Mechanism of molnupiravir-induced SARS-CoV-2 mutagenesis

1Coronaviruses use an RNA-dependent RNA polymerase (RdRp) for the replication and transcription of their RNA genome–. RdRp is an important target for the development of antiviral drugs against coronaviruses,–. Structures of RdRp have been reported for SARS-CoV-1 and SARS-CoV-2– and provide insights into the mechanisms of RNA-dependent RNA synthesis. The structures also enable mechanistic studies that can rationalize the molecular processes underlying the antiviral activity of compounds targeting RdRp.

2Antiviral drugs often target viral polymerases and function as nucleoside analogs that terminate RNA chain elongation. However, such chain-terminating antivirals are generally not effective against SARS-CoV-2 because coronaviruses carry an exonucleolytic proofreading activity that can remove misincorporated nucleotides from the nascent RNA 3′ end–. The nucleoside analog remdesivir can circumvent proofreading because its incorporation does not terminate elongation but only stalls RdRp after the addition of three more nucleotides,–. Remdesivir was the first FDA-approved drug for the treatment of patients with COVID-19–, but its effectiveness is disputed, emphasizing the need to develop new antiviral drugs.

3Another promising drug candidate for the treatment of patients with COVID-19 is molnupiravir (or EIDD-2801), which also targets the RdRp of SARS-CoV-2. Molnupiravir is an isopropylester prodrug of the nucleoside analog β-d-N4-hydroxycytidine (NHC or EIDD-1931),. Molnupiravir interferes with the replication of various viruses–, including SARS-CoV-2,. It inhibits SARS-CoV-2 replication in human lung tissue, blocks SARS-CoV-2 transmission in ferrets and reduces SARS-CoV-2 RNA in patients. In contrast to approved drugs such as remdesivir that are administered by infusion, molnupiravir is orally available. Molnupiravir has been tested in phase I trials for safety, tolerability and pharmacokinetics, and phase II/III studies are currently ongoing (NCT04405739, NCT04405570 and NCT04575597). Available data suggest that molnupiravir acts as a mutagenizing agent that causes an ‘error catastrophe’ during viral replication,,. Indeed, NHC can introduce mutations into viral RNA, as shown for Venezuelan equine encephalitis virus. Also, the sequencing of influenza virus populations has indicated that NHC causes G-to-A and C-to-U transitions in viral RNA, and the same transitions have been found for SARS-CoV-2.

4Despite this progress, a systematic biochemical and structural analysis of molnupiravir- or NHC-induced RNA mutagenesis by viral RNA polymerases is lacking. In this Article, we quantify the effects of molnupiravir or NHC on the RNA synthesis activity of SARS-CoV-2 RdRp using a purified biochemical system and defined synthetic RNAs. Together with structural analysis, we establish the molecular mechanism of molnupiravir-induced RNA mutagenesis. Our results provide detailed insights into the mechanism of action of molnupiravir, which is entirely distinct from that of remdesivir or chain-terminating nucleoside analogs.

5We first tested whether purified SARS-CoV-2 RdRp can use the active form of molnupiravir, NHC triphosphate (‘MTP’) (Fig. 1a,b), as a substrate for RNA synthesis. We conducted RNA elongation assays in a defined biochemical system using recombinant RdRp and synthetic RNA template–product duplexes (Methods). We used four different RNA duplexes that differed at position +1 of the template strand (Supplementary Table 1), which directs binding of the incoming nucleoside triphosphate (NTP) substrate (Fig. 1c). The RNA product strand contained a fluorescent label at its 5′ end that allowed us to monitor and quantify RNA elongation.

6When nucleotides G or A were present at the RNA template position +1, NHC monophosphate (‘M’) was readily incorporated instead of C or U, respectively (Fig. 1d,e). Time-dependent RNA elongation experiments showed that M was slightly less efficiently incorporated than the cognate nucleotide C (Fig. 1f). M incorporation opposite A was also observed, but was substantially reduced compared to incorporation of the cognate nucleotide U (Fig. 1g). These results can be explained by base pairing of an incoming MTP substrate with either G or A in the RNA template strand. Consistent with this model, NHC adopts different tautomeric forms that have been predicted to allow for base pairing with either G or A.

7We next tested whether the incorporation of NHC monophosphate (M) into nascent RNA interferes with further RNA extension. We first conducted RNA elongation assays with a scaffold that allowed for RNA extension by four nucleotides (nt) (Fig. 2a). We observed that incorporation of M instead of the cognate C or U did not prevent incorporation of three subsequent nucleotides (Fig. 2b). Furthermore, we tested RNA extension with a scaffold that allowed for the incorporation of 11 nt (Fig. 2c). In this case too, the RdRp reached the end of the template when uridine triphosphate (UTP) or cytidine triphosphate (CTP) was replaced by MTP, although, again, incorporation of M instead of U was less efficient than incorporation of M instead of C (Fig. 2d). These results demonstrate that M incorporation into nascent RNA does not prevent further RNA elongation. Thus, longer RNA products containing M nucleotides may be synthesized by the RdRp in the presence of MTP. This posed the question of what happens when M-containing RNA is used as a template in a second step of RNA synthesis.

8To investigate the templating properties of NHC, we prepared an M-containing RNA by solid-phase synthesis using the phosphoramidite building block M-PA, which we synthesized using the convertible nucleoside approach from a ribose-protected O4-chlorophenyluridine (Fig. 3a, Methods, Extended Data Fig. 1 and Supplementary Dataset 1). The presence of M in the obtained RNA as well as RNA purity were confirmed by denaturing HPLC and HR-ESI-MS (Fig. 3b). The M-containing RNA oligo was annealed with a fluorescently labeled product RNA such that the M nucleotide occupied templating position +1 (Fig. 3c and Supplementary Table 1).

9Elongation assays showed that the M residue at the +1 position of the template strand directed incorporation of either G or A into nascent RNA, but not C or U (Fig. 3d,e). This can be explained by the formation of M-GTP or M-ATP base pairs in the RdRp active center. Consistent with this, thermal melting experiments with RNA duplexes containing M-G or M-A base pairs located at terminal or internal positions showed similar RNA duplex stabilities that were slightly lower than for duplexes containing a C-G base pair (Extended Data Fig. 2 and Supplementary Table 2). Thus, when the RdRp uses RNA containing NHC monophosphate as a template, either the correct or the incorrect nucleotide is incorporated into the RNA product, and thus mutagenesis will occur.

10The above data indicate that the key aspect of the mutagenesis mechanism is the formation of stable M-G and M-A base pairs in the RdRp active center. To investigate this, we solved two structures of RdRp–RNA complexes that correspond to mutagenesis products after M-templated incorporation of either G or A (Methods). We formed RdRp–RNA complexes containing M in the template strand and either G or A at the 3′ end of the product strand. This was predicted to result in the formation of nascent M-G or M-A base pairs in position −1, which is occupied after successful M-templated nucleotide incorporation and RdRp translocation. We prepared RNA duplex scaffolds with M-containing oligonucleotides (Extended Data Fig. 3), formed RdRp–RNA scaffold complexes, and subjected these to cryo-EM analysis as described in ref..

11We indeed obtained RdRp–RNA structures that contained either an M-A or an M-G base pair at position −1 (Fig. 4 and Table 1). The structures showed an overall resolution of 3.3 Å and 3.2 Å, respectively, with the active center region resolved at ~2.9 Å in both cases (Extended Data Fig. 4). As expected from the scaffold design, the structures showed the post-translocation state with a free NTP-binding site at position +1 (Fig. 4b,c). Comparison of the two structures with each other and with our original RdRp–RNA structure and with remdesivir-containing RdRp–RNA structures did not reveal major differences, neither in the protein subunits nor in the nucleic acids, except that the protruding, second turn of RNA and the sliding poles of the nsp8 subunits were poorly ordered and not retained in the final model.

12The cryo-EM densities at the −1 position of the structures could readily be interpreted by modeling M-A or M-G base pairs (Fig. 4b,c). The densities were so detailed that we could clearly distinguish G and A bases (Fig. 4d). The densities were also consistent with the proposed base pairing that is enabled by different tautomeric forms of NHC (Fig. 4e). However, the observed hydrogen-bonding geometries were not optimal, possibly explaining our biochemical observations that suggest that M can mimic C well, and mimic U to some extent, but neither mimicry is perfect (Figs. 1 and 2). These results represent the first direct visualization of NHC in a polymerase enzyme and show that stable M-G and M-A base pairs can be formed and accommodated in the RdRp active center, readily explaining our biochemical results.

13Our systematic biochemical analysis suggests a two-step model for the mechanism of molnupiravir-induced coronavirus RNA mutagenesis (Fig. 5). When the molnupiravir prodrug enters the cell, it is converted to NHC triphosphate (MTP), which can be used by the RdRp of SARS-CoV-2 as a substrate instead of CTP or UTP. Therefore, in a first step, the RdRp is predicted to frequently incorporate M instead of C or U when it uses the positive-strand genomic RNA (+gRNA) as a template to synthesize negative-strand genomic (−gRNA) and subgenomic RNA (−sgRNA). In a second step, the resulting M-containing RNA can be used as a template for the synthesis of +gRNA or positive-strand subgenomic mRNA (+sgmRNA). The presence of M in the −gRNA then leads to mutations in the positive-strand RNA products, which do not support formation of intact new viruses, as predicted by the ‘error catastrophe’ model,,.

14Our structural studies confirm the key aspect of this model, namely the formation of M-G and M-A base pairs in the RdRp active center. When we interpreted our cryo-EM densities, we assumed that one stable tautomer predominates in each case; that is, the amino-M tautomer forms a base pair with G and the imino-M tautomer forms a base pair with A. These base pairs do not impair RdRp progression, explaining the observed mutagenic pattern seen in in vivo studies of NHC-exposed coronaviruses. This antiviral mechanism is conceptually similar to the recently suggested mutagenesis mode of action of favipiravir,, but is entirely distinct from that of remdesivir, which impairs RdRp progression. However, like remdesivir, molnupiravir escapes viral RNA proofreading because M incorporation and M-directed misincorporation are apparently not recognized by the viral exonuclease,. Such proofreading escape may also be due to the stability of the M-G and M-A base pairs, which are predicted not to induce or favor backtracking of RdRp, which is probably required for exposing the RNA 3′ end to the proofreading exonuclease,.

15Finally, the two-step model can explain how molnupiravir or NHC monophosphate leads to RNA mutagenesis by polymerases of other viruses. For influenza, a possible two-step mutagenesis mechanism had been inferred from sequencing a molnupiravir-experienced virus population. Also consistent with the model, MTP does not inhibit RNA synthesis by hepatitis C polymerase or the RdRp of respiratory syncytial virus. Furthermore, the reverse transcriptase of human immunodeficiency virus can incorporate G or A opposite the NHC located in the template. The two-step mutagenesis model rests on the base-pairing properties of NHC that we have structurally defined here and can explain why molnupiravir and NHC exhibit broad-spectrum antiviral activity against a wide variety of RNA viruses.

16Beside the high antiviral potency of NHC, potential risks have to be considered. Host RNA polymerases may use MTP as substrate, and indeed the mitochondrial DNA-dependent RNA polymerase can use EIDD-1931 and incorporate NHC monophosphate into RNA in vitro. In addition, possible mutagenic effects of NHC were recently described in mammalian cells. Therefore, it will be important to characterize the effects of molnupiravir and NHC on cellular polymerase function in future studies.

17After the submission of our manuscript, independently derived biochemical data about the molnupiravir mechanism of action became available that are consistent with our findings. On the basis of steady-state kinetic measurements, the authors suggest the formation of M-A and M-G base pairs, a longstanding hypothesis that could now be confirmed by the structural data presented in our study.

18No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

19Preparation of SARS-CoV-2 RdRp, composed of nsp12, nsp7 and two copies of the nsp8 subunits, was carried out as described in ref., with some modifications. Nsp12 protein was expressed from pFastBac vector 438C (Addgene 154759) in Hi5 insect cells. Production of the bacmid, V0 and V1 viruses was carried out as described in ref.. Sixty hours after transfection with the corresponding V1 virus, the cells producing nsp12 with a 6xHis-MBP N-terminal tag were collected by centrifugation (3,000 r.p.m., 10 min at 4 °C) and lysed by sonication in lysis buffer A1 (400 mM NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 30 mM imidazole pH 8.0, 5 mM β-mercaptoethanol, 0.284 μg ml−1 leupeptin, 1.37 μg ml−1 pepstatin, 0.17 mg ml−1 PMSF and 0.33 mg ml−1 benzamidine). The lysate was clarified by centrifugation at 74,766g for 60 min and ultracentrifugation at 100,000g at 4 °C for 60 min, followed by filtration through a 0.45-μm filter (Amicon Ultra centrifugal filter, Merck). The protein was bound to HisTrap HP prepacked columns (GE Healthcare) pre-equilibrated in lysis buffer, washed with high salt buffer B2 (1 M NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 5 mM β-mercaptoethanol) followed by washing with buffer A1, and eluted with a gradient of 0–80% A1-B1 buffers over 30 column volumes (CV) (B1 buffer: 200 mM NaCl, 25 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 400 mM imidazole pH 8.0, 3 mM MgCl2 and 5 mM β-mercaptoethanol). Fractions containing nsp12 were pooled and the tag was cleaved with His-tagged tobacco etch virus (TEV) protease overnight during dialysis against buffer D (200 mM NaCl, 25 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 5 mM β-mercaptoethanol). The protein solution was applied to a HisTrap HP prepacked column to remove uncleaved protein, tag and TEV protease. The flow-through containing nsp12 was further purified by ion exchange chromatography using Hi TRAP Q HP and SP HP prepacked columns (GE Healthcare) equilibrated with buffer D. The unbound protein was concentrated and further purified via size exclusion chromatography on a HiLoad S200 16/60 column (GE Healthcare) in buffer A2 (300 mM NaCl, 20 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 1 mM MgCl2, 1 mM Tris(2-carboxyethyl)phosphine (TCEP)). The peak contained monomeric fractions of nsp12 that were pooled, concentrated to 55 μM, aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

20Nsp7 and nsp8 were prepared as described in ref.. Briefly, both proteins were expressed in Escherichia coli BL21(DE3) RIL from the pET-derived vector 14-B (a gift from S. Gradia; Addgene 48308) in LB medium individually. Cells were grown to an optical density at 600 nm of 0.4 at 30 °C and protein expression was induced with 0.5 mM IPTG at 18 °C for 16 h. After collection, cells were resuspended in lysis buffer A1 (400 mM NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 30 mM imidazole pH 8.0, 5 mM β-mercaptoethanol, 0.284 μg ml−1 leupeptin, 1.37 μg ml−1 pepstatin, 0.17 mg ml−1 PMSF and 0.33 mg ml−1 benzamidine). Nsp8 and nsp7 were purified separately using the same purification procedure. The cells were lysed using a French press and three cycles. Lysates were subsequently cleared by centrifugation (74,766g, 4 °C, 30 min). The supernatant was applied to a HisTrap HP column (GE Healthcare), pre-equilibrated in lysis buffer. The column was washed with high-salt buffer B2 (1 M NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 5 mM β-mercaptoethanol) and with buffer A1 (400 mM NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 30 mM imidazole pH 8.0 and 5 mM β-mercaptoethanol). The protein sample was then eluted using a gradient of 0–80% B1 over 30 CV (B1 buffer: 300 mM NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 400 mM imidazole pH 8.0, 3 mM MgCl2, 5 mM β-mercaptoethanol). The protein sample was dialysed in buffer D (200 mM NaCl, 50 mM Na-HEPES pH 7.4, 10% (vol/vol) glycerol, 5 mM β-mercaptoethanol) in the presence of His-tagged TEV protease at 4 °C. The dialysed sample was subsequently applied to a HisTrap HP column (GE Healthcare), pre-equilibrated in A1 buffer. The flow-through that contained the untagged protein of interest was applied to an ion exchange chromatography HiTrap Q column (GE Healthcare). The unbound protein sample, containing nsp8 or nsp7, was concentrated using a 10,000 molecular weight cutoff Amicon Ultra Centrifugal Filter (Merck) and applied to a HiLoad S200 16/600 system (GE Healthcare) equilibrated in buffer A2 (300 mM NaCl, 20 mM Na-HEPES pH 7.4, 5% (vol/vol) glycerol, 1 mM TCEP). Peak fractions containing proteins were pooled, concentrated to 430 μM (nsp7) and 400 μM (nsp8), aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

21RNA scaffolds were designed according to published SARS-CoV-2 RNA extension assays,. All assays except those shown in Fig. 3c were performed with RNA duplexes composed of short RNA template and product strands. The 5′ end of the RNA product strands contained a 6-carboxyfluorescein (FAM) fluorescent label that allowed us to monitor RNA extension. The 3′ end of the RNA template strands harbored a mismatched dideoxy-C to prevent extension. The RNA extension assay shown in Fig. 3c was performed with a single-stranded looped RNA scaffold that contained a FAM fluorescent label at the 5′ end. For all assays that required synthesis of M-containing oligonucleotides, RNA strands had to be kept short for technical reasons.

22All unmodified and 5′ FAM-labeled RNA oligonucleotides (Supplementary Table 1) were purchased from Integrated DNA Technologies (IDT). MTP was purchased from MedChemExpress and NTPs from Thermo Scientific. The assay was performed as described in ref., except for the following changes. The final concentrations of nsp12, nsp8, nsp7 and RNA were 3 μM, 9 μM, 9 μM and 3 μM, respectively, except that 1 μM of RNA was used in the assays in Figs. 1f,g and 2d. The concentration of NTPs was 37.5 μM except for the assays shown in Figs. 1f,g and 2d, where 4 μM was used. RNA in annealing buffer (50 mM NaCl, 10 mM Na-HEPES pH 7.5) was annealed by heating it to 75 °C for 1 min and gradually cooling to 4 °C. Annealed RNA and pre-mixed RdRp were incubated in reaction buffer (100 mM NaCl, 20 mM Na-HEPES pH 7.5, 5% (vol/vol) glycerol, 10 mM MgCl2, 5 mM β-mercaptoethanol) for 10 min at 30 °C and the reactions were started by addition of NTPs. After 20 min of incubation at 30 °C, the reactions were stopped with 2× stop buffer (7 M urea, 50 mM EDTA, 1× TBE buffer). RNA products were resolved on 20% denaturing polyacrylamide-urea gels in 0.5× TBE running buffer and visualized with a Typhoon 95000 FLA Imager (GE Healthcare Life Sciences). Bands were quantified with Image Lab (Bio-Rad). The obtained data were plotted and analyzed with GraphPad Prism version 9.

235′-O-DMT-2′-O-TOM-O4-chlorophenyluridine was prepared as described in ref. and converted to the corresponding N4-hydroxy-N4-benzoylcytidine 3′-(2-cyanoethyl) diisopropyl phosphoramidite (M-PA) in three steps. Details of the synthetic procedures are provided in the following, and NMR spectra of the isolated compounds are provided in the Supplementary Information. RNA oligonucleotides were then prepared by solid-phase synthesis on a controlled pore glass support (0.6-µmol scale) using 2′-O-TOM-protected ribonucleoside phosphoramidites (70 mM in CH3CN) and ethylthiotetrazol (ETT, 250 mM in CH3CN) as activator, with 4-min coupling time, as previously described,. The oligonucleotides were deprotected with 25% NH4OH/EtOH 3/1 at 55 °C for 6 h, followed by 1 M TBAF in THF for 12 h, and carefully purified by denaturing polyacrylamide gel electrophoresis to remove a minor fraction, in which NHC was converted to U during deprotection.

24The purity and identity of the RNA oligonucleotides were analyzed by anion-exchange HPLC (Dionex DNAPac PA200, 2 × 250 mm, at 60 °C; solvent A: 25 mM Tris-HCl (pH 8.0), 6 M urea; solvent B: 25 mM Tris-HCl (pH 8.0), 6 M urea, 0.5 M NaClO4; gradient: linear, 0–48% solvent B, 4% solvent B per 1 CV) and HR-ESI-MS (Bruker micrOTOF-Q III, negative ion mode, direct injection).

25Thermal melting experiments were performed in 10 mM sodium phosphate buffer pH 7.0, 100 mM NaCl, at an RNA duplex concentration of 20 µM. Absorbance versus temperature profiles were recorded at 260 nm on a Varian Cary 100 spectrometer equipped with a Peltier temperature controller, at a heating rate of 0.5 °C min−1, for two heating and two cooling ramps between 10 and 90 °C. Melting curves were normalized to the absorbance at 95 °C, fitted to a two-state transition model with linearly sloping lower and upper baselines, and the melting temperatures were determined at the inflection point of the curves.

26All reactions were performed under an inert nitrogen atmosphere with dry solvents (CH2Cl2, CH3CN). For workup and purification, distilled solvents (technical quality) were used. Column chromatography was performed on silica gel (Kieselgel 60, Merck) with a particle size of 0.040–0.063. Thin-layer chromatography was performed on Alugram aluminium sheets (Machery-Nagel, UV visualization, 254 nm). NMR spectra were recorded using Bruker Avance III (400 MHz) spectrometers. Chemical shifts (δ) are given in ppm, relative to the residual solvent signals as internal standards (CDCl3: 1H = 7.26, 13C = 77.16). Data are reported as s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad. Coupling constants (J) are given in Hz. HR-ESI-MS results were recorded on a Bruker micrOTOF-Q III spectrometer. The detected mass-to-charge ratio (m/z) is given, as well as the calculated monoisotopic mass (Supplementary Information, Extended Dataset 1 and Extended Data Fig. 1).

27To synthesize compound 1 (5′-O-(4,4′-dimethoxytrityl)-N4-hydroxy-2′-O-(triisopropyl-silyloxymethylcytidine), 5′-O-DMT-2′-O-TOM-O4-chlorophenyluridine (400 mg, 474 µmol, 1.0 equiv.) was dissolved in anhydrous CH3CN (4 ml) under a nitrogen atmosphere. 4-Dimethylaminopyridine (DMAP; 174 mg,1.42 mmol, 3.0 equiv.) and NEt3 (331 µl, 2.37 mmol, 5.0 equiv.) were added, followed by hydroxyl amine hydrochloride (165 mg, 2.37 mmol, 5.0 equiv.). After stirring for 21 h at ambient temperature, the reaction mixture was diluted with CH2Cl2 and washed with saturated aqueous NaHCO3 (2×). The organic phase was dried over Na2SO4 and the solvent was removed under reduced pressure. The crude product was purified by column chromatography (n-hexane:EtOAc + 1% NEt3 1:2 to 2:1) to yield the product (compound 1, 230 mg, 308 µmol, 64%) as a colorless foam.

28To synthesize compound 2 (N4-benzoyl-5′-O-(4,4′-dimethoxytrityl)-N4-hydroxy-2′-O-(tri-isopropylsilyloxy)methylcytidine), a solution of compound 1 (200 mg, 267 µmol, 1.0 equiv.) in anhydrous CH2Cl2 (4 ml) was treated with DMAP (65.3 mg, 535 µmol, 2.0 equiv.) and NEt3 (149 µl, 1.07 mmol, 4.0 equiv.) under a nitrogen atmosphere. Benzoic anhydride (59.9 mg, 265 µmol, 0.99 equiv.) was added in three portions within 3 h and the resulting reaction mixture was stirred for one more hour at ambient temperature. Volatiles were removed under reduced pressure. The crude residue was purified by column chromatography (n-hexane:EtOAc + 1% NEt3 1:2) to yield compound 2 (183 mg, 215 µmol, 80%) as a colorless foam.

29To synthesize compound 3, (N4-benzoyl-5′-O-(4,4′-dimethoxytrityl)-N4-hydroxy-2′-O-(tri-isopropylsilyloxy)methylcytidine 3′-cyanoethyl-N,N-diisopropyl phosphoramidite), compound 2 (150 mg, 176 µmol, 1.0 equiv.) was dissolved in anhydrous CH2Cl2 (2 ml) and cooled to 4 °C. 2-Cyanoethyl N,N,N′,N′-tetraisopropyl phosphoramidite (67.3 µl, 212 µmol, 1.2 equiv.) and 4,5-dicyanoimidazol (23 mg, 194 µmol, 1.1 equiv.) were added in two portions within 1 h. After one additional hour at 4 °C, the reaction mixture was allowed to warm to ambient temperature, then stirred for one more hour. The solvent was evaporated and the crude residue was purified by column chromatography (n-hexane:EtOAc + 1% NEt3 2:1 to 1:1) to yield compound 3 (75.0 mg, 71.3 µmol, 40%) as a colorless foam.

30RNA scaffolds for structural studies were prepared by annealing two RNA oligonucleotides, because the length of the NHC-containing RNA was limited for technical reasons. The first RNA oligo was designed to form a template–product hybrid with a hairpin leaving a 10 nt product overhang at the 3′ end. A second oligo containing 8 nt complementary to the overhang was annealed to obtain an RNA template–product scaffold with a single nick at position −10 in the template RNA (Extended Data Fig. 3 and Table 1). The second RNA oligo had NHC incorporated at position −1 and a short 5′ overhang of three G nucleotides. RNA scaffolds for RdRp–RNA complex formation were prepared by mixing equimolar amounts of two RNA strands in annealing buffer (10 mM Na-HEPES pH 7.4, 50 mM NaCl) and heating to 75 °C, followed by stepwise cooling to 4 °C. RdRp–RNA complexes were formed by mixing purified nsp12 (1.25 nmol) with an equimolar amount of annealed RNA scaffold and threefold molar excess each of nsp8 and nsp7. After 10 min of incubation at 30 °C, the mixture was applied to a Superdex 200 Increase 3.2/300 size exclusion chromatography column (GE Healthcare), equilibrated in complex buffer (20 mM Na-HEPES pH 7.4, 100 mM NaCl, 1 mM MgCl2, 1 mM TCEP) at 4 °C. Peak fractions corresponding to RdRp–RNA complex were pooled and diluted to 1.5 mg ml−1. An additional 0.2 nmol of the annealed RNA scaffold were spiked into each sample before grid preparation. A 3 µl volume of the concentrated RdRp–RNA complex was mixed with 0.5 µl of octyl β-d-glucopyranoside (0.003% final concentration) and applied to freshly glow-discharged R 2/1 holey carbon grids (Quantifoil). The grids were blotted for 7 s with blot force 5 using a Vitrobot Mark IV instrument (Thermo Fischer Scientific) at 4 °C and 95% humidity and plunge-frozen in liquid ethane.

31Cryo-EM data were collected with SerialEM on a Titan Krios transmission electron microscope (Thermo Fischer Scientific) operated at 300 keV. Inelastically scattered electrons were filtered out with a GIF Quantum energy filter (Gatan) using a slit width of 20 eV. Images were acquired using a K3 direct electron detector in counting mode (non-super resolution) at a nominal magnification of ×105,000, resulting in a calibrated pixel size of 0.834 Å px−1. Images were exposed for a total of 2.0 s with a dose rate of 20.5 e− px−1 s−1, resulting in a total dose of 59.6 e−/Å2 that was fractionated into 80 frames. Our previous cryo-EM analysis of SARS-CoV-2 RdRp–RNA complexes showed strong preferred particle orientation in ice. To obtain more particle orientations, all data were collected with 30° stage tilt. Motion correction, CTF estimation and particle picking and extraction were performed using Warp. A total of 11,060 and 10,230 videos were collected for M-A- and M-G-containing structures, respectively.

32For the RdRp–RNA complex containing the M-A base pair, 2.5 million particles were extracted using Warp 1.0.9. Particles were imported to cryoSPARC 2.15 and subjected to two-dimensional (2D) classification. 2D classes representing contamination or broken particles were selected and used for calculating two ab initio structures. All particles were then used for supervised 3D classification against five references, where four were originating from the ab initio reconstruction of contamination and broken particles and one was a previous RdRp–RNA complex structure (EMD-11995). The class containing RNA-bound RdRp yielded ~800,000 particles, which were subjected to homogeneous 3D refinement. The refined particles were then exported to RELION 3.1 and focus-refined in 3D with an initial local angular sampling of 3.7° and a mask around RdRp that omitted the nsp8 sliding poles and the upstream, second RNA turn. To improve the quality of the density in the active site, particles were 3D-classified without image alignment (T = 4, four classes) using the same mask. The best class was focus-refined to an overall resolution of 3.3 Å. Local resolution was estimated with Relion 3.1 using a kernel size of 10 Å. For the RdRp–RNA complex containing the M-G base pair, 2.2 million particles were extracted using Warp 1.0.9. Further processing was as for the first complex, except that the class containing RNA-bound RdRp yielded 850,000 particles and the refinement resulted in a reconstruction at an overall resolution of 3.2 Å.

33Atomic models were built using our previously published SARS-CoV-2 RdRp–RNA complex structure (PDB 7B3D) as starting model. The model was first rigid-body-fitted into the density and then manually adjusted in Coot. The protruding second RNA turn and the nsp8 extensions were removed from the model due to discontinuous cryo-EM density for these regions, indicating mobility. Restraints for molnupiravir monophosphate (M) were generated in phenix.elbow and the structures were refined using phenix.real_space_refine with restraints imposed on secondary structure and the backbone phosphate geometry of M. Model quality was assessed using MolProbity within Phenix, which revealed excellent stereochemistry for both structural models (Table 1). The figures and the Supplementary video were prepared with PyMol (Schrödinger) and ChimeraX.

34Further information on research design is available in the Nature Research Reporting Summary linked to this Article.

35Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41594-021-00651-0.