|Home | About | Journals | Submit | Contact Us | Français|
Dengue virus (DENV) is the most prevalent mosquito-borne viral pathogen in humans. Neither vaccine nor antiviral therapy is currently available for DENV. We report here that N-sulfonylanthranilic acid derivatives are allosteric inhibitors of DENV RNA-dependent RNA polymerase (RdRp). The inhibitor was identified through high-throughput screening of one million compounds using a primer extension-based RdRp assay [substrate poly(C)/oligo(G)20]. Chemical modification of the initial “hit” improved the compound potency to an IC50 (that is, a concentration that inhibits 50% RdRp activity) of 0.7 μM. In addition to suppressing the primer extension-based RNA elongation, the compound also inhibited de novo RNA synthesis using a DENV subgenomic RNA, but at a lower potency (IC50 of 5 μM). Remarkably, the observed anti-polymerase activity is specific to DENV RdRp; the compound did not inhibit WNV RdRp and exhibited IC50s of >100 μM against hepatitis C virus RdRp and human DNA polymerase α and β. UV cross-linking and mass spectrometric analysis showed that a photoreactive inhibitor could be cross-linked to Met343 within the RdRp domain of DENV NS5. On the crystal structure of DENV RdRp, Met343 is located at the entrance of RNA template tunnel. Biochemical experiments showed that the order of addition of RNA template and inhibitor during the assembly of RdRp reaction affected compound potency. Collectively, the results indicate that the compound inhibits RdRp through blocking the RNA tunnel. This study has provided direct evidence to support the hypothesis that allosteric pockets from flavivirus RdRp could be targeted for antiviral development.
The family Flaviviridae consists of three genera: Flavivirus, Pestivirus, and Hepacivirus. The genus Flavivirus contains about 73 viruses, many of which are arthropod-borne and pose major public health threats worldwide (15). The four serotypes of dengue virus infect 50 to 100 million people each year, with approximately 500,000 cases developing into life-threatening dengue hemorrhage fever (DHF) and dengue shock syndrome (DSS), leading to about 20,000 deaths. In addition to DENV, West Nile virus (WNV), Japanese encephalitis virus (JEV), yellow fever virus (YFV), and tick-borne encephalitis virus (TBEV) also cause significant human diseases. No antiviral therapy is currently available for treatment of flavivirus infections. Human vaccines are only available for YFV, JEV, and TBEV (15). Development of antiviral therapy and new vaccines is urgently needed for flaviviruses.
The flavivirus genome is a single-stranded RNA of plus-sense polarity. The genomic RNA contains a 5′ untranslated region (UTR), a single open reading frame, and a 3′ UTR. The single open reading frame encodes a long polyprotein that is processed by viral and host proteases into 10 mature viral proteins. Three structural proteins (Capsid [C], premembrane [prM], and envelope [E]) are components of virus particles. Seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) are responsible for viral replication (40), virion assembly (19, 21, 24, 33), and innate immunity antagonism (4, 16, 23, 29, 30). Two viral proteins encode enzymatic activities that have been targeted for antiviral development. NS3 functions as a protease (with NS2B as a cofactor), helicase, 5′-RNA triphosphatase, and nucleoside triphosphatase (7, 14, 42). The N-terminal part of NS5 is a methyltransferase that methylates the N7 and 2′-O positions of the viral RNA cap structure (13, 18, 37); the C-terminal part of NS5 has an RNA-dependent RNA polymerase (RdRp) activity (1, 39). The RdRp activity is unique to RNA viruses and therefore represents an attractive antiviral target.
Two types of inhibitors could be developed to suppress viral polymerases. Type 1 inhibitors are nucleoside/nucleotide analogs that function as RNA or DNA chain terminators; about half of the current antiviral drugs are nucleotide analogs (10). For flaviviruses, a nucleoside analog (7-deaza-2′-C-methyl-adenosine), originally developed for hepatitis C virus (HCV) RdRp, showed anti-DENV activity (32, 38). We recently reported a similar adenosine analog (7-deaza-2′-C-acetylene-adenosine) that potently inhibited DENV both in cell culture and in mice; unfortunately, this compound showed side effects during a 2-week in vivo toxicity study (44). Nevertheless, these studies have proved the concept that nucleoside analogs could potentially be developed for flavivirus therapy. Type 2 inhibitors are non-nucleoside inhibitors (NNI) which bind to allosteric pockets of protein to block enzymatic activities; the mechanism of action of NNI includes structural alteration of polymerase to an inactive conformation, blocking the conformational switch from polymerase initiation to elongation, or impeding the processivity of polymerase elongation (11). A broad range of chemical classes have been identified as NNI, including inhibitors of HIV (9, 35) and HCV (3, 5, 11, 25).
In the present study, we performed high-throughput screening (HTS) to search for NNI of DENV RdRp. The HTS and chemistry synthesis led to the identification of N-sulfonylanthranilic acid derivatives as inhibitors of DENV RdRp. The compounds specifically inhibit DENV RdRp. UV cross-linking experiments mapped the compound binding site to the RdRp domain of DENV NS5. Amino acid Met343, located at the entrance of RNA template tunnel of the DENV RdRp, was cross-linked to the compound. These results, together with biochemistry experiments, suggest that the compound blocks the RdRp activity through binding to the RNA template tunnel of the polymerase.
Compound synthesis and structure-activity relationship (SAR) studies were recently published (45). Compound NITD-1 was an initial “hit” identified from a DENV RdRp HTS. Over one million compounds from the Novartis library collection were screened. Compound NITD-2 was a “lead” obtained after an SAR study. NITD-29 contained a photoalkylating group benzophenone (2) and was synthesized for a UV cross-linking experiment to map the compound-binding site on the DENV NS5 protein. The synthesis of NITD-29 was described elsewhere (45).
Three types of primer extension-based RdRp assays were performed. All three assays were in a scintillation proximity assay (SPA) format using poly(C) (Sigma-Aldrich) as a template and a 5′-biotinylated oligo(G)20 (Sigma) as a primer. The RdRp activity was quantified by the incorporation of [3H]GTP using SPA beads coated with streptavidin (28, 43). All 50% inhibitory concentrations (IC50s) reported in the present study were calculated using Prism (Graphpad) or ActivityBase software (ID Business Solutions). Full-length NS5 of DENV-2 and DENV-4, each with an N-terminal His6 tag, were generated using a standard baculovirus expression system and Escherichia coli expression system, respectively; the recombinant proteins were purified using a HiTrap chelating column according to the manufacturer's protocol (GE Healthcare).
The first assay type was used for HTS in 384-well white plates. Compounds from the library were spotted onto the plates followed by assembly of reactions (30 μl) containing 50 mM Tris-HCl (pH 7.0), 10 mM KCl, 2 mM MgCl2, 2 mM MnCl2, 0.05% CHAPS, 0.05 U of RNasin/μl, 50 nM protein, 0.8 μM GTP, 0.5 μCi of [3H]GTP, 4 μg of poly(C)/ml, and 0.2 μg of oligo(G)20/ml. After incubation at 23°C for 20 min, the reactions were terminated by adding 150 μg of streptavidin-coated SPA beads to stop solution (75 mM Tris-HCl [pH 7.0], 37.5 mM EDTA, 225 mM NaCl). The plates were further incubated at 23°C for 30 min and subjected to quantification on LEADSeeker (GE Healthcare).
The second assay type was used to allow a single cycle of RNA synthesis. This assay used heparin to trap unbound DENV RdRp and was performed in 96-well plates. Briefly, RNA-NS5 complex was preformed in the HTS assay buffer (described above) at 23°C for 1 h, after which various concentrations of heparin (0 to 96 ng/ml) were added. The reaction mixtures were incubated for another 15 min. Compound NITD-2 and [3H]GTP were then added to the reactions (final volume of 50 μl). After incubation at 23°C for 1 h, the reactions were terminated by adding an equal volume of 2× stop buffer containing 3-mg/ml SPA beads in 50 mM Tris-HCl (pH 7.5), 40 mM EDTA, and 150 mM NaCl. The incorporation of [3H]GTP was measured by a MicroBeta counter (Perkin-Elmer).
The third assay type was used to examine the effect of order of addition of RNA template and inhibitor (during reaction assembly) on the compound potency. Different schemes of RNA template/inhibitor addition are described in Results. For each experimental scheme, various concentrations of NITD-2 were tested to derive the IC50s. The assay buffer and the RNA template concentration were as described for the type 1 assay.
For DENV, the de novo RdRp reaction (25 μl) contained 50 mM HEPES (pH 8.0), 10 mM KCl, 5 mM MgCl2, 2 mM MnCl2, 10 mM dithiothreitol (DTT), 0.5 mM ATP, 0.5 mM UTP, 0.5 mM CTP, 2.5 μCi of [α-33P]GTP (10 μCi/μl, 3,000 Ci/mmol; Perkin-Elmer), 0.25 μg of NS5, 0.5 μg of RNA template, and indicated concentrations of compound NITD-2. The RNA template was in vitro transcribed from a PCR product using a T7 MEGAscript kit (Applied Biosystems). The PCR template contained a T7 promoter, followed by a cDNA fragment representing a DENV-2 subgenome with a deletion from nucleotides 169 to 10263 (GenBank accession number AF038403.1). The de novo RdRp reactions were incubated at 23°C for 30 min; the mixtures were passed through a MicroSpin G-25 column (GE Healthcare) to remove unincorporated nucleoside triphosphates (NTPs); and the RNA elute was extracted with phenol-chloroform and precipitated with ethanol. The RNA pellet was dissolved in 10 μl of RNase-free water, mixed with an equal volume of denaturing Gel Loading Buffer II (Applied Biosystems), and loaded onto a 10% denaturing polyacrylamide gel with 7 M urea. A PhosphorImager was used to quantify the 33P-labeled RNA products.
For WNV, the RdRp assay (20 μl) was performed in 50 mM Tris-HCl (pH 8.0), 5 mM MgCl2, 20 mM NaCl, 2 mM DTT, 0.5 mM ATP, 0.5 mM UTP, 0.5 mM CTP, 5 μM GTP, 15 μCi of [α-33P]GTP (10 μCi/μl, 3,000 Ci/mmol), 1 μg of WNV subgenomic RNA, and 2 μg of WNV full-length NS5. The reactions were incubated at 33°C for 2 h and analyzed on a denaturing PAGE as described for the DENV RdRp assay (above). The expression and purification of the WNV NS5, and the preparation of the WNV subgenomic RNA (containing the 5′-terminal 269 nucleotides directly connected to the 3′-terminal 622 nucleotides of the WNV genome) were reported previously (46).
Activity profiling against HCV NS5B, DNA polymerase α and β were similarly examined using SPA-based assays in 96-well half-area white plates. The HCV NS5B (from genotype 1B strain Con1) contained a deletion of the C-terminal 21 amino acids. The protein (with a C-terminal His tag) was generated in an E. coli expression system and purified through a Ni2+ column (47). The HCV RdRp reaction (40 μl) contained 50 nM NS5B, 50 mM HEPES (pH 7.3), 10 mM KCl, 5 mM MgCl2, 0.5 U of RNasin/μl, 5 mM DTT, 3.6 μg of poly(C)/ml, 0.36 μg of 5′-biotinylated oligo(G)13/ml, 1.33 μM GTP, and 0.16 μCi of [3H]GTP. The reaction was incubated at 30°C for 2 h and stopped by the addition of an equal volume of 2× stop solution (containing 61.2 mM potassium oxalate [pH 1.68] and 50 mM EDTA) and 5-mg/ml streptavidin SPA beads. The plate was quantified by using a MicroBeta counter.
Human polymerase α (Chimerx) and β (Trevigen) were commercially purchased. The polymerase α assay was performed in 56.25 mM Tris-HCl (pH 8.0), 2.5 mM NaCl, 10 mM MgCl2, 0.05% Tween 20, 2.5% glycerol, 5 mM DTT, 3.7 μg of poly(dA)/ml, 0.47 μg of 5′-biotinylated oligo(dT)18/ml, 2 μM total dTTP (with a 1:10 ratio of [3H]dTTP [0.9 μCi/reaction] to cold dTTP), and 0.2 U of polymerase α. The reaction was incubated at 37°C for 60 min. The polymerase DNA β assay was performed in 56.25 mM Tris-HCl (pH 8.0), 150 mM KCl, 2.5 mM NaCl, 10 mM MgCl2, 0.05% Tween 20, 2.5% glycerol, 10 mM DTT, 3.7 μg of poly(dA)/ml, 0.47 μg of 5′-biotinylated oligo(dT)18/ml, 2 μM total dTTP (with a 1:10 ratio of [3H]dTTP [0.9 μCi/reaction] to cold dTTP), and 0.2 U of polymerase β. The reaction was incubated at 37°C for 60 min. Both reactions were stopped and quantified as described for the HCV RdRp assay.
Virus titer reduction assays were performed on Vero cells. Cells were seeded in a 96-well plate (2 × 104 cells per well). At 24 h post-seeding, the cells were infected with DENV-2 (NGC strain), YFV (17D vaccine strain), chikungunya virus at a multiplicity of infection (MOI) of 0.1. The infected cells were immediately treated with compound at the indicated concentration. Culture fluids were collected at 44 h postinfection (p.i.) for DENV-2 and YFV and at 22 h p.i. for chikungunya virus. Virus titers were quantified by standard plaque assays on BHK cells (36).
Labeling of the photoreactive derivative (NITD-29) was performed by incubating 20 μM full-length DENV-2 NS5 protein with 200 μM NITD-29 in a total volume of 30 μl. The sample was exposed to a table top UV light source (UVitec) at a wavelength of 365 nm. The reaction assembly and UV cross-linking were performed in a 4°C room. Samples were withdrawn at 15, 30, and 60 min, and a small aliquot (5 μl) was diluted and used to assess the RdRp activity using the SPA assay. The SPA assay was performed to ensure that the inhibitor had irreversibly inactivated the RdRp by forming a covalent bond with the protein; the non-cross-linked compound was sufficiently diluted such that it would not affect the enzyme activity. The remaining sample was subjected to size exclusion spin column to remove the unincorporated inhibitor and subjected to mass spectrometric (MS) analysis (41).
Samples of untreated and compound-cross-linked NS5 were subjected to SDS-PAGE on 4 to 20% Tris-glycine gels and stained with Coomassie blue. Protein bands at ~100 kDa were excised and in-gel digested with trypsin as described previously (41). After in-gel digestion, tryptic hydrolysates were eluted from the bands with 5% formic acid and collected in a microtiter plate. For matrix-assisted laser desorption ionization (MALDI)-MS and tandem MS (MS/MS) analyses, the tryptic peptides were purified on ZipTips (Millipore Corp., Bedford, MA) by using a Tecan Genesis ProTeam 150 system (Tecan, Maennedorf, Switzerland). After two washes with 5 μl of 80% acetonitrile (ACN)-0.1% trifluoroacetic acid (TFA), the tips were equilibrated twice with 5 μl of 0.1% TFA, and the hydrolysate was applied; after four washes with 5 μl of 0.1% TFA, the peptides were directly eluted or spotted onto ABI 4700 MALDI targets (100-well plates) with 2 μl of a solution of α-cyano-4-hydroxycinnamic acid (5 mg/ml in 50% ACN and 0.1% TFA containing 2 mM NH4H2PO4) applied to the backend of the ZipTips. MALDI spots were analyzed on the Applied Biosystems 4700 Proteomics Analyzer (ABI, Framingham, MA) in MS and MS/MS mode. Both MS and MS/MS data were acquired with a Nd:YAG laser with 200 Hz repetition rate; 2,000 shots were accumulated for each spectrum in MS mode and 4,000 shots were accumulated for each precursor ion in MS/MS mode. MS/MS mode was operated with 1 keV, and the products of metastable decomposition at elevated laser power were detected. MS data were acquired with proximal external calibration and MS/MS data using default instrument calibration.
Peptides with potential cross-links were identified by comparing MALDI-MS spectra of control and compound-cross-linked NS5. MS/MS spectra of parent ions with crosslink-specific mass shifts were analyzed manually to confirm and localize the cross-linked residue.
We identified NITD-1 (Fig. (Fig.1A)1A) as a DENV-2 RdRp inhibitor after screening a library of over one million compounds in a SPA-based RdRp HTS. The HTS assay measured the DENV-2 RdRp-mediated primer extension activity using poly(C)/oligo(G)20 as a RNA template. The HTS results showed a Z′ factor of 0.73 and a hit rate of 0.7%. We focused on NITD-1 after analyzing the “hit” list for drug-like properties such as potency, aqueous solubility, and metabolic stability (data not shown). NITD-1 contained N-sulfonylanthranilic acid and suppressed DENV-2 RdRp activity in a dose-responsive manner, with an IC50 of 7.2 μM (Fig. (Fig.1B).1B). Chemistry synthesis was performed to study the SAR of the “hit.” The details of the SAR results were recently published (45). The SAR study generated NITD-2 which had an improved IC50 of 0.7 μM (Fig. (Fig.1B).1B). Compared to NITD-1, NITD-2 has a replacement of dihydropyrazolone with N-benzyl-pyrazole. In addition to improving the compound potency, the SAR results allowed us to synthesize a photoreactive derivative (NITD-29) that could be used to map the compound-binding site of NS5 through UV-cross-linking (see below).
To examine the spectrum of the anti-polymerase activity, we tested NITD-2 against three other polymerases, including HCV NS5B RdRp, and human DNA polymerases α and β (Fig. (Fig.1C).1C). At 100 μM concentration, NITD-2 reduced 32, 50, and 23% of the HCV NS5B, polymerase α, and polymerase β activities, respectively. Therefore, the IC50s of the compound in these polymerases were ≥100 μM. In contrast, 33 μM NITD-2 suppressed ca. 86% of the DENV-2 RdRp activity. As controls (Fig. (Fig.1D),1D), compounds with known activities against HCV RdRp (LCY967) (8), polymerases α (aphidicolin) (17), and polymerases β (lithocholic) (31) showed expected anti-polymerase results, with IC50s of 50 nM, 40 μM, and 40 μM, respectively. These results indicate that NITD-2 selectively inhibits DENV-2 polymerase.
The HTS assay used poly(C)/oligo(G)20 as a RNA template to measure DENV-2 RdRp-mediated elongation activity; no initiation event was involved. Thus, NITD-2 was assumed to block RNA elongation. However, because multiple rounds of RNA synthesis could occur in the HTS RdRp assay, it was possible that the compound inhibits the transition between multiple rounds of RNA synthesis rather than through a direct inhibition of RNA elongation. To address this question, we examined the activity of NITD-2 in a single round of RNA synthesis by incubating the RdRp reaction with heparin, a highly sulfated glycosaminoglycan polymer that can trap the unbound RdRp. First, we determined the optimal heparin concentration required for the single round RNA synthesis. This was accomplished by incubating various concentrations of heparin (0 to 96 ng/ml) with the binary NS5-RNA complex, which was preformed by incubating NS5 (50 nM) with poly(C)/oligo(G)20 RNA (0.25 μg/ml) for 60 min. After the mixtures were incubated for 15 min, [3H]GTP was added to initiate the RNA synthesis. As shown in Fig. Fig.2A,2A, the addition of heparin to the RdRp reactions reduced the [3H]GTP incorporation in a dose-responsive manner. The reduction of RdRp activity was presumably due to the heparin-mediated trapping of the unbound RdRp after its completion of the first round of RNA synthesis. A saturated concentration of heparin was reached at 50 ng/ml; at this concentration, heparin decreased the total GTP incorporation by two-thirds. Based on this result, we decided to use 50 ng of heparin/ml in our single-round RNA synthesis experiments.
Next, we determined the IC50s of NITD-2 in the presence or absence of heparin (Fig. (Fig.2B).2B). The compound clearly inhibited RdRp activity in the single-round RNA synthesis. However, the presence of heparin reduced the IC50 from 0.7 μM (without heparin) to 0.32 μM (with heparin). As a control, 3′ddGTP, a known RNA chain terminator, also decreased the IC50 from 2 nM (without heparin) to 0.4 nM (with heparin) (Fig. (Fig.2C).2C). Overall, the results clearly demonstrate that NITD-2 inhibits the DENV RdRp-mediated RNA elongation.
Since the above assays used an artificial RNA template [poly(C)/oligo(G)20] and DENV-2 NS5, it is important to validate the compound activity in an authentic de novo RdRp assay. In addition, we sought to determine whether the compound inhibits NS5 from other serotype of DENV. To address these questions, we prepared recombinant full-length NS5 from both DENV-2 and DENV-4 (Fig. (Fig.3A).3A). The NS5 from both serotypes generated an RNA product using the DENV-2 subgenomic RNA (containing the first 168 nucleotides and the last 461 nucleotides of the viral genome) (Fig. (Fig.3B).3B). The length of the RNA product was similar to that of a 5′-end 33P-labeled input RNA template, indicating that the RNA product was derived from the de novo RdRp activity. The DENV-4 NS5 exhibited more robust RdRp activity than the DENV-2 NS5; the difference between the two polymerase activities could be due to a difference in protein folding during the NS5 preparation. Nevertheless, NITD-2 reduced the amount of RNA products in a dose-responsive manner for both DENV-2 and DENV-4 NS5, with IC50s of 4.6 and 5.2 μM, respectively (Fig. (Fig.3B).3B). These results demonstrate that NITD-2 inhibits de novo RNA synthesis (Fig. (Fig.3B),3B), although the potency is reduced compared to the potency derived from the poly(C)/oligo(G)20 elongation RdRp assay (Fig. (Fig.1B).1B). The discrepancy in IC50s between the two assays is most likely due to the difference in experimental conditions.
To test whether NITD-2 could inhibit other flavivirus RdRp, we expressed full-length NS5 of WNV as described before (46). SDS-PAGE analysis of the purified protein showed a major band of ~100 kDa, representing WNV NS5 (Fig. (Fig.3A).3A). A de novo RdRp assay, using a subgenomic RNA template containing the 5′ terminal 269 nucleotides directly connected to the 3′-terminal 622 nucleotides of the WNV genome, was performed. As shown in Fig. Fig.3C,3C, addition of NITD-2 to the WNV RdRp reactions did not affect the yield of de novo RNA product. Overall, these results suggest that the compound specifically inhibits DENV RdRp.
To map the compound-binding site on the protein, we synthesized a photoreactive derivative of the inhibitor, NITD-29. Compared to NITD-2, NITD-29 contained an extra photoalkylating group benzophenone (Fig. (Fig.4A).4A). NITD-29 inhibited the RdRp activity of DENV-2 NS5 [using the poly(C)/oligo(G)20 primer extension HTS assay], with an IC50 of 1.5 μM (Fig. (Fig.4B).4B). Cytotoxicity assay indicated that NITD-29 is not toxic up to 50 μM (Fig. (Fig.4C).4C). Importantly, virus titer reduction experiments showed that NITD-29 at 6 and 17 μM suppressed DENV-2 yields by 3.4- and 4.5-fold, respectively (Fig. (Fig.4D).4D). In contrast, the compound only slightly inhibited YFV at 17 μM; almost no inhibition was observed when chikungunya virus (a plus-strand alphavirus)-infected cells were treated with the compound (Fig. (Fig.4D).4D). These results indicate that NITD-29 selectively inhibits DENV through suppression of the viral RdRp activity.
NITD-29 was UV cross-linked to DENV-2 NS5. The cross-linked complex was subjected to trypsin digestion, followed by MALDI MS analysis. The NITD-29/NS5 complex without UV cross-linking was similarly analyzed as a negative control. A search of the peptide mass fingerprints of the cross-linked sample versus the non-cross-linked sample for a mass difference of 656.17 (NITD-29) revealed a peptide ion at m/z 2,048.87, which was specific for the cross-linked sample (Fig. (Fig.5A,5A, left panel); this ion was not observed in the non-cross-linked analog (right panel). The peptide mass matched the sequence of amino acids 336 to 349 of DENV-2 NS5 (QTGSASSMVNGVVR) modified by 656.17 atomic mass units (amu). The MS/MS spectrum of m/z 2,048.87 from the cross-linked sample showed 6 unmodified C-terminal fragment ions (Fig. (Fig.5B,5B, labeled as y1 to y6 in red) and one unmodified N-terminal b5 ion, all of which match the sequence QTGSASSMVNGVVR, as shown in the bottom panel of Fig. Fig.5B.5B. These results indicate that the modification is located in the sequence stretch SSM of amino acids 341 to 343.
If the Met343 residue is modified, two isomers could provide a possible explanation of the additional fragments (1,344.75, 1,376.73, and 1,392.76, labeled in green in Fig. Fig.5B)5B) observed in the MS/MS spectrum of m/z 2048.87; these isomers were depicted in the upper inset of Fig. Fig.5B.5B. However, a proportion of the C-terminal y5 and y6 ions was shifted by one mass unit [Fig. [Fig.5B,5B, unshifted peaks labeled as y5(a) and y6(a), shifted peaks as y5(b) and y6(b)], indicating partial hydrolysis of Asn345 to Asp345; this reaction is frequently observed for the sequence element NG and is further supported by the pronounced formation of y4 promoted by Asp345 (Fig. (Fig.5B).5B). Collectively, the results strongly indicate that NITD-29 was cross-linked to Met343 within the RdRp domain of NS5. Sequence alignment showed that Met343 is absolutely conserved among the four serotypes of DENV as well as YFV, but it is not conserved in the WNV and JEV NS5.
Based on the crystal structure, Met343 is located next to the thumb subdomain of the DENV RdRp (Fig. (Fig.6A).6A). Docking and molecular simulation suggest that the compound binds to the RNA template formed between the fingers and thumb subdomains; the detailed methods for modeling was recently reported (45). Two RdRp residues, Arg737 and Thr413, form distinct hydrogen bond interactions with the inhibitor; the observed interactions were maintained during the course of the dynamics simulations, indicating their roles in stabilizing the compound in the allosteric binding pocket (45). Sequence alignment showed that residue Arg737 is conserved among mosquito-borne flavivirus NS5, whereas residue Thr413 is divergent among various members of flavivirus.
These results suggest that the compound inhibits RdRp activity through competing with RNA template in the RNA template tunnel. If this is the case, the order of addition of inhibitor and RNA template during the assembly of RdRp assay should affect the compound potency. We performed two sets of experiments to test this hypothesis (Fig. (Fig.6B).6B). For the first set of experiments, we preincubated NS5 with RNA template for 1 h before adding the inhibitor (NITD-2) and [3H]GTP to initiate the reaction; under this assay scheme, the compound displayed an IC50 of 1.80 μM. Reversing the order of addition between the RNA template and inhibitor significantly improved the IC50 to 0.48 μM (p value of <0.05 [Student t test]). For the second set of experiments, we incubated NS5 with RNA template for 1 h, followed by the compound addition. After incubating the mixture for another 1 h, [3H]GTP was added to start the reaction. This experimental scheme yielded an IC50 of 4.63 μM. Reversing the order of addition between the RNA template and inhibitor significantly improved the IC50 to 0.66 μM (P value of <0.05). These results clearly indicate that the compound and RNA template compete against each other during the RdRp reactions.
To develop flavivirus antivirals, we have taken both nucleoside analog and NNI approaches to target DENV RdRp. As the first step toward the NNI approach, we identified an N-sulfonylanthranilic acid derivative (NITD-1) as an inhibitor of DENV-2 RdRp through HTS. SAR study improved the compound potency. We used two distinct assays to demonstrate the anti-DENV RdRp activity: assay 1 measured the activity against RNA elongation of poly(C)/oligo(G)20, and assay 2 measured the activity against de novo RNA synthesis using a DENV-2 subgenomic RNA template. The compound inhibited RNA production in both assays, indicating that the anti-RdRp activity is not template specific. In contrast, the compound (NITD-2) selectively inhibited DENV RdRp; it did not inhibit a closely related WNV RdRp (Fig. (Fig.3C),3C), despite the structural similarity of the two polymerases (26, 43). Furthermore, we showed that the compound did not inhibit HCV RdRp or human DNA polymerase α and β at higher concentrations (Fig. (Fig.1C).1C). The observed DENV-specific activity is desirable for antiviral development because it avoids potential side effects from nonspecific inhibition of host polymerases.
Our biochemical and structural experiments strongly indicate that the compound inhibits RdRp through binding to the RNA template tunnel of the polymerase. This conclusion was supported by three lines of evidence. First, UV cross-linking and MS analyses showed that NITD-29 could be cross-linked to residue M343 within the RdRp domain of NS5 (Fig. (Fig.5).5). This residue is located at the entrance of the RNA template tunnel. Computational docking and simulation suggest that the compound binds to the RNA template tunnel (Fig. (Fig.6A)6A) (45). Second, crystal structure analysis of the DENV-3 RdRp in complex with the compound revealed an extra electron density in the RNA template tunnel (J. Lescar et al., unpublished results). Unfortunately, the current cocrystal form did not allow us to trace the complete compound structure in the RdRp structure. Lastly, biochemical RdRp assay showed that the order of addition of RNA template and inhibitor affected the IC50 of the compound. Addition of the compound before the RNA template during assay assembly displayed a better potency; reversing the order of addition reduced the compound potency (Fig. (Fig.6B6B).
How does the identified compound inhibit RdRp activity through binding to the RNA template tunnel? Crystal structures show that RdRp from DENV-3 and WNV are in a “closed” conformation (26, 43). The proteins are expected to change to an “open” conformation during RNA initiation and elongation. Such conformational switches are required to allow (i) the single-stranded RNA template to enter the template tunnel and (ii) the newly synthesized RNA (together with the template strand) to exit the RNA output channel. The compound identified in the present study may interfere with the conformational change, leading to the reduction of RNA synthesis. Alternatively, the compound binding to the RNA tunnel may sterically block the entrance of the RNA template to the polymerase. Structural studies are needed to further define the mechanism of inhibition.
Development of NNI is often faced with the challenge of virus heterogeneity. The amino acids that form an allosteric binding site for NNI are often under less selective pressure, resulting in quick emergence of resistance strains. In fact, allosteric inhibitors that were developed against one genotype of HCV were often not potent against another genotype (34); in addition, resistant isolates may already exist in the clinical quasispecies (20), which can quickly dominate viral population upon monotherapy of NNI. For flaviviruses, NS5 is the most conserved viral protein (22), among which the NS5 from the four serotypes of DENV has an amino acid homology of 73%. Although we showed that NITD-2 inhibited RdRp activities of both DENV-2 and DENV-4 NS5, the compound did not inhibit WNV RdRp (Fig. (Fig.3).3). The selective inhibition is most likely due to a subtle conformational difference and amino acid variation between the DENV and WNV enzymes.
Inhibitors identified from target-based HTS are often inactive in cell culture. Although NITD-29 inhibited DENV-2 in a virus titer reduction assay (Fig. (Fig.4D),4D), neither NITD-1 nor NITD-2 exhibited any antiviral activity in cell culture (data not shown). Three common reasons could account for the discrepancy between the enzyme activity and the cell culture activity: (i) lack of cell permeability, (ii) metabolic instability, and (iii) inaccessibility of the binding site (due to steric hindrance of other components in the replication complex). In vitro pharmaco-analysis showed that NITD-2 had a logPe value of <−6.0 in the PAMPA (for parallel artificial membrane permeability assay), suggesting that the compound penetrates cell membrane poorly. Chemistry effort is needed to improve the cell permeability of the current lead compound NITD-2.
To our knowledge, the present study reports the first allosteric inhibitor of RdRp from members of the genus Flavivirus. For development of allosteric inhibitors, Malet et al. recently proposed two allosteric pockets that were conserved in the crystal structures of DENV-3 and WNV RdRp (27). Both pockets are located in the thumb subdomain and could be targeted for rational design of small molecular inhibitors. However, before launching an antiviral effort, it would be important to validate the biological relevance of these pockets. This could be accomplished by mutagenesis of recombinant NS5 and analysis of the RdRp/MTase activities of the mutant proteins; the role of these allosteric pockets in viral replication could be examined by mutagenesis of an infectious cDNA clone of DENV. These studies should provide new opportunities for development NNI of DENV.
We thank Subhash Vasudevan, Julien Lescar, Thai Leong Yap, and Wan Yen Lee for helpful discussions and technical support. We also thank Ka Yan Chung and Siew Pheng Lim for providing the plasmid template used for DENV-2 subgenomic RNA synthesis.
Published ahead of print on 17 March 2010.