Cryo-EM structures of the SARS-CoV-2 endoribonuclease Nsp15 reveal insight into nuclease specificity and dynamics

Abstract

Nsp15, a uridine specific endoribonuclease conserved across coronaviruses, processes viral RNA to evade detection by host defense systems. Crystal structures of Nsp15 from different coronaviruses have shown a common hexameric assembly, yet how the enzyme recognizes and processes RNA remains poorly understood. Here we report a series of cryo-EM reconstructions of SARS-CoV-2 Nsp15, in both apo and UTP-bound states. The cryo-EM reconstructions, combined with biochemistry, mass spectrometry, and molecular dynamics, expose molecular details of how critical active site residues recognize uridine and facilitate catalysis of the phosphodiester bond. Mass spectrometry revealed the accumulation of cyclic phosphate cleavage products, while analysis of the apo and UTP-bound datasets revealed conformational dynamics not observed by crystal structures that are likely important to facilitate substrate recognition and regulate nuclease activity. Collectively, these findings advance understanding of how Nsp15 processes viral RNA and provide a structural framework for the development of new therapeutics.

Fig. 1. Architecture of hexameric SARS-CoV-2 Nsp15.

a Domain organization of Nsp15. The numbering corresponds to the amino acid residues at the domain boundaries. The Nsp15 N-terminal domain (ND) is shown in orange, the middle domain (MD) is shown in red, and the poly-U-specific endonuclease domain (endoU) is shown in beige. b Orthogonal views of the cryo-EM map reconstruction of UTP-bound Nsp15 (left) and its corresponding model shown as a cartoon (right). Each Nsp15 protomer is colored as green (P1), purple (P2), coral (P3), tan (P4), gray (P5), and blue (P6). c Model of Nsp15 protomer shown as a cartoon and colored as seen in (a). The position of the Nsp15 endoU active site is highlighted with a black box and the modeled nucleotide is shown as sticks (cyan). N and C mark the N- and C-termini, respectively.

Fig. 2. 5′-UMP coordination by the Nsp15 endoU active site.

a Amino acid sequence alignment of endoU active site residues from Nsp15 homologs. Secondary structure motifs observed in the nucleotide-bound Nsp15 cryo-EM structure are shown above with their corresponding amino acid residue boundaries. b Nucleotide-bound Nsp15 cryo-EM map reconstruction with protomers colored as seen in Fig. 1b. Excess UTP was added to the sample resulting in additional density within all six endoU active sites. The nucleotide density is colored in cyan and the black box demarcates the endoU active site. c Nsp15 coordination of 5′-UMP ligand. Due to the poor density of the UTP β- and γ-phosphates, 5′-UMP was modeled into the active site. Cartoon model of the 5′-UMP-bound Nsp15 endoU active site where 5′-UMP (cyan) and individual residues H235, H250, K290, S294, and Y343 are shown as sticks (top). Model of uracil base discrimination shown as sticks (bottom). d RNA cleavage activity of Nsp15 variants (2.5 nM) incubated with FRET RNA substrate (0.8 μM) over time. RNA cleavage was quantified from three technical replicates. The mean and standard deviation are plotted and P values of wt-Nsp15 compared to H235A (blue; P < 6.4 × 10⁻⁹) and H250A (red; P < 6.2 × 10⁻⁹) are reported from two-tailed Student’s t tests.

Fig. 3. Extracted ion chromatograms and MS spectra of Nsp15 RNA cleavage products.

a Extracted ion chromatograms of the 3′-TAMRA labeled AA-TAMRA RNA cleavage product. 5′-HO-AA-TAMRA is readily observed in the presence of Nsp15 (+Nsp15) as a singly charged ion at m/z 1463.42, but is undetectable in the absence of enzyme (-Nsp15). NL stands for normalized level. b MS spectrum confirms the identity of 5′-HO-AA-TAMRA and does not detect the presence of alternative 5′-cleavage products. c Extracted ion chromatograms of the 5′-fluorescein labeled (FI) FI-AAAU cleavage product. FI-AAAU is only detected in the presence of Nsp15. d MS spectrum confirms the identity of FI-AAAU. The 3′-product is observed primarily as a doubly charged ion at m/z 914.14, which corresponds to a 2′3′-cP terminated moiety. A doubly charged ion at m/z 923.14 is also present and corresponds to a 3′-P terminated species. MSMS spectra of m/z 1463.14 and m/z 914.14 unambiguously confirm the identity of these ions (Fig. S5) e Extracted ion chromatograms of the 5′-FI-AAAU-2′3′-cP cleavage product (top) and the 5′-FI-AAAU-3′-P cleavage product (bottom) set to the same scale to demonstrate that under the conditions employed, the majority of the cleavage product is terminated by a cyclic phosphate. The graphical representation of the areas under the curve of the extracted ion chromatograms shows that approximately 80% of the cleavage product is the cyclic phosphate (assuming similar ionization efficiencies of the two species). f Cartoon schematic of RNA cleavage by the uridine-specific Nsp15 endoribonuclease. Nsp15 catalyzes a transesterification reaction that results in 2′3′-cyclic phosphate (2′3′-cP) and 5′-hydroxyl (5′-OH) RNA ends. Nsp15 can further catalyze a subsequent hydrolysis reaction resulting in the conversion of the 2′3′-cyclic phosphate to a 3′-phosphate (3′-P), however, the rate of hydrolysis is slow. R denotes the 5′ and 3′-ends of the phosphodiester backbone.

Fig. 4. SARS-CoV-2 Nsp15 active site superimposition with metal-independent endoribonucleases.

a SARS-CoV-2 Nsp15 catalytic residues are shown as sticks (PDB ID 7K0R; beige) along with its bound 5′-UMP nucleotide (cyan) and model 3′-phosphate (3′-PO₄). b Bos taurus (Bt) RNase A (PDB ID: 1O0N; magenta) bound to 3′-UMP (cyan), c Homo sapiens (Hs) angiogenin (PDB ID: 1HBY; brown) bound to 3′-phosphate (3′-PO₄), d SARS-CoV-1 Nsp15 (PDB ID: 2H85; orange), e MERS Nsp15 (PDB ID: 5YVD; green), and f PRRSV Nsp11 (PDB ID: 5EYI; gray) are superimposed with the SARS-CoV-2 Nsp15 active site.

Fig. 5. 3D Variability of the endoU Domain.

a 3D variability analysis of cryo-EM reconstructions of Nsp15 apo-states. Frame 1 (gray) and frame 20 (blue) of a 20-frame set describing the motion by the first eigenvector of the cryo-EM reconstructions. Black outlines demarcate the unique views of the endoU domains and dotted arrows demarcate regions of conformational heterogeneity. b Superposition of rigid-body docked models of the Nsp15 hexamer into the Apo Nsp15 H235A (dataset ii) reconstructions derived from 3D variability analysis. c View of the individual P1 protomer from (b) illustrating the conformational dynamics of the endoU domain. The linker between the endoU and MD is colored in gray.

Fig. 6. Molecular dynamics reveal UMP locks the endoU domain.

a Dynamical cross-correlation matrix (DCCM) analysis of a 5′-UMP-bound protomer and its 5′-UMP-free form from hexamer systems. Both inter- and intra-domain positive correlations are significantly enhanced due to 5′-UMP binding. The correlations were calculated from the equally spaced 100-configurations extracted from the last 500 ns of each simulation. Black boxes highlight the positively correlated motions between the endoU domains within the hexameric assembly (box i) and the endoU domain relative to the N-terminal half of the protomer (box ii). b Cartoon illustrating the dynamics of Nsp15 and the mechanism of RNA cleavage. Monomeric Nsp15 is stabilized by the formation of a hexameric assembly. In the absence of ligand the endoU domain exhibits conformational dynamics; however, upon binding to the RNA the endoU domain becomes stabilized. The fixed endoU active site cleaves viral RNA and primarily generates products with a cyclic phosphate.