidine. In addition, RO-0622 was also a substrate of human thymidine kinase 2. RO-9187 was phosphorylated by human dCK in vitro with an efficiency similar to R1479. However, phosphorylation of RO-9187 in human hepatocytes was higher than that of previously studied ribonucleosides, achieving a mean triphosphate concentration of 15.6 pmol/million cells from 2 M RO-9187 within 24 h, which is 3.3-20-fold higher than that reported for NM107 and 6.5-fold higher than that achieved from PSI-6130 under similar conditions (20,27). It is therefore possible that additional enzymes are involved in the phosphorylation of RO-9187 or that phosphorylation of the intermediates RO-9187-MP or RO-9187-DP by the respective nucleoside monophosphate and diphosphate kinases could be more efficient than that of ribonucleoside-MP and -DP analogs.
The novel nucleoside analog RO-9187 was derived from ara-C (cytarabine), a cytotoxic nucleoside analog, which forms the backbone of induction therapy for the treatment of acute myelogenous lymphoma. Ara-C is a potent inhibitor of cell proliferation in cell culture and also inhibited proliferation of Huh-7 hepatoma cell-derived 2209-23 HCV replicon cells with an IC 50 of 29 nM, measured as inhibition of [ 3 H]thymidine incorporation into cellular DNA. HCV replication in this cell line is dependent on cell proliferation, such that the antiviral effect of cytostatic agents cannot be assessed (41). The introduction of a 4Ј-azido group dramatically increased the selectivity of RO-9187 as compared with ara-C, with no inhibitory effect on [ 3 H]thymidine incorporation measurable in HCV replicon cells at concentrations up to 1 mM. This correlates to an increase in antiviral selectivity over cell proliferation inhibition of more than 34,000-fold achieved through nucleoside 4Ј-substitution. Similarly, RO-8013 (2Ј-␣-deoxy-2Ј-␤-fluorocytidine) was a potent inhibitor of cell proliferation with an IC 50 of 174 nM. The introduction of a 4Ј-azido group to generate RO-0622 also substantially increased the antiviral selectivity. These results are consistent with previous findings that R1479 (4Јazidocytidine) showed high selectivity for the inhibition of HCV replication and R1479-TP did not inhibit the related RNAdependent RNA polymerase of influenza virus.
RO-9187 and RO-0622 inhibited HCV replicon variants resistant to inhibition by R1479 (S96T) and 2Ј-C-methyl nucleosides (S282T) with similar potency as wild-type replicons. The absence of a 2Ј-␣-substituent may therefore allow these compounds to bind to NS5B in a conformation not affected by the increased bulk of S282T in the NS5B active site. The mechanism of resistance to R1479 conferred by the S96T mutation in NS5B has not been determined yet, but the availability of novel 4Ј-substituted analogs with high inhibitory potency against S96T replicons may help to resolve the molecular mechanism of the S96T effect on R1479.
The safety profile of RO-9187 was assessed in a 2-week toxicity study in Hanover-Wistar rats, in comparison with ribavirin, a nonselective nucleoside analog, currently forming part of standard of care therapy for HCV-infected patients. There were no treatment related toxicity findings in rats treated with RO-9187 at dose levels up to 2000 mg/kg/day for 14 days, consistent with the apparent high selectivity ratio observed for RO-9187 in cell culture. Plasma concentrations in rats treated with RO-9187 reached levels in excess of 16 g/ml (Ͼ50 M), i.e. ϳ300-fold above the antiviral potency in HCV replicon cells. In contrast, treatment of rats with 200 mg/kg ribavirin for 14 days was associated with significant toxic changes in Hanover-Wistar rats affecting multiple organ systems, similar to pre-viously described findings (42). Plasma concentrations of ribavirin reached mean C max values of 1.7-3.2 g/ml over the 14-day dosing period. The time-dependent increase in plasma concentration was likely related to accumulation of ribavirin-TP in red blood cells, leading to an increased half-life of ribavirin in rat plasma (43).
These studies establish 4Ј-substituted, 2Ј-deoxynucleoside analogs as a new class of antiviral agents with high potency and selectivity for the inhibition of HCV replication, differentiated resistance profiles, exceptionally high phosphorylation efficiency in human target cells, and potential for achieving good safety and tolerability profiles. Further assessment of the preclinical safety profile is warranted for individual compounds from this series, which has the potential to provide novel improved medicines for the treatment of HCV infection.