Novel cis-Acting Element within the Capsid-Coding Region Enhances Flavivirus Viral-RNA Replication by Regulating Genome Cyclization

Cell culture.

BHK-21 cells were grown in Dulbecco modification of minimal essential medium (DMEM; pH 7.5; Gibco) supplemented with 5% fetal bovine serum (FBS; PAA, Morningside, Queensland, Australia) at 37°C in 5% CO₂.

DNA constructs.

The p4-Rluc-Rep replicon was reported previously (31, referred to as DEN-R.luc2A-RP in the original publication). The infectious clone p4 (32) and the host Escherichia coli strain for all of the p4 and replicon constructs, BD1528, were kindly provided by Stephen S. Whitehead. To construct the p4-cHPstop-SP-IRES-Rluc-Rep replicon, the encephalomyocarditis virus (EMCV) internal ribosome entry site (IRES) sequence was fused upstream of the Renilla luciferase (Rluc) gene by overlapping PCR, and the resulting fragment was then fused downstream of the spacer sequence (positions 67 to 707 of the enhanced green fluorescent protein [EGFP] gene ORF). The AscI-XhoI fragment of p4-Rluc-Rep was replaced by an artificial multiple cloning sequence: 5′-GGCGCGCCAAACTTTCCCCTTACGTATTAACATGCATCTTGCGGCCGCGTACAGAGTCTCGAG--3′ (restriction sites for AscI, SnaBI, NotI, and XhoI are underlined from the left to right). The resulting construct was used as a backbone for the assembly of a new replicon. First, a hepatitis D virus ribozyme sequence, followed by a unique AgeI site, was inserted downstream of the 3′ end of DENV cDNA in the backbone construct. The spacer-IRES-Rluc fragment and the AscI-5′UTR-cHPstop-SnaBI fragment, which contains a stop codon in the cHP sequence, were subsequently subcloned into the backbone to obtain p4-cHPstop-SP-IRES-Rluc-Rep.

A cassette cloning strategy was then used to introduce mutations into p4-cHPstop-SP-IRES-Rluc-Rep. The AscI-5′UTR-cHPstop-SnaB I fragment was engineered by overlapping PCR so that its sequence corresponding to position 54 to 90 in the capsid ORF was replaced by the following sequence: 5′-CTCCTCACACACTACGTAAATGGGGAGGAG-3′. The underlined sequence indicates the BseRI restriction site. The resulting construct was named pBseRI-03. Oligonucleotides containing the desired mutations were annealed and cloned into BseRI-treated pBseRI-03 to generate AscI-5′UTR-cHPstop-SnaB I fragments containing different mutations. The corresponding fragments were then subcloned into p4-cHPstop-SP-IRES-Rluc-Rep to obtain replicon mutants.

To introduce mutations into the p4 infectious clone, mutations were introduced into the AscI-NheI-XhoI fragment from p4-Rluc-Rep using a cassette cloning method similar to that described above. The NheI-XhoI fragments were then replaced by the NheI-XhoI fragment from p4. The resulting constructs contained AscI-XhoI fragments from p4 with the desired mutations. The corresponding fragments were subcloned into the p4 plasmid by Eco91I/PauI digestion, and p4 constructs carrying various mutations in DCS-PK were generated. To generate GVD-containing p4 mutants, the SacII-MluI fragment of p4 was substituted with homologous fragments containing GVD mutations. The Eco91I-PauI fragment of the p4-GVD plasmid was substituted with homologous fragments containing mutations. The sequences of the primers and oligonucleotides are available upon request.

RNA structure prediction.

RNA secondary structures were predicted using the mfold online server (33, 34). The pknotsRG (35, 36) and PseudoViewer3 (37, 38) online versions were used to predict and illustrate, respectively, the RNA structures containing pseudoknots.

In vitro transcription.

The RiboMAX large-scale RNA production SP6 kit (Promega) was used for in vitro transcription according to the manufacturer's instructions, with minor changes. For the replicon and infectious RNA assay, linearized DNA was extracted with phenol-chloroform, ethanol precipitated, and dissolved in RNase-free water. Topically, 2 to 3 μg of linearized DNA, 6 μl of 5× SP6 buffer, and 2.5 μl of SP6 enzyme mix were mixed with 6 mM ATP, 6 mM CTP, 6 mM UTP, 2 mM GTP, and 0.83 mM m⁷GpppG cap analog (Promega) in a 30-μl reaction. The reactions were incubated at 37°C for 3.5 h, and then RQ1 RNase-free DNase I (Promega) was added to the reaction and incubated at 37°C for 20 min to remove the DNA template. For RNA structure probing and RNA binding assays, DNA fragments containing the SP6 promoter and respective viral sequences were amplified by high-fidelity PCR and purified from agarose gels to prepare templates for the transcription of DEN-4 5′ 300-nt RNA, DEN-4 3′UTR RNA, JEV 5′ 320-nt RNA, and WNV 5′ 320-nt RNA. In vitro transcription reactions were topically performed at 37°C for 4.5 h in a 60-μl mixture containing 1 to 2 μg of template DNA, 12 μl of 5× SP6 buffer, 6 μl of SP6 enzyme mix, and 5 mM each of the four ribonucleoside triphosphates, followed by the removal of DNA templates as described above. All in vitro-transcribed RNAs were purified using the RNeasy minikit (Qiagen) according to the RNA cleanup protocol. Purified RNA was quantified by an ND-1000 spectrophotometer (NanoDrop) and stored at −80°C before use.

RNase probing.

Various 5′ end RNAs of flaviviruses were incubated at 85°C for 5 min and slowly cooled to room temperature. The RNAs were then digested by 0.01 U of RNase T1, 0.001 U of RNase V1, or 0.01 ng of RNase A (Ambion). The reactions were performed at 25°C for 15 min in a 10-μl mixture containing 10 mM Tris-HCl (pH 7.0), 100 mM KCl, 10 mM MgCl₂, 10 μg of yeast RNA, 1 μg of sample RNA, and the respective RNase. Control reactions without RNase were performed in parallel. Forty microliters of precipitation/inactivation buffer with ethanol was added to stop the reactions. Reactions were incubated at −80°C for 30 min, followed by centrifugation at 15,700 × g and 4°C for 15 min. Pellets were washed with 120 μl of 70% ethanol, air dried, and dissolved in 5 μl of RNase-free water.

NMIA modification.

Various RNAs corresponding to the 5′ end sequences of DENV, JEV, and WNV were modified with N-methylisatoic anhydride (NMIA; Sigma-Aldrich). Experiments were performed as previously described (39). Reactions were incubated at 37°C for 45 min, and then the modified RNAs were purified by using the RNeasy MinElute cleanup kit (Qiagen).

Primer extension.

Purified RNase-treated and NMIA-modified RNA samples were reverse transcribed into cDNA using carboxyfluorescein (FAM)-labeled primers. The primer sequences were as follows: 5′-FAM-GTGGGATGGAAAGGACTC-3′ or 5′-FAM-TTCCCAGAAAAAAGTCC-3′ for DENV, 5′-FAM-TCGGGGCTAATGCTGTAAACT-3′ for JEV, and 5′-FAM-TGCTGTGAACCTGAAGAAC-3′ for WNV. Reverse transcription was performed using Superscript II reverse transcriptase (Invitrogen) in a 20-μl reaction according to the manufacturer's instructions. The FAM-labeled cDNA samples were analyzed by capillary electrophoresis using a 3730xl DNA analyzer (Applied Biosystems). Dideoxy sequencing reactions were performed in parallel. The detection of NMIA reactivity by primer extension is generally known as the SHAPE (selective 2′-hydroxyl acylation analyzed by primer extension) assay (39).

Replicon assay.

A total of 2 × 10⁴ BHK-21 cells were seeded into each well of 48-well plates 1 day prior to transfection and incubated at 37°C in 5% CO₂. Transfection was performed with Lipofectamine 2000 reagent (Invitrogen) in triplicate according to the manufacturer's instructions. A total of 250 ng of RNA was transfected into each well, except for p4-Rluc-Rep, of which 500 ng of RNA was transfected. Cell lysates were collected at 6, 24, 48, 72, and/or 96 h after transfection, and Renilla luciferase activity was measured using the Renilla luciferase assay system (Promega) in a GloMax-96 microplate luminometer (Promega).

Electroporation of full-length vRNA.

Portions (5 μg) of each RNA transcript was electroporated into 400 μl of BHK-21 cells at 10⁷ cells/ml in 0.4-cm cuvettes (Bio-Rad) using a Gene Pulser Xcell electroporation system (Bio-Rad) at 270 V and 975 μF. The cells were resuspended in 14 ml of DMEM containing 5% FBS. A total of 0.5 ml of the cell suspension was seeded into each well of 24-well plates, and 0.3 ml of the cell suspension was seeded into each well of 8-well Lab-Tek II chamber slides (Thermo Scientific). Plates and slides were incubated at 37°C in 5% CO₂. The medium in the 24-well plates was changed 4 h postelectroporation.

qRT-PCR.

At 4, 24, 48, 72, and 96 h postelectroporation, electroporated BHK-21 cells were washed thoroughly with phosphate-buffered saline (PBS), and then total RNA was isolated using the RNeasy minikit (Qiagen) in triplicate except for the experiment accessing the stability of WT, XT, S1X, S3X, 75.78, and 91.93-GVD RNAs, in which RNA was isolated in duplicate. Quantitative reverse transcription-PCR (qRT-PCR) was performed on a LightCycler 2.0 (Roche) system with the One Step PrimeScript RT-PCR kit (TaKaRa). Probe (5′-FAM-CAGGAGGATTGTGACCATAGAGGCCCA-TAMRA-3′) and primers (forward, 5′-CCCCGGAACAACAGTCACA-3′; reverse, 5′-TGCAGTGGTGGTCCTCAAAG-3′) targeting the NS1 gene were used to detect DENV RNA. Purified full-length RNA transcript was 10-fold serial diluted and used as standards for calculation of RNA copy number.

Immunofluorescence assay (IFA).

Electroporated BHK-21 cells grown in chamber slides were washed with PBS and fixed with cold acetone at 24, 48, 72, and 96 h after electroporation. The fixed cells were incubated with anti-DENV envelope monoclonal antibody (2D73; 20 μg/ml, 40 μl per well) at 37°C for 1 h, followed by three 10-min PBS washes. Goat anti-mouse IgG (1:200 dilution; Alexa Fluor 488-conjugated [Zsbio]) was added, and the slides were incubated at 37°C for 1 h, followed by three 10-min washes with PBS. The cells were then stained with 40 μl of 0.5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) for 5 min and washed with PBS. The fluorescence images were captured with an Olympus BX51 microscope.

Plaque assay.

Culture supernatants of electroporated BHK-21 cells were collected at 48, 72, and 96 h postelectroporation. The virus titers of the supernatants were determined by plaque assay as described before (40).

RNA binding assay.

A total of 0.3 μg of DEN-4 3′UTR RNA and various amounts (0.24, 0.6, 1.2, or 2.4 μg) of different 5′ 300-nucleotide (nt) RNAs were mixed and diluted to 14-μl reaction mixes, which also contained 5 mM Tris-HCl (pH 8.0) and 0.5 mM EDTA, the 5′:3′ RNA molar ratio in the samples is 1:1, 2.6:1, 5:1, and 10:1, respectively. The samples were then heated at 95°C for 2 min and immediately placed in an ice bath. Next, 6 μl of 3.3× RNA folding buffer (333 mM HEPES [pH 8.0], 20 mM MgCl₂, 333 mM NaCl) was added to each sample and mixed well. The RNA samples were incubated at 37°C for 30 min to allow folding into complexes, and then 2.5 μl of gel loading solution (Ambion) was added to each sample. The samples were loaded into 5% native polyacrylamide gels, which had been pre-run in 0.5× Tris-borate-EDTA buffer at 4°C for 30 min, and electrophoresis was conducted at 60 V at 4°C for 4 h. The gels were then stained with SYBR Gold nucleic acid gel stain (Invitrogen) for 20 min and directly visualized with a UV transilluminator. The data were analyzed using the ImageJ package.

Conserved three-stem pseudoknot is located downstream of the 5′CS in mosquito-borne flavivirus.

To identify potential cis-acting element(s) downstream of the 5′CS in mosquito-borne flaviviruses, consensus sequences of the capsid-coding region of each flavivirus species were generated and subjected to the mfold web server (33, 34) to obtain structure models. The base-pairing patterns of the two most stable structure models implied that they were possibly coexistence by formation of a pseudoknot (data not shown). Sequence comparison also indicated that the downstream of 5′CS region's base-pairing potential of folding into a pseudoknot is conserved in all mosquito-borne flaviviruses except YFV (Fig. 1A). To address the possibility that the cis-acting element downstream of the 5′CS has a pseudoknot structure, the sequences of downstream of 5′CS regions of different flaviviruses were subjected to pknotsRG (35, 36). As we hypothesized, similar pseudoknot structures were predicted (Fig. 1B), which were designated DCS-PK. The DCS-PK has three stem regions (Fig. 1B), and stem 1 and loop 2 are highly conserved among mosquito-borne flaviviruses (Fig. 1A), whereas the stem 2-loop 3 hairpin, which links the 3′ of stem 1 and 5′ of loop 2, has variable sequences between different virus species. The sequences of loop 1 and stem 3 are also not conserved. These bioinformatic analyses indicate that a conserved RNA pseudoknot may be formed in the capsid-coding region of mosquito-borne flavivirus.

Structure probing of the DCS-PK structure.

To confirm the physical existence of the predicted DCS-PK structure, RNase V1/T1/A and NMIA were used to probe the 5′ end structure of DENV serotype 4 (DEN-4), WNV, and JEV (Fig. 2 and data not shown). RNase V1, which cleaves double-stranded or base-stacking regions, detected all three DCS-PK stems from DEN-4 and JEV, and stems 1 and 2 from WNV were also RNase V1 sensitive. Single-strand-specific RNase T1/A-sensitive nucleotides were mainly located in loops and junctions. NMIA selectively adds 2′-O adducts to structural flexible nucleotides, and its reactive sites in DCS-PK were mainly located in loop regions and loop-stem interfaces, whereas most nucleotides in the stems showed no SHAPE activity. In summary, the structure probing results support the predicted three-stem pseudoknot structure of DCS-PK.

DCS-PK is required for efficient vRNA replication and virus propagation.

Reverse genetic approaches were used to investigate the role of DCS-PK in virus life cycle. Silent mutations targeting DCS-PK (Fig. 3A) were introduced into the infectious clone p4 of DEN-4 (32). The XT mutant was designed to disrupt the overall secondary structure of DCS-PK. Stem 1 was disrupted in S1X, and S3X included a disrupted stem 3. Mutations were also introduced into the nonconserved loop 3 to generate mutant 75.78, which retained the original secondary structure of DCS-PK. The codon usage frequencies in the above mutants were left unchanged to avoid potential effects on viral translation efficiency. Mutant 91.93, which targeted the bulged A93 of stem 3 and A91 of loop 2 but did not affect DCS-PK secondary interactions, corresponded to a previously reported DEN-2 mutant (28) with reduced vRNA replication. Detection of viral antigens in vRNA-transfected BHK-21 cells by IFA demonstrated that all of the above-mentioned mutant RNAs were infectious. There were fewer DENV-positive cells in XT, 91.93, S1X, and S3X mutant RNA-transfected cells than in wild-type (WT) or 75.78 mutant RNA-transfected cells, demonstrating that the disruption of RNA structure in DCS-PK impairs viral replication (data not shown). To directly assess the effect of DCS-PK on viral genome RNA replication, vRNA levels in transfected cells were determined by qRT-PCR. The vRNA levels in the XT, S1X, and S3X mutant RNA-transfected cells were approximately 1 order of magnitude lower than those of WT and 75.78 RNA-transfected cells at 48 to 72 h after transfection, whereas the latter two exhibited similar vRNA levels at these time points (Fig. 3B). The vRNA level of 91.93 also showed a moderate reduction of ca. 50% at 48 to 72 h, which is in agreement with previous findings (28). The viral loads in the supernatants of the WT-, 75.78-, and S3X-transfected BHK-21 cells were compared. At 48 to 72 h after transfection, the virus titers of S3X were significantly lower than those of WT or mutant 75.78 (Fig. 3C). To exclude the possible impact of mutations on RNA stability, RNAs containing mutations in the conserved GDD motif of NS5 were electroporated into BHK-21 cells. Monitoring of intracellular vRNA levels indicated that the stability of mutant RNAs was similar to that of WT RNA (Fig. 3D). Taken together, the results described above provide strong evidence that the presence of DCS-PK structure enhances vRNA replication and thus facilitates virus propagation. These results also suggest that the function of DCS-PK depends on its secondary structure, as well as some conserved primary sequences.

Secondary structure of DCS-PK is required for its function.

The function of DCS-PK and its dependence on the secondary structure was further investigated with another reverse genetic approach. For each DCS-PK stem region, a pair of mutants was constructed to bypass potential influences of changes in the capsid sequence and/or viral translation on vRNA replication (Fig. 4A). One mutant contained a disrupted stem region (S1NEG, S2NEG, and S3NEG), and the corresponding stem region was restored in the other construct (S1POS, S2POS, and S3POS). The paired mutants encoded the same capsid sequence, which differed from the WT capsid by a few residues. The mutations were carefully chosen to minimize the changes in conserved residues in the capsid sequence, and codon usage frequencies of the paired mutants were remained the same. The vRNA levels of S1POS and S1NEG were similar at all time points after transfection (Fig. 4B). In contrast, at 48 to 96 h after transfection, the vRNA level in the S2POS-transfected cells was significantly higher than that of S2NEG, indicating that S2POS replicated more efficiently (Fig. 4C). Similar results were observed for S3POS and S3NEG (Fig. 4D). The RNA stability was similar between each pair of mutants, as determined by electroporation of NS5 GVD mutant RNAs (Fig. 4B to 4D). IFA detection of viral antigens in vRNA-transfected cells (data not shown) agreed with the results of vRNA replication. Taken together, these results indicate that secondary structure is required for DCS-PK function. For stem 2 and stem 3, maintaining their secondary structures is most likely sufficient for the function of DCS-PK, whereas in the case of stem 1, its highly conserved sequence might be required for DCS-PK function.

The function of stem 2-loop 3 hairpin depends on its secondary structure.

To characterize the structure and function of DCS-PK in detail, a novel DEN-4 replicon was designed (Fig. 5) to bypass the coding role of DCS-PK sequence. In this replicon, the translation of viral protein is under the control of EMCV IRES, and the 0.6-kb spacer sequence separates the 5′UTR-capsid region and IRES to ensure the correct folding of both (Fig. 5A). To abort 5′ cap directed translation of capsid ORF, one stop codon was introduced into the loop sequence of the cHP element (18) without changing its secondary structure. SHAPE analysis showed that the stop codon introduced into cHP did not change the overall secondary structure of DEN-4 5′ end (data not shown). The p4-cHPstop-SP-IRES-RLuc-Rep replicated similarly to p4-Rluc-Rep (Fig. 5B) and was suitable for further mutation experiments.

Sequence alignment showed that the primary sequences of the stem 2-loop 3 hairpin are not conserved among flaviviruses, suggesting that the secondary structure is crucial in determining the function of the stem 2-loop 3 hairpin. To verify this hypothesis, a set of mutations (Fig. 6A) were introduced into p4-cHPstop-SP-IRES-RLuc-Rep, and their replication efficiency was monitored by a transient luciferase assay. As shown in Fig. 6B, the role of the stem 2-loop 3 hairpin in vRNA replication was confirmed, as deletion of the entire stem 2-loop 3 hairpin sharply decreased vRNA replication. Randomization of the stem 2 sequence did not impair vRNA replication, whereas randomization of either the upstream or downstream strand sequence, which disrupted stem 2 secondary structure, significantly reduced vRNA replication 48 h posttransfection. Lengthening or shortening stem 2 by 2 bp did not adversely affect vRNA replication. Randomization of 4 of 5 nt in loop 3, lengthening loop 3 by 2 nt, or shortening it by 2 nt also had no harmful effects on vRNA replication. In summary, the above results demonstrate that the function of stem 2-loop 3 hairpin depends on its secondary structure and is independent of primary sequence.

Sequence requirements of stem 1-loop 2 for efficient vRNA replication.

Loop-helix interactions have been found in many H-type pseudoknots (41, 42, 43, 44). Generally, loop 2 crosses the minor groove of stem 1, and loop 1 crosses the major groove of stem 2 (corresponding to stem 3 of DCS-PK), which facilitates the formation of tertiary contacts. Although DCS-PK is not a classical H-type pseudoknot, the conservation of stem 1 and loop 2 sequences (Fig. 1) and the low SHAPE activity of loop 2 (Fig. 2) imply that loop-helix interactions might play roles in its function. To investigate the sequence requirements of stem 1 and loop 2 for vRNA replication, two panels of mutagenesis were performed based on p4-cHPstop-SP-IRES-RLuc-Rep (Fig. 7). First, randomizing, lengthening, or shortening stem 1 and the disruption or deletion of the stem 1 secondary structure significantly reduced vRNA replication (Fig. 7A and B), demonstrating the high sequence dependence on conserved stem 1 for optimal vRNA replication. Next, five base pairs of stem 1 were substituted with different Watson-Crick base pairs (Fig. 7C). Base pair substitutions (1.2GC, 1.3CG, and 1.4GC) significantly enhanced replication compared to single mutations aiming to disrupt the corresponding base pairs (G64C, C63G, and C51G) (Fig. 7D), confirming the importance of secondary interactions in stem 1 for DCS-PK function; however, the replication of all of the base pair-substituted replicons was reduced to some extent compared to the WT replicon. For A48-U65, replacing it with a G-C base pair reduced vRNA replication to ca. 15% relative to the level of the WT replicon 48 h posttransfection (referred to as the percentage of WT below). For G50-C63, the only mutant able to reach 50% of the level of WT was the C63U mutant (G50-C63 Watson-Crick to G50-U63 wobble), whereas mutants containing other base pairs replicated poorly, indicating that G50 and the base-pairing interactions with its partner are absolutely required for DCS-PK function. All of the G52-C61 base pair substitutions reduced vRNA replication to ca. 15 to 30% of the WT. Substitutions of the C49-G64 and C51-G62 pairs had less effect on vRNA replication than substitutions of the other three base pairs. The above results demonstrate that both the base composition and orientation of the base pairs in stem 1 are important for DCS-PK function.

Loop 2 of DCS-PK has consensus sequence G/AAAGA. Shortening or lengthening loop 2 significantly reduced vRNA replication (Fig. 7D), indicating the importance of the length of loop 2. All single-base substitutions reduced vRNA replication >2-fold compared to WT, except for G87A, which replicated to ca. 70% of WT. These results agree with sequence alignments, which show that G and A exist at the 5′ first position of DCS-PK loop 2 in different flaviviruses, whereas the other nucleotides are highly conserved (Fig. 1). Thus, the loop 2 nucleotide sequence is critical for optimal vRNA replication. These results indicate that most of the conserved nucleotides in stem 1 and loop 2 are indispensable for optimal vRNA replication. The above results also agree with the common loop-helix interaction feature of classical pseudoknots, suggesting the presence of tertiary interactions between DCS-PK stem 1 and loop 2.

Loop 1-stem 3 interactions are important for the structure and function of DCS-PK.

To investigate the role of loop 1 in DCS-PK function, mutagenesis was performed based on the above-mentioned replicon system (Fig. 8A). First, A53G mutation or A53 deletion significantly reduced vRNA replication (Fig. 8B). In contrast, A53U only had little effect on vRNA replication, and the insertion of 2 nt into loop 1 also did not significantly change vRNA replication (Fig. 8B). Thus, the exact length of loop 1 is not required for DCS-PK function, which is different from loop 2. The results of the mutagenesis analysis fit well with the sequence comparison, and loop 1 should contain at least one non-G residue (Fig. 1A).

SHAPE analysis was then performed for the loop 1 mutants A53G and A53U (Fig. 8C and D). A53G mutation led to a dramatic increase in the SHAPE activity of stem 3, whereas substituting A53 with U53 did not change the overall SHAPE activity of DCS-PK. Thus, loop 1 is important for the stability of DCS-PK structure. These results suggest that, similar to the loop 1-stem 2 major groove interactions in classical pseudoknots, loop 1 interacts with stem 3 to maintain the structure and function of DCS-PK because the A53G mutation greatly reduced the stability of stem 3 without influencing its secondary interactions. Importantly, substitution of the DCS-PK loop 1 sequence of DEN-4 with the corresponding sequence from JEV (L1J) reduced vRNA replication, whereas replacing loop 1 and stem 3 with homologous regions from JEV (S3L1J) restored vRNA replication. Although the S3J mutant's replication efficiency was similar to WT, these data provide functional evidence that loop 1 interacts with stem 3 (Fig. 8A and B). Taken together, the highly consistent results from SHAPE and mutagenesis demonstrate that the interaction between loop 1 and stem 3 is critical for the structure and function of DCS-PK.

Structural requirement of stem 3 for efficient vRNA replication.

Since the sequence of stem 3 is not conserved among flaviviruses, mutagenesis was performed to investigate the structural requirement of stem 3 for DCS-PK function. As shown in Fig. 9, changing the length of stem 3 by 1 bp or replacing the 5′ helix of stem 3 with the corresponding region from JEV did not affect vRNA replication, but randomization of the 5′ helix of stem 3 or half of it significantly reduced vRNA replication (Fig. 9B). Mutagenesis analysis of bulged A59 and A93, the A55-U97, G56-C96, A58-U94, and C60-G92 bp (Fig. 9C) was performed based on the replicon system. Deletion of A59 or A93 or the combination of both did not affect vRNA replication, and substitution of the A59-A93 bulge with a U-A base pair also had little effect on vRNA replication, suggesting that the bulges in stem 3 are dispensable for DCS-PK function; however, the A59G and A93G mutants showed reduced replication compared to WT (Fig. 9D). Substitution of A55-U97, G56-C96, and A58-U94 with other base pairs did not impair vRNA replication, suggesting that these base pairs are not strictly required for DCS-PK function. In contrast, all of the mutants that deleted, disrupted, or substituted the C60-G92 bp with other base pairs replicated less efficiently than WT (Fig. 9D). These results suggest that the C60-G92 bp is the most likely determinant of stem 3's ability to form tertiary interactions with loop 1 to stabilize the structure of DCS-PK.

Localization of DCS-PK and its distance relative to the 5′CS.

In most flaviviruses, the DCS-PK element is located just 2 to 3 nt downstream of the 5′CS, with two or three adenosine (A) nucleotides as the spacer sequence between them. Mutagenesis using the DEN-4 replicon demonstrated that insertion of more adenosines into the spacer sequence had only a slight effect on vRNA replication, whereas substitutions of adenosines with other nucleotides (G or U) significantly reduced vRNA replication (Fig. 10A and C). Inserting either WT or mutated DCS-PK (DCS-PKX-PK or DCS-PKX-PKX) into the downstream region of the mutated DCS-PK in the original position failed to restore vRNA replication (Fig. 10B and C), confirming that proximity between DCS-PK and 5′CS is required for DCS-PK function. Importantly, the DCS-PK of DEN-4 could be functionally replaced by the JEV DCS-PK but not by the SL6 element of TBEV (27) or the H-type pseudoknot (45) from phage T2 gene 32 mRNA (Fig. 10B and C).

DCS-PK modulates vRNA cyclization.

The dependence of DCS-PK function on its proximity to the 5′CS suggests that DCS-PK plays a role in vRNA cyclization. Thus, RNA binding assays were performed to test this hypothesis. The 5′ 300-nt RNAs of WT, S1X, S3X, 75.78, S2POS, and S2NEG were used to bind the 3′UTR RNA (Fig. 11). The percentage of 5′-3′ complexes increased and the percentage of free 3′ RNA decreased as the concentration of 5′ RNA increased. No difference in complex formation was found between 75.78, S2POS, and WT; however, the percentages of 5′-3′ complexes formed by S1X, S3X, and S2NEG 5′ RNA mutants were significantly lower than that formed by WT when the lowest concentration of 5′ RNAs was used (Fig. 11, lane 1×). At a higher 5′ RNA concentration (Fig. 11, lane 10×), the percentage of the remaining free 3′ RNA was apparently higher for S1X, S3X, and S2NEG than for WT 5′ RNA. These results demonstrate that the disrupted DCS-PK structure (Fig. 11, S1X, S3X, and S2NEG) hindered the formation of the 5′-3′ complex and that an intact DCS-PK structure is required for efficient vRNA genome cyclization. The above results agree with the vRNA replication results (Fig. 3 and 4), indicating that DCS-PK enhances vRNA replication by modulating genome cyclization.