Identification and biological characterization of heterocyclic inhibitors of the hepatitis C virus RNA-dependent RNA polymerase.

The hepatitis C virus (HCV) NS5B protein encodes an RNA-dependent RNA polymerase (RdRp), the primary catalytic enzyme of the HCV replicase complex. We established a biochemical RNA synthesis assay, using purified recombinant NS5B lacking the C-terminal 21 amino acid residues, to identify potential polymerase inhibitors from a high throughput screen of the GlaxoSmithKline proprietary compound collection. The benzo-1,2,4-thiadiazine compound 1 was found to be a potent, highly specific inhibitor of NS5B. This agent interacts directly with the viral polymerase and inhibits RNA synthesis in a manner noncompetitive with respect to GTP. Furthermore, in the absence of an in vitro-reconstituted HCV replicase assay employing viral and host proteins, the ability of compound 1 to inhibit NS5B-directed viral RNA replication was determined using the Huh7 cell-based HCV replicon system. Compound 1 reduced viral RNA in replicon cells with an IC(50) of approximately 0.5 microm, suggesting that the inhibitor was able to access the perinuclear membrane and inhibit the polymerase activity in the context of a replicase complex. Preliminary structure-activity studies on compound 1 led to the identification of a modified inhibitor, compound 4, showing an improvement in both biochemical and cell-based potency. Lastly, data are presented suggesting that these compounds interfere with the formation of negative and positive strand progeny RNA by a similar mode of action. Investigations are ongoing to assess the potential utility of such agents in the treatment of chronic HCV disease.

The hepatitis C virus (HCV) NS5B protein encodes an RNA-dependent RNA polymerase (RdRp), the primary catalytic enzyme of the HCV replicase complex. We established a biochemical RNA synthesis assay, using purified recombinant NS5B lacking the C-terminal 21 amino acid residues, to identify potential polymerase inhibitors from a high throughput screen of the Glaxo-SmithKline proprietary compound collection. The benzo-1,2,4-thiadiazine compound 1 was found to be a potent, highly specific inhibitor of NS5B. This agent interacts directly with the viral polymerase and inhibits RNA synthesis in a manner noncompetitive with respect to GTP. Furthermore, in the absence of an in vitro-reconstituted HCV replicase assay employing viral and host proteins, the ability of compound 1 to inhibit NS5Bdirected viral RNA replication was determined using the Huh7 cell-based HCV replicon system. Compound 1 reduced viral RNA in replicon cells with an IC 50 of ϳ0.5 M, suggesting that the inhibitor was able to access the perinuclear membrane and inhibit the polymerase activity in the context of a replicase complex. Preliminary structure-activity studies on compound 1 led to the identification of a modified inhibitor, compound 4, showing an improvement in both biochemical and cell-based potency. Lastly, data are presented suggesting that these compounds interfere with the formation of negative and positive strand progeny RNA by a similar mode of action. Investigations are ongoing to assess the potential utility of such agents in the treatment of chronic HCV disease.
Hepatitis C virus (HCV), 1 a positive strand RNA virus of the Flaviviridae family, is the major etiological agent of post-trans-fusion and sporadic non-A, non-B hepatitis (1). An estimated 2-3% of the world population is chronically infected with HCV, which causes significant liver disease, cirrhosis, and can eventually lead to the development of hepatocellular carcinoma. In infected cells, translation of the viral RNA yields a 3011-residue polyprotein chain (2)(3)(4), which is subsequently cleaved to generate envelope and core proteins, for assembly of new virus particles and nonstructural enzymes essential for viral replication (5)(6)(7). Studies using recombinant NS5B polymerase have provided direct evidence for RNA-dependent RNA polymerase activity (8,9), and this catalytic activity has been confirmed to be required for infectivity in chimpanzees (10).
NS5B polymerase contains a hydrophobic C-terminal domain thought to be responsible for anchoring the protein to mammalian cell membranes. Removal of the C-terminal 21 residues has been reported to facilitate protein isolation from Escherichia coli without compromising RdRp activity (11). The HCV RdRp initiates RNA synthesis preferentially from the 3Ј terminus of the template RNA (12,(13)(14)(15) but lacks specificity for HCV RNA in vitro, because it readily utilizes heterologous nonviral templates (8). Based on crystallographic studies of the enzyme containing C-terminal truncations (16,17), the hydrophobic tail present in the full-length enzyme could be predicted to partially occupy the palm domain, and hence may interfere with identification of active site inhibitors in biochemical assays. Although modeling experiments suggest that a templateprimer can be readily accommodated into the truncated polymerase without altering the global folding, it is unclear whether this model can be extrapolated to the full-length RdRp that exists within the functional replicase complex.
A detailed understanding of the mechanisms regulating HCV replication has been plagued by the lack of an efficient virus culture system. Recently, however, cell-based replicon systems for HCV were developed in which the nonstructural proteins stably replicate subgenomic viral RNA in Huh7 cells (18,19). In combination with studies using recombinant RdRps, such systems are proving to be invaluable in developing a better insight into the mechanisms of Flaviviridae RNA synthesis. This report describes the development of a robust RdRp assay using poly(rC) as template and oligo(rG) 13 as primer to identify novel, small molecule NS5B inhibitors from high throughput screening of compound libraries. We confirm herein that use of a C-terminal truncated NS5B represents a viable approach to identify highly selective HCV NS5B inhibitors that also interfere with viral RNA synthesis in the cellbased HCV replicon system.

EXPERIMENTAL PROCEDURES
Protein Purification-Expression of His-tagged C-terminal 21-amino acid deleted NS5B (J4, genotype 1b, HCV ⌬21) in E. coli was performed as described previously (20). For purification, E. coli cell lysate was applied to Talon metal affinity resin (CLONTECH), and the bound proteins were step-eluted with 30 and 200 mM imidazole. The 200 mM Talon eluate was dialyzed against buffer A containing 20 mM Tris-Cl, pH 8.0, 3 mM DTT, 150 mM NaCl, and 10% glycerol and applied to poly(U)-Sepharose 4B (Amersham Biosciences) column. After washing with 10 column volumes with buffer A, the bound protein was eluted with buffer A plus 1 M NaCl. NS5B containing fractions were pooled and concentrated to 30 mg/ml using a centricon10 concentrator (Amicon). Purity of NS5B as judged by SDS-PAGE was Ͼ95%.
Biochemical RdRp Assays-The high throughput RdRp assay was carried out in 384-well plates using 50 nM enzyme, 0.2 Ci of [␣-33 P]GTP, 0.6 M GTP, 250 nM 5Ј-biotinylated oligo(rG 13 )/poly(rC) in 20 mM Tris-Cl, pH 7.5, 5 mM MgCl 2 , 25 mM KCl, 3 mM DTT, and 0.05% bovine serum albumin. The 25-l reaction was terminated after 2 h at 25°C upon addition of equal volume of 100 mM EDTA and transferred to a streptavidin-coated FlashPlate. After incubation at 25°C for 30 min, and the plate was washed extensively and counted using a Packard TopCount microplate reader (n ϭ 4 for secondary screening). Additionally, viral RdRps (20,21) and DNA polymerase ␣ and ␤ (ChimeRx, Milwaukee, WI) were used to assess the selectivity of inhibition for the compounds according to a protocol similar to that described previously (20). Recombinant baculovirus expressing human DNA pol-␣ (kindly obtained from Dr. Teresa Wang, Stanford University School of Medicine, Stanford, CA) was used to infect Sf9 cells for purification. After 60 h, cell pellets were frozen at Ϫ80°C. Thawed pellets were resuspended in 50 mM Tris-Cl, pH 8.0, 100 mM NaCl, 1 mM DTT, and complete protease inhibitors. The suspension was briefly sonicated and cell debris was removed by centrifugation at 14K ϫ g in a microcentrifuge. The resultant supernatant was further cleared by centrifugation at 100K ϫ g for 60 min, captured on a 1.0-ml HiTrap heparin column (Amersham Biosciences), and eluted with a 1 M NaCl wash. The eluent was diluted to 200 mM NaCl with 50 mM Tris, pH 8.0, and captured on a 1.0-ml HiTrap N-hydroxy-succinimide (NHS) column coupled with 10 mg of SJK237-71 anti-pol-␣ monoclonal. The protein was eluted from the immunoaffinity column using 0.1 M glycine, pH 3.0, and neutralized using 1 M Tris, pH 9.0.
Gel-based RdRp assays used the negative strand RNA of the 5Ј-NTR and positive strand RNA of the 3Ј-NTR. Two primers (T7/BB7570: TAATACGACTCACTATAGGCCAGCCACGATAGCCGCGCTG and 5Ј-NTR: GCCAGCCCCCGATTGGGGGCGACACT) were used to amplify pBB7 DNA (C. Rice, Rockefeller University, NY) to amplify a DNA fragment containing a T7 promoter. In vitro transcription reactions (Ambion) were performed using T7 RNA polymerase to generate a 600-base RNA representing the 3Ј-end of the negative strand from the 5Ј-NTR. To generate 3Ј-NTR RNA, a 696-bp DNA fragment containing 3Ј-NTR and portions of NS5B was cloned into pGEM vector (pGEM92f 3Ј-NCR, kindly provided by P. Thommes, GlaxoSmithKline, Stevenage, UK). Plasmid DNA was linearized by ScaI for in vitro transcription to generate a 696-base RNA representing the positive strand of the 3Ј-NTR. After phenol chloroform extraction, the RNAs were precipitated and analyzed on a 6% PAGE gel containing 7 M urea. The RNA was excised and eluted into 10 mM Tris-Cl, pH 8.0, 1 M EDTA overnight at 37°C and further purified using an RNeasy column (Qiagen). 25-l biochemical RdRp reactions containing 20 mM Tris-Cl, pH 7.5, 25 mM KCl, 3 mM DTT, 2.5 mM MgCl 2 , 1.6 unit of RNasin, 0.6 M GTP, 0.4 mM each of ATP, CTP, and UTP, 0.25 l of [␣-32 P]GTP, 9 nM RNA template, and 10 nM NS5B. 3-fold serial dilutions of compounds were prepared in 50% Me 2 SO, and 1 l was added to the reactions. For controls, 1 l of 50% Me 2 SO was added. The reactions were incubated at 25°C for 2 h before terminating the reaction with 5 l of 100 mM EDTA. After phenol/chloroform extraction, free nucleotides were removed by applying the reaction mixtures to a G-50 spin column. Approximately onesixth of the RNA products were separated by electrophoresis on 8% Tris-Borate EDTA gel. The labeled RNA products were quantified by phosphorimaging.
Cytotoxicity Assays-XTT cytotoxicity assays were performed in Huh7 cells. Briefly, 8000 cells/well were cultured overnight in 96-well plates at 37°C and 5% CO 2 with Dulbecco's modified essential medium containing 10% fetal calf serum and 1% nonessential amino acids. Test compounds (10 mM stocks) were diluted serially 1:2 in Me 2 SO, and 100 l of working dilution was added to a final concentration of 50 M compound, 0.5% Me 2 SO. The plates were incubated an additional 40 h, followed by addition of 50 l of XTT/phenazine methosulfate (Sigma) solution to a final concentration of 400 ng of XTT and 3 ng of phenazine methosulfate per well. Plates were further incubated up to 4 h until the A 460 value was between 1.5 and 1.8 to allow IC 50 calculation.
Positive Strand Replicon RNA Detection-Replicon cells were plated at 3 ϫ 10 3 cells per well in a 96-well plate plates at 37°C and 5% CO 2 in Dulbecco's modified essential medium containing 10% fetal calf serum, 1% nonessential amino acids and 1 mg/ml Geneticin. After allowing 4 h for cell attachment, 1 l of compound dilution was added to the medium (n ϭ 8 wells per dilution). Briefly, 11 2.5-fold dilutions of 1 mM stock test compound in Me 2 SO were prepared with final concentration ranging from 10,000 to 1.0 nM. Plates were incubated for 40 h, until they reached 80% confluence. After removal of medium, 150 l of guanidine salt-based RLT buffer (Qiagen) was added to each well and RNA was purified according to the manufacturer's recommendations (Qiagen RNeasy) and were eluted twice in 45 l of dH 2 O prior to RT-PCR. Approximately 40 l of TaqMan EZ RT-PCR (Applied Biosystems) master mix (1ϫ TaqMan EZ buffer, 3 mM Mn(OAc) 2 , 0.3 mM dATP, 0.3 mM dCTP, 0.3 mM dGTP, 0.6 mM dUTP, 0.2 mM neo-forward, 0.2 mM neo-reverse, 0.1 mM neo-probe, 1ϫ cyclophilin mix, 0.1 unit/l of rTth DNA polymerase, 0.01 unit/l AmpErase UNG, and H 2 O to 40 l) was added to each tube of 96-tube optical plate along with 10 l of RNA elution. Primers and probes specific for the positive-strand RNA detection of neomycin gene were: neo-forward, 5Ј-CCGGCTACCTGC-CCATTC-3Ј; neo-reverse, 5Ј-CCAGATCATCCTGATCGACAAG-3Ј; neoprobe, 5ЈFAM-ACATCGCATCGAGCGAGCACGTAC-TAMRA3Ј. Additionally, the PDAR control reagent human cyclophilin was used for normalization. Samples were mixed briefly and placed in an ABI7700 (Applied Biosystems) at 50°C, 2 min; 60°C, 30 min; and 95°C, 5 min; with cycling parameters set to 94°C, 20 s; 55°C, 1 min for 40 cycles. The reaction is monitored in real-time over 40 cycles to generate a raw amplification plot on a logarithmic scale. The plot is analyzed during the exponential phase, at ϳ25 cycles, to determine the quantities of cDNA. The relative cDNA levels for neo and cyclophilin were determined compared with controls treated only with Me 2 SO, and the ratio of neo:cyclophilin was used for IC 50 calculation (n ϭ 8). The exponential phase provides very consistent data and a theoretical sensitivity range of 7 logs.
Negative Strand Replicon RNA Detection-To achieve strand-specific detection, a primer containing HCV RNA sequences and an 18-base tag of nonrelated sequence at the 5Ј-end was for the reverse transcription (RT) reaction, 5Ј-ACATGCGCGGCATCTAGACCGGCTACCTGCCCAT-TC-3Ј. A Thermoscript-RT-PCR system (Invitrogen) was used for the RT reaction according to the manufacturer's protocol, with ϳ9 l of the cell-harvested RNA and 1 l of primer (10 M) incubated with RT at 60°C for 1 h. Following that incubation, 2 l of cDNA product containing the 5Ј tag was amplified for TaqMan quantification using the 48 l of TaqMan Universal Master Mix (Applied Biosystems) as well as primers: neo-tag, 5Ј-ACATGCGCGGCATCTAGA-3Ј; neo reverse, 5Ј-C-CAGATCATCCTGATCGACAAG-3Ј; and neo probe, 5ЈFAM-ACATCGC-ATCGAGCGAGCACGTAC-TAMRA3Ј. Samples were mixed briefly and placed in an ABI7700 (Applied Biosystems) at 50°C, 2 min; 95°C, 10 min, with cycling parameters set to 94°C, 15 s; 55°C, 1 min for 40 cycles. The negative strand copy number in each reaction was determined using linear regression analysis based on the slope and intercept generated with a negative strand copy standard curve. The negative strand copies per cell were determined by dividing the total negative strand copies per reaction by the total cells per reaction.
Circular Dichroism (CD) Spectroscopy-CD analysis was performed on purified HCV ⌬21 NS5B (0.3 mg/ml in 25 mM Tris, pH 7.5, 25 mM KCl, 5 mM MgCl 2 , and 0.8% Me 2 SO) at 20°C in the presence (40 M) and absence of the inhibitors. Far-UV CD spectra were recorded on a Jasco J-710 CD instrument using a water-jacketed cell of 0.1-cm path length. Data (10 accumulations) were collected using a time constant of 2 s with a 1-nm constant spectral bandwidth at 50 nm/min (CD wavelength scans not shown). Thermal stability of HCV ⌬21 NS5B was measured by scanning temperature at 1°C/min while monitoring the CD signal at 220 nm. Thermal stability data were then analyzed using GraFit and a fitting equation adapted from Ref. 22 to predict the midpoint transition temperature (T m ). Competition with GTP-NS5B assays for secondary confirmation and mechanistic studies were performed similar to the high throughput assay described above, with the following modifications. Briefly, 5 l of 5ϫ concentrated inhibitor solutions or water control was added followed , with doubling dilutions made in reaction buffer, and started with 5 l of enzyme mix (250 nM NS5B in reaction buffer). Assay plates were incubated at room temperature for 30 min (reactions were linear over this period at low GTP concentrations) and halted with 35 l of stop buffer (phosphate-buffered saline containing 0.1 M EDTA). Next, 50 l of terminated reactions was transferred into Streptavidin FlashPlate wells (PerkinElmer Life Sciences), and binding was allowed to proceed at room temperature for 45 min with plate shaking. Well contents were removed by pipette and wells were washed 3ϫ with 200 l of wash buffer (5 min per wash in phosphate-buffered saline containing 0.5% Tween 20). Incorporated radioactive GTP was determined using a Packard TopCount microplate reader.
The benzo-1,2,4-thiadiazine 1 demonstrated potent inhibition of the HCV ⌬21 RdRp with a biochemical IC 50 of around 0.1 M (Table I). Moreover, preliminary structure activity relationships indicated a critical role for the quinolone N-1 substituent for biological activity. Whereas both the N-unsubstituted 2 and N-methyl 3 analogs were significantly weaker NS5B inhibitors, the N-butyl and N-isopentyl derivatives (1 and 4) were potent inhibitors (Table I). Compound 4 showed an IC 50 of around 0.08 M, whereas little inhibition was observed with compounds 2 and 3 up to 50 M (13 and 35% inhibition at 50 M, respectively). Compounds 1 and 4 retained activity against the full-length form of the enzyme, with IC 50 values ϳ2-fold higher (data not shown).
Biophysical Characterization of Benzo-1,2,4-thiadiazines-Mechanistic studies regarding enzyme selectivity and mode of action were performed to characterize the differential inhibition profile of derivatives 1-4 for the HCV RdRp.
Selectivity-Similar IC 50 values for compound 1 and 4 were obtained when the RNA substrate for the viral RdRp was changed to oligo(U)-primed homopolymeric poly(rA), or unprimed heteropolymeric RNA (data not shown), indicating the absence of substrate-specific inhibition. Compounds 1-4 were tested for selectivity against closely related viral RdRps (BVDV and GBV-B) and mammalian polymerases (DNA pol ␣ and DNA pol 〉), and all IC 50 values were Ͼ50 M with a selectivity index above 500.
Polymerase Interaction-CD spectroscopy was used to monitor thermal denaturation of NS5B in the presence of either Me 2 SO or 40 M inhibitor. Notably, the polymerase-compound 1 complex melted over a broad temperature range (T m ) resulting in a 3°C increase (Table II), whereas compound 3 resulted in a 0.7°C increase. Slow kinetics for compound 1-polymerase complex unfolding under the experimental assay conditions could explain this delayed transition. The polymerase-compound 4 complex resulted in a similar shift in T m (⌬T m ϭ 3.4°C), although with a lower magnitude than compound 1 (Fig. 1). Additionally, the affinity of these derivatives for the viral RdRp remained unchanged in the presence of poly(rC) 25-mer or a 3Ј-ribodeoxy GTP chain terminator (data not Stability was evaluated with a Jasco J-710 spectropolarimeter by monitoring the ellipticity at 220 nm. As temperature is increased and protein denatures, a loss in ellipticity is observed. Representative data compares the midpoint melting transition temperature (T m ) of NS5B only (circles, 46.4°C), to NS5B with 40 M derivative 4 (squares, 49.8°C). The increased shift in T m and change in the unfolding enthalpy is directly consistent with binding to NS5B, because it has been previously shown that thermal stabilities of proteins, in the presence of their ligands, are enhanced as the affinity constant increases (22).  shown). Studies using analytical ultracentrifugation and isothermal titration microcalorimetry further confirmed these observations (data not shown). Nucleic Acid Interaction-Aminoquinoline agents have been shown previously to interact with nucleic acid (24), and a similar compound known to bind to both single-strand and double-strand nucleic acid 2 was used as a positive control. In the PicoGreen displacement assay, this aminoquinoline had a C 50 value of 0.03 M (Fig. 2). The active inhibitors 1 and 4 had C 50 values for double-strand RNA of about 50 and 0.5 M, respectively, indicating that the reduction in polymerase activity most likely did not occur by titration or binding of the RNA substrate. The selectivity index comparing nucleic acid interaction with biochemical potency was 500 for inhibitor 1, and 6 for inhibitor 4.
Solubility limitations precluded testing of inhibitor 4 at higher concentrations in this RNA displacement assay. Compounds 2 and 3 had C 50 values Ͼ 50 M (data not shown). Consistent with these observations, similar results were obtained in the presence and absence of 1 mM magnesium ions or when using alternative nucleic acid substrates such as poly(A)-poly(U) RNA or with double-stranded DNA (data not shown).
Noncompetitive with GTP-Consistent with a previous report (25), the K m for GTP using the ⌬21 NS5B preparation on a poly(rC):oligo(rG) RNA substrate was shown to be 0.73 M (data not shown). Furthermore, mechanistic enzymology studies with compound 1 suggest that this inhibitor exhibits a kinetic behavior consistent with a reversible, noncompetitive mechanism of inhibition with respect to GTP. Consistent with this observation, compound 1 was unable to block the binding of GTP to the viral polymerase. In this study, the K m for GTP remained unchanged while the V max decreased with increasing concentration of compound 1 (Fig. 3A). However, as expected, the K m for GTP significantly increased upon titration of 3Јribodeoxy GTP while the V max remained unchanged (Fig. 3B). Fluorescence excitation and emission wavelengths were 480 and 520 nm, respectively. Three compounds tested, control (circle), derivative 1 (square), and derivative 4 (triangle) with increasing concentrations as shown on the x-axis. A decrease in fluorescence (control compound) indicates binding to RNA. Maximum fluorescence in the absence of control or derivative was taken to be 100%, and minimum fluorescence in the absence of RNA was taken to be 0%. The C 50 value is defined as the concentration of control or derivative that produces a reduction of 50% displacement of bound probe or a 50% net reduction of fluorescence signal. Therefore, inhibition by 3Ј-ribodeoxy GTP was competitive with respect to GTP, and the Lineweaver-Burk plots indicated that compound 1 most likely interacts with the viral polymerase at a site distinct from the GTP binding site. Consistent with these data, compound 4 was also shown to be noncompetitive with GTP (data not shown).
For viral reduction assays, TaqMan was utilized to monitor both cellular and viral RNA. Cyclophilin RNA levels were normalized to positive-strand HCV viral RNA to allow an accurate IC 50 determination upon compound titration, and no changes in the level of cellular RNA were evident upon addition of up to 10 M of compound 1, further confirming the lack of cytotoxicity in these cells. Cell-based inhibition of viral replication was confirmed with compound 1 in the HCV replicon system, with an IC 50 of 0.55 M (Table III, n ϭ 8) and a therapeutic index relative to cytotoxicity of ϳ100. Expectedly, benzo-1,2,4-thiadiazines 2 and 3 did not exhibit the ability to inhibit viral replication in the replicon system, when tested at concentrations at or below 20 M (data not shown). Consistent with the similarity in biochemical potency between compounds 1 and 4, compound 4 showed activity in the replicon system with an IC 50 of 0.52 M for reduction in positive strand viral RNA. Percent reduction in viral RNA was 80% for compound 1 and 91% for compound 4 at 10 M.
Impact of Benzo-1,2,4-thiadiazines on Replication Intermediates-Positive strand viral RNA represents the nucleic acid strand that is translated and initially copied to generate the negative strand replicative intermediate. Negative strand RNA is the template used to generate the positive strand message, which is generally packaged into productive virions. Although the replicon system does not generate infectious particles, because the coding elements for the structural polypeptides have been removed from the replicon construct, monitoring a reduction in positive strand RNA represents a facile method for quantifying activity of polymerase inhibitors. However, the formation of positive and negative strand RNAs would not necessarily require identical replicase conformations, mechanistic requirements, or similar sets of cofactors. To perform mode of action studies and potentially differentiate inhibitors, it is useful to confirm whether antiviral agents inhibit both positive and negative strand RNA synthesis. To that end, the activity profile for compounds 1 and 4 was assessed in a biochemical gel-based RdRp assay using the native termini of the viral genome (3Ј-NTR RNA and the negative strand of 5Ј-NTR RNA) as well as in the cell-based replicon by monitoring the reduction in negative strand replicon RNA.
The inhibition profile for compounds 1 and 4 in an RdRp assay using the native HCV termini as RNA substrate, showed a 3-fold increase in potency for compound 4 relative to compound 1 (Fig. 4), compared with less than a one-fold difference in IC 50 between these compounds when using homopolymeric RNA. Furthermore, since the IC 50 values for a single compound were similar for both positive or negative strand RNAs (5Ј RNA IC 50 ϭ 0.17 M for 1 and 0.06 M for 4; 3Ј RNA IC 50 ϭ 0.14 M for 1 and 0.04 M for 4), this suggested that the mechanism of inhibition for positive strand RNA synthesis is likely to be similar to that for inhibition of negative strand RNA formation for these compounds. The inhibition profile from compounds 2 and 3 in this gel-based assay was similar to that reported for the primer-extension assay detailed in Table I (e.g. less than 30% inhibition at 50 M). Interestingly, the ⌬21 HCV was less efficient at generating the full-size 600-base product from the 5Ј-NTR negative strand RNA, as shown by the presence of smaller sized termination products (ϳ400 -500 bases), compared with products from the 3Ј-NTR substrate (Fig. 4).
To evaluate whether this shared similarity in strand-specific inhibition occurs in the replicon cells, this system was first validated as a surrogate model for viral infection by confirming the presence of a disparate ratio of positive to negative strand RNA. Levels of positive strand RNA have been reported to be at least 10-fold higher than negative strand in infected liver tissues (26). In the replicon cells, TaqMan analysis showed that ϳ2200 positive strand replicon RNA copies per cell were present, whereas 200 copies per cell of negative strand were detected (data not shown). Using this assay, compounds 1 and 4 showed a similar inhibition profile for reductions in either   (Table III), with IC 50 values ranging from 0.4 to 0.55 M. These data are consistent with the suggestion that these inhibitors exert an equivalent effect on replication of both positive and negative strand RNA synthesis.

DISCUSSION
The HCV RNA-dependent RNA polymerase, a central catalytic enzyme of replication, represents a viable target for identification of antiviral agents to treat chronic HCV infections. In this study, we report the development of a practical RdRp assay suitable for high throughput screening. We have demonstrated herein that screening of panels of drug-like molecules with soluble recombinant HCV NS5B in a ribonucleotide incorporation assay resulted in the identification of inhibitors of viral RNA synthesis. Furthermore, the benzo-1,2,4-thiadiazine derivatives were shown to be highly selective for the HCV RNA polymerase, failing to inhibit other viral and mammalian polymerases. Significantly, these compounds retained absolute selectivity for HCV, because they were completely inactive at inhibiting GBV-B virus NS5B RNA synthesis even though the two enzymes share 37% identity and 52% similarity.
Mechanistic studies confirmed that the reduction in viral RNA synthesis by the benzo-1,2,4-thiadiazines was most likely due to direct catalytic inhibition, rather than titration of nucleic acid or competition with nucleotides. Consistent with this observation, compound 1 demonstrated no interaction with nucleic acid at concentrations up to 50 M. The exact mechanism of inhibition remains unclear. Because compound 4 shares similar structure as well as biochemical and cell-based activity with compound 1, the apparent enhanced interaction with RNA for compound 4 is unclear. The solubility limitations of compound 4 may indirectly affect interpretation of the RNA binding data.
Despite the use of a C-terminally truncated polymerase for high throughput screening, the benzo-1,2,4-thiadiazines were capable of inhibiting polymerization of the full-length NS5B polymerase in biochemical assays with a relatively similar potency profile. Most importantly, two of the analogs in the series were able to inhibit the full-length enzyme in the context of the replicase complex expressed in HCV replicon cells. This replication complex most likely consists of cellular and nonstructural viral polypeptides. The implication associated with activity in the replicon cells is that the inhibitors were able to access the perinuclear membrane, the site of viral RNA synthesis. In addition to the reduction in viral RNA, the cytotoxicity and artificial membrane permeability data for these compounds are consistent with cellular penetrance. Furthermore, these compounds were unlikely to disrupt replicon RNA synthesis by altering local membrane integrity, because they did not affect virus production in a BVDV plaque assay when tested up to 10 M, which requires membrane-localized replication for virus propagation (data not shown).
Because the virus generates a negative strand replicative intermediate to use as a template for producing positive strand RNA, we investigated whether these compounds were equally capable of inhibiting both viral replication processes. Individually, compounds 1 and 4 shared similar activities of inhibition for both negative and positive strand RNA replication in the biochemical assay. Furthermore, data from the cell-based replicon system was consistent with these observations and indicate that the mode of action does not appear to be distinct for these two replication processes. Preliminary structure activity relationships on screening hit 1 led to the identification of analog 4 showing similar potency in both biochemical and cell-based assays. Investigations are ongoing to assess the potential utility of such agents in the treatment of chronic HCV disease.