A Novel in Vitro Replication System for Dengue Virus

Positive strand viral replicases are membrane-bound complexes of viral and host proteins. The mechanism of viral replication and the role of host proteins are not well understood. To understand this mechanism, a viral replicase assay that utilizes extracts from dengue virus-infected mosquito (C6/36) cells and exogenous viral RNA templates is reported in this study. The 5′- and 3′-terminal regions (TR) of the template RNAs contain the conserved elements including the complementary (cyclization) motifs and stem-loop structures. RNA synthesis in vitro requires both 5′- and 3′-TR present in the same template molecule or when the 5′-TR RNA was added in trans to the 3′-untranslated region (UTR) RNA. However, the 3′-UTR RNA alone is not active. RNA synthesis occurs by elongation of the 3′-end of the template RNA to yield predominantly a double-stranded hairpin-like RNA product, twice the size of the template RNA. These results suggest that an interaction between 5′- and 3′-TR of the viral RNA that modulates the 3′-UTR RNA structure is required for RNA synthesis by the viral replicase. The complementary cyclization motifs of the viral genome also seem to play an important role in this interaction.

triad of trypsine-like serine proteases (9,10). The serine protease domain of NS3, present within the N-terminal 180 amino acid residues, requires NS2B for activity (11)(12)(13)(14)(15)(16)(17)(18). The crystal structure of the NS3 protease domain has recently been reported (19). The C-terminal region of the NS3 protease domain contains conserved motifs found in the NTP-binding proteins and DEXH family of RNA helicases (20 -22). The nucleoside triphosphatase and RNA helicase activities of DEN2 NS3 have recently been demonstrated by expression and purification of recombinant NS3 in Escherichia coli (23). The RNA helicase activity is thought to be involved in unwinding of a doublestranded RNA replicative intermediate formed during replication of the flavivirus genome (8,24). NS5, the largest of the flaviviral proteins, contains conserved motifs found in many viral RNA-dependent RNA polymerases (RdRP) 1 (25,26), implicating a role of NS5 in viral replication.
Template-dependent and template-specific in vitro replication systems have been developed to study mechanisms of some plant and a few eukaryotic positive strand RNA viruses. These studies have revealed that viral replicases function as membrane-bound complexes of both viral and cellular proteins and viral RNA (Refs. 27-35 and for a review see Ref. 36). However, the mechanism of initiation of viral RNA synthesis as well as RNA genome replication in general and the role of host proteins in this process remain unclear for positive strand RNA viruses.
In flavivirus-infected cells, three RNA species have been detected: a genomic RNA of 40 -44 S; a double-stranded, RNase-resistant, replicative form of 20 -22 S; and a partially RNase-sensitive, replicative intermediate of 20 -28 S RNA (37)(38)(39)(40). The in vitro RdRP assays that have been developed to study flavivirus replication utilize membrane-bound complexes isolated from the infected cell lysates (37)(38)(39)(40). These studies have examined incorporation of radiolabeled nucleotides into the three RNA species on endogenous viral RNA templates. In these replication complexes, the endogenous replicative intermediates present in the infected cell extracts were shown to form a double-stranded replicative form. The replicative form produced endogenously from replicative intermediate or added exogenously gave rise to a 44 S genome-size RNA species, suggesting that initiation of (ϩ)-strand RNA synthesis occurred in these in vitro assays (39,40).
To study the mechanism of viral replication in molecular detail, it is crucial to develop an in vitro RdRP assay that can utilize exogenous RNA templates containing essential regulatory elements of the viral genome involved in viral replication. For initiation of (Ϫ)-and (ϩ)-strand RNA synthesis, conserved RNA sequences with intrinsic stem-loop structures from the 3Ј and 5Ј regions of many viral genomes are thought to play an important role in RNA replication (41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52)(53)(54)(55). In flavivirus genomes, such conserved elements have been noted previously; their functions in replication remain to be established. For example, the 3Ј-terminal 96 nucleotides of the mosquito-borne flaviviruses within the 3Ј-UTR form a conserved and stable stem-loop structure, although the primary sequence is not conserved (for a review see Ref. 8). Formation of such a stable secondary structure within the 3Ј-terminal 373 nt of the 3Ј-UTR was shown to exist in solution (56). A potential secondary structure with a lesser predicted stability near the 5Ј termini of several flavivirus RNAs including DEN2 was reported (57). There are also two short conserved sequences (CS1 and CS2) in 3Ј-UTR shared by all mosquito-borne flaviviruses; CS1, 26 nucleotides in length, is located 5Ј to the stem-loop structure of the 3Ј-UTR (3Ј-CS1). A portion of the 3Ј-CS1 is complementary to a conserved element is located within the N-terminal coding region of the capsid protein, C, in the 5Ј-TR of the viral genome (5Ј-CS). It has been proposed that these complementary sequences might play a role in cyclization of the viral genome (Ref. 58 and for a review see Ref. 8). However, cyclization of any of the flavivirus genome has not been observed to date. Thus the role of cyclization motifs in genome replication is unknown at present.
In this study, we report the development of the first in vitro RdRP assay that utilizes cell-free extracts of DEN2-infected mosquito (C6/36) cells and exogenous subgenomic RNA templates containing 5Ј-and/or 3Ј-terminal regions of the viral genome. The results indicate that there is an interaction between 5Ј-and 3Ј-terminal regions of the viral RNA that is required for RNA synthesis in vitro. This interaction is modulated by the complementary cyclization motifs. RNA synthesis occurs by the 3Ј-end elongation of the template RNA to yield a predominantly double-stranded RNA hairpin with a limited single-stranded loop region. The kinetics of the formation of the RNA hairpin product indicates that the template RNA is first modified yielding a RNase A-sensitive intermediate that is then converted to the hairpin product twice the size of the template RNA. This in vitro RdRP assay will be useful to study the sequence and protein requirements for RNA synthesis in vitro.

Preparation of DEN2 Viral Replicase Complex from the Infected Mosquito (C6/36) Cells
Dengue virus type 2 (New Guinea Strain C) was propagated as described (59,60). To obtain the DEN2 viral replication complex, C6/36 cells were infected with DEN2 virus (multiplicity of infection ϭ 5) in T-150 cm 2 flasks for 36 h at 28°C. Uninfected C6/36 cells were used as controls. Cells were harvested by centrifugation at 800 ϫ g for 10 min, and the cell pellet was resuspended in 0.5 ml of TNMg buffer (10 mM Tris-HCl, pH 8.0, 10 mM sodium acetate, 1.5 mM MgCl 2 ) per T-150 cm 2 flask. After passing the cell suspension 20 times through a syringe (one ml capacity), fitted with a 27-guage needle, the cell lysates were centrifuged at 800 ϫ g for 10 min at 4°C to fractionate the cytoplasmic and nuclear fractions as described (39). The protein concentrations of the cytoplasmic extracts were determined as described (61), and the extracts were stored in aliquots at Ϫ70°C until used.

Western Blot Analysis
An aliquot of the cytoplasmic extract was mixed with an equal volume of a 2ϫ sample loading buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 0.7 M ␤-mercaptoethanol, 10% glycerol, 0.05% bromphenol blue). Proteins were separated by SDS-PAGE (10%) and transferred electrophoretically to a polyvinylidene difluoride membrane. The membrane was treated with nonfat dry milk (5%) for 2 h in Tris-buffered saline containing 0.1% Tween-20, pH 7.5, and then incubated overnight with polyclonal rabbit anti-NS3 or anti-NS5 antibodies (1:2000). The membranes were washed three times for 20 min each with this buffer and incubated with horseradish peroxidase-labeled goat anti-rabbit anti-body (1:20000) for 2 h at room temperature. Chemiluminescence detection of the immunoreactive bands were performed as described by the manufacturer (ECL system; Amersham Pharmacia Biotech).

Plasmid Constructs
3Ј-UTR 373nt /wtCYC Plasmid-The p190-24 plasmid contains the 3Јterminal 373 nt from the DEN2 cDNA and 26 nt from the pGEM7Zf ϩ vector (TCAAGCTATGCATCCAACGCGTTGGG) under the control of SP6 promoter (56). When this plasmid is linearized with XbaI, the cDNA of (Ϫ)-strand polarity was extended by 4 additional nucleotides of the XbaI-5Ј overhang. The RNA synthesized from this template strand would include the 4 additional nucleotides complementary to this overhang at its 3Ј-end. To generate 3Ј-UTR 373nt RNA, which contains the wild type cyclization sequence but without these 4 additional nucleotides, two approaches were followed. Firstly, a PCR approach was followed using the two oligonucleotide primers: the upstream 5Ј-primer (A), 5Ј-TGACCATGATTACGCCAAGCTATTTAGGTG-3Ј anneals a region upstream of the SP6 promoter and a part of the promoter (underlined); the downstream 3Ј primer (B), 5Ј-AGAACCTGTTGATTCAA-CAGCACC-3Ј is complementary to the 3Ј-end of the viral genome. The PCR fragment was purified and used for in vitro transcription (see below). In the second approach, the XbaI-linearized p190-24 plasmid (10 g in each reaction) was treated with increasing amounts of S1 nuclease (0, 3, 10, 30, 75, and 150 units) in 100 l of reaction mixtures containing 50 mM sodium acetate, pH 4.5, 28 mM NaCl, and 4.5 mM ZnSO 4 at room temperature for 30Ј, extracted with phenol, and precipitated with ethanol. The S1 nuclease-treated plasmids were used for the in vitro transcription catalyzed by SP6 RNA polymerase. To ensure that the S1 nuclease precisely removed the XbaI overhang, the blunt ends were ligated by T4 DNA ligase and cloned. DNA sequencing was carried out using an automated DNA sequencer (Biotech Facility, University of Kansas Medical Center).
cDNA Constructs Encoding Subgenomic RNA Containing the 5Ј-and 3Ј-Terminal Regions of DEN2 Genome-A 230-nt DNA fragment from the 5Ј-terminal region that includes the 5Ј-UTR (96 nt), and the 5Ј-CS1 was generated by PCR using pMK4 as a template and the upstream primer (C), 5Ј-CGGAATTCGGATCGATCCCCCCTAATAC-3Ј (containing the T7 promoter), and the downstream primer (D) 5Ј-CAGTTCCT-GAGGTCCTCGTCCCTGCAG-3Ј (complementary to 217-233 nt of the viral genome, which includes the 5Ј-CS1 element). The primers, C and D, contain the EcoRI and Bsu36I sites (underlined), respectively. The PCR product was digested with EcoRI and Bsu36I and purified using a Qiagen cartridge. The 3Ј-terminal fragment (800 nt) that includes the 3Ј-UTR was derived from the plasmid clone pGEM-PCR1.3 (4) by digestion with Bsu36I (nt position 9885) and XbaI (nt position 10723). The pSP64 vector plasmid was digested at the multiple cloning site with EcoRI and XbaI. A three-fragment ligation and cloning yielded pSY-1 plasmid containing DEN2 cDNA sequences under the control of T7 promoter. To construct a plasmid encoding a shorter subgenomic RNA, the pSY-1 plasmid was partially digested with XmnI, followed by digestion with Bsu36I. The overhang from the Bsu36I-cut plasmid was blunt-ended by treatment with E. coli Klenow DNA polymerase fragment and was cloned to yield the pSY-2 plasmid (720 nt) in which 348 nt from the pSY-1 was deleted but retaining all of 3Ј-UTR region. The pSY-1 and pSY-2 plasmids were used to produce the subgenomic RNAs of 1.0 kilobase and 720 nt, respectively, by in vitro transcription (Fig. 1).
5Ј-TR/mutCYC and 3Ј-UTR 373nt /mutCYC Plasmids-To construct a PCR fragment of 5Ј-TR/mutCYC (which contains substitution mutations within the cyclization sequence), a PCR-based mutagenesis protocol was followed. Two PCR products were generated with overlapping sequences using two sets of primers. PCR1 was obtained using the 5Ј-primer (E), containing the EcoRI site, 5Ј-AGCTATGACCATGAT-TACGAATTC-3Ј that corresponds to the upstream region of the T7 promoter in the pSP64 vector and the 3Ј-primer (F), 5Ј-TTTCACAGA-GAGAGAAGGCGTATTTCTCGCCTTT-3Ј. PCR2 was produced using the 5Ј primer (G), 5Ј-GCCTTCTCTCTCTGTGAAACGCGAGAGAAAC-CG-3Ј, and the 3Ј primer (H), 5Ј-TGAGGTCCTCGTCCTG-3Ј. The underlined complementary sequences in F and G primers represent the mutated cyclization element (mutCYC). The primer H shares identical sequences with pSY-1 plasmid in the vicinity of Bsu36I site (underlined). The two products, PCR1 and PCR2, produced using the primer sets E/F and G/H were purified, mixed, and used for a third PCR in the presence of the primer set E/H. This final PCR product was purified and used for in vitro transcription to generate 5Ј-TR/mutCYC RNA.
To construct a PCR fragment containing the 3Ј-UTR 373nt /mutCYC that contains mutations within the cyclization sequence complementary to that in 5Ј-TR/mutCYC), a PCR-based mutagenesis protocol was followed. PCR1 was obtained using the 5Ј-primer (A), the 3Ј-primer (I), 5Ј-CAGCGCTCTCTCTGTGTTTTTTGTTTTGGGGGGG-3Ј. PCR2 was produced using 5Ј primer (J), 5Ј-AAACACAGAGAGAGCGCTGGGAAA-GACCAGAGAT-3Ј and the 3Ј-primer (B). The underlined complementary sequences in I and J primers represent mutated cyclization element (mutCYC). p190-24 plasmid was used as the template for PCR. The two products, PCR1 and PCR2, produced using the primer sets A/I and J/B were purified, mixed, and used for a third PCR in the presence of the primer set A/B. This final PCR product was purified and used for in vitro transcription to generate the 3Ј-UTR/mutCYC RNA.

Preparation of RNA Templates
The RNA templates used in the in vitro RdRP assays are shown in Fig. 1. To synthesize RNA templates containing 3Ј-terminal 373 nt (3Ј-UTR 373nt ) or the subgenomic RNA templates, the plasmid constructs, p190.24 (56) and pSY-2, respectively, were linearized with XbaI. The supercoiled plasmids were used as templates for PCR. The PCR products or the XbaI-linearized plasmids were used in the in vitro transcription reaction catalyzed by either SP6 (for the 3Ј-UTR 373nt RNA) or T7 RNA polymerase (for subgenomic RNAs) using conditions supplied by the manufacturer (Promega). Briefly, the in vitro transcription reactions were carried out at 37°C for 3 h in reaction mixtures (100 l) containing the 1ϫ buffer (40 mM Tris-HCl, pH 7.5, 6 mM MgCl 2 , 2 mM spermidine, 10 mM NaCl), 5 g of the DNA templates, a mixture of four rNTPs (0.5 mM each), 100 units of RNasin, 10 mM dithiothreitol, and 5 units of T7 or SP6 RNA polymerase. Then the reaction mixtures were digested with 5 units of RNase-free DNase I (Promega) at 37°C for 20 min to remove the DNA templates. The RNA transcripts were extracted with phenol at pH 4.5 and precipitated with ethanol. The quality of RNA products was analyzed electrophoretically and quantified spectrophotometically. Because of the 26 nt from the vector sequence at the 5Ј-end, the 3Ј-UTR 373nt RNA is actually 399 nt in length but is referred to in this study as 3Ј-UTR 373nt . Nonviral exogenous RNAs were prepared by linearizing three different vector plasmids (pSP64, pSP70, and pTM1) with six different restriction enzymes. RNAs were prepared by in vitro transcription and purified as described above.

In Vitro RdRP Assay
The in vitro RdRP assay was performed in 50-l reaction mixtures containing 50 mM Hepes, pH 7.3, 3 mM magnesium acetate, 6 M zinc acetate, 25 mM potassium acetate, 60 units of RNase inhibitor, 10 mM ␤-mercaptoethanol, 0.5 mM each ATP, GTP, and UTP, and 10 M of CTP mixed with 10 Ci of [␣-32 P]CTP (800 Ci/mmol), 0.1 mg/ml actinomycin D, 10 g of cytoplasmic extract from infected cells, and 5 g of an exogenous RNA template. As a control, a parallel reaction mixture containing all the components except the cytoplasmic extract from the uninfected C6/36 cell lysate (10 g) was included. Controls in which cytoplasmic extracts from infected or uninfected cells but without any exogenous RNA were also included. Reaction mixtures were incubated at 30°C for 1.5 h. To carry out the kinetics of RdRP reaction, seven separate RdRP reactions that included DEN2-infected cytoplasmic extracts and the subgenomic RNA template were initiated. At time intervals of 0, 5, 10, 15, 20, 30, and 40 min, reactions were terminated by the addition of EDTA (10 mM). The samples were extracted with phenol, pH 4.5, and precipitated by the addition of 0.1 volume of 3 M sodium acetate, pH 5.2, and 2.5 volumes of 100% ethanol in the presence of 5 g of yeast tRNA (Ambion). The RNA products were analyzed using either polyacrylamide, 7 M urea or by formaldehyde-agarose gel electrophoresis, followed by autoradiography.

Analysis of RdRP Products by RNase A Digestion
The RdRP products separated on PAGE, 7 M urea gels were cut out and eluted in 400 l of elution buffer (2 M ammonium acetate, 1% SDS, 1 mM EDTA, and 25 g/ml tRNA) either at 37°C for overnight or at 65°C for 4 h. After briefly spinning down gel slices, the supernatants were precipitated overnight with 1 ml of 100% ethanol at Ϫ20°C. The eluted products were treated with or without RNase A (Sigma; 5 ng/l) in 20 l of either 2ϫ SSC (300 mM NaCl and 30 mM sodium citrate, pH 7.2) or 0.1ϫ SSC (15 mM NaCl and 1.5 mM sodium citrate, pH 7.2) at 37°C for 30 min. The reactions were stopped by adding 30 l of TES stop buffer (10 mM Tris-HCl, pH 8.0, 50 mM EDTA, and 0.2% SDS), followed by phenol extraction and ethanol precipitation in the presence of 5 g of yeast tRNA. The RNase A-treated samples were analyzed on formaldehyde agarose gel, followed by autoradiography.

Sodium Periodate Oxidation of RNA
In vitro transcribed RNAs were dissolved in 200 l of 50 mM NaOAc, pH 5.0. 50 l of 0.1 M NaIO 4 (20 mM) was added and then incubated for 1 h at room temperature. Lysine (60 mM) was added to saturate the excess periodate and further incubated for 3 h at room temperature. The reactions were phenol extracted and precipitated with ethanol. The pellets were washed with 70% EtOH and dissolved in water and subsequently desalted with Bio-Gel P gel column (Bio-Rad). The periodatetreated RNAs were quantified spectrophotometrically and visualized on PAGE, 7 M urea gel by acridine orange.

Cytoplasmic Extracts from DEN2-infected Mosquito (C6/36) Cells Are Active in the Synthesis of RNA from the Exogenous
Templates-Previous studies showed that flavivirus RdRP activity was tightly associated with intracellular membranes in the cytoplasmic fractions of flavivirus-infected mammalian (monkey kidney cell line (Vero), or baby hamster kidney cell line (BHK-21)) cells (37-40, 62, 63). The earlier studies examined the incorporation of labeled nucleotides into the three RNA species that were synthesized from endogenous viral RNA templates. We sought to develop an in vitro RdRP assay that could utilize exogenous RNA templates that contain either 3Јor 5Ј-untranslated regions or both to determine the sequence requirements for RNA synthesis and to characterize the proteins that interact with these elements. We chose to use an established Aedes albopictus (C6/36) cell line for preparation of cytoplasmic extracts from DEN2-infected cells (see "Experimental Procedures") because the viral titers are significantly higher in C6/36 cells than in vertebrate cells (64).
We constructed cDNA clones containing the putative regulatory regions of the viral genome at the 5Ј-TR and 3Ј-UTR under the control of either the T7 or SP6 promoter as described under "Experimental Procedures." pSY-2 contains the subgenomic cDNA (720 nt) corresponding to both 5Ј-TR 230nt and 3Ј-UTR (full-length) of the genome linked together in the same molecule. The 5Ј-TR 230nt /wtCYC plasmid contains the cDNA corresponding to the 5Ј-terminal region (230 nt) of the genome; both cDNAs were downstream of the T7 promoter. The 3Ј-UTR 373nt /wt CYC plasmid contains the SP6 promoter along with 26 nt from the vector followed by the 3Ј-UTR 373nt . Plasmids containing mutations in the 5Ј-CYC or 3Ј-CYC motif within either the 5Ј-TR 230nt or the 3Ј-UTR 373nt were constructed as described under "Experimental Procedures" (Fig. 1).
First, we analyzed whether our cytoplasmic extracts contained the putative viral replicase components, NS3 and NS5, by Western blot analysis using rabbit polyclonal anti-NS3 and -NS5. Both NS3 and NS5 proteins were readily detected in the cytoplasmic extracts prepared from DEN2-infected cells but not from the uninfected cells (Fig. 2).
For the in vitro RdRP assay, exogenous RNAs were synthesized as run-off transcripts in the in vitro transcription reactions on either XbaI-linearized plasmids or the PCR DNAs as described under "Experimental Procedures." The 3Ј-UTR 373nt RNA synthesized from the XbaI-linearized plasmid contains 4 additional nucleotides complementary to the 5Ј overhang (5Ј-CTAG-3Ј) extending the authentic 3Ј-end of viral RNA (referred as 3Ј-UTR 373ϩ4nt RNA-XbaI). However, the RNA synthesized by the SP6 RNA polymerase on the PCR template would lack this 3Ј extension of 4 nt (3Ј-UTR 373nt RNA-PCR) and maintain the authentic 3Ј-end of viral genome. Both 3Ј-UTR RNAs include the highly conserved cyclization motif located from nucleotide positions 90 to 100 from the 3Ј-end (3Ј-wtCYC). The subgenomic RNA containing both 5Ј-TR and full-length 3Ј-UTR was synthesized as a run off transcript by T7 RNA polymerase either from the XbaI-linearized pSY-2plasmid (720 ϩ 4 nt) or the PCR product (720 nt). The subgenomic RNA includes the wild type cyclization motifs, which are complementary to each other (Fig. 1).
The RdRP reactions were carried out using the 3Ј-UTR 373ϩ4nt RNA or the 3Ј-UTR 373nt RNA as the exogenous RNA template and cytoplasmic lysates from either the DEN2infected or uninfected cells (Fig. 3A, lanes 2 and 3). Control RdRP assays lacking the exogenous RNA template were also included (Fig. 3A, lanes 4 and 5). The labeled RdRP products were analyzed on partially denaturing 4% polyacrylamide gels containing 7 M urea (PAGE, 7 M urea). The results of the in vitro RdRP assay shown in Fig. 3 indicate that both uninfected and DEN2-infected cell lysates gave rise to a labeled product of the same size as the input 3Ј-UTR 373ϩ4nt RNA (Fig. 3A, lanes 2 and 3, respectively, and also Fig. 3B, lane 1). However, DEN2infected cell lysates produced a distinctly labeled product that migrated faster than the size expected for the input 3Ј-UTR 373ϩ4nt RNA (Fig. 3, A, lane 2, and B, lane 1). In contrast, the RdRP assays carried out in the absence of an exogenous RNA template but in the presence of cytoplasmic extracts from infected or uninfected cells did not produce either of the products, indicating that these two products were formed only when the exogenous RNA template was added (Fig. 3A, compare  lanes 4 and 5 with lanes 2 and 3). Analysis of six nonviral RNA templates generated from vector plasmids as controls indicated that only two produced faster migrating RNA species like the 3Ј-UTR 373ϩ4nt RNA (data not shown). The labeled products with mobilities slower than the input exogenous RNA were also formed with cytoplasmic extracts from DEN2-infected cells (Fig. 3A, lanes 2 and 4). These are likely to be the products formed by the viral replicase utilizing the endogenous viral RNA templates as reported previously (37). However, these products did not fractionate into distinct species observed previously (37), probably because the gel system we used is selected to fractionate products formed from the small exogenous RNA templates.
However, the 3Ј-UTR 373nt RNA template generated from the PCR product did not produce the faster migrating species in the RdRP assay with DEN2-infected extracts (Fig. 3B, lane 2). The faster migrating species were also not formed when uninfected cell lysates were used with either of the 3Ј-UTR RNA templates (Fig. 3A, lane 3, and B, lanes 3 and 4). In contrast, the subgenomic RNA template, which contained the 4 additional nucleotides complementary to the XbaI-5Ј overhang or generated from the PCR DNA lacking these 4 nt at its 3Ј end, formed the faster migrating species (marked as 2ϫ) as well as the unitsized RNA product (marked as 1ϫ). The faster migrating RNA 1ϫ denotes the labeled RNA product which is of the same size as the input template RNA. 2ϫ denotes the RNA product with a faster mobility in this gel system. Later experiments showed that this RNA has a double-stranded hairpin-like structure with a size twice the size of input RNA. A, exogenous 3Ј-UTR 373nt RNA produced from the XbaI-linearized plasmid which would contain 4 additional nucleotides complementary to the XbaI overhang was used. Lane 1, labeled 3UTR 373ϩ4nt RNA as a marker; lanes 2 and 3, exogenous 3Ј-UTR 373ϩ4nt RNA added; lanes 4 and 5, no exogenous RNA added. B, the exogenous 3Ј-UTR 373nt RNA (lanes 1-4) or the subgenomic RNA (lanes 5-8) was synthesized from either XbaI-linearized plasmids (marked as XbaI in lanes 1, 3, 5, and 7) or from the PCR-generated DNAs (marked as PCR in lanes 2, 4, 6, and 8).
product was observed in the RdRP reaction that contained only the DEN2-infected cell lysates (Fig. 3B, lanes 5 and 6). These experiments have been repeated six times with consistent results.
Because both DEN2-infected and uninfected cell lysates produced RNA products of unit size (1ϫ) with the exogenous RNA templates, it is possible that this RNA was the product of 3Ј-terminal labeling by the host-encoded terminal nucleotidyl transferase in both cases. A second possibility is that de novo initiation by the DEN2-infected lysate at the 3Ј-end of template strand produced the (Ϫ)-strand RNA, which could migrate as 1ϫ in the partially denaturing gel electrophoresis system. On the other hand, the faster migrating RNA could be either a prematurely terminated (Ϫ)-strand product or an RNA that is highly compact in structure, such as a double-stranded RNA hairpin.
To characterize the labeled RNA products of the RdRP assays further, the faster migrating product formed with 3Ј-UTR 373ϩ4nt RNA template was eluted from the PAGE, 7 M urea gel and digested by RNase A under conditions of high ionic strength (see "Experimental Procedures"). The products of RNase A digestion were analyzed by electrophoresis on a formaldehyde-agarose gel (Fig. 4A). Under conditions of high ionic strength, the double-stranded regions in RNA are resistant to RNase A, whereas single-stranded regions are susceptible. The results showed that the faster migrating RNA product of 3Ј-UTR 373ϩ4nt RNA template had twice the size of the input RNA template (2ϫ) prior to RNase A digestion (Fig. 4A, lane 3) but migrated as a unit-sized RNA (1ϫ) after RNase A digestion under conditions of high ionic strength (Fig. 4A, lane 4). These results indicated that the faster migrating product is a doublestranded hairpin that had a single-stranded region susceptible to RNase A. The RNA size markers and an aliquot of the RdRP reaction mixture were loaded in lanes 1 and 2 of Fig. 4A, respectively. The faster migrating RNA products formed from the subgenomic RNA templates (generated from the PCR DNA and XbaI-linearized plasmid) also had twice the size (2ϫ) of input RNAs before RNase A digestion (Fig. 4B, lanes 3 and 7) but migrated as a unit-sized (1ϫ) product after RNase A digestion (lanes 4 and 8).
Next, we sought to analyze the 1ϫ product, which was formed in reactions containing lysates from either uninfected or DEN2-infected C6/36 cells. The in vitro assays using uninfected cell lysates with the 3Ј-UTR 373nt and the subgenomic RNA templates did not produce appreciable 1ϫ labeled product. The 1ϫ labeled RNA products formed from the 3Ј-UTR 373ϩ4nt and the subgenomic RNAs as templates with DEN2-infected cell lysates were isolated from the PAGE/urea gel and digested with RNase A. The results showed that the 1ϫ products formed with the subgenomic RNA templates synthesized from XbaI-linearized plasmid and the PCR DNA were resistant to RNase A under conditions of high ionic strength (Fig. 4B, lanes 2 and 6). This result suggested that the 1ϫ product contained the input template RNA and the complementary (Ϫ)-strand RNA synthesized in the RdRP reaction, which would anneal when isolated from the gel and become resistant to RNase A. These results did not distinguish between the possibilities that the 1ϫ product was formed by de novo initiation of RNA synthesis at the 3Ј-end of the template or arising from the digestion of the 2ϫ hairpin product by a nuclease (see below). However, the 1ϫ product formed with the 3Ј-UTR 373ϩ4nt RNA template and uninfected cell lysate (Fig.  3A, lane 3) was sensitive to RNase A under these conditions, suggesting that it is a single-stranded RNA (data not shown). This RNA product is likely to be formed by the action of the host terminal nucleotidyl transferase, which has been observed in earlier studies as a membrane-associated enzyme (38,65,66).
Kinetics of the Formation of RNA Products in the RdRP Assays-To determine the pathway for the formation of the double-stranded (2ϫ) hairpin product and whether the unitsized RNA was formed independent of the hairpin product, a time course of RdRP reaction was performed. The standard RdRP reactions were initiated and at the end of each indicated time point, EDTA was added (10 mM) to terminate the reaction. The RNA products formed at various time points were analyzed by formaldehyde-agarose gel electrophoresis prior to or after digestion with RNase A (Fig. 5). The results shown in Fig. 5A indicate that in 10 min of incubation time, the 1ϫ product was formed first at detectable levels that increased gradually with time. In contrast, the 2ϫ product was formed with a lag of 15 min, and the levels increased significantly over the time course of incubation. In addition, there were other RNA products formed starting at 10 min whose mobilities were between those of 1ϫ and 2ϫ products. In comparison, the results obtained after RNase A digestion of these time course samples (Fig. 5B) FIG. 4. Analysis of RNA products formed in vitro by RNase A digestion. The faster migrating (2ϫ) and the input template-sized RNA (1ϫ) products shown in Fig. 3 were eluted from the PAGE/urea gel and digested with RNase A under high salt conditions. The digested products and the controls were analyzed by formaldehyde-agarose gel as described under "Experimental Procedures." A, lane 1, RNA markers; lane 2, the products formed in the in vitro assay; lanes 3 and 4, eluted 2ϫ product before and after RNase A digestion, respectively. To detect the low amounts of radioactivity in lanes 3 and 4, the same autoradiogram was developed by longer exposure. A composite picture is shown. B, an in vitro assay was performed with subgenomic RNA (720 nt) synthesized from either XbaI-linearized or PCR-generated DNA, and the RNA products were analyzed by RNase A digestion under high salt conditions. 1ϫ and 2ϫ are as defined in Fig. 3.

FIG. 5.
Kinetics of RdRP activity with the exogenous subgenomic RNA template. A, seven RdRP reactions were initiated using DEN2-infected cell lysates and exogenous subgenomic RNA template as described under "Experimental Procedures." Each of the reactions was terminated by the addition of EDTA (10 mM). RNA products were purified by phenol extraction and ethanol precipitation. One-half of each sample was digested with RNase A. The untreated (A) and RNase A-treated (B) samples were analyzed by electrophoresis on two separate formaldehyde-agarose gels, followed by autoradiography. The first lane indicates the labeled marker subgenomic RNA (720 nt).
indicated that the 1ϫ product formed at 10 min of incubation was predominantly sensitive to RNase A, suggesting that this RNA species is likely to be the product of a terminal nucleotidyl transferase. However, the 2ϫ product formed between 15 and 40 min, upon digestion with RNase A, was converted predominantly to RNase A-resistant 1ϫ product. The RNA species of intermediate mobilities were essentially sensitive to RNase A, which yielded two small RNA fragments as seen first with the 10-min sample. These results revealed a pathway for the formation of the double-stranded hairpin product as follows. The RNA template is initially modified by the 3Ј-end elongation to an RNase A-sensitive product by either terminal nucleotidyl transferase or the viral replicase or both. This first intermediate is further elongated to a second intermediate (with a mobility between 1ϫ and 2ϫ), which is predominantly a RNase A-sensitive product but contains some double-stranded region to yield small RNase A-resistant products (Fig. 5B, 10 min lane). Further incubation of the reaction between 15 and 40 min results in a RNA product having a hairpin (2ϫ) structure. This RNA product has a predominantly RNase A-resistant, double-stranded region and a limited single-stranded loop region sensitive to RNase A. These results suggest that a small amount of the RNase A-resistant 1ϫ product, formed in a 90-min standard RdRP reaction with the subgenomic RNA template (Fig. 4B), originated from the double-stranded hairpin (2ϫ) product during the course of incubation. Figs. 3 and 4 suggested that self-priming at the 3Ј-end of the RNA template by viral RdRP activity to produce a hairpin-like structure was dependent on the nature of the RNA template used. The 3Ј-UTR 373ϩ4nt RNA was active, whereas the 3Ј-UTR 373nt generated from the PCR fragment was essentially inactive. This result was independently confirmed by removal of the XbaI 5Ј overhang from the linearized plasmid by digestion with S1 nuclease at two different concentrations (3 and 10 units/5 g). Sequence analysis at the 3Ј-end of the S1 nucleasetreated plasmids indicated that at 3 units of enzyme/5 g of plasmid DNA, the 3Ј-UTR 373nt cDNA end corresponded to the authentic 3Ј terminus, whereas at 10 units of the enzyme/5 g of DNA, it had lost three additional nucleotides from the 3Ј terminus. The 3Ј-UTR 373nt RNA templates produced from the S1 nuclease-treated plasmids by the SP6 RNA polymerasecatalyzed in vitro transcription were also inactive like the RNA produced from the PCR product (data not shown). On the other hand, the subgenomic RNAs generated from either the S1 nuclease-treated plasmids under identical conditions used for the 3Ј-UTR 373ϩ4nt plasmid or from the PCR fragment (which would correspond to the authentic 3Ј terminus) were active in the RdRP assays in producing the hairpin-like products (data not shown).

5Ј-terminal Region of DEN2-RNA Is Required for Selfprimed RNA Synthesis at the 3Ј-End-Our results presented in
These results indicated that the subgenomic RNA is intrinsically active in the RdRP reaction to produce the self-primed product, whereas the 3Ј-UTR 373nt RNA is not. The data suggested that the 5Ј-terminal sequences of DEN2 RNA are likely to play a role in modulating the structure at the 3Ј-end to confer a self-priming activity to the subgenomic RNA in the RdRP assay. To address this hypothesis, we tested whether the addition of 5Ј-terminal RNA containing 230 nt (5Ј-TR 230nt ) would confer the self-priming activity to the 3Ј-UTR 373nt RNA in our RdRP assay. The RdRP assay was carried out using the 3Ј-UTR 373nt RNA alone or in the presence of 5Ј-TR 230nt RNA as templates. The results shown in Fig. 6 indicated that the 3Ј-UTR 373nt RNA alone was inactive as a template in the RdRP assay (lane 2), confirming the previous observations; however, the 5Ј-TR 230nt added in trans could activate the 3Ј-UTR 373nt RNA to have the self-primed RNA synthesis activity in the RdRP reaction to produce the faster migrating 2ϫ product (lane 3). Moreover, the 5Ј-TR 230nt RNA by itself had self-priming activity (lane 4). The faster migrating products were confirmed to be 2ϫ by running the samples in formaldehydeagarose gels (data not shown and Fig. 7).
The conserved sequence elements within the 5Ј-and 3Јterminal regions of flavivirus RNA genomes (5Ј-CS1, 3Ј-CS1, 3Ј-CS2, and RCS2) include the complementary motifs, "cyclization sequences" (5Ј-CYC and 3Ј-CYC; Fig. 1) (58). The role of 5Јand 3Ј-CYC motifs or the other conserved sequence elements in the viral life cycle is unknown at present. The results shown thus far indicated that the 3Ј-UTR 373nt RNA, which includes the 3Ј-CYC element, is inactive in the RdRP assay; however, the presence of the 5Ј-terminal region containing the entire 5Ј-UTR (96 nt) with conserved stem-loop structure (57) and the 5Ј-CS1 element along with the 3Ј-UTR in the same molecule (for example, in the subgenomic RNA) or its addition in trans to the 3Ј-UTR RNA activated RNA synthesis at the 3Ј-end of the 3Ј-UTR RNA.
These results raised the interesting question of whether the cyclization motifs play any role in conferring the template activity of 3Ј-UTR 373nt RNA in the RdRP assay. To address this question, a mutant cDNA containing point mutations in the 5Ј-CYC motif that was designed to disrupt base pairing with the 3Ј-CYC motif was constructed (5Ј-TR/mutCYC; Fig. 1). A second construct containing complementary mutations in the 3Ј-CYC motif within the 3Ј-UTR 373nt cDNA (3Ј-UTR 373nt /mut-CYC) was also made. These mutant cDNAs were used to generate the 5Ј-TR 230nt /mutCYC and 3Ј-UTR 373nt /mutCYC RNAs, which were then used as templates in the RdRP assays. The results shown in Fig. 7 indicate that the wild type or mutant 3Ј-UTR 373nt RNA was inactive in the self-priming activity (lanes 2 and 3; lane 1 contained the control, 3Ј-UTR 373ϩ4nt RNA, that has this activity). However, the addition of 5Ј-TR 230nt RNA containing the wild type 5Ј-CYC element activated the 3Ј-UTR 373nt RNA containing the wild type 3Ј-CYC element (lane 4) but not the mutant 3Ј-CYC element (lane 5). Similarly, the mutant 5Ј-TR 230nt /mutCYC RNA could not transactivate the wild type 3Ј-UTR 373nt RNA (lane 7), whereas it could transactivate the mutant 3Ј-UTR 373nt /mutCYC RNA (lane 8). We also observed that both mutant and wild type 5Ј-TR 230nt RNAs were active by themselves in the RdRP assay (lanes 6 and 9). This template activity was probably due to a specific structure of these RNAs recognized by the RdRP.
These results taken together indicated that the 5Ј-and 3Јterminal regions of the viral genome do functionally interact with each other. This interaction is required for the template activity of 3Ј-UTR 373nt RNA in self-primed RNA synthesis by a copy-back mechanism at the 3Ј-end of viral RNA catalyzed by the viral-specific RdRP. In addition, the CYC motifs, which are complementary to each other, are important in modulating this interaction.
The Role of 3Ј-OH Moieties of 3Ј-UTR/wtCYC and 5Ј-TR/ wtCYC RNAs in Self-primed RNA Synthesis in Vitro-We sought to determine whether the 3Ј-OH group of the 5Ј-TR/ wtCYC RNA was required for activation of the template activity of the 3Ј-UTR/wtCYC RNA in the self-primed RNA synthesis. The 3Ј-OH group of each of the RNA template was oxidized by sodium periodate followed by treatment with lysine to remove the excess periodate. The RNAs were purified and used in the RdRP assays. The results shown in Fig. 8 indicate that the 5Ј-TR/wtCYC and 3Ј-UTR/wtCYC RNAs, when present together, are active in self-primed RNA synthesis from the 3Ј-OH moiety of each template molecule (Fig. 8, A, lane 2, and B, lane  5). When the 3Ј-UTR/wtCYC RNA was treated with periodate, its self-priming activity was nearly abolished in the presence of the untreated 5Ј-TR/wtCYC RNA, whereas the untreated 5Ј-TR/wtCYC RNA by itself was still active to produce the 1ϫ and 2ϫ products (Fig. 8, A, lane 3, and B, lane 7). Similarly, when the 5Ј-TR/wtCYC RNA was treated with periodate, its selfpriming activity was nearly abolished, but its ability to activate the 3Ј-UTR/wtCYC RNA in trans for self-primed RNA synthesis was still intact (Fig. 8, A and B, lanes 4 and 6, respectively). When both RNA templates were treated with periodate, the self-priming activity was severely reduced (Fig. 8, A and B,  lanes 5 and 8, respectively). Similar results were observed by treatment of the subgenomic RNA with sodium periodate (Fig.  8A, lanes 7 and 8). These results suggested that the selfpriming RNA synthesis at the 3Ј-OH end of the RNA templates is very specific and is not modified by the cell extracts during the in vitro RdRP reactions. In addition, the trans activation of 3Ј-UTR/wtCYC RNA by 5Ј-TR/wtCYC RNA for RNA synthesis does not require the 3Ј-OH group of the latter. These results are also consistent with the conclusion derived from the results of Fig. 5 because the formation of both 1ϫ and 2ϫ products were abolished by blocking the 3Ј-OH group of the subgenomic RNA (Fig. 8A, lanes 7 and 8). DISCUSSION In this study, we describe the first flaviviral replicase assay system that utilizes cytoplasmic extracts from dengue virusinfected mosquito (C6/36) cells and exogenously added subgenomic RNA templates to study the mechanism of viral replication in vitro. The RNA templates used in this in vitro system contain the conserved sequence (CS) elements found near or within the 5Ј-and 3Ј-UTR of many flavivirus RNA genomes examined to date (Ref. 58 and for a review see Ref. 8). These sequences also include those that are predicted to form potential stem-loop (5Ј-and 3Ј-SL) structures (56 -58, 67, 68). Biochemical evidence for the formation of SL structures was provided by enzymatic RNA structure probing studies (56,68) and from the sequence data of some rare cDNA clones derived from three flaviviral RNAs, which could only be explained by self-primed cDNA synthesis by reverse transcriptase at the 3Ј-end of these viral RNAs (58).
Our results shown in this study indicate that the self-primed FIG. 7. Role of cyclization motifs in RNA synthesis mediated by the 5-TR 230nt RNA at the 3-end of 3-UTR 373nt RNA template. The 9-nt cyclization motifs (CYC) were mutated either individually to disrupt or together to restore complementarity. The mutant plasmids were used for in vitro transcription to synthesize 5Ј-TR 230nt and 3Ј-UTR 373nt RNAs containing either wild type or mutant CYC motifs (wtCYC or mutCYC). RNAs were used in the in vitro assays, and the products were analyzed by formaldehyde-agarose gel system. RNA with mutCYC motif alone. In this gel system, the labeled input template-sized (1ϫ) 3Ј-UTR 373nt RNA, which has an additional 26 nt from the vector sequences, appears to have migrated close to the 2ϫ of 5Ј-TR 230nt RNA product.
FIG. 8. 3-hydroxyl group of the subgenomic RNA or the 3-UTR 373nt RNA is required for RNA synthesis mediated by the 5-TR 230nt sequences. RNAs were synthesized by in vitro transcription and treated with sodium periodate (indicated as ࡔ) followed by lysine to block the 3Ј-OH group. Purified RNAs were used in the in vitro RdRP assays as described under "Experimental Procedures." Lane 1 in A and lane 9 in B refer to the periodate-untreated 3Ј-UTR 373ntϩ4nt RNA as control. RNAs 3Ј and 5Ј refer to 3Ј-UTR 373nt RNA and 5Ј-TR 230nt RNA, respectively. A, analysis of products was carried out by PAGE/ urea system. B, analysis of products by formaldehyde-agarose gel system. The products were detected by autoradiography.
RNA synthesis producing a hairpin-like structure (2ϫ) occurs in the viral replicase-catalyzed reaction, which appears to be the major species formed in vitro. A minor product formed in a RdRP assay carried out using either infected or uninfected cell extracts has the same mobility as the input RNA in a partially denaturing gel electrophoresis system. However, analysis of the product formed with uninfected cell lysates indicated that it was sensitive to RNase A digestion, whereas the product formed with infected cell lysates was RNase-resistant, suggesting that the latter product was double-stranded RNA. These results suggested that the RNase A-sensitive product formed in the reaction containing the uninfected cell lysates is likely to be formed by the host terminal nucleotidyl transferase by addition of nucleotides to the 3Ј-end of the RNA template.
The kinetics of the RdRP reaction catalyzed by the viral replicase reveals that the initial RNA product formed is also sensitive to RNase A, suggesting that it was the product of limited 3Ј-end elongation by either the host terminal nucleotidyl transferase or by the viral replicase or both. This RNA intermediate is then converted to the hairpin-like product by a copy-back mechanism. The results of sodium periodate treatment that abrogated the formation of both 1ϫ and 2ϫ products are also consistent with this conclusion. The RNase A-resistant 1ϫ product observed in a 90-min RdRP reaction could arise from digestion of the single-stranded region of the 2ϫ product by nuclease(s) during the RdRP reaction. However, formation of this product at least in part by de novo synthesis by the viral replicase at the 3Ј-end of RNA template could not be ruled out.
The role of the conserved sequence elements 3Ј-CS1, 3Ј-CS2, 3Ј-RCS2, and 5Ј-CS1 in flaviviral life cycle is unknown. The 9-nt complementary sequence elements, known as cyclization sequences (3Ј-CYC and 5Ј-CYC), are thought to cyclize viral RNA to form a panhandle structure with 11-12 contiguous base pairs that include the adjacent sequences (58). The 3Ј-UTR 373nt RNA, as well as the subgenomic RNA used in this study, include these conserved sequence elements. Whether these CYC elements in flaviviral RNAs are involved in formation of cyclic intermediates during replication or in some other event in the viral life cycle is yet to be determined. The free energy of cyclization was calculated to be Ϫ9 to Ϫ11 kcal at 25°C, supporting the possibility that cyclization could indeed occur (58). We analyzed the potential secondary structure within the 3Ј-UTR and predicted base pairing between 5Ј-and 3Ј-CYC motifs in the 0.7-kilobase subgenomic RNA using the MFOLD program (69) (data not shown). The predicted structure for the 3Ј-SL alone up to CS1 showed that the 3Ј-CYC motif is in a loop region, whereas the predicted structure for the subgenomic RNA showed the 5Ј-and 3Ј-CYC motifs form base pairings in a stable stem containing 11 base pairs (from 134 to 144 nt from the 3Ј-end). Studies using RNA structure probing methods are essential to determine the secondary structure of the subgenomic template RNA active in our in vitro assay. Interestingly, Alphaviral and Bunyaviral RNAs have been reported to cyclize under physiological conditions (70,71). It was proposed that the cyclization event would fulfill a putative requirement for binding of viral RNA replicase to both 5Ј-and 3Ј-terminal regions prior to initiation of viral replication (58).
The results of our study suggest that interactions between the 5Ј-and 3Ј-terminal regions, mediated by conserved CYC motifs, result in a conformational change that is a prerequisite for self-primed RNA synthesis at the 3Ј-end of the 3Ј-UTR by the viral replicase. Cellular proteins that are shown to bind the 3Ј-stem-loop structure of a flavivirus RNA (48,49,72) could play an important role in modulating the stability of this interaction. Although the complementarity between the CYC motifs might be the driving force in bringing the ends of the genome together, it is possible that there are additional interactions, mediated by other conserved motifs such as 3Ј-CS2 and RCS2, which could play a role in conferring a specific conformation for initiation of RNA synthesis.
In a recent report, a functional interaction in vivo between two RNA components, RNA-1 and RNA-2 of the plant virus, red clover necrotic mosaic virus genome, was demonstrated (73). In that study, a 34-nt segment of RNA-2 component was shown to be required for subgenomic RNA synthesis from genomic RNA-1 component in vivo. Interestingly, within the RNA-1 and RNA-2 components, two 8-nt motifs complementary to each other were shown to be important for subgenomic RNA synthesis. Again, mutations that maintained complementarity were active, suggesting that base pairing interactions between the two motifs are likely to be involved in subgenomic RNA synthesis (73).
Complete replication of RNA genomes of the positive strand viruses, cucumber mosaic virus, tobacco mosaic virus, and nodavirus by either a partially purified or by a crude membranebound replicase complex have been reported (28,34,74). In other in vitro template-dependent replication systems, only synthesis of the negative strand to form the double-stranded RNA intermediate has been demonstrated (35). In some in vitro RdRP reactions, the template RNA gave rise to not only the complementary (Ϫ)-strand RNA but also RNA species larger than the input RNA templates. For example, poliovirus RdRP gave rise to dimer-sized RNA products in vivo and in vitro (75)(76)(77)(78). The evidence suggested that the 3Ј-ends were generated by endonucleolytic attack of the 3Ј-SL structure(s) of polioviral RNA, which then served as templates for the RdRP giving rise to a variety of self-primed products. Moreover, blocking the 3Ј-OH by sodium periodate did not have any effect on the generation of dimer-sized products, because new 3Ј-OH was presumably generated by the endonucleolytic cleavage (77). However, the RNA synthesis at the 3Ј-end of the RNA templates used in our system appears to be specific to RNA structure. In contrast, purified preparations of viral RdRP of poliovirus (79), hepatitis C virus (80) or dengue virus (81) can utilize exogenous viral or synthetic RNA templates for RNA synthesis in vitro but without any template specificity or structure dependence.
A few other viral RdRP such as cucumber mosaic virus (28), brome mosaic virus (31,82), and turnip crinkle virus (32) showed template and structure-specific RNA synthesis. The turnip crinkle virus RdRP utilized the (Ϫ)-strand of satellite C RNA as a template and produced two non-template-sized products, a panhandle-like structure and a hairpin-like structure (L-and S-RNAs, respectively) by terminal elongation of the 3Ј-end. The formation of the L-RNA had specific sequence and structural requirements, whereas the S-RNA was formed from all (Ϫ)-strand RNAs tested (50). These RdRP preparations had some cellular proteins that perhaps contributed to the template specificity (for a review see Ref. 36). Cellular proteins from the DEN2-infected extracts are also likely to contribute to the exogenous template specificity observed in this study.
Other studies have shown that cis-acting signals required for viral RNA replication are noncontiguous in the viral genome of the brome mosaic virus (44,83,84), tobacco mosaic virus (85), and the double-stranded RNA virus of yeast (L-A virus) (Ref. 41 and for a review see Ref. 36). Moreover, in the single-stranded negative sense RNA genome of the Rhabdoviridae family (the vesicular stomatitis virus RNA (86)) and in the segmented influenza virus RNA genome (87), interactions between 5Ј-and 3Ј-terminal sequences involve self-complementary sequences. The mechanisms of viral replication and the specificity of interactions of the viral replicases and viral RNA genomes are likely to be distinct in these different viruses. Development of an in vitro RdRP assay for a flavivirus that can utilize exogenous RNA templates and identification of a functional interaction between conserved sequence elements within the 5Ј-and 3Ј-terminal regions that play a role in viral RNA synthesis in vitro are important steps toward unraveling the mechanism of flavivirus replication.