Structural Analysis of the OC43 Coronavirus 2′-O-RNA Methyltransferase

ABSTRACT

The OC43 coronavirus is a human pathogen that usually causes only the common cold. One of its key enzymes, similar to other coronaviruses, is the 2′-O-RNA methyltransferase (MTase), which is essential for viral RNA stability and expression. Here, we report the crystal structure of the 2′-O-RNA MTase in a complex with the pan-methyltransferase inhibitor sinefungin solved at 2.2-Å resolution. The structure reveals an overall fold consistent with the fold observed in other coronaviral MTases. The major differences are in the conformation of the C terminus of the nsp16 subunit and an additional helix in the N terminus of the nsp10 subunits. The structural analysis also revealed very high conservation of the S-adenosyl methionine (SAM) binding pocket, suggesting that the SAM pocket is a suitable spot for the design of antivirals effective against all human coronaviruses.

IMPORTANCE Some coronaviruses are dangerous pathogens, while some cause only common colds. The reasons are not understood, although the spike proteins probably play an important role. However, to understand the coronaviral biology in sufficient detail, we need to compare the key enzymes from different coronaviruses. We solved the crystal structure of 2′-O-RNA methyltransferase of the OC43 coronavirus, a virus that usually causes mild colds. The structure revealed some differences in the overall fold but also revealed that the SAM binding site is conserved, suggesting that development of antivirals against multiple coronaviruses is feasible.

INTRODUCTION

Prior to the COVID-19 pandemic, only six other human coronaviruses were known: severe acute respiratory syndrome coronavirus (SARS-CoV), Middle East respiratory syndrome coronavirus (MERS-CoV), and four other coronaviruses, OC43 (organ culture 43), 229E, NL63, and HKU1, which are responsible for up to 30% of mild respiratory diseases (1). All of these are members of the subfamily Orthocoronavirinae of the Coronaviridae family. The Orthocoronavirinae subfamily is further divided into four genera: Alpha-, Beta-, Gamma-, and Deltacoronavirus. 229E-CoV and NL63-CoV are members of the genus Alphacoronavirus, while OC43-CoV, HKU1-CoV, SARS-CoV, SARS-CoV-2, and MERS-CoV are all members of the genus Betacoronavirus (2).

The OC43 coronavirus has a 30.7-kbp, positive-sense, single-stranded-RNA (+RNA) genome. That is unusual for a +RNA virus but similar to other coronaviruses (3). It was transmitted to humans recently, in the 19th century, probably from cattle; the bovine CoV is its closet relative. It has also been speculated that OC43-CoV might be responsible for the 1889–1890 pandemic (4), which has usually been attributed to the H2N2 influenza strain (5). Interestingly, while OC43-CoV usually infects the upper respiratory tract and causes respiratory diseases, it is also neurotropic and can be neuroinvasive (6, 7).

The coronaviral genome usually encodes two polyproteins (pp1a and pp1ab), the spike (S), envelope (E), membrane (M), and nucleocapsid (N) proteins, and several accessory proteins. OC43, in addition, encodes hemagglutinin esterase (HE), which dramatically increases its infectivity (8). The pp1a and pp1ab polyproteins are autocatalytically cleaved into 16 nonstructural proteins (nsp1 to nsp16) that ensure many enzymatic activities needed for viral replication, most notably the RNA replication machinery (9). nsp7, nsp8, and nsp12 form the viral RNA-dependent RNA polymerase (RdRp), nsp13 functions as a helicase, and nsp10, nsp14, and nsp16 are RNA methyltransferases (MTases). nsp14 is responsible for proofreading during RNA synthesis; it also has an additional enzymatic activity as an exoribonuclease (10).

In this study, we focused on the 2′-O MTase of OC43-CoV. RNA methylation is important for (viral) RNA stability; it protects the RNA from degradation (innate immunity shielding) and facilitates its efficient translation (11–13). Installation of the 5′ cap in coronavirus-infected cells is a four-step process: (i) the 5′ gamma phosphate of the nascent RNA is hydrolyzed, probably by the coronaviral nsp13 helicase; (ii) GMP is transferred to the 5′ end by an unknown guanylyltransferase; (iii) the nsp10:nsp14 protein complex methylates the N-7 position of the newly attached guanosine; and (iv) the nsp10:nsp16 protein complex methylates the 2′-O of the first nucleotide ribose. The goal is to create viral RNA that is stable in human cells, is efficiently translated, and does not induce the innate immune response. Nonmethylated RNA is recognized by the RIG-I (retinoic acid-inducible gene I) pattern recognition receptor and is also recognized by and binds to interferon (IFN)-induced protein with tetratricopeptide repeats 1 and 5 (IFIT1 and IFIT5), which efficiently prevent its translation (14, 15). These facts imply that inhibiting the nsp10:nsp16 MTase is a valid strategy to fight coronaviral infections.

Here, we present the crystal structure of the OC43-CoV 2′-O MTase (nsp10:nsp16 complex) to understand how different coronaviruses evade innate immunity. The structure revealed relatively high conservation of the architecture of the active site between OC43-CoV and SARS-CoV-2 and suggested that the design of MTase inhibitors targeting multiple human coronaviruses is feasible.

RESULTS

Crystallization of the OC43-CoV nsp10:nsp16 MTase.

We analyzed the sequences of all human coronaviral nsp10 and nsp16 proteins (Fig. 1). The OC43 nsp10 has 53% sequence identity and 68% sequence similarity to the nsp10 of SARS-CoV-2, and similarly, the OC43 nsp16 displays 66% sequence identity and 76% sequence similarity to the nsp16 of SARS-CoV-2, which are not high degrees of identity and similarity for viruses of the same genus. We aimed to solve the crystal structure of the nsp10:nsp16 MTase of OC43 to learn the degree of conservation of the three-dimensional (3-D) structure of an essential enzyme between these two coronaviruses. We used full-length nsp16 and slightly truncated nsp10 (residues 10 to 131) for crystallization trials (Fig. 2A). Both proteins were expressed recombinantly and purified to homogeneity, and the enzymatic activity of the recombinant nsp10:nsp16 protein complex was verified (Fig. 2B). Next, we started the crystallization trials. Eventually, we obtained crystals that diffracted to 2.2 Å. We solved the structure by molecular replacement and refined it to good R factors and geometry (details in Table 1 and Materials and Methods).

FIG 1.

Sequence alignment of all human coronaviral nsp10 and nsp16 proteins. (A) nsp10 and (B) nsp16. Identical residues are highlighted in red fields, while conserved residues are in red font. Important residues involved in coordination of zinc are marked by asterisks below the alignments, and whole zinc-binding regions are in dashed-line boxes, while residues coordinating sinefungin are marked by empty triangles. Secondary-structure features of OC43-CoV are symbolized by helices and arrows for beta sheets.

FIG 2.

The crystal structure of the methyltransferase complex (nsp10:nsp16) from OC43-CoV. (A) Schematic representation of the crystallized and full-length nsp10:nsp16 proteins. (B) SDS-PAGE gel illustrating the purity of the recombinant nsp10:nsp16 complex and a graph illustrating its enzymatic activity. Values are for the reaction performed with all components except the enzyme (1st bar), with all components except for the RNA (2nd bar), and with all components (3rd bar). Error bars show standard deviations. A.U., arbitrary units. (C) Overall crystal structure of the nsp10:nsp16 complex (ribbon) with a sinefungin molecule (white) bound in the active site (SFG) and a “kick-out” omit map (F_o − F_c) of electron density (green) at 3.5 sigma with SFG excluded from the calculation. (D) Topological representation of the secondary-structure features of the complex. (E and F) Zn²⁺ coordination centers in nsp10 in kick-out omit maps (F_o − F_c) of electron density (green) at 10 sigma with Zn²⁺ excluded from the calculation.

TABLE 1.

Crystallographic data collection and refinement statistics

Parametera	Value(s) for OC43 nsp10:nsp16b
PDB accession code	7NH7
Data collection and processing
Space group	P31
Cell dimensions—a, b, c (Å)	61.9, 61.9, 109.7
X-ray source	Home source
Wavelength (Å)	1.5419
Resolution range (Å)	36.57–2.2 (2.28–2.2)
No. of unique reflections	23,828 (2,385)
Completeness (%)	99.74 (99.13)
Multiplicity	5.1 (4.7)
Mean I/σ〈I〉	11.66 (2.48)
CC1/2	0.981 (0.687)
CC*	0.995 (0.902)
Rmerge (%)	13.05 (62.57)
Structure solution and refinement
Rwork (%)	16.41 (21.51)
Rfree (%)	21.01 (29.29)
RMSD bond length (Å)/bond angle (°)	0.005/0.86
Avg B factors (Å2)	28.24
Clashscore	7.84
Ramachandran plot—favored regions/outlier regions (%)	97.32/0

^aCC_1/2, correlation of one half of the observations with the other half; CC*, correlation of the observed data with the unknown “true” intensities.

^bNumbers in parentheses refer to the highest-resolution shell.

The overall fold of OC43-CoV nsp10:nsp16 MTase.

The overall fold is consistent with the fold of previously analyzed coronaviral nsp10:nsp16 MTases from SARS-CoV-2, SARS-CoV, and MERS-CoV (16–19). We could trace the entire protein chain except for the very last five residues of nsp16 and the last residue of nsp10. The nsp16 subunit is composed of 10 β-sheets and 11 helices (Fig. 2). It exhibits a Rossmann fold where the β-sheets are arranged in a central β-motif in the shape of a letter “J” that is surrounded by α-helices (Fig. 2C and D). The small nsp10 subunit is composed only of three small β-sheets and five α-helices (Fig. 2C and D). The nsp10 fold is stabilized by two zinc fingers. The first zinc finger stabilizes the central part of the nsp10 molecule and is formed by Cys74, Cys77, His83, and Cys90. The second zinc finger stabilizes the very C terminus and is formed by Cys166, Cys119, Cys127, and Cys129 (Fig. 2E and F). Notably, all the residues forming both zinc fingers are absolutely conserved among human coronaviruses, highlighting their importance for the function of the nsp10 protein.

S-Adenosyl methionine (SAM) binding pocket.

Prior to crystallization trials, the nsp10:nsp16 2′-O MTase was supplemented with 1 mM sinefungin, an adenosine derivative originally isolated from Streptomyces griseolus by Eli Lilly and Co. as a potential antibiotic (20) that is a pan-methyltransferase inhibitor. The electron density for sinefungin is clearly visible (Fig. 2C). Detailed analysis revealed residues that are responsible for sinefungin binding. Asp114 binds the adenine amino group, while Asp99 forms hydrogen bonds with both ribosyl hydroxyl groups. The amino acid moiety of sinefungin is coordinated by hydrogen bonds to Asp130 and Tyr47, while Asn43 and His69 contribute to ligand binding via water bridges (Fig. 3). All of these residues are absolutely conserved in human coronaviruses (Fig. 1), even Asn43 and His69, which contribute to ligand binding indirectly.

FIG 3.

Detail of sinefungin binding to nsp16. (A) Overall surface representation of the nsp10:nsp16 complex of OC43-CoV. (B) Detailed view of the sinefungin interaction with the active site of the enzyme. The amino acids involved in the interaction are shown as sticks, the waters are shown as red spheres, and the hydrogen bonds are depicted as dashed lines. (C) Structural comparison of OC43-CoV, SARS-CoV, SARS-CoV-2, and MERS-CoV nsp16 sinefungin binding sites.

Comparison to SARS-CoV-2 nsp10:nsp16 MTase.

We compared the structures of OC43 and SARS-CoV-2 nsp10:nsp16 MTases to gain insight into the structural conservation of this enzyme within human coronaviruses. The overall fold remains conserved, with root mean square deviations (RMSD) of 0.99 Ų for the nsp16 and 1.03 Ų for the nsp10 protein (Fig. 4A), as does the sinefungin binding pocket, with the exception of Tyr101, where most human CoVs (SARS-CoV, SARS-CoV-2, MERS-CoV, NL63-CoV, and HKU1-CoV) have an Asn residue instead (Fig. 1 and 3C).

FIG 4.

Structural alignment of the OC43-CoV and SARS-CoV-2 2′-O-RNA MTases. (A) The overall fold of structurally aligned SARS-Cov-2 and OC43-CoV MTases. (B) The N-terminal region of nsp10 of OC43-CoV (green) contains an additional α-helix. (C to F) The C terminus of nsp16 of OC43-CoV (magenta) lacks the final β-sheet of nsp16 SARS-CoV-2 (cyan), and it is driven toward nsp10, where it is a part of the nsp10:nsp16 interface (C), while the SARS-CoV-2 C terminus of nsp16 is involved in crystal contacts just as the SARS-CoV-2 N terminus is (D to F).

However, the structural alignment also revealed significant differences. The C terminus of OC43-CoV nsp16 is located at the nsp10:nsp16 interface and contributes to nsp10:nsp16 binding. While the SARS-CoV-2 nsp16 C terminus is folded away from the interface and actually forms crystal contacts, which the SARS-CoV-2 nsp16 N terminus does as well (Fig. 4D to F). SARS-CoV-2 nsp16 also has a final β-sheet (β-12) that is antiparallel to β-5. Surprisingly, this β-sheet (β-12) is missing on the OC43 nsp16. On the other hand, the OC43 nsp10 has an additional α-helix (residues 11 to 20) that is not present in the SARS-CoV-2 nsp10 (Fig. 4B).

DISCUSSION

Before the appearance of SARS and MERS, the coronaviruses were rather an insignificant group that did not attract that much attention (21), despite some of the most important discoveries, such as proteolytical processing of the spike protein (at that time called the E2 glycoprotein) or that the leader sequence of subgenomic RNAs comes from the 5′ end of the genome, having already been made (22, 23). Upon the appearance of MERS and SARS, the coronaviruses attracted significant scientific interest that led to the molecular description of their life cycle (reviewed in Snijder et al. [9]), including a structural understanding of their key enzymes (19, 24–30). However, the four human coronaviruses that generally do not cause any serious illness (OC43-, 229E-, NL63-, and HKU1-CoVs) are still understudied. Nonetheless, it would be interesting and important to understand why certain coronaviruses cause deadly diseases and others just mild colds. Certainly, the spike protein is the primary determinant of pathogenicity: both the SARS-CoV and the SARS-CoV-2 spike proteins recognize angiotensin-converting enzyme 2 (ACE2) as the receptor (31, 32), while MERS-CoV uses dipeptidyl peptidase 4 (33), OC43- and HKU1-CoV use 9-O-acetylated sialic acids (34), and E229-CoV uses aminopeptidase N (35). However, NL63-CoV also uses ACE2 as its receptor (36, 37) and causes only common colds, suggesting that other determinants of coronaviral perilousness also exist.

Here, we report the structure of the human OC43-CoV 2′-O MTase, a complex of two nonstructural proteins, nsp10 and nsp16. The main function of this complex is to create the cap structure on the 5′ end of viral RNA. The structure revealed that despite significant sequence differences between OC43 and other human coronaviruses, the overall fold is well preserved: the biggest differences are an additional helix (α1) in the nsp10 subunit that is not observed in the SARS-CoV and SARS-CoV-2 coronaviruses and a different conformation of the OC43 nsp16 C terminus. The C terminus is also missing a β-sheet compared to its SARS-CoV-2 counterpart.

The development of inhibitors of coronaviral MTases (both nsp14 and nsp16) has just begun. The compounds discovered so far are potent in vitro; however, most of them suffer from low membrane permeability, hindering their direct testing in antiviral assays (38–40). Several compounds inhibiting the flaviviral MTase (the N-terminal domain of the NS5 protein) are known, and some of them are potent in cell-based antiviral assays; however, none of these inhibitors is approved as a drug (41–44).

Importantly, the structure revealed high conservation of the SAM binding pocket. All residues that make direct contact with the ligand (sinefungin) are absolutely conserved among human coronaviruses, suggesting that mutations of these residues are not tolerated. This is definitely good news for drug design, because compounds that target the SAM pocket will likely be effective against all human coronaviruses and the evolution of resistance against them is unlikely.

MATERIALS AND METHODS

Protein expression and purification.

The genes encoding OC43 nsp10 and nsp16 proteins were codon optimized for Escherichia coli and commercially synthesized (Thermo Fisher Scientific). The genes were cloned in a home-made pSUMO-Kan plasmid (a pRSFD-derived vector already encoding a His_8x-SUMO solubility and purification tag). The proteins were purified using our established protocols for viral enzymes (45, 46). Briefly, E. coli BL21 cells were transformed with nps10 and nsp16 expression vectors and cultivated at 37°C in LB medium with the addition of ZnSO₄ (10 μM) and kanamycin (50 μg/ml). The production of nsp16 and nsp10 proteins was induced by the addition of IPTG (isopropyl-β-d-thiogalactopyranoside) (300 μM) at the late exponential growth phase (optical density [OD] = 0.6 to 0.8), and the temperature was lowered to 18°C for 18 h. The bacterial cells were collected by centrifugation, resuspended in lysis buffer (50 mM Tris, pH 8, 300 mM NaCl, 20 mM imidazole, 5 mM MgSO₄, 3 mM β-mercaptoethanol, 10% glycerol) and lysed by sonication. The lysate was centrifuged, the supernatant was incubated with Ni²⁺ agarose (Macherey-Nagel) and washed with lysis buffer, and the recombinant proteins were eluted by lysis buffer supplemented with 300 mM imidazole. The His_8x-SUMO tag was cleaved by the Ulp1 protease during dialysis against lysis buffer. After dialysis, the His_8x-SUMO tag was removed by a second round of affinity chromatography. The proteins were further purified by gel filtration on a HiLoad 16/600 Superdex 75 column (GE Healthcare) in size exclusion buffer {50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM TCEP [Tris(2-carboxyethyl)phosphine hydrochloride], 5% glycerol}. The purified proteins were mixed in a 1:1 molar ratio, concentrated to 5 mg/ml, and immediately used for crystallographic trials. After purification, proteins were concentrated to 5 mg/ml and stored at −80°C.

Crystallization and structural analysis.

The nsp10:16 protein complex was supplemented by 1 mM sinefungin. Crystals of the nsp10:nsp16 protein complex grew in 3 days in a sitting drop that was created by mixing 300 nl of protein solution and 300 nl of mother liquor (0.1 M MES [morpholineethanesulfonic acid], pH 6.5, 15% [vol/vol] polyethylene glycol monomethyl ether 500 [PEG 500 MME]). They were cryoprotected in mother liquor supplemented with 20% glycerol and flash frozen in liquid nitrogen. The crystals diffracted to 2.2-Å resolution and belonged to the trigonal P3₁ space group. Data were collected using a home source (Rigaku) equipped with a Pilatus detector (Dectris). The data were indexed, scaled, and integrated using XDS (47). The structure was solved by molecular replacement using the structure of the SARS-CoV-2 nsp10:nsp16 (PDB accession code 6YZ1) as the search model in Phaser (48) and was refined in Phenix (49), together with manual building in Coot (50). The structure was refined to good R factors (R_work = 16.41% and R_free = 21.01%) and good geometry as summarized in Table 1. RMSDs were also calculated in Coot using the secondary-structure-matching algorithm.

RNA preparation.

m⁷GpppA was prepared chemically according to a published protocol (51). A 35-mer m⁷GpppA-capped RNA was prepared by in vitro transcription using an DNA template (5′-CAGTAATACGACTCACTATAGGGGAAGCGGGCATGCGGCCAGCCATAGCCGATCA-3′) and the TranscriptAid T7 high-yield transcription kit (Thermo Scientific). The reaction was performed in a 50-μl mixture containing 1× TranscriptAid reaction buffer, 7.5 mM nucleoside triphosphates [NTPs], 6 mM cap analog, 1 μg template DNA, and 1× TranscriptAid enzyme mix. The mixture was incubated for 8 h at 37°C. Next, the RNA was purified using RNA Clean & Concentrator-5 from Zymo Research. DNase I treatment was performed directly on the column according to the purification protocol for 15 min at room temperature.

Analysis of the nsp10:nsp16 enzymatic activity.

The reaction was performed in a total volume of 15 μl. The starting reaction mixture contained 10 μM SAM (S-adenosyl methionine), 5 μM m7GpppA-capped RNA, and 1 μM nsp10 in the MTase reaction buffer (5 mM Tris, pH 8.0, 1 mM MgCl₂, 3 mM dithiothreitol [DTT]). The reaction was initiated by adding 500 nM nsp16 and incubated at 30°C while shaking at 350 rpm. After 15 min of incubation, 15 μl of 40% methanol was added and the samples were analyzed on an Echo mass spectrometry system coupled with a Sciex 6500 triple-quadrupole mass spectrometer operating with an electrospray ionization source. The spectrometer was run in the multiple-reaction-monitoring (MRM) mode with the interface heated to 350°C. The declustering potential was 20 V, the entrance potential 10 V, and the collision energy 28 eV. Ten nanoliters was injected in the mobile phase (flow rate of 0.46 ml/min; 70% acetonitrile with 0.1% formic acid). The characteristic product ion of S-adenosylhomocysteine (SAH), m/z 385.1 > 134.1, was used for quantification.