Perturbation in the Conserved Methyltransferase-Polymerase Interface of Flavivirus NS5 Differentially Affects Polymerase Initiation and Elongation

ABSTRACT

The flavivirus NS5 is a natural fusion of a methyltransferase (MTase) and an RNA-dependent RNA polymerase (RdRP). Analogous to DNA-dependent RNA polymerases, the NS5 polymerase initiates RNA synthesis through a de novo mechanism and then makes a transition to a processive elongation phase. However, whether and how the MTase affects polymerase activities through intramolecular interactions remain elusive. By solving the crystal structure of the Japanese encephalitis virus (JEV) NS5, we recently identified an MTase-RdRP interface containing a set of six hydrophobic residues highly conserved among flaviviruses. To dissect the functional relevance of this interface, we made a series of JEV NS5 constructs with mutations of these hydrophobic residues and/or with the N-terminal first 261 residues and other residues up to the first 303 residues deleted. Compared to the wild-type (WT) NS5, full-length NS5 variants exhibited consistent up- or downregulation of the initiation activities in two types of polymerase assays. Five representative full-length NS5 constructs were then tested in an elongation assay, from which the apparent single-nucleotide incorporation rate constant was estimated. Interestingly, two constructs exhibited different elongation kinetics from the WT NS5, with an effect rather opposite to what was observed at initiation. Moreover, constructs with MTase and/or the linker region (residues 266 to 275) removed still retained polymerase activities, albeit at overall lower levels. However, further removal of the N-terminal extension (residues 276 to 303) abolished regular template-directed synthesis. Together, our data showed that the MTase-RdRP interface is relevant in both polymerase initiation and elongation, likely with different regulation mechanisms in these two major phases of RNA synthesis.

IMPORTANCE The flavivirus NS5 is very unique in having a methyltransferase (MTase) placed on the immediate N terminus of its RNA-dependent RNA polymerase (RdRP). We recently solved the crystal structure of the full-length NS5, which revealed a conserved interface between MTase and RdRP. Building on this discovery, here we carried out in vitro polymerase assays to address the functional relevance of the interface interactions. By explicitly probing polymerase initiation and elongation activities, we found that perturbation in the MTase-RdRP interface had different impacts on different phases of synthesis, suggesting that the roles and contribution of the interface interactions may change upon phase transitions. By comparing the N-terminal-truncated enzymes with the full-length NS5, we collected data to indicate the indispensability to regular polymerase activities of a region that was functionally unclarified previously. Taken together, we provide biochemical evidence and mechanistic insights for the cross talk between the two enzyme modules of flavivirus NS5.

INTRODUCTION

The Flavivirus genus of family Flaviviridae includes more than 70 viruses which often cause human encephalitis and hemorrhagic diseases. The majority of the flaviviruses are transmitted by an infected mosquito or tick, including the well-known dengue virus (DENV), yellow fever virus (YFV), West Nile virus (WNV), Japanese encephalitis virus (JEV), and tick-borne encephalitis virus (TBEV). Currently, vaccines are lacking for DENV and WNV, and there is an urgent need for new antiflavivirus drugs. The 10- to 11-kb flavivirus RNA genome is positive sense and contains a single open reading frame (ORF) that is translated into a large polyprotein, from which, subsequently, three structural and seven nonstructural proteins are produced through proteolytic processing. Among those, the largest and most conserved nonstructural protein 5 (NS5) plays essential roles in viral genome capping and replication processes through its N-terminal S-adenosyl-l-methionine (SAM)-dependent methyltransferase (MTase) and C-terminal RNA-dependent RNA polymerase (RdRP), respectively. As a key player in the formation of the cap 1 structure (^N7MeG5′-ppp-5′A_2′OMe) at the 5′ end of the flavivirus genome (1), the MTase is believed to catalyze both guanine N7 and 2′-O methylation steps and may possess guanylyltransferase (GTase) activity in the formation of the G5′-ppp-5′A linkage (2, 3). As with its counterpart NS5B from the Hepacivirus genus (type species, hepatitis C virus [HCV]) or the Pestivirus genus (type species, bovine viral diarrhea virus [BVDV]) of the family Flaviviridae, the flavivirus RdRPs carry out de novo RNA synthesis in viral genome replication (4). Structurally, the thumb domain of these and other de novo RdRPs contains a priming element that reaches the vicinity of the active site to facilitate initiation (5–8). Similar to DNA-dependent RNA polymerases (9, 10), an initiation complex (IC) of de novo RdRPs is unstable and the polymerase needs to make a transition to become a stable and processive elongation complex (EC) in order to complete the synthesis of genome-length products (11).

Although the evolutionary linkage between MTase and RdRP of flavivirus and the regulation between these enzyme modules have been attractive topics in the field, structural and in vitro functional studies have been largely limited to characterization of MTase or RdRP individually. The crystal structures of the ≈260-residue flavivirus MTase domain and the ≈630-residue RdRP region have been reported in isolation (12–16). The aforementioned methylation and guanylyltransfer activities of the MTase have been analyzed in multiple flaviviruses (2, 3, 12, 17, 18). To date, the full-length NS5 has not been shown to behave differently from the MTase domain alone in capping activities. Similarly, the RdRP region itself has been documented to carry out polymerase activities in various types of assays (15, 16, 19). Although different levels of polymerase activities of the RdRP region and the full-length NS5 have been reported, the impact of the MTase domain on the polymerase function remains unclear, in part due to controversial observations and nonuniformity of polymerase assays (16, 20).

Recently, we reported the first crystal structure of a full-length NS5 from JEV. The structure revealed a medium-size MTase-RdRP interface that features a six-residue hydrophobic network in the central region and a GTR sequence at the edge, both of which are highly conserved in flaviviruses (6). Despite the seemingly independent activities of both enzyme modules of NS5 described above, this structure has provided explicit clues in further dissection of the functional regulation between MTase and RdRP. By introducing point mutations of the conserved residues in the MTase-RdRP interface in the context of reverse genetics systems for both JEV and DENV serotype 2 (DENV-2), we recently collected evidence that the interface does play essential roles for flavivirus replication (21), thus functionally validating the biological relevance of the intramolecular interface of NS5 observed in the full-length crystal structure. With an aim to gain further mechanistic insights into whether and how the interface regulates the polymerase function, we designed point mutations in the hydrophobic network and/or progressive deletions in the N-terminal one-third of NS5 to alter the MTase-RdRP interactions. We established a primer-dependent assay and a de novo assay, both of which were used to assess polymerase initiation activities, and we used a high-salt chase experiment to characterize polymerase elongation. Our data showed that perturbation into the interface affected both initiation and elongation of RNA synthesis, but the actual effects brought by the mutations on each phase were different, or even opposite in a couple of cases. Moreover, the N-terminal extension (residues 276 to 303) appeared to be indispensable for normal polymerase activity, in contrast to the MTase and linker regions (residues 1 to 265 and 266 to 275, respectively) that are not required for NS5 to retain the capability in both primer-dependent and de novo RNA synthesis.

MATERIALS AND METHODS

Cloning and protein expression.

Six full-length NS5 constructs with point mutations M_113, M_113/121, M3, R_351/467, R3, and M3-R3 were made using the QuikChange site-directed mutagenesis method and pET26b-Ub-JEV-NS5 as the parental plasmid (6, 22). The N-terminal deletion mutants D261, D273, D303 were made by using a site-directed, ligase-independent mutagenesis method (SLIM) with WT pET26b-Ub-JEV-NS5 as the parental plasmid. The N-terminal deletion constructs with point mutations D261_R3, D273_R3, and D303_R3 were made by using the SLIM method and pET26b-Ub-JEV-NS5-R3 as the parental plasmid (23). The resulting plasmids were transformed into Escherichia coli strain BL21(DE3)/pCG1 for expression of NS5 variants with a C-terminal Gly-Ser-Ser-Ser-His₆ tag as previously described (6, 24). Cell growth and protein induction procedures were previously described except that the temperature for overnight culture growth was 30°C and for the N-terminal deletion constructs, the postinduction time before harvesting was 8 h.

Purification of JEV NS5 and its variants.

Cell lysis, subsequent purification, and storage procedures were as previously described (6), except that a HiTrap Q HP column (GE Healthcare) was used in the second chromatographic purification step for the D303 and D303-R3 constructs. This column was equilibrated with a buffer containing 50 mM NaCl, 25 mM Tris (pH 9.0), 0.1 mM EDTA, 20% (vol/vol) glycerol, and 0.02% (wt/vol) NaN₃, and the protein was eluted on a linear gradient to 620 mM NaCl. Briefly, resuspended cells were lysed by passage through an AH-2010 homogenizer (ATS Engineering Ltd.), and the lysate were supplemented with IGEPAL CA-630 (Sigma-Aldrich) and polyethylenimine (PEI) prior to centrifugation for 40 min at 17,000 rpm in an SS-34 rotor (Thermo Scientific). The NS5 protein was purified sequentially using Ni affinity, ion-exchange, and gel filtration chromatography. Purified protein samples were supplemented with Tris-(2-carboxyethyl)phosphine (TCEP), concentrated, flash-frozen in 5-, 10-, or 20-μl aliquots, and stored at −80°C for single use. The final buffer was 300 mM NaCl, 5 mM MES (pH 6.0), 20% (vol/vol) glycerol, 0.02% (wt/vol) NaN₃, and 5 mM TCEP. The molar extinction coefficient for each NS5 construct was calculated based on the protein sequence using the ExPASy ProtParam program (http://www.expasy.ch/tools/protparam.html). The yield was in the range of 1 to 15 mg of pure protein per liter of bacterial culture.

RNA preparation.

Template strand RNAs were obtained by in vitro T7 RNA polymerase transcription using a parental plasmid pRAV23 and approaches modified from protocols described previously (25, 26). Briefly, synthetic DNAs (Integrated DNA Technologies) that contained sequences to generate the target RNAs were used in PCRs to generate T7-promoter-containing DNA constructs, which in turn were used to generate RNA transcripts in 10-ml T7 RNA polymerase transcription reaction mixtures. Because the RNA transcript contained a glmS ribozyme in its 3′ portion, glucosamine-6-phosphate was then added to allow self-cleavage to generate RNA templates with homogeneous 3′ ends prior to an annealing process to ensure proper folding of the ribozyme. The target RNAs were purified using 12% polyacrylamide–7 M urea gel electrophoresis, excised from the gels, and electro-eluted by using an Elu-Trap device (GE Healthcare). Purified RNAs were stored in RNA annealing buffer (RAB; 50 mM NaCl, 5 mM Tris [pH 7.5], 5 mM MgCl₂) at −80°C after a self-annealing process (a 3-min incubation at 95°C followed by snap-cooling to minimize intermolecular annealing). For primer-dependent assays, the 8-mer and 10-mer primer RNAs (Integrated DNA Technologies) provided at 10% molar excess were annealed to the 31-mer and 33-mer template RNAs, respectively, via a 3-min incubation at 45°C followed by slow cooling to room temperature in RAB to yield the T31/P8 and T33/P10 constructs. For de novo polymerase assays, a GG dinucleotide (TaKaRa) was mixed at a 1.25:1 molar ratio with the 32-mer RNA template, and the same annealing procedures were applied to produce the T32/P2 construct. The assembled RNA constructs were stored at −20°C.

Primer-dependent polymerase assay.

For the primer-dependent polymerase assay, unless otherwise indicated a typical 20-μl reaction mixture containing 6 μM NS5 or its variant, 4 μM RNA construct (T33/P10 or T31/P8), 300 μM GTP, and 300 μM ATP in buffer P (50 mM Tris-HCl [pH 7.5], 75 mM NaCl, 5 mM MnCl₂, 2 mM MgCl₂, 5 mM TCEP) was incubated at 35°C for 5, 15, or 45 min before being quenched by an equal volume of stop solution (95% [vol/vol] formamide, 20 mM EDTA [pH 8.0], 0.02% [wt/vol] bromphenol blue, 0.02% [wt/vol] xylene cyanol). The quenched samples were heated at 100°C for 45 to 60 s. RNA species were resolved by 20% polyacrylamide–7 M urea gel electrophoresis until bromphenol blue migrated to about three-fourths of the vertical gel dimension and then visualized by staining with Stains-All (Sigma-Aldrich). Color images obtained by scanning the stained gels were converted to gray-scale images prior to quantitation by using ImageQuant (GE Healthcare).

De novo polymerase assay.

For the de novo polymerase assay, unless otherwise indicated a typical 20-μl reaction mixture containing 6 μM NS5 or its variant, 4 μM RNA construct T32/P2, 300 μM UTP, and 300 μM ATP in buffer D (50 mM Tris-HCl [pH 7.5], 20 mM NaCl, 5 mM MnCl₂, 5 mM TCEP) was incubated at 30°C for 45, 90, or 120 min. Then, the reaction was quenched by an equal volume of stop solution and the mixture was heated at 100°C for 45 to 60 s. The procedures for denaturing polyacrylamide gel electrophoresis (PAGE), gel staining, and quantification were those described above for the primer-dependent assay.

Single-nucleotide polymerase elongation assay.

For single-nucleotide polymerase elongation studies, reactions were first carried out as described for the de novo polymerase assay. For data presented below in Fig. 7B, the reaction proceeded for 45 min before NaCl and CTP were added to the reaction mixture, to reach a final volume of 22 μl and final concentrations of 210 mM and 300 μM, respectively. The reaction was then carried out at 35°C and was quenched by an equal volume of stop solution at 5, 10, or 30 min after the addition of CTP. For the control experiment, the reaction was carried out as described for the de novo assays, except that the NaCl concentration was raised to 210 mM, CTP was provided at 300 μM along with UTP and ATP, and the reaction temperature was 35°C. As shown with the data presented below in Fig. 7C, for comparison of the properties of EC9 and IC9, the concentration of NS5 was reduced to 3.2 μM (lower than the 4 μM concentration of the P32 RNA template) to highlight the role of competitor. A chemically synthesized 9-mer (csP9) with a sequence identical to P9 was used in the IC9 assembly, where it was mixed at a 0.25:1 molar ratio (similar to the estimated P9:T32 molar ratio after EC9 assembly) with the T32 RNA prior to the standard annealing procedure described above to make the T32/csP9 construct. The EC9 construct was assembled in a 20-μl reaction mixture at 30°C for 60 min prior to supplementation with NaCl and CTP as described above. The reaction proceeded for another 10 or 45 min at 35°C before being quenched by an equal volume of stop solution. Except for the usage of T32/csP9, the 22-μl IC9 reaction mixture was otherwise identical to the EC9 final reaction mixture, with the limited consumption of ATP and UTP neglected. For data presented below in Fig. 8, the reaction mixture was centrifuged at 13,633 rpm on an Eppendorf 5418 centrifuge for 5 min. The supernatant was removed, and the pellet was subjected to two rounds of mild washes with buffer D before resuspension in a high-salt buffer, followed by the addition of CTP while on ice. The final mixture contained 50 mM Tris-HCl (pH 7.5), 210 mM NaCl, 5 mM MnCl₂, 5 mM TCEP, and 300 μM CTP. After the addition of CTP, the reaction mixture was brought to a 35°C water bath, and the reaction proceeded for another 1, 2, 5, 10, or 30 min before being quenched by an equal volume of stop solution. The procedures for denaturing PAGE, gel staining, and quantitation were those described for the primer-dependent assay. To obtain the apparent polymerase elongation rate constants (k_pol^app) corresponding to the conversion from EC9 to EC10, the values representing the fraction of 10-mer intensity (f) at all five times (t) was fitted to a single exponential rise equation: f = offset + amplitude [1 − exp(−k_pol^app × t]), where offset represented the fraction of 10-mer contributed by misincorporation prior to the chase reaction and the amplitude considered the possibility of a 9-mer RNA that eventually failed to extend to a 10-mer. For all k_pol^app values reported, extra data points at 0 min were added to the fitting routine using the 10-mer intensity fraction obtained in the absence of CTP (−CTP). For the k_pol^app value reported in Fig. 8E, t was replaced with the expression (t − t₀); the t₀ term was used to compensate the CTP-induced EC9-to-EC10 elongation that occurred during the incubation on ice.

FIG 7.

A single-nucleotide extension assay suitable for characterization of NS5 elongation activities. (A) Reaction flow charts of the dinucleotide (P2)-derived reactions (left) and the 9-mer (csP9)-derived reactions (right). The boxed GC sequence in T32 is inverted in the template used in the competitor RNA construct in panel C. (B) The EC9 was resistant to 210 mM NaCl and was able to extend to EC10 upon CTP addition. In the control experiment, 210 mM NaCl blocked the production of EC10 by inhibiting the initial steps of synthesis. (C) The 9-mer–to–10-mer conversion from EC9 was more rapid than that from an IC9 assembled using the T32 template and a csP9, while both EC9 and IC9, to some extent, showed resistance to the challenge of a competitor RNA construct.

FIG 8.

Characterization of elongation kinetics on representative full-length NS5 constructs. (A) Comparison of EC9 and EC10 reactions among five NS5 constructs. (B to E) Fraction of 10-mer intensity (10mer_int.) as a function of time measured for five NS5 constructs under standard conditions (B and D), with two CTP concentrations (C) or with a 50 mM NaCl concentration (E). Values for each data set in panels B and D were taken from three individual experiments. The average 10mer_int. fraction and standard deviations are shown. For each data set, the gel of one experiment is shown in panel A. Fitting errors are shown for k_pol^app values overlaid in the plots.

RESULTS

The establishment of two distinct polymerase assays for flavivirus NS5.

Currently, in vitro polymerase assays for de novo viral RdRPs can be classified into three main categories. A typical assay in the first category uses a long homopolymeric template in the presence or absence of a homopolymeric primer usually about 10 to 20 nucleotides (nt) in length and assesses polymerase activity based on nucleotide triphosphate (NTP) consumption (16, 27, 28); this assay can therefore only provide a crude measure of catalysis. The second category features a subgenomic RNA template (typically several hundred nucleotides) that contains both 5′- and 3′-untranslated regions (UTRs) of the viral genome to sustain in cis RNA-RNA interactions and RNA-polymerase interactions that have been documented to ensure efficient replication and precise initiation site selection (29, 30). Although these assays provide the best situation mimicking RNA synthesis in vivo, they often cannot provide explicit information for polymerase initiation and/or elongation, since the amount of the full-length product is typically the only measurable factor. The assays we established in this work fall into the third category, which utilize either genome-derived or artificial heteropolymeric template sequences normally shorter than 100 nt. By using different NTP combinations as the primary strategy to allow the monitoring of RNA polymerization down to the single-nucleotide level, these assays offer the best power in dissecting the mechanism of RNA synthesis (19, 31–33).

It is generally accepted that RdRPs of the family Flaviviridae carry out de novo initiation in viral genome replication. However, primer-dependent activities of these enzymes have been reported from various in vitro assays (19, 33), providing valuable measures of RNA synthesis. We first established a primer-dependent assay based on what had been utilized in poliovirus (PV) polymerase systems (26). Two RNA constructs, T31/P8 and T33/P10 (Fig. 1A), were used for the optimization of the assay conditions. The T31/P8 construct contained a 31-mer template strand and an 8-mer primer strand. The 3′ portion of the template strand formed a 6-bp duplex with the primer, while the 5′ portion of the template could fold back on itself to form a hairpin with a 6-bp duplex and a GAAA tetraloop (Fig. 1A). Except for having a longer 8-bp template-primer duplex, the T33/P10 construct was otherwise identical to the T31/P8 construct. These and similar constructs were successfully used to assemble stably stalled polymerase ECs in PV and related virus systems (26, 34). When only GTP and ATP were provided, the primer strand was expected to elongate by 4 nucleotides to produce a +4-nt product (12-mer for T31/P8; 14-mer for T33/P10). We carefully examined the solution stability of several of the NS5 variants and found that the removal of the MTase domain drastically improved NS5's solubility under low-ionic-strength and low-glycerol conditions (data not shown). Hence, the standard assay conditions was determined by a systematic examination of the assay parameters, using N-terminal-truncated forms of NS5 (Fig. 1B to F). First, manganese ion (Mn²⁺) was required for the polymerase activity of JEV NS5, and replacing dithiothreitol (DTT) with TCEP as the reducing agent moderately enhanced the utilization of the primer (Fig. 1B). Second, the highest yield of the +4-nt product was achieved at pH 7.5 among conditions tested (Fig. 1C). In the time course experiments carried out at 30°C, a significant amount of +1- to 3-nt products appeared continuously, and the distribution of the products slowly progressed toward the longest +4-nt species (Fig. 1D). Slow nucleotide extension and/or fast product dissociation likely accounted for this phenomenon, indicating that the polymerase had not entered the elongation phase. Hence, this assay was only suitable to provide a measure of polymerase initiation. Moreover, an increase of the reaction temperature to 35°C improved the conversion rate of the +4-nt product (Fig. 1E). Finally, polymerase activity was not very sensitive to the concentration of NaCl in the tested range of 20 to 100 mM (Fig. 1F).

FIG 1.

A primer-dependent assay characterizing polymerase initiation activities of JEV NS5. (A) Diagram of constructs T31/P8 and T33/P10 used in the primer-dependent assays. The template strand (cyan) and the primer strand (green) form a 6- or 8-bp duplex in the upstream direction, and the downstream template folds back to form a stem-loop structure. (B to F) Analysis for optimal assay conditions, including the following variables: the combination of reducing agent (5 mM DTT or 5 mM TCEP) and divalent metal ions (2 mM Mg²⁺ or 5 mM Mn²⁺) (B); pH (Tris) (C); incubation time (D); reaction temperature and the RNA constructs (E); concentration of NaCl (F). Square boxes indicate the selected variables for the standard reaction in the studies. Solid triangles indicated the +1-nt products that were likely dissociated from the polymerase complex and failed to be extended further, and the appearance of these products was more obvious when T33/P10 was used. For panels B to E, a truncated form of NS5 with residues 274 to 905 (D273 here) was used; for panel F, another truncated form of NS5 with residues 262 to 905 (D261 here) was used.

Recently, a de novo polymerase assay was developed using HCV NS5B polymerase (31). Building on this study, we designed a 32-mer template RNA (T32) for our de novo polymerase assays (Fig. 2A). This template RNA contained the same downstream hairpin as the T31-P8/T33-P10 constructs, while its 3′ region included a templating sequence derived from the HCV study. Such a sequence allowed the assembly of a stable EC of HCV NS5B by using a GG dinucleotide primer (P2). We followed this strategy and assembled an NS5/T32/P2 initiation complex (IC2). When ATP and UTP were supplied, the polymerase was expected to produce a 9-mer product. In our search for optimal assay conditions, first we found that Mn²⁺ was again required for polymerase activity, and the replacement of DTT with TCEP reduced the amount of misincorporation products (Fig. 2B and C). The primary misincorporation product was a 10-mer and was likely related to incorporation of a UMP on a templating guanidine through a G:U mispairing, based on the template sequence and the NTP species supplied. Second, a temperature of 30°C and a pH of 7.5 were chosen to achieve a balance of the large amount of the regular 9-mer products and relatively low level of misincorporation (Fig. 2C and E). In contrast to the observations in the primer-dependent assay, regular polymerase activity and misincorportation were sensitive to the concentration of NaCl and the reducing agent used, respectively, with a larger amount of the 9-mer observed at the lower NaCl concentration (20 mM) (Fig. 2F). Although 6- to 8-mer RNAs were stainable in our experiments, these intermediates were not detected at all time points, strongly suggesting that the polymerase had entered the elongation phase of RNA synthesis. The rapid elongation and slow dissociation of the RNA product are the two hallmark properties of a polymerase EC. Once it had proceeded from initiation to elongation, a polymerase complex exhibited fast kinetics with few accumulated intermediates. Presumably, slow de novo initiation and the subsequent transition to rapid elongation had both occurred on the way to producing the 9-mer product. Therefore, this assay has the capacity to assess the de novo initiation efficiency. On the other hand, by probing the polymerase–template–9-mer EC (EC9), the dissection of the polymerase elongation step becomes possible.

FIG 2.

A de novo polymerase assay for JEV NS5. (A) Diagram of constructs T32/P2 used in the de novo assays. When ATP and UTP were supplied as the only NTP substrates, the template-strand T32 (cyan) directed a 7-nucleotide (red) extension of the dinucleotide primer P2 (green) to produce a 9-mer product. (B to F) Analysis for optimal assay conditions, including the following variables: the combination of reducing agent (5 mM DTT or 5 mM TCEP) and divalent metal ions (2 mM Mg²⁺ or 5 mM Mn²⁺) (B); reducing agent and temperature (C); pH (Tris) (D); temperature and incubation time (E); concentration of NaCl (F). Boxes indicate the selected variables for the standard reaction used in the studies. The solid triangles indicate misincorporated products.

Altering the MTase-RdRP interface interactions led to different polymerase initiation levels of NS5.

In the recently reported crystal structure of the full-length JEV NS5, a conserved medium-size interface between the MTase and the RdRP was identified (6). Lying in the heart of this interface are a set of six hydrophobic residues, three from each side (Fig. 3A and B). We carried out a conservation analysis using NS5 sequences from a collection of flaviviruses (41 species and 4 serotypes of DENV). These represented all viral species assigned as flaviviruses by the International Committee on Taxonomy of Viruses (ICTV) in 2012 for which the complete NS5 sequence was available (http://www.ictvonline.org). It turned out that P113, W121, F351, and P585 were invariant, F467 had only one tyrosine (Y) variant, and L115 was also largely conserved as a hydrophobic residue (41 out of 45 sequences had L/M/P/V/I at this position) (Fig. 3A). Structurally, the core enzymatic modules of MTase and RdRP are connected sequentially by three elements: an invariant GTR sequence (residues 263 to 265) at the C terminus of MTase, a 10-residue linker (residues 266 to 275) with low sequence conservation, and a moderately conserved N-terminal extension (residues 276 to 303) of core polymerase (Fig. 3A and C).

FIG 3.

A conservation analysis of the MTase-RdRP interface and the design of NS5 variants. (A) Sequence logos generated using a multiple-sequence alignment of 45 NS5 sequences of flavivirus (http://weblogo.berkeley.edu/). (Top panel) The six key residues (indicated by solid triangles) that comprised the hydrophobic network at the MTase-RdRP interface and their surrounding residues. (Bottom panel) Three elements connecting the MTase and RdRP cores. (B) Structure of the interface, showing the spatial organization of the six-residue hydrophobic network (side chains are shown as large spheres) and the GTR tripeptide (the α-carbons are shown as small brown spheres). The main chain coloring scheme is the same as for panel A (bars above the logo). (C) Diagram showing the design of N-terminal-truncated NS5 constructs. (D) A list of all NS5 constructs used in this study, with abbreviations and full descriptions, including residue range, mutation site, and mutation type.

To further our understanding of the function of the MTase-RdRP interface and the regions connecting the two enzyme modules, we made a collection of NS5 constructs and compared them with WT NS5 in the two assays described above (Fig. 3C and D). First, point mutations in the hydrophobic network were made in the context of full-length NS5. The rationale for the mutations was to modulate the MTase-RdRP interface by replacing one or several hydrophobic residues in the network with aspartic acid (D), which has a negatively charged side chain (Fig. 3D). Second, several versions of N-terminal deletions were made to test whether the presence of the MTase, the linker, or the N-terminal extension affected polymerase activities. Third, the point mutations in the RdRP side of the hydrophobic network were introduced to the truncated NS5 constructs. If the hydrophobic network at the interface or the MTase-RdRP connecting regions play roles in RNA synthesis, either at initiation or elongation, the established polymerase assays may be able to provide informative dissection by comparing the RNA product profiles and/or polymerization kinetics between the WT NS5 and its variants.

We first compared the WT enzyme with all full-length variants. The T31/P8 construct was used in the primer-dependent assay, and RNA synthesis was monitored at three incubation time points (5, 15, and 45 min) with GTP and ATP supplied as the only NTP substrates. For each enzyme, a clear ladder was formed by the primer and the extension products with a length of 8 to 12 nt at 5 min (Fig. 4A, lanes 1 to 7). As the reaction proceeded to 15 and 45 min (Fig. 4A, lanes 8 to 14 and 15 to 21), the conversion of the primer was nearly finished and the 12-mer became the dominant product, indicating the completion of primer extension by 4 nt. Due to the basic characteristics and sensitivity limitation of the staining technique used in this study (Fig. 5), we chose the percentage of 12-mer intensity at 5 min (designated 12mer_int.%) to obtain semiquantitative information for the overall velocity of the initiation-mode RNA synthesis, rather than pursuing particular kinetics parameters that would require multiple measurements at time points around 5 min. More explicitly, the 12mer_int.% value is the percentage of the 12-mer intensity among the total intensities of all primer/product RNAs (i.e., 8- to 12-mers). Comparison with the WT enzyme (12mer_int.% of 26), M_113 (12mer_int.% of 17) and R_351/467 (12mer_int.% of 14) mutants showed slower production of 12-mers in 5 min (compare lane 1 to lanes 2 and 5), while the synthesis by R3 (12mer_int.% of 43) and M3-R3 mutants (12mer_int.% of 44) was obviously faster (lanes 6 and 7). This suggested that the initiation activities of JEV NS5 were sensitive to the perturbation in the hydrophobic network at the MTase-RdRP interface. For the de novo assay, the production of the 9-mer was much slower than the 12-mer production in the primer-dependent assay, and only a portion of the short P2 primer was utilized (Fig. 4B). Different from the P8 primer that could form a relatively stable 6-bp duplex with its template favoring nucleotide extension, the extension of the short primer/product was likely at a disadvantage in early stages of synthesis when competing with dissociation from the template. Hence, RNA synthesis was monitored at 45, 90, and 120 min in the de novo assay. The total amount of 9-mer product and the 10-mer misincorporation product (referred to here as the 9-10-mer) at the 45-min time point was used to semiquantitatively assess polymerase activity, i.e., the activity of the polymerase needed to make the transition from dinucleotide-directed initiation to elongation in generating these products. Therefore, the production of 9-10-mer was related to both phases of synthesis. Given the fact that initiation has significantly slower kinetics than elongation, the 9-10-mer production was also expected to provide a reasonable measure of polymerase initiation. The results were largely consistent with those obtained in the primer-dependent assay, with M_113 and R_351/467 producing smaller amounts of 9-10-mer and M3-R3 producing a larger amount of 9-10-mer at the 45-min time point than the WT NS5 (Fig. 4B, compare lanes 32, 35, and 37 with lane 31).

FIG 4.

Mutations of the hydrophobic network modulated the initiation activities of JEV NS5. A total of six full-length NS5 mutants were compared with the WT NS5 in two types of polymerase assays. (A) In the primer-dependent assay, the percentage of 12-mer intensity among all 8- to 12-mer species (12mer_int.%) at the 5-min time point was used to evaluate the polymerase initiation activities. Up or down arrows indicate higher or lower values of 12mer_int.% compared to the WT. (B) In the de novo assay, the relative intensity of 9-10-mer at the 45-min time point was used to estimate polymerase activities (the value for WT was set to 100). Up or down arrows indicate higher or lower levels of the 9-10-mer compared to the WT. Stars indicate misincorporation products, with lengths of 11 nt or longer.

FIG 5.

The Stains-All dye is suitable for semiquantitative analysis. (A to C) Linear relationship analysis. A series of samples with different concentrations (0.44, 0.88, 2.2, 4.4, 8.8, 22, and 44 μM) of the P8 RNA used in the primer-dependent assays were prepared. The loading amount of the 4.4 μM sample of P8 (equivalent to the concentration used in the primer-dependent assays) was set as 1 (solid squares in panels B and C). Each sample was mixed with an equal volume of stop solution, loaded with an equal volume as in the primer-dependent assays, and resolved by 20% polyacrylamide–7 M urea gel electrophoresis (A). Color images obtained by scanning the stained gels were converted to gray-scale images prior to quantitation using ImageQuant (GE Healthcare). Within the concentration range between 0.44 and 8.8 μM (0.1 and 2 relative amounts), a very good linear relationship between the amount of RNA and the measured intensity was observed (B and C). With a concentration of 22 μM or higher, the intensity became underestimated (empty circles in panel B). Panel C is an enlarged version of the low-concentration part of panel B. Values in panels B and C were taken from three individual experiments and were normalized to the 4.4 μM samples, with the average intensity value shown in both panels and standard deviations provided in panel C. (D) Intensity-RNA length dependency analysis. The P10 and P8 RNAs used in the primer-dependent assay were prepared at four different concentrations, and each sample was mixed with an equal volume of stop solution. With an equal molar amount, the intensity of 10-mer was, overall, moderately higher than that of the 8-mer. In our quantitation analysis for all polymerase assays, as the RNA species had a typical relative amount between 0.1 and 1 and a length variation between 8 and 12 nucleotides, the Stains-All dye afforded at least the capacity for semiquantitative analysis.

We next tested the N-terminal-truncated NS5 variants. In the primer-dependent assay, D261, D273, and their point mutation variants exhibited apparently slower production of 12-mer after 5 and 10 min than the WT NS5 did (Fig. 6A, compare lanes 1 and 8 with lanes 2 to 5 and 9 to 12, respectively). For each form of the truncation, the introduction of three point mutations further slowed the synthesis (Fig. 6A, compare lanes 2, 4, 9, and 11 with lanes 3, 5, 10, and 12, respectively). For the de novo assay, a smaller amount of 9-10-mer product was found at 45 min (Fig. 6B, compare lane 31 with lanes 32 to 35), although the differences between the WT and the mutants were not as obvious as in the primer-dependent assay. First, these data suggested that the presence of the MTase module and/or the linker region was not essential for regular polymerase activities, consistent with previous data showing the relative functional independence of the polymerase region (15, 16, 19). More importantly, removal of the MTase and/or the linker had an overall more disruptive impact on polymerase initiation activities compared to introduction of point mutations in the full-length protein, indicating that the conformation observed in the full-length NS5 crystal structure maintaining the MTase-RdRP interface likely represented the ground-state conformation in early stages of RNA synthesis (see the Discussion). With the N-terminal extension further removed, D303 and its variant, D303-R3, lost the capability of regular template-directed synthesis (Fig. 6A, lanes 6 and 7, 13 and 14, and 20 and 21), indicating that this region does play auxiliary roles in RNA polymerization and should be considered part of the polymerase module. Rather than adding a GAGA tetranucleotide to the 3′ end of the primer, continuous addition of AMP yielded a ladder profile in the denaturing polyacrylamide gels (Fig. 6A, lanes 6 and 7, 13 and 14, and 20 and 21 and data not shown). Such off-pathway activities have been observed with other polymerases. For instance, T7 RNA polymerase makes G-ladder products during the early stages of the RNA synthesis on certain DNA templates, owing to product dissociation and subsequent realignment to the template strand (35). In contrast to the D261 and D273 constructs, the same set of R3 mutations actually led to more profound off-pathway synthesis in D303. In the de novo assay, polymerase activity was not evident for D303 or D303-R3 (Fig. 6B, lanes 36 and 37, 43 and 44, and 50 and 51). The accumulation of RNA species that migrated faster than the template was likely related to the basal activity of possibly contaminated nucleases, which only became evident if NS5 could not efficiently utilize the RNA template.

FIG 6.

Comparison of N-terminal-truncated forms of NS5 and the full-length protein (WT) in RNA synthesis. A total of six N-terminal-truncated NS5 variants were compared with the WT NS5 in two types of polymerase assays. (A) In the primer-dependent assay, the 12-mer_int.% values at the 5-min and 15-min time points were used to evaluate the polymerase initiation activities. Arrows indicate different extents (lower values) of the 12mer_int.% compared to the WT. (B) In the de novo assay, the relative 9-10-mer_int. values at the 45-min and 90-min time points were used to estimate polymerase activities. Arrows indicate different extents of levels of 9-10-mer intensity compared to the WT. Solid triangles indicate the RNA species possibly related to basal level of nuclease activities. Stars indicate the approximate migrating positions of the dominant products made by the D303-R3 constructs. Horizontal bars between lanes 19/20 and 20/21 indicate the migration differences between regular template-directed synthesis in lane 19 and the off-pathway synthesis in lanes 20 and 21.

Two full-length NS5 variants exhibited different polymerase elongation kinetics.

The successful production of EC9 (EC with a 9-mer product) in the de novo assay provided the capacity to assess elongation-mode synthesis by using a “challenge and chase” strategy, which has been widely used in nucleic acids polymerase studies. Typically, high salt or heparin is introduced upon the formation of an EC to block further initiation events, and then certain NTP or NTP combinations are added to probe any EC-derived activities (31, 36). After an incubation period to allow EC9 formation with ATP and UTP, NaCl was supplemented to a final concentration of 210 mM, and CTP was subsequently added to allow possible single-nucleotide incorporation, generating an EC with a 10-mer product (EC10) (Fig. 7A). The accumulation of the 10-mer product was indeed observed during the 30-min period (Fig. 7B, lanes 2 to 4). In the control experiment, WT NS5 failed to produce any 10-mer product in the same period when ATP, UTP, and CTP were introduced together at initiation with an equivalent NaCl concentration of 210 mM (Fig. 7B, lanes 5 to 7). These data clearly showed that the EC10 produced in the chase experiment was not derived from dinucleotide-driven synthesis.

To further explore the properties of EC9, we first compared 9-10-mer production from EC9 and from an “IC9,” which was assembled using NS5 and an RNA construct comprising the T32 template and a chemically synthesized 9-mer (csP9) identical to P9 in EC9. Although under the high-salt condition the IC9 was capable of making 10-mer products, the synthesis was notably much slower than observed in the parallel EC9 experiment (Fig. 7C, compare lanes 12 to 14 with lanes 2 to 4). We next assessed the stability of EC9 and IC9 by using a competition assay, in which an RNA construct competitor containing the csP9 and an RNA template with the immediate templating GC sequence in T32 (Fig. 7A, boxed) inverted were provided at a 2:1 molar ratio to the T32/P9. Such a competitor is able to bind NS5 but will only allow a minimum level of extension from csP9 in the presence ATP, CTP, and UTP (Fig. 7C, lane 8). Therefore, this competitor can compete with the dissociated T32/P9 for rebinding NS5. The template was provided at a molar ratio of 1:0.3 to the csP9, not only to mimic the situation in the EC9 experiments but also to offer the capability to trap free P9 for reassembling with T32 and NS5. In the presence of the competitor, the production of 10-mer was evident but reduced compared to that in the absence of competitor in both the EC9 and IC9 experiments (Fig. 7C, compare lanes 5 to 7 with lanes 2 to 4 and compare lanes 15 to 17 with lanes 12 to 14), suggesting both complexes readily dissociated during the reaction incubation period. However, further stability comparison of EC9 and IC9 may be difficult with the detection method employed in the current study. Taken together, the EC9 derived from the P2 primer had gained competency of elongation but was not as stable as was observed in the HCV study (31).

We then used this high-salt chase assay to compare the elongation kinetics of NS5 constructs (Fig. 8). Four full-length variants were chosen to represent those with slower (M_113 and R_351/467) or faster (R3 and M3-R3) 12-mer production at initiation. Similar to what was found in the HCV study, the majority of EC generated by all full-length NS5 constructs was in a precipitated form in the preassembly reaction mixture (31). To facilitate the quantitation efforts, the preassembled EC was pelleted and washed to remove the initiation NTPs and the unbound dinucleotide primer, and it was then resuspended in high-salt buffer to carry out a time course experiment for CTP incorporation (Fig. 8A). Given the fact that 300 μM CTP was well above the concentration of previously documented K_m values for viral RdRP elongation (37–39), and considering that a small amount of RNA could dissociate and reassemble with NS5 during the incubation, the data were fitted to a single-step irreversible model (EC9 to EC10) to estimate the apparent polymerase elongation rate constant (k_pol^app). Under the tested condition, the WT NS5 had a k_pol^app of 0.14 min⁻¹ (Fig. 8A and B). To further confirm our assumption of NTP saturation, we compared the incorporation under CTP concentrations of 100 and 300 μM, and indeed the two reactions produced very similar product conversion profiles (Fig. 8C). We then measured the apparent elongation rate constants of the four full-length variants (Fig. 8A and D). Very interestingly, the slower initiation M_113 mutant showed faster elongation kinetics (k_pol^app, 0.20 min⁻¹) than the WT enzyme did, while the faster initiation variant M3-R3 was slower at elongation (k_pol^app, 0.11 min⁻¹) (Fig. 8D). Although behaving differently from the WT enzyme at initiation, the R_351/467 and R3 mutants actually had about the same elongation rate constants as the WT enzyme (Fig. 8D). Taken together, these point mutations had different impacts on polymerase initiation and elongation, implying that the hydrophobic network or the entire MTase-RdRP interface are relevant in both phases of synthesis but possibly with different behaviors regarding the dynamics and regulation of polymerase catalysis.

DISCUSSION

Functional relevance of the MTase-RdRP interface in polymerase catalysis.

Structural organization of the viral RdRPs is highly homologous, having seven polymerase motifs, A to G, surrounding the active site within the encircled right-hand architecture (6, 26). Although other viral and host factors are important components of the viral replication complex, many viral RdRPs, such as PV 3D^pol and HCV NS5B, can function quite independently in vitro (40, 41). These RdRPs comprise a ≈500-residue polymerase core without other enzyme modules integrated into the same polypeptide chain. In fewer cases, another viral enzyme(s) or functional domain(s) is naturally fused to RdRP, providing the capacity for regulation and cooperation to take place intramolecularly between RdRP and its fusion partners (42–44). Among these, the flavivirus NS5 is very unique in having an MTase as the only partner of the RdRP immediately fused to its N terminus. There had been radical evidence for the intramolecular interaction or regulation between MTase and RdRP, but the recent report of the full-length JEV NS5 crystal structure provided the first high-resolution view of the MTase-RdRP interface with highly conserved features (6). Interestingly, the WNV and DENV RdRP crystal structures in isolation are partially disordered, in particular in the index, ring, and pinky finger subdomains, which are known to contribute to RdRP catalysis and stability (33, 45, 46). In the full-length NS5 structure, the entire fingers domain is well-ordered and adopts canonical folds compared to RdRPs without fusion partners (5, 47). The index and ring fingers both have direct and conserved interactions with the MTase, strongly indicating that the maintenance of the interface likely plays roles in polymerase function.

The presence of the interface buries regulation elements, including a nuclear localization signal and the putative NS3 binding site (48, 49), and small-angle X-ray scattering data are consistent with a more elongated shape of NS5 (6, 50). Together these data suggest that the interface may unravel to allow NS5 to adopt other conformations. Also, taking the structural observations of the apo full-length NS5 and the apo RdRP into consideration (6, 15, 16), we propose that three primary states of apo NS5 may exist (Fig. 9A). The dominant state, a1, is an NS5 that maintains the interface interactions with a properly folded RdRP conformation for polymerase initiation (i.e., the conformation observed in the full-length NS5 crystal structure). Interface interactions are unraveled in states a2 and a3, with state a2 having a proper fold of RdRP that is also capable of initiation and state a3 having a partially disordered fingers domain that is initiation incompetent (i.e., the conformations observed in DENV and WNV RdRP crystal structures). In our in vitro assays, the majority of the point mutations and the deletion variants had obvious impacts on initiation activities. In our model, mutations that either perturb the interface or affect the proper folding of RdRP would change the equilibrium of the three states, and therefore influence initiation. Following initiation, the polymerase makes a transition to form a stable EC (Fig. 9B and C); the interactions between the polymerase motifs and the RNA scaffold likely take over control of the conformational stability of RdRP. Overall, NS5 variants tested in our elongation assay had comparable apparent elongation rate constants as the WT NS5. It is conceivable that the equilibrium of the interface formation and unraveling still exist in EC (Fig. 9B, states e1 and e2), but this would have a lower impact on polymerase catalysis. The differences in conformational state versatility and different impacts on polymerase catalysis brought by the equilibrium between conformational states likely accounts for the obvious differences in initiation and elongation kinetics of the full-length NS5 constructs tested in our studies. To specifically test the conformational states hypothesis that we propose here would require the collection of either new structural evidence, or kinetics data, or both, using different forms of the NS5 proteins.

FIG 9.

Cartoon illustration of possible conformational/functional states at polymerase initiation and elongation of flavivirus NS5. (A) Three states of a proposed apo NS5 based on previously reported structural and functional studies. No matter whether the interface interactions were present, states a1 and a2 are initiation competent, as the RdRP region adopts a proper fold. (B) Two states of an NS5 EC; both are elongation competent. (C) A minischeme of RNA synthesis by flavivirus NS5.

In a very recent study in which we addressed the function of the MTase-RdRP interface with single-point serine/arginine/aspartic acid (S/R/D) mutations of the hydrophobic network introduced into infectious virus clone or replicon systems, virus replication levels were dramatically impaired, with a significant portion of mutations having a lethal phenotype (21). In the current study, even if the D mutation with the highest impact on virus replication was utilized and the combination of up to six point mutations was applied, the overall impact on polymerase activities was not as apparent. Several factors could account for the different extent of impacts observed between the virology approaches and our in vitro polymerase studies. First, impairment of either initiation or elongation, once placed into the context of full-genome synthesis coupled with viral protein translation, could be greatly amplified. Second, the mutations could potentially perturb the capping process through regulation of the MTase region of NS5. Moreover, altering the interface interactions may have an impact on the interactions between NS5 and other viral/host proteins which also contribute to the virus life cycle. Further investigations are necessary for a more comprehensive understanding of the roles of the MTase-RdRP interface observed in the full-length NS5 structure.

Indispensability of the N-terminal extension in polymerase initiation.

The N-terminal extension of flavivirus NS5 is a quite unique element whose function remains elusive. Except for pestivirus NS5B, which contains a similarly folded region, all other viral RdRPs for which the structure is known do not have equivalents to it (6). In the current study, when the MTase region and/or the linker region was removed (constructs D261 and D273), NS5 retained its polymerase function in both types of initiation assays, albeit with lower activities. Interestingly, when the N-terminal extension was further removed (D303), the polymerase lost the ability to carry out regular template-directed synthesis in the primer-dependent assay and de novo synthesis activity was completely abolished. These data suggest that the N-terminal extension may not be dispensable, at least in the initiation events in RNA synthesis by flavivirus NS5. Due to the fact that the sequence of this region is only moderately conserved (Fig. 3A) in flaviviruses and lacks similarity to its pestivirus counterpart (6), it may be difficult to locate a specific site(s) responsible for the impairment of polymerase initiation. Nevertheless, further clarification of its contribution in initiation and similar testing in pestivirus NS5B would be of interest. Structurally, the N-terminal extension interacts with the junction regions between the palm and the index/ring fingers. It has been shown that the dynamics between the fingers and thumb domains is integrated with the polymerase catalytic cycle (26). Thus, the N-terminal extension could possibly contribute to polymerase initiation events by structurally stabilizing fingers-thumb interactions.

Compared to the N-terminal extension, the linker region has even lower sequence conservation and interacts with a junction region (residues 299 to 308) of the N-terminal extension and the index finger (6). Other than tethering the two enzyme modules together, its function remains unclear. In a very recent study, addition of the linker to the N-terminal end of the N-terminal extension of DENV serotype 4 (DENV-4) RdRP enhanced de novo synthesis activity of the polymerase (20). However, the D261 (with intact linker) and D273 (with 8 out of 10 linker residues removed) constructs in our study exhibited very similar levels of synthesis in both types of initiation assays. This apparent discrepancy could be related to the different nature of assay conditions, as the polymerase activity in the DENV study was derived from the bulk consumption of NTP over a period of 3 h. Multiple turnovers of RNA synthesis could happen, and any off-pathway synthesis could not be distinguished from regular synthesis in such an assay, bringing further complications to the nature of the values. On the other hand, removal of the linker did not abolish polymerase activity in either study, indicating that it does not play a critical role in polymerase function.

Toward the structure of an NS5 elongation complex.

De novo viral RdRPs share a “priming element” that projects from the thumb domain to the vicinity of the active site (5–7), partially occupying the front channel that accommodates the template-product RNA duplex in the elongation phase presumably for all viral RdRPs (26). With supporting evidence from both crystal structures and biochemical data (8, 51, 52), it is generally accepted that the priming element facilitates initial steps of de novo synthesis and needs to completely withdraw from the active site when the template-product duplex reaches a length of 7 to 8 bp, with its upstream end exiting via the front channel (34). Although whether and how the priming element helps to control the precise and relatively efficient initiation at the 3′ terminus of the genomic or the antigenomic RNA needs to be further clarified, a more attractive aim is to solve a high-resolution crystal structure of a de novo viral RdRP EC, which has been a challenge for about 15 years, since the report of the first de novo RdRP crystal structure (5) followed by the report of the first and only de novo RdRP IC crystal structure (8). The EC structure would reveal the fate of the priming element: whether it is dispensable at elongation or it has new missions. Comparing to cocrystallization and apo RdRP crystal soaking strategies, assembling a stable EC prior to crystallization trials would likely provide better opportunities toward resolving an EC structure. By adapting the recently reported methods for HCV NS5B EC assembly into the current study (31), we have made a solid step toward an EC structure by assembling a “metastable” JEV NS5 EC through de novo synthesis. Although the k_pol^app values obtained in our study are significantly lower than the typical 10 to 100 s⁻¹ reported for picornavirus RdRP ECs (37–39, 53), at least to some extent they represent elongation-mode synthesis as the 9-mer to 10-mer conversion is obviously faster than that in the parallel experiment starting with IC9 (Fig. 7C). When the NaCl concentration was reduced from 210 mM to 50 mM (a concentration similar to that used in the picornavirus EC studies), the k_pol^app value increased about 10-fold (1.5 min⁻¹), further suggesting that the EC9 obtained in our de novo assay has elongation competency (Fig. 8E). In contrast, the current format of our primer-dependent assays may not be able to produce an EC. The complete usage of the 8-mer primer provided in 10% excess to the T31 template indicated that the RNA products readily dissociated from the complex, allowing extra primer to anneal on the template RNA and then extend (Fig. 4 and 6). This instability of the complex and the slow kinetics suggested by the aforementioned accumulation of 9- to 11-mer intermediates together indicated that an EC was not formed in such an assay.

Flavivirus genome replication in vivo requires the participation of RNA elements in both 5′ and 3′ regions for efficient initiation and precise start site selection (30, 54), and other viral and host members of the membrane-associated replication complex may regulate the catalytic activity of NS5 RdRP. Compared to what happens in vivo, the EC9 assembly strategy in the current study is much simplified, and the polymerase complex might have not undergone all critical steps en route to a bona fide EC. Note that all polymerase catalytic activity in the current study requires the presence of manganese ion. As documented in previous studies (53, 55), manganese ion often diminishes stringency of RdRP catalysis and thus may have compensated the RNA substrate simplification in the current study to ensure initiation and subsequent transition toward elongation.

In summary, the current study built on the recently revealed MTase-RdRP interface of JEV NS5 and provided mechanistic dissections of how the interface-regulated NS5 polymerase activities via assays that specifically probed the initiation or the elongation phase of synthesis. Our data indicated that polymerase initiation activities were more susceptible to the perturbation in the interface, and the same NS5 variant could bring opposite effects on initiation- and elongation-mode synthesis, implying the different relevance of the interface in different stages of synthesis. Data from N-terminal-truncated NS5 constructs suggest that a flavivirus/pestivirus-specific N-terminal extension region is likely indispensable for JEV polymerase function. The work improved our understanding of the regulatory roles of the MTase-RdRP interface, in particular for the polymerase function of NS5, and future studies probing the interface in a broader context of both viral replication and capping will further benefit our understanding of these roles.