Dengue virus serotypic replacement of NS3 protease or helicase domain causes chimeric viral attenuation but can be recovered by a compensated mutation at helicase domain or NS2B, respectively

ABSTRACT Mosquito-borne dengue viruses (DENVs) have evolved to four serotypes with 69%–78% amino acid identities, resulting in incomplete immunity, where one serotype’s infection does not cross-protect against secondary infections by other serotypes. Despite the amino acid differences, structural and nonstructural (NS) proteins among serotypes play similar functions. NS3 is an enzyme complex: NS3 has N-terminal protease (PRO) and C-terminal helicase (HEL) activities in addition to 5’ RNA triphosphatase (5’RTP), which is involved in the RNA capping process. In this study, the effects of NS3 replacements among serotypes were tested. The replacement of NS3 full-length (FULL), PRO or HEL region suppressed viral replication in BHK-21 mammalian cells, while the single compensatory mutation improved the viral replications; P364S mutation in HEL revived PRO (DENV3)-replaced DENV1, while S68T alteration in NS2B recovered HEL (DENV1)-replaced DENV2. The results suggest that the interactions between PRO and HEL as well as HEL and NS2B are required for replication competence. Lower-frequency mutations also appeared at various locations in viral proteins, although after infecting C6/36 mosquito cells, the mutations’ frequencies changed, and/or new mutations appeared. In contrast, the inter-domain region (INT, 12 amino acids)-replaced chimera quickly replicated without mutation in BHK-21 cells, although extended cell culture accumulated various mutations. These results suggest that NS3 variously interacts with DENV proteins, in which the chimeric NS3 domain replacements induced amino acid mutations, irrespective of replication efficiency. However, the viral sequences are further adjusted for replication efficiency, to fit in both mammalian cells and mosquito cells. IMPORTANCE Enzyme activities for replicating DENV 5’ cap positive (+) sense RNA have been shown to reside in NS3 and NS5. However, it remains unknown how these enzymes coordinately synthesize negative (-) sense RNA, from which abundant 5’ cap (+) sense RNA is produced. We previously revealed that NS5 dimerization and NS5 methyltransferase(MT)–NS3HEL interaction are important for DENV replication. Here, we found that replication incompetence due to NS3PRO or HEL replacement was compensated by a mutation at HEL or NS2B, respectively, suggesting that the interactions among NS2B, NS3PRO, and HEL are critical for DENV replication.


The replacement of NS3FULL, PRO, HEL, or INT caused viral attenuation
The sequence identities of DENV polyproteins among four serotypes are lowest between DENV2 and DENV4 (Table 1). NS3FULL, PRO, HEL, or INT in DENV2 full-length cDNA was replaced by each corresponding sequence from DENV4 (Fig. 1), creating chimeric proteins that consist of lower amino acids. Each chimeric cDNA was constructed through the yeast recombination method and was used for synthesizing each infectious RNA by in vitro transcription reaction with a cap analog. The synthesized RNA was electroporated into BHK-21 mammalian cells and monitored for replication efficiency by immunofluores cent (IF) staining against viral NS1. The wild-type (WT) DENV2 RNA replicated quickly at 2 days p.e. and induced cell death, accompanied with fragmented nuclei (cell death) shown by Dapi staining at 6 days p.e. (Fig. 2). The WT DENV2 RNA made it difficult to continue culturing in BHK-21 cells beyond 10 days p.e. In contrast, PRO, HEL, or INT-replaced chimeric RNA induced fewer viral positive cells compared with WT DENV2 and delayed its spread to all cells (Fig. 2). Among these chimera, PRO-replaced virus did not expand through the extended cell culture beyond 10 days p.e. and disappeared, while INT chimera quickly expanded during the 6-10 days p.e. (Fig. 2). Among triplicated repetitions, the different features of viral spreads were observed in INT-chimeric RNAs. In one experiment, it was difficult to continue cell culture beyond 10 days due to more severe cell death and aggressively reduced cell numbers (INT (3), Fig. 2), while the other two experiments enabled extended cultures due to less cytotoxicity and stalled expansion of the viral progenies at later time points (INT (1), Fig. 2). HEL-chimeric viruses showed slower expansion than INT chimera in all triplicated experiments. HEL chimeras gradually expanded to BHK-21 cells but lacked cytotoxicity to the cells (Fig. 2). In contrast, NS3FULL-replaced chimera did not show NS1 positivity in the cytoplasm by IF staining through all time points (Fig. 2).

NS3FULL or PRO viral RNA copy numbers were significantly lower than those in WT or INT, while HEL viral copy numbers slowly increased
INT chimera showed the highest copy numbers among NS3 chimera at 6 days p.e. (Fig.  3A). HEL chimera's copy numbers were the next most abundant and reached its highest copy numbers at 26 days p.e., comparable to those of WT at 6 days p.e. (Fig. 3A). In contrast, PRO and FULL chimeras produced significantly lower copy numbers and decreased at 26 days p.e. (Fig. 3A).

Infection of HEL-chimeric viruses to C6/36 mosquito cells showed lower replication than those with infections by WT or INT, did not induce C6/36syncytial change, and lacked the plaque-forming cytotoxicity to LLC/MK2 mammalian cells
The equal viral copy numbers among WT, INT, and HEL chimera viruses (collected from BHK-21 cell culture supernatants at 6, 6, and 26 days p.e., respectively) were used for infecting C6/36 mosquito (Aedes albopictus) cells. INT-chimeric viruses replicated at lower copy numbers compared with WT at 5 days post infection (p.i.) but reached the WT DENV2 copy numbers at 10 days p.i. (Fig. 3B). In contrast, HEL-chimeric viruses replicated with the lowest copy numbers through all time points (Fig. 3B). DENV2 is known to induce syncytium-forming morphological change of C6/36 cells (24). WT-or INT-replaced DENV2 induced syncytial change of C6/36 cells at ~5 days or ~10 days p.i., respectively, while HEL-replaced DENV2 did not show a clear morphological change (Fig. 4A). WT-or INT-chimeric viruses (collected at 5 and 10 days p.i., respectively) induced cell death in LLC/MK2 mammalian cells in plaque assays, while HEL-replaced chimeric viruses did not (Fig. 4B). These results indicate that HEL-replaced DENV2 possesses significantly less replication competence and lowered cytotoxicity, compared with WT or INT chimera.
Reporter-DENV2 replicon RNA assay showed that the FULL-, PRO-, or HELchimeric replicon did not efficiently replicate, compared with the WT-or INTchimeric one Renilla luciferase (Rluc)-contained DENV2 replicon cDNA was constructed to solely evaluate the RNA replication efficiency and to distinguish it from other processes, including the particle formation by structural proteins and the packaging of the synthesized RNA into the particle. The structural polyprotein regions in those infectious cDNAs were replaced by Rluc. Then, IRES sequence was inserted prior to coding NS polyproteins (Fig. 5A). The in vitro transcribed 5' cap RNAs were electroporated into BHK-21 cells and Rluc activities were measured periodically. At 4 hours after electropora tion, all replicon RNAs could induce the initial Rluc activity peaks derived from the translation of the input RNAs (Fig. 5B). At 24 hours, this peak decreased due to the decays of the input RNA. After that, Rluc activities of WT DENV2 replicon increased as a result of the increased newly synthesized replicon RNAs. In contrast, INT-chimeric replicon showed less Rluc activities, reflecting the less synthesized replicon RNAs. Rluc activities of FULL-, PRO-, or HEL-chimeric replicon were notably even lower. This replicon data indicated that all NS3-chimeric replacements except INT significantly lowered the synthesis of the replicon RNA.
The PRO chimera, replaced by DENV1 or DENV3, as well as the HEL chimera by DENV1 were also unable to expand after the initial replication, while HEL chimera by DENV3 could expand slowly In order to confirm whether the chimeric viral expansion depends on the region of the replacement, the infectious DENV2 cDNA was alternately replaced in its PRO or HEL region by the corresponding one from DENV1 or DENV3. PRO-chimeric RNA from DENV1 Full-Length Text or DENV3 showed the disappearance of the viral protein NS1 after 6 days p.e. of the initial replication (Fig. 6). HEL-chimeric RNA from DENV1 stalled the viral expansion and disappeared at later time points (> 26 days) (Fig. 6), while HEL chimera by DENV3 could gradually expand, resembling HEL replacement by DENV4 ( Fig. 2 and 6). These results suggest that the PRO region is more serotypic sequencespecific than the HEL sequence for its role in the replication process.

Combined replacement of NS2B and PRO did not revive replication compe tence
Since NS2B-NS3PRO functions as a cofactor-protease complex, it was considered that the same serotypic NS2B-NS3PRO sequence may be required for efficient viral RNA replica tion. The same serotypic replacements by DENV4 sequences were applied to create NS2B-NS3PRO-replaced infectious DENV2 cDNA. The in vitro transcribed 5' cap chimeric RNA was electroporated into BHK-21 cells. The viral replication was not observed through all time points (Fig. 6).

Full-Length Text
The PRO replacement using the corresponding from DENV3 for DENV1 infectious RNA gained replication efficiency at a later time point In order to improve the viral replication of PRO-replaced RNA, PRO-replacement was designed between DENV1 and DENV3, both of which possess the highest NS3 amino acid identity (85%) among DENV serotypes (Table 1). DENV3 PRO-replacing DENV1 fulllength cDNA was constructed. The transcribed 5' cap chimeric RNA initially replicated in BHK-21 cells during 2-6 days p.e., although after that it became undetectable in IFA (Fig.  6). However, the virus replication reappeared at a later time point (~22 days p.e.) and Full-Length Text gradually expanded its spread through cells (Fig. 6). This result indicated that the virus started to grow only after it acquired higher replication competency at a later time point.

Reduced replacement region of HEL from DENV1 revived the chimeric DENV2 expansion at a later time point
It was also considered that a smaller replacement region of NS3 may improve viral progeny replication. The infectious DENV2 cDNA, in which partly N-terminal HEL (pHEL, 170-514 a.a) ( Fig. 1) was replaced by the corresponding from DENV1, was used for in vitro transcription reaction. This 5' capped pHEL-chimeric RNA showed stalled viral expansion and few NS-1 positive cells until 26 days p.e., similar to the complete HEL (DENV1) replacement, although after that the viral progenies gradually expanded (Fig.  6). This result indicated that the same serotypic C-terminal HEL to the backbone DENV serotype did not improve the stalled replication from the beginning but helped the expansion at a later time point. Partial HEL replacements in DENV2 with the correspond ing from DENV3 (170-510 a.a.) or DENV4 (170-573 a.a.) showed the similar contribution to specially accelerate the expansion at later time points, compared to the fully HELreplaced chimeric RNAs (Fig. 2, 6, and 7).

INT-chimeric viruses could efficiently replicate without acquiring mutations but accumulated various mutations at later time points
One of the INT-chimeric RNA electroporations showed more severe cytotoxicity to BHK-21 cells than the other two INT chimeras and did not induce a mutation at 10 days p.e. (Table 5). After the former INT virus was infected to C6/36 cells, only small numbers (6%-8%) of E mutations occurred and no amino acid alterations in NS proteins were observed, indicating that there is no requirement to change the chimera RNA sequence for efficient replication in both BHK-21 and C6/36 cells. In contrast, the latter two INT-chimeric viruses could be cultured for longer times and acquired mutations in a variety of regions, including E, NS2A, NS2B, NS3HEL, NS4B, NS5MT, and 3'UTR (Table 5). Between these two less cytotoxic INT-chimeric viruses, the common mutations were only seen at NS3HEL (R186K). Since the INT-chimeric virus without acquiring mutations could expand efficiently and have strong cytotoxicity, the mutations in the latter two INTchimeric viruses are not purportedly required for the compensatory mechanism. Those mutations are potentially by-products induced by the INT replacement for maintaining the relationship among viral proteins.
DENV3 PRO-replacing DENV1 induced a high-frequency mutation (99.5%) at NS3HEL (P364S) RNAseq revealed that P364S at HEL was induced as a single high-frequency mutation in DENV3 PRO-replacing DENV1-chimeric viruses ( Table 6). Other amino acid muta tions occurred at NS2B (15%) and NS4A (17%). Therefore, P364S at HEL mutation was potentially a compensatory mutation to improve the replication efficiency. We created DENV3 PRO-replacing DENV1 cDNA containing the P364S mutation in HEL. The transcribed 5' cap RNA was electroporated into BHK-21 cells. It was revealed that the reproduced viruses continuously grew without disappearance after initial replication (~6 days) and quickly expanded to all cells at 18 days p.e. (Fig. 7). With the exception of the mutation at E (E204K), all were lower-frequency mutations (6%-18%) in C, NS2B, NS4B, and NS5MT. These did not have any overlap with the mutations in the PRO (DENV3)-DENV1-chimeric viruses, except the NS2B (I114T) mutation (Table 6). After infecting the PRO-chimeric P364S-derived viruses to C6/36 cells, the I114T mutation disappeared ( Table 6). These results confirm the sole compensatory role of P364S mutation in HEL in reviving the replication competency of the PRO-replaced chimeric RNA.

3'UTR replacements in DENV2 by the corresponding from DENV3 or DENV4 retained efficient replication and even increased cytotoxicity in the plaque assay, compared with WT DENV2
The comparison of 3'UTR nucleotides at 10414 and its surroundings (~35 nucleotides) among serotypes showed that these regional sequences are fully conserved between DENV3 and DENV4, while DENV2 has seven different nucleotides and DENV1 has two differences (Fig. 8A). DENV1 or DENV4 HEL-replacement caused A to T mutation at 10414 (Tables 3 and 7), while DENV3 HEL-replacement induced mutations T to C at 10398, A to G at 10399 or T to C at 10423 (Table 4). All these mutations except 10423 correspond to the 3' UTR sequences in the HEL-replaced serotypes. Therefore, the interaction between 3'UTR and NS3HEL is considered. In order to examine the role of 3'UTR in interaction with NS3HEL, the whole 3'UTR from DENV3 or DENV4 was replaced in the infectious WT DENV2. These in vitro transcribed 5' cap RNAs efficiently replicated after being electropo rated into BHK-21 cells, comparable to WT DENV2 ( Fig. 2 and 7). These 3'UTR-chimeric viruses induced syncytial change in C6/36 cells at 5 days and cytotoxicity to LLC/MK2 cell in plaque assay (Fig. 8B). In particular, the DENV3-3'UTR-chimeric virus made larger plaques than WT DENV2 did ( Fig. 8B and C), suggesting that the 3'UTR-replacement induced more replication competence. RNAseq analysis showed that no amino acid alteration through NS proteins as well as no nucleotide mutation at 3'UTR occurred in 3'UTR-replacing chimeric viruses from DENV3 or DENV4 during replication in BHK-21 cells (Table 8). After infecting these viruses to C6/36 cells, DENV3-3'UTR chimera only showed low frequencies of amino acid mutations at NS5MT (12%-15%) ( Table 8). The data suggests that the replaced 3'UTR is not affecting NS3HEL protein sequences. Among WT DENV1-4, the largest plaque forming ability was shown by WT DENV3 and the smallest was by WT DENV2 (Fig. 8C). It is conceivable that 3'UTR RNA itself affects the replication efficiency and DENV3-3'UTR contributed to more efficient replication, resulting in the larger plaque formation via more cytopathic effects.

DISCUSSION
The details remain unknown on how NS3, NS5, and membrane integral NS2A, 2B, 4A, and 4B proteins coordinately work for a series of replication processes. In this study, the NS3FULL replacement did not show replication through all time points, while either PROor HEL-replaced chimeras replicated from earlier time points (2-6 days p.e.), albeit with delayed viral expansion or the disappearance of the infected cells after the initial replication period. This is a contrary result to NS5 replacement, in which NS5FULL replacement is less attenuated than either MT or POL domain replacement (23), suggest ing that NS3 may be more involved in interactions with other NS proteins, while NS5 may play a more independent role in the replication process. Since NS5 forms dimer (25), the full-length NS5 replacement does not cause a problem for dimerization. HEL (DENV4)-replaced chimeric DENV2 RNAs could replicate slowly and gradually expand to cells, while PRO (DENV4)-replaced DENV2 caused the disappearance of the chimeric virus progenies after 6 days p.e. However, the other serotypic HEL or PRO replacements showed different results; HEL (DENV1)-replaced DENV2 could not expand the viral progenies and disappeared eventually, whereas PRO (DENV3)-replacing DENV1 expanded after 22 days p.e. It was further found that the partial HEL replacement (pHEL) by DENV1 (N-terminal replacement; 170-514 a.a., excluding C-terminal 515-619 a.a.) could make the progeny expand after 26 days p.e. These sudden viral expansions at later time points may indicate the involvement of compensatory mutations in the recovery of replication competence. The RNAseq data revealed the high-frequency NS2B mutation (S68T), with which the pHEL(DENV1)-chimeric RNA replicated efficiently. The pHELchimeric RNA with S68T at NS2B showed faster replication, confirming the role of the NS2B mutation in compensation for the N-terminal HEL replacement. Viral progenies from HEL (DENV3 or DENV4)-replaced infectious RNA could be slowly reproduced, accompanied by diversely localized and various frequency mutations in NS proteins, in addition to prM and E proteins. A possible explanation for the variously located mutations with diversified frequencies is a sequential occurrence of the muta tions, induced by protein-protein interactions. The first mutation leads to a second one, which further causes a third one, and so on, in order to correct the protein-protein interactions. In this regard, all observed mutations are not necessary to regain the impaired functions by compensatory mutations. It is considered that the accumulated mutations are either (1) necessary as a compensatory function to recover the viral life cycle or (2) induced by protein-protein interactions or mechanism(s) other than those required for replication fitness.
The PRO (DENV3)-replaced viral progeny in DENV1 RNA could expand at a later time point by acquiring the compensatory mutation at P346S in HEL. The relationship between PRO and HEL was also observed in the HEL replacement. NS3PRO mutations at K112T (99.8%) and K112Q (28%) were seen in DENV3 HEL-replacing DENV2 (Table 4). P364S in HEL is localized at the dip area from the protein surface area (Fig. 9A), suggest ing that the mutation may not directly affect the binding to the counterpart, but may change the HEL domain structure. S68T at NS2B is positioned at the surface of the hydrophilic region but distant from PRO-binding region (Fig. 9B), suggesting no correla tion with PRO function, although since NS2B joins the replication compartment with the ER membrane, the NS2B hydrophilic region's mutation may affect communication between NS3HEL and the replication compartment formed by NS2B.
Full-Length Text sequence affects HEL and NS4B interaction. Further analysis, such as mutagenesis-based positional analysis, is necessary for clarifying the interaction between NS3 and NS4B.
INT-replaced chimeric RNA could replicate more efficiently than other chimeric RNAs without inducing a mutation. However, mutations were accumulated through the extended culturing INT-chimeric viruses. The results raise a question of whether various mutations could simply be accumulated by culturing for a longer time. Since WT DENV2 is difficult to continue culturing in BHK-21 cells due to strong cytotoxicity, the sequence data at later time points is not available. However, WT DENV1 collected at an earlier time point (6 days after the electroporation into BHK-21 cells plus 5 days infection to C6/36 cells) showed accumulations of various amino acid alterations in E protein, including a high-frequency one at 68% ( Table 6), suggesting that regardless of time points, high-frequency mutations can be quickly accumulated in viral populations. Conversely, the long-time cultured (50 days p.e.) chimeric viruses (DENV1 pHEL-replac ing DENV2) showed few amino acid alterations (Table 7), also suggesting that longer culturing viruses do not simply accumulate broad mutations. Therefore, the mutations Replicon data showed that NS3-chimeric replacements except INT chimera are barely able to replicate (~7 days) (Fig. 5B). However, the infectious HEL or PRO replacement could replicate at 2 days p.e. HEL-replaced infectious RNA gradually expanded without a compensatory mutation, while the PRO-replaced chimeras disappeared after 6 days p.e. and a compensatory mutation was required for regaining replication expansion at a later time point. Therefore, the replicon data is not simply applicable to the mechanism of the HEL-or PRO-replaced infectious RNA. It was reported that the major difference in replication characteristics of infectious viruses compared with replicons was the distribution of the viral double-stranded RNA, which was correlated to the rearrange ment of the ER and Golgi apparatus (28,29). It was suggested that the replicon could induce less rearrangement of the ER and Golgi apparatus, compared with infectious RNA, and therefore, it was suggested that the replicon data only reflects the latent period of the initial replication process prior to the later stage of excessive (+) sense RNA synthesis (28). It is speculated that NS3FULL, PRO, and HEL replacement were all less efficient in the initial replication (latent period) than WT or NS3INT chimera. Since PRO (DENV4)-or HEL (DENV4)-chimeric DENV2 RNAs can be characterized differently based on the viral expansion or disappearance course after the initial replication, the difference between these chimeras must rather exist in rearranging the ER-Golgi apparatus to form the replication compartment. It is considered that the more severe replication incompetence (the viral disappearance) was due to incomplete building of the replication compart ment with rearranged ER and Golgi apparatus and that NS2B-HEL and NS3PRO-HEL interactions are involved in it. The compensatory mutation was required for recovering this ER-Golgi rearrangement in PRO (DENV3)-replaced DENV1 as well as pHEL (DENV1)replaced DENV2, while the ER-Golgi rearrangement could be slowly induced without compensatory mutation in HEL (DENV4 or DENV3)-replaced chimeric RNAs. Reproduced viral progenies from PRO-or HEL-replaced chimera still did not cause cytotoxicity, while the INT-replaced one induced syncytial change of C6/36 cells as well as plaque forming cytotoxicity in LLC/MK2 cells, suggesting the importance of the initial latent period in viral replication efficiency.
The varied frequency A to T nucleotide mutations at 10414 position in 3'UTR were induced in HEL (DENV1 or DENV4)-replaced chimeric DENV2. 10398 and 10399 mutations in 3'UTR were also induced in HEL (DENV3)-replaced DENV2. These 3'UTR mutations may not be important as a compensated role for the HEL replacement since they were not so high-frequency in many recovered progeny viruses and even disappeared after infecting C6/36 cells. The addition of NS2B S68T mutation in the pHEL (DENV1) chimera RNA showed a faster replication without a significant induction of 10414 in 3' UTR (6%, Table 7). It is possible that these mutations in 3'UTR do not correlate to NS3HEL but increase replication competence under another mechanism. The replacement of 3'UTR with corresponding DENV3 or DENV4 did not induce an amino acid mutation at NS proteins nor any nucleotide mutation at 3'UTR, but induced syncytial morphological change in C6/36 cells as well as increased cytotoxicity in the plaque assay, reflecting increased replication efficiency.

Construction of full-length DENV cDNA, in which NS3FULL, PRO, HEL, or INT domain was replaced
The yeast/E. coli shuttle vector plasmid pRS424 was used for cloning the full-length DENV1-4 cDNAs (PRS424-FLDV1-4, gifts from Dr. Falgout, FDA). The replacement (NS3FULL, PRO, HEL, or INT) regions were amplified by PCR, consisting of one to three fragments, including 5' and 3' adjacent homologous sequences (>30 nucleotides, primer details are described in the Table S1). The amplified PCR products were mixed with the linearized pRS424-FLDV2 by XhoI (at 5427-5430 nt position) or pRS424-FLDV1 by NheI (at 5877-5890 nt position) to form Saccharomyces cerevisiae YPH857 competent cells. The homologous recombination-occurred plasmid-contained S. cerevisiae was selected and amplified in Tryptophan-lacked yeast media, followed by transforming and expanding E. coli stbl2 competent cells in ampicillin-contained LB media (30,31). The designed homologous recombination as well as the incomplete replacements happened by chance in yeast cells. These cDNAs sequences were verified by Sanger (Genewiz) and Nanopore sequencings (Plasmidsaurus).

In vitro RNA transcription and electroporation into BHK-21 mammalian cells
Infectious NS3-chimeric DENV cDNAs were linearized by SacII (for DENV1 backbone), or BcgI or SacI (for DENV2 backbone) at the 3' ends. The linearized cDNAs were used as templates for in vitro transcription reaction to synthesize 5' cap infectious RNAs with SP6 RNA polymerase and m 7 GpppG cap analog (30,31). Each synthesized RNA (~3 µg) was electroporated into ~1 x 10 6 BHK-21 cells (American Type Culture Collection), using Amaxa Nucleofector II system (Amaxa). The pulsed cells were immediately spread on a T-25 flask, followed by splitting into T-75 flask at 2 days p.e. Afterward, these cells were continuously cultured by repeated splitting every 4 days using one-third of the trypsinized cells and fresh medium (23).

IF staining against NS1 for detecting WT or chimeric DENV
The productions of the chimeric viruses were periodically monitored by IF staining against viral NS1 (IFA) using 7E11, a monoclonal anti-DENV NS1 antibody (1:200 dilution, gift from Dr. Falgout, FDA). FITC-labeled goat anti-mouse IgG conjugate was used as a secondary antibody (1:100 dilution, Seracare) to visualize viral NS1 under epifluorescent microscope (Olympus IX-71). Nuclear staining was performed with DAPI (1:400 dilution, KPL) and was compared with cytoplasmic localization of NS1. The rate of viral NS1-posi tive cells versus total cell numbers in more than six microscopic fields was counted under NIH image software and compared among samples of different NS3 chimeras and testing dates.

Measurement of viral copy numbers (RT-qPCR) and cytopathic viral titer (plaque assay)
WT or chimeric DENVs in the supernatants of cell cultures were analyzed for their viral copy numbers. Viral RNAs were extracted, using Quick-RNA TM -Viral kit (Zymo Research). RT reaction was performed to make the viral cDNAs by ProtoScript II reverse transcriptase (NEB) with random primer and dNTPs (NEB) at 42°C for 2 hours. The copy numbers of cDNAs were measured by qPCR with primer pair amplifying capsid or NS1 region (Table  S1) and iTaq Universal SYBR green Supermix (Bio-Rad) in Mic qPCR cycler (Bio Molecular System). The average threshold cycle (Ct) values were converted to viral copy numbers by comparing them with the standard amounts of DENV cDNAs.
Plaque assay was performed in paired wells of six well plates, using LLC/MK2 cells, which were infected for 2 hours by WT or chimeric DENV from serially diluted super natants in the infected C6/36 cell cultures. After removal from medium, cells were overlaid with DMEM containing 0.9% SeaPlaque agarose (Lonza) and were continuously incubated for 9-15 days at 37°C. Plaques were fixed with 37% formaldehyde (VWR) and were visualized by crystal violet staining.

RNA extraction from viral particles and RNAseq analysis
Supernatants (~10 mL) were collected from BHK-21 or C6/36 cell cultures in T-75 flasks. Viral particles were precipitated by centrifugation at 15,000 g for 45 min in 40% PEG8000 solution containing 10 mM Tris (pH 8.0), 120 mM NaCl, and 1 mM EDTA (23). After removing the supernatants, the remaining 0.5 mL solutions at the bottom of the tubes, containing the precipitates, were repeatedly mixed with 1.5 mL Trizol-LS (Ambion) and 0.4 mL chloroform. After the centrifugation of these mixed solutions at 12,000 g for 15 mins, the upper layer separated from Trizol/chloroform (bottom layer) was used for ethanol precipitation to collect viral RNA. The ethanol-precipitated pellet was dissolved by Tris (10 mM, pH 8.0)buffered DEPC water and was purified with RNA Clean & Concentrator TM kit (Zymo Research). Library preparation and RNAseq sequencing were performed at Zymo Research Inc. (Irvine, CA), where each RNA sample (260 ng) was purified using Zymo-Seq RiboFree Total RNA Library kit (Zymo Research). Ribosomal RNA was removed from RNA samples, which were then reverse-transcribed to cDNAs. The produced cDNAs were ligated with P7 adaptor sequence at 3' end, followed by second strand synthesis and P5 adaptor ligation to the opposite sites of the double-stranded DNAs. After purification by DNA size (300-600 bp) with beads in the kit, index PCR was performed; initial denaturation at 95°C for 10 mins; 16 cycles of denaturation at 95°C for 30 secs, annealing at 60°C for 30 secs, and extension at 72°C for 60 secs; and final extension at 72°C for 7 mins. Successful library construction was confirmed with Agilent's D1000 ScreenTape Assay on TapeStation. Libraries were sequenced on an pHEL: partly HEL-replaced (NS3 170-571 a.a.). c The cutoff for inclusion is >5% abundance. The numbers refer to the nucleotide localization (1 to 10723) in the chimeric DENV2 RNA, followed by the corresponding original amino acid, location, mutation, and frequency of occurrence shown within parentheses. Bold numbers and letters means >50% frequency of amino acid alteration. ND, not detected.
Illumina Novaseq to a sequence depth of >30 million read pairs (150 bp paired-end sequencing).
Raw Fastq files created from RNAseq reactions were used for alignment analysis of DENV sequences using Geneious Prime software (Biomatters). Paired-end sequences were made from R1 and R2 Fastq files. After trimming the 3' end by BBDuk plugin, the paired-end sequences were aligned to the reference sequence. The created contigs were analyzed for SNPs. Nucleotide alterations at more than 5% frequency were shown.

Statistical analysis
Measurements of viral copy numbers, viral titers, and Rluc replicon assays were performed with duplicated samples and the average numbers and SD were calculated. Viral NS1-posite cell rates were calculated with six microscopic fields, and the average numbers and SD was calculated. Variation analyzed as significant difference was assessed with one-way ANOVA with Bonferroni post-test, using Prism (Graphpad software).

DATA AVAILABILITY
The RNAseq data as raw reads are available as fastq files in the NCBI Short Read Archive (SRA) Data/Download Web page (BioProject accession number, PRJNA916891).

ADDITIONAL FILES
The following material is available online.