Interactions between the Dengue Virus Polymerase NS5 and Stem-Loop A

ABSTRACT

The process of RNA replication by dengue virus is still not completely understood despite the significant progress made in the last few years. Stem-loop A (SLA), a part of the viral 5′ untranslated region (UTR), is critical for the initiation of dengue virus replication, but quantitative analysis of the interactions between the dengue virus polymerase NS5 and SLA in solution has not been performed. Here, we examine how solution conditions affect the size and shape of SLA and the formation of the NS5-SLA complex. We show that dengue virus NS5 binds SLA with a 1:1 stoichiometry and that the association reaction is primarily entropy driven. We also observe that the NS5-SLA interaction is influenced by the magnesium concentration in a complex manner. Binding is optimal with 1 mM MgCl₂ but decreases with both lower and higher magnesium concentrations. Additionally, data from a competition assay between SLA and single-stranded RNA (ssRNA) indicate that SLA competes with ssRNA for the same binding site on the NS5 polymerase. SLA₇₀ and SLA₈₀, which contain the first 70 and 80 nucleotides (nt), respectively, bind NS5 with similar binding affinities. Dengue virus NS5 also binds SLAs from different serotypes, indicating that NS5 recognizes the overall shape of SLA as well as specific nucleotides.

IMPORTANCE Dengue virus is an important human pathogen responsible for dengue hemorrhagic fever, whose global incidence has increased dramatically over the last several decades. Despite the clear medical importance of dengue virus infection, the mechanism of viral replication, a process commonly targeted by antiviral therapeutics, is not well understood. In particular, stem-loop A (SLA) and stem-loop B (SLB) located in the 5′ untranslated region (UTR) are critical for binding the viral polymerase NS5 to initiate minus-strand RNA synthesis. However, little is known regarding the kinetic and thermodynamic parameters driving these interactions. Here, we quantitatively examine the energetics of intrinsic affinities, characterize the stoichiometry of the complex of NS5 and SLA, and determine how solution conditions such as magnesium and sodium concentrations and temperature influence NS5-SLA interactions in solution. Quantitatively characterizing dengue virus NS5-SLA interactions will facilitate the design and assessment of antiviral therapeutics that target this essential step of the dengue virus life cycle.

INTRODUCTION

Dengue virus (DENV) is a positive-sense, single-stranded RNA (ssRNA) virus and is a member of the Flavivirus genus within the Flaviviridae family. DENV causes a wide range of diseases in humans, ranging from self-limiting dengue fever to life-threatening dengue hemorrhagic fever and dengue shock syndrome (1–5). Recent estimates indicate that more than 300 million dengue virus infections occur each year, 96 million of which clinically manifest. Currently, it is estimated that 3.9 billion people living in 128 countries are at risk of infection by DENV (6, 7). Four serotypes of the virus have been identified (DENV serotype 1 [DENV1] to DENV4), and most severe cases of hemorrhagic fever are associated with secondary infections by a serotype different from that of the first infection (8). The flavivirus genus also includes other human pathogens, such as West Nile virus, tick-borne encephalitis virus, yellow fever virus, and Japanese encephalitis virus (JEV), as well as the rapidly spreading Zika virus.

The 11-kb DENV genome encodes 10 proteins, including three structural proteins (capsid, premembrane, and envelope proteins) and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (9, 10). NS5 is a viral replicative RNA-dependent RNA polymerase (RdRp) consisting of 900 amino acids distributed across two functional domains (11–14). The N-terminal methyltransferase (MTase) domain (263 amino acids long) functions as a guanylyltransferase and methyltransferase whose activity is necessary for 5′-RNA cap synthesis and methylation (15, 16); cap methylations are important for viral RNA recognition and for hijacking the host cell translational apparatus. The C-terminal domain possesses RdRp activity, including the active site for RNA synthesis. Crystal structures of DENV and JEV NS5 have been determined (17–19). These structures showed different interdomain interactions within the NS5 monomer and indicated that NS5 can adopt a number of conformations with different relative orientations of the MTase and RdRp domains. Additionally, interdomain interactions between monomers in an NS5 dimer suggest the coordination of RdRp and MTase activities across the NS5 monomer and dimer (17). Data from small-angle X-ray scattering (SAXS) studies indicate that NS5 proteins of all four DENV serotypes can adopt a more elongated shape than that in the crystal structure (20, 21), suggesting that flexibility between the MTase and RdRp domains is likely essential for the viral replication process. Interactions between the MTase and RdRp domains are also important for the polymerase activity of NS5, since full-length NS5 has higher polymerase activity than does the RdRp domain alone (22).

In addition to encoding viral proteins, the DENV genome contains tertiary RNA structures at the 5′ and 3′ untranslated regions (UTRs) that regulate different viral processes (23). In particular, the 5′ UTR contains two stem-loops, stem-loop A (SLA) and stem-loop B (SLB), and SLA has been shown to function as a promoter for RNA synthesis during the replication process (24–26). The importance of SLA in viral RNA replication has been demonstrated for several flaviviruses (24, 27, 28). Despite the importance of SLA in initiating dengue virus RNA replication, interactions between NS5 and SLA have not been quantitatively analyzed, and thus, fundamental aspects of the NS5-SLA interaction, such as stoichiometry and energetics of intrinsic affinities, are unknown. We previously analyzed NS5 interactions with ssRNA and double-stranded RNA (dsRNA), as well as ssDNA and dsDNA, using fluorescence titration techniques (29). Here, we report studies of the interaction between NS5 and SLA and describe the effects of salt, magnesium, and temperature on NS5-SLA binding affinity. Moreover, a competition assay shows that SLA and ssRNA compete for the same binding site on NS5 polymerase and indicates that there is a 1:1 stoichiometry of NS5 to SLA in the NS5-SLA complex.

RESULTS

SLA₈₀ forms a stable monomer in solution and adopts a more compact conformation with higher magnesium concentrations.

Flavivirus polymerase specifically recognizes the viral 5′ UTR during viral RNA replication (24, 25). In particular, SLA within the 5′ UTR is predicted to form a “Y”-shaped secondary structure and serves as a promoter for NS5 polymerase binding and activity (25, 27). Specific binding between NS5 and SLA was observed when SLA (70 nucleotides [nt]) had an additional ssRNA region at the 3′ end (Fig. 1) (25, 27). We thus constructed SLA with an additional 10 nt, containing the first 80 nt of the dengue virus genome, referred to here as SLA₈₀, to study the interaction between SLA and NS5. We hypothesized that the stem-loop structure of SLA₈₀ would be stabilized by cations that neutralize the negatively charged phosphate backbone, thus allowing close packing of the RNA. Therefore, we first determined the changes in the shape of SLA₈₀ in the presence of Mg²⁺ by analytical ultracentrifugation (AUC). Sedimentation velocity AUC experiments measure the rate at which molecules move through solution and thus reflect hydrodynamic properties of macromolecules that are related to the size and shape of molecules. The size and shape of SLA₈₀ were determined in the presence of 0, 1, and 10 mM MgCl₂ (Table 1). The overall size and shape of the molecule were similar with both 0 and 1 mM MgCl₂, with a Stokes radius of ∼3.5 nm, and low magnesium concentrations (1 mM) did not have any effect on the overall shape of the molecule. However, at 10 mM magnesium, SLA₈₀ acquires a more compact shape, reflected in a ∼20% decrease in the Stokes radius from ∼3.5 nm to ∼2.7 nm. The frictional ratio also decreased from 1.8 to 1.6, indicating that SLA₈₀ becomes more compact in the presence of 10 mM magnesium. The behavior of SLA₈₀ in the presence of different concentrations of divalent cations is similar to those of previously studied structured RNAs such as the turnip yellow mosaic virus, tobacco mosaic virus, and brome mosaic virus RNAs, all of which become more compact at higher (10 mM) Mg²⁺ concentrations (30).

FIG 1.

Structure of SLA at the 5′ UTR of the DENV genome. The sequences for the first 80 nt from DENV1 to -4 genomes are shown, and nonconserved nucleotides are shown in blue. The secondary structures of DENV2 and DENV3 5′ UTRs were predicted by using the Mfold Web server (42).

TABLE 1.

Effect of magnesium on the size and shape of SLA₈₀ as measured by AUC

Mg concn (mM)	S20	f/f0	RH (nm)a
0	2.17	1.8	3.53
1	2.14	1.8	3.51
10	2.32	1.6	2.74

^aR_H, Stokes radius.

SLA₈₀ binds specifically to DENV NS5.

The binding constant for NS5 and SLA₈₀ was determined by using SLA₈₀ labeled with fluorescein at the 3′ end (SLA₈₀-F) at 10°C. The dependence of the relative fluorescence intensity of SLA₈₀-F (ΔF_obs) on the NS5 concentration is shown in Fig. 2. Binding of NS5 to SLA₈₀-F (1.5 × 10⁻⁸ M) in buffer B1 (50 mM Tris [pH 7.5], 100 mM NaCl, 1 mM dithiothreitol [DTT], and 10% glycerol) led to quenching of the marker emission intensity. The maximum observed value for fluorescence quenching is ∼17%. The titration curve was analyzed by using the approach described in Materials and Methods (equation 5). The nonlinear least-squares fit using K₁ and ΔF_max as the only two fitting parameters provided a binding constant, K₁, equal to 1.8 × 10⁶ M⁻¹ and thus a dissociation constant (K_d) of 0.55 μM (Fig. 2, blue circles).

FIG 2.

Interactions of fluorescein-labeled SLA₈₀ with dengue virus NS5. Fluorescence titrations of fluorescein-labeled SLA₈₀ with DENV3 NS5 (λ_ex = 480 nm; λ_em = 520 nm) were carried out by using standard buffer (100 mM NaCl, 50 mM Tris [pH 7.5], 10% glycerol, and 1 mM DTT at 10°C) containing the following different MgCl₂ concentrations: 0 mM, 0.4 mM, 1.0 mM, 1.3 mM, and 2.0 mM. The concentration of fluorescein-labeled SLA₈₀ is 1.5 × 10⁻⁸ M. The solid lines are nonlinear least-squares fits of the titration curve with a K of 1.8 × 10⁶ M⁻¹ (0 mM MgCl₂), a K of 2.2 × 10⁶ M⁻¹ (0.4 mM MgCl₂), a K of 5.5 × 10⁶ M⁻¹ (1.0 mM MgCl₂), a K of 2.3 × 10⁶ M⁻¹ (1.3 mM MgCl₂), and a K of 1.8 × 10⁶ M⁻¹ (2.0 mM MgCl ₂).

Next, we tested how the magnesium concentration affects the formation of the NS5-SLA₈₀ complex, since previous binding studies of NS5 and ssRNA showed that magnesium affects the affinity of NS5 for ssRNA (29). Fluorescence titrations of SLA₈₀-F with NS5 in buffer B1 containing 0, 0.4, 1.0, 1.3, and 2.0 mM MgCl₂ at 10°C are shown in Fig. 2. An increase of the magnesium concentration from 0 to 1 mM enhanced the binding constant from a K_{0 mM Mg} of 1.8 × 10⁶ M⁻¹ to a K_{1 mM Mg} of 5.5 × 10⁶ M⁻¹ (K_d of 181 nM). However, a further increase to 2 mM MgCl₂ did not enhance NS5-SLA₈₀ affinity but rather reduced it to the value measured with 0 mM MgCl₂. Furthermore, NS5-SLA₈₀ interactions were not detectable at a higher magnesium concentration of 10 mM; i.e., no fluorescence change was observed upon the addition of NS5. SLA₈₀ was shown to adopt a more compact conformation with 10 mM MgCl₂ by AUC (Table 1), and thus, the compact form of SLA observed in the presence of 10 mM MgCl₂ may not be able to bind NS5. Our results indicate that the optimal magnesium concentration for interactions between NS5 and SLA₈₀ is ∼1 mM.

Effect of salt on NS5-SLA₈₀ interactions.

To determine the salt dependence of NS5 binding to SLA₈₀, fluorescence titrations of SLA₈₀-F with NS5 were carried out in buffer B1 containing 80, 100, 130, 170, and 200 mM NaCl and 1 mM MgCl₂ at 10°C (Fig. 3A). As the salt concentration increased, the maximum fluorescence change at saturation, ΔF_max, decreased from ∼19% at 80 mM to ∼10% at 200 mM NaCl. The affinity of NS5 for SLA₈₀ decreased with increasing salt concentrations, from a K of 7.0 × 10⁶ M⁻¹ (K_d of 142 nM) for 80 mM NaCl to a K of 2.0 × 10⁶ M⁻¹ (K_d of 500 nM) for 200 mM NaCl. The dependence of the logarithm of the intrinsic binding constant on the logarithm of [NaCl] (log-log plot) is shown in Fig. 3B (29, 31). This plot is linear in the studied salt concentration range, and the slope, ∂logK/∂log[NaCl], is −1.3, which indicates that the release of ∼1 Na⁺ ion accompanies the intrinsic interactions between NS5 and SLA₈₀ (31).

FIG 3.

Effect of salt on SLA₈₀-NS5 binding. (A) Fluorescence titrations of fluorescein-labeled SLA₈₀ with NS5 (λ_ex = 480 nm; λ_em = 520 nm) in buffer B1 (50 mM Tris [pH 7.5] at 10°C, 1 mM MgCl₂) containing the following different NaCl concentrations: 80 mM, 100 mM, 130 mM, 170 mM, and 200 mM. The concentration of fluorescein-labeled SLA₈₀ is 1.5 × 10⁻⁸ M. Increasing salt concentrations decrease the affinity of the polymerase for SLA₈₀. The solid lines are nonlinear least-squares fits of the titration curve with a K of 7.0 × 10⁶ M⁻¹ (80 mM NaCl), a K of 5.5 × 10⁶ M⁻¹ (100 mM NaCl), a K of 4.0 × 10⁶ M⁻¹ (130 mM NaCl), a K of 3.5 × 10⁶ M⁻¹ (170 mM NaCl), and a K of 2.0 × 10⁶ M⁻¹ (200 mM NaCl). (B) Dependence of the logarithm of the intrinsic binding constant, K, on the logarithm of NaCl concentrations. The solid line is the linear least-squares fit of the NaCl concentration regions of the plot, which provide the slope, ∂logK/∂log[NaCl] = −1.3.

Effect of temperature on NS5-SLA₈₀ interactions.

To further characterize the nature of the SLA₈₀-NS5 interaction, we examined the effect of temperature on complex formation. Fluorescence titrations of SLA₈₀-F with NS5 were performed in buffer B1 in the presence of 1 mM MgCl₂ at 10, 20, and 25°C (Fig. 4). The titration curves were identical for all three temperatures, and nonlinear least-squares fitting, using equation 5 with K₁ and ΔF_max as the only two fitting parameters, yielded a K value of 5.5 × 10⁶ M⁻¹ (K_d of 181 nM). Thus, neither the intrinsic affinity nor the value of ΔF_max is affected by temperature in the range from 10°C to 25°C. Thus, the interaction of NS5 with SLA₈₀ is accompanied by an apparent enthalpy change, ΔH°, of ∼0. Hence, the intrinsic interactions are independent of temperature and are predominantly driven by the apparent entropy change, ΔS°.

FIG 4.

Effect of temperature on SLA₈₀-NS5 binding. Shown are fluorescence titrations of fluorescein-labeled SLA₈₀ with NS5 (λ_ex = 480 nm; λ_em = 520 nm) in buffer (50 mM Tris [pH 7.5], 100 mM NaCl, 1 mM MgCl₂) at 10, 20, and 25°C. The concentration of fluorescein-labeled SLA₈₀ is 1.5 × 10⁻⁸ M. The solid line is a nonlinear least-squares fit of both titration curves with a K of 5.5 × 10⁶ M⁻¹.

SLA₈₀ and single-stranded εA(pεA)₁₉ compete for the same binding site on NS5.

Although the DENV NS5 polymerase is known to specifically recognize SLA to initiate RNA replication, the SLA binding site on NS5 is not known. To determine whether the SLA binding site on NS5 overlaps the template binding site of the RdRp domain, direct competition studies of the binding of NS5 to the fluorescent etheno-adenosine 20-mer εA(pεA)₁₉ were carried out in the presence of unlabeled SLA₈₀. The etheno-derivative of A(pA)₁₉, εA(pεA)₁₉, was obtained by modification of A(pA)₁₉ with chloroacetaldehyde (see Materials and Methods). Binding of NS5 to εA(pεA)₁₉ enhanced the fluorescence emission intensity of the nucleic acid. The relative fluorescence signal (ΔF_obs) of εA(pεA)₁₉ (8.0 × 10⁻⁷ M) as a function of the logarithm of the NS5 concentration is shown in Fig. 5A (blue). The maximum relative change of the fluorescence signal is 68%. The binding of εA(pεA)₁₉ to NS5 in the absence of SLA₈₀ was analyzed by using equation 5 and yielded a K_ssRNA of 2.2 × 10⁶ M⁻¹. When DENV3 SLA₈₀ (8.0 × 10⁻⁷ M) was present, the titration curve shifted toward higher NS5 concentrations, but the maximum value of the relative fluorescence signal remained unchanged (Fig. 5A, green). This result indicates competition between εA(pεA)₁₉ and SLA₈₀ for the same binding site on NS5. Analysis of the binding curve was performed by using equations 6 and 7 and yielded a K_SLA80 of 7.0 × 10⁶ M⁻¹ (K_d of 142 nM).

FIG 5.

SLA₈₀ and εA(pεA)₁₉ compete for the same binding site on dengue virus NS5. (A) Fluorescence titrations of εA(pεA)₁₉ with DENV3 NS5 (λ_ex = 325 nm; λ_em = 410 nm) in standard buffer (100 mM NaCl, 50 mM Tris [pH 7.5], and 10% glycerol at 10°C) in the absence and presence of DENV3 SLA₈₀ at two concentrations, 8.0 × 10⁻⁷ M and 2.5 × 10⁻⁶ M. The concentration of εA(pεA)₁₉ is 8.0 × 10⁻⁷ M. The solid lines are nonlinear least-squares fits of the titration curve with a K of 2.2 × 10⁶ M⁻¹ (no SLA₈₀), a K of 7.0 × 10⁶ M⁻¹ (0.8 μM SLA₈₀), and a K of 7.5 × 10⁶ M⁻¹ (2.5 μM SLA₈₀). (B) Fluorescence titrations of εA(pεA)₁₉ with DENV3 NS5 (λ_ex = 325 nm; λ_em = 410 nm) in the absence and presence of DENV2 SLA₈₀ at two concentrations, 8 × 10⁻⁷ M and 2.5 × 10⁻⁶ M. The solid lines are nonlinear least-squares fits of the titration curve with a K of 2.2 × 10⁶ M⁻¹ (no SLA₈₀), a K_SLA80 of 1.0 × 10⁷ M⁻¹ (0.8 μM SLA₈₀), and a K_SLA80 of 1.2 × 10⁷ M⁻¹ (2.5 μM SLA₈₀).

At higher SLA₈₀ concentrations {2.5 × 10⁻⁶ M [3-fold molar excess over εA(pεA)₁₉]}, the titration curve was further shifted toward higher NS5 concentrations (Fig. 5A, red). Analysis of the binding reaction (equations 6 and 7) yielded a K_SLA80 of 7.5 × 10⁶ M⁻¹ (K_d of 133 nM), the same as that obtained in the presence of 8.0 × 10⁻⁷ M SLA₈₀ within experimental accuracy. Thus, the change of the SLA₈₀ concentration did not affect the affinity of NS5 for SLA₈₀, indicating that the 1:1 stoichiometry of the NS5-SLA₈₀ complex remained unchanged.

SLAs at the 5′ end of the DENV1 to -4 genomes share between 75% (between DENV2 and DENV4) and 96% (between DENV1 and DENV3) sequence identities (Fig. 1). All SLA sequences are predicted to have a similar Y-shaped stem-loop structure (32). It was shown previously that DENV2 NS5 can use DENV1 SLA as a promoter to initiate RNA synthesis (32), suggesting that NS5 can bind SLA from different serotypes. We thus tested whether DENV3 NS5 binds DENV2 SLA₈₀ using the competition titration assay described above. NS5-εA(pεA)₁₉ complex formation was measured at two different concentrations of DENV2 SLA₈₀, 8.0 × 10⁻⁷ and 2.5 × 10⁻⁶ M (Fig. 5B). Similar to the NS5 interaction with DENV3 SLA₈₀ (Fig. 5A), the titration curves were shifted toward higher NS5 concentrations, indicating that DENV3 NS5 binds DENV2 SLA. DENV3 NS5 has a slightly higher affinity for DENV2 SLA₈₀ than for its own serotype, with a K_SLA80 of 1.0 × 10⁷ M⁻¹ (K_d of 100 nM with 8.0 × 10⁻⁷ SLA₈₀). Further increasing the DENV2 SLA₈₀ concentration to 2.5 × 10⁻⁶ M did not increase the binding constant (K_SLA80 = 1.2 × 10⁷ M⁻¹; K_d of 83 nM), indicating that DENV3 NS5 interacts with DENV2 SLA with 1:1 stoichiometry. Therefore, DENV3 NS5 recognizes SLAs of other DENV serotypes and thus most likely can replicate the RNA genomes from other serotypes, consistent with data from previous reports (32).

SLA₇₀ binds specifically to full-length DENV NS5.

So far, we examined the interactions between NS5 and SLA₈₀, which has the 10-nt ssRNA region at the 3′ end of SLA. To determine whether DENV NS5 recognizes SLA itself, we analyzed interactions between SLA₇₀ (the stem-loop only) (Fig. 1) and the enzyme using a competition assay. The dependence of the relative fluorescence intensity of εA(pεA)₁₉ (ΔF_obs) on the logarithm of the NS5 concentration, in the absence and presence of SLA₇₀, is shown in Fig. 6. The binding of NS5 to εA(pεA)₁₉ in the absence of SLA₇₀ has a K_ssRNA value of 2.2 × 10⁶ M⁻¹ (Fig. 6, blue squares). The presence of SLA₇₀ clearly shifted the binding curve toward higher NS5 concentrations, while the maximum value of the relative fluorescence signal remained unchanged (Fig. 6, green squares). The nonlinear least-squares fit provided a K_SLA70 of 1.2 × 10⁷ M⁻¹ (equations 6 and 7) and, thus, a K_d of 83 nM. This result indicates that SLA₇₀ efficiently competes with εA(pεA)₁₉ for the same binding site, and NS5 recognizes and binds to SLA₇₀ as well as SLA₈₀.

FIG 6.

SLA₇₀ binds to full-length NS5. Shown are fluorescence titrations of εA(pεA)₁₉ with DENV3 NS5 (λ_ex = 325 nm; λ_em = 410 nm) in standard buffer (100 mM NaCl, 50 mM Tris [pH 7.5], and 10% glycerol at 10°C) in the absence and presence of DENV3 SLA₇₀. The concentration of εA(pεA)₁₉ is 8.0 × 10⁻⁷ M. The solid lines are nonlinear least-squares fits of the titration curve with a K of 2.2 × 10⁶ (no SLA₇₀) and a K_SLA70 of 1.2 × 10⁷ M⁻¹ (0.8 μM SLA₇₀).

DISCUSSION

Replication of the viral genome is the primary goal of any viral infection, and thus obtaining quantitative information on the initial step of binding between NS5 and the RNA promoter is essential for understanding virus replication. We have characterized DENV3 NS5-SLA₈₀ interactions under various solution conditions. DENV3 NS5 binds SLA₈₀ with a dissociation constant of 130 to 140 nM, measured by either direct binding or competition assays (Fig. 2 and 5). The NS5-SLA₈₀ interaction is affected by MgCl₂ concentrations, and binding is strongest when the Mg²⁺ concentration is ∼1 mM. At higher Mg²⁺ concentrations (>1 mM), the affinity of NS5 for SLA₈₀ is decreased. Further increases of Mg²⁺ concentrations to 10 mM, where SLA adopts a more compact form, diminish the polymerase affinity to hardly detectable levels in standard fluorescence titration experiments. This result suggests that there are two types of Mg²⁺ binding sites in the NS5-SLA₈₀ complex, one that promotes the interaction between NS5 and SLA₈₀ and the other that inhibits the interaction.

The NS5-SLA interaction is also influenced by the NaCl concentration, and ∼1 Na⁺ ion is released upon NS5-SLA₈₀ complex formation (Fig. 3B). This result is in contrast to the NS5-ssRNA interaction, where the release of ∼6 Na⁺ ions during complex formation was observed (29). The weaker effect of NaCl observed for the NS5-SLA interaction indicates that the binding-site size (the numbers of nucleotides and amino acids involved in interactions) for SLA is smaller than that for ssRNA (Fig. 7). Unlike ssRNA, SLA acquires a defined structure, and thus, the data strongly suggest that SLA has a lower number of direct contacts with (but a higher affinity for) NS5 than does the ssRNA. Additionally, the NS5-SLA₈₀ interaction is not influenced by temperature (Fig. 4), while the binding of NS5 and ssRNA is affected (29). The fact that the association reaction between NS5 and SLA₈₀ is driven by entropy (Fig. 4) suggests that the interacting surfaces of the protein and SLA are not affected by complex formation; i.e., no additional structural adjustment is required as a result of complex formation, and the solvent and ions released from the complex drive the reaction.

FIG 7.

Model of the NS5-RNA interaction. NS5 consists of the MTase domain (yellow) and the RdRp domain (blue). The arrangement of the MTase and RdRp domains is depicted based on the crystal structures of DENV NS5 (PDB accession no. 5CCV). The template binding channel in RdRp is formed by the inner surfaces of the fingers and the thumb subdomains and leads to the active site in the palm subdomain (indicated by the GDD motif). An ssRNA binds in the template binding channel of NS5 with multiple interactions (left). The SLA₇₀ or SLA₈₀ binding site overlaps the template binding channel but involves less contact with NS5 than with ssRNA. The RNA molecules are shown in red, and interaction sites are depicted with dashed lines.

SLA is predicted to interact with NS5 via the RdRp domain, since full-length NS5 and the isolated RdRp domain can promote RNA synthesis and have similar binding affinities for SLA in electrophoretic mobility shift assays (EMSAs) (24, 25). However, it is not known where SLA binds on the RdRp domain. It seems unlikely that SLA binds in the template binding channel of RdRp, because crystal structures of NS5 and RdRp domains show a narrow template binding channel that is large enough to accommodate only an ssRNA. Competition assays between ssRNA (20-mer) and SLA₈₀ showed that SLA₈₀ competes with the ssRNA and that NS5 and SLA₈₀ bind with a 1:1 stoichiometry (Fig. 5). This result indicates that the binding site for SLA₈₀ on NS5 likely overlaps the template binding site in the RdRp domain, although direct contacts between the protein and the nucleic acid are likely different in the two complexes (Fig. 7). Recent studies of the interaction of DENV RdRp with the RNA genome suggest that the 3′ stem-loop is recognized by the thumb subdomain of RdRp, and the polymerase mutant that abolished the 3′ stem-loop interaction also eliminated 5′ SLA interactions (33). In the crystal structure of DENV NS5, the residues implicated in 3′ stem-loop/SLA binding in the thumb subdomain (R770, R773, R856, Y838, and K841) are exposed to the solvent and are away from the MTase domain. Thus, the proposed Arg-rich site in the thumb subdomain could be the SLA binding site in RdRp (Fig. 7).

DENV3 NS5 is able to bind DENV2 SLA₈₀ with a slightly higher affinity (83 nM) than that for DENV3 SLA₈₀ (Fig. 5). This affinity is slightly lower than the value reported for NS5 and SLA in the DENV2 system (10 to 20 nM), as measured by using EMSAs (24). Since DENV3 NS5 recognizes both DENV2 and DENV3 SLAs, the NS5 proteins likely recognize the fold and the sequence of SLA (Fig. 1). The DENV2 SLA regions important for NS5 binding and viral replication were previously identified (24, 25, 32). The base pairings at the bottom stem-loop (⁶UUA/⁶⁵UAA and ¹²UAC/⁵⁸GUA) (Fig. 1) and at the side stem-loop (⁴⁴GAGC/⁵¹GCUC) are required for both NS5 binding and viral replication, whereas a disruption of base pairing at the top stem-loop (²⁶AAG/³⁶CUU) did not significantly affect NS5 binding or viral replication (21). The major difference between DENV2 and DENV3 SLA structures lies in the top stem-loop containing nt 22 to 41 (Fig. 1), which does not affect the association with NS5. Thus, DENV NS5 likely recognizes both DENV2 and -3 SLA₈₀s, consistent with our results.

Furthermore, DENV NS5 binds SLA₇₀ with an affinity similar to that determined for NS5-SLA₈₀ binding (Fig. 6). The SLA₇₀ construct encompasses the entire stem-loop but does not possess the additional 10 nucleotides at the 3′ end of SLA₈₀. The ssRNA region at the 3′ end following SLA was shown to be required for the initiation of the polymerization reaction (25). The RdRp domain of NS5 does not bind to SLA without the ssRNA tail (i.e., SLA₇₀) but binds SLA containing the 3′ ssRNA tail (such as SLA₁₀₀) in EMSAs. Our results indicate that DENV3 NS5 binds both SLA₇₀ and SLA₈₀ with similar affinities, and thus, the NS5 polymerase recognizes the secondary structure and sequence of SLA, rather than the 10-nt ssRNA region at the 3′ end of SLA₈₀. The reason for the discrepancy between data from previous studies and those from our studies could be due to different NS5 constructs (the RdRp domain versus full-length NS5), experimental methods (EMSA versus solution fluorescence), or the length of RNA used (100 nt versus 80 nt). Considering the similarities between the UTRs and NS5 sequences among all flaviviruses, our analysis of DENV3 NS5-SLA complex formation is likely applicable to other flaviviruses and should help in the development of antiviral therapeutics against flaviviral diseases.

MATERIALS AND METHODS

Reagents and buffers.

DENV3 SLAs (GenBank accession no. KU725665) containing the first 70 nucleotides (SLA₇₀) and the first 80 nucleotides (SLA₈₀) with and without a fluorescein tag at the 3′ end, DENV2 SLA₈₀ (GenBank accession no. KU725663), and an adenosine 20-mer, A(pA)₁₉, were synthesized (Midland Certified Reagents, Midland, TX). The concentrations of the nucleic acids were determined by using the following extinction coefficients: ϵ₂₆₀ of 10,300 cm⁻¹ M⁻¹ for poly(A) and ϵ₂₆₀ of 685,200 cm⁻¹ M⁻¹ for SLA. The etheno-derivative of A(pA)₁₉, εA(pεA)₁₉, was obtained by modification of A(pA)₁₉ with chloroacetaldehyde. This modification goes to completion and provides a fluorescent derivative of the nucleic acid (29, 34, 35).

Expression and purification of DENV3 NS5.

Full-length DENV3 NS5 with a C-terminal hexahistidine tag was expressed in Escherichia coli BL21-CodonPlus-RIL cells (20). Briefly, the cells were grown at 37°C in Luria broth (LB) containing 34 μg/ml kanamycin and 30 mg/ml chloramphenicol to an absorbance at 600 nm of 0.6 to 0.7. The expression of the protein was induced by the addition of 1 mM isopropyl-β-d-1-thiogalactopyranoside. Cell growth was continued overnight at 18°C. For protein purification, the cells were suspended in lysis buffer (100 mM sodium phosphate [pH 8.0], 0.5 M NaCl, 2 mM β-mercaptoethanol, and 1 tablet of an EDTA-free protease inhibitor [Roche Applied Science, Penzberg, Germany]) and sonicated. After centrifugation, the protein in the soluble fraction was purified by Talon metal affinity chromatography (Clontech, Mountain View, CA). NS5 was eluted with a 5 to 100 mM gradient of imidazole in 100 mM sodium phosphate (pH 7.0) buffer containing 0.5 M NaCl and 2 mM β-mercaptoethanol. The fractions containing NS5 were collected and concentrated to 1 to 2 ml by using an Ultrafree centrifugal filter device (Millipore, Billerica, MA). Subsequently, the pooled fractions were loaded onto a Superdex 200 size exclusion column (GE Healthcare, Little Chalfont, UK) equilibrated with a solution containing 20 mM Tris-HCl (pH 7.0), 0.4 M NaCl, and 2 mM DTT. DENV3 NS5 elutes as a monomer. The purity of the protein, estimated by SDS-PAGE, was >95%. The concentration of NS5 was spectrophotometrically determined with an extinction coefficient, ϵ₂₈₀, of 21.2198 × 10⁴ cm⁻¹ M⁻¹ obtained by using an approach based on the method of Edelhoch (36).

Fluorescence measurements.

Steady-state fluorescence titrations were performed by using an ISS PC1 spectrofluorometer (ISS, Urbana, IL). In order to avoid possible artifacts due to fluorescence anisotropy of the sample, polarizers were placed in excitation and emission channels and set at 90° and 55° (magic angle), respectively. The binding of DENV NS5 to the RNA was monitored by the fluorescence of either εA(pεA)₁₉ (excitation wavelength [λ_ex] = 325 nm; emission wavelength [λ_em] = 410 nm) or fluorescein attached to SLA₈₀ (λ_ex = 480 nm; λ_em = 520 nm). Binding curves were fit by using KaleidaGraph software (Synergy Software, PA). In the case of titrations using competition with εA(pεA)₁₉, the relative increase in the fluorescence of the nucleic acid (ΔF) upon binding of DENV NS5 is defined as ΔF_obs = (F_i − F₀)/F₀, where F_i is the fluorescence of the nucleic acid solution at a given titration point, i, and F₀ is the initial fluorescence of the sample (29, 37–39). In the case of direct titrations of fluorescein-modified SLA, the relative fluorescence change is defined as F_i/F₀. Both F₀ and F_i are corrected for background fluorescence at the applied excitation wavelength (29, 37, 39). Standard B1 buffer, containing 50 mM Tris-HCl adjusted to pH 7.5 with HCl at a given temperature, 100 mM NaCl, 1 mM DTT, and 10% (wt/vol) glycerol, was used for all binding experiments. All experiments were performed in triplicate.

Fluorescence signal analysis.

The binding constant, K₁, characterizing the association of SLA₈₀ with NS5, is defined as

$$K_1=\frac{{[C_1]}_F}{{[{\text{SLA}}_{80}]}_F{[\text{NS}5]}_F}$$ (1)

where [C₁]_F is the concentration of the formed complex.

The observed fluorescence of the sample at any point of the titration is defined as

$$F_{\text{obs}}=F_F{[{\text{SLA}}_{80}]}_F+F_C{[C_1]}_F$$ (2)

where F_F and F_C are the molar fluorescence intensities of the free SLA₈₀ and the formed complex, respectively. Thus,

$$F_{\text{obs}}=F_F{[{\text{SLA}}_{80}]}_F+F_C{K}_1{[{\text{SLA}}_{80}]}_F{[\text{NS}5]}_F$$ (3)

The mass conservation equation for the total SLA₈₀ concentration, [SLA₈₀]_T, in the sample is

$$\begin{array}{c}{[{\text{SLA}}_{80}]}_T={[{\text{SLA}}_{80}]}_F+{[C_1]}_F={[{\text{SLA}}_{80}]}_F\left(1+K_1{[\text{NS}5]}_F\right)\end{array}$$ (4)

Using equations 3 and 4, the relative observed change of the SLA₈₀ fluorescence, ΔF_obs, is then

$$\Delta F_{\text{obs}}=\frac{F_{\text{obs}}}{F_F{[{\text{SLA}}_{80}]}_T}=\frac{1}{1+K_1{[\text{NS}5]}_F}+\Delta F_{\max \hspace{0.17em}}\left(\frac{K_1{[\text{NS}5]}_F}{1+K_1{[\text{NS}5]}_F}\right)$$ (5)

where ΔF_max = F_C/F_F is the maximum value for the observed relative fluorescence quenching.

Competition assay.

Binding of εA(pεA)₁₉ and NS5 was carried out in buffer B1 containing an additional 1 mM MgCl₂ at 10°C by monitoring the changes of the fluorescence signal originating from εA(pεA)₁₉ (λ_ex = 325 nm; λ_em = 410 nm) (29). The binding constant was determined by using equation 5. In the competition assay, labeled εA(pεA)₁₉ (each 8.0 × 10⁻⁷ M) and unlabeled SLA₈₀ (8.0 × 10⁻⁷ M or 2.5 × 10⁻⁶ M) or SLA₇₀ (8.0 × 10⁻⁷ M) were mixed in buffer B1 containing 1 mM MgCl₂ and titrated with NS5 at 10°C. The change of the fluorescence signal originates only from the binding of εA(pεA)₁₉ to the protein. The shift of the titration curve is a result of competition between εA(pεA)₁₉ and either SLA₈₀ or SLA₇₀ for the binding site on NS5. Competition was analyzed by using equations 6 and 7, where K₁ and K₂ correspond to the binding constants characterizing the association of εA(pεA)₁₉ with NS5 and the binding constant characterizing the association of SLA₈₀ (or SLA₇₀) with NS5, respectively (38). Experiments were repeated three times.

$$\Delta F_{\text{obs}}=\frac{F_{\text{obs}}}{F_F{[{\text{SLA}}_{80}]}_T}=\frac{1}{1+K_1{[\text{NS}5]}_F}+\Delta F_{\max \hspace{0.17em}}\left(\frac{K_1{[\text{NS}5]}_F}{1+K_1{[\text{NS}5]}_F}\right)$$ (6)

where the total NS5 concentration is

$${[\text{NS}5]}_{\text{total}}={[\text{NS}5]}_F+\frac{K_1{[\text{εA}(\text{εA})}_{19}{]}_T{[\text{NS}5]}_F}{1+K_1{[\text{NS}5]}_F}+\left(\frac{K_2{[{\text{SLA}}_{80}]}_T{[\text{NS}5]}_F}{1+K_1{[\text{NS}5]}_F}\right)$$ (7)

Analytical ultracentrifugation measurements.

The size and shape of SLA₈₀ were analyzed by analytical ultracentrifugation using an Optima XL-A analytical ultracentrifuge (Beckman Inc., Palo Alto, CA), as we described previously (29, 40, 41). Sedimentation equilibrium scans at different MgCl₂ concentrations (0, 1, and 10 mM) were collected at the absorption band of SLA₈₀ (260 nm). Time-derivative analyses of the sedimentation scans were performed with the software supplied by the manufacturer, using averages of 8 to 15 scans for each concentration.