Characterization of interactions between hepatitis C virus NS5B polymerase, annexin A2 and RNA – effects on NS5B catalysis and allosteric inhibition

The ectodomain of NS5B (NS5BΔ21) from genotype 1b was produced as earlier described, with the details of collection of blood samples, viral RNA extraction, cDNA formation, and cloning of NS5B 1b DNA into an expression vector published in [20] and methods for protein expression and purification published in [21]. HCV NS3 and HCV NS3-4A genotype 1a were produced as previously described [20].

Bovine AnxA2 and an engineered variant containing mutations in helix C and in the CD loop (Ser substitutions of Lys308, Lys309 and Lys310 in helix C, Lys313 in the CD loop and Tyr317 and Gln321 in helix D) (mAnxA2) were expressed and purified as previously described [19].

The 31-mer RNA (5’CGAUACUCCCUUUAUAUAACCAUCAAUCGCC 3′) used in SPR polymerase assay [22], was synthesized as previously described [23]. Briefly, the DNA oligos:

5′ ATTCGTTAATACGACTCACTATAGGG 3′ and.

5′ GGCGATTGATGGTTATATAAAGGGAGTATCGCCCTATAGTGAGTCGTATTA 3′.

were used as a template for the 31 bp RNA synthesis. The primers were annealed by heating to 100 °C for 1 min in 50 mM Tris-HCl pH 7.4 and 100 mM KCl buffer. The RNA was synthesized using Riboprobe T7 kit (Promega, Fitchburg, WI, USA) according to the manufacturer’s instructions. The synthesized RNA was purified with standard phenol/chlorophorm extraction procedure. Unincorporated nucleotides were removed using PD SpinTrap G-25 spin columns (GE Healthcare, Uppsala, Sweden).

The 8-mer RNA (5’GGG GAU UG-3′) used in interaction studies with AnxA2 and NS5B was purchased from Eurofins Genomics.

SPR measurements were performed at 25 °C using a Biacore S51 or T200 instrument and the data was analyzed using Biacore T200 evaluation software version 1.0 (GE Healthcare, Uppsala, Sweden). Experiments were performed using HBS-EP running buffer (0.01 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) pH 7.4, 0.15 M NaCl, 3 mM ethylene diamine tetra acetic acid (EDTA), 0.005% v/v Surfactant P20) in the presence of 70 μM Ca²⁺ and 1 mM Mg²⁺, using a flow rate of 30 μl/min. AnxA2 or NS5B (genotype 1b BK) were immobilized on a CM5 chip by standard amine coupling using 5 mM maleate buffer pH 6 and NaAc buffer pH 6, respectively, as immobilization buffer. Multi cycle experiments with regeneration after each injection (with 2 M NaCl and 2 M MgCl₂), or single cycle kinetic experiments, without regeneration between the injections, were performed with four or five concentrations of the analyte in HBS-EP running buffer. In experiments investigating the interplay between AnxA2, NS5B and RNA, AnxA2 was immobilized as above, and either 8-mer RNA or NS5B was injected to form a stable binary complex, followed by injection of the third interaction partner.

The sensorgrams were corrected for buffer bulk effects and unspecific binding of the samples to the chip matrix by blank and reference surface subtraction (subtraction of inactivated and deactivated flow cell channel or where NS5B was immobilized and surface inactivated by 3 × 30 s injections of 6 M guanidine-HCl).

The association rates (k _a), dissociation rates (k _d), dissociation constants (K_D) and maximum binding responses (R_max) were estimated by global non-linear regression analysis and a reversible 1-step interaction reaching steady-state at the end of the analyte injection or by fitting the sensorgram to a reversible 1-step interaction model. The analysis was based on report points taken at the end of analyte injections even if steady state had not been reached. In cases where the interaction mechanism was not established, the analysis is purely qualitative and any estimated K_D-values are termed “apparent” (K_D ^app), only useful for discussions on the possible order of magnitude of affinities.

A continuous de novo HCV NS5B polymerase assay, using surface plasmon resonance biosensor technology (Biacore T200, GE Healthcare, Uppsala, Sweden), was developed, as illustrated in Fig. 3a (from [23]). Streptavidin (Sigma-Aldrich, St. Louis, MO, USA) was diluted to 100 μg/ml in the buffer containing 10 mM NaAc pH 5, 0.1 mM EDTA, 1 mM NaCl, 1 mM dithiothreitol (DTT) and immobilized on the CM5 chip surface by standard amine coupling procedure, resulting in 4500 Refractive units (RU). The surface was washed 3 times with conditioning solution (1 M NaCl, 50 mM NaOH) to remove excess unbound streptavidin. Subsequently, about 500–700 RU of biotinylated ssDNA oligo diluted to 660 nM in the running buffer (10 mM Tris-HCl pH 7.4, 50 mM NaCl, 1 mM EDTA, 1 mM β-mercaptoethanol and 0.005% Tween 20) was captured by streptavidin-biotin interaction. This DNA oligo has a complementary sequence to the in vitro synthesized single-stranded (ss) 31-mer RNA, which was hybridized in the next step to an approximate level of 200–300 RU. The ss31-mer RNA, an oligomer with at sequence originally designed for a de novo polymerase assay [22], was diluted in the same buffer as ssDNA to a final concentration of 20 ng/μl. The NS5B Δ21 polymerase and/or NS5B Δ21 supplemented with 750 μM ribo nucleotide triphosphates (rNTPs) (Promega, Fitchburg, WI, USA) were injected over the surface containing the DNA/RNA hybrid and over the reference surface with immobilized streptavidin. The effect of AnxA2 or mAnxA2 on polymerase activity was tested by adding 100 nM or 200 nM of AnxA2/mAnxA2 to the NS5B/rNTPs mixture. The assay was performed using HBS-EP running buffer (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v Surfactant P20) in the presence of 70 μM Ca²⁺ and 1 mM Mg²⁺. The data was analyzed using Biacore T200 evaluation software version 1.0 (GE Healthcare). The NS3 based enzyme activity assay was performed according to the protocol described previously [24].

To investigate a possible direct interaction between AnxA2 and NS5B polymerase, AnxA2 was immobilized on an SPR sensor chip and NS5B was run as analyte (Fig. 1a). The experiment demonstrated that AnxA2 and NS5B formed a stable complex with a very slow rate of dissociation. The interaction between the two proteins had to be actively broken for the signal to return to baseline before a new sample of analyte could be injected. A regeneration solution of high ionic strength (2 M NaCl and 2 M MgCl₂) was found to be suitable.

To confirm the interaction, the reverse experiment was performed, i.e. with NS5B immobilized on the chip and AnxA2 used as analyte. Moreover, to avoid artefacts potentially induced by a regeneration solution, single cycle kinetic (SCK) experiments were performed. This experimental approach confirmed the binding of AnxA2 to NS5B (Fig. 1b). To eliminate the possibility that the binding to NS5B was unspecific, the interaction was also analyzed using a reference flow cell where NS5B was first immobilized and subsequently inactivated (denatured) by multiple injections of 6 M guanidine-HCl (Fig. 1c). Successful inactivation of NS5B in the reference flow cell was confirmed by showing that the HCV polymerase inhibitor filibuvir did not interact with the denatured protein in the reference cell. Similar SPR sensorgrams of the interaction of filibuvir to immobilized NS5B was achieved for the experiment with an inactivated NS5B immobilized on the reference cell (Fig. 2a) compared to the experiment with a blank reference cell. The interaction seen for AnxA2 in the assay where unspecific binding was subtracted, confirmed a strong interaction between NS5B and AnxA2.

The interaction between AnxA2 and NS5B may be influenced by the fact that both are viral RNA-binding proteins. For example, it is possible that AnxA2 bound to host mRNA is not able to bind NS5B or that AnxA2 bound to NS5B is not able to bind RNA, indicating an overlap of the binding sites. In addition, if mAnxA2 interacts with NS5B, this would exclude helices C-D as the site of interaction between the two proteins. The non-RNA binding engineered AnxA2 variant, mAnxA2, was therefore also tested for its ability to interact with NS5B (Fig. 1d). Although the experiment confirmed a strong interaction between NS5B and immobilized mAnxA2, it was weaker than that observed between NS5B and wild type (wt) AnxA2 and dissociated more rapidly (Fig. 1a). The non-ideal shape of the sensorgrams indicated that the interaction was not stable, potentially due to a lower stability of the mAnxA2 surface. It should be noted that the far-UV CD spectrum of mAnxA2, when compared to wt AnxA2, indicated that the overall helical secondary structure was conserved, while the thermal denaturation profiles indicated small structural perturbations in mAnxA2 due to the mutation of six exposed amino acids that could explain the differences in interaction with NS5B [25]. The lower amount (1041 RU) of mAnxA2 immobilized to the chip compared to AnxA2 (3595 RU) was attributed to the immobilization procedure and the lower content of lysine residues in mAnxA2, having 4 out of 32 lysine residues replaced by serine (i.e. K308S, K309S and K310S, K313S). However, the lower surface density and stability have no practical consequence for the interpretation of the data.

Non-linear regression analysis of the sensorgrams for the interaction between NS5B and AnxA2 showed that it could not be well described by any simple mechanistic model, preventing the determination of kinetic parameters. Instead, an approximate steady state analysis of the SPR sensorgrams was performed in order to quantify the affinity for the interaction (Fig. 1e-h). The apparent dissociation constants (K_D ^app) (Table 1) are thus approximations of the affinity.

The mean K_D ^app for the interaction between AnxA2 and NS5B, based on two independent multicycle experiment where AnxA2 was immobilized and NS5B was run as analyte, was 30 nM. In the reverse experiment, the concentration of AnxA2 was too low relative to the K_D ^app for the affinity to be estimated.

To investigate if AnxA2 also interacts directly with the HCV NS3 protease, NS3 or the NS3/NS4A complex were run as analytes over sensor surfaces with immobilized AnxA2. However, no interactions were detected with either the NS3 protein at concentrations up to 270 nM or with the NS3/NS4A complex up to 500 nM (Additional file 1: Figure S1).

As a control, the effect of AnxA2 on the activity of NS3 protease was tested using a FRET-based enzyme activity assay, although no direct interaction was detected between NS3 and AnxA2 using the SPR assay. No effect on the enzymatic activity was detected when AnxA2 was present in a 5 to 1 M ratio to NS3 (Additional file 1: Figure S2, Table S1). The two different experimental approaches both indicate that AnxA2 does not interact directly with the NS3 protein of HCV.

To obtain an understanding of the functional implications of the interaction between AnxA2 and NS5B, we investigated the effect of allosteric NS5B inhibitors on the interaction.

When injecting AnxA2 over an NS5B surface, alone or in the presence of 500 nM filibuvir it was evident that filibuvir had little, if any, effect on the interaction (Fig. 2c). However, when filibuvir was injected over an NS5B-AnxA2 surface, it resulted in a significantly lower binding response (Fig. 2b), compared to when injected over an NS5B surface (Fig. 2a). When injecting NS5B concomitantly with filibuvir over immobilized AnxA2, a significantly reduced binding of NS5B to AnxA2 was also observed (Fig. 2d). This was not seen when performing the corresponding experiment with the polymerase inhibitor dasabuvir (Fig. 2d). These results show that the binding of filibuvir to NS5B impairs the interaction between NS5B and AnxA2. Reciprocally, when NS5B has formed a complex with AnxA2, its ability to interact with filibuvir is impaired (Table 2). However, the binding of AnxA2 to NS5B is not affected when NS5B is subjected to AnxA2 and filibuvir simultaneously, possibly attributable to a kinetic effect.

The potential effect of AnxA2 on the polymerase activity of NS5B was investigated using an SPR-de novo polymerase assay, in which a 31-mer RNA, designed to be used for de novo polymerase assay [22] was captured on a partly complementary ssDNA immobilized on a sensor chip, and NS5B and rNTPs were injected as analytes in the absence or presence of AnxA2 (Fig. 3a).

The assay monitors the formation of the complementary RNA chain or nucleotides incorporation in real time. Reference experiments included the control experiment in which NS5B was injected alone, i.e. without rNTPs or AnxA2, or in which AnxA2 and mAnxA2 were injected without NS5B.

NS5B interacted with the 31-mer RNA (Fig. 3b-d). In contrast, neither AnxA2 nor mAnxA2 were observed to interact with the 31-mer RNA at concentrations up 1000 nM or 200 nM, respectively (Fig. 3d). The activity of NS5B was not affected by the presence of AnxA2 at equimolar amounts (Fig. 3b). However, when AnxA2 was added at a 2:1 M ratio to NS5B, the nucleotide incorporation rate was significantly reduced (Fig. 3c). mAnxA2 reduced the polymerase activity of NS5B more efficiently than AnxA2 (Fig. 3b, c).

It was unexpected that AnxA2 did not interact with the 31-mer RNA captured by ssDNA (Fig. 3d). To confirm that the AnxA2 used in our experiments was functional and could bind RNA, a control experiment was performed. As it has been shown that AnxA2 has an intrinsic poly(G)-binding activity [25], an 8-mer RNA (5’GGG GAU UG-3′) containing a poly(G) sequence was injected as analyte over immobilized AnxA2 (Fig. 4a). It was confirmed that AnxA2 interacted with this RNA (Fig. 4a), although it was not able to bind to the immobilized 31-mer RNA (Fig. 3d). The K_D was found to be in the low nanomolar range, as determined by steady state affinity analysis and by global kinetic fit using a 1:1 model (Fig. 4c-d, Table 3). As a control, it was confirmed that the 8-mer RNA did not interact with the mAnxA2 surface (Fig. 4b).

To visualize the interplay between AnxA2, the 8-mer RNA and NS5B, the formation of a ternary complex between the three entities was explored using surfaces with immobilized AnxA2 and then injecting either 8-mer RNA or NS5B to form a stable binary complex, followed by injection of the third binding partner, thus potentially forming a ternary complex. Eight-mer RNA (50 nM) was first injected over AnxA2, providing a reference point for RNA-binding to AnxA2 (Fig. 5a). After regeneration of the surface, NS5B (50 nM) was injected over AnxA2, resulting in a stable AnxA2-NS5B complex, as indicated by a stable new baseline after injection (Fig. 5a). The injection of the 8-mer RNA over the AnxA2-NS5B surface resulted in the same signal as when 8-mer RNA was injected over AnxA2 alone (Fig. 5a). This suggests that 8-mer RNA binds to the surface via AnxA2, not NS5B (Fig. 5b).

To confirm this interpretation, the opposite experiment was performed, i.e. NS5B (50 nM) was injected over AnxA2 alone to get a reference value. Eight-mer RNA (100 nM) was then injected over the regenerated AnxA2 surface, creating a stable AnxA2-RNA complex (Fig. 5c). NS5B was subsequently injected, resulting in NS5B binding to the AnxA2-RNA complex. This injection of NS5B over the AnxA2-RNA complex only resulted in a slightly higher signal than for NS5B binding to AnxA2 alone (Fig. 5c). This confirms the formation of an RNA-AnxA2-NS5B complex (Fig. 5d).

Finally, injection of the 8-mer RNA and NS5B concomitantly (in ratio 1:1 or 2:1) over AnxA2 was compared to the response observed when injecting NS5B and the 8-mer RNA separately (data not shown). The response seen for the NS5B-RNA complex was identical to that resulting from the injection of NS5B alone, suggesting that the binding of the NS5B complex to the 8-mer RNA is disrupted upon NS5B binding to AnxA2.