Successful Propagation of Flavivirus Infectious cDNAs by a Novel Method To Reduce the Cryptic Bacterial Promoter Activity of Virus Genomes

Abstract

Reverse genetics is a powerful tool to study single-stranded RNA viruses. Despite tremendous efforts having been made to improve the methodology for constructing flavivirus cDNAs, the cause of toxicity of flavivirus cDNAs in bacteria remains unknown. Here we performed mutational analysis studies to identify Escherichia coli promoter (ECP) sequences within nucleotides (nt) 1 to 3000 of the dengue virus type 2 (DENV2) and Japanese encephalitis virus (JEV) genomes. Eight and four active ECPs were demonstrated within nt 1 to 3000 of the DENV2 and JEV genomes, respectively, using fusion constructs containing DENV2 or JEV segments and empty vector reporter gene Renilla luciferase. Full-length DENV2 and JEV cDNAs were obtained by inserting mutations reducing their ECP activity in bacteria without altering amino acid sequences. A severe cytopathic effect occurred when BHK21 cells were transfected with in vitro-transcribed RNAs from either a DENV2 cDNA clone with multiple silent mutations within the prM-E-NS1 region of dengue genome or a JEV cDNA clone with an A-to-C mutation at nt 90 of the JEV genome. The virions derived from the DENV2 or JEV cDNA clone exhibited infectivities similar to those of their parental viruses in C6/36 and BHK21 cells. A cis-acting element essential for virus replication was revealed by introducing silent mutations into the central portion (nt 160 to 243) of the core gene of DENV2 infectious cDNA or a subgenomic DENV2 replicon clone. This novel strategy of constructing DENV2 and JEV infectious clones could be applied to other flaviviruses or pathogenic RNA viruses to facilitate research in virology, viral pathogenesis, and vaccine development.

The Flavivirus genus consists of more than 70 members that are categorized into several antigenic groups (46). Most flaviviruses are transmitted by mosquito or tick vectors and cause serious human and animal diseases (46). They include dengue virus (DENV), Japanese encephalitis virus (JEV), West Nile virus (WNV), yellow fever virus (YFV), and tick-borne encephalitis virus (TBEV). DENV and JEV cause some of the most serious arthropod-borne viral illnesses. There are four different serotypes of dengue virus, DENV1, DENV2, DENV3, and DENV4. Dengue cases have been reported in over 100 countries, and an estimated 2.5 billion people live in areas in which dengue is epidemic (26, 27, 49). DENV infection often leads to dengue fever, dengue hemorrhagic fever, and dengue shock syndrome (24, 28, 48). JEV transmission has been observed in the Southern Hemisphere and has the potential to become a worldwide public health threat. JEV can cause permanent neuropsychiatric sequelae and is sometimes fatal in children (56, 60, 61).

Flaviviruses are enveloped RNA viruses that consist of single-stranded, positive-sense, 10.5- to 11-kb genomic RNA. The genome is associated with multiple copies of capsid proteins that are translated as a single polyprotein. After entering a host cell, the translated polyprotein is then cleaved into three structural proteins (C, prM, and E) and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) by host proteases and a single virus-encoded protease to initiate viral replication (12, 18). Introduction of flavivirus genomic RNA into susceptible cell lines can result in the production of infectious virus particles (1). This phenomenon has prompted the study of flavivirus virology via introduction of flavivirus genomic RNA that has been transcribed in vitro from full-length flavivirus infectious cDNA.

Reverse genetics is a powerful method for studying the viral replication of positive-strand RNA viruses (8). Unfortunately, the instability of full-length flavivirus cDNA in Escherichia coli has been a major hurdle in a attempt to construct flavivirus cDNAs (reviewed in references 4, 54, and 63). Several strategies have been developed to avoid or overcome the instability of infectious flavivirus cDNA. The instability of plasmids containing full-length YFV was avoided using an in vitro ligation approach involving two plasmids (53). The in vitro ligation method has also been successfully applied to other flaviviruses, such as JEV (57), DENV2 (31), DENV3 (5), and TBEV (50). A second approach involves using a special E. coli strain to construct stable full-length DENV4 cDNA (40). This method has been used to construct cDNAs of a DENV1-4 chimera (9), DENV2 (6), DENV3 (5), and TBEV (50), but it has failed in the preparation of DENV3 cDNA (14). A third approach involves the use of medium- or low-copy-number plasmids to stabilize full-length flavivirus cDNAs. This strategy has been employed with different strains of DENV2 cDNAs (23, 36, 51, 71), West Nile virus cDNA (55, 66), Kunjin virus cDNA (34, 35), and TBEV cDNA (43). Two other modified methods have been derived from this strategy to facilitate the construction of flavivirus cDNAs. One is based on homologous recombination of yeast for the assembly of full-length DENV2 (51), DENV1 (52), and DENV4 (33) cDNAs. The other modification uses low-copy-number vectors to construct TBEV infectious clones in AbleK cells by ligating an 11-kb PCR fragment of full-length TBEV cDNA into a vector (22). Despite reports of successful construction of flavivirus cDNAs from the above-described strategies, many of these plasmids are still deleterious for E. coli and result in slow growth and low yield of cDNAs (reviewed in references 4, 54, and 63).

Several attempts to construct genetically stable JEV cDNAs using low-, medium-, and high-copy-numbers of vectors or special bacterial hosts have failed (45, 57, 58). One special design inserts an intron containing a stop codon into the C terminus of a JEV core gene that markedly stabilizes the JEV infectious cDNA clone in bacteria (65). In another approach, full-length JEV cDNAs were successfully harbored by cloning the JEV genome into bacterial artificial chromosome (BAC) vectors (70). YFV (62), DENV1 (59), and DENV2 (47) infectious cDNAs can be assembled into BAC vectors, which suggests that bacteria can withstand any toxicity that arises from flavivirus cDNA cloned into a BAC vector. Similarly, the BAC vector has also been used to construct other RNA virus cDNAs, e.g., transmissible gastroenteritis coronavirus (TGEV) cDNA (2), with poison sequences that make the infectious cDNA clone plasmid unstable. However, the genetic stability of TGEV cDNA in a BAC vector lasts for about 80 generations in E. coli (20). This indicates that the intact full-length TGEV cDNAs in the BAC vector still possess some intrinsic toxicity to E. coli cells. The fact that BAC plasmids were not commonly used for cloning most coronavirus infectious cDNA clones, e.g., TGEV (67), avian infectious bronchitis virus (11), mouse hepatitis virus (69), and severe acute respiratory syndrome coronavirus (68), suggests that use of the BAC vector is not feasible to overcome the intrinsic toxicity in the sequences of most coronavirus genomes.

Little is known about what causes the toxicity of flavivirus cDNAs in E. coli. We sought to develop a reliable and convenient method for constructing stable full-length flavivirus or other RNA virus cDNA by reducing the intrinsic toxicity of the viral genome sequence in E. coli. In the present study, we took a novel approach to assembling infectious DENV2 and JEV cDNA clones by introducing silent mutations into the putative E. coli promoter (ECP) sequences within the DNEV2 and JEV genomes. A cis-acting element located at the central portion of the DENV2 core gene was revealed to be essential for DENV2 replication. Thus, a feasible reverse genetics method was established to prepare full-length infectious DENV2 and JEV cDNA clones. This method will greatly facilitate the manipulation, e.g., sited-directed mutagenesis, deletion, and insertion, of pathogenic flavivirus or other RNA virus cDNA clones that are difficult to manipulate. The feasibility and convenience of handling toxic flavivirus cDNA clones in bacteria by use of our methodology will speed up the study of virology, virus pathogenesis, and vaccine development.

MATERIALS AND METHODS

Reagents.

Dulbecco's modified Eagle's medium (DMEM), alpha-modified minimal essential medium (alpha-MEM), fetal bovine serum (FBS), and RPMI 1640 were obtained from Invitrogen (Carlsbad, CA). Unless otherwise specified, all chemicals were purchased from Sigma (Poole, United Kingdom).

Cell lines and virus strains.

Baby hamster kidney (BHK21) clone 15 cells were kindly provided by P. R. Beatty (Department of Molecular and Cell Biology, University of California at Berkeley) and cultured in alpha-MEM supplemented with 5% FBS at 37°C in a 5% CO₂ incubator. Aedes albopictus C6/36 cells (ATCC CRL-1660) were cultured in RPMI 1640 supplemented with 10% FBS and 1% nonessential amino acids at 28°C in a 5% CO₂ incubator. Virus-infected BHK21 cells were grown in their respective media supplemented with 2% FBS. DENV2 (Taiwanese PL046 strain) (41) and JEV (RP-9 strain) (13) were both kindly provided by C. L. Liao (Institute of Biomedical Sciences, National Defense Medical Center, Taiwan). A virus stock was prepared in C6/36 cells by infecting at an appropriate multiplicity of infection (MOI) in RPMI 1640 medium containing 2% FBS and incubated at 28°C until the appearance of cytopathic effects (CPE). The supernatant was then harvested and stored in 20% FBS at −80°C. Virus titers were determined using a plaque-forming assay in BHK21 clone 15 cells.

E. coli, yeast methods, and strains.

Frozen, competent E. coli strain C41, a derivative of BL21(DE3) (44), was purchased from OverExpress Inc. C41 was cultured in 2YT (16 g of Bacto tryptone, 10 g of yeast extract, and 5 g of NaCl in 1 liter of distilled water [pH 7.2]) medium or agar plates. Standard yeast medium and methods were used (10). Saccharomyces cerevisiae YPH857 was purchased from ATCC. The genotype of YPH857 is MATα ade2-101 lys2-801 ura3-52 trp1-Δ63 HIS5 CAN1 his3-Δ200 leu2-Δ1 cyh2. Competent yeast cells were prepared using the lithium acetate procedure (10).

Plaque-forming assay.

BHK21 clone 15 cells were plated at a density of 8 × 10⁴ cells per well in 12-well plates containing 1 ml of medium and cultured overnight. Then, 0.1 ml of serially diluted virus solution was added to ∼70 to 80% confluent BHK21 cells. After a 2-h adsorption period, the virus solutions were replaced with either 0.75% methyl cellulose (M-0512; Sigma)-containing DMEM and 2% FBS for DENV2-infected cells or 1.2% methyl cellulose-containing DMEM and 2% FBS for JEV-infected cells. On the fifth or third day after DENV or JEV infection, respectively, the methyl cellulose solution was removed from the wells and the cells were fixed and stained with a crystal violet solution (1% crystal violet, 0.64% NaCl, and 2% formaldehyde) (42). The PFU per milliliter of DENV2 or JEV were then determined.

Prediction of E. coli promoter regions in the DENV2 and JEV genomes.

To search for potential E. coli promoter regions within the DENV2 or JEV genome, the website (http://www.fruitfly.org/seqtools-/promoter.html) search program (Neural Network promoter prediction) from the Berkeley Drosophila Genome Project was used to analyze the complete DENV2 or JEV genome sequence.

Preparation of viral RNAs and viral cDNAs by RT.

To prepare viral RNAs, DENV2 and JEV were grown and amplified in C6/36 cells. The amplified DENV2 or JEV was harvested and measured using a plaque assay to determine the titers. We applied approximately 200 μl PL046 or RP9 virus (1 × 10⁶ PFU/ml) to purify viral RNA using the Qiagen RNeasy kit, as described in the manufacturer's protocol. Viral RNAs served as templates for the reverse transcription (RT) of viral RNAs using the Transcriptor first-strand cDNA synthesis kit (Roche Biochemicals) with primer 5′-AGAACCTGTTGATTCAACA-3′ for DENV2 or primer 5′-AGATCCTGTGTTCTTCCT-3′ for JEV, according to the manufacturer's protocol. The RT products of the DENV2 and JEV RNAs served as templates for the synthesis of viral cDNAs by PCR.

Construction of DEN-Luc, JEV-Luc, DENV2 mutant, JEV-A90C, and DENV2 and mutant replicon clones.

The detailed methods for the construction of all clones are described in Materials and Methods in the supplemental material.

In vitro transcription and transfection.

Full-length DENV2 and JEV cDNA clones were linearized with XbaI or KpnI, respectively, extracted with phenol-chloroform, ethanol precipitated, and redissolved in RNase-free water. The in vitro transcription reaction mixtures consisted of 2 μg linearized DNA, 0.5 mM (each) ATP, CTP, and UTP, 0.1 mM GTP, 0.5 mM cap analog m7G(5′)ppp(5′)G (Ambion), 10 mM dithiothreitol (DTT), 40 U RNasin, 30 U SP6 RNA polymerase, and 1× SP6 RNA polymerase buffer in a total volume of 30 μl. The reaction mixtures were incubated at 40°C for 3 h, and 2 μl was loaded in an agarose gel for electrophoresis to determine if the reaction was working. The remaining reaction mixture was aliquoted and stored at −70°C. The typical yields of the transcribed RNAs were approximately 5 μg. To transfect the viral RNA into BHK21 cells, approximately 0.5 μg full-length in vitro-transcribed DENV2 or JEV RNA was incubated with 1 μl Lipofectamine 2000 (Invitrogen) in 75 μl Opti-MEM medium. The mixture was then added to BHK21 cells in 12-well plates and incubated at 37°C for 4 h. The Lipofectamine-RNA mixture was removed, and the cells were fed with alpha-MEM maintenance medium containing 5% FBS. The synthesized virus particles were harvested from the supernatant of the transfected BHK21 cells posttransfection. The harvested viral particles were amplified in C6/36 cells for two passages and then applied to naïve BHK21 cells to test for virus titer.

Indirect immunofluorescence assay (IFA) to detect viral antigens.

BHK21 cells were cultured and manipulated in 12-well plates. BHK21 cells transfected with in vitro-transcribed DENV2 or JEV RNAs were fixed with 1% paraformaldehyde for 1 h at 4°C and then permeabilized with methanol for 15 min at −20°C. The cells were blocked with 2% horse serum, and each well was stained with 0.5 μg/ml anti-DENV prM protein antibody (mouse monoclonal, HB-114; ATCC) (32) or with a 1:3,000 dilution of anti-JEV envelope protein antibody (mouse monoclonal, 4G2; ATCC) (29). The cells were then incubated with an AlexaFluor488 anti-mouse IgG antibody (Molecular Probes, Invitrogen). The nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI), and images were acquired using a Leica DMIRB microscope equipped with CoolSNAP cooled charge-coupled device (CCD) cameras and Empix Northern Eclipse image software.

Replicon RNA transcription and transfection.

Replicon plasmids were linearized by XbaI for DENV2. DNA was phenol-chloroform extracted, precipitated, and used as a template for in vitro transcription using a SP6 Message mMachine kit (Ambion). The RNA was quantified by spectrophotometry and stored at −80°C. RNAs were transfected into BHK21 cells with Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol.

Transient dengue virus replicon assay.

The transient replicon assay was performed for quantification of viral translation and viral RNA synthesis. BHK21 cells were seeded in 24-wells plates (1 × 10⁴ cells per well) and incubated overnight. A 0.5-μg amount of RNA was transfected for each well. Triplicate wells were lysed for luminometry at 4, 8, 12, 24, 36, 48, 60, 72, and 96 h posttransfection. A Renilla luciferase assay kit (Promega) was used to analyze dengue virus replicon activity.

Quantitative RT-PCR.

To quantify replicated viral positive-strand RNA in replicon-transfected BHK21 cells, the RNA was isolated from the above-mentioned cell samples using the Qiagen RNeasy kit as described in the manufacturer's protocol. The primers used for quantitative RT-PCR are listed in Table S3 in the supplemental material. Equal quantities of viral RNAs were reverse transcribed to cDNA using the Invitrogen Thermoscript RT kit with specific primers DV2.PS-R for detection of positive- and negative-strand viral RNA basing on the manufacturer's protocol and as described previously (64), with slight modification. Real-time quantitative PCR was used to measure the viral RNA as described previously (30). The reaction was conducted in a 20-μl volume comprising 5 μl of cDNA, 1× LightCycler TaqMan master mix (Roche), 200 nM (each) primers DV2.U2-F and DV2.L1-R, and 50 nM hydrolysis probe DV2.P1 (TIB MOLBIOL). The LightCycler TaqMan master kit (Roche Biochemicals) and LightCycler 1.5 instrument (Roche Biochemicals) were utilized in this study under the following conditions: preincubation at 95°C for 10 min followed by 45 cycles of three-step incubations at 95°C for 15 s (denaturation), 60°C for 30 s (annealing and elongation), and 72°C for 1 s (complete elongation with a single fluorescence measurement). A linear relationship between RNA copy numbers per milliliter and corresponding threshold cycle (C_T) values over 7 log units of RNA concentration was established (correlation coefficient, r = 0.99). The copy number of RNA was determined by using standards that were measured by spectrophotometry.

Growth curve for virus replication.

C6/36 or BHK21 cells were infected with DENV2 PL046 strain parental virus or with transcript-derived virus at an MOI of 0.01 or 0.1 PFU/cell in six-well plates. The supernatant of infected cells was removed daily and stored at −70°C. DENV2 titers in each collected sample were determined through serial titration on BHK21 cells.

Statistical analysis.

Data were analyzed with SPSS Statistics, version 19. ECP activities of wild-type and silent mutant fragments of DENV2 or JEV were tested for significant differences by using a t test. Viral RNA levels in cells transfected with wild-type and mutant replicons were also tested for significant differences by t test.

RESULTS

Identification of active E. coli promoter sequences within the DENV2 genome.

Due to the known toxicity of flavivirus cDNA to E. coli (54), we initially decided to use either low-copy vectors (pBR322) for cloning in E. coli or a yeast shuttle vector (pRS313) for cloning in yeast cells to reduce the toxicity derived from the DENV2 (PL046 strain) and JEV (RP9 strain) genomes. Unfortunately, we were not able to obtain full-length DENV2 and JEV cDNA clones using either approach. We therefore hypothesized that the toxicity of the DENV2 and JEV cDNAs originates from the cryptic expression of viral proteins in E. coli.

To examine the possibility of cryptic expression of DENV2 and JEV viral proteins in E. coli, a DNA fragment corresponding to the core-prM-E-NS1 genes (nucleotides [nt] 1 to 3000) of DENV2 or JEV was used. The DENV2 or JEV cDNA fragment (nt 1 to 3000) was synthesized by RT-PCR from parental DENV2 (PL046 strain) or JEV (RP9 strain) virus stocks. The RT-PCR-synthesized DENV2 or JEV cDNA fragment (nt 1 to 3000) was subjected to sequencing and analyzed for the prediction of putative ECPs. Surprisingly, 14 putative ECPs with scores higher than 0.9 were found in the DENV2 genome sequences using the Neural Network promoter program from the Berkeley Drosophila Genome Project (http://www.fruitfly.org/seq_tools/promoter.html) (data not shown). In contrast, only four putative ECPs with scores higher than 0.9 were predicted in the JEV genome sequences by the Neural Network promoter program (Table 1).

TABLE 1.

Predicted E. coli promoters within nt 1 to 3000 of the DENV2 and JEV genomes

Virus and E. coli promoter	Virus sequencea	Scoreb
DENV2
ECP1
Wild type	160 …ctgacaaagagattctcact… 205	0.93
Mutant	160 …ctgacGaagCgGttctcact… 205	NDc
ECP2
Wild type	198 …ggaccattaaaactgttcat… 243	0.95
Mutant	198 …ggaccaCtGaaGctgttcat… 243	ND
ECP3
Wild type	376 …actgcaggcatgatcattat… 421	0.94
Mutant	376 …actgcaggcCtgatcattat… 421	ND
ECP4
Wild type	633 …ccacatgggtaacttatggg… 678	0.97
Mutant	633 …ccacatgggtGacttatggg… 678	ND
ECP5
Wild type	1059 …ccaaacaacctgccactcta… 1104	0.95
Mutant	1059 …ccaaGcaacctgccacCcta… 1104	ND
ECP6
Wild type	2104 …tctatcggcaaaatgcttga… 2149	0.98
Mutant	2104 …tctatcggcaGaatgcttga… 2149	ND
ECP7
Wild type	2582 …acaagactggaaaacctgat… 2627	0.96
Mutant	2582 …acaagactggaGaacctgat… 2627	ND
ECP8
Wild type	2615 …acaccagaattgaatcacat… 2660	1.00
Mutant	2615 …acaccagaGCtgaaCcacat… 2660	ND
JEV
ECP9
Wild type	60 …aacggaagataaccatgact… 105	0.94
Mutant	60 …aacggaagCtaaccatgacg… 105	ND
ECP10
Wild type	72 …catgactaaaaaaccaggag… 117	1.00
Mutant	72 …catgacGaaGaaGccaggag… 117	ND
ECP11
Wild type	676 …ctacgtccaatatggacggt… 721	0.90
Mutant	676 …ctacgtccaGtaCggacggt… 721	ND
ECP12
Wild type	1352 …ttgggagaacaatccagccagaaaacatcaaat… 1397	0.94
Mutant	1352 …tCgggagaacaatccagccagaaaacatcaaGt… 1397	ND

^aMutations introduced into ECPs are indicated by uppercase letters.

^bThe score for the predicted prokaryotic promoter was calculated by a website (http://www.fruitfly.org/seq_tools/promoter.html) search program (Neural Network promoter prediction) from the Berkeley Drosophila Genome Project.

^cND, not detectable.

The prediction of 14 putative ECPs with scores higher than 0.9 within nt 1 to 3000 of the DENV2 genome led us to determine if nt 1 to 3000 of the DENV2 genome have active ECPs in E. coli. Segments of DENV2 nt 1 to 3000 were fused to Renilla luciferase and evaluated for fusion protein expression in E. coli. The DENV2 cDNA fragment (nt 1 to 3000) was divided into 10 fragments (nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 1201 to 1500, 1501 to 1800, 1801 to 2100, 2101 to 2400, 2401 to 2700, and 2701 to 3000). Each fragment was fused in frame with the Renilla luciferase gene and inserted into plasmid pRS313 through homologous recombination. Individual constructs containing the DENV2-luciferase gene were transformed into E. coli Stbl2 cells. The Stbl2 cells harboring the DENV2-luciferase gene were cultured and harvested, and cell extracts were subjected to measure luciferase enzymatic activity. As expected, several DENV2-Luc constructs (nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 1201 to 1500, 2101 to 2400, and 2401 to 2700) have relatively higher luciferase enzymatic activity in bacteria than control plasmid pRS-Luc, which contains only the luciferase gene without a DENV2 sequence (Fig. 1 A).

FIG. 1.

Identification of functional E. coli promoters within the DENV2 genome in E. coli. (A) Nucleotides (nt) 1 to 3000 of DENV2 posses different prokaryotic promoter activities. An empty vector plasmid, pRS313, was used to harbor DENV2-Luc fusion constructs or a construct with Luc alone (RS-Luc). Ten fragments corresponding to nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 1201 to 1500, 1501 to 1800, 1801 to 2100, 2101 to 2400, 2401 to 2700, and 2701 to 3000 of the DENV2 genome were fused in frame with the Renilla luciferase gene and transformed into E. coli. Cell lysates from E. coli transformed with various DENV2-Luc constructs or the control plasmid, RS-Luc, were tested for luciferase activity. The error bars represent the SEMs from four independent experiments (n = 4). (B) The silent mutations affect the luciferase activity of DENV2-Luc constructs. Six fragments corresponding to nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 2101 to 2400, and 2401 to 2700 of the DENV2 genome were fused in frame with the Renilla luciferase gene and transformed into E. coli. Cell lysates from E. coli transformed with various DENV2-Luc constructs or the control plasmid, RS-Luc, were tested for luciferase activity. The RS-Luc plasmid served as an empty vector control construct. The error bars represent the SEMs from four independent experiments (n = 4). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. (C) DENV2-Luc fusion protein expression in E. coli cells. Lysates from E. coli cells transformed with various DENV2-Luc constructs with or without mutations were separated by SDS-PAGE, transferred to a polyvinylidene difluoride (PVDF) membrane, and detected using a monoclonal antibody against Renilla luciferase.

To predict the locations of active ECPs within the DENV2 segments of DENV2-Luc constructs (nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 1201 to 1500, 2101 to 2400, and 2401 to 2700), we used the Neural Network promoter program to look for ECPs with higher scores (>0.9) within the DENV2 segments of DENV2-Luc constructs with luciferase activity. Also, we focused on the ECPs that have an optimal start codon located downstream from the ECP initiation site and in frame with the virus coding region (38). We predicted ECP1 (nt 160 to 205) and ECP2 (nt 198 to 243) within the DENV2 segment from nt 1 to 300, ECP3 (nt 376 to 421) within the DENV2 segment from nt 301 to 600, ECP4 (nt 633 to 678) within the DENV2 segment from nt 601 to 900, ECP5 (nt 1059 to 1104) within the DENV2 segment from nt 901 to 1200, ECP6 (nt 2104 to 2149) within the DENV2 segment from nt 2101 to 2400, and ECP7 (nt 2582 to 2627) and ECP8 (nt 2615 to 2660) within the DENV2 segment from nt 2400 to 2700 (Table 1). Since the luciferase activity of DENV2-Luc(1200-1500) is low and there is no optimal start codon (in frame with the dengue virus coding region) downstream of the only predicted ECP (nt 1349 to 1394), we decided to ignore the region from nt 1200 to 1500 of DENV2 genome.

To reduce the likelihood that these putative E. coli promoters were contained in the DENV2 (nt 1 to 3000) genomes, silent mutations were created in the predicted DENV2 ECPs 1 to 8 (Table 1) to reduce the scores without altering the amino acid sequences. The effects of all possible silent mutations on the E. coli promoter prediction scores of DENV2 were then evaluated with the Neural Network promoter program. The silent mutations predicted to severely reduce the scores of ECPs were then chosen for creating DENV2-Luc mutant constructs (Table 1).

To further determine if the chosen silent mutations that are predicted to reduce E. coli promoter activity of ECP1 to -8 of the DENV2 genome (Table 1) indeed affect E. coli promoter activity in bacteria, the silent mutations were individually introduced into the wild-type DENV2-Luc constructs (nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, 1201 to 1500, 2101 to 2400, and 2401 to 2700). Compared to the luciferase activities of the various wild-type DENV2-Luc constructs, most of the silent mutations significantly decreased the luciferase activities of all DENV2-Luc constructs in E. coli (Fig. 1B).

In addition to luciferase activity, Western blotting with anti-Renilla luciferase monoclonal antibody demonstrated DENV2-Luc fusion protein expression. As expected, there was no band appearing in the lane of the RS-Luc control even after longer exposure of the Western blot (Fig. 1C). However, apparent expression of the DENV2-Luc proteins was found in the wild-type DENV2-Luc constructs containing nt 1 to 300, 301 to 600, and 2401 to 2700. Moderate expression of luciferase protein was shown in the wild-type DENV2-Luc constructs containing nt 601 to 900 and 901 to 1200. Several DENV2-Luc constructs with silent mutations exhibited substantially lower DENV2-Luc fusion protein expression (e.g., nt 1 to 300, 301 to 600, 601 to 900, 901 to 1200, and 2101 to 2400), which is consistent with the observation that the luciferase activity of DENV2-Luc constructs in E. coli was significantly reduced by the silent mutations introduced into ECP1 to -8 (Fig. 1B).

Identification of active E. coli promoter sequences within the JEV genome.

To search for putative ECPs within nt 1 to 3000 of the JEV genome, we used the Neural Network promoter program to predict ECPs with scores higher than 0.9. In contrast to the 14 putative ECPs within nt 1 to 3000 of the DENV2 genome, only four ECPs (nt 60 to 105, 72 to 117, 676 to 721, and 1352 to 1397) with scores higher than 0.9 were found (Table 1). Consequently, three JEV-Luc constructs (nt 1 to 150, 600 to 899, and 1200 to 1499) were directly constructed to determine if the predicted ECP9 to -12 are active and synthesize JEV-Luc fusion proteins in E. coli.

Similar to the construction of DENV2-Luc constructs, three JEV segments, i.e., nt 1 to 150 containing ECP9 and ECP10, nt 600 to 899 containing ECP11, and nt 1200 to 1499 containing ECP12, were fused to the empty vector reporter gene (Renilla luciferase) to determine if there was JEV-Luc fusion protein expression in E. coli. Those DNA segments of JEV were inserted in frame with the downstream sequence of the Renilla luciferase gene. Luciferase enzymatic activity was used to measure the promoter activities of various JEV-Luc constructs in E. coli. The wild-type JEV-Luc(1-150) construct displayed strong luciferase activity, indicating strong intrinsic E. coli promoter activity in E. coli (Fig. 2 A). In contrast, much lower luciferase activity was observed in the other two JEV-Luc constructs (600-899 and 1200-1499).

FIG. 2.

Identification of functional E. coli promoters within the JEV genome in E. coli. (A) Nucleotides (nt) 1 to 3000 of JEV posses different prokaryotic promoter activities. An empty vector plasmid, pRS313, was used to harbor JEV-Luc fusion constructs or a construct with Luc alone (RS-Luc). Three fragments corresponding to nt 1 to 150, 600 to 899, and 1200 to 1499 of the JEV genome were fused in frame with the Renilla luciferase gene and transformed into E. coli. Cell lysates from E. coli transformed with various JEV-Luc constructs or the control plasmid, RS-Luc, were tested for luciferase activity. Bars represent wild-type fragments that harbor no mutations. The RS-Luc plasmid served as an empty vector control construct. The error bars represent the SEMs from five independent experiments (n = 5). (B) Effects of silent mutations on the luciferase activity of the JEV-Luc(1-150) construct. Three types of silent mutations, A90C, M1, and DM, were introduced into ECP9, ECP10, and ECP9+10 within the JEV-Luc(1-150) construct. Cell lysates from E. coli transformed with wild-type JEV-Luc(1-150), various JEV-Luc(1-150) mutant constructs, and the control plasmid, RS-Luc were tested for luciferase activity. Bars represent wild-type fragments that harbor no mutations. The RS-Luc plasmid served as an empty vector control construct. The error bars represent the SEMs from five independent experiments (n = 5). **, P ≤ 0.01; ***, P ≤ 0.001. (C) Fusion protein expression of wild-type JEV-Luc(1-150) and various JEV-Luc(1-150) mutants in E. coli cells. Lysates from E. coli cells transformed with the wild-type JEV-Luc(1-150) and various JEV-Luc(1-150) mutant constructs were separated by SDS-PAGE, transferred to a PVDF membrane, and detected using a monoclonal antibody against Renilla luciferase.

Since the JEV-Luc(1-150) construct displayed strong promoter activity in bacteria than JEV-Luc(600-899) and JEV-Luc(1200-1499), we focused on ECP9 and ECP10, which are located at nt 1 to 150 of the JEV genome. To reduce the prokaryotic promoter activity of the JEV-Luc(1-150) construct and determine whether ECP9 and/or ECP10 is responsible for the strong prokaryotic promoter activity of the JEV-Luc(1-150) construct in bacteria, several mutations were designed by the Neural Network promoter program and introduced into ECP9 and/or ECP10 of the JEV-Luc constructs (Table 1). The luciferase activity was greatly reduced when an A-to-C mutation at nt 90 was introduced into ECP9 within nt 1 to 150 of the JEV-Luc(1-150) construct (Fig. 2B). The designed silent mutations (M1) predicted to reduce E. coli promoter activity of ECP10 were also able to significantly reduce the luciferase activity of JEV-Luc(1-150). The luciferase activity was abolished to an undetectable level when mutations were introduced into both ECP9 and ECP10 within the JEV-Luc(1-150) construct. Western blotting was performed to determine if the JEV-Luc(1-150) fusion protein was expressed. Consistent with the enzymatic luciferase activity of wild-type JEV-Luc(1-150) and various mutant constructs in bacteria, introduction of the A90C mutation into ECP9 or of mutations into both ECP9 and ECP10 of the JEV-Luc(1-150) construct decreased JEV-Luc(1-150) fusion protein expression (Fig. 2C).

Stabilization of full-length DENV2 cDNA with silent mutations that reduce E. coli promoter activity within the DENV2 genome.

To prepare DENV2 cDNA fragments for the construction of full-length DENV2 cDNA, we first synthesized three fragments of DENV2 cDNAs (nt 1 to 3000, 2850 to 8000, and 7000 to 10724) derived from RT-PCR of DENV2 viral RNAs. The three individual DENV2 cDNA fragments covered the whole DENV2 genome and were sequenced directly, serving as a reference for later cloning steps. Instead of using conventional molecular cloning technology, we took advantage of the homologous recombination mechanism in yeast cells to clone full-length DENV2 and JEV cDNAs. Initially, we used yeast cells to assemble full-length DENV2 PL046 cDNA without introducing any mutations in the L fragment of the DENV2 genome (Fig. 3). We were not able to obtain the correct full-length pRS/FLDen plasmid purified from His⁺ yeast colonies in E. coli.

FIG. 3.

Construction of a full-length DENV2 cDNA clone with the yeast shuttle vector pRS313. (A) The yeast shuttle vector contained a bacterial replication origin (ori) and selection marker (Amp^r) and a yeast replication origin (CEN) and selection marker (His3). (B) The DENV2 cDNA fragment L was amplified from DENV2 viral RNA by RT-PCR, and eight sets of silent mutations were sequentially introduced into fragment L to create fragment L*. Fragment L* possessed short overlaps with the termini of the linearized pRS313 vector. (C) SacI-linearized pRS313 was mixed with fragment L* and used to transform yeast strain YPH857 to His⁺. In yeast, recombination occurred between the short homologous regions at the termini of the polylinker of pRS313 with fragment L to generate the pRS/DenL plasmid. (D, E, F, and G) The XhoI-linearized pRS/DenL* was incubated with fragment R synthesized from DENV2 viral RNA and transformed into yeast cells to generate the pRS/DenL*R plasmid through homologous recombination between the termini of fragment R and the linear pRS/DenL*. (H, I, and J) The pRS/DenL*R plasmid was linearized using NotI and mixed with fragment M (obtained by RT-PCR of DENV2 viral RNA). It was then transformed into yeast cells to generate full-length pRS/FLDen cDNA. Regions of crossing over (X) are indicated.

After encountering the difficulty faced by most labs in cloning flavivirus cDNAs, we decided to determine if lowering E. coli promoter activity in the viral genome by inserting silent mutations (Table 1) could stabilize DENV2 and JEV cDNA clones in bacteria. Since it has been shown that the N terminus of the dengue virus core gene affects virus replication via functioning as a cis element that enhances recognition of the 5′ core start codon and efficient RNA synthesis (15, 17), we suspected that the silent mutations within ECP1 and ECP2 may affect virus replication, as they are located in the middle of DENV2 core coding region. Thus, we decided to construct four different full-length DENV2 cDNA mutants. One mutant, designated DENV2-8M, contains eight sets of silent mutations within ECP1 to ECP8 of the DENV2 genome. Six sets of silent mutations were introduced into ECP3 to -8 of the DENV2 genome to construct the other mutant, DENV2-6M, The other two mutants, designated DENV2-7M-1 and DENV2-7M-2, contain silent mutations within ECP1 and ECP3 to -8 and within ECP2 to -8, respectively.

To construct the pRS/DenL* as shown in Fig. 3A to D, six to eight sets of silent mutations were sequentially introduced into the L fragment of the dengue virus genome. In general, E. coli colonies with the correct pRS/DenL* plasmid exhibited a relatively normal grow rate, indicating that the DENV2 L* fragment had no apparent toxicity. In contrast, E. coli colonies harboring pRS/DenL grew very slowly, and no correct pRS/DenL plasmid could be consistently obtained.

We continued to use the pRS/DenL* plasmid to finish assembling full-length DNEV2 cDNA clones with the R fragment (nt 7000 to 10724) and M fragment (nt 2850 to 8000) of the DENV2 genome (Fig. 3E to J). Yeast colonies carrying the full-length DENV2 cDNA clone, pRS/FLDen (DENV2-6M, DENV2-7M-1, DENV2-7M-2, and DENV2-8M), were identified by colony PCR to detect the M fragment, and all six yeast colonies were positive. After the purified yeast DNA was transformed into C41, the C41 colonies carrying full-length DENV2 cDNA clones (DENV2-6M, DENV2-7M-1, DENV2-7M-2, and DENV2-8M) were generally homogeneous in size, unlike the pRS/DenL plasmid, which resulted in various sizes. Four out of six positive yeast DNAs were chosen and transformed into C41. The DENV2-6M, DENV2-7M-1, DENV2-7M-2, and DENV2-8M plasmids purified from the C41 colonies displayed good DNA yields and correct restriction enzyme digestion patterns. The DNA yield of the DENV2-8M construct (1,604 ng/10⁹ cells) was slightly better than those of the DENV2-6M, DENV2-7M-1, and DENV2-7M-2 constructs (1,530 ng/10⁹ cells, 1,560 ng/10⁹ cells, and 1,560 ng/10⁹ cells, respectively) in bacteria.

Since we were not able to obtain a wild-type DENV2 cDNA clone, it is impossible to quantitatively show the difference in plasmid stability between the wild-type and various mutant DENV2 cDNA clones. Instead, we used a pRS313 vector (the same vector used to harbor DENV2 mutant clone) to harbor the widely used stable hepatitis C virus 1b replicon reporter sequence (7) in order to construct the pRS313-HCV1b clone, which has a size similar to that of DNEV2 mutant clones. The plasmid stability of the HCV1b replicon construct was compared with those of various DENV2 mutant clones. In general, the number of bacterial colonies formed represents the degree of toxicity or stability of plasmids in bacteria. Thus, the number of bacterial colonies formed was used to assess the plasmid stability of DENV2 mutant constructs. One nanogram of pRS313-HCV1b plasmid transformed into C41 bacterial cells resulted in 4,287 ± 1,051 colonies (mean ± standard error of the mean [SEM]), 1,443 ± 390 colonies were detected in the bacteria transformed with 1 ng of DENV2-6M plasmid, and 2,147 ± 265, 1,903 ± 203, and 3,200 ±565 colonies were found in the bacteria transformed with 1 ng of DENV2-7M-1, DENV2-7M-2, and DENV2-8M plasmids, respectively. The colony formation efficiency indicates that silent mutations indeed increase the stability of DENV2 mutant plasmids in bacteria. The stability of cloned DENV2-6M infectious cDNAs was further assessed by repeated subculture in E. coli strain C41. C41 was transformed with DENV2-6M, and the grown colonies were selected on a 2YT agar plate containing ampicillin (25 μg/ml). Four colonies were picked, mixed, and incubated at 30°C in liquid 2YT medium containing ampicillin (25 μg/ml) for an additional 1 day. The saturated 2YT medium containing bacteria harboring DENV2-6M plasmids was restreaked onto a 2YT-ampicillin agar plate. Four colonies grown on the agar plate were picked and mixed for the same growth testing cycle for five times. After five growth cycles, four clones containing DENV2-6M plasmid were tested, and the integrity was confirmed by using two sets of restriction enzyme digestion to make sure that no transposon insertion and DNA rearrangement occurred during the passages. The integrity of DENV2-6M could be maintained for at least five passages of subcultures. DENV2-6M, DENV2-7M-1, DENV2-7M-2, and DENV2-8M plasmids obtained from C41 colonies were verified by DNA sequencing to check the integrity of the cloned DENV2 genomic cDNA, and correct plasmids were subjected to further analyses.

Previously, we intended to directly assemble full-length DENV2 cDNA in bacteria without doing any mutagenesis within the DENV2 genome. We failed to obtain the correct clone but obtained an incorrect full-length DENV2 plasmid, DENV2-def (defective cDNA clone), with several mutations within the poison DNA fragment, core-prM-E-NS (data not shown). After we successfully assembled DENV2-6M infectious cDNA clones in yeast cells and amplified them in bacteria, we wanted to know if our previous failed cloning strategy (direct cloning in bacteria) could assemble the poison region containing six sets of silent mutations to obtain the correct full-length infectious clone. We digested DENV2-def with SacI (just upstream of the SP6 promoter sequence) and SmaI (nt 3554 within the DENV2 genome) restriction enzymes to remove the DNA fragment (nt 1 to 3554) containing poison sequences to serve as a vector to assemble the corresponding DNA fragment (nt 1 to 3554) with six sets of silent mutations by ligation in bacteria. As expected, we consistently obtained five correct DENV2-6M clones out of six bacterial colonies. Six sets of silent mutations were verified in those correct clones by DNA sequencing analyses. Thus, we can obtain the DENV2-6M infectious cDNA clone by assembling in either yeast or bacterial cells.

Assembly of a full-length JEV cDNA by reducing E. coli promoter activity within the JEV genome.

As we had successfully cloned the full-length DENV2 cDNA, we used a similar cloning strategy to construct a full-length JEV cDNA clone (Fig. 4) by introducing silent mutations into ECP9 and -10 (Table 1). The results shown in Fig. 2B indicated that moderate luciferase activity was found in JEV-Luc(600-899) and JEV-Luc(1300-1599) and strong luciferase activity in JEV-Luc(1-150). Thus, we decided to determine if the mutations introduced into ECP9, ECP10, or ECP9+10 stabilize the full-length JEV cDNAs in E. coli.

FIG. 4.

Construction of a full-length JEV cDNA clone with the yeast shuttle vector pRS313. (A) The yeast shuttle vector contained a bacterial replication origin (ori) and selection marker (Amp^r) and a yeast replication origin (CEN) and selection marker (His3). (B) JEV cDNA fragment M was amplified from JEV viral RNA by RT-PCR. Fragment M possesses short overlaps with the termini of the linearized pRS313 vector. (C) BamHI-linearized pRS313 was mixed with fragment M and used to transform yeast strain YPH857 to His⁺. (D) In yeast, recombination occurred between the short homologous regions at the termini of the polylinker of pRS313 with fragment M to generate the pRS/JEVM plasmid. (E, F, and G) ClaI-linearized pRS/JEVM was incubated with the R fragment synthesized by RT-PCR from viral RNA and transformed into yeast cells to generate the pRS/JEVMR plasmid through homologous recombination. (H, I, and J) The L fragment synthesized from viral RNA by RT-PCR was first mutated to create the L* fragment. The L* fragment was incubated with the XhoI-linearized pRS/JEVMR plasmid and transformed into yeast cells to generate the full-length pRS/FLJEV infectious cDNA. Regions of crossing over (X) are indicated.

In general, the strategy consisted of assembling three JEV cDNA fragments, fragments L, R, and M, into a pRS313 shuttle vector through homologous recombination (Fig. 4). Since ECP9 is located in the-well conserved 5′ noncoding region of the JEV genome, introduction of mutations into ECP9 may impair virus replication. Thus, we decided to construct three types of JEV mutations by introducing mutations into ECP9 or ECP10 alone or into both ECP9 and ECP10. To introduce those three types of silent mutations within the L fragment, site-directed mutagenesis was performed to generate the L* fragment (Fig. 4H). Yeast colonies that harbored the assembled full-length JEV cDNA clones containing the L fragment grew much more slowly than those containing the L* fragment. Yeast colonies containing pRS/JEVFL (L* or L) were chosen, amplified, and transformed into bacteria. After several attempts, we were not able to obtain the correct full-length JEV cDNA clones containing the L fragment from E. coli. This phenomenon was similar to the construction of the full-length DENV2 cDNA clones with no mutations within ECPs of the DENV2 genome. We also observed that C41 cells harboring the full-length JEV cDNA clone with silent mutations within ECP10 alone were not able to propagate. In contrast, the full-length JEV cDNA clones with silent mutations either at ECP9 alone or at both ECP9 and -10, designated JEV-A90C and JEV-DM, respectively, were easily obtained, ampli- fied, and purified from C41 cells. After restriction digestion pattern and sequencing analyses, most colonies carrying the JEV-A90C and JEV-DM plasmids were found to be correct and to have decent DNA yields, although the DNA yields of the JEV-DM cDNA clone were slightly better than those of the JEV-A90C cDNA clone. DNA sequencing analyses revealed that all JEV-A90C and JEV-DM cDNA clones had the same nucleotide sequences except that the JEV-DM clone had additional mutations within ECP10 of the JEV genome.

The stability of cloned JEV-A90C infectious cDNAs was assessed by repeated passage in E. coli strain C41. C41 was transformed with JEV-A90C, and the grown colonies were selected on a 2YT agar plate containing ampicillin (25 μg/ml). Four colonies were picked, mixed, and restreaked on a 2YT agar plate. The passage cycle for growing C41 cells harboring the JEV-A90C plasmid was repeated five times. Four clones containing the JEV-A90C plasmid were tested, and the integrity was confirmed by using two sets of restriction enzyme digestion to make sure that no transposon insertion and DNA rearrangement occurred during the passages. The integrity of JEV-A90C could be maintained for at least five passages of cultures on agar plates.

To show that the JEV-A90C plasmid can be assembled and easily manipulated in bacteria as shown for the DNA manipulation of the DENV2-6M plasmid in bacteria, we wished to know if the poison region (core-prM-E-NS1) containing A90C silent mutation could be cloned into the plasmid by direct cloning in bacteria. Previously, we failed to obtain correct full-length JEV cDNA clones by direct cloning in bacteria. We often obtained JEV cDNA clones with a 1-nt deletion, e.g., JEV-dl2201 (with a 1-nt deletion at nt 2201). We digested the JEV-dl2201 clone with BssHII (∼1 kb upstream of the SP6 promoter sequence) and BspEI (nt 3446 of the JEV genome) restriction enzymes to remove the region from nt 1 to 3446 to serve as a vector to assemble the corresponding DNA fragment (nt 1 to 3446) with an A90C mutation. As expected, we consistently obtained two correct JEV-A90C clones out of six colonies, which were verified by DNA sequencing analyses. Thus, we can obtain the JEV-A90C infectious cDNA clone by assembling in either yeast or bacterial cells.

Silent mutations introduced into the central portion of the DENV2 core gene affect the infectivity of DENV2.

To determine if full-length DENV2 cDNA clones harboring different sets of silent mutations are infectious, full-length DENV2 cDNA clones were used as templates to produce full-length synthetic RNAs by in vitro transcription and transfected into BHK21 cells. There was substantial cytopathic effect (CPE) as a result of efficient DENV2 viral infection 3 days after transfection into BHK21 cells with RNA transcripts derived from the DENV2-6M and DENV2-7M-2 cDNA clones with silent mutations introduced into ECP3 to -8 and ECP2 to -8, respectively. There was no apparent CPE in the BHK21 cells even 5 days after transfection with RNA transcripts derived from the DENV2-8M and DENV2-7M-1 cDNA clones with silent mutations within ECP1 to -8 and ECP1 and 3 to -8, respectively.

To further confirm the infectivities of the various DENV2 mutant viruses derived from DENV2-6M, DENV2-7M-1, DENV2-7M-2, and DENV2-8M cDNA clones, an indirect immunofluorescence assay (IFA) was used to detect cells expressing DENV2 prM protein in transfected BHK21 cells. DENV2 prM proteins could be detected in cells at 2 days after transfection with in vitro-transcribed DENV2-6M or DENV2-7M-2 RNA (Fig. 5 A). More cells with strong fluorescence staining were observed from cells transfected with DENV2-6M RNA transcripts than from those transfected with DENV2-7M-2 RNA transcripts at 3 days after transfection. There was an apparent decrease of the cell population with strong IFA staining in the cells transfected with DENV2-6M transcripts at 4 days after transcription due to cell death resulting from CPE caused by virus infection. Almost every cell was immunopositive on day 5 after transfection with DENV2-6M or DENV2-7M-2 RNA transcripts, although a more severe CPE (30% of BHK21 cells remaining) was seen in cells transfected with DENV2-6M than in those transfected with DENV2-7M-2. Therefore, transcripts were efficiently expressed in the cells transfected with DENV2-6M or DENV2-7M-2 synthetic RNA, and the virus progeny were able to efficiently replicate and infect cells. In contrast, DENV2 prM was detected in more than 50% of BHK21 cells only at 4 days after transfection with DENV2-7M or DENV2-8M RNA transcripts. Strong immunostaining with prM antibody occurred 5 days after transfection with DENV2-7M or DENV2-8M RNA transcripts.

FIG. 5.

DENV2 infectivity is affected by the silent mutations inserted into the central region of the core gene within the DENV2 genome. (A) Immunofluorescence analysis of cells transfected with various DENV2 RNA transcripts. BHK21 cells were transfected in vitro with transcripts derived from a DENV2-6M, DENV2-7M-1, DENV2-7M-2, or DENV2-8M plasmid or were mock transfected as a control. Monoclonal antibody against the DENV2 prM antigen was used to detect the infected cells by indirect immunofluorescence at 2, 3, 4, and 5 days after transfection with in vitro-derived transcripts. (B) Determination of titers of virions from cells transfected with various DENV2 mutant RNA transcripts. BHK21 cells were transfected in vitro with transcripts derived from a DENV2-6M, DENV2-7M-1, DENV2-7M-2, or DENV2-8M plasmid. Plaque assay was performed to determine the virus titer derived from the supernatants of BHK21 cells transfected with in vitro-derived transcripts at 1, 2, 3, 4, and 5 days after transfection. The error bars represent the SEMs from three independent experiments.

To further quantitatively evaluate the difference in virus spreading activity between various full-length DENV2 mutant cDNA clones harboring silent mutations, the titers of virions synthesized from BHK21 cells transfected with various DENV2 mutant RNAs were measured by plaque formation assay. The titers of virions derived from the DENV2-6M cDNA clone were highest (around 1 × 10⁶ PFU/ml) on day 3 after transfection (Fig. 5B). There was a dramatic drop in the titers (from 1 × 10⁶ to 1 × 10⁴ PFU/ml) of virions derived from the DENV2-6M cDNA clone at 5 days after transfection, which is consistent with the observation that only around 30% of BHK21 cells were alive at 5 days after transfection with DENV2-6M RNA transcripts. A delay in reaching virus titers of 1 × 10⁶ PFU/ml occurred in the cells at 5 days after transfection with DENV2-7M-2 RNA transcripts. The virions derived from DENV2-7M or DENV2-8M slowly reached virus titers of around 1 × 10⁴ PFU/ml at 5 days after transfection.

A cis-acting element essential for virus replication of DENV2 is located in the central portion of the DENV2 core gene.

Since the silent mutations introduced into the DENV2 core gene affected spreading of viruses derived from DENV2-7M-1, DENV2-7M-2, and DENV2-8M RNA transcripts (Fig. 5), we decided to utilize subgenomic dengue virus replicons to further determine the effect of those silent mutations on the translation or replication capacity of dengue virus replicon RNAs in BHK21 cells. The same silent mutations introduced into ECP1 or/and ECP2 within the core gene of the full-length DENV2 cDNA clone were constructed in the core gene of a wild-type subgenomic dengue virus replicon with a reporter luciferase gene, and these constructs were designated mECP1, mECP2, and mECP1+2 (Fig. 6 A). We transfected BHK21 cells with wild-type or various DENV2 mutant replicons RNAs (mECP1, mECP2, and mECP1+2) (Fig. 6A) and monitored the luciferase signal at different time points. The luciferase activities of the wild-type and various mutant DENV2 replicons were similar at 4, 8, and 12 h posttransfection, indicating that the mutations within ECP1, -2, and -1+2 apparently did not affect the translation efficiency of viral RNAs (Fig. 6B). However, the mECP1 and mECP1+2 replicons showed lower luciferase activity at 36 h posttransfection than the wild-type and mECP2 replicons. The dramatic difference in luciferase activity between the wild-type dengue virus replicon and mECP1 or mECP1+2 was observed after 48 h posttransfection. A slight reduction of luciferase activity was found in the mECP2 replicon compared to wild-type dengue replicon after 48 h posttransfection.

FIG. 6.

DENV2 replicon activity is affected by the silent mutations inserted into the central region of the core gene within the DENV2 genome. (A) Schematic diagram of the DENV2 reporter replicon. The 5′ UTR (black line), the N-terminal 102 amino acids of the C protein (C102), the Renilla luciferase gene (Rluc), the FMDV2A cleavage site (black box), a neomycin resistance gene (Neo), an encephalomyocarditis virus (EMCV) internal ribosome entry site (IRES) element (gray box), the C-terminal 24 amino acids of E (E24), the entire NS regions (NS1∼NS5), and the 3′ UTR (black line) are indicated. The replicon was used to quantify the effects of silent mutations on viral RNA replication of DENV2. (B) Kinetics of transient expression of wild-type and various dengue virus mutant (mECP1, mECP2, and mECP1+2) replicons in BHK21 cells. The luciferase activity in cytoplasmic extracts prepared from BHK21 cells transfected with wild-type or various dengue virus mutant replicon RNAs was measured at different time points (4, 8, 12, 24, 36, 48, 60, 72, and 96 h). (C) Quantitation of various dengue virus mutant replicon RNAs by real time RT-PCR. Positive-strand viral RNA copy numbers at 48 and 72 h after transfection of BHK21 cells with wild-type or various dengue virus mutant replicon RNAs are shown. The error bars represent the SEMs from three independent experiments (n = 3). *, P ≤ 0.05; **, P ≤ 0.01.

To further determine if the mutations in the dengue virus core gene affect the level of viral RNAs of the dengue virus replicon, the viral RNAs were measured by RT-PCR derived from cells transfected with wild-type or various dengue mutant replicons at either 48 or 72 h posttransfection. Consistent with the results from the reporter luciferase activity, there was also a significant reduction in the level of positive-stranded viral RNAs in mECP1 and mECP1+2 dengue virus replicons compared to the wild-type dengue replicon. There was a slight reduction in the level of positive-stranded viral RNAs derived from the mECP2 virus dengue replicon compared to that for the wild-type dengue virus replicon at either 48 or 72 h posttransfection.

Virions derived from the DENV2-6M cDNA clone exhibit a plaque morphology and growth curve similar to those for virions derived from the DENV2 parental virus.

Since the silent mutations introduced into the central portion of the core gene caused a slower replication phenotype in the spreading of DENV2-7M-1-, DENV2-7M-2-, and DENV2-8M cDNA-derived viruses, we decided to determine if the virions derived from DENV2-6M cDNA clone (with six sets of silent mutations within ECP3 to -8) possess virus spreading activity similar to that of the parental DENV2 virus stock. First, we compared the plaque morphology of viruses derived from DENV2-6M to that of parental DENV2 virus stocks. As shown in Fig. 7 A, the plaque size in BHK21 cells infected with DENV2-6M virus is quite similar to that in cells infected with parental DENV2 viruses. To evaluate the effect of the silent mutations introduced into ECP3 to -8 of the DENV2 genome on DENV2 growth, the growth kinetics of synthetic DENV2 viruses in BHK21 cells were analyzed and compared to those of DENV2 parental virus stocks. BHK21 cells were infected at the same MOI (either 0.01 or 0.1) with DENV2 viruses derived from the DENV2-6M cDNA clone and DENV2 parental virus stocks. The supernatants of infected BHK21 cells were removed daily, and the titers of viruses were subsequently determined. The growth curves of viruses derived from the DENV2-6M cDNA clone and DENV2 parental viruses were indistinguishable in BHK21 cells (Fig. 7B). To further determine if the grow curve of viruses derived from DENV2-6M in BHK21 cells is similar to that in mosquito cells (C6/36 cells), the growth kinetics of synthetic DENV2 viruses in C6/36 cells were analyzed and compared to those of DENV2 parental virus stocks. C6/36 cells were infected at the same MOI (either 0.01 or 0.1) with DENV2 viruses derived from the DENV2-6M cDNA clone and DENV2 parental virus stocks. The supernatants of infected C6/36 cells were removed daily, and the titers of viruses were subsequently determined. The growth curves of viruses derived from the DENV2-6M cDNA clone and DENV2 parental viruses were indistinguishable (Fig. 7C). These results suggest that the promoter-silencing mutations introduced into ECP3 to -8 of the DENV2 genome did not perturb the normal functions of the DENV2 proteins and that the cloned full-length DENV2 cDNA retained its integrity in our cloning strategy.

FIG. 7.

Plaque morphologies and growth curves of DENV2-6M transcript-derived viruses are similar to those of parental DENV2 stocks. (A) Plaque morphologies of viruses derived from DENV2-6M transcripts and parental DENV2. BHK21 cells were infected with the parental DENV2 PL046 strain viruses or infected with DENV2-6M transcript-derived viruses, overlaid with methylcellulose, and stained for 5 days with crystal violet. (B) Growth kinetics of virions derived from DENV2-6M RNA transcripts and parental DENV2 in BHK21 cells. BHK21 cells were infected with parental DENV2 virus stock or with DENV2-6M transcript-derived viruses at MOIs of 0.01 and 0.1 PFU/cell. Virus samples from the medium of infected BHK21 cells were harvested daily, and the viral titer of each sample was determined in BHK21 cells. (C) Growth kinetics of virions derived from DENV2-6M RNA transcripts and parental DENV2 in mosquito cells (C6/36 cells). C6/36 cells were infected with parental DENV2 virus stock or with DENV2-6M transcript-derived DENV2 at MOIs of 0.01 and 0.1 PFU/cell. Virus samples from the medium of infected C6/36 cells were harvested daily, and the viral titer of each sample was determined in BHK21 cells. The error bars represent the SEMs from three independent experiments.

Virions derived from the JEV-A90C cDNA clone exhibit plaque morphology and growth curve phenotypes similar to those of the JEV parental virus.

Since it has been shown that an A-to-C mutation at nt 90 of the JEV genome reduced the promoter activity of ECP9 and stabilized the JEV-A90C plasmid in bacteria (Fig. 2B and C), we sought to evaluate if the JEV-A90C cDNA clone is an infectious cDNA clone and if recombinant JEV-A90C viruses have infectivity similar to that of parental JEV. First, the virus infectivity of recombinant JEV-A90C virions was determined by IFA staining to detect JEV envelope protein expression in BHK21 cells transfected with JEV-A90A RNA transcripts. IFA staining of JEV E protein was detected in transfected BHK21 cells at 2 days after transfection (Fig. 8 A). Strong IFA staining and severe CPE were observed in transfected BHK21 cells at 3 days after transfection. Next, the plaque-forming ability was also used to compare the virus replication capacities of parental JEV and recombinant JEV-A90C. Similar plaque morphology was shown in BHK21 cells infected with either recombinant JEV-A90C or parental JEV (Fig. 8B). The growth kinetics of parental JEV or recombinant JEV-A90C were also compared. As shown in Fig. 8C, there was no apparent difference in the growth curves of virus replication in BHK21 cells infected with parental JEV and recombinant JEV-A90C at an MOI of 0.01 or 0.1. Similarly, there was no apparent difference in the growth curves of virus replication in C6/36 cells infected with parental JEV and recombinant JEV-A90C at a MOI of 0.01 or 0.1 (Fig. 8D).

FIG. 8.

Plaque morphologies and growth curves of JEV-A90C transcript-derived viruses are similar to those of parental JEV stocks. (A) Immunofluorescence analysis of cells transfected with JEV-A90C RNA transcripts. BHK21 cells were transfected in vitro with transcripts derived from JEV-A90C or mock transfected as a control. Monoclonal antibody against the JEV envelope antigen was used to detect the infected cells by indirect immunofluorescence at 1, 2, 3, and 4 days after transfection with in vitro-derived transcripts. (B) Plaque morphology of virus derived from JEV-A90C transcripts and parental JEV. BHK21 cells were infected with the parental JEV RP9 strain viruses or JEV-A90C transcript-derived viruses, overlaid with methylcellulose, and stained at 3 days postinfection with crystal violet. (C) Growth kinetics of virions derived from JEV-A90C RNA transcripts and parental JEV in BHK21 cells. BHK21 cells were infected with parental JEV stock or with JEV transcript-derived viruses at MOIs of 0.01 and 0.1 PFU/cell. Virus samples from the medium of infected BHK21 cells were harvested daily, and the viral titer of each sample was determined in BHK21 cells. (D) Growth kinetics of virions derived from JEV-A90C RNA transcripts and parental JEV in mosquito cells (C6/36 cells). C6/36 cells were infected with parental JEV virus stock or with JEV-A90C transcript-derived JEV at MOIs of 0.01 and 0.1 PFU/cell. Virus samples from the medium of infected C6/36 cells were harvested daily, and the viral titer of each sample was determined in BHK21 cells. The error bars represent the SEMs from three independent experiments.

DISCUSSION

The novel methods described here are convenient approaches for cloning full-length DENV2 and JEV cDNA clones with little toxicity in E. coli. For years, attempts have been made to develop a feasible method for constructing full-length flavivirus cDNA clones, which are essential to a successful reverse genetics system. Most previous attempts encountered the intrinsic toxicity of the flavivirus genome sequences in E. coli. We first demonstrated that the instability of the DENV2 and JEV cDNA clones in bacteria is due to the toxicity of the cryptic expression of viral proteins by multiple ECP sequences that are embedded in the viral genome. Furthermore, we showed that silent mutations, which were introduced to reduce ECP activities, stabilized the full-length DENV2 and JEV cDNA clones in E. coli. The infectious virions were efficiently produced in cells transfected with in vitro-transcribed RNAs derived from the full-length DENV2 and JEV cDNA clones with mutations. Interestingly, a cis-acting element essential for DENV2 replication was discovered and located in the central portion (nt 160 to 243) of the core gene within DENV2 genome. Thus, the methods described in this paper provide a feasible way to construct full-length flavivirus cDNA clones. Our methodology should greatly facilitate the study of the molecular mechanism of flavivirus replication and pathogenesis, as well as future vaccine development.

We primarily used the homologous recombination cloning approach, or recombineering, in yeast cells to construct unstable full-length DENV2 and JEV cDNA clones because recombineering not only has a much higher cloning efficiency but also is more tolerant than bacteria to the toxicity of poisonous sequences. Yeast cells can easily assemble full-length DENV2 cDNA clones through homologous recombination, without the tedious conventional cloning process. Unlike the vulnerability of bacteria to the poisonous flavivirus sequences, yeast cells serve as perfect hosts for cloning DENV2 (NGC strain) (51) and DENV1 (Western Pacific strain) (52) infectious cDNA clones because they are more tolerant than bacteria to the toxicity of DENV cDNA. The full-length DENV2 (NGC strain) infectious clone was successfully obtained from E. coli, but the DENV2 cDNA toxicity still persisted, resulting in slow bacterial growth. Using a similar strategy, we were able to construct and obtain full-length DENV2 (PL046 strain) cDNA clones from yeast cells without any modification of the DENV2 genome. However, we were not able to amplify full-length DENV2 (PL046 strain) cDNA plasmids in bacteria, indicating that yeast cells can endure the toxicity of DENV2 cDNA as they presumably promote less cryptic viral protein expression than bacteria. We suspect that our failure to obtain the full-length DENV2 (PL046 strain) cDNA clone from E. coli may have stemmed from differences in the genomic sequences between the DENV2 NGC and PL046 strains. Furthermore, we observed that yeast cells are able to carry full-length DENV2 cDNA plasmids from another strain (DENV2 16681), although they grew very slowly (data not shown). Like for the DENV2 PL046 strain, we were not able to recover full-length DENV2 (16681 strain) cDNA clones in E. coli. This observation is consistent with the speculation that various degrees of toxicity are derived from sequence variations among the different DENV2 strains (NGC, PL046, and 16681). This speculation may also explain why infectious cDNA clones were obtained only for certain strains of DENV2. Interestingly, unlike construction of the DENV2 cDNA clone, assembling the full-length JEV (RP9 strain) cDNA clone in yeast cells was not successful when no silent mutations were introduced into the JEV genome. Full-length JEV cDNA clones (JEV-A90C and JEV-DM) were obtained from yeast and bacteria when the silent mutations were introduced into ECP9 and ECP9+10 of the JEV genome, respectively (Table 1). This clearly suggests that yeast cells were not able to circumvent the JEV cDNA toxicity that results from inappropriate expression of JEV cDNAs, even though they were able to tolerate the toxicity due to DENV2 (NGC, PL046, and 16681 strain) cDNAs.

Several previous reports have speculated that cryptic expression of viral proteins occurs in bacteria harboring flavivirus cDNA. The aberrant expression of flavivirus proteins in bacteria is thought to be responsible for their toxicity in E. coli. First, insertions and deletions are usually found and mapped to the E-NS1-NS2A region during the construction of a DEN4 infectious clone, which may be due to the secondary structure or to the adventitious expression of some toxic product (51). Furthermore, the introduction of a stop codon at residue 147 of the NS1 protein stabilizes DEN3/4 and DENV1/4 chimera infectious cDNAs, which clearly indicates that the toxicity is derived from the spurious expression of dengue virus structural genes (core-prM-E-NS1) (5). Moreover, spontaneous nonsense mutations often occur during cloning of the JEV 5′ half plasmid (57, 65). It is likely that spurious transcription is initiated from a prokaryotic promoter-like sequence that is located somewhere in the JEV E gene, as nonsense mutations (e.g., a stop codon) prevent the expression of the problem genome region at the level of translation (65). Two of our own observations also support this speculation. One observation is that deletion (either a large deletion or a single-nucleotide deletion), insertion (transposon insertion), and premature stop codons frequently occurred in the region of the DENV2 (PL046 strain) and JEV (RP9 strain) structural genes when the full-length DENV2 or JEV cDNA clone was constructed without any silent mutations introduced into ECP1 to -10 as shown in Table 1 (data not shown). Taken together, our observations are consistent with several previous findings (5, 51, 57, 66) demonstrating that the well-known toxicity of flavivirus cDNAs may arise from inappropriate expression of the flavivirus genome in bacteria.

An important phenomenon disclosed by this report is that the cryptic expression of the DENV2 and JEV viral proteins, which accounted for the instability, was initiated from multiple ECPs within the DENV2 and JEV genomes. Active ECPs within the DENV2 and JEV genomes were demonstrated by viral sequence-driven DENV2- or JEV-luciferase fusion protein expression and luciferase enzyme activities (Fig. 1 and 2). The activity of ECPs in bacteria was further evidenced by the fact that promoter-silencing mutations could reduce DENV2 and JEV cDNA fragment reporter expression in bacteria (Fig. 1 and 2). The observation that DENV2-6M (silent mutations within ECP3 to -8) and DENV2-8M (silent mutations within ECP1 to -8) cDNA clones were consistently able to be constructed in bacteria supports the idea that the instability of DENV2 cDNA clones in bacteria results from multiple active ECPs within the DENV2 genome. In contrast, four sets of silent mutations introduced either into ECP1, -2, -7, and -8 or into ECP3 to -6 of the DENV2 genome were not able to circumvent the instability of full-length DENV2 cDNA in bacteria (data not shown), supporting the toxicity of DENV2 cDNA resulting from multiple active ECPs in bacteria. Another line of evidence supporting that more silent mutations inserted into ECP1 and/or ECP2 of DENV2-6M infectious cDNA increase its plasmid stability comes from the result that bacterial colony numbers for DENV2-7M-1, -7M-2, and -8M were better than those for DENV2-6M. Similarly, ECP9 to -12 within the JEV genome were shown to be active in bacteria (Fig. 2A). As expected, we were not able to construct a full-length JEV cDNA clone without inserting silent mutations into the JEV genome. However, full-length JEV cDNA clones, JEV-A90C and JEV-DM, were consistently obtained when the silent mutations were inserted into ECP9 alone and ECP9+10 of the JEV genome, respectively. Insertion of silent mutations at ECP10 alone within JEV genome was not able to stabilize the stability of JEV cDNA in bacteria, although silent mutations inserted into ECP10 of JEV-Luc(1-150) significantly reduced promoter activity in bacteria (Fig. 2B). The results clearly indicated that ECP9 is the key element responsible for the instability of full-length JEV cDNA in bacteria, although the promoter activity of ECP10 to -12 can be detected in bacteria. The other line of evidence supporting the multiple-ECP viral protein expression of DENV2 or JEV cDNA in E. coli came from our successful cloning of full-length DENV1 and DENV3 cDNAs using the same rationale by introducing multiple silent mutations into the predicted ECPs (different from the ECPs within the DENV2 genome) within the DENV1 and DENV3 genomes (nt 1 to 3000) (unpublished results). Our results strongly suggest that multiple, active E. coli promoters exist within nt 1 to 3000 of the DENV1, -2, and -3 and JEV genomes, although the ECPs among DENV1, -2, and -3 are all different. Therefore, the toxicity of the DENV1, -2, and -3 and JEV cDNAs in bacteria is very likely due to the cryptic viral protein expression that is initiated from multiple ECPs of the DENV and JEV genomes in bacteria.

The involvement of ECPs within the DENV2 and JEV genomes in the toxicity to E. coli occurs not only in flavivirus cDNAs but also in other nonprokaryotic genes. It has been noted that some nonprokaryotic genes are difficult to clone into plasmids, presumably due to the toxicity derived from those genes. This includes the human cystic fibrosis transmembrane conductance regulator (hCTFR) (21), human growth hormone (hGH) receptor (3), and potato virus X (25) genes. In these cases, the cryptic promoter activity of hCTFR or potato virus X in E. coli can be detected either by the chloramphenicol acetyltransferase (CAT) reporter assay or by Northern blotting. The instability and toxicity of hCTFR, hGH receptor, and potato virus X full-length clones in E. coli are greatly reduced by site-directed mutagenesis of one cryptic promoter sequence within those full-length clones, which decreases the prokaryotic promoter activity. The reports described above are in agreement with our results (Fig. 1 and 2) that the stability of the DENV2 and JEV cDNA clones is increased by reducing the activity of multiple prokaryotic promoters in bacteria. Interestingly, most of the toxic genes mentioned above encode membrane proteins. The toxic sequences within the flavivirus cDNAs are located mostly in the viral gene region that encodes membrane proteins, i.e., core-prM-E-NS1. Therefore, it follows that our novel method for cloning DNEV2 and JEV cDNAs may provide a convenient way to construct toxic genes encoding membrane proteins in E. coli.

The introduction of silent mutations into ECP3 to -8 of the DENV2 genome increased the stability of the DENV2 cDNA clone in bacteria and did not apparently affect the transmission of virus from cells transfected with RNA transcripts derived from DENV2-6M cDNA clones (Fig. 5A and 7). IFA staining was detected very rapidly (2 days after transfection) and virus titers reached around 1 × 10⁶ PFU/ml (3 days after transfection) in BHK21 cells transfected with DENV2-6M RNA transcripts. The virions derived from DENV2-6M RNA transcripts replicated as efficiently as parental virus stocks in either C6/36 cells or BHK21 cells (Fig. 7A and B), which strongly indicates that the silent mutations within ECP3 to -8 did not affect virus replication. In contrast, a delay in the IFA staining and reduction of virus titers (10- to 100-fold reduction) was observed in BHK21 cells transfected with DENV2-7M-1 or DENV2-7M-2 RNA transcripts (Fig. 5), which contain additional silent mutations either within ECP1 or ECP2 of DENV2-6M RNA transcripts, respectively. This suggested that the silent mutations introduced into ECP1 (nt 160 to 205) or ECP2 (nt 198 to 243) located in the dengue virus core gene, but not ECP3 to -8 within prM-E-NS1 of the DENV2 genome, affect virus replication. DENV2 reporter replicon experiments further showed that the mutations in either ECP1 or ECP2 reduce the viral replication of the dengue virus replicon but have no apparent effect on the translation of viral RNAs (Fig. 6A and B). We suspected that the silent mutations introduced into either ECP1 or ECP2 within the middle region of the DENV2 core gene may interfere with DENV2 replication by disrupting the cis-acting element located at the core-coding region essential for virus replication. It has been reported that a core hairpin (cHP) (nt 114 to 134) within the N terminus of the dengue virus core-coding region serves as a cis-acting element essential for virus replication (15, 16). The hairpin structure cHP affects the virus life cycle by enhancing recognition of the 5′ core start codon (16) and efficient RNA synthesis (15). It was proposed that cHP may stabilize the overall 5′-3′ panhandle structure or participate in recruitment of factors associated with the replicase machinery (15). The other element, the 5′ downstream AUG region (nt 100 to 115), located at the N terminus of the core-coding region was also found to affect virus replication by functioning as a cis-acting element that possibly is involved in genome circularization (19). Taking the findings together, we uncovered a cis-acting element located in the central portion (nt 160 to 243) of the core gene, which extends our understanding of the role of the DENV2 core gene serving a cis-acting element essential for virus replication. The detailed mechanism by which nt 160 to 243 of the core gene serve as a cis element is needed to investigate whether the region interacts with the 3′ end of the dengue virus genome and participates in genome circularization.

A full-length JEV cDNA was consistently obtained when site-directed mutagenesis was performed to introduce a mutation (A to C) at nt 90 within ECP9 of the 5′ untranslated region (5′ UTR) of the JEV genome, which reduced the toxicity resulting from cryptic expression of JEV proteins in bacteria (Fig. 2). Severe CPE was observed in BHK21 cells at 3 days after transfection with JEV-A90C RNA transcripts (Fig. 8A), suggesting that the JEV-A90C cDNA clone is infectious. Since the A90C mutation is located at very well conserved 5′ UTR of the JEV genome, there is a concern that the produced virions derived from JEV-A90C RNA transcripts may have a reversion (C to A) at nt 90 in the virus genome. Amplified virions from virions synthesized from cells transfected with JEV-A90C RNA transcripts were subjected to RT-PCR, and the A90C mutation was found to be in the viral RNAs of the amplified virions (data not shown). Amplified JEV-A90C virions containing an A-to-C mutation at nt 90 in the JEV genome were shown to have infectivity similar to that of parental JEV viruses, demonstrating that a single mutation (A90C) made the manipulation of a notoriously unstable JEV infectious cDNA clone much more convenient and feasible.

The novel approach presented here will facilitate basic and applied viral research. First, our method for cloning DENV2 and JEV cDNA clones demonstrates that silent mutations in multiple ECPs stabilize the DENV2 and JEV cDNA clones in bacteria and explains for the first time the long-standing unsolved mystery of the toxic nature of flavivirus cDNAs in E. coli. This knowledge can be applied to other RNA viruses or any unstable cDNA. Second, an in-depth understanding of flavivirus virology at the molecular level has not been achieved due to difficulties in manipulating flavivirus infectious cDNA clones. Easy manipulation of stable, infectious cDNA clones will be immensely useful for providing insights into viral replication at the molecular level. Third, it is known that ease of handling infectious cDNA clones is important to provide a simple, reliable, and cost-effective method for maintaining viruses and vaccine depositories (37, 39). In particular, large-scale production of live attenuated vaccine cDNA clones can be easily obtained by our approach and used to test the efficacy of vaccine candidates. Our method will therefore facilitate vaccine development. Studies are in progress to improve our established methodology, including designing stable DNA-based infectious cDNA clones (eukaryotic promoter driven) and other ways to reduce the intrinsic toxicity of flavivirus cDNA and to stabilize flavivirus cDNA clones. Those improvements will further facilitate our manipulation of flavivirus infectious cDNA clones and broaden our understanding of flavivirus replication and pathogenesis.