Interactions of Dengue Virus NS5 and NS3 with the 3′ End of Its Negative-Strand RNA

TL;DR

This study explores how dengue virus proteins interact with RNA structures to initiate replication, revealing key differences in their recognition of RNA ends.

Contribution

The study identifies the functional role of the 3′SLA RNA structure in dengue virus replication and its interaction with NS5 and NS3 proteins.

Findings

NS5 polymerase efficiently synthesizes RNA from both 5′SLA- and 3′SLA-containing templates.

NS5 and NS3 show reduced binding to a dsRNA intermediate formed by 5′SLA and 3′SLA.

NS3 is unable to unwind the dsRNA intermediate, suggesting additional factors are needed for replication.

Abstract

Dengue virus is an important human pathogen that infects over 400 million people each year. Despite its global health significance, several essential aspects of the viral replication mechanism remain poorly understood. Flaviviruses carry out asymmetric viral RNA synthesis, wherein positive-strand RNA is synthesized in excess over negative-strand RNA. The template for positive-strand synthesis is the negative strand in a double-stranded RNA intermediate, yet little is known about how positive-strand RNA synthesis is initiated. Orthoflaviviruses, including dengue virus, require an RNA promoter, stem–loop A (SLA) at the 5′ end of the viral genome for negative-strand RNA synthesis. Consequently, a complementary stem–loop structure is predicted at the 3′ end of the negative strand (3′SLA), where positive-strand synthesis is initiated. To understand the functional role of 3′SLA, we investigated its structure and examined its interaction with the viral replication proteins NS5 polymerase and NS3 helicase. NS5 and NS3 differentially recognize the stem–loop structures of the positive and negative strands (5′SLA and 3′SLA, respectively), yet NS5 polymerase efficiently synthesizes RNA from both 5′SLA- and 3′SLA-containing templates. We further show that the stable 5′ and 3′SLA elements readily form a duplex that mimics the replication intermediate under our in vitro conditions. Both NS5 and NS3 showed reduced binding to this dsRNA intermediate and NS3 was unable to unwind it, suggesting that additional factors may be required to regulate viral replication in infected cells.

Figures

Full text

1. Introduction

Orthoflaviviruses are positive-strand RNA viruses, belonging to Flaviviridae. Members of the orthoflavivirus genus include medically important human pathogens such as dengue, Zika (ZIKV), West Nile (WNV), yellow fever, and Japanese encephalitis virus [1]. Dengue virus (DENV) is transmitted by Aedes aegypti or Aedes albopictus mosquitoes, which are primarily found in tropical and subtropical regions. While DENV infection typically causes dengue fever with flu-like symptoms, in some cases, it can result in life-threatening conditions such as dengue hemorrhagic fever and dengue shock syndrome [2]. The CDC estimates that DENV infects over 400 million humans worldwide each year, resulting in ~40,000 deaths from severe dengue disease (https://www.cdc.gov/dengue/about/index.html). The DENV genome shares a similar organization with other orthoflaviviral genomes. The 11 kb positive-sense RNA consists of 5′ and 3′ untranslated regions (UTRs) flanking one open reading frame (ORF). The ORF encodes a polyprotein that is subsequently processed into ten proteins, three structural (C, prM, and E) and seven non-structural (NS) proteins. The NS proteins, together with the viral RNA, form the viral replication complex responsible for RNA replication [3,4]. Although all seven NS proteins are involved in the formation of a viral replication complex, only two, NS5 and NS3 have enzymatic activities responsible for the genome replication process. Crystal structures of both NS5 and NS3 have provided mechanistic insights into their enzymatic functions [5,6,7]. NS5 comprises two domains. The smaller N-terminal methyltransferase (MTase) domain is involved in 5′-RNA cap formation and methylation, modifications that are critical for viral RNA stability and translation [8,9]. The larger C-terminal RNA-dependent RNA polymerase (RdRp) domain is responsible for viral RNA synthesis [10,11]. NS3 also has two domains, an N-terminal serine protease domain and a C-terminal helicase domain [12]. The NS3 protease requires NS2B as a cofactor and is responsible for cleaving the viral polyprotein. The C-terminal helicase domain possesses helicase and RNA triphosphatase activities [13,14].

The viral 5′ and 3′ UTRs also play critical roles in regulating RNA replication. It is proposed that the viral genome undergoes circularization, mediated by the complementary cyclization sequences located at the 5′ and 3′-ends of the RNA [15,16]. At the 5′ end, a 70-nucleotide (nt) long stem–loop structure known as stem–loop A (5′SLA) functions as a promoter by recruiting the viral polymerase NS5 [15,17]. Recent studies suggest that the 3′-end of the genome also contains RNA elements recognized by NS5 [18,19]. NS5 first synthesizes the negative-strand RNA complementary to the genomic positive strand, leading to the formation of a replicative double-stranded RNA (dsRNA) intermediate [20,21,22]. This dsRNA intermediate then serves as a template for synthesis of new positive-strand RNAs, which are subsequently capped and methylated to generate functional viral genomes (Figure 1A) [20].

During positive-strand RNA synthesis, the dsRNA intermediate must be separated, as NS5 can only use single-stranded RNA (ssRNA) as a template [23]. The viral helicase NS3 is likely involved in unwinding of the replicative dsRNA intermediate [11]. However, NS3 has not been shown to unwind blunt-ended dsRNA and thus, the mechanism by which the dsRNA intermediate is separated is unclear. Upon strand separation, stable stem–loop structures are likely to form at both termini, one at the 5′ end of the positive strand (5′SLA) and a complementary structure at the 3′ end of the negative strand (3′SLA) (Figure 1). It is currently not known whether 3′SLA folds into a stable RNA structure and is involved in replication. NS5 and NS3 likely recognize these RNA elements (5′SLA and 3′SLA) to initiate positive-strand RNA synthesis [11,24]. We have previously investigated NS5 interaction with 5′SLA and other RNAs, and determined the crystal structures of 5′SLA from DENV and ZIKV (Figure 1B) [24,25,26,27]. Here, we investigate the stability of the 5′ and 3′SLA elements, their recognition by the viral proteins NS3 and NS5, and how these interactions regulate RNA synthesis and the formation of replication intermediates.

2. Materials and Methods

2.1. Expression and Purification of DENV NS5 and NS3 Proteins

DENV3 NS5 and NS3 were expressed with N-terminal hexa-histidine tags and purified as previously described [28]. DENV3 NS2B-NS3 was engineered by fusing the 47 cytoplasmic amino acids of NS2B to the N-terminus of NS3, separated by a GGGGSGGGG linker [29]. To prevent cleavage between NS2B and NS3, the active site residue His51 was mutated to Ala in NS3. All proteins were expressed in Rosetta (DE3) Escherichia coli cells. The cells were grown at 37 °C in Luria Broth (LB) containing 100 μg/mL ampicillin and 25 μg/mL chloramphenicol to an OD_600_ of 0.7–0.8. Protein expression was induced by the addition of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and cell growth continued overnight at 18 °C. For NS5 protein purification, cells were resuspended in lysis buffer containing 50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 2 mM β-mercaptoethanol, and EDTA-free protease inhibitor (Roche Applied Science, Penzberg, Germany), or 50 mM phosphate, pH 7.0, 500 mM NaCl, 1 mM DTT, and 1 μg/mL each of DNase and RNase. After cell lysis by sonication and centrifugation to remove insoluble debris, the soluble fraction was applied to Talon metal affinity column (Clontech, Mountain View, CA, USA). NS5 was eluted with a 10–150 mM imidazole gradient in 50 mM Tris-HCl, pH 7.0, 400 mM NaCl and 2 mM β-mercaptoethanol. Fractions containing NS5 proteins were concentrated and applied to a Superdex 200 size-exclusion column (GE Healthcare Life Sciences, Marlborough, MA, USA), equilibrated with buffer containing 20 mM Tris-HCl, pH 7.0, 500 mM NaCl, 2 mM dithiothreitol (DTT), and 10% glycerol. NS3 and NS2B-NS3 proteins were similarly purified using Talon metal affinity and Superdex 200 size exclusion chromatography using the buffer containing 20–50 mM Tris-HCl, pH 7.0, 500 mM NaCl, 2 mM β-mercaptoethanol or DTT, and 20% glycerol. Eluted proteins were analyzed by SDS-PAGE. DENV3 proteins were used for interaction studies with DENV2 RNAs, because the structure of DENV2 5′SLA was available, and cross-serotype SLA-NS5 interactions have been shown to be functionally compatible [26,30].

2.2. Circular Dichroism Spectroscopy

Fluorescein-labeled RNAs (5′SLA-F and 3′SLA-F) were used for CD and fluorescence measurements. DENV 5′SLA-F, corresponding to the 5′-terminal 80 nucleotides of the viral genome, and 3′SLA-F, corresponding to the 3′-terminal 73 nucleotides of the negative strand were synthesized with a 3′ and 5′ fluorescein tag, respectively (Midland Certified Reagents, Midland, TX, USA) (Table 1). The dsRNA hairpin (60 nt) and single-stranded RNA (18 nt) were also synthesized as controls (Integrated DNA Technologies, Coralville, IA, USA) (Table 1). Far-UV CD spectra of RNAs (5′SLA-F, 3′SLA-F, ssRNA, and dsRNA hairpin) were recorded using a Jasco J-815 spectrometer in CD buffer containing 10 mM phosphate, pH 7.0, 10 mM NaCl, and 0.1 mM EDTA. Measurements were performed using a 0.1 cm path length cuvette. To measure thermal stability, CD spectra were recorded every 5 °C from 5 to 95 °C by heating the cuvette at 1 °C/min. Ellipticity was measured between 200 and 290 nm at each temperature. Each experiment was performed in duplicate.

2.3. Fluorescence Measurements

Steady-state fluorescence titrations were performed using an ISS PC1 spectrofluorometer (ISS, Urbana, IL, USA). Polarizers were placed in the excitation and emission channels and set at 90° and 55° (magic angle), respectively, to minimize artifacts due to fluorescence anisotropy. Experiments were conducted by adding DENV NS5 or NS3 proteins to solutions containing fluorescein-labeled 5′SLA or 3′SLA RNA. The RNA concentration was within the linear range of fluorescence signal. All fluorescence measurements were carried out in buffer B1, composed of 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM MgCl_2_, 2 mM β-mercaptoethanol, and 10% glycerol. Each experiment was repeated three times with proteins from a single protein preparation. Binding of DENV NS5 and NS3 to 5′SLA or 3′SLA was monitored by measuring changes in fluorescein emission (λ_ex_ = 480 nm, λ_em_ = 520 nm). Titration curves were fitted using KaleidaGraph software (Synergy Software, Reading, PA, USA). The relative fluorescence change (ΔF_obs_) was defined as (F_i_ − F0)/F0, where F_i_ is the fluorescence at a given titration point and F0 is the initial fluorescence value of the sample. Both F0 and F_i_ were corrected for background fluorescence at the excitation wavelength. Fluorescence titration curves were fitted using Equations (1) and (2) to determine the binding constants for NS5–SLA and NS3–SLA interactions:

$$[eqn]$$ $$[eqn]$$

2.4. In Vitro Transcription of RNAs

The 72 nt 5′SLA was generated by in vitro transcription using the template generated by PCR using the primers in Table 2. The other RNA constructs (5′SLA_135_, 5′SLA_135_NHP, 5′SLA_100_, 5′SLA_90_, 3′SLA_135_, 3′SLA_135mut_, noSLA_135_) were generated by in vitro transcription using plasmids as templates. These T7 transcription template plasmids were generated using overlapping oligonucleotides (Table 2) that were 3′ extended by PCR to introduce the T7 promoter and restriction sites, and cloned into the pUC19 vector between HindIII and KpnI sites. Purified plasmids were linearized with KpnI prior to transcription. The RNA sequences are listed in Table 1.

T7 RNA polymerase was purified as previously described [31], and T7 transcription was carried out following a published protocol [32]. In brief, 100 μg of linearized pUC19 vector or 3 μg of PCR product was mixed with 2 μg of T7 RNA polymerase in a 10 mL reaction buffer containing 50 mM Tris-HCl, pH 7.5, 15 mM MgCl_2_, 5 mM DTT, 2 mM spermidine, and 1 mM of each NTP. Reactions were incubated at 37 °C for 3 to 6 h, after which RNase-free DNase I (New England Biolabs, Ipswich, MA, USA) was added to degrade the DNA template. In vitro transcribed RNAs were purified by Mono Q ion-exchange column (Cytiva, Wilmington, DE, USA), equilibrated with 20 mM Tris-HCl, pH 7.5 and 2 mM MgCl_2_, and eluted with a 0–1 M NaCl gradient. RNA-containing fractions were analyzed by denaturing PAGE (8 M urea, 8% acrylamide). Fractions containing desired RNA were pooled, concentrated to ~20 μM in water, and stored at −80 °C until use. The purity of the target RNA was greater than 90%, as determined by denaturing PAGE. If necessary, RNAs were further purified by denaturing PAGE (8 M urea, 8%) as described previously [33]. Following additional purification, RNA was refolded by heating to 80 °C and slowly cooled in the presence of 5 mM MgCl_2_.

2.5. Fluorescein Labeling of RNA

5′SLA was labeled with fluorescein at the 3′ end. To oxidize the 3′OH to aldehyde, 5′SLA (10 nM) was incubated with 2.5 mM NaIO_4_ and 100 mM sodium acetate, pH 5.2, followed by a 50 min incubation on ice in the dark [34]. The reaction was quenched with 6.9 mM KCl, and the RNA was precipitated using 3M sodium acetate, pH 5.2 and 100% ethanol. The pellet was then washed with 70% ethanol and dissolved in water. The RNA was labeled with fluorescein by mixing with 100 mM sodium acetate, pH 5.2, and 2 mM fluorescein-5-thiosemicarbazide (Cayman chemical) in amber tube, and the reaction mixture was gently rocked at 4 °C for 16 h. RNA was then ethanol precipitated and dissolved in water. Excessive dye was removed using NAP-5 column (Cytiva Life Sciences). RNA was precipitated with ethanol and then dissolved in water.

2.6. Electrophoretic Mobility Shift Assay (EMSA)

For binding assays between viral proteins (NS5 or NS3) and RNA, 1 μM RNA was incubated with 0.5, 1, 2, 4, or 8 μM of NS3 or NS5 in binding buffer containing 25 mM HEPES-KOH, pH 7.0, 40 mM KCl, 5% glycerol, 0.5 mM MgCl_2_, and 2 mM DTT. The mixtures were incubated for 30 min at room temperature (RT), then briefly chilled on ice before loading onto a 1.5 or 1% agarose gel prepared in 0.5× TBE buffer, pH 9.2. RNA bands were visualized using SYBR Safe (Thermo Fisher Scientific, Waltham, MA, USA) or ethidium bromide staining. To visualize protein bands, gels were stained with Coomassie staining solution.

2.7. NS5 Polymerase Assay

For NS5 polymerization reactions, in vitro transcribed RNA (0.5 μM) was mixed with NS5 (0.5 μM) in the presence of NTPs (100 μM) in reaction buffer containing 40 mM Tris-HCl, pH 7.5, 24 mM NaCl, 3 mM DTT, 25% glycerol, and 2 mM MnCl_2_ in a total volume of 50 μL. Mn^2+^ was used in the polymerase assay because it supported higher NS5 activity than Mg^2+^. Similar product patterns were observed when Mg^2+^ was used instead (data not shown). Reactions were incubated at 37 °C for 2 h. To test terminal nucleotidyl transfer activity of NS5, polymerase reactions were carried out with UTP (16.8 μM) mixed with fluorescein-12-UTP (8.2 μM, Jena Bioscience, Jena, Germany). Reactions were terminated by the addition of 5 mM EDTA, followed by phenol extraction and ethanol precipitation of RNA. RNA pellets were resuspended in 5 μL of 80% formamide and analyzed by denaturing PAGE (8% acrylamide, 8 M urea) in 1× TBE buffer. Total RNA was visualized by SYBR Safe or ethidium bromide staining, and fluorescence from incorporated fluorescein was detected using a Bio-Rad ChemiDoc XRS+ imaging system (Bio-Rad, Hercules, CA, USA).

2.8. Cryo-Electron Microscopy of the Annealed 5′SLA135 and 3′SLA135

The 5′SLA_135_ and 3′SLA_135_ were mixed in equimolar concentration (10 µM each) in the presence of 5 mM MgCl_2_ and annealed. Quantifoil^®^ (Electron microscopy sciences, Hatfield, PA, USA) R 2/1 (200 mesh) holey carbon grids glow-discharged for 1 min were used for cryo-EM sample preparation. The annealed RNA sample (3.5 µL) was applied to the grids at 20 °C and 100% humidity, blotted and plunge frozen in liquid ethane using Vitrobot Mark IV (Thermo Scientific, Waltham, MA, USA). Grids were imaged on the Talos Arctica 200 KeV electron cryo-microscope (Thermo Scientific, Waltham, MA, USA), equipped with a Falcon-3 Direct electron detector in the counting mode at Indiana University.

2.9. RNA Helicase Assay and Data Analysis

RNA substrates used in helicase assays were prepared by annealing synthetic RNA oligonucleotides (Integrated DNA Technologies, Coralville, IA, USA). An 18 nt RNA oligonucleotide labeled with fluorescein at the 3′ end (5′-GCCUCGCUGCCGUCGCCA-fluorescein-3′) was used to generate dsRNA substrates containing either 18 or 10 nt 3′ overhangs. The substrate with an 18 nt 3′ overhang was formed by annealing the labeled RNA with a 36 nt complementary strand (3′-ACGAGGGAGACGAGGAGACGGAGCGACGGCAGCGGU-5′). The substrate with 10 nt 3′ overhang was generated similarly by annealing to a 28 nt complementary strand (3′-GACGAGGAGACGGAGCGACGGCAGCGGU-5′). Oligonucleotides were mixed, heated to 90 °C, and slowly cooled to room temperature over 2 h to promote duplex formation. Helicase assays were carried out at 37 °C in a final volume of 20 μL. RNA substrates (10 nM) and NS3 protein (200 nM) were pre-incubated for 10 min in reaction buffer (25 mM HEPES-KOH, pH 7.0, 40 mM KCl, 5% glycerol, 0.5 mM MgCl_2_, 2 mM DTT, and 0.01 mM EDTA). Unwinding reactions were initiated by the addition of 2 mM ATP and 50 nM unlabeled RNA trap to prevent re-annealing of the probe. Reactions were performed at 37 °C in a water bath and incubated for time points ranging from 30 s to 120 min. Reactions were quenched by the addition of 0.2 M EDTA and resolved on 20% polyacrylamide gels in TBE buffer. Fluorescent signals from fluorescein-labeled RNA were detected by scanning the gels using a UVP ChemStudio imaging system (Analytik Jena, Upland, CA, USA). The amounts of ssRNA and dsRNA were quantified based on the intensity of the corresponding bands using GelAnalyzer 19.1 software (http://www.gelanalyzer.com). Observed rate constants and amplitudes (k_obs_ and A, respectively) for RNA unwinding were determined by fitting the time-course data to a single exponential function of time (t):

$$[eqn]$$

Helicase assays using the fluorescein-labeled 5′SLA and unlabeled 3′SLA duplex (10 nM) were carried out under the conditions described above, except that 40 mM KCl was omitted and a 10-fold molar excess of RNA trap was included.

3. Results

3.1. 3′SLA Adopts a Stable dsRNA Structure

The 3′ end of the negative-strand RNA contains a sequence complementary to the 5′SLA and is predicted to form a stem–loop structure. We modeled the secondary structure of DENV2 3′SLA based on the base pairing pattern observed in the crystal structure of DENV2 5′SLA (Figure 1B), consisting of a top stem, side loop, and bottom stem [26]. The 3′SLA structure shared a similar bottom stem but exhibited a reversed orientation of the side loop and top stem compared to the 5′SLA (Figure 1B). The calculated free energies (ΔG) for the 5′SLA and 3′SLA structures using the RNAeval program [35] were −20.2 and −13.6 kcal/mol, respectively, indicating that 3′SLA was likely less thermodynamically stable. The reduced stability predicted for 3′SLA could be attributed to the absence of G•U wobble base pairs present in 5′SLA, as well as differences in directional base-pair stacking energies used in the nearest neighborhood energy method [36].

To assess whether the 3′SLA in the negative-strand RNA adopted a stable secondary structure, we performed circular dichroism (CD) spectroscopy [37]. Two RNAs were synthesized, one comprising the first 80 nt at the 5′ end of the genome (5′SLA) and the other comprising the last 73 nt at the 3′ end of the negative strand (3′SLA) (Table 1). In addition, a 60 nt dsRNA hairpin and 19 nt ssRNA were synthesized as positive and negative controls, respectively (Table 1). CD spectra were recorded at 5 °C in CD buffer (10 mM phosphate pH 7.0, 10 mM NaCl, and 0.1 mM EDTA). Both dsRNA and ssRNA controls displayed expected CD spectra [37]. The dsRNA control showed a CD spectrum with a maximum peak around 260–270 nm, while the ssRNA control exhibited no prominent peaks above 250 nm (Figure 2A). Both 5′SLA and 3′SLA RNAs produced CD spectra resembling that of the dsRNA control, with maximum peaks at 267 and 264 nm, respectively, indicating that both RNAs formed stable dsRNA structures (Figure 2B,C).

We next determined the thermal stability of 5′SLA and 3′SLA. CD spectra of 5′SLA, 3′SLA and two controls were measured at 5 °C intervals from 5 °C to 95 °C in CD buffer. As the temperature increased, the peak near 260–270 nm in the dsRNA control gradually decreased in intensity and shifted towards the longer wavelength (red shift), suggesting that the A-form helical RNA structure is disrupted (Figure 2A). In contrast, the ssRNA control showed no major peak above 250 nm across the temperature range, as expected (Figure 2A). Both 5′SLA and 3′SLA exhibited CD spectral changes similar to the dsRNA control, including a red shift with increasing temperature (Figure 2B,C). To quantify thermal stability, we calculated the melting temperatures (T_m_) as the mid-log of the transition phase by plotting the CD signal at 267 nm for 5′SLA and 264 nm for 3′SLA, the wavelength with maximal peak signal, against temperature (Figure 2B,C). The resulting melting temperatures were 51 °C for 5′SLA and 55 °C for 3′SLA. These results indicated that 3′SLA formed a structure at least as stable as that of 5′SLA.

3.2. NS5 Preferentially Binds to 5′SLA in the Positive Strand over 3′SLA in the Negative Strand

During viral replication, the NS3 helicase is proposed to unwind the dsRNA intermediate, exposing the 5′ end of the positive strand and the 3′ end of the negative strand. This unwinding enables the NS5 polymerase to bind the 3′ end of the negative strand and initiate synthesis of a new positive strand (Figure 1A). At the terminus of the separated dsRNA intermediate, the 5′SLA and 3′SLA structures are expected to fold independently (Figure 1). Orthoflaviviral NS5 specifically recognizes 5′SLA on the positive strand [15,16,17], but it remains unclear whether NS5 also recognizes 3′SLA on the negative strand. To address this, we measured DENV NS5 binding to 5′SLA and 3′SLA using a fluorescence titration assay [24,25]. The 5′SLA and 3′SLA RNAs were labeled with fluorescein at their 3′ and 5′ ends, respectively (5′SLA-F and 3′SLA-F, Table 1). Binding of DENV NS5 to 5′SLA resulted in a decrease in fluorescence intensity, with a maximum quenching of ~15% (Figure 3A), as previously reported [24]. The binding constant was determined using nonlinear least-squares analysis (see Methods). DENV NS5 bound 5′SLA with a binding constant (K_5′SLA_) of 5.5 (±0.8) × 10^6^ M^−1^ (dissociation constant K_d_ = 0.18 ± 0.03 μM). This value was consistent with the previously reported affinities for DENV and ZIKV NS5 binding to 5′SLA [24,25]. In comparison, NS5 binding to 3′SLA led to a larger fluorescence quenching, with a maximum decrease of ~42%, likely due to the shorter distance between the SLA structure and the fluorophore (Figure 3A). DENV NS5 bound 3′SLA with a binding constant (K_3′SLA_) of 2.5 (±0.4) × 10^6^ M^−1^ (K_d_ = 0.4 ± 0.06 μM). Thus, NS5 bound 5′SLA with ~2.2-fold higher affinity than 3′SLA, indicating a preference for the positive-strand RNA promoter.

Next, we evaluated NS5 interactions with SLA elements using electrophoretic mobility shift assays (EMSAs) with in vitro-transcribed RNAs containing 5′SLA or 3′SLA. First, RNAs corresponding to the first 72 nts of the DENV genome (5′SLA) and the last 72 nts of the negative strand (3′SLA) were used to assess their binding to NS5 (Figure 3B and Table 1). Each RNA (1 μM each) was incubated with increasing NS5 concentrations (0.5- to 8-fold molar excess). NS5 bound 5′SLA_72_ with higher affinity than 3′SLA_72_, as indicated by a clear shift in input RNA at the lowest NS5 concentration tested (0.5 μM) (Figure 3B and Figure S1). With increasing NS5 concentrations, the RNA-protein complexes shifted further, forming higher molecular weight species for both 5′SLA and 3′SLA. During viral replication, NS5 would interact with 5′SLA or 3′SLA within longer RNA molecules. To better mimic these interactions, we next used in vitro-transcribed RNAs corresponding to the first 135 nts of the DENV genome (5′SLA_135_) and the last 135 nts of the negative strand (3′SLA_135_) (Table 1 and Figure 3C). A 135 nt RNA lacking the SLA secondary structure (noSLA_135_) was used as a control in EMSA. The gel shift data again showed that NS5 bound 5′SLA_135_ with higher affinity than 3′SLA_135_ (Figure 3C), consistent with the fluorescence titration data. Interaction with 5′SLA_135_ produced at least three shifted bands, including an initial fast migrating band (marked with asterisk on Figure 3C). NS5 forms multiple complexes with 5′SLA-containing RNAs [26,30,38]. NS5 can engage single-stranded RNA in its template-binding channel or recognize the 5′SLA promoter on the thumb subdomain [24,26,39]. Accordingly, NS5 likely forms multiple complexes with 5′SLA_135_ by interacting with both the 5′SLA structure and the adjacent single-stranded region. In contrast, 3′SLA_135_ and noSLA_135_ exhibited similar binding patterns, and the faster-migrating complex observed with 5′SLA_135_ was absent. These results suggested that NS5 did not engage the 3′SLA structure specifically, and its binding affinity for 3′SLA resembled that for nonspecific RNA.

3.3. NS3 Does Not Distinguish Between 5′SLA or 3′SLA for Binding

We next examined whether DENV NS3 distinguishes between 5′SLA and 3′SLA by measuring its binding affinities to fluorescein-labeled 5′SLA and 3′SLA. Fluorescence titrations with DENV NS3 were carried out similarly as described above (Figure 4A). Binding of NS3 to labeled 5′SLA resulted in a maximum fluorescence quenching of ~12%, yielding a binding constant (K_5′SLA_) of 0.9 (±0.2) × 10^6^ M^−1^ (K_d_ = 1.1 ± 0.25 μM) (Figure 4A). NS3 binding to labeled 3′SLA caused a larger fluorescence quenching with the maximum value of ~45% (Figure 4A). DENV NS3 bound 3′SLA with a binding constant (K_3′SLA_) of 2.1 (±0.3) × 10^6^ M^−1^ (K_d_ = 0.48 ± 0.07 μM). Although the calculated affinity of NS3 for 3′SLA was approximately two-fold higher than for 5′SLA, both values were comparable to previously reported NS3 binding affinities for short 5–10 nt RNAs [40].

We further examined NS3 interactions with in vitro transcribed viral RNAs, 5′SLA_135_, 3′SLA_135_, and noSLA_135_ using EMSA. All three RNAs, 5′SLA_135_, 3′SLA_135_, and noSLA_135_ interacted similarly with DENV NS3, suggesting that NS3 primarily recognized single-stranded regions present in each RNA rather than the folded SLA (Figure 4B). Taken together, the EMSA results were consistent with the fluorescence titration data and indicated that NS3 lacks structural specificity in binding to SLA elements.

3.4. DENV NS5 Can Synthesize RNA from 5′SLA or 3′SLA-Containing RNA Templates

Orthoflavivirus NS5 uses the genomic 5′SLA as a promoter to initiate negative-strand RNA synthesis [15,16,41]. In contrast, it is not clear whether positive-strand RNA synthesis requires an RNA promoter. During positive-strand synthesis, the negative strand of the dsRNA intermediate serves as the template, and 5′SLA and 3′SLA structures are likely to form as a result of the dsRNA unwinding. To determine whether the structure of 5′SLA or 3′SLA influences RNA synthesis by NS5, we performed polymerase reactions using 5′SLA_135_, 3′SLA_135_ or noSLA_135_ as RNA templates. The RNA templates were incubated with DENV NS5 for 2 h and RNA products were analyzed by denaturing PAGE [42] (Figure 5A).

Orthoflavivirus NS5 can synthesize RNAs that are either template-length through de novo initiation or twice the template length through elongation [22,43,44]. The elongation product is generated by a copy-back mechanism, in which NS5 polymerase extends a fold-back RNA template. The 5′SLA_135_ and noSLA_135_ templates are predicted to form a 5 bp hairpin at their 3′ end (fold-back RNA), which could support elongation reaction (Figure 5A). When NS5 was incubated with 5′SLA_135_, four major RNA products were observed, ranging in size from ~80 to 250 nt (products 1–4 in Figure 5A, left). Based on their size, products 1 and 2 likely represented elongation products. Product 1 (~250 nt) was consistent with the calculated length of the fully elongated product (253 nt), whereas the smaller product 2 (~180 nt) likely reflected incomplete elongation (Figure 5A). To test this, we performed an NS5 polymerase assay using a mutant 5′SLA_135_ construct in which the 3′-end hairpin was disrupted (5′SLA_135_NHP, Table 1 and Figure 5B). The polymerase reaction using the 5′SLA_135_NHP template resulted a significant reduction in RNA products larger than the template, whereas shorter RNA products remained unaffected. This indicates that the higher molecular weight bands from the 5′SLA_135_ template (products 1 and 2 in Figure 5A) primarily resulted from elongation initiated at the 3′ hairpin. We next tested whether the incomplete elongation product 2 (~180 nt) arose from NS5 stalling at the 5′SLA structure and failing to synthesize full-length RNA (Figure 5A). To test this idea, we designed two additional templates with reduced spacing between the 3′ hairpin and 5′SLA: one containing a 10 nt spacer (5′SLA_100_) and another lacking a spacer, in which the 3′ hairpin was directly adjacent to the 5′SLA (5′SLA_90_) (Table 1 and Figure 5C). As expected, NS5 produced a shorter elongation product from 5′SLA_100_, migrating slightly above the input RNA, and no elongation products from 5′SLA_90_. The results indicated that the proximity of the 5′SLA structure to the 3′ end of the template impaired elongation by NS5 and that the 5′SLA structure could act as a physical barrier to RNA synthesis.

Product 3 from the polymerase reaction from the 5′SLA_135_ template was similar in size to the template RNA, which could result from either de novo initiation, terminal nucleotidyl transfer, or both. Terminal nucleotidyl transferase activity, which adds non-templated nucleotides at the 3′ end, is commonly observed in RNA polymerases [45,46]. To determine whether DENV NS5 possesses such activity, we performed a polymerase reaction using only fluorescein-labeled UTP (Figure 5A). Fluorescence imaging showed a prominent band corresponding to the template size, indicating that NS5 can catalyze template-independent terminal uridylylation (U^Flu^ lane in Figure 5A). Thus, the template-sized product likely reflected both de novo RNA synthesis and the terminal nucleotidyl transferase activity of NS5. Product 4 (~90 nt), which is shorter than the input template, could result from either template degradation or de novo RNA initiation. The addition of RNase inhibitor had no effect on its formation, suggesting that product 4 did not arise from nuclease-mediated degradation. Interestingly, product 4 was absent when either the noSLA_135_ template (which lacks the 5′SLA structure) or shorter 5′SLA_90_ RNA was used (Figure 5A,C). This suggested that the structure or spatial arrangement of the SLA contributed to its generation. These observations indicate that product 4 likely represents a de novo RNA product initiated at the 3′ end of the template but prematurely terminated, possibly due to stalling of NS5 at the 5′SLA structure.

When 3′SLA_135_ was used as the template in the polymerase reaction, three major RNA products (1, 2, and 3 in the middle panel of Figure 5A) were observed, ranging in size from ~80 to 200 nt. Product 1 (~190 nt) likely represented an elongation product and matched the predicted size expected for a copy-back mechanism initiated at the 3′ end (Figure 5A). The 3′ end of the 3′SLA stem–loop structure could serve as an elongation primer. Product 2 corresponded to the template size (~135 nt), while product 3 (~80 nt) was smaller than the input RNA. The mechanism generating product 3 remains unclear. One possibility is that NS5 initiates RNA synthesis at a single-stranded region immediately downstream of the 3′SLA. This would be similar to observations in hepatitis C virus, where the NS5B polymerase can initiate RNA synthesis on circular RNA templates [47]. Thus, despite its lower binding affinity for 3′SLA, NS5 efficiently carried out RNA polymerization on 3′SLA-containing template. Taken together, these results indicate that NS5 can utilize both 5′SLA- and 3′SLA-containing RNAs for RNA synthesis.

Next, we used noSLA_135_ as a template to determine whether the absence of 5′SLA or 3′SLA influences RNA synthesis by NS5. The noSLA_135_ template lacks a secondary structure but contains a predicted hairpin at the 3′ end that can serve as a primer for elongation (Table 1). NS5 predominantly generated elongation products that are longer than the template, including species exceeding three times the template length (~500 nt), suggesting multiple rounds of elongation (Figure 5A, right). Thus, while NS5 was capable of synthesizing RNA from templates lacking SLA elements, the presence of 5′SLA and 3′SLA appeared to inhibit aberrant elongation under in vitro conditions, thereby promoting de novo RNA synthesis.

3.5. 5′SLA and 3′SLA Are Fully Annealed in the dsRNA Replication Intermediate

Orthoflavivirus NS3 possesses helicase activity that unwinds dsRNA in the 3′ to 5′ direction and is thought to unwind the dsRNA replicative intermediate. However, NS3 requires a 3′-overhang for helicase activity and is therefore unable to unwind the blunt-ended dsRNA intermediate (closed dsRNA intermediate in Figure 6A) [48,49,50]. Alternatively, the positive- and negative-strand termini may not be completely double-stranded due to the formation of highly stable RNA structures, such as 5′SLA and 3′SLA, and could ‘breathe’ to allow strand separation at the termini (open dsRNA intermediate in Figure 6A) [51]. To investigate the nature of the dsRNA ends, we generated an RNA duplex mimicking the end of the replication intermediate by annealing complementary 5′SLA_135_ and 3′SLA_135_. The two RNAs readily formed a duplex even at room temperature, indicating that they were at least partially annealed (Figure 6B). To test whether the 5′SLA or 3′SLA were folded in this model replication intermediate, we performed a DENV NS5 binding assay. If the 5′SLA is folded in the model replication intermediate, NS5 would be expected to bind the duplex as effectively as it binds 5′SLA_135_. Compared to its interaction with 5′SLA_135_, NS5 showed significantly reduced binding to the duplex RNA (Figure 6C). This suggested that the 5′SLA structure was not retained and thus the two strands were fully base-paired, forming a closed dsRNA intermediate (Figure 6A). These findings were further supported by cryo-electron microscopy (cryo-EM) imaging of the 5′SLA-3′SLA duplex. The individual images showed linear structures of ~350 Å, consistent with fully base-paired 135nt dsRNA (Figure 6D).

We next tested whether NS3 could unwind the model dsRNA intermediate. If the dsRNA intermediate is fully base-paired, NS3 would be unable to separate the 5′SLA_135_-3′SLA_135_ duplex (Figure 6E). As a control, we first evaluated NS3 helicase activity using two dsRNA substrates. The 18 nt RNA labeled with fluorescein at the 3′ end was annealed with either 36 or 28 nt RNA, generating dsRNA with 18 or 10nt 3′ overhangs, respectively (18bp-18ss and18bp-10ss substrates Figure 6E). Helicase activity was measured under single-turnover conditions, with NS3 provided in 20-fold molar excess over the RNA substrate. An unlabeled 18nt RNA was included as a trap to prevent reannealing of the separated strands. NS3 and the substrate were preincubated for 10 min, and the reaction was initiated by the addition of ATP and the RNA trap (Figure 6E). The unwound, labeled RNA strand was quantified over time, and the data were fitted to a single-exponential decay (see Method). DENV NS3 unwound the dsRNA containing 18nt or 10nt 3′ overhangs (18bp-18ss or -10ss) with an observed rate constants of k_unw_ = 0.60 ± 0.04 min^−1^ and 0.44 ± 0.08 min^−1^, respectively (Figure 6E). Thus, NS3 helicase activity decreased as the 3′ overhang shortens from 18 to 10 nt. These results are consistent with the previous reports showing that the length of the 3′ overhang influences NS3 helicase activity with 12–18 nt required for optimal unwinding [13,52].

We next tested NS3 helicase activity on the dsRNA, composed of fluorescein-labeled 5′SLA_135_ annealed to unlabeled 3′SLA_135_ (Figure 6F). Unlabeled 5′SLA_135_ was used as a trap. NS3 was unable to unwind the SLA duplex, suggesting that the 3′ ends of both strands are inaccessible to the helicase. NS3 was unable to unwind the SLA duplex, suggesting that the 3′ ends of both strands were inaccessible to the helicase. To assess whether NS5 binding to 5′SLA could promote strand separation, we co-incubated NS5 with 5′SLA_135_ prior to forming a duplex with 3′SLA_135_ (Figure 6F). However, NS3 remained unable to unwind the duplex. During viral replication, NS3 forms a complex with its cofactor NS2B for the proteolytic activity [3,4,53]. Although the isolated helicase domain of NS3 exhibits unwinding activity in vitro [29], its interaction with NS2B may enhance its processivity. To test this, we generated an NS2B-NS3 fusion protein in which NS3 is N-terminally linked to the 47 cytosolic residues of NS2B via a flexible linker [29]. While the NS2B-NS3 fusion protein exhibited helicase activity comparable to NS3 alone on the 18bp-18ss substrate, the protein was unable to unwind the SLA duplex (Figure 6F).

Next, we generated an open duplex at one end of the 5′SLA-3′SLA dsRNA by introducing mutations in the last 12 nt of 3′SLA_135_, preventing base-pairing with 5′SLA_135_ (Figure 6G). Despite the presence of a 3′ ssRNA stretch, NS3 was unable to unwind this mutant duplex, suggesting that the strand separation is limited by NS3 processivity. Collectively, these results indicated that neither SLA structure (5′SLA or 3′SLA) was likely to form in the replication intermediate in the absence of additional cofactors. Furthermore, DENV NS3, whether alone or in complex with NS5 or NS2B, was insufficient to unwind the model replication intermediate.

4. Discussion

NS5 catalyzes both negative- and positive-strand RNA synthesis, but the mechanism of positive-strand RNA synthesis remains poorly understood. During viral replication, positive-strand RNA is produced at 10- to 50-fold higher levels than negative-strand RNA [54], and thus NS5 polymerase spends the majority of its time synthesizing positive-strand RNA. The requirements for positive- and negative-strand synthesis differ with respect to their templates (dsRNA vs. genomic ssRNA), products (single-stranded positive-strand RNA vs. dsRNA), and downstream modifications (type 1 cap formation vs. no modification for negative-strand). It is currently unclear how NS5 recognizes the negative strand as a template in the context of a dsRNA intermediate. During positive-strand RNA synthesis, the NS3 helicase is thought to unwind the dsRNA replicative intermediate, exposing the 3′ end of the negative strand. This exposed terminus then serves as the template for NS5 polymerase to synthesize new RNA. Given that synthesis initiates at the 3′ end of the negative strand, any RNA structures present at this terminus are likely to influence the process. To investigate this, we characterized the structure of DENV 3′SLA at the 3′ end of the negative strand, which is complementary to the 5′SLA promoter in the positive-strand genome, and examined the binding and enzymatic activities of NS5 and NS3 with 3′SLA.

CD analysis showed that DENV 3′SLA adopts a stable tertiary structure. The thermal stability of 3′SLA is comparable to that of 5′SLA, with 3′SLA exhibiting a slightly higher melting temperature (55 °C vs. 51 °C). Because both 5′SLA and 3′SLA form stable RNA folds, we examined whether viral NS5 and NS3 proteins differentially recognize the SLA RNAs. Both fluorescence-binding assays and EMSA indicated that NS5 has higher affinity for 5′SLA than for 3′SLA. This higher affinity supports the model in which 5′SLA acts as a promoter element for NS5-mediated synthesis of the negative-strand RNA [15,16,55]. Current studies for DENV 3′SLA align with recent findings for ZIKV 3′SLA, which adopts a folded stem–loop structure [56]. ZIKV NS5 also exhibits roughly two-fold higher binding to the 5′SLA than to 3′SLA. In contrast, NS3 showed similar binding to 5′SLA, 3′SLA, and non-specific RNA, indicating that it does not specifically recognize either SLA structure. This implies that NS3 primarily interacts with the 3′ ends of single-stranded RNAs in a sequence- or structure-independent manner. Previous structural studies have proposed that DENV NS3 recognizes the first 12 nucleotides of the viral genome [57]. However, we did not observe specific interactions between NS3 and 5′SLA, likely due to differences between the linear 12 nt ssRNA used in structural studies and the folded conformation of the 5′SLA.

Next, we tested whether NS5 polymerase activity is influenced by the presence of 5′SLA or 3′SLA structures within RNA templates. Polymerase assays were performed using 135 nt RNA templates corresponding to either the 5′ end of the positive strand (5′SLA_135_), the 3′ end of the negative strand (3′SLA_135_), or a control lacking the SLA structure (noSLA_135_). NS5 was able to utilize both 5′SLA_135_ and 3′SLA_135_ as templates. However, NS5 also generated smaller, aborted RNA products, suggesting that the polymerase alone could not efficiently unfold the stable secondary structures of 5′SLA or 3′SLA. In contrast, when noSLA_135_ RNA was used as a template, NS5 generated long elongation products (>2–3 times the template length). These results indicate that the presence of SLA elements in RNA templates limits RNA synthesis by elongation, likely promoting de novo initiation. Although 5′SLA in the viral genome (positive strand) serves as an RNA promotor for negative-strand RNA synthesis [41], it is unclear whether a similar promoter is required for positive-strand RNA synthesis. Our findings suggest that positive-strand RNA synthesis from the 3′SLA-containing template does not require a promoter, at least for 135 nt long RNA. This is consistent with previous reports that WNV NS5 synthesizes RNA from the 3′-terminal 230 nts of the negative strand, which includes the 3′SLA element [58].

Finally, we investigated the nature of the dsRNA ends using an RNA duplex that mimics the replication intermediate and examined whether NS3 helicase could unwind this substrate. Although NS3 is widely assumed to unwind the dsRNA intermediate during positive-strand RNA synthesis, previous studies have shown that NS3 cannot separate blunt-ended dsRNA. Instead, NS3 requires a 3′ single-stranded overhang to initiate helicase activity [48,49,50]. Based on cryo-EM images and the NS5 binding assay, the dsRNA formed by annealing complementary 5′SLA_135_ and 3′SLA_135_ strands is largely double-stranded. NS5 does not bind to the duplex, suggesting that the 5′SLA structure is not present. Further, recombinant NS3 protein alone was unable to unwind this dsRNA duplex. To mimic strand ‘breathing’ at the terminus of the dsRNA intermediate, mutations were introduced at the 3′ end of 3′SLA_135_ to prevent annealing with 5′SLA_135_. However, NS3 still did not unwind the mutated duplex, suggesting that it lacks the processivity to separate a 135 bp duplex. This observation is consistent with previous reports that NS3 exhibits low processivity on long (>36 bp) RNA duplexes [59]. Thus, under the in vitro conditions tested, a 135 nt RNA from the 5′ end of the DENV genome and its complementary negative strand form a fully annealed duplex, which NS3 helicase cannot unwind. These findings suggest that, in virus-infected cells, additional factors are likely required to initiate positive-strand RNA synthesis. In vivo, the termini of the viral replicative intermediate may be protein-associated within membrane-bound replication complexes, where host helicases or viral cofactors could promote local strand separation.