Zika Virus Methyltransferase: Structure and Functions for Drug Design Perspectives

The ZIKV genome encodes the bifunctional NS5 protein, with an ∼264-amino-acid (aa) MTase domain at the N terminus followed by an RNA-dependent RNA polymerase (RdRp) domain (Fig. 1A). The MTase domain contains the canonical 2′-O-MTase (K-D-K-E) catalytic tetrad and possesses ∼65% and 70% similarity to those of the four DENV serotypes and WNV, respectively. In this study, different constructs were designed. For functional assays, a construct, aa 4–278, encompassing the putative MTase domain from aa 4 to aa 278 of NS5, and full-length NS5 (aa 4 to 903) were produced and purified. From NS5 preparations, both the full-length protein and the MTase domain resulting from cleavage by bacterial proteases (here called NS5-MTase_C) were recovered after size exclusion chromatography (SEC). The purified proteins were detected as single bands at their expected molecular masses (∼30 to 35 kDa and ∼100 kDa, respectively) (Fig. 1B) upon SDS-PAGE analysis. The apparent molecular mass observed for NS5-MTase_C is similar to those of the ZIKV and DENV MTase domains (Fig. 1B).

As the original MTase construct (aa 4–278) did not crystallize, we generated another that includes the complete N terminus but is devoid of the linker between the MTase and RdRp domains (aa 1–264; PDB code 5M5B ). This construct allowed determination of the crystal structure of the ZIKV MTase domain. The structure was solved by molecular replacement using the DENV3 NS5 MTase/SAM (PDB code 3P97 ) as the template model and refined to a resolution of 2.01 Å, with R_work and R_free values of 16.0% and 19.1%, respectively. The crystal belongs to space group P1 with the following unit cell parameters: a = 37.6 Å, b = 64.1 Å, and c = 72.0 Å and α = 113.1°, β = 97.8°, and γ = 92.0°. Data collection and refinement statistics are reported in Table 1. The ZIKV MTase domain crystallized with two molecules in the asymmetric unit (Fig. 2A), whereas size exclusion chromatography indicates that the protein behaves as a monomer in solution. The low-interaction surface area (about 500 Ų) suggests that dimerization occurs during crystallization as previously observed for Modoc virus (MODV) MTase (32). The two protomers (A and B chains) are similarly folded, with significant differences arising from disorder in residues 43 to 58 of chain B (Fig. 2A). A superimposition of 244 Cα atoms in the A and B chains of the protein resulted in an RMSD of 0.52 Å. Chain A consists of a canonical MTase core (residues 61 to 229, construct numbering PDB code 5M5B ) folded into a seven-stranded β-sheet (β1 to β7) surrounded by four α-helices (αX, αA, αD, and αE), as shown in Fig. 2A and C. Appended to the core are an N-terminal extension (residues 8 to 60) and a C-terminal extension (residues 230 to 272). The N-terminal subdomain comprises a helix-turn-helix motif followed by a β-strand and an α-helix (A1, A2, B1, and A3). The C-terminal subdomain consists of an α-helix followed by a β-strand (A4 and B2). Inspection of the electron density maps reveals strong additional electron density in the SAM binding pocket, allowing the unambiguous modeling of SAM. Because no SAM is added during purification or crystallization, its presence in the structure must have originated from E. coli. The SAM molecule is bound in the central cleft formed of β-strands β1, β2, and β4 and α-helices αX and αA (Fig. 2A and B). The SAM adenine base is accommodated within a hydrophobic pocket defined by the side chains of Val138, Phe139, and Ile153 and stabilized by hydrogen bonds with a carboxylic oxygen from Asp137 and the nitrogen main-chain atoms of Lys111 and Val138 (Fig. 2B, left panel). The ribose moiety is hydrogen bonded to the Glu117 carboxyl group. The sugar is also bridged via a water molecule to the side chain of Glu117. The methionine tail of the SAM molecule is positioned by interactions with the side chains of Ser62, Trp93, and Asp152, as well as with the main chain of Gly92. The methionine carboxylate is also bridged via a single water molecule to the side chain of Glu117 and to the backbone of Arg90. Three molecules of sulfate interact with chain A. One of them is bound via a cluster formed by arginines Arg43, Arg47, Arg63, and Arg90 (Fig. 2D). The structure of chain A is highly similar to those of DENV-3 MTase (RMSD of 0.79 Å with PDB code 3P97 chain A) and of ZIKV MTase recently determined by Coloma et al. (RMSD of 0.29 Å with PDB code 5KQR chain A) (22). There is no difference in SAM binding with the latter structure (PDB code 5KQR ), and the sulfate bound via a cluster formed by arginines Arg43, Arg47, Arg63, and Arg90 overlaps perfectly with the phosphate described by Coloma et al. (22). Structural differences from DENV-3 MTase are limited to small changes in the conformation of three solvent-exposed loops (residues 55 to 58, 178 to 183, and 251 to 254), as shown by superimposition of several structures (Fig. 2C).

We observe significant conformational changes between chain A and chain B of ZIKV MTase (RMSD of 0.52 Å) likely favored by their different crystal environments. Indeed, the chain B environment does not allow this loop to adopt the conformation observed in chain A. However, it is unlikely that crystal packing is the only factor involved since similar disorder has been observed, under different crystal packing, for the MODV MTase (PDB code 2WA2 ) (32). We conclude that the intrinsic flexibility of this loop, at least in the absence of the RNA substrate, has also contributed to building of this crystal form. As illustrated in Fig. 2A and D, three main differences are noteworthy in chain B, as follows. (i) There is no supporting electron density for residues 43 to 58 (part of the B1 β-strand, the A3 α-helix, and the following loop (Fig. 2A]). A similar absence of electron density has already been described for one of the two chains of the MODV MTase (32), suggesting a certain degree of flexibility in this region. (ii) Downstream of the aa 43–58 flexible region, the αX α-helix is kinked (Fig. 2D), with its N-terminal part (residues 61 to 67) taken away from this helix axis. (iii) There is no additional electron density characterizing the methyl donor SAM in the SAM binding site. Interestingly, in chain B of the MODV MTase, a SAM molecule is present, even if the homologous region aa 35–56 appears to also be disordered. However, the kink observed in the αX α-helix of ZIKV MTase chain B is absent from MODV MTase chain B. The kink has two main structural consequences. First, the side chain of Ser62 cannot interact anymore with the methionine tail of the SAM, as observed in chain A (Fig. 2B, left panel). Second, a cluster formed by arginines of the RNA binding groove (Arg43, Arg47, Arg63, and Arg90) in chain A is distorted in chain B and does not accommodate any sulfate molecule (Fig. 2D). In the absence of SAM, two polar molecules, a sulfate and a glycerol close by (∼3.2 Å), are held in the SAM binding pocket of chain B (Fig. 2A and B). The sulfate molecule is positioned by hydrogen bonds with the side chain of His116, the main chain of Glu117, and with the bound glycerol nearby. It is also bridged via a single water molecule to the main chains of Gly112 and Glu117, as well as to the side-chain hydroxyl of Thr110. Moreover, the glycerol molecule participates in van der Waals interactions with the main chains of Gly87 and Lys111 and is stabilized by hydrogen bonds with the sulfate and the main-chain amide of Lys111. The glycerol is also bridged via two water molecules to the main chains of Asp137 and Val138 for the first one and to the main chains of Gly87 and Asp152 for the second one. Such a structure with compounds bound in the SAM binding site could serve as a basis for structure-based drug design.

The 2′-O-MTase activity was measured comparatively using the three ZIKV proteins (Fig. 1B). Proteins were incubated with a short RNA substrate bearing an unmethylated cap structure, GpppAC₄, in the presence of a radiolabeled methyl donor ([³H]SAM). The amount of [³H]CH₃ transferred onto RNA was quantified with a DEAE filter binding assay. ZIKV NS5, NS5-MTase_C, and the ZIKV NS5-MTase domain exhibit similar enzymatic activities (Fig. 3A). The level of detected enzymatic activities is about 30% lower than the control activity of DENV NS5-MTase. ZIKV NS5-MTase activity is thus not stimulated by the presence of the RdRp domain in cis. In addition, we compared the enzymatic activity of the ZIKV MTase domain using the unmethylated substrate GpppAC₄ and a substrate methylated at the N-7 position, ^mGpppAC₄. The methyl transfer efficiency is similar (Fig. 3B), suggesting that both substrates can be methylated at the 2′-O position of the adenosine ribose, as already reported for DENV and WNV MTases (19, 33). Additionally, the ZIKV MTase domain shows an optimal activity between pH 8 and 9 (Fig. 3C). Time course experiments indicate that the 2′-O-MTase functions without additional magnesium (not shown). We next addressed sequence specificity using short RNAs consisting of the 5′ ends of various viral RNAs starting either with A or G (Fig. 3D). Although RNA methylation efficiency by ZIKV MTase varies, we could not observe any obvious preference for Flavivirus GpppA-RNAs or GpppG-RNAs.

An SLA hairpin structure is present at the 5′ end of Flavivirus genomes. This structure is required for the detection of the N-7-MTase activity of WNV and DENV (7, 20). First, we compared the 5′ ends of ZIKV and DENV. The seven first residues of the ZIKV RNA are strictly identical to those of DENV, and ZIKV RNA forms a hairpin structure similar to that of DENV. N-7-methylation assays driven by ZIKV NS5-MTase were thus conducted on SLA-containing DENV RNAs with various 5′-capped ends (GpppA-RNA₇₅, ^mGpppA-RNA₇₅, GpppA_m-RNA₇₅, and ^mGpppA_m-RNA₇₅). ^mGpppA-DENV₇₅ and GpppA_m-DENV₇₅ are both significantly better methylated than ^mGpppA_m-DENV₇₅ (P < 0.001), indicating that N-7- and 2′-O-methylations occur (Fig. 3E). In addition, the high methylation level observed with ^mGpppA-DENV₇₅ suggests that the 2′-O-methylation is more efficient than the N-7-methylation under the applied experimental conditions. Interestingly, MTase activity using RNA already methylated at the N-7- and 2′-O positions (^mGpppA_m-DENV₇₅) was also detected. This observation suggests that additional methylations downstream of the cap structure might occur (described below). Altogether, these results show that, as observed for other flaviviruses, the ZIKV MTase can perform cap 2′-O-methylation on short RNAs without sequence specificity, whereas cap N-7-methylation does indeed require longer RNAs containing the Flavivirus conserved SLA structure.

Since the ZIKV MTase is active on double-methylated cap-1 RNAs (Fig. 3E), the methyl transfer on internal residues of RNAs was further investigated using uncapped RNAs in the presence of a 2-times-higher ZIKV MTase concentration than in the previous experiments. Figure 3F indicates that a 13-mer RNA bearing a 5′-triphosphate corresponding to the nascent 5′ end of DENV RNA can be methylated, confirming that ZIKV MTase methylates uncapped RNA putatively on internal positions. In order to further characterize this activity, homopolymeric RNAs bearing a 5′-OH group were tested (Fig. 3F). ZIKV NS5-MTase can methylate a 5′-OH-A₂₇ RNA, but not 5′-OH-U₂₇, -G₂₇, and -C₂₇ RNAs, indicating that ZIKV MTase methylates adenosine residues. Since ZIKV harbors a 2′-O-MTase catalytic tetrad, we inferred that the MTase activity targets the 2′-O of the ribose of the adenosine residues. This activity was confirmed by demonstrating that an RNA previously methylated at all 2′-O positions (A_m27), is not methylated by the ZIKV MTase (Fig. 3G). In addition, the internal adenosine 2′-O-methylation does not specifically occur in the vicinity of the RNA 5′ end since RNAs already methylated on the first (A_mA₂₆) or the two first (A_mA_mA₂₅) adenosines were methylated (Fig. 3G).

The 5′ ends of Flavivirus mRNAs are organized in an SLA structure, and many adenosines are engaged in double-strand stretches. We thus investigated whether the internal methylation could be selective for single-strand RNA (ssRNA) or double-strand RNA (dsRNA). We compared the MTase activities on A₂₇ and on A₂₇ previously annealed with a complementary U₂₇ RNA. No MTase activity was detected using dsRNA (A₂₇ + U₂₇) (Fig. 3H) confirming that ZIKV MTase methylates preferentially ssRNA. In conclusion, the ZIKV MTase preferentially methylates the 2′-O position of internal adenosine residues that are not involved in dsRNA structures.

Our structural and biochemical analyses indicate that the ZIKV MTase shares common structural and functional characteristics with DENV MTase (RMSD of 0.79 Å for PDB code 3P97 chain A). We thus tested whether some compounds already documented to inhibit DENV MTase might also be active against the ZIKV MTase. SAM analogues and cap analogues were tested as well as allosteric inhibitors designed from a fragment-based screening campaign performed against DENV MTase (31). For this purpose, both internal methylation and cap 2′-O-methylation were monitored on GpppAC₄ and A₂₇ RNA substrates, respectively, with increasing concentrations of the investigated inhibitors. We found that SAM analogues such as SAH and sinefungin inhibit cap-dependent ZIKV 2′-O-MTase activity with 50% inhibitory concentrations (IC₅₀) in the low micromolar range (Fig. 4A; Table 2). These compounds also abrogate internal methylation, but SAH shows a lower efficacy against internal methylation than sinefungin, which inhibits both methylations in a similar manner. Cap analogues (GpppA, ^mGpppA, GpppG, and ^mGpppG) bind to the DENV cap binding site by stacking interactions (34) and might therefore block RNA cap recognition, thus limiting cap methylation. This might also interfere with recognition and binding of uncapped RNA for internal methylation since the cap binding site might extend the RNA binding zone. GpppG and ^mGpppG show more potent inhibitions than GpppA and ^mGpppA on cap 2′-O-methylation, with IC₅₀ of 72 μM and 184 μM, respectively, versus 491 μM and 293 μM (Fig. 4B and C; Table 2). In addition, cap analogues inhibit internal methylations, to a lesser extent though, suggesting that the MTase cap binding site forms a part of the RNA binding zone but is not essential for internal methylation of an A₂₇ RNA.

Finally, we tested whether six compounds (1 to 6 in Table 2) targeting an allosteric site of DENV MTase (30, 31) could inhibit ZIKV MTase. All of the compounds selected on DENV MTase inhibit ZIKV cap 2′-O-MTase activity (Fig. 4D and Table 2). Compounds 1 to 6 exhibit lower IC₅₀ for ZIKV 2′-O-MTase activity than those for DENV MTase, suggesting that they are more potent on ZIKV MTase. In addition, all compounds inhibiting the cap-2′-O-MTase activity also inhibit the internal methylation, but with higher IC₅₀ (Table 2), indicating that internal and cap 2′-O-MTase activity use the same catalytic mechanism. These data also indicate that the structural and functional similarities between ZIKV and DENV MTase pave the way for the development of pan-viral inhibitors targeting the members of the Flavivirus genus.

For the functional study, the coding sequence of ZIKV NS5 (aa 4–903) and the ZIKV NS5 MTase (aa 4–278) domain were synthesized (Genscript) based on the sequence of the ZIKV strain H/PF/2013 (GenBank accession no. KJ776791.2 ) and then cloned into a pQE30 (Qiagen) plasmid with an N-terminal His₆ tag. For the structural study, the coding sequence of ZIKV NS5 MTase (aa 1–264) was amplified by PCR starting from the synthetic gene and subcloned in pDEST14 (Thermo Fisher Scientific) with an N-terminal His₆ tag.

The MTase domains (aa 4–278 and aa 1–264) were produced in Escherichia coli T7 Express Iq (New England BioLabs). Cells were grown in Terrific Broth at 37°C until the optical density at 600 nm (OD₆₀₀) reached 0.6. Protein expression was then induced by 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at 17°C overnight. Bacteria were harvested by centrifugation. The bacterial pellets from a 2-liter bacterial culture were resuspended in 100 ml lysis buffer (50 mM Tris-HCl [pH 8], 300 mM NaCl, 5% glycerol, 0.1% Triton, 10 μg/ml DNase I, 2 tablets of EDTA-free antiprotease cocktail [Roche], 0.25 mg/ml lysozyme). After 30 min of incubation at 4°C, the cells were sonicated and clarified by centrifugation prior to immobilized metal affinity chromatography (IMAC) purification on a 5-ml His prep column (GE Healthcare), with elution in 50 mM Tris-HCl, 300 mM NaCl, and 250 mM imidazole (pH 8.0). The eluted protein was then loaded on a 16/60 Superdex 200 (GE Healthcare) equilibrated in a mixture of 10 mM HEPES, 500 mM NaCl, glycerol 5%, and 1 mM dithiothreitol (DTT [pH 7.5]).

The protocol for the production and purification of the ZIKV NS5 was adapted from the protocol developed for DENV NS5 (40). After the IMAC purification, a size exclusion chromatography (SEC) step was applied to separate the cleaved MTase and the full-length NS5 using a Superdex S75 HR 16/20 column (GE Healthcare) preequilibrated in a mixture of 50 mM HEPES (pH 7.5), 750 mM NaCl, 10% glycerol, and 10 mM DTT. Proteins were concentrated and stored at −20°C after adding glycerol to a final concentration of 40%.

The enzymatic assays were carried out in 40 mM Tris-HCl (pH 8.0), 1 mM DTT, 2 μM SAM, and 0.33 μM [³H]SAM (PerkinElmer) in the presence of 0.7 μM synthetic RNAs with various 5′-end modifications (triphosphorylated, pppRNA; unmethylated cap, GpppARN; cap-0, ^mGpppARN, or 2′-O-methylated cap, GpppG_mRNA). Purified ZIKV NS5 and NS5-MTase were added to final concentrations of 0.75 μM, except for the internal methylation assay, which was performed in the presence of 1.5 μM MTase.

Reaction mixtures were incubated at 30°C and stopped after 30 min except for the time course experiments by a 10-fold dilution of the reaction mixture in 20 μM ice-cold SAH. Samples were then transferred to DEAE cellulose filters (PerkinElmer) by using a Filtermat Harvester apparatus (Packard Instruments). The unincorporated [³H]SAM was removed from the filter by several washing steps with 10 mM ammonium formate (pH 8.0), with H₂O, and with absolute ethanol, before drying the DEAE filters. The filters transferred into plastic bags, BetaplateScint (Wallac) scintillation fluid was added before quantification of [³H]methyl groups transferred onto RNA substrates in counts per minute using a Wallac 1450 MicroBetaTriLux liquid scintillation counter.

For the MTase inhibition assays, 0.75 μM ZIKV NS5-MTase was mixed with the inhibitor candidate before the addition of SAM and RNA substrate to start the reaction. The final concentration of dimethyl sulfoxide (DMSO) in the reaction mixtures was below 5%, and control reactions were performed in the presence of similar DMSO concentrations. Reaction mixtures were incubated at 30°C for 30 min and analyzed by filter binding assay as described above.

The IC₅₀s of SAH, sinefungin, cap analogues, and compounds 1 to 6 were determined with GraphPad Prism using the log (inhibitor) versus response variable slope equation.

Various short RNA sequences were chemically synthesized on a solid support using an ABI 394 synthesizer. After RNA elongation with 2′-O-pivaloyloxymethyl phosphoramidite monomers (42) (Chemgenes, USA), the 5′-hydroxyl group of RNA still anchored to the solid support was phosphorylated, and the resulting H-phosphonate derivative was oxidized and activated into a phosphoroimidazolidate derivative to react with either pyrophosphate to give a 5′-triphosphate RNA (pppRNA) (43) or GDP to give GpppRNA (44). After deprotection and release from the solid support, the GpppRNA sequences were purified by ion-exchange–high-performance liquid chromatography (IEX-HPLC) and validated to be >95% pure by IEX-HPLC analysis, and they were unambiguously characterized by matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry. Subsequent N-7-methylation of the purified GpppRNAs was performed enzymatically using N-7-hMTase (44).

The ZIKV MTase (aa 1–264) was concentrated to 5 mg/ml for crystallization screenings in sitting drops. After 2 weeks, crystals were obtained and optimized in 150 mM sodium citrate (pH 5.6) and 1.5 M ammonium sulfate. Crystals were transferred to a cryoprotectant solution of the same reservoir solution containing 10% glycerol for a few seconds and flash-cooled in liquid nitrogen.

Diffraction data were collected at the SOLEIL synchrotron (Proxima 1). We used AutoPROC (45) for data processing. Images were processed and scaled using XDS (46) and Aimless (47). Phaser (48) was used for molecular replacement. The structure of DENV3 NS5-MTase/AdoMet (PDB code 3P97 ) was used as the template model. The first ZIKV MTase model was built using ARP/wARP (49), and autoBUSTER (50) was used for structure refinement. Iterative cycles of model building were done with the program COOT (51). The quality of the refined structures was assessed with MOLPROBITY (52). The crystal structure was determined at a resolution of 2.01 Å and belonged to space group P1 with the following unit cell parameters: a = 37.6 Å, b = 64.1 Å, and c = 72.0 Å and α = 113.1°, β = 97.8°, and γ = 92.0°. Gesamt (53) was used to superimpose the obtained model with DENV3 MTase (PDB code 3P97 ), ZIKV MTase (PDB code 5KQR ), and MODV MTase (PDB code no. 2WA1 ) and to determine RMSD values. Figures were generated with Chimera.

The crystal structure of the Zika virus MTase construct generated in this study was deposited in the Protein Data Bank (PDB) under code 5M5B .