Novel Broad Spectrum Inhibitors Targeting the Flavivirus Methyltransferase

The flavivirus methyltransferase (MTase) is an essential enzyme that sequentially methylates the N7 and 2’-O positions of the viral RNA cap, using S-adenosyl-L-methionine (SAM) as a methyl donor. We report here that small molecule compounds, which putatively bind to the SAM-binding site of flavivirus MTase and inhibit its function, were identified by using virtual screening. In vitro methylation experiments demonstrated significant MTase inhibition by 13 of these compounds, with the most potent compound displaying sub-micromolar inhibitory activity. The most active compounds showed broad spectrum activity against the MTase proteins of multiple flaviviruses. Two of these compounds also exhibited low cytotoxicity and effectively inhibited viral replication in cell-based assays, providing further structural insight into flavivirus MTase inhibition.


Introduction
The genus Flavivirus in the family Flaviviridae is composed of about 53 arthropod-borne viruses [1][2][3]. The four serotypes of dengue virus (DENV), yellow fever virus (YFV), West Nile virus (WNV), Japanese encephalitis virus (JEV), and Tick-borne encephalitis virus (TBEV) are categorized as global emerging pathogens that can cause serious human disease, including meningitis, myelitis, encephalitis, and hemorrhagic disease [4][5][6][7]. DENV infection threatens approximately 2.5 billion people around the world. Since 1999, WNV has spread rapidly throughout the Western Hemisphere, including the contiguous United States, Canada, Mexico, the Caribbean, and into parts of Central and South America [8]. Although vaccines for humans are currently available for YFV, JEV, and TBEV [6,7], no clinically approved vaccine or antiviral therapy for humans is available for WNV and DENV. Therefore, it is a public health priority to develop and improve vaccines and antiviral agents for prevention and treatment of flavivirus infections.
Various inhibitors of flavivirus MTases have been found through the use of a variety of techniques including cell-based assay, virtual screening, and structure-based design [15,21,22,[24][25][26][27][28][29][30]39]. Although many inhibitors were found to inhibit the N7 and/or 2'-O MTase activities with IC 50 values in the micromolar or nanomolar range (IC 50 : compound concentration required to inhibit 50% of enzyme activity), the majority of these compounds have not shown antiviral efficacy. Only a few of these compounds were found to inhibit the growth of various flaviviruses with an EC 50 in the low micromolar range (EC 50 : effective concentration of compound to inhibit virus growth by 50%) [27,28,30]. However, they display relatively low potency, high cytotoxicity, and/or low therapeutic index.
To search for novel and potent MTase inhibitors, we performed virtual screening of the Diversity Set II library of 1,364 compounds from the National Cancer Institute Developmental Therapeutics Program (NCI DTP). Functional analysis indicated that two compounds, NSC 306711 and NSC 610930, inhibited both the N7 and 2'-O MTase functions. Cytotoxicity and antiviral analyses indicated that they also inhibited the virus growth with low micromolar IC 50 in cell culture. Particularly, compound NSC306711 displayed high therapeutic index.

Virtual screening to identify novel potent inhibitors of flavivirus MTase
A suitable ligand binding pocket for virtual screening (VS) is provided by the crystal structures for SAH and 36A ligands bound to the DENV3 MTase (PDB ID: 3P8Z) [39]. The DENV3 MTase-inhibitor co-structure was chosen because the SAH-derivative inhibitor occupied a flavivirus-conserved pocket [34] and clearly defined the co-factor binding pocket [39]. We first optimized the docking parameters for AutoDock Vina by re-docking SAH and 36A into the SAM-binding site of the MTase. The root-mean-square deviation (RMSD) between the redocked and crystallography-determined conformations of SAH and 36A was 1.2 Å and 1.7 Å, respectively ( fig 1). These numbers are comparable to the ones published previously, by using different structures as models [25][26][27]. We then applied these optimized parameters to dock the NCI diversity set II library into the binding sites of both monomers in the DENV3 MTase structure, using AutoDock Vina. We selected 42 top-ranked compounds with better scores than the SAH control for further investigation ( fig 2).  [26,30]. For example, SAH was reported to require 6-fold lower IC 50 concentration for inhibitions of 2'-O than of N7 [39]. As only the N7 MTase activity is essential for the virus replication [10,33], these 13 compounds were chosen for further analyses, although some of them showed no inhibition towards the 2'-O MTase activity.
We carried out detailed inhibition analyses of these compounds to determine their IC 50 values for both the N7 and 2'-O activities of the WNV MTase (Table 1, Fig 4). In the absence of detergent, the anti MTase potency (IC 50 ) for these compounds ranged from 0.87 μM to 95 μM for the N7 inhibition. To rule out non-specific promiscuous inhibitors [40,41], we also carried out the N7 inhibition experiment for selected non-toxic compounds (see CC 50 below) in the presence of detergent CHAPS ( Table 1). The 2'-O inhibition was only performed in the presence of CHAPS, resulting IC 50 from 4.3 μM to over 300 μM. Fig 4 shows the results of an example dose-response experiment of the best inhibitor, NSC 306711, for both N7 and 2'-O inhibitions (both with CHAPS). Two compounds (NSC 23128 and 115448) were excluded from further analyses as the IC 50 values of these compounds in the presence of detergent were significantly higher than those in the absence of detergent. In addition, compound NSC35489 was also excluded due to the weak inhibition activity. All other compounds, including the most active compound NSC 306711, showed similar IC 50 values with/without CHAPS, indicating that they are likely specific inhibitors. They were chosen for further investigations including cell-based cytotoxicity and antiviral potency analyses. The MTase was in cartoon representation in grey color with representative contact residues in stick representation. Ligands (SAH or 36A) were in stick representation. Colors for atoms unless specified: oxygen, red; nitrogen, blue; carbon for MTase residues, grey; carbon for ligands (crystallography-determined), magenta; carbon for ligands (docked), cyan.

Cytotoxicity and antiviral analyses
Cell-based assays were next performed to evaluate the biological activities of the selected compounds. The cytotoxicity of these compounds was first evaluated by using a MTT cell proliferation assay with a BHK-21 cell line (Table 1, fig 5), as we described previously [20,30]. As shown in fig 5 and Table 1, several compounds were quite toxic to the cells with the CC 50 values similar or less than their in vitro IC 50 values. The rest of the compounds, including NSC 36806, 322921, 306711, and 610930, showed much less toxicity, with CC 50 values nearly 10 times Inhibitions of the N7 and 2'-O methylation activities of the WNV MTase were analyzed on TLC plates. The N7 methylation was measured by conversion of G*pppA-RNA!m 7 G*pppA-RNA; the 2'-O methylation was measured by conversion of m 7 G*pppA-RNA!m 7 G*pppAm-RNA (the asterisk indicates that the following phosphate is 32 P labeled; the RNA represents the first 90 nucleotides of the WNV genome). The spots representing different cap structures on TLC plates were quantified by a PhosphorImager. The relative methylation activity without compounds was set at 100%, and the relative methylation activity with a particular compound was defined as specific activity (compound)/specific activity (no compound) * 100. higher than those of IC 50 values. Therefore, these compounds were further investigated for their in vitro antiviral efficacy.
Viral titer reduction assays were used to evaluate the compounds' antiviral efficacy. As shown in fig 5 and Table 1, compounds NSC 36806 and 322921 did not inhibit the WNV titer  5). Compared to their CC 50 values, the low EC 50 values indicated that these two compounds display relatively good therapeutic window (Table 1). In addition, the antiviral potency of these compounds are consistent with their IC 50 values.  Broad spectrum anti-MTase activity Since the SAM-binding VS target site is conserved among flavivirus MTases [34], a nanomolar inhibitor targeted to this site has the potential to show broad spectrum anti-MTase activity. Therefore, we carried out inhibition assays using the recombinant MTases from DENV2, DENV3, and YFV. We noticed that the 2'-O reaction product m 7 G Ã pppAm migrated to different positions in these experiment ( fig 6A, 6B and 6C). This was due to a known effect of nuclease P1 used in the experiment [20,42]. Due to unknown reasons, when nuclease P1 from US Biological was used, the double methylated product would migrate to a position between G Ã pppA and m 7 G Ã pppA as shown in fig 6B and 6C, whereas it would migrate to a position above m 7 G Ã pppA as shown in fig 6A when    sidechains of residues Ser56, Lys61, and Ser159, and the backbone of residues Gly58, Cys82, Gly86, and Asp146 ( fig 7B). The larger size of this compound allows it to extend out of the pocket and drape over a helical scaffold, making close contacts with 14 amino acid residues from the enzyme (shown as sticks and surfaces in fig 7B). NSC610930 has six electrostatic contacts with the MTase, with sidechains of residues Ser56 and Thr104, and backbones of residues Gly81, Asp146, Glu149, and Arg160 ( fig 7D). Due to its smaller size, it is nestled in the binding pocket and makes close contact with only 9 enzyme residues (shown as sticks and surfaces in fig 7D). The larger number of electrostatic and non-polar contacts between NSC306711 and the enzyme can explain its higher inhibitory capacity as compared to NSC610930. There are commonalities and differences between the backbone and sidechain motif binding to the DENV3 MTase for the four inhibitors: SAH, 36A, NSC306711, and NSC610930. Two of these inhibitors have electrostatic contacts with the Gly86, Trp87, Lys105, Lys130, and Asp146 backbone atoms. In addition, the backbone atoms of Gly58, Cys82, Val132, Glu149, and Arg160 form an electrostatic contact in at least one inhibitor. The common feature of all four inhibitors is an electrostatic contact with the sidechain of Ser56. Two inhibitors show electrostatic contacts with the Asp131 and Asp146 sidechains. In addition, the sidechains of residues Lys61, Thr104, His110, and Ser159 formed electrostatic contacts in at least one inhibitor. The first step in designing new inhibitors using the presently identified compounds as scaffolds could therefore use simple substitutions that can generate additional contacts with this pool of backbone and sidechain motif contacts in the DENV3 MTase SAM-binding pocket.

Analysis of NSC306711 and NSC610930 binding to the WNV MTase
We noticed that although the compounds were initially identified through docking into the SAM-binding pocket of the DENV3 MTase, it appears that the compounds are overall less active against the DENV3 MTase than against the others ( Table 2). One explanation could be that because the substrate used in the assays was an authentic sequence of the WNV and it might not be optimal for the DENV3. This is particularly reasonable as the N7 function of flavivirus MTase requires distinct viral stem-loop structure for optimal reaction [43]. An alternative explanation is that the compounds may bind the MTases differently. To address this concern, we independently docked these two compounds into the WNV MTase (fig 8). The docking conformations were quite different from those for the Dengue MTase, suggesting that one explanation for the differences in activity could be attributed to different binding poses of the molecules in the two binding sites. Whether the compounds bind similarly or differently to these MTases will require mutational and biochemical experiments and/or co-crystal structure with bound inhibitor. However, these are outside the scope of the present study.

Discussion
In this study we have identified potential inhibitors of flavivirus MTase using a virtual screening method, and further examined the efficacy of these compounds using in vitro and cellbased assays. Two of these compounds, NSC306711 and NSC610930, inhibited the MTase proteins of multiple flaviviruses, reduced WNV replication in a dose-dependent fashion, and were relatively non-toxic to BHK-21 cells. The comparatively larger size of NSC306711, and its predicted interaction with MTase residues outside of the SAM binding pocket, may be responsible for its high potency. It is possible that these "extra" interactions outside of the SAM binding pocket could be used as virtual screening parameters to identify inhibitors specific for flavivirus, but not host, MTase proteins.
A challenge to developing inhibitors specific to flavivirus MTase enzymes is the similarity between flaviviral MTases and those of the host cell. Due to the similarity of RNA, GTP, and SAM binding sites of flavivirus and host MTases, inhibitors targeted towards any of these sites may also inhibit host cell MTases and result in toxicity [44]. One difference from host MTases is the presence in flavivirus MTase proteins of an extended cleft continuing from the SAM binding pocket [34]; several inhibitory compounds that project into this cleft have been described [39]. Additionally, residues outside of the SAM binding site may confer specificity as appears to be the case with NSC306711.
A second difference is that host cells divide the N7 and 2'-O methylations among multiple enzymes, whereas flavivirus MTase proteins carry out both functions. One model of flavivirus MTase function posits a translocation of the RNA from an N7 binding position to 2-O' binding position on the same MTase molecule during the methylation process [2,44]. If such a translocation does occur, a small molecule or RNA analogue that blocks this process could prove a viable inhibitor. A previous study exploring compounds that bind in one of the two identified MTase RNA binding sites identified compounds with potency, but not specificity [26].
A potential third route of flavivirus MTase inhibition is to target the GTP binding site using nucleoside analogs to prevent the binding of the capped portion of the viral RNA and its subsequent methylation. Ribavirin, a nucleoside analog used clinically to treat various RNA virus infections, has been shown to bind to the DENV MTase GTP binding site and inhibit RNA cap methylation in vitro [22]. Interestingly, we have identified nucleoside analogs that appear to bind to both the GTP binding site as well as the SAM binding pocket, inhibiting MTase activity in vitro and viral replication [30]. These compounds, along with those identified in this study, give us further insight into the chemical scaffolds most likely to inhibit flavivirus MTase proteins.

Virtual screening
The program Autodock Vina [45] was used for the molecular docking of the NCI diversity set II library obtained from the http://dtpsearch.ncifcrf.gov/FTP/DIVERSITY web address in January 2011. The sdf format library was converted to pdb format using the program babel [46]. The two DENV3 MTase monomers bound to either SAH or 36A (an SAH-derivative inhibitor, PDB ID 3P8Z) [39] were used as the target proteins. A ligand box extending 30 Å in each direction with its center located at the SAH binding site, and an exhaustiveness parameter of 8 was used for the docking. These parameters were chosen based on their ability to dock SAH or 36A (a SAH-based inhibitor) into their correct binding orientations in the target site. The predicted binding energy for SAH according to the Autodock Vina scoring function (-7.2 kcal/mol) was used as a cutoff for top-scoring compounds to test experimentally.

In vitro MTase inhibition assay
The in vitro MTase inhibition assay was performed, using the 5'-end-labeled substrates G Ã pppA-RNA and m 7 G Ã pppA-RNA, representing the first 90 nucleotides of the WNV genome (the asterisk indicates that the following phosphate is 32 P labeled), as described previously [20,30]. The N7 and 2'-O methylation inhibition assays were performed as described previously with the addition of 0.05% CHAPS [10,21]. To rule out none specific inhibitors, N7 inhibition experiment without CHAPS was also performed. The N7 methylation was evaluated by conversion of G Ã pppA-RNA!m 7 G Ã pppA-RNA. The 2'-O methylation was assayed by conversion of m 7 G Ã pppA-RNA!m 7 G Ã pppAm-RNA. The specific activity (%) for N7 was defined as Intensity (m 7 G Ã pppA)/(Intensity (G Ã pppA)+Intensity (m 7 G Ã pppA)) Ã 100). The specific activity (%) for 2'-O was defined as Intensity (m 7 G Ã pppAm)/(Intensity (m 7 G Ã pppA)+Intensity (m 7 G Ã pppAm)) Ã 100). The relative methylation activity without compounds was set at 100%, and the relative methylation activity with a particular compound was defined as specific activity (compound)/specific activity (no compound) Ã 100. The IC 50 value, unless specified, was determined by fitting of the dose-response curve using the ORIGIN software package.

Antiviral assay
A viral titer reduction assay was used to determine the compounds' effect on WNV, as described previously [20,30].

Author Contributions
Conceived and designed the experiments: MB HC BL ZL HL NB LDK. Performed the experiments: MB HC BL NKB SAJ JZ ZL. Analyzed the data: MB HC BL NKB SAJ JZ ZL LDK HL. Contributed reagents/materials/analysis tools: NKB LDK HL. Wrote the paper: MB HC BL NKB SAJ JZ ZL LDK HL.