The C-terminal 50 Amino Acid Residues of Dengue NS3 Protein Are Important for NS3-NS5 Interaction and Viral Replication*

Background: NS3-NS5 interaction is important for the dengue virus life cycle. Results: NS3 residue Asn-570 is essential for its interaction with NS5; mutation in an infectious cDNA abolished virus production and reduced positive-strand RNA synthesis. Conclusion: NS3-NS5 interaction may be required for coordinated positive- and negative-strand RNA synthesis. Significance: NS3-NS5 interaction may be a target for rational design of antiviral drugs. Dengue virus multifunctional proteins NS3 protease/helicase and NS5 methyltransferase/RNA-dependent RNA polymerase form part of the viral replication complex and are involved in viral RNA genome synthesis, methylation of the 5′-cap of viral genome, and polyprotein processing among other activities. Previous studies have shown that NS5 residue Lys-330 is required for interaction between NS3 and NS5. Here, we show by competitive NS3-NS5 interaction ELISA that the NS3 peptide spanning residues 566–585 disrupts NS3-NS5 interaction but not the null-peptide bearing the N570A mutation. Small angle x-ray scattering study on NS3(172–618) helicase and covalently linked NS3(172–618)-NS5(320–341) reveals a rigid and compact formation of the latter, indicating that peptide NS5(320–341) engages in specific and discrete interaction with NS3. Significantly, NS3:Asn-570 to alanine mutation introduced into an infectious DENV2 cDNA clone did not yield detectable virus by plaque assay even though intracellular double-stranded RNA was detected by immunofluorescence. Detection of increased negative-strand RNA synthesis by real time RT-PCR for the NS3:N570A mutant suggests that NS3-NS5 interaction plays an important role in the balanced synthesis of positive- and negative-strand RNA for robust viral replication. Dengue virus infection has become a global concern, and the lack of safe vaccines or antiviral treatments urgently needs to be addressed. NS3 and NS5 are highly conserved among the four serotypes, and the protein sequence around the pinpointed amino acids from the NS3 and NS5 regions are also conserved. The identification of the functionally essential interaction between the two proteins by biochemical and reverse genetics methods paves the way for rational drug design efforts to inhibit viral RNA synthesis.

involved in formation of mature virion and viral RNA replication, respectively. Among the NS proteins, NS3 and NS5 contain the enzymatic activities that are essential for DENV replication (3).
During viral RNA replication within the RC, many critical RNA-RNA, RNA-protein, and protein-protein interactions occur to synthesize both positive-and negative-strand viral RNA (31,32). There have been several reports of NS3-NS5 interactions that include biochemical pulldown assays from infected cell extracts (28,(32)(33)(34) and two-hybrid (Y2H) studies that mapped the interaction to the C-terminal region of NS3 helicase (residues 303-618) and the N-terminal region of NS5 RdRP (residues 320 -368; known as bNLS (nuclear localization sequence)) (35,36). The NS5-binding site appears to be centered at residue Lys-330 because the mutation to alanine disrupted its interaction with NS3 and abolished RNA replication, although the in vitro RdRP activity was unaffected (37). Based on available crystal structures of the RdRP domain of NS5, it has been proposed that the cavity occupied by Lys-330 may be a potential target for antiviral drug design by blocking NS3-NS5 interaction (37). However, the details of the interaction from the NS3 perspective is missing to fully exploit structure-guided drug design.
In this study, through the use of both NS3 WT and mutant peptides in competitive NS3-NS5 interaction ELISA, we identified a conserved amino acid in subdomain III of DENV NS3 protein, Asn-570, as being critical for its interaction with NS5. Mutation of NS3:Asn-570 to alanine in the DENV2 cDNA clone abolished infectious virus production and reduced viral protein production and RNA replication. This mutation also suggests that the NS3-NS5 interaction is essential for viral RNA replication by possibly coordinating positive-and negativestrand synthesis. Small angle x-ray scattering (SAXS) data of NS3 helicase (residues 172-618) covalently linked to NS5(320 -341) supports the observation that physical interaction occurs in the region of interaction between NS3 and NS5.
DENV2 of cosmopolitan genotype (GenBank TM number EU081177.1) that was used in this study was grown in C3/36 cells and titered in BHK-21 cells before storage at Ϫ80°C. This virus was isolated during a local dengue outbreak that occurred in 2005 as part of Early Dengue Infection and Outcome (EDEN) Study in Singapore (38).
Mutation of NS3 Asn-570 to alanine in DENV3 NS2B 18 NS3 was done using QuikChange II XL site-directed mutagenesis kit (Stratagene), according to the manufacturer's protocol. The following primers were used: NS3 N570A forward (5Ј-GATGGGC-AACGCAATGCTCAAATTTTAGAGGAG-3Ј) and reverse (5Ј-CTCCTCTAAAATTTGAGCATTGCGTTGCCCATC-3Ј). The underlined nucleotide corresponds to the mutation that was being made. Mutations were confirmed by automated DNA sequencing.
Peptide Synthesis-Peptides were synthesized at the Nanyang Technological University peptide synthesis core facility (Singapore).
Viral Inhibition Assay-2 ϫ 10 5 Huh-7 cells were seeded into a 12-well plate and incubated overnight at 37°C with 5% CO 2 . Cells were infected with DENV2 at a multiplicity of infection of 1 for 1 h, after which the virus inocula were then removed and replaced with 5% FBS/DMEM maintenance media. At 6 h postinfection, the infected cells were treated with 7.5 M NS3 peptides complexed with 22.5 M penetratin in a molar ratio of 1:3 or 7.5 M NS5-penetratin fusion peptides (Table 1). At 24 h post-infection, cells were washed once with PBS prior to lysis by TRIzol for cellular viral RNA quantification by real time RT-PCR analysis with primers that binds to the NS1 gene (forward 5Ј-CCGCTGACATGAGTTTTGAGTC-3Ј and reverse 5Ј-CATGACAGGAGACATCAAAGGA-3Ј) (40).
ATPase Activity Assay-The assay was carried out as described (13), with slight modifications. Purified NS2B(320 -368)NS3 WT or N570A protein of 2.5 nM was preincubated at 37°C with poly(U) (10 g/ml) in 40 l of reaction buffer (50 mM Tris/HCl, pH 7.5, 2 mM MgCl 2 , 1.5 mM dithiothreitol, 0.05% Tween 20, 0.25 g/ml bovine serum albumin (Sigma)) for 5 min. The reaction was initiated by the addition of 10 l of varying ATP concentrations (2-fold serial dilution, starting from 2000 M) and carried out for 10 min at 37°C. 10 l of malachite green reagent (BioAssay Systems) was added to stop the reaction. Absorbance was read at 635 nm after 30 min at room temperature. The K m of the protein was determined with GraphPad Prism 5, with Michaelis-Menten Equation 1, SAXS-SAXS data of the NS3(172-618) and NS3(172-618)-NS5(320 -341) were measured by the NanoStar TM instrument (Bruker), equipped with a METALJET TM x-ray source and Vantec 2000-detector system. The METALJET TM source uses the liquid gallium source to deliver a high intensity x-ray beam at the wavelength of ϭ 1.34 Å. The SAXS measurements were carried out with the source to sample distance of 145 cm, a two-pinhole collimation system, and the sample to detector distance of 67 cm (41). SAXS experiments of both proteins were carried out at 1.2, 2.2, and 4 mg/ml in a sample volume of 40 l at 15°C. For each sample, a total of nine measurements at 5-min intervals were recorded. The data were flood-field and spatially corrected, and processed using the built-in SAXS software. We tested the possible radiation damage by comparing all data sets, and no changes were detected. The scattering intensity of the buffer was subtracted, and the difference curves were scaled for the concentration. All the data processing steps were performed automatically using the program package PRIMUS (42). The forward scattering I(0) and the radius of gyration R g were evaluated using the Guinier approximation (43). These parameters were also computed from the entire scattering patterns using the indirect transform package GNOM (44), which also provides the distance distribution function p(r). Ten low resolution models of NS3(172-618) or NS3(172-618)-NS5(320 -341) were independently built by the program GASBOR (45). The spatial discrepancy (NSD), which is a measure of similarity between sets of three-dimensional points, was computed between all 10 reconstructions using the DAMAVER program (46). The reconstruction with the least NSD was selected for NS3(172-618) or the fusion protein NS3(172-618)-NS5(320 -341). The ensemble optimization method (EOM) suite was used to select an ensemble of conformations that best fit the experimental data, and the dimensions of selected conformations were compared with the random pool to evaluate the flexibility and compactness of NS3(172-618)-NS5(320 -341) (47,48).
DENV2 Full-length cDNA Clone Construction and Site-directed Mutagenesis-To construct a full-length DENV2 cDNA clone (GenBank TM accession EU081177.1, cosmopolitan genotype; Fig. 4A), low passage virus stock was subjected to viral RNA extraction by RNeasy kit (Qiagen), and three cDNA fragments (fragment boundaries indicated by nucleotide numbers; Fig. 4A) covering the complete genome were amplified from viral RNA by RT-PCR using SuperScript III one-step RT-PCR kits (Invitrogen). Fragment 1 contained the SphI restriction site, a T7 promoter sequence, and DENV2 cDNA nucleotides 1-4498, which also contained the KpnI restriction site. Fragment 2 spanned from the KpnI site (nucleotide position 4493) to the XbaI site (nucleotide position 6008). Fragment 3 spanned from XbaI site (nucleotide position 6003) to the 3Ј end of the genome (nucleotide position 10,723), containing the SacI site at the end of the genome. The PCR product of each cDNA fragment was digested and cloned into a pre-digested and modified low copy number plasmid pWSK29 (49). The plasmid was modified by the replacement of the BssHII site that was located before the T7 promoter with the SphI site by site-directed mutagenesis of the plasmid with the following primer: forward 5Ј-GGCCAGTGAGCATGCGTAATACGAC-3Ј and reverse 5Ј-GTCGTATTACGCATGCTCACTGGCC-3Ј. The underlined nucleotide corresponds to the mutation that was being made. The subclone that maintained each fragment was named accordingly as follows: pWSK29 D2 fragment 1, fragment 2, and fragment 3, respectively, and each subclone was validated by DNA sequencing by 1st BASE DNA Sequencing Services (Singapore) before proceeding for assembly. Subsequently, fragment 2 was inserted into the subclone pWSK29 D2 fragment 1 at the KpnI and XbaI site to generate subclone pWSK29 D2 fragment 1 ϩ 2. Finally, fragment 3 was inserted to generate subclone pWSK29 D2 fragment 1 ϩ 2 at XbaI and SacI sites to generate the full-length cDNA clone, pWSK29 D2 full length. The E. coli XL-1 Blue chemically competent cell (Stratagene) was used for construction and propagation of the cDNA clones. Standard cloning procedures were performed with the exception that the cDNA clones were propagated at 30°C for at least 20 h. All restriction enzymes were purchased from New England Biolabs.
The genome-length cDNA clones with NS3:N570A and NS5: K330A mutations were constructed using the subclone pWSK29 D2 fragment 3. The mutations were generated using QuikChange II XL site-directed mutagenesis kit (Stratagene) and performed according to the manufacturer's protocol. The following primers were used for the generation of both mutants: NS3:N570A forward (5Ј-CTTTGATGGAGTCAAG-AACGCCCAAATCTTGGAAGAAAATG-3Ј) and reverse (5Ј-CATTTTCTTCCAAGATTTGGGCGTTCTTGACTCCAT-CAAAG-3Ј); NS5:K330A forward (5Ј-GTGGTTAGGCTGCT-AACAGCACCTTGGGATGTCATCCCC-3Ј) and reverse (5Ј-GGGGATGACATCCCAAGGTGCTGTTAGCAGCCTAAC-CAC-3Ј). The underlined nucleotides correspond to the mutation that was being made. Mutations were confirmed by automated DNA sequencing, and fragment 3 bearing the mutation was excised from the plasmid by XbaI and SacI and inserted into subclone pWSK29 D2 fragments 1 ϩ 2 that were similarly cut with XbaI and SacI.
In Vitro Transcription, RNA Electroporation, Plaque Assay, Real Time RT-PCR (Reverse Transcription-PCR), Immunofluorescence Assay (IFA), and Western Blot-BHK-21 cells were trypsinized, washed twice with cold PBS, and resuspended in Opti-MEM (Invitrogen) at a cell density of 1 ϫ 10 7 cells/ml. 10 g of in vitro transcribed RNA with T7 mMESSAGE mMACHINE kit (Ambion) of DENV2 WT and mutants were mixed with 800 l of cell suspension in a pre-chilled 0.4-cm cuvette and electroporated at settings of 850 V and 25 microfarads, 2 pulses with an interval of 3 s. Electroporated cells were allowed to recover at room temperature for 10 min before resuspending in complete RPMI 1640 medium for cell recovery. Cells (3 ϫ 10 5 ) were then seeded into a 12-well plate and incubated at 37°C in the presence of 5% CO 2 . Media were changed to 2% FBS maintenance media after 6 h post-transfection. Samples were harvested every 24 h post-transfection until 120 h. Supernatants were collected and clarified for titering of the infectious virus particle by standard plaque assay and extracellular viral RNA quantification by real time RT-PCR analysis (40). Cells were then washed once with PBS prior to lysis by TRIzol reagent (Invitrogen) or 1ϫ SDS-PAGE reducing loading dye for cellular viral RNA quantification and Western blot, respectively.
For cellular viral RNA quantification, total RNA was isolated using the TRIzol extraction method from the cell lysate. 500 ng of total RNA was used for cDNA synthesis using the Improm II reverse transcription system (Promega) with random primer in accordance to manufacturer's instructions. 40 ng of cDNA was used for real time RT-PCR analysis of viral RNA in Bio-Rad iQ-5 real time thermal cycler with the use of SYBR Green supermix (Bio-Rad) as described previously with primers that bind to the NS1 gene (forward 5Ј-CCGCTGACATGAGTTTTGAGTC-3Ј and reverse 5Ј-CATGACAGGAGACATCAAAGGA-3Ј) (40). Absolute numbers of intracellular viral RNA genome copy were quantitated using the DENV standard curve, normalized to actin levels, and reported as absolute number of viral RNA genome copy per g of RNA used for real time RT-PCR.
For intracellular viral RNA genome copy of both positive and negative strands, 5Ј-tagged primers that bind to the E gene (Eden2_forward_RT, 5Ј-GGCCGTCATGGTGGCGAATAACA-GGCTATGGCACTGTCACGAT-3Ј, and Eden2_reverse_RT, 5Ј-GGCCGTCATGGTGGCGAATAACCATTTGCAGCAA-CACCATCTC-3Ј) were used for transcribing cDNAs of both polarities (50,51). The forward primer was used to transcribe cDNA from the negative-strand RNA, whereas the reverse primer was used to transcribe cDNA from positive-strand RNA by using Improm II reverse transcription system (Promega). 40 ng of cDNA was used for real time RT-PCR analysis of viral RNA in Bio-Rad iQ-5 real time thermal cycler with the use of SYBR Green supermix (Bio-Rad) with the appropriate primer pair for either negativestrand (forward 5Ј-GGCCGTCATGGTGGCGAATAA-3Ј and reverse 5Ј-CCATTTGCAGCAACACCATCTC-3Ј) or positivestrand (forward 5Ј-GGCCGTCATGGTGGCGAATAA-3Ј and reverse 5Ј-CAGGCTATGGCACTGTCACGAT-3Ј) detection. The underlined sequence corresponds to the 5Ј-tagged sequence (51). Absolute positive-and negative-strand copy numbers were quantitated and reported as described above.
For extracellular viral RNA quantification, viral RNA from the supernatant was extracted by Qiagen viral RNA extraction kit according to the manufacturer's instructions, and for its quantification, a SYBR Green one-step real time RT-PCR (Bio-Rad) was conducted using the same PCR conditions as cellular RNA quantification, with primers that bind to the NS1 gene (as mentioned above). Absolute viral RNA genome copy was calculated based on the DENV standard curve generated and reported as absolute number of viral RNA genome copy per ml of supernatant. The detection limit of each real time RT-PCR assay is indicated as a gray line in the graph.
IFA against E protein by anti-E mouse antibody 4G2, NS3 protein by anti-NS3 human antibody 3F8, and dsRNA by anti-dsRNA mouse antibody J2 (Scicons), and Western blot against NS3 protein by anti-NS3 3F8 were performed as described previously (33). IFA images were captured on an inverted fluorescence microscope (Olympus IX71, Center Valley) at ϫ20 magnification, and image analysis was performed with ImageJ software (52).
Statistical Analysis-Student's t test was used to determine statistical significance. p values of Յ0.05 were considered as significant.
NS3 Asn-570 Is Critical for the Interaction-To map the NS3 sequence that interacts with NS5 more precisely, we proceeded to screen an array of overlapping 15-mer peptides (Mimotopes) (53,54) that spanned subdomain III in the same competitive ELISA format ( Fig. 2A), and we identified two peptides (NS3(566 -580) and -(571-585)) that moderately (p value ϭ 0.06 and 0.009, respectively) blocked NS3-NS5 interaction, thus narrowing down the interaction region to residues 566 -585 of NS3. We next tested the synthetic peptide NS3(566 -585) in ELISA and showed that it could also disrupt NS3-NS5 interaction in a dose-dependent manner with an IC 50 value of 128.8 Ϯ 2.57 M (Fig. 2B). At the same time, another peptide, NS3(86 -100) with a similar charge as NS3(566 -585), did not compete. This indicates NS3-NS5 interaction involves sequence-specific residues on the NS3 protein. Additionally, the same peptide also inhibited NS3-NS5 interaction in the AlphaScreen assay format where the interacting partners were synthesized in vitro using the wheat germ expression system (data not shown) (55).
Next, we also used the peptide in the viral inhibition assay to determine whether blocking the interaction site could reduce viral replication (Fig. 2C). Uptake of NS3(566 -585) peptide into infected cells was facilitated by a well characterized cellpenetrating peptide, penetratin (56), which forms a nonconva-lent complex with the peptide (57). As shown in Fig. 2C, NS3(566 -585) peptide could reduce viral replication by ϳ33% (p value ϭ 0.0195, Fig. 2C, panel i). As a positive control, we also tested the NS5(320 -341) peptide that was covalently linked to penetratin, which can also facilitate the uptake of the peptide into infected cells (58). NS5(320 -341) peptide could also reduce viral replication by ϳ33% (p value ϭ 0.0043, Fig. 2C,  panel ii). This suggests that blocking interaction between NS3 and NS5 could be a potential therapeutic target.
Sequence alignment of NS3(566 -585) of DENV1-4 and other flaviviruses (Fig. 2D) suggested that Asn-570 (highlighted in gray) is highly conserved within this region and may be critical for the NS3-NS5 interaction. To test this, we synthesized NS3(566 -585)(N570A) peptide and carried out the same competitive ELISA. We found that the replacement of asparagine by alanine at position 570 of NS3 resulted in a null-peptide with respect to its ability to block the NS3-NS5 interaction (Fig. 2B). Next, we expressed and purified the NS3:N570A full-length protein to measure its ATPase activity (59), and we found that it was comparable with the WT NS3 protein (Fig. 2E). The affinity of ATP for NS3 WT and NS3 N570A was similar (K m ϭ 185.8 Ϯ 10.86 and 176.1 Ϯ 8.73 M, respectively). Both proteins also had similar turnover numbers (k cat ϭ 3.15 and 3.11 s Ϫ1 , respectively). From these results, we surmised that Asn-570 of NS3 helicase subdomain III appears to be critical for the NS3-NS5 interaction without affecting the in vitro ATPase activity.

NS3 Residue Asn-570 Is Required for Interaction with NS5
like ATPase (63,64) and the NS3 protease-helicase complex of DENV (11). SAXS data of both proteins at three different concentrations were collected to yield the final composite scattering curves shown in Fig. 3, B and C, which indicate that both proteins are monodispersed in solution. Inspection of the Guinier plots at low angles revealed good data quality and no protein aggregation (Fig. 3, B and C, insets). NS3(172-618) has a radius of gyration (R g ) of 25.16 Ϯ 0.6 Å and a maximum dimension (D max ) of 72.95 Å (Fig. 3D), whereas the NS3(172-618)-NS5(320 -341) has an R g of 25.42 Ϯ 0.6 Å and a D max of 75.64 Å (Fig. 3D). Comparison of the forward scattering of both proteins with the values obtained from a reference solution of NS3(86 -100), which has a similar net charge as NS3(566 -585) and NS3(571-585), was included as negative control. Data are shown as the mean Ϯ S.D. of triplicates from two independent experiments. C, viral inhibition assay was performed with NS3 (panel i) or NS5 (panel ii) peptide that spanned the NS3-NS5 interaction site. For NS3 peptides, they formed a nonconvalent complex with penetratin peptide, which also enables NS3 peptide to be transported into the cells (57). For NS5 peptides, they were synthesized as penetratin fusion peptide as penetratin has been shown to have cell-penetrating property (73), and this enables the NS5 peptide to be transported into the cells. 6 h post-infection, the cells were treated with the peptides. Infected cells were harvested at 24 h post-infection for cellular viral RNA quantification by real time RT-PCR analysis. Fold-change was normalized to 24-h control (penetratin alone) and was plotted, and data are shown as the mean Ϯ S.D. of duplicate from one independent experiment. The x axis labels are as follows: penetratin (p), penetratin and NS3(566 -585) complex (NS3 566 -585 ϩp), and penetratin and NS3(86 -100) complex (NS3 86 -100 ϩp) for panel i and penetratin (p), penetratin fused to NS5(320 -341) (pNS5 320 -341 ), and penetratin fused to scrambled NS5(320 -341) (pNS5 s320 -341 ) for panel ii. D, sequence alignment of NS3 residues 566 -585 of DENV2 with other DENV serotypes and representative members of the Flavivirus genus (74). NS3 residue Asn-570 that is critical for NS3-NS5 interaction is highlighted in gray and bold. The alignment was performed using ClustalW (74).  The arrows indicate the two protrusions, one at the bottom of the NS3(172-618)-NS5(320 -341) solution shape, which leads to a more elongated shape in this protein, reflected by the increased D max value (see Fig. 3D). This protrusion is in proximity to the Asn-570 (blue sticks) and NS3 peptide region, NS3(566 -585) (black schematic) (inset in F). The second protrusion of the NS3(172-618)-NS5(320 -341) may be caused by a conformational alteration due to NS3(172-618) and NS5(320 -341) interactions. G, EOM R g distribution of the selected ensemble (red line) contains a narrow peak at 24.6 Å, which is slightly smaller than the center R g of the random pool (gray filled area), 24.8 Å, suggesting NS3(172-618)-NS5(320 -341) is rigid and compact in solution, and peptide NS5(320 -341) is bound to NS3(172-618).
To eliminate the possibility that the peptide may be flexible in solution, the x-ray scattering data set of NS3(172-618)-NS5(320 -341) had been further analyzed using the Ensemble Optimization Method (EOM) (47,48) to assess the compactness and flexibility of NS3(172-618)-NS5(320 -341) in solution. Based on the width and position of the selected ensemble peak relative to random pool in R g distribution, the flexibility and compactness of the protein in solution can be determined. In the case of NS3(172-618)-NS5(320 -341), the R g distribution of the ensemble contains a narrow peak, indicating that NS3(172-618)-NS5(320 -341) is rigid in solution. This peak is centered at 24.6 Å, which is slightly smaller than the center R g of a random pool, 24.8 Å, suggesting that NS3(172-618)-NS5(320 -341) is compact in solution. The EOM data demonstrate that peptide NS5(320 -341) is bound to NS3(172-618).
NS3:N570A Mutant Has Reduced Infectious Virus Production and Viral Protein Synthesis-To study the impact of NS3: N570A mutation on NS3-NS5 interaction during the virus life cycle, we first generated a DENV2 WT cDNA clone by standard molecular cloning techniques. DENV2 of strain D2/SG/ 05K3295DK1/2005 of the cosmopolitan genotype (GenBank TM accession number EU081177.1) was isolated during a local dengue outbreak that occurred in 2005 as part of Early Dengue Infection and Outcome (EDEN) Study in Singapore, and it was picked as the template for construction of the DENV2 fulllength cDNA clone because of the well documented history of the patients (38). The overall schematic representation of cloning strategy is shown in Fig. 4A. The cDNA clone was subdivided into three fragments based on unique restrictions that are present within the genome to facilitate the assembly of the fulllength clone. To prevent unwanted rearrangement of the clone, both full-length cDNA and cDNA fragments were maintained in low copy plasmid pWSK29 and E. coli XL-1 Blue that were transformed with these plasmids grown at 30°C (49). The assembled full-length cDNA clone contained a T7 promoter at the 5Ј end for in vitro transcription and a SacI site at the 3Ј end for linearization of cDNA. The use of SacI site generated the in vitro RNA transcript, which contains an additional nonviral nucleotide at the 3Ј end, instead of two additional nonviral nucleotides for DENV cDNA that used the XbaI site (65). Prior to linearization for in vitro transcription, the full-length cDNA was checked for possible rearrangement by EcoRI digestion, and the predicted fragments of 9609, 4015, 1284, and 1122 bp were observed on 0.6% agarose gel (data not shown), indicating no sign of recombination (65).
IFA of BHK-21 cells that were transfected with in vitro transcribed genome-length RNA showed increasing numbers of cells expressing E protein from days 1 to 3 post-transfection, with ϳ50 -60% of cells being E-positive on day 3 (Fig. 4B). Plaque assay for supernatant showed an increase in infectious viral particles from days 1 to 3, peaking at 1 ϫ 10 5 pfu/ml on day 3 (Fig. 4C). The size of the plaque on BHK-21 cells, which was produced by the supernatant of BHK-21-transfected cells, was comparable with those produced by the supernatant of C6/36-transfected cells (Fig.  4C, left and right insets, respectively). These results show that the RNA transcribed from the cDNA clone is highly infectious and that infectious virus can be produced in both BHK-21 mammalian cells and C6/36 mosquito cells.
After determining the growth kinetics of cDNA clone-derived DENV2 WT virus, the DENV2 NS3:N570A cDNA mutant clone was generated by site-directed mutagenesis, and its phenotype in infectious virus production and viral protein synthesis was examined over the course of 5 days' post-transfection. For comparison, we also generated the NS5:K330A mutant (37), a known NS3-NS5 interaction-defective mutant that failed to replicate by reverse genetics studies, as a control. Similar to the NS5:K330A mutant, no infectious virus was recovered from the NS3:N570A mutant even when neat supernatant was used for titering (Fig. 5B). In agreement with the lack of infectious virus, extracellular viral RNA levels for both mutants were at the limit of detection of DENV RNA by RT-PCR and did not change over 5 days, suggesting that the mutants are inviable (Fig. 5A). The RNA detection for WT increased over 72 h when it reached maximal virus RNA detection that stabilized until 120 h. Next, to determine whether polyprotein synthesis from the transfected RNA is occurring and that the absence of infectious virus particles may be due to a lack of packaging and/or release of infectious virion, we checked the level of NS3 protein by both Western blot and IFA. Surprisingly, for the cells transfected with NS3:N570A mutant transcript, we could detect NS3 protein using humanized monoclonal antibody 3F8 (33) in the 24-and 48-h samples (Fig.  5C, red arrows) but not in the 72-h sample. The highest percentage of NS3-positive cells (5-10%) was observed on day 2 post-transfection, which declined to Յ5% of NS3-positive cells on day 3 post-transfection (Fig. 5D), and finally to an undetectable level at later time points on days 4 and 5 (data not shown). On the other hand, for the NS5:K330A mutant, the NS3 protein was almost undetectable by Western blot and IFA (Յ1% of NS3-positive cells) at all time points post-transfection (Fig. 5, C  and D). Taken together, the results indicate that reduced NS3-NS5 interaction in NS3:N570A mutant impairs infectious virus production and viral protein synthesis.
NS3:N570A Mutant Showed Accumulation of Negativestrand RNA-The reduction in viral protein synthesis for the NS3:N570A mutant and the lack of detectable infectious plaques suggest that attenuated RNA replication could occur at an early time point for this mutant. Therefore, we examined the RNA replication kinetics by real time RT-PCR and IFA for detection of dsRNA. The results showed that intracellular RNA can be detected for NS3:N570A and WT but, intriguingly, not for NS5:K330A-transfected cells (Fig. 6A and B). The RNA copy for WT increased from around 10 6 copies/g at 6 h post-trans-fection to 10 9 copies/g of RNA at 120 h, as detected by realtime RT-PCR. Although less than WT, a 10-fold increase in intracellular viral RNA copy was observed from day 1 to 2 (p value ϭ 0.008) for NS3:N570A mutant and thereafter, the level declined till day 5. The RNA copy for the NS5:K330A mutant declined from 6 to 48 h post-transfection and remained relatively stable from 48 to 120 h, suggesting that there was no viral RNA replication, and only the input RNA that was transfected by electroporation was being detected. Interestingly, the detection of the peak intracellular RNA level for NS3:N570A mutant on day 2 correlated with the highest level of NS3 protein detected by Western blotting (Fig. 5C). To explore this further, we quantified the level of intracellular positive-and negative-strand viral RNA with strand-specific tagged primers in the real time RT-PCR experiment (Fig. 6C). Essentially, the tag sequence that was described by Plaskon et al. (51) was fused to the sequence that binds to E gene (50) to create a primer that can bind to either the positive or negative strand during the cDNA synthesis step. During real time RT-PCR, the tag and E-specific primers were used to distinguish cDNAs that were transcribed from either the positive-or negative-strand template for accurate quantification of both strands. As expected, for WT viral RNA-transfected cells, the level of both positive-and negative-strand RNAs increased over time until day 3 and declined at slightly different rates after that until day 5. The quantification was consistent with previous data that an excess of positive-over negative-strand RNA can be detected in WT virus-infected cells (66). The RNA quantification for the NS3:N570A mutant revealed a different trend to WT and NS5: K330A mutants. Overall, the level of positive-and negativestrand RNA synthesis for NS3:N570A was lower than WT. However, the level of negative-strand RNA synthesis (rate ϭ 2.06 ϫ 10 4 Ϯ 9.77 ϫ 10 2 negative-strand/h) for NS3:N570A appeared to increase faster than positive-strand RNA (rate ϭ Ϫ2.17 ϫ 10 5 Ϯ 5.82 ϫ 10 4 positive-strand/h) during a 6 -24-h period, and at a somewhat similar rate for both strands (rate ϭ 3.55 ϫ 10 5 Ϯ 3.74 ϫ 10 3 negative-strand/h and 7.48 ϫ 10 5 Ϯ 1.28 ϫ 10 5 negative-strand/h) during the 24 -48-h period, suggesting that the negative-strand RNA may be accumulating as a double-stranded replicative form by base-pairing with the transfected positive-sense RNA. Interestingly, beyond 48 h, the rates at which both positive-and negative-strand RNA accumulated for NS3:N570A mutant over time were almost identical. For NS5:K330A mutant, the level of detected positive-and negative-strand RNA corresponds to the amount of transfected positive-and negative-strand RNA that degraded over time.
(Note: T7 transcribed positive-sense RNA used for transfection contained ϳ0.1-0.3% negative-sense RNA that could be detected by the primers that bind to E gene and is the basis for the gray cutoff line for the residual negative-strand from transfected RNA in Fig. 6C.) The NS3:N570A mutant probably results in either reduced or completely abolished NS3-NS5 interaction and does not support further RNA synthesis from the dsRNA template that is formed (67). This is consistent with the detection of highest percentage of dsRNA-positive cells on day 2 (5-10%) by intracellular RNA staining that does not accu-

NS3 Residue Asn-570 Is Required for Interaction with NS5
mulate as seen in WT at the later time points (Fig. 6B). Interestingly, our detailed analysis suggests that NS5:K330A mutant cannot synthesize negative-strand RNA, although the mutation has been shown to have no effect on the in vitro enzymatic activities (37). These results suggest that the impairment of infectious virus production and viral protein synthesis in the NS3:N570A mutant may be due to the inability of the mutant NS3 to engage with NS5 and use the dsRNA template to produce more positive-strand genomic RNA that can be translated and processed.

DISCUSSION
Flavivirus NS3 and NS5 proteins are recognized as attractive targets for antiviral drug development because of their important functional roles in viral replication (68 -71). Because of the potential interdependence of the two proteins in orchestrating viral genome replication in the RC, the interaction between NS3 and NS5 has been proposed as a promising new target (28,29,37,71). It was previously demonstrated by the Y2H study that NS3(303-618) of the helicase domain interacts with NS5(320 -368) of the RdRP domain. Subsequently, based on available crystal structures, subdomain III of NS3 helicase (residue 483-618) was suggested to be involved in the interaction with NS5 because it has a large protein surface area, located away from the region of main catalytic active sites of NS3 (13,22,36). The NS5 residue Lys-330, which is located on the surface of NS5 thumb subdomain, has been identified as a critical residue in interacting with NS3 helicase (37).
In this study, we focused our attention on identifying the NS3 residues involved in the NS3-NS5 interaction, and we used biochemical, genetic, and biophysical approaches to investigate the functional relevance of the proposed interaction site. Through the use of truncated NS3 protein constructs, overlapping peptides, and peptide phage display, we had fine-mapped the interaction region to residues 566 -585 for NS3 and residues 320 -341 for NS5. Our attempts to obtain crystals of NS3 and NS5 in complex were not successful, and therefore, we employed SAXS to obtain a solution shape that supports the physical interaction between NS3(172-618) and NS5(320 -341). Through EOM analysis of the SAXS data, we were able to demonstrate that NS3(172-618)-NS5(320 -341) is compact and rigid, indicating that peptide NS5(320 -341) makes specific and discrete interactions with the helicase domain, NS3(172-618).
Electrostatic potential surface analysis of NS3(566 -585) and NS5(320 -341) shows a negatively charged contiguous surface on NS3 and a positively charged contiguous surface on NS5, which are charge-complementary to each other, and this further supports the results that the two regions can indeed interact with one another via charge interaction (Fig. 7A). Inspection of the amino acid sequence in NS3(566 -585) region within flavivirus (Fig. 2D) indicated a high conservation of sequence, and we identified a conserved amino acid within this region, residue Asn-570, which we showed to be important for NS3-NS5 interaction; the mutation of this residue to alanine disrupts the in vitro NS3-NS5 interaction. When the same mutation was engineered into the DENV2 cDNA clone, it abolished infectious virus production, similar to NS5:K330A mutant (37). However, unlike the NS5:K330A mutant, the NS3:N570A mutant was able to synthesize low but unsustainable amounts of viral RNA and proteins. Comparison of the pattern of positive-and negative-strand RNA synthesis between WT, NS3: N570A, and NS5:K330A confirmed that the NS5:K330A mutant is completely inactive since transfected positive-strand RNA degraded over time, and the negative-strand RNA detected corresponded to background. The WT RNA-transfected cells showed synchronized synthesis of positive-and negative-strand RNA, which is in agreement with previous studies that have shown an excess of positive-to negativestrand RNA in DENV-infected cells (66). However, it is possible that the long term coordinated synthesis of positive-and negative-strand RNA requires a functional RC with optimum protein interaction affinities between NS3 and NS5 (Fig. 7B). This is supported by the NS3:N570A mutant that showed fairly robust negative-strand synthesis using the transfected RNA as a template, and it is probably analogous to the situation at the early stages of viral replication in an infected cell. The weakened or abolished interaction between NS3 and NS5 in the NS3: N570A mutant does not support the replication of new positive-strand RNA from 6 to 24 h when compared with WT virus, which fits rather well with the carefully conducted real time RT-PCR quantifications. Surprisingly, the NS5:K330A mutant that has been shown to be enzymatically active in vitro and demonstrated to have no interaction with NS3 did not show a FIGURE 7. Schematic representation of NS3-NS5-RNA interaction. A, surface electrostatic potential presentation of NS5(273-900) (RdRP) and NS3(172-618) (helicase). The protein backbones of NS5(320 -341) and NS3(566 -585) are shown in ribbon presentation. The side chains of NS5 Lys-330 and NS3 Asn-570 are displayed. B, simplified schematic model of NS3-NS5 interaction complex with RNA. Dashed line (green) is used to denote that the uncertain path that exiting template RNA from NS3 takes to enter NS5 for complementary daughter strand synthesis. Dark blue line denotes the unwound parental strand; green line denotes the parental strand that serves as template; gray line denotes the newly synthesized daughter strand. similar increase in negative-strand RNA as compared with NS3: N570A mutant. There may be two possible reasons for the inactivity of the NS5:K330A mutant in making negative-strand RNA. First, it may be due to impaired intramolecular signaling between the two NS5 domains that probably coordinate the two functional activities of NS5 (72) required for in vivo polymerase activity. Second, its weak/abolished interaction with NS3 may affect the unwinding of secondary structures in the transfected positive-sense RNA that is needed for negativestrand synthesis. Taken together, our study indicates that different NS3-NS5 interaction-defective mutants can impair infectious virus production, viral protein synthesis, and RNA replication to varying degrees, which is likely to be dependent on the importance of the amino acids that are involved in NS3-NS5 interaction, and also possibly the intramolecular interactions in NS5. It is interesting to note that the coordinated synthesis of positive-and negative-strand RNA at the early stages of replication can be further explored by studying NS3-NS5 interaction mutants displaying varying strengths/degrees of binding and also the contribution of intramolecular cross-talk between the domains of NS5 in strand-specific RNA synthesis in the RC (Fig. 7B). Importantly, we also note that although the NS3 residue Asn-570 is conserved among the four DENV serotypes and several members of flavivirus genus, it is not conserved in yellow fever virus NS3, which has a histidine in place of asparagine at this position in NS3 and tyrosine instead of lysine at position 330 of NS5 (for sequence alignment, refer to Figs. 1B and 5 in Ref. 36).
Overall, this study has identified a potentially new druggable target for the development of antiviral drugs to block NS3-NS5 interaction that is essential for viral replication. The available crystal structures of NS3 and NS5 together with in vitro assays for interaction can be used for in silico and high throughput screening campaigns to find lead molecules for antiviral drug development. Alternatively, because the NS3:N570A mutant genome can be translated to a low level and is able to synthesize negative-strand RNA, it may serve as a potential RNA-mediated vaccine, although the basis for this requires development of new technology platforms.