Biochemical Characterization of Emerging SARS-CoV-2 Nsp15 Endoribonuclease Variants

Graphical abstract


Introduction
Coronaviruses are a family of large, positive stranded RNA viruses that have known zoonotic potential 1 . The Middle Eastern Respiratory Syndrome (MERS-CoV) and Severe Acute Respiratory Syndrome (SARS-CoV-1) coronaviruses have caused several epidemics in the early 2000s. Currently a novel coronavirus, SARS-CoV-2, is responsible for a multi-year pandemic . Coronaviruses are also responsible for 10-30% of common colds 2 . RNA viruses, which replicate with an RNA-dependent-RNA polymerase are known to mutate frequently 3 . Mutations that confer a benefit to the virus are useful, but many mutations are neutral or deleterious. Evolution of RNA viruses is generally accepted to be driven by both genetic drift and negative selection (removing deleterious mutations) 4 . Coronaviruses encode proof-reading machinery that reduces mutation rate, but it is still very high compared to the mutation rate of cellular replication of DNA. Having low replication fidelity (i.e. a high error rate) allows coronaviruses to sample the mutational space by supporting a wide range of sequences during an infection, which leads to fast adaptation in response to selective pressures 5 . Mutations in viral RNA can also arise through damage or the action of RNA editors 6 . The enormous sequencing effort throughout the COVID-19 pandemic has presented the opportunity to analyze patterns of mutations and investigate the functional effects of predominant mutants, lending insight into important structural and functional regions of viral proteins that may otherwise have remained uncharacterized [7][8][9][10][11][12][13][14][15] .
The first two-thirds of the $30 kb coronavirus genome encodes two open reading frames, pp1a and pp1ab, that encode 16 non-structural proteins (nsps) via a ribosomal frame-shifting mechanism 1,16 . The latter one-third encodes structural and accessory proteins, such as the spike protein. Many of the nsps function in the replication transcription complex (RTC) and are necessary for viral replication 17 . Studies have shown mutational frequency varies across the SARS-CoV-2 genome 18 ; mutation frequency seems to be higher in regions encoding structural proteins, and lower in the region encoding nsps [18][19] . While structural proteins, especially the spike protein, have been extensively studied, nsps have had less focus [20][21][22] . However, given their conservation across coronaviruses and lower mutational rates compared to structural proteins, nsps make attractive anti-viral targets.
One such target is Nsp15, a uridine-specific endoribonuclease that processes viral RNA to prevent the activation of the dsRNA sensor MDA5 [23][24][25] . In vivo studies in animal models infected with coronaviruses with inactivated Nsp15 show lower mortality and reduced pathogenicity 23,[25][26][27] . Recent structural studies of Nsp15 have revealed important molecular details regarding how it processes both single and double stranded (ds) RNA [28][29][30][31] . SARS-CoV-2 Nsp15 is a hexamer formed from a dimer of trimers; monomeric Nsp15 is inactive [29][30][32][33] . Each protomer consists of an N-terminal (NTD), middle (MD), and catalytic endoribonuclease (EndoU) domain. Nsp15 is only active as a hexamer and the hexamer is critical for supporting dsRNA binding 28 . Beyond Nsp15, EndoU domains are found in all kingdoms of life and share a catalytic triad with RNase A 34 . Like RNase A, Nsp15 uses a two-step transesterification mechanism to cleave RNA 3 0 of uridines 29 . While much progress has been made to understand how Nsp15 functions to processes viral RNA, many open questions remain. Moreover, in comparison with the EndoU domain the NTD and MD's from SARS-CoV-2 Nsp15 remain poorly characterized.
Increased understanding of the molecular details of Nsp15 structure and function provides a platform to better understand the effect of mutations in the Nsp15 coding sequence. Interestingly, in certain variants, Nsp15 mutations have been noted as clade defining markers [35][36] . For example, in an analysis of different Delta variant clades, Nsp15 H235Y is a marker for Delta subclade C and K260R is a marker for Delta clade E. Nsp15 K13N was identified as a marker for a subclade of the B.1.1.33 lineage in Brazil 37 . Additionally, Nsp15 T113I is a marker for Omicron 38 and was computationally predicted to have the greatest impact on fitness of Orf1b (Nsp12-16) mutations analyzed across 6.4 million SARS-CoV-2 genomes. 39 However, it remains unknown how any of these specific mutations impact Nsp15 activity.
Our goal was to use the Nsp15 mutational information from the first year of the COVID19 pandemic to characterize how mutations affect its endoribonuclease activity, either directly or through changes in its oligomerization state. We analyzed Nsp15 mutations found in sequences deposited in the GISAID database as of June 2021 40 . We selected mutations from each domain (NTD, MD and EndoU) based on frequency of mutation and reports of specific variants in the literature, as well as our knowledge of active site residues. We then carried out in vitro biochemical assays of the recombinantly purified mutants to evaluate their oligomeric state and test nuclease activity compared to WT. Through this approach we have established how these amino acids changes seen in SARS-CoV-2 patient samples impact Nsp15 function. Moreover, through this approach we have gained additional knowledge about important regions of Nsp15.

Results
Identifying Nsp15 mutants from SARS-CoV-2 genome sequencing Approximately 1.9 million full length sequences of Nsp15 were extracted from the EpiCov section (allnuc0614) of the GISAID database, which is a worldwide repository of viral isolates. The sequence of Nsp15 of each viral isolate was compared to the original Wuhan isolate. In line with the relatively low replication fidelity of coronaviruses, single nucleotide variants were present across all three domains of the endoribonuclease. Of the full-length Nsp15 sequences (1038 bases), 1025 positions were observed to have nucleotide substitutions (Supplemental File 1), which were corresponding to changes in 341 out of 346 amino acids from Nsp15. However, 87% of these variants were observed in less than 100 isolates (Supplemental File 1). Six amino acid variants were observed in over 10,000 isolates, where five of them (N74N,  43,964; D79D, 15,684; L214L, 40,846; L217L,  74,888; N278N, 10,718) are synonymous and one (D220Y, 10,323) is non-synonymous ( Figure 1 (A)). For subsequent biochemical analysis we focused only on non-synonymous variants that led to a change in the protein sequence (Figure 1(B)). As expected, surface residues were more frequently mutated than core residues 19 .
Four of the Nsp15 non-synonymous substitutions show multiple base substitutions, G18R, R207S, K290N, and W333C. In each case there is a definite skewing of the abundance of one of these mutations. For G18R, the G to A substitution is 152-fold more prevalent than the G to C substitution. A simple explanation for this could be the elevated transition mutations relative to transversions, which was noted long ago 44 . However, the G to A transition on the positive strand of SARS-CoV-2 is one of the least frequent transition  substitutions observed in SARS-CoV-2 45 . The other three notable transversions are all either G to T or G to C transversions. The G to T transversion is 152to 3087-fold more prevalent in these cases. This is substantially different from what is seen in either SARS-CoV-2 or human genomes 46 where G to C transversions are more common. The overrepresentation of the G to T events could also be due to a jackpot effect. While these observations were made with the GISAID allnuc0614 dataset (Supplemental File 1), these differences persist in newer versions (allnuc0215; Supplemental File 2). There has been significant recent analysis of genome-wide mutational spectra, especially as it relates to human cancer. Of particular interest has been the distribution of trinucleotide-centered mutational motifs and the mechanism(s) by which they occur. The only statistically significant trinucleotide motif found mutated in SARS-CoV-2 genomes was uCn 45 . None of the codons in the substitutions of interest within Nsp15 match that motif.
While the initial analysis of SARS-CoV-2 sequences in June 2021 provided the rationale for selecting Nsp15 mutations to characterize biochemically, we have continued to monitor SARS-CoV-2 variant lineages to understand how the Nsp15 mutations we selected appear in Delta and Omicron (Supplemental File 3). Delta, the predominant SARS-CoV-2 VOC through most of 2021, was replaced by the Omicron VOC beginning in late 2021. Omicron is not a direct descendant of delta; its origin is completely independent 47 . One can see this in our analysis of the mutational frequencies of the non-synonymous amino acid mutations within Nsp15 (Supplemental Table 2, Supplemental File 4). If Delta was the progenitor of Omicron, some of the mutations present in Delta would be present in nearly all of the GISAID Omicron genomes yet we see that for all nonsynonymous substitutions only a fraction of the Omicron genomes carry any one substitution seen in the Delta genomes. Subsets of both Delta and Omicron genomes carry nearly all of these substitution mutations suggesting convergent evolution of these mutations in different SARS-CoV-2 lineages. Their persistence in both lineages suggests that any functional alteration to Nsp15 is well tolerated, perhaps due to compensation by other non-structural proteins.

N-terminal domain variants
We assessed the impact of mutations on K13, G18, and T34I ( Figure 3(A)) from the NTD of Nsp15, which has been shown to be critical for oligomerization 30,43,48 . In particular, K13 sits at the interface between neighboring protomers, and participates in water-mediated interactions with RNA 30 Supplemental Figure 2). G18 is also positioned near the dsRNA binding platform within the Nsp15 hexamer 28 . Given the importance of the NTD in oligomerization, we compared the amount of hex-amer and monomer from each purified variant. WT Nsp15 has a hexamer/monomer ratio around 1 (Table 1, Figure 3(B)). K13R had a slightly greater amount of hexamer, while T34I had a ratio similar to WT. The other variants all had increased amounts of monomer. This provides additional evidence that disturbances in the NTD affect hexamer stability and suggests that these variants would result in less active Nsp15 in virus infected cells since monomeric Nsp15 is inactive.
Using an established FRET nuclease assay 29,42 the cleavage activity of the hexameric forms of the K13N, K13R, G18E, G18R, and T34I variants was measured (Figure 3(B)). Nsp15 K13N showed the largest decrease in cleavage activity, while K13R, G18E, and T34I also had less activity compared to WT Nsp15 (Figure 3(B)). Nsp15 G18R nuclease activity was unchanged compared to WT Nsp15 (Figure 3(B)). Given the high number of isolates with T34I, its decrease in activity contradicts the notion that variants with high amounts of isolates are more advantageous for the endoribonuclease. T34 is a core residue in the NTD, surrounded by non-polar residues (Supplementary Figure 2). As a core residue, it is likely important for maintaining the correct fold and stability of the NTD. Analysis of T34 variants in the SARS-Cov-2 Molecular Dynamics database 58 suggests that T34I would have a stabilizing effect on Nsp15 while mutation to alanine, arginine, or lysine at the same position would have a destabilizing effect. Thus, we assume that introduction of a non-polar isoleucine at this position would likely not disturb the fold of the NTD. The stabilizing effect of T34I could impact nuclease activity in other ways such as altering conformation dynamics within the hexamer.
Overall, this FRET data supports the hypothesis that the NTD provides charge-mediated stability to RNA substrates 30 . The addition of a positive charge (G18R) did not negatively impact nuclease activity, while the loss of a positive residue (K13N) or the introduction of a negative residue (G18E) decreased activity. We also tested nuclease activity on a longer, more physiologically relevant RNA substrate corresponding to the Transcriptional Regulatory Sequence (TRS) for the SARS-CoV-2N protein using a gel-based cleavage assay. The trends seen in the FRET assay were confirmed by gel-based cleavage (Table 1, Supplemental Figure 3).

Middle domain variants
We analyzed middle domain variants from two distinct faces of Nsp15 0 s MD (Figure 4(A)). T113, V128, D133, and L163 lie on one face of this domain, while A93 and T115 sit on the opposite face. All the MD variants had increased amounts of monomer, highlighting that the MD also plays an important role in stabilizing the hexamer. Previous work on MERS Nsp15 characterized several residues in the MD involved in oligomerization 41 . When we mapped these variants onto our SARS-CoV-2 Nsp15 structure, we see T115 and L163 facilitate interactions with neighboring protomers. T113 occupies the same loop as T115 but faces the opposite direction in our model and is further removed from the hexamer interface. Interestingly, T113I is the only MD variant in our analysis which showed increased hexamer formation. A recently determined structure of Nsp15 bound to dsRNA revealed that T113 forms part of a large platform spanning three protomers from the hexamer that supports dsRNA engagement 28 . T115 lies near the interface with a protomer from the bottom trimer, while L163 sits at the interface with a neighboring protomer in the same (top) trimer (Supplemental Figure 2). L163F dramatically reduced the amount of purified hexamer, suggesting this residue is particularly important for oligomerization (Table 1 , Figure 4(B)).
Our FRET activity assay revealed a large, significant increase in activity with the T113I and V128F mutants, and a significant decrease in activity for L163F (Figure 4(C)). The remaining variants A93S, T115A, and D133Y showed no significant changes in cleavage activity. The activity trends seen by FRET were again confirmed using a longer RNA substrate in a gelbased cleavage assay (Table 1, Supplemental Figure 3-4).
Located in the middle domain, V128 resides in a surface exposed loop away from protomer interfaces; therefore, structural analysis does not provide a clear reason for this residue affecting cleavage significantly-Nsp15 V128F is twofold more active than WT Nsp15. However, the change from a small hydrophobic residue to a large hydrophobic residue could possibly lead to changes in the domain conformation to avoid surface exposure. Similarly, T113I also lies in a surface exposed loop that would not be predicted to lead to an increase in activity in the cleavage of single stranded RNA. This is not seen when T115, two residues down on the same loop, is mutated. These two variants also show opposite trends in oligomerization; T113I yields increased hexamer, while T115A leads to increased monomer. These data suggest surface exposed loops can alter structural conformations to affect cleavage. In virus, increased nuclease activity could potentially tip the balance from being protective (cleaving viral RNA to avoid dsRNA sensors) to harmful (preventing viral replication due to increased cleavage). Thus, our in vitro nuclease assays suggest these mutations could be detrimental to viral proliferation, yet they were found in numerous isolates. In fact, T113I was computationally predicted to increase viral fitness 39 . Finally, in a recently determined structure of Nsp15 bound to a dsRNA substrate, T113 forms part of a large platform for dsRNA binding, thus T113 likely facilitates the binding of longer RNA substrates (Supplemental Figure 5) 28 . In the future to better understand the impact of the T113I mutation and the other variants, mutational studies in virus would be beneficial.
Along with a significant reduction in hexamer, the remaining L163F hexamer was less active than WT. The combined oligomerization and nuclease activity results for this variant suggest this mutant would produce very little active enzyme in virus. This residue is located near residues N164 in SARS-CoV-2 (equivalent to N157 in MERS) 41 . Mutation of MERS N157 decreased RNA binding affinity, reduced nuclease activity, decreased oligomeriza-tion, and decreased the thermal stability of the endoribonuclease 41 . Our analysis of L163F in SARS-CoV-2 supports earlier work that this interface is critical for nuclease activity and protein stability.

EndoU domain variants
Finally, we analyzed specific variants from the EndoU domain, which contains the uridine binding pocket and catalytic core ( Figure 5(A)). P206 (which exists at the end of the long linker between the MD and EndoU domains), R207, K260 and V321 occupy the face of the EndoU domain opposite of the active site. H235, K290, and W333 lie in the active site on the other side of the EndoU domain and have been well-characterized biochemically [29][30][41][42] . H235 and K290, along with H250, form the catalytic triad and interact with the scissile phosphate. It is therefore somewhat surprising that H235Y would be found in so many isolates (3630)-in fact, recent work characterized this mutation as a lineage marker for a clade of the Delta variant, Delta D. W333 forms important p-stacking interactions with the base 3 0 of the cleaved uridine. D220 occupies the same face as the active site but is farther removed from the RNA binding sites. It is surface exposed, and the side chain does not form any interactions with other residues (defined by a distance of <5 A). In contrast to the other domains, several EndoU variants showed an increased hexamer/monomer ratio: P206S, H235Y, K260R, and K290N. R207S, Table 1 Summary of Nsp15 variants analyzed in this study. A summary table of the variants categorized by domain of Nsp15 (light orange, ND; medium orange, MD; dark orange, EndoU). The rate refers to mutation rate as a percentage. The impact score q value describes its position in the distribution of all variants where the range is 0-1 and 0.5 is the mean q score 58 . The hexamer/monomer ratio is displayed along with the activity of the variant compared to WT as a percentage (based on FRET endpoint data; NSC, no significant change). Residues with asterisks (*) denote variants below the 1,000 threshold in GISAID that were selected based on previous biochemical data. D220Y, V321M and W333C showed a decreased hexamer/monomer ratio indicative of increased monomer ( Figure 5, Table 1). As expected, our FRET endoribonuclease assay confirmed that the catalytic residue variants H235Y and K290N were inactive ( Figure 5(C)). The other active site residue W333C also showed significantly reduced cleavage activity, confirming its importance in pi-stacking with the RNA substrate. R207S and K260R also showed significantly reduced activity, while P206S and D220Y were not significantly different and V321M had increased activity. Gel-based cleavage assays with a longer RNA substrate showed similar trends to the FRET assay data (Table 1, Supplemental Figure 6).
V321M appears to be involved in the structural integrity of the EndoU fold, as part of a b-sheet packed against an a-helix on the side opposite the active site. Our cleavage data suggests a larger sidechain at that position may change the shape of the EndoU domain to promote more efficient cleavage. This is interesting because V321M was also identified in the same computational fitness analysis as T113I, making it another great candidate for future studies in virus.
P206S and D220Y having similar activity as WT aligns with our initial hypothesis that mutants with high numbers of isolates would not significantly change nuclease function. Too much viral RNA cleavage and too little viral RNA cleavage are both likely to be detrimental to the virus, thus we would expect the more common Nsp15 variants to retain relatively the same nuclease activity.

Discussion
In this work we biochemically characterized Nsp15 variants extracted from the GISAID database of SARS-CoV-2 sequences acquired during the pandemic. Nsp15 variants are unlikely to drive viral evolution and become lineage defining markers because the protein is not involved directly in viral entry or replication. Several studies have identified Nsp15 mutations that may have increased fitness and clinical significance, likely by suppressing interferon I response 39,49 . Thus, characterizing variants seen in patient samples provides new knowledge about protein function beyond characterizing the active site residues directly involved in catalysis. Identifying new regions of Nsp15 that affect oligomerization and nuclease activity builds a base for mapping regulatory regions and additional drug design targets. We specifically analyzed changes in oligomerization state and nuclease activity. The majority of the variants studied resulted in decreased Nsp15 activity. This work highlights the importance of combining bioinformatics and structural data to predict mutation effects and then testing those hypotheses in vitro. For example, based on Nsp15 structures alone, we would not have predicted a large impact of the T34I mutation, as it is a core residue that only interacts with other residues in the N-terminal domain. Indeed, the oligomeric state is not perturbed; however, our nuclease assays reveal a significant decrease in activity. Similarly, V128F would also not be predicted to have a great impact based solely on structural information, since it is not directly involved in any protomer interfaces and far from any EndoU active sites. However, our assays showed it decreases the formation of hexamer, but the hexamer formed has a substantial increase in nuclease activity. This suggests that structural conformations in that region could be manipulated to regulate nuclease activity, perhaps making it a surface where protein-protein interactions take place inside infected cells. Biochemical characterization is not high throughput, therefore identification and selection of variants to test is important. Mutations with a high number of isolates are an easy selection criterion; our selection of K13, T34, T115, and R207 agree with bioinformatics analysis of main mutated residues across the SARS-CoV-2 proteome published in early 2021 18 . Knowledge of molecular mechanisms is also important; this led us to select N-terminal residue and active site residue mutations to further test the established roles of these amino acids 30,[41][42]48 . We recently deter- (purple), Nsp15 K290N (green), Nsp15 V321A (green) and Nsp15 W333C (light blue). Asterisk (*) signifies mutants were separated using a different AKTA system which affected elution volume. ***p < 0.001, ****p < 0.0001. mined a structure of Nsp15 bound to dsRNA 28 ; of the Nsp15 mutants we characterized here, the only non-active site residues in the area of the dsRNA are T113 ($4 A), D133 ($5 A) and V128 ($9 A) (Supplemental Figure 5).
Coronavirus genomes have a bias for uridines [50][51] so the evolution of a uridine-specific nuclease is interesting. Clearly, regulation of viral RNA through cleavage of uridines both in the polyU sequence of the negative strand (complementary to the polyA tail) 23 and throughout the positive strand 52 is important for the coronavirus life cycle. Except for the Nsp15 H235Y mutation, which serves as a marker of a Delta viral clade 35 , none of the active site residues accumulated mutations above the 1000 isolate threshold we used. The lack of nuclease dead isolates signifies the importance of nuclease activity for the virus. Thus, Nsp15 H235Y stands out as a candidate to further study in viruses.
A recent computational modeling study suggested Nsp15 serves as a scaffold for the assembly of the Replication Transcription Complex (RTC) 53 . Thus, some of the variants may disrupt RTC protein-protein interactions. In the MD, Nsp15 T115 appears to interact with Nsp8, while D133 is positioned near Nsp10 (Supplemental Figure 7). Nsp15 EndoU residues D220 and V3231 neighbor the Nsp16/10 complex in the RTC, with V321 at the end of Nsp10 not interfacing with Nsp16 and D220 close to Nsp16 (Supplemental Figure 7). In this function of scaffolding the RTC, nuclease activity may not be necessary. Thus, the H235Y and K290N mutations which kill catalytic activity but increase hexamer formation (Table 1), could support the RTC model. The Delta clade with Nsp15 H235Y has adaptations in other virallyencoded immune evasion proteins that may fill the gap of a nuclease dead Nsp15. However, a recent preprint analyzed transmission dynamics of SARS-CoV-2 variants and found the H235Y mutation along with several other mutations in that Delta clade led to transmission suppression compared to other lineages 54 , evidence that a lack of Nsp15 activity may contribute to poor viral transmission.
Work to develop virus-like particles to study SARS-CoV-2 variants found that an important RNA packaging signal overlaps with the Nsp15 coding sequence 55 ; therefore, variants that do not affect oligomerization or activity nonetheless could affect RNA packaging. Similarly, synonymous amino acid mutations may also result in changes to RNA packaging. More research is needed to understand the determinants of that cis-acting element 55 . Additionally, a recent preprint proposes an RTC model that positions Nsp15 to discriminate cleavage sites based on RNA structure as opposed to sequence 56 . This may be another way that variants impact function at the RNA level as opposed to the protein level.
Biochemical and in vivo studies of SARS-CoV-2 protein variants beyond the spike protein will contribute to a fuller picture of how the coronavirus proteome functions and changes over time. Recent work analyzing RdRp variants and the effect on remdesivir inhibition serve as one example 57 . Earlier this year a SARS-CoV-2 molecular dynamics database was created that takes the wealth of bioinformatics data and structural information we have on SARS-CoV-2 and applied molecular dynamics simulations to estimate the impact of variants 58 . This high-throughput source of information is a great starting point to better understand viral protein evolution. The addition of in vitro data points, like the analyses presented in this paper, as well as in vivo analyses, to repositories like this database could facilitate a holistic understanding of the molecular impacts of viral mutations.

Nsp15 variants analysis
A total of 51,536,473 entries of the SARS-CoV-2 nucleotide sequences were downloaded from GISAID (version: allnuc0614). The downloaded data contains the coding sequences for genes including Nsp15 based on the hCoV-19/Wuhan/ WIV04/2019 reference. Only full-length Nsp15 sequences (1038 bp) were retained for the downstream analysis, so 1,905,411 entries were included in the mutational analysis. The sequences of Nsp15 isolates were then aligned to that of the original Wuhan isolate (GenBank NC_045512.2) using the nucmer command from MUMmer 4.0 package 59 with default parameters. Variant data from the nucmer alignments were generated using show-snps command and the output was parsed directly using a custom perl script to convert MUMmer snps file to a tab delimited vcflike table. Protein annotation was included in the final table (Supplemental File 1). Variants selected for testing were primarily at the amino acid level. Mutations that did not result in a specific amino acid change were discarded. The snps caused the same amino acid change (i.e. H235Y) were then summed to provide the total number of isolates per amino acid mutation. Visualizations of the variants on the Nsp15 DNA and protein sequences were generated via the lollipop function in trackViewer R package 60 .

Protein expression of Nsp15 variants
WT Nsp15 was previously synthesized by Genscript (Piscataway, NJ), and contains an Nterminal His tag with thrombin and TEV cleavage sites in pET14b 29 . For this study, Genscript mutated the WT sequence to our variants of interest (Supplemental Table 1). WT Nsp15 and Nsp15 variants were overexpressed in E. Coli C41 (DE3) competent cells cultured in Terrific Broth supplemented with 100 lg/mL ampicillin. Transformed cell cultures were grown to an optical density (600 nm) of 0.6-1.0 at 37°C prior to induction with 0.2 mM IPTG and overnight incubation at 16°C. Harvested cells were stored at À80°C until needed.

Protein purification
Protein purification was carried out as previously described 29 . Cells were resuspended in lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5% glycerol, 5 mM b-ME, 5 mM imidazole) and then supplemented with complete EDTA-free protease inhibitor tablets (Roche). They were disrupted by sonication and the lysate was clarified by centrifugation at 15,000 rpm for 50 min at 4°C, followed by incubation for 45 minutes with a TALON metal affinity resin (Clontech). His-Thrombin-TEV-Nsp15 variants were eluted with 250 mM imidazole and incubated with thrombin (Sigma) at room temperature in Thrombin Cleavage Buffer (50 mM Tris pH 8.0, 150 mM NaCl, 5% glycerol, 2 mM b-ME, 2 mM CaCl 2 ) for a 4 h time period. Thrombin cleavage was quenched by the addition of 1 mM PMSF (phenylmethylsulfonyl fluoride). The cleavage reactions were incubated with TALON metal affinity resin, and tagless protein was eluted in batch and resolved by gel filtration using a Superdex-200 column equilibrated with SEC buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM MnCl 2 , 5 mM b-ME). The peak fraction corresponding to the hexamer was used in subsequent assays. SDS-PAGE was used to assess protein purity (Supplemental Figure 1).

Nsp15 endoribonuclease FRET assay
Real-time Nsp15 RNA cleavage was monitored as previously described [29][30] with minor modifications to the protocol. The 5 0 -fluorescein (FI) label on the RNA substrate is quenched by its 3 0 -TAMRA label (5 0 -FI-AAAUAA-TAMRA-3 0 ). A negative control substrate (5 0 -FI-AAAAAA-TAMRA-3 0 was also used (no cleavage observed; data not shown). The FRET RNA substrate (0.8 lM) was then incubated with a constant amount of Nsp15 variant (2.5 nM) in RNA cleavage buffer (20 mM HEPES pH 7.5, 75 mM NaCl, 5 mM MnCl 2 , 5 mM b-ME) at 25°C for a 60 min time period. RNA cleavage was measured as an increase in fluorescein fluorescence. The fluorescence was measured every 2.5 min using a POLARstar Omega plate reader (BMG Labtech) set to excitation and emission wavelengths of 485 ± 12 nm and 520 nm, respectively. Three technical replicates were performed to calculate the mean, standard deviation, and pairwise comparison test. (Dunnett's T3 multiple corrections test, Prism/Graphpad).

Gel-based endoribonuclease assay
Gel-based cleavage assays were performed as described previously 30 . Double labeled RNA sub-strates (5 0 -FI and 3 0 -Cy5, 500 nM) were incubated with Nsp15 (50 nM) in RNA cleavage buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM MnCl 2 , 5 mM DTT, 1 u/ll RNasin ribonuclease inhibitor) at room temperature for 30 min with the reaction samples collected at 0, 1, 5, 10 and 30 min. The reactions were quenched with 2x urea loading buffer (8 M urea, 20 mM Tris pH 8.0, 1 mM EDTA). Loading buffer without dye was used due to the expected size of cleavage products and the size of bromophenol blue. To monitor the gel front, control lanes of protein only with bromophenol blue were run. To generate a ladder, alkaline hydrolysis of the RNA was carried out for 15 min at 90°C using 1 lM RNA in alkaline hydrolysis buffer (50 mM sodium carbonate pH 9.2, 1 mM EDTA) and quenched with 2x urea loading buffer. The cleavage reactions were separated using 15% TBE-urea PAGE gels and visualized with a Typhoon RGB imager (Amersham) using Cy2 (k ex = 488 nm, k em = 515-535 nm) and Cy5 (k ex = 635 nm, k em = 655-685 nm) channels. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.