2.1. General information

Crystallization reagents and supplies were purchased from Hampton Research, Aliso Viejo, California, USA. Crystallographic supplies for crystal mounting and X-ray diffraction data collection at room temperature were purchased from MiTeGen, Ithaca, New York, USA. The Ac-SAVLQSGF-CONH₂ peptide used in this study was custom-synthesized by Biomatik Corp., Kitchener, Ontario, Canada.

2.2. Expression and purification of M^pro/C145A

A DNA insert encoding a 6×His tag followed by a TEV protease cleavage site and SARS CoV-2 M^pro (306 amino acids) bearing an active-site C145A mutation was cloned into the pJ414 vector (ATUM) and transformed into Escherichia coli BL21(DE3) cells (Agilent). The last residue in the TEV protease recognition sequence (ENLYFQG/S) was chosen to be a serine, such that upon cleavage the M^pro/C145A construct starts with the native N-terminal Ser1 residue. Cells were grown in Luria–Bertani medium, and the M^pro/C145A fusion protein was expressed using established protocols. The fusion protein was purified under native conditions by lysis and Ni–NTA affinity chromatography in 20 mM Tris–HCl pH 8, 150 mM NaCl, 1 mM TCEP, 20 mM imidazole. This step was followed by size-exclusion chromatography on a Superose 12 column (16 × 60 cm, Cytiva) to remove minor high-molecular-weight contaminants and to exchange the buffer to 25 mM Tris–HCl pH 8, 100 mM NaCl, 0.5 mM EDTA, 1 mM DTT, 20 mM imidazole, which is suitable for cleavage by TEV protease at a molar ratio of 75:1 fusion protein:TEV protease. TEV protease was prepared as described by Lucast et al. (2001). Upon completion of the cleavage and adjustment of the EDTA concentration to 0.25 mM, M^pro/C145A was collected in the flowthrough from the Ni–NTA column, concentrated and then subjected again to isocratic fractionation on Superose 12 in 25 mM Tris pH 7.6, 150 mM NaCl, 1 mM TCEP. Peak fractions were combined, concentrated and aliquots of the protein were stored at − 20 C. Purity was verified both by SDS–PAGE and electrospray ionization mass spectrometry.

2.3. Crystallography of the M^pro/C145A–substrate complex

We have previously published detailed methods for crystallizing high-quality M^pro crystals using the sitting-drop vapor-diffusion methodology (Kneller, Galanie et al., 2020; Kneller, Phillips, Kovalevsky et al., 2020). Crystallization conditions for the M^pro/C145A mutant were discovered by automated high-throughput screening at the Hauptman–Woodward Medical Research Institute (Luft et al., 2003). The protein was concentrated to 6 mg ml⁻¹. Initially, flower-like crystal aggregates unsuitable for room-temperature X-ray diffraction grew using a sample of 270 µl protein solution mixed with 1.6 mg lyophilized peptide in 20% PEG 3350, 0.1 M bis-Tris pH 7.0. Subsequently, the protein was mixed in a 1:5 molar ratio with peptide from a 50 mM stock in 100% DMSO and used for crystallization with 20% PEG 3350, 0.1 M bis-Tris pH 6.5. This condition produced a final pH in the crystallization drop of 6.5 as measured by a microelectrode. Drops were seeded using a cat whisker and flower-shaped aggregates grew with thicker `petal' substituents after one week of incubation at 14°C. A single crystal from the aggregate was harvested and mounted on loops and capillary setups for room-temperature diffraction.

2.4. Room-temperature X-ray diffraction data collection and structure refinement

Room-temperature X-ray diffraction data were collected on a Rigaku HighFlux HomeLab instrument equipped with a MicroMax-007 HF X-ray generator, Osmic VariMax optics and a Dectris EIGER R 4M hybrid photon-counting detector. Data were integrated using the CrysAlis^Pro software suite (Rigaku, The Woodlands, Texas, USA) and were reduced and scaled using AIMLESS (Evans & Murshudov, 2013) from the CCP4 suite (Winn et al., 2011). The M^pro/C145A–substrate structure was solved by molecular replacement using PDB entry 6wqf (Kneller, Phillips, O'Neill et al., 2020) with Phaser (McCoy et al., 2007) from CCP4. The model was iteratively refined using phenix.refine from the Phenix suite (Liebschner et al., 2019) and Coot (Emsley et al., 2010; Casañal et al., 2020) aided by MolProbity (Chen, Arendall et al., 2010) for geometry validation. Final data-collection and refinement statistics are given in Supplementary Table S1.

3.1. Binding of peptide substrate

In M^pro/C145A–substrate, a substrate occupies each of the two active-site grooves in the M^pro homodimer structure (Fig. 1). There is unambiguous electron density for all residues except for the P3′ Phe side chain. Interactions between M^pro and substrate are nearly identical in both protomers; hence, subsequent details of our analysis will focus on protomer A. Main-chain interactions between the peptide and M^pro are consistent with adding a β-strand to the antiparallel β-sheets flanking the scissile bond [Fig. 2(a)]. There are eight hydrogen bonds between the main chains of the peptide substrate and M^pro, which involve residues Thr190, Glu166, His164, Ser144, Gly143 and Thr26. In the catalytic site, P1 Gln is driven into the S1 subsite via a 2.7 Å hydrogen bond to the His163 imidazole and a water-mediated interaction with Asn142, whereas its main-chain carbonyl O atom is centered into the characteristic oxyanion hole [Fig. 2(b)]. The bulky hydrophobic P2 Leu side chain fills its respective hydrophobic S2 subsite, whereas the smaller P4 Ala juts into the shallow S4 pocket capped by the flexible S5 loop. P2′ Gly contributes to substrate binding via backbone hydrogen-bond interactions with Thr26, whereas P1′ Ser makes no hydrogen bonds to M^pro, as is the case for P5 Ser. The P5 Ser, P3 Val, P2′ Ser and P3′ Phe side chains face the solvent space between crystallo­graphic symmetry-related molecules, with the P3′ Phe phenyl ring being dynamic and poorly defined in the electron-density maps.

Figure 1 Mpro/C145A–substrate complex. Mpro/C145A (protomer A is shown as a purple cartoon and protomer B as an orange surface) was co-crystallized with an octapeptide (green sticks and surface) corresponding to the nsp4/nsp5 autocleavage site. The top inset shows the peptide in protomer B in conventional orientation, while the bottom inset shows protomer A rotated by 180°. Fo − Fc omit difference electron-density maps for each peptide are contoured at 3σ.

Figure 2 Mpro/C145A–substrate interactions for induced fit of substrate. (a) Main-chain hydrogen-bond interactions between Mpro and the peptide substrate (shown as orange and green C atoms, respectively). Hydrogen bonds are shown as dotted lines. Distances are in Å. (b) Hydrogen-bond interactions between peptide P1 Gln with the S1 subsite and the catalytic site. Water molecules are shown as red spheres. (c) Superposition of Mpro/C145A–substrate with the substrate-free Mpro neutron crystal structure (light blue; PDB entry 7jun) showing the scissile bond in relation to the catalytic Cys145–His41 zwitterion. Superposition was by a least-squares fit to Cα atoms of protomer A. (d) Induced fit of Mpro/C145A to the peptide substrate by the rearrangements of side chains in the S2 and S4 subsite.

The C-terminal residues of M^pro flank the last helix in domain III. Here, residues 302–306 constitute a flexible loop (Kneller, Phillips, O'Neill et al., 2020) and are typically disordered in crystal structures of SARS-CoV-2 M^pro to the extent that some crystallographers choose not to model the C-terminus (Dai et al., 2020; Zhang et al., 2020). In this context, it is noteworthy that electron density for the C-terminal residues up to the very last Gln306 amine is clearly observed in M^pro/C145A–substrate (Supplementary Fig. S1). When the C-terminal residues of M^pro/C145A–substrate and the product complex reported by Lee et al. (2020) are superimposed, the main-chain atoms align with an r.m.s.d. of 0.7 Å for residues 303–306 (Supplementary Fig. S2), demonstrating that the C-terminus of M^pro is pre-organized to bind into the active site during M^pro autoprocessing.

3.2. Comparison with substrate-free M^pro

Superposition of M^pro/C145A–substrate and our recently determined substrate-free neutron crystallographic structure of wild-type SARS-CoV-2 M^pro (PDB entry 7jun; Kneller, Phillips, Weiss et al., 2020) offers unique insights into catalysis because the Cys145–His41 dyad was observed as a zwitterionic species with the Cys145 thiolate S^γ atom 3.8 Å away from His41 N^ɛ2, already primed for proteolysis [Fig. 2(c)]. When the two structures are superimposed, the Cys145 thiolate is positioned 1.8 Å away from the scissile-bond carbonyl C atom with an ideal distance and orientation for a nucleophilic attack. The scissile-bond amide N atom is positioned 2.7 Å away from the N^ɛ2 hydrogen on the catalytic His41, which does not significantly change its position in the complex, poised for proton transfer to the amino-terminus that is created in the cleavage reaction. The M^pro/C145A–substrate structure therefore underscores the suggestion that the zwitterionic catalytic dyad is present prior to substrate binding (Kneller, Phillips, Weiss et al., 2020) instead of requiring reorganization of the catalytic site to facilitate proton transfer from Cys145 to His41 in the catalytically resting (nonreactive) state (i.e. a protonated Cys145 thiol and singly protonated His41 imidazole) to create the reactive zwitterionic species (Paasche et al., 2014; Ramos-Guzmán et al., 2020; Díaz & Suárez, 2021).

In subsite S1, Glu166 accommodates P1 Gln by moving ∼0.9 Å away from the substrate and simultaneously rotating to form a 2.8 Å hydrogen bond to His172 that is absent in substrate-free M^pro. This conformational change of the Glu166 side chain is due to the binding of a ligand to the active site and is consistent with the changes observed in M^pro–inhibitor complexes (Kneller, Galanie et al., 2020; Dai et al., 2020). Meanwhile, the Asn142 side chain flips ∼180° and shifts 2.6 Å from its position in the substrate-free structure to produce water-mediated contacts with P1 Gln. This comparison indicates that the catalytic dyad is already pre-organized and primed for catalysis in the substrate-free structure, requiring minimal rearrangements to initiate the peptide-hydrolysis reaction. Induced fit of the peptide substrate to M^pro is achieved primarily through the continuation of β-sheet structures and key rearrangements of the large side chains Met165, Met49 and Gln189 from their positions in substrate-free M^pro to create subsites S2 and S4 [Fig. 2(d)]. The substrate P2 Leu forms the S2 subsite by a >3 Å steric swing of the Met49 side chain located on the short S2 helix (residues 46–50). P2 Leu also acts in consortium with the methyl C atom of P4 Ala to produce the shallow S4 subsite by initiating a new rotational isomer for Met165 and shifting the position of the S4 β-hairpin loop by as much as 1.5 Å. The peptide residue arrangement is stabilized by Gln189 rotating to hydrogen-bond to the main-chain NH of P3 Val. P5 Ser does not interact directly with M^pro. However, the hydroxymethyl side chain might act to nullify repulsion of the Gln189 backbone by intramolecular hydrogen bonding to the carbonyl of P3 Val. The active site around P4 and P5 expands by >1.5 Å from the substrate-free structure to accommodate the substrate, as measured by the C^α distances between Pro168/168′ and Thr190/190′. Interestingly, the S2 subsite helix of the complex shows only minor distal expansion, unlike in M^pro–inhibitor complexes where the S2 subsite helix can shift by >1.5 Å (Jin et al., 2020; Kneller, Galanie et al., 2020). Similarly, the S5 loop (residues 189–194) is only minimally disturbed by the substrate, shifting by <1.0 Å.

3.3. Comparison with SARS-CoV M^pro/H41A

It is instructive to compare our room-temperature M^pro/C145A–substrate structure from SARS-CoV-2 with a low-temperature structure of the SARS-CoV main protease H41A mutant, M^pro/H41A, soaked with an 11-amino-acid peptide substrate spanning the same nsp4/nsp5 autocleavage site (Xue et al., 2008; Fig. 3; PDB entry 2q6g). The peptide substrate in both structures bows into the active site by conforming to the antiparallel β-sheets flanking the catalytic site. Overall, the orientations of the substrate residues in both protomers are in agreement or can be explained by a lack of electron density, with one exception. The C^γ atoms of P2 Leu deviate by ∼1 Å, which is coordinated with the rotational isomer differences observed for the Met49 and Met165 side chains. In the SARS-CoV M^pro/H41A–substrate complex, the substrate pushes the catalytic Cys145 side chain away from the scissile bond towards the vacated space left by the H41A mutation to a position where the S atom is 4.5 Å from the scissile carbonyl C atom. M^pro possesses 96% sequence identity between SARS-CoV and SARS-CoV-2. The nearest mutation to the active site in SARS-CoV-2 M^pro, A46S, does not introduce appreciable differences despite minor differences in the S2 subsite helix.

Figure 3 Comparison of substrate-bound Mpro/C145A from SARS-CoV-2 (PDB entry 7n89) with Mpro/H41A from SARS-CoV (PDB entry 2q6g). (a) Protomer A in Mpro/C145A–substrate from SARS-CoV-2 (orange protein, green substrate) superimposed on the SARS-CoV Mpro/H41A–substrate complex (cyan protein, slate substrate). Superposition was by a least-squares fit of Cα atoms of protomer B in PDB entry 2q6g to protomer A in PDB entry 7n89. (b) Protomer B in Mpro/C145A–substrate from SARS-CoV-2 (orange protein, green substrate) superimposed on the SARS-CoV Mpro/H41A–substrate complex. Superposition was by a least-squares fit of Cα atoms of protomer A in PDB entry 2q6g to protomer B in PDB entry 7n89.