SARS-CoV-2 nsp15 preferentially degrades AU-rich dsRNA via its dsRNA nickase activity

Abstract It has been proposed that coronavirus nsp15 mediates evasion of host cell double-stranded (ds) RNA sensors via its uracil-specific endoribonuclease activity. However, how nsp15 processes viral dsRNA, commonly considered as a genome replication intermediate, remains elusive. Previous research has mainly focused on short single-stranded RNA as substrates, and whether nsp15 prefers single-stranded or double-stranded RNA for cleavage is controversial. In the present work, we prepared numerous RNA substrates, including both long substrates mimicking the viral genome and short defined RNA, to clarify the substrate preference and cleavage pattern of SARS-CoV-2 nsp15. We demonstrated that SARS-CoV-2 nsp15 preferentially cleaved pyrimidine nucleotides located in less thermodynamically stable areas in dsRNA, such as AU-rich areas and mismatch-containing areas, in a nicking manner. Because coronavirus genomes generally have a high AU content, our work supported the mechanism that coronaviruses evade the antiviral response mediated by host cell dsRNA sensors by using nsp15 dsRNA nickase to directly cleave dsRNA intermediates formed during genome replication and transcription.


Introduction
Severe acute respiratory syndrome coronavirus 2 (S AR S-CoV-2) is the pathogen responsible for the coronavirus disease 2019 (COVID-19) pandemic ( 1 ), which, since emerging in late 2019, has resulted in over 650 million infections and over 6.5 million deaths worldwide ( 2 ), severely impacting the global economy .Currently , new strains of S AR S-CoV-2 continue to arise ( 3 ), resulting in new cases and recurring infections.At the same time, the issue of long COVID cannot be disregarded ( 4 ,5 ).
Coronaviruses are a diverse class of single-stranded (ss) positive-sense RNA viruses with a genome size of approximately 30 000 nucleotides, a fact that makes them the largest of the known RNA viruses ( 6 ).Coronaviruses encode 15-16 nonstructural proteins (nsp) to facilitate genome replica-tion and transcription.Among these, nsp7-16 are considered key components of the viral replication-transcription complex (RTC), while nsp12-16 are primarily involved in manipulating the viral RNA ( 7 ).Nevertheless, our understanding of these proteins remains inadequate, and the lack of indepth biochemical and enzymological studies have impeded our understanding of the molecular mechanism of coronaviral genome replication and transcription.Nsp15 is one of the nsps for which the function and exact role in genome replication and transcription remain elusive.
The EndoU domain is found in all kingdoms of life ( 12 ), and in RNA viruses, it is chiefly found in nidoviruses.However, not all nidoviruses have an EndoU domain; only those infecting vertebrates, such as coronaviruses and arteriviruses, have evolved EndoU domains ( 14 ).One speculation about this evolutionary phenomenon is that vertebrate nidoviruses evolved EndoU domains in order to counteract the vertebratespecific innate immune responses against viral RNA, such as the interferon system.This hypothesis is primarily based on several studies that demonstrated that the EndoU activity of coronavirus nsp15 did not directly affect coronavirus replication (15)(16)(17), but instead suppressed the antiviral response of the host cell (16)(17)(18)(19)(20).It was suggested that the EndoU activity of coronavirus nsp15 mediates evasion of host cell double-stranded (ds) RNA sensors by reducing dsRNA produced by viral genome replication and transcription ( 16 ,17 ), and two explanations have been proposed to clarify this process.A study by Hackbart et al. ( 21 ) concluded that coronavirus nsp15 uses its EndoU activity to limit the abundance and length of the 5 -polyU of the viral negative-sense strand RNA, which can fold back and generate stem-loop structures by hybridizing with an A / G-rich domain on the negativesense strand RNA.This stem-loop structure may be recognized as dsRNA by host cell dsRNA sensors.A study by Ancar et al. ( 22 ) concluded that coronavirus nsp15 used its En-doU activity to cleave U ↓ A and C ↓ A sequences within the viral positive-sense strand RNA, including the pre-3 -polyA site, thereby inhibiting viral negative-sense strand RNA synthesis and dsRNA accumulation.Although dsRNA intermediates formed during viral genome replication and transcription are widely acknowledged to be the most probable pathogenassociated molecular patterns (PAMPs) to activate host cell dsRNA sensors ( 16 , 17 , 23-25 ), it is unclear whether coronavirus nsp15 directly cleaves such intermediates.One study examining the biochemical properties of coronavirus nsp15 found greater efficacy in the cleavage of dsRNA than ssRNA ( 26 ), while other studies reported that coronavirus nsp15 was more efficient at cleaving ssRNA than dsRNA ( 27 ,28 ).Furthermore, it has been found that when ssRNA forms secondary structures, nsp15 shows a preference for cleaving more flexible Us, particularly those located on loops ( 29 ,30 ).However, studies of the S AR S-CoV-2 nsp15 structure have revealed specific interactions between nsp15 hexamers and dsRNA ( 28 , 31 , 32 ), but no significant interactions between nsp15 hexamers and ssRNA, except for the active site pocket ( 13 ,33 ).This discrepancy existing between coronavirus nsp15 biochemical properties and structural studies-as well as our interest in the mechanism of nsp15-mediated immune evasionmotivated us to further investigate the coronavirus nsp15 substrate preference, and therefore in this study, we focused on S AR S-CoV-2 nsp15.In previous work, the RNA substrates tested were limited, which may hinder the full characterization of nsp15 activity.In the present work, we prepared a large variety of RNA substrates, including both long substrates mimicking the viral genome and short defined RNA, to clarify the substrate preference and cleavage pattern of S AR S-CoV-2 nsp15.
We initially created RNA substrates to mimic ssRNA and dsRNA present during S AR S-CoV-2 genome replication and transcription, and found that with RNA substrates containing a high AU content, nsp15 exhibited a significant preference for cleaving dsRNA, with distinct cleavage sites compared to ssRNA.We then identified that nsp15 preferentially cleaved consecutive Us located in AU-rich areas in dsRNA, and further demonstrated that the relaxed structure of AU-rich areas facilitated U-flipping, making it easier for nsp15 to recognize and cleave the U sites within AU-rich areas.Furthermore, we found that in addition to the AU content and distribution in RNA, the RNA length also affected the substrate preference of nsp15, likely due to differences in the manner by which nsp15 binds to ssRNA and dsRNA.Finally, we clarified that nsp15 cleaved dsRNA as a dsRNA nickase.On the whole, our work supported the mechanism that coronaviruses evade the antiviral response mediated by host cell dsRNA sensors by using nsp15 dsRNA nickase to directly cleave dsRNA intermediates formed during genome replication and transcription.

Protein expression and purification
DNA fragments encoding wild-type S AR S-CoV-2 nsp15 were amplified by PCR from a pET-32a(+) plasmid carrying the nsp15 gene ( 34 ) and inserted into pET-28a(+) vectors harboring an N-terminal 6 × His tag using a ClonExpress II One Step Cloning kit (Vazyme, China) (complete sequence of the nsp15 expression plasmid is listed in Supplementary Table S1 ).The constructs were transformed into Esc heric hia coli BL21(DE3)pLysS cells (AngYuBio, China).The cells were cultured in 6 l of LB medium containing 50 μg / ml kanamycin and 34 μg / ml chloramphenicol at 37 • C until the OD 600 reached approximately 0.8.The flasks containing the culture were then placed at 4 • C for 1 h.Then, protein expression was induced by addition of 0.2 mM IPTG, and incubation continued at 16 • C for 16 h.The cells were harvested and then stored at −80 • C.
A single active site mutation (H234A) was introduced using the Gibson assembly method (primers for cloning are listed in Supplementary Table S1 ).The H234A mutant was expressed and purified with the same procedure as detailed above.
The VSW-3 RNA polymerase and the T7 RNA polymerase S43Y mutant used in IVT were expressed and purified as described previously ( 35 ,36 ).
Protein concentrations were determined with a Bradford Protein Quantitative kit (Bio-Rad, USA), with bovine serum albumin as a standard.Protein purity was analyzed by SDS-PAGE with Coomassie blue staining.Protein size was determined with Precision Plus Protein Standards (Bio-Rad, USA) in SDS-PAGE analysis.

RNA substrates preparation
For long RNA substrates ( > 500 nt or bp) The S AR S-CoV-2 genome sequence was obtained from the NCBI (NC_045512.2).The DNA templates were synthesized by GenScript or constructed via the Gibson assembly method (primers for cloning are listed in Supplementary Table S2 ) and cloned into a pUC19 vector (sequences of plasmids are listed in Supplementary Table S2 ).The transcription templates (sequences are listed in Supplementary Table S3 ) containing VSW-3 RNA polymerase promoter for positive-sense and negative-sense strand ssRNA were amplified by a threestep PCR using PrimeSTAR Max DNA Polymerase (Takara, Japan) (primers are listed in Supplementary Table S3 ) and then purified with an PCR Cleanup kit (Axygen, USA).For the IVT reaction, 35 ng / μl template DNA was incubated with 1.5 U / μl murine RNase inhibitor (New England Biolabs, USA), 0. ).The RNA substrates related to the ORFs of the S AR S-CoV-2 S protein gene (NC_045512.2),E. coli ung or malE genes (NC_000913.3), or Taq DNA Pol I gene (J04639.1)(sequences of plasmids containing these genes are listed in Supplementary Table S2 ) were prepared with the same procedure as detailed above (transcription template sequences and primers are listed in Supplementary Table S3 ).

For short RNA substrates ( ≤200 nt or bp)
The transcription templates (sequences are listed in Supplementary Table S3 ) containing T7 RNA polymerase promoter for the positive-sense and negative-sense strand ssRNA were amplified by two-step PCR using PrimeSTAR Max DNA Polymerase and the pUC19 plasmid containing the DNA template of the S AR S-CoV-2 mini-genome or E. coli ung gene with corresponding primers (listed in Supplementary Table S3 ), and then purified with a Monarch PCR & DNA Cleanup kit (New England Biolabs, USA).For IVT reactions, 35 ng / μl template DNA was incubated with 1.5 U / μl murine RNase inhibitor, 0.2 μM inorganic pyrophosphatase, and 0.15 μM T7 RNA polymerase S43Y mutant at 37 • C for 1 h in an IVT buffer.Then, 2 U of DNase I-XT was added into 10 μl of the reaction mixture and incubation was extended for 30 min at 37 • C to remove the template DNA.The transcripts were purified with a Monarch RNA Cleanup kit.The RNA annealing reaction was carried out as described above.
The Gibson assembly method was employed to construct the template-containing plasmids needed for preparing the transcription templates for the positive-sense and negativesense strand ssRNA of the variants of the mL100, m100, S100, E100 and mLLL substrates (primers for cloning are listed in Supplementary Table S2 ).The RNA substrates were prepared with the same procedure as detailed above (transcription template sequences and primers are listed in Supplementary Table S3 ).To prepare the nick-containing variant of the mL100R80 substrate, approximately 25 μM positive-sense strand ssRNA synthesized by IVT was annealed with 25 μM each of the two RNA oligonucleotides (synthesized chemically by Gen-Script) corresponding to the two segments of the negativesense strand ssRNA.
For short 6-FAM-labeled RNA substrates ( ≤44 nt or bp) The 6-FAM-labeled and non-labeled RNA oligonucleotides were synthesized chemically by GenScript.The RNA annealing reaction containing 10 mM Tris-HCl (pH 7.4 at 25 • C), 50 mM KCl, 25 μM 6-FAM-labeled RNA oligonucleotides and 25 μM non-labeled RNA oligonucleotides was carried out on a PCR instrument with the specific annealing program detailed above.For preparing the RNA-DNA hybrid substrates, the non-labeled RNA oligonucleotides were replaced with the corresponding DNA oligonucleotides in the annealing reaction.

RNA cleavage assays
To confirm dsRNA substrates using RNase I and RNase A, 300 ng of RNA substrates were incubated with the indicated amount of RNase I f (New England Biolabs, USA) or RNase A (Thermo Scientific, USA) to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 8.0 at 25 • C), 50 mM NaCl (for RNase I) or 500 mM NaCl (for RNase A), 1 mM DTT, 0.5 mM EDTA].Reactions were performed at 37 • C for 20 min and stopped by adding 0.8 U Proteinase K (New England Biolabs, USA), followed by 15 min incubation at 37 • C. Then 2 μl of 6 × T riT rack DNA loading dye (Thermo Scientific, USA) was added into reaction mixture and reaction samples were analyzed by 1% TAE agarose gel electrophoresis (AGE).The gel was stained with ethidium bromide and RNA was visualized with a UVsolo Touch system (Analytik Jena, Germany).Image processing was conducted using Im-ageJ software.
To investigate the substrate preference of nsp15 using the RNA substrates shown in Figure 1 , 600 ng of substrate RNA was incubated with the indicated concentration of nsp15 or the H234A mutant to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, and indicated concentrations of EDTA, MnCl 2 , MgCl 2 , CaCl 2 , ZnCl 2 and CuCl 2 ].Then, 40 U of murine RNase inhibitor were added in the indicated reactions.Reactions were performed at 37 • C for the indicated time (30 min if not otherwise specified) and then terminated by the addition of 10 μl of 2 × RNA loading dye (New England Biolabs, USA).Reaction samples were heated at 85 • C for 2 min, immediately placed on ice for 2 min, and then analyzed by 1% TAE AGE.The gel was stained with ethidium bromide and the RNA was visualized with a UVsolo Touch system.The image was processed, and grey values of the gel bands corresponding to the full-length RNA substrates were quantified with ImageJ software.
To detect the specific cleavage of dsRNA substrates by nsp15, 300 ng of dsRNA substrates were incubated with the indicated concentration of nsp15 or the H234A mutant (5 nM if not otherwise indicated) to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT and indicated concentrations of EDTA, MnCl 2 , MgCl 2 and CaCl 2 (0.5 mM MnCl 2 if not otherwise indicated)].Reactions were performed at 37 • C for the indicated time (30 min if not otherwise specified) and then inhibited by adding 2 μl of 6 × T riT rack DNA loading dye and placing the reaction samples on ice.Reaction samples were analyzed by 1% TAE AGE (for dsRNA substrate length > 500 bp) or 12% TBE polyacrylamide gel electrophoresis (PAGE) (for dsRNA substrate length ≤ 200 bp).The gel was stained with ethidium bromide and RNA was visualized with a UVsolo Touch system.Image processing was conducted using ImageJ software.
To identify nsp15 cleavage sites using 6-FAM-labeled RNA substrates, 1 μM substrate RNA was incubated with 5 nM nsp15 or H234A mutant to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, and 0.5 mM MnCl 2 ].Reactions were performed at 37 • C for the indicated time (30 min if not otherwise specified) and then terminated by the addition of 10 μl of 2 × RNA loading dye.Reaction samples were heated at 85 • C for 2 min, immediately placed on ice for 2 min, and then analyzed by 20% TBE-urea PAGE (8 M urea).To generate RNA size ladders, alkaline hydrolysis of the corresponding 6-FAM-labeled ssRNA substrates at a concentration of 6 μM to a final volume of 10 μl in an alkaline hydrolysis buffer [50 mM sodium carbonate (pH 9.4) and 1 mM EDTA] was performed for 15 min at 90 • C and quenched with 10 μl of 2 × RNA loading dye.6-FAM-labeled RNA was visualized with a ChemiScope imager (CLiNX, China) using the Cy2 (Ex470BL, Em525 / 30F) channel.Image processing was performed using ImageJ software.
To verify the absence of single-stranded regions in 6-FAMlabeled dsRNA substrates using RNase A, 1 μM substrate RNA was incubated with 0.5 μM RNase A to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 8.0 at 25 • C), 500 mM NaCl, 1 mM DTT, 0.5 mM EDTA].Reactions were performed at 37 • C for 20 min and stopped by adding 0.8 U Proteinase K, followed by 15 min incubation at 37 • C. Then 11 μl of 2 × RNA loading dye was added into reaction mixture.Reaction samples were heated at 85 • C for 2 min, immediately placed on ice for 2 min, and then analyzed by 20% TBE-urea PAGE (8 M urea).6-FAM-labeled RNA was visualized with a ChemiScope imager using the Cy2 (Ex470BL, Em525 / 30F) channel.Image processing was performed using ImageJ software.
To determine the nsp15 cleavage efficiency of various 6-FAM-labeled dsRNA substrates by native PAGE analysis, 1 μM substrate RNA was incubated with the indicated concentration of nsp15 or H234A mutant (20 nM if not otherwise indicated) to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, and 0.5 mM MnCl 2 ].Reactions were performed at 37 • C for 30 min and then inhibited by adding 2 μl of 6 × Tri-Track DNA loading dye and placing the reaction samples on ice.Reaction samples were analyzed by 20% TBE PAGE.6-FAM-labeled RNA was visualized with a ChemiScope imager using the Cy2 (Ex470BL, Em525 / 30F) channel.The image was processed and grey values of the gel bands corresponding to the full-length dsRNA substrates were quantified with ImageJ software.
To determine the nsp15 cleavage efficiency of various 6-FAM-labeled RNA substrates by denaturing PAGE analysis, 1 μM RNA substrates were incubated with (unless otherwise stated) 10 nM nsp15 or H234A mutant to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, and unless otherwise stated, 0.5 mM MnCl 2 ].Reactions were performed at 37 • C for the indicated time (30 min if not otherwise specified) and then terminated by the addition of 10 μl of 2 × RNA loading dye.Reaction samples were heated at 85 • C for 2 min, immediately placed on ice for 2 min, and then analyzed by 20% TBE-urea PAGE (8 M urea).6-FAM-labeled RNA was visualized with a ChemiScope imager using the Cy2 (Ex470BL, Em525 / 30F) channel.The image was processed and grey values of the gel bands corresponding to the full-length RNA substrates were quantified with ImageJ software.
To demonstrate the fragmentation of the long ( ∼1 kb) dsRNA substrates mimicking S AR S-CoV-2 genome replication intermediates by nsp15, 300 ng of dsRNA substrates were incubated with 0, 50, 150 or 450 nM of nsp15 to a final volume of 10 μl in an RNA cleavage buffer [50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT and 0.5 mM MnCl 2 ].Reactions were performed at 37 • C for 1 h and then inhibited by adding 2 μl of 6 × T riT rack DNA loading dye and placing the reaction samples on ice.Reaction samples were analyzed by 2.5% TBE AGE.The gel was stained with ethidium bromide and RNA was visualized with a UVsolo Touch system.Image processing was conducted using ImageJ software.

Differential scanning fluorimetry
25 μl of sample containing 50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, 5 mM EDTA or indicated concentrations of MnCl 2 , 5 × S YPR O Orange dye (Sigma-Aldrich, USA) and 2 μM nsp15 was loaded into a 96-well PCR plate.Differential scanning fluorimetry experiments were performed using a CFX Connect qPCR instrument (Bio-Rad, USA).After an initial equilibration step at 25 • C for 10 min, the temperature was increased by 0.5 • C every 30 s until it reached 95 • C. Fluorescence intensity was measured after each cycle using the FRET channel.The final data were analyzed using Bio-Rad CFX Maestro software.Each sample produced a melt curve, and the first derivative of the melt curve produced a peak, which provided the melting temperature ( T m ).

Evaluation of dsRNA thermodynamic stability by denaturing PAGE analysis
First, 10 μl of 2 × RNA loading dye was added to 10 μl of a sample containing 50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, 0.5 mM MnCl 2 and 1 μM 6-FAMlabeled substrate RNA.The sample was heated at 85 • C for 2 min, immediately placed on ice for 2 min, and then analyzed by 20% TBE-urea PAGE (8 M urea).6-FAM-labeled RNA was visualized with a ChemiScope imager using the Cy2 (Ex470BL, Em525 / 30F) channel.Image processing was performed using ImageJ software.

SARS-CoV-2 nsp15 preferentially degrades dsRNA with sequences from the SARS-CoV-2 genome
Most prior studies on the biochemical properties of coronavirus nsp15 employed short RNA substrates ranging from a few to tens of nucleotides, which might not present full recognition elements for nsp15.To closely mimic the naturally occurring RNA during coronaviral genome replication and transcription, we designed an approximately 1-kb chimeric RNA sequence to mimic the S AR S-CoV-2 genome by splicing three distinguishing regions from the S AR S-CoV-2 genome (Figure 1 A; Supplementary Table S3 ).This chimeric S AR S-CoV-2 mini-genome contains multiple characteristic sequences.The 5 and 3 untranslated regions (UTRs) of the coronavirus genomes have conserved secondary structures and play a crucial role in genome replication and transcription ( 37 ).The transcription regulatory sequences (TRSs) are located upstream to most open reading frames (ORFs) and direct the leader-body junction during synthesis of coronavirus subgenomic RNA ( 6 ).The consensus TRS core of S AR S-CoV-2 is 5 -A CGAA C-3 , but the sequence upstream of ORF6 and the sequence upstream of ORF10 lack the intact 5 -A CGAA C-3 sequence.The S AR S-CoV-2 mini-genome also contains partial ORF sequences neighboring these characteristic sequences to minimize the disruption to the integrity of the characteristic sequences.Moreover, the S AR S-CoV-2 mini-genome includes the UA and CA sequences proposed by Ancar et al. ( 22 ).We aimed to determine whether S AR S-CoV-2 nsp15 specifically cleaves certain sequences in the S AR S-CoV-2 mini-genome and identify any differences in nsp15 cleavage of the positivesense and negative-sense strand RNA and dsRNA.The S AR S-CoV-2 mini-genome dsRNA was prepared by in vitro transcription (IVT) of the positive-sense and negative-sense strand RNA, following by annealing both strands into dsRNA.To prevent interference from dsRNA by-products generated during in vitro transcription, we utilized VSW-3 RNA polymerase ( 36 ), which has previously been demonstrated to effectively reduce these by-products.To confirm that the annealed 1-kb product was dsRNA and void of single-stranded regions or loops, we used RNase I and RNase A for examination.RNase I preferentially degrades ssRNA in the absence of calcium ( 38 ), while RNase A specifically degrades ssRNA at NaCl concentrations of 0.3 M or higher ( 39 ).Under reaction conditions containing 0.5 mM EDTA, the annealed product showed minimal cleavage while the positive-sense strand RNA was completely cleaved by RNase I ( Supplementary Figure S1 A, left panel).Similarly, under reaction conditions containing 500 mM NaCl, the annealed product showed almost no cleavage while the negative-sense strand RNA was completely cleaved by RNase A ( Supplementary Figure S1 A, right panel).These results confirmed that the annealed product was dsRNA.Furthermore, to exclude the effect of potential nuclease contamination, we purified both the wild-type S AR S-CoV-2 nsp15 with a 6 × His tag at the N-terminus and its active-site mutant, H234A, as a control for the cleavage reaction, by metal affinity chromatography and gel filtration chromatography ( Supplementary Figure S1 B).
Previous studies reported that nsp15 required Mn 2+ for optimal EndoU activity ( 26 , 27 , 40 , 41 ) and a Mn 2+ concentration of 5 mM was commonly used.As 5 mM Mn 2+ is physiologically irrelevant ( 42 ,43 ), we conducted the cleavage reaction at Mn 2+ concentrations of 0.5 mM and zero in addition to 5 mM.Surprisingly, we observed that nsp15 efficiently cleaved both ssRNA and dsRNA substrates at 5 mM Mn 2+ , whereas nsp15 cleaved the dsRNA substrates significantly more efficiently than the ssRNA substrates at Mn 2+ concentrations of 0.5 mM and zero (Figure 1 B; Supplementary Figure S1 C, D).When approximately 60% of the full-length dsRNA substrates were cleaved, only around 20% of the full-length ss-RNA substrates were cleaved under the latter two conditions, demonstrating a 3-fold difference in cleavage efficiency.To verify the altered substrate preference of nsp15 under different Mn 2+ concentrations and eliminate the possible effect of nsp15 concentration changes, we conducted the cleavage reaction with the same nsp15 concentration but with various reaction times at various Mn 2+ concentrations to facilitate cleavage of dsRNA substrates to a similar extent.We found that the cleavage preference of nsp15 for dsRNA was weakened as the Mn 2+ concentration increased beyond 0.5 mM ( Supplementary Figure S1 E).In contrast, no significant change occurred in the cleavage preference of nsp15 for dsRNA when the Mn 2+ concentration was lower than 0.5 mM ( Supplementary Figure S1 F).However, no significant differences were observed when examining the RNA products produced by nsp15 cleavage in the absence and presence of 5 mM Mn 2+ (Figure 1 C).The 1% TAE agarose gel electrophoresis results indicated that nsp15 cleaved both positive-sense and negative-sense strand RNA without any apparent site preference, as shown by the smear in the gel.The gel electrophoresis results of the cleavage products of the S AR S-CoV-2 minigenome dsRNA substrate showed prominent bands (indicated by black pentagrams in Figure 1 C) with mobilities close to those of the 500-bp DNA marker, suggesting that nsp15 displays a cleavage preference for certain sites in the middle of this dsRNA substrate.Given the presence of a TRS core sequence (TRS-CS) in the middle of the S AR S-CoV-2 minigenome RNA (Figure 1 A), we investigated whether nsp15 preferentially cleaved this TRS-CS in the form of dsRNA.We prepared a variant of the S AR S-CoV-2 mini-genome dsRNA substrate with this TRS-CS deleted ( Supplementary Table S3 ) and found that specific cleavage of this variant by nsp15 was also detected, with no significant difference in the cleavage efficiency between this variant and the original S AR S-CoV-2 mini-genome dsRNA substrate (Figure 1 D), suggesting that the cleavage preference of nsp15 is TRS-CS-independent.
To investigate whether nsp15 had a similar cleavage preference for dsRNA as other sequences, we extracted two sequences from either the S AR S-CoV-2, E. coli , or Thermus aquaticus ( Taq ) genome (Figure 1 E; Supplementary Table S3 ), and prepared ssRNA and dsRNA substrates by the same method as previously described.Results from the cleavage of these substrates by nsp15 showed a minimal variance in the nsp15 cleavage efficiency of dsRNA substrates derived from the same genome, but a significant difference in the cleavage efficiency of dsRNA substrates derived from different genomes; however, there was no significant difference in the cleavage efficiency of ssRNA substrates derived from different genomes (Figure 1 F).We discovered that one of the most significant differences among various genome-derived RNA substrates consists in their AU content (Figure 1 A,E), with S AR S-CoV-2 RNA possessing a significantly higher AU content ( > 60%) than that of the RNA from other genomes, signifying that the cleavage preference for dsRNA of nsp15 is positively related to the AU content of RNA substrates.The AU distribution and RNA structure of these dsRNA sub-strates may also affect the cleavage preference of nsp15.When nsp15 cleaved these six groups of RNA substrates, its substrate preference was altered under 5 mM Mn 2+ conditions ( Supplementary Figure S1 G), which is consistent with the cleavage of the S AR S-CoV-2 mini-genome RNA substrates by nsp15 (Figure 1 B), indicating that this phenomenon was unrelated to specific RNA sequence.We investigated the impact of varying Mn 2+ concentrations on nsp15 stability by differential scanning fluorimetry.The results indicated that nsp15 stability remained consistent across different Mn 2+ concentrations ( Supplementary Figure S1 H), which agrees with structural studies reporting no metal-binding site of S AR S-CoV-2 nsp15 ( 12 , 13 , 33 , 44 ).It has been reported that the folding of RNA into stable tertiary structures is highly sensitive to the concentration and type of cations ( 45 ).Therefore, it is possible that the high concentration of Mn 2+ altered the ssRNA structure and the interaction between nsp15 and ssRNA, thereby increasing the cleavage efficiency of nsp15 on ssRNA.
We also examined the influence of four additional divalent metal ions on the cleavage activity of nsp15 at a concentration of 0.5 mM.It was found that Mg 2+ and Ca 2+ also slightly increased the cleavage activity of nsp15, which agrees with previous studies ( 26 , 27 , 40 ), whereas Zn 2+ and Cu 2+ inhibited the cleavage activity of nsp15, which has not been reported before ( Supplementary Figure S1 I).Moreover, Mg 2+ and Ca 2+ did not alter the cleavage preference of nsp15 for dsRNA ( Supplementary Figure S1 I), as well as the cleavage preference of nsp15 for certain sites in the middle of S AR S-CoV-2 mini genome dsRNA substrate ( Supplementary Figure S1 J).A Mn 2+ concentration of 0.5 mM was used in most of the subsequent RNA cleavage experiments.

Identification of the cleavage sites of SARS-CoV-2 nsp15 on the dsRNA substrates derived from the SARS-CoV-2 mini-genome
To locate the preferred cleavage sites of nsp15 on the 967bp S AR S-CoV-2 mini-genome dsRNA (hereinafter O967), we created several variants of this substrate by deleting specific sequences from the original substrate (Figure 2 A, top; Supplementary Table S3 ).Based on the size of the fragments after the cleavage of O967 by nsp15, we deduced that the preferred cleavage site was in the middle region of O967.We selected the middle 200 bp of O967 and divided it into left and right halves (Figure 2 A, colored in blue and yellow, respectively).First, the middle 200-bp region was deleted from O967 to produce substrate dm200.When dm200 was treated with nsp15, the prominent product bands (marked by black pentagrams in Figure 2 B) indicating specific cleavage products disappeared.Although certain nonspecific cleavage still occurred, as indicated by the smear in the gel, the overall cleavage efficiency significantly decreased with dm200.However, when the left or right half 100 bp of the middle 200-bp region was deleted, as in the dmL100 and dmR100 substrates, cleavage by nsp15 still produced specific fragments, as revealed by the prominent product bands (indicated by black pentagrams in Figure 2 B), suggesting the existence of at least one nsp15 preferred cleavage site within both the left and right halves of this 200-bp region.
After that, we prepared shorter dsRNA substrates to focus on the middle 200-bp region ( Supplementary Figure S2 A; Supplementary Table S3 ).Here, the corresponding ssRNA for annealing and construction of dsRNA substrates were synthe-sized utilizing the T7 RNA polymerase S43Y mutant, which was previously found to reduce undesired termination in runoff RNA synthesis and produce RNA with higher terminal homogeneity ( 35 ).Specific cleavage by nsp15 was observed for the m200 (entire middle 200 bp as described above), mL100 (left half of m200) and mR100 (right half of m200) substrates, but not the m100 (middle 100 bp of m200) substrate ( Supplementary Figure S2 B).These results suggested that the preferred cleavage sites for nsp15 are located in both the 386-485-bp (mL100) and 486-585-bp (mR100) regions of the O967 substrate.The cleavage of the 100-bp mL100 substrate produced two sets of fragments, approximately 60 bp and 40 bp in size.Similarly, cleavage of the 100-bp mR100 substrate yielded two sets of fragments, approximately 55 bp and 45 bp in size (Figure 2 C).The sizes of the cleaved fragments indicated two possible cleavage sites in each substrate.To further locate the cleavage sites for nsp15, we prepared dsRNA substrates that corresponded to 80 bp on the left and 80 bp on the right of the mL100 and mR100 substrates (Figure 2 A, bottom; Supplementary Table S3 ), respectively.After analyzing the sizes of the cleavage products of these substrates (Figure 2 C) and comparing them with those from the mL100 and mR100 substrates, we estimated the cleavage sites for nsp15 to be within the 441-450-bp sequence (designated as Lpca, Figure 2 A, colored in green) and 526-535-bp sequence (designated as Rpca, Figure 2 A, colored in green) of the O967 substrate.To confirm that the originally observed specific cleavage by nsp15 on the O967 substrate occurred in the Lpca and Rpca, we prepared three variants of the O967 substrate with either the Lpca or Rpca or both removed (Figure 2 A, top; Supplementary Table S3 ).Deletion of the Lpca or Rpca still maintained the specific cleavage of the O967 substrate by nsp15, while deletion of both abolished the specific cleavage (Figure 2 B), confirming that the preferred cleavage sites for nsp15 were indeed located in the Lpca and Rpca.
Interestingly, although the m100 substrate contained both an Lpca and Rpca ( Supplementary Figure S2 A), nsp15 did not demonstrate preferential cleavage of this substrate ( Supplementary Figure S2 B).Given that both the Lpca and Rpca were close to the end of the m100 substrate, we hypothesized that flanking sequences with adequate length on both sides of the Lpca or Rpca were also required for cleavage by nsp15.We prepared six dsRNA substrates of identical length, all including an Lpca but with various distances from the Lpca to the end of the dsRNA ( Supplementary Figure S2 A; Supplementary Table S3 ).Results from these substrates showed that as the Lpca approached the end of the dsRNA in either direction, cleavage of Lpca by nsp15 was weakened.Furthermore, the preferential cleavage of Lpca by nsp15 is no longer observed when Lpca was located 5 bp away from the end of the dsRNA ( Supplementary Figure S2 C).These results suggested that the lengths of flanking sequences on either side of the preferred cleavage sites of nsp15 affected the cleavage efficiency of nsp15.
To locate the exact position of the specific cleavage on the O967 dsRNA substrate by nsp15, we employed short 6-FAMlabeled dsRNA substrates, which were formed by annealing chemically synthesized RNA oligonucleotides (Figure 2 A, bottom).These dsRNA substrates contained either an Lpca or Rpca and had adequate flanking sequences to ensure the preferential cleavage of Lpca and Rpca by nsp15.The cleavage products of these dsRNA substrates by nsp15 were analyzed by denaturing PAGE, and the precise sizes of the cleavage products were determined using the alkaline hydrolysis products of the corresponding ssRNA substrates as ladders.Three to four sites with the highest cleavage efficiencies on every strand of the dsRNA are indicated in Figure 2 A and D, signifying that nsp15 cleaved at the 3 -side of uridylate in a stretch of consecutive Us.To confirm that cleavage of these sites by nsp15 occurred in the double-stranded region, we validated these dsRNA substrates using RNase A, which was applied to detect single-base mismatches in dsRNA and RNA-DNA hybrid ( 46 ,47 ).The RNase A cleavage experiments suggested the absence of single-stranded region in these dsRNA substrates ( Supplementary Figure S2 D).Interestingly, we found that a ss-RNA with the same sequence as one of the two strands of the above dsRNA was cleaved by nsp15 in a completely different manner.Nsp15 cleaved almost all U sites and a few C sites in these ssRNA substrates, although the cleavage efficiencies at these sites varied ( Supplementary Figure S2 G, H).The cleavage at U and C sites in ssRNA by nsp15 was reported previously ( 28 , 29 , 33 ).However, the cleavage at consecutive Us in dsRNA observed in this study was much more specific and efficient.
In the above assays, the dsRNA substrates were labeled with 6-FAM at the 5 terminus.To obtain a full observation of the cleavage by nsp15, the dsRNA substrates containing an Lpca as described above were also labeled with 6-FAM at the 3 terminus ( Supplementary Figure S2 E).The preferred cleavage sites for nsp15 deduced on these substrates were consistent with those identified on the 5 -labeled dsRNA substrates.
Although the results based on the 5 -labeled dsRNA substrates demonstrated that the most efficiently cleaved site was U24, the results based on the 3 -labeled dsRNA substrates showed that the most efficiently cleaved site was U25 (Figure 2 A, D; Supplementary S2 E, F), indicating that sequential cleavages might occur in adjacent U sites.Notably, the alkaline hydrolysis products of the ssRNA substrates labeled with 6-FAM at the 3 terminus did not form standard ladders in the smallsized (1-6 nt) range ( Supplementary Figure S2 F).Therefore, the sizes of the cleavage products from these 3 -labeled dsRNA substrates were assigned based on both the alkaline hydrolysis ladder and the cleavage products of the corresponding ssRNA substrates.

dsRNA cleavage by SARS-CoV-2 nsp15 is sensitive to AU arrangement
Both the Lpca and Rpca were located in areas highly rich in AU content (indicated by orange dashed boxes in Figure 2 A).We refer to such areas exceeding 10 bp in length and containing at most one GC base pair as AU-rich areas.We investigated the effect of the AU-rich areas on nsp15 cleavage by replacing various sequences in the Lpca-containing mL100 substrate with GC-rich sequences ( Supplementary Figure S3 A; Supplementary Table S3 ).When the sequences out of the AU-rich area were substituted by GC-rich sequences, cleavage by nsp15 was not affected ( Supplementary Figure S3 B, top).However, when the AU-rich sequences adjacent to the 10-bp core sequence were replaced with GC-rich sequences, we found that not only was the cleavage efficiency of nsp15 reduced, but also the cleavage sites shifted ( Supplementary Figure S3 B, bottom).As judged by the sizes of the cleavage products ( Supplementary Figure S3 B, bottom), when the sequences on the left side of the AU-rich area were replaced by GC-rich sequences, the cleavage sites shifted toward the right, and vice versa.When the 10-bp or 20-bp AU-rich sequences containing nsp15 preferred cleavage sites from the mL100 substrate were inserted into two other 100-bp dsRNA substrates, they were cleaved similarly by nsp15 ( Supplementary Figure S3 C, D; Supplementary Table S3 ).Despite that the cleavage sites were all within the 10-bp region (Figure 2 A), the 20-bp sequences produced stronger cleavages than the 10bp sequences ( Supplementary Figure S3 D).These results suggested that the AU-rich sequences adjacent to the nsp15 cleavage sites facilitated its cleavage.A more quantitative demonstration of the effect of AU-rich sequences around the cleavage sites was shown by short 6-FAM-labeled dsRNA substrates, the −F-L substrate and its variants (Figure 3 A).GC replacement outside of the 20-bp AU-rich area showed no significant effect on cleavage within the 10-bp core sequence (Figure 3 B).However, GC replacement within the 20-bp AU-rich area and adjacent to the 10-bp core sequence reduced the cleavage of nsp15 significantly (Figure 3 B, C).
We also investigated the effect of GC replacement within the 10-bp core sequence (Figure 3 D; Supplementary Table S3 ).When the AU sequences on either side were substituted by GC sequences, cleavage by nsp15 decreased, and the cleavage sites were 'pushed' toward the other side of the 10-bp core sequence, as judged by the sizes of the cleavage products (Figure 3 E).These results suggested that nsp15 preferentially cleaved AU-rich areas in dsRNA.
The observed cleavage sites indicated that nsp15 preferred to cleave consecutive Us rather than individual Us in dsRNA (Figure 2 A, D).To investigate this property of nsp15, we created the mL100 substrate variants 1U, 2U and 3U containing no consecutive Us, 2-nt consecutive Us, and 3-nt consecutive Us in the middle of the AU-rich area, respectively (Figure 3 F; Supplementary Table S3 ).The cleavage of 1U by nsp15 was very weak, while the cleavage efficiency of 3U was significantly higher than that of 2U (Figure 3 G).Moreover, stronger cleavage was observed on substrates containing longer consecutive U sequences ( Supplementary Figure S3 E, F; Supplementary Table S3 ).When an AU-rich area containing consecutive Us ≥3 nt in both strands was introduced into another substrate (m100 substrate) by replacing two CG base pairs with UA base pairs (Figure 3 H; Supplementary Table S3 ), nsp15 cleaved the variant much more efficiently than the original substrate (Figure 3 I).
dsRNA thermodynamic stability and length modulate the cleavage of SARS-CoV-2 nsp15 S AR S-CoV-2 nsp15 has been reported to cleave dsRNA via a base-flipping mechanism in two studies ( 28 ,32 ).It has also been shown that when ssRNA forms secondary structures, nsp15 prefers to cleave flexible Us located in loops ( 29 ,30 ).The AU-rich areas in dsRNA may have a more relaxed structure, as AU base pairs are less thermodynamically stable than GC base pairs, which may facilitate U-flipping that nsp15 can recognize and cleave.We prepared various 42-bp substrates with different AU distributions or content (Figures 2 A  and 4A ) to examine their thermodynamic stability.The +F-42.1 and −F-42.1 substrates had the same AU content as the +F-L and −F-L substrates, but the AU distribution was more scattered, whereas the +F-42.2and −F-42.2substrates had a significantly lower AU content.We initially analyzed the thermodynamic stability of these substrates using denaturing PAGE (Figure 4 B).The results showed that the +F-L and −F-L substrates exhibited the lowest thermodynamic stability and completely denatured in denaturing PAGE.In contrast, the +F-42.2and −F-42.2substrates demonstrated the highest thermodynamic stability and resisted the denaturation condition.The +F-42.1 and −F-42.1 substrates exhibited partial denaturation in denaturing PAGE, suggesting intermediate thermodynamic stability.To further demonstrate the differences in the thermodynamic stability between these dsRNA substrates, we conducted a thermal melting analysis on these dsRNA substrates (unlabeled, designated as L, 42.1 and 42.2) using EvaGreen dye (Figure 4 C).The melt peak results indicated that the L substrate had a melting temperature ( T m ) of 74.5 • C, the 42.1 substrate had a T m of 79 • C, and the 42.2 substrate had a T m of 91 • C. Consistently, nsp15 cleaved the −F-L substrates more efficiently than the −F-42.1 substrates, and only displayed minimal cleavage of the −F-42.2substrates under identical reaction conditions (Figure 4 D; Supplementary Figure S4 A), suggesting that low thermodynamic stability of consecutive AU base pairs in dsRNA substrates facilitated cleavage by nsp15.Notably, results obtained on these short RNA substrates revealed that nsp15 cleaved all ssRNA substrates with a similar efficiency, unrelated to their AU distribution or content ( Supplementary Figure S4 A).
To further evaluate the effect of the thermodynamic stability of the dsRNA substrate on its cleavage by nsp15, we created variants of the +F-42.1 and −F-42.1 substrates by replacing a G in the middle with an A to introduce a singlebp AC mismatch (Figure 4 A).Compared with the partial  denaturing of the original dsRNA, denaturing PAGE analysis revealed that the dsRNA substrates harboring the single mismatch were denatured completely (Figure 4 B).Thermal melting analysis also demonstrated a reduction in the thermodynamic stability of dsRNA by the mismatch (Figure 4 C).Consistent with our hypothesis, the mismatch enhanced the cleavage efficiency of nsp15 ( Supplementary Figure S4 B), and nsp15 showed a strong preference to cleave sites close to the mismatch (Figure 4 E).On the +F-42.1 substrate, nsp15 did not cleave the U21 site adjacent to the GC base pair; however, with the introduction of the AC mismatch in the +F-42.1vsubstrate, nsp15 cleaved the U21 site adjacent to the mismatch position, and cleavage at U22 and U16 sites was also enhanced (Figure 4 E).On the −F-42.1 substrate, nsp15 cleaved U26 most efficiently.When the GC base pair was substituted by an AC mismatch in the −F-42.1vsubstrate, the strongest cleavage switched to C23, right in the mismatch.These results suggested that the mismatch reduced the thermodynamic stability of dsRNA to facilitate base-flipping within the area.It is noteworthy that although A23 was also in the mismatch, no cleavage was observed (Figure 4 E), verifying that nsp15 only recognized pyrimidines for cleavage.
We previously found that as the Lpca approached the end of the dsRNA, its cleavage by nsp15 diminished ( Supplementary Figure S2 C).Additionally, a previous structural study of the nsp15-dsRNA complex reported that dsRNA of approximately 35 bp was sufficient to interact across the nsp15 hexamer and the EndoU domain interacts with the middle of the dsRNA ( 28 ).Hence, we hypothesized that an adequate length of dsRNA on both sides of the cleavage sites was necessary for binding to nsp15 and for localization of the cleavage sites into the active site of nsp15.We designed variants of the −F-L substrate by lengthening or shortening the flanking sequences on both sides of the cleavage sites (Figure 4 F).The substrates +F-L30 and −F-L30 with the 11bp shortest flanking sequences were still cleaved specifically by nsp15 (Figure 4 G).However, the cleavage efficiency increased with an extension of the flanking sequences, and reached a maximum with a 40-bp substrate (Figure 4 H; Supplementary Figure S4 C), suggesting that approximately 16-bp flanking sequences at both sides were adequate to provide optimal binding to nsp15.In contrast, the nsp15 cleavage efficiency of the corresponding ssRNA substrates did not vary similarly ( Supplementary Figure S4 C), confirming that nsp15 bound to ssRNA or dsRNA using different modes.To examine the disparity in the ability of nsp15 to bind ss-RNA and dsRNA, we performed an electrophoretic mobility shift assay, using the active-site mutant of nsp15, H234A, which was used in the structural study of the S AR S-CoV-2 nsp15 dsRNA complex ( 28 ).However, this assay did not reveal significant binding of H234A to either ssRNA or dsRNA ( Supplementary Figure S4 D).This indicates that the electrophoretic mobility shift assay might not be sensitive enough for detecting the interactions between nsp15 and RNA, which is likely weak or transient.Additionally, our attempts to study the binding of nsp15 to RNA using wild-type nsp15 and modified RNA substrates were unsuccessful.The phosphorothioate modification on RNA attenuated but not block nsp15 cleavage ( Supplementary Figure S4 E).Nsp15 was unable to cleave 2 -methoxy-modified RNA, but the electrophoretic mobility shift assay also failed to detect the binding of wild-type nsp15 to this RNA ( Supplementary Figure S4 F).

SARS-CoV-2 nsp15 is a dsRNA nickase
Although we observed dsRNA breaks resulting from the cleavage by nsp15, the cleavage efficiency for every strand of the dsRNA often varied (Figures 2 A and 5A , H; Supplementary Figure S2 E).Moreover, we observed one or two additional bands above the full-length substrate band in the native PAGE analysis of certain cleavage reactions (Figure 5 B; Supplementary Figure S5 A,B; Supplementary Table S3 ), and these bands were still observed after removal of nsp15 by an RNA purification kit (Figure 5 B), indicating that the additional bands did not result from protein binding.We hypothesized that nsp15 may cleave dsRNA in a nicking manner and the bands above the full-length substrate band were formed due to dsRNA nicking.To test this hypothesis, we created two variants of the 40LR −15 substrate ( Supplementary Figure S3 A), in which the consecutive Us were only present on one strand within the AU-rich area (Figure 5 C; Supplementary Table S3 ).In contrast to the original 40LR −15 substrate, double-strand breaks were not observed for these two variants.Instead, formation of gel bands above the substrate bands were observed (Figure 5 D).We also created a variant of the mL100R80 substrate (Figure 2 A) containing a pre-existing nick in one strand by annealing three corresponding ssRNA (Figure 5 E).We found that this nicked variant mL100R80v was more susceptible to break by nsp15 than the mL100R80 substrate (Figure 5 F).Furthermore, the cleavage products of the mL100R80v variant formed a band above the full-length substrate band, indicating that the prenicked strand of the variant was further cleaved by nsp15, which was consistent with observations on the cleavage of short dsRNA substrates labeled with 6-FAM (Figure 2 A, D; Supplementary Figure S2 E,F).All of these results support that nsp15 cleaved dsRNA in a nicking manner.The bands above the full-length substrate band represent the cleavage products in which only one strand is cleaved by nsp15, and the observed dsRNA breaks were due to close nicks in both strands.Despite being an efficient dsRNA nickase, nsp15 was not able to cleave the RNA strand in the RNA-DNA hybrid under identical reaction conditions (Figure 5 G, H).

Discussion
We found that dsRNA length and thermodynamic stability modulated the dsRNA cleavage efficiency of S AR S-CoV-2 nsp15 (Figure 4 ).First, it was important that the dsRNA was of sufficient length to ensure optimal binding to the nsp15 hexamer and to localize the nsp15 preferred cleavage sites to the active site pocket of nsp15.Furthermore, pyrimidines located in less thermodynamically stable areas in dsRNA, such as AU-rich areas and mismatch-containing areas, are more prone to flipping, which allowed for recognition by nsp15 and subsequent cleavage.However, nsp15 accessed most pyrimidines in ssRNA directly without base flipping, which may partially explain the discrepancy between the ssRNA and dsRNA cleavage by nsp15.The thermodynamic stability of ssRNA was mainly determined by its secondary structure, and was not directly correlated with its AU content or AU distribution.Moreover, ssRNA did not exhibit specific binding to the nsp15 hexamer compared to dsRNA.Hence, as the AU content or RNA length decreased, nsp15 exhibited a noticeable decline in cleavage efficiency of dsRNA substrates, but not of the corresponding ssRNA substrates ( Supplementary Figure S4 A,C).Interestingly, when nsp15 cleaved short RNA substrates, its substrate preference was not altered under 5 mM Mn 2+ conditions ( Supplementary Figure S4 A).The inconsistent findings in previous studies on nsp15 substrate preference ( 26-28 ) might be due to the different RNA substrates used.We examined the RNA substrates used in these studies and noticed that the only previous study reporting that nsp15 cleaved dsRNA more efficiently ( 26 ) had used a long dsRNA substrate of 1 kb and a dsRNA substrate with a single base mismatch, supporting our points.Overall, our study provided sufficient insights into the factors of RNA substrates that determine the substrate preference of coronavirus nsp15.
Coronavirus infection results in the formation of perinuclear double-membrane structures derived from endoplasmic reticulum, including double-membrane vesicles (DMVs), convoluted membranes, and double-membrane spherules, which make up the replication organelles where the viral RTC locates and viral RNA synthesis occurs ( 6 ).Coronavirus dsRNA, commonly considered as genome replication intermediates, mainly localizes to the DMV interior, a process that is thought to evade detection by cytosolic innate immune sensors (48)(49)(50)(51).Previous studies reported that coronavirus nsp15 localized with viral RTCs and replicating viral RNA ( 16 ,52-54 ), suggesting that nsp15 interacted with viral dsRNA intermediates.Furthermore, compared to wild-type virus infection, EndoU-deficient virus infection resulted in increased cytosolic dsRNA ( 17 ) or yielded more dsRNA that did not localize with the viral RTC during the early phase ( 16 ), thus activating host cell dsRNA sensors, such as Mda5, PKR and OAS ( 16 ,17 ).This suggests that early during EndoU-deficient virus infection, viral dsRNA intermediates are released to the DMV exterior and detected by cytosolic dsRNA sensors, whereas in wild-type virus infection, nsp15 prevents this from happening using its EndoU activity.Based on our findings, we developed a model showing how nsp15 mediates this process (Figure 6 E).
We found that S AR S-CoV-2 nsp15 preferentially degraded dsRNA with sequences from the S AR S-CoV-2 genome, which has a high AU content (62%) (Figure 1 ).It was determined that S AR S-CoV-2 nsp15 preferentially cleaved consecutive Us in AU-rich areas of dsRNA (Figures 2 and 3 ) via its dsRNA nickase activity (Figure 5 ).Interestingly, we found that coronavirus genomes generally have a high AU content, ranging from 55% to 68%, with an average value of 61% (Figure 6 A; Supplementary Table S4 ).This indicates the potential physiological significance of nsp15 to directly degrade coronavi-   ral dsRNA intermediates.It was reported that OAS 3 selectively binds long dsRNA ( > 50 bp) ( 55 ), PKR prefers to dimerize upon binding to a similar sized dsRNA ( > 60 bp) ( 56 ,57 ), and Mda5 is most efficiently activated by even longer dsRNA ( > 2 kb) ( 58 ).Based on the coronavirus genome size of approximately 30 000 nucleotides, unguarded or unprocessed dsRNA intermediates have the ability to strongly activate these dsRNA sensors.Our study showed that nsp15 cleaved dsRNA ≥40 bp with optimal activity (Figure 4 H; Supplementary Figure S4 C), and we demonstrated that nsp15 was able to cleave long dsRNA with a high AU content into fragments shorter than 50 bp (Figure 6 B).Therefore, we hypothesized that after processing by the nsp15 dsRNA nickase, the degradation products of viral dsRNA intermediates are not able to efficiently activate the dsRNA sensors, even if they are released from the DMVs.A previous study reported that the 5 -polyU of the coronavirus negative-sense strand RNA acted as an Mda5-dependent PAMP ( 21 ).Interestingly, in our study, we found that nsp15 did not preferentially target consecutive Us when cleaving ssRNA ( Supplementary Figure S2 H).However, we demonstrated that nsp15 cleaved the dsRNA form of a 39-bp 5 -polyU-containing sequence derived from the negative-sense strand of S AR S-CoV-2 genome with significantly higher cleavage efficiency compared to cleaving the corresponding ssRNA (Figure 6 C, D).These results suggested that nsp15 is more likely to remove the 5 -polyU of negative-sense strand RNA by directly cleaving the dsRNA intermediate.Overall, our work supported the mechanism that coronaviruses evade the antiviral response mediated by host cell dsRNA sensors by using nsp15 dsRNA nickase to directly cleave dsRNA intermediates formed during genome replication and transcription (Figure 6 E).

Figure 1 .
Figure 1.SARS-CoV-2 nsp15 preferentially degrades dsRNA with sequences from the SARS-CoV-2 genome.( A ) Schematic showing the SARS-CoV-2 mini-genome construction.1-265, 27 137-27 447 and 29 480-29 870 indicate the corresponding sequence ranges in the SARS-CoV-2 genome.+, − and ds refer to positive-sense ssRNA, negative-sense ssRNA, and dsRNA, respectively; TRS-CS, the core sequence of the transcription regulatory sequence; 5 UTR (no cap), 5 untranslated region without a 5 cap; −M, right part of the ORF encoding the membrane protein; ORF6, the sixth ORF of the SARS-CoV-2 genome; ORF7a −, left part of the seventh ORF of the SARS-CoV-2 genome; −N, right part of the ORF encoding the nucleocapsid protein; ORF10, the last ORF of the SARS-CoV-2 genome; 3 UTR (no polyA), 3 untranslated region without the polyA tail.( B ) Clea v age of the ssRNA and dsRNA substrates related to the SARS-CoV-2 mini-genome (as shown in A) by nsp15 in the presence of 5 mM Mn 2+ , 0.5 mM Mn 2+ or 5 mM EDTA with various reaction times.Remaining full-length RNA substrates after nsp15 clea v age w ere quantified as % FL.T he full gel images are sho wn in Supplementary Figure S6 A. ( C ) Clea v age of the ssRNA and dsRNA substrates related to the SARS-CoV-2 mini-genome by various concentrations of nsp15 in the presence of 5 mM EDTA or 5 mM Mn 2+ (0, 15, 30, 60 and 120 nM nsp15 for 5 mM EDTA; 0.125, 0.25 and 0.5 nM nsp15 for 5 mM Mn 2+ ).( D ) Clea v age of the SAR S-CoV-2 mini-genome dsRNA substrates with or without TR S-CS b y nsp15.In (C and D), the prominent gel bands indicating specific clea v age are mark ed with blac k pentagrams.( E ) Sc hematic representation of RNA substrates 1-6.+, − and ds refer to positive-sense ssRNA, negative-sense ssRNA, and dsRNA, respectively.( F ) Cleavage of RNA substrates 1-6 by nsp15 in the presence of 0.5 mM Mn 2+ .The nsp15 H234A mutant was used as a negative control.Reduction of the full-length RNA substrates by nsp15 cleavage was quantified as % FL.The average and standard deviation for at least two independent reactions are graphed.The full gel images are shown in Supplementary Figure S6 F. All samples were analyz ed b y 1% TAE AGE (nativ e gel).
20 μl of sample containing 50 mM Tris-HCl (pH 7.4 at 25 • C), 140 mM KCl, 1 mM DTT, 0.5 mM MnCl 2 , 1 × EvaGreen dye (Biotium, USA) and 10 μM dsRNA was loaded into a 96-well PCR plate.Thermal melting experiments were performed using a CFX Connect qPCR instrument.After an initial equilibration step at 25 • C for 3 min, the temperature was increased by 0.5 • C every 15 s until it reached 95 • C. Fluorescence intensity was measured after each cycle using the SYBR Green channel.The final data were analyzed using Bio-Rad CFX Maestro software.Each sample produced a melt curve, and the first derivative of the melt curve produced a peak, which provided the melting temperature ( T m ).

Figure 2 .
Figure 2. Identification of the clea v age sites of SARS-CoV-2 nsp15 on the dsRNA substrates derived from the SARS-CoV-2 mini-genome.( A ) Schematic showing the dsRNA substrates derived from the SARS-CoV-2 mini-genome (the original 967-bp dsRNA substrate is designated as O967).Three to four sites identified with the strongest clea v age in e v ery strand of the 6-FAM-labeled dsRNA substrates are marked with blue triangles, and the cleavage efficiency is indicated by the height of the triangle.The AU-rich areas containing these sites are marked by orange dashed boxes.Lpca, left preferred clea v age area; Rpca, right preferred clea v age area.( B ) Clea v age of the O967 substrate and its v ariants b y nsp15.Samples w ere analyz ed b y 1% TAE AGE (nativ e gel).T he prominent gel bands indicating specific clea v age are mark ed b y black pentagrams.( C ) Clea v age of the mL1 00 and mR1 00 substrates and their 80-bp variants by nsp15.Samples were analyzed by 12% TBE PAGE (native gel).The approximate sizes of the cleavage products are labeled.( D ) Identification of the clea v age sites of nsp15 in the Lpca-containing and Rpca-containing short dsRNA substrates labeled with 6-FAM at the 5 terminus.Samples were analyzed by 20% TBE-urea PAGE (denaturing gel).The three or four strongest cleavage sites in every dsRNA substrate are noted.

Figure 3 .
Figure 3. dsRNA clea v age b y SAR S-CoV-2 nsp15 is sensitiv e to AU arrangement.( A ) Schematic representation of the −F-L substrate and its v ariants used in (B and C).The three U sites with the strongest cleavage in every strand of the −F-L substrate identified previously are shown in blue.The AU-rich areas containing nsp15 preferred clea v age sites are indicated using orange dashed bo x es.T he GC-rich sequences replacing the original sequences are shown in red.The black lines represent the original sequences.( B ) Cleavage of the −F-L substrate and its variants as shown in (A) by nsp15.The nsp15 H234A mutant was used as a negative control.Reduction of the full-length substrate in every reaction was calculated as % FL.The a v erage and standard deviation of three independent reactions are graphed.Student's t-test was performed.ns, not significant, P > 0.05; * * * P < 0.001.( C ) Clea v age of the −F -L and −F -L#10 substrates b y v arious concentrations (0, 2.5, 5, 10, 20, 40 and 80 nM) of nsp15.R eduction of the full-length substrate in e v ery reaction was calculated as % FL.In (B and C), samples were analyzed by 20% TBE PAGE (native gel).( D ) Schematic representation of the mL100 substrate and its variants used in (E).The blue arrows indicate the shift direction of the dsRNA cleavage sites in the variants as compared with the mL100 substrate.( E ) Shift of the dsRNA clea v age sites in the substrates shown in (D), indicated by the sizes of the cleavage products.An increased distance between the two prominent gel bands indicates a shift to the right, while a decreased distance indicates a shift to the left.( F ) Schematic representation of the mL100 substrate and its variants used in (G).( G ) Impact of consecutive Us on the cleavage efficiency of nsp15.( H ) Schematic representation of the m100 substrate and its variant used in (I).The two UA base pairs replacing the original CG base pairs are shown in red.The AU-rich area containing consecutive Us ≥ 3 nt in both strands in the variant is marked with orange dashed boxes.( I ) Enhancement of nsp15 clea v age b y introduction of an AU-rich area containing consecutiv e Us ≥ 3 nt in both strands into dsRNA.In (G and I), the concentration of nsp15 or H234A was 10 nM.In (E, G and I), samples were analyzed by 12% TBE PAGE (native gel).

Figure 4 .
Figure 4. dsRNA thermodynamic st abilit y and length modulate the clea v age of SARS-CoV-2 nsp15.( A ) Schematic showing the +F-42.1 and −F-42.1 substrates, their variants containing a single base mismatch, and the +F-42.2and −F-42.2substrates.The corresponding ssRNA substrates are marked b y gre y dashed bo x es.T he preferred clea v age sites of nsp15 on the +F -42.1 and −F -42.1 substrates and their v ariants are sho wn in blue.T he A base replacing the original G base in the variants of the +F-42.1 and −F-42.1 substrates are shown in red.( B ) Denaturing PAGE analysis of the indicated RNA substrates.P ro ximity of the dsRNA gel bands to the corresponding ssRNA gel bands indicates the e xtent of dsRNA denaturation.( C ) T hermal melting analysis of unlabeled alternatives to the +F-L, −F-L, +F-42.1v,−F-42.1v,+F-42.1,−F-42.1,+F-42.2 and −F-42.2substrates (designated as L, 42.1v, 42.1 and 42.2).The first derivative of the melt curve produced a peak, which provided the melting temperature ( T m ).RFU refers to relative fluorescence unit.( D ) Clea v age of the −F -L, −F -42.1 and −F -42.2 substrates b y nsp15.( E ) Identification of the preferred clea v age sites of nsp15 on the +F -42.1, +F -42.1v, −F-42.1 and −F-42.1vsubstrates.( F ) Schematic representation of the variants of the −F-L substrate with various lengths.The three U sites with the strongest clea v age in e v ery strand of the 30-bp v ariants are sho wn in blue.T he black lines represent the sequence of the −F-L30 substrates.( G ) Identification of the preferred clea v age sites of nsp15 on the +F-L30 and −F-L30 substrates.The three strongest cleavage sites in every labeled strand are shown.In (E and G), samples were analyzed by 20% TBE-urea PAGE (denaturing gel).( H ) Clea v age of the variants of the −F-L substrate with various lengths by nsp15.In (D and H), samples were analyzed by 20% TBE PAGE (native gel).Reduction of the full-length dsRNA caused by nsp15 cleavage was quantified as % FL.The nsp15 H234A mutant was used as a negative control.The average and standard deviation for two (D) or three (H) independent reactions are graphed.Student's t -test was performed in (H).ns, not significant, P > 0.05; * P < 0.05; * * P < 0.01; * * * * P < 0.0 0 01.

Figure 5 .
Figure 5. SARS-CoV-2 nsp15 is a dsRNA nickase.( A ) Clea v age of the 3 -6-FAM-labeled positive-sense or negative-sense strand of the dsRNA substrates shown in Supplementary Figure S2 D by nsp15.Reduction of the full-length substrate in every reaction was calculated as % FL.The nsp15 H234A mutant was used as a negative control.The average and standard deviation of three independent reactions are graphed.Student's t -test was performed.* * * P < 0.001.( B ) Clea v age of the mL100 substrate by various concentrations (0, 5, 10 and 20 nM) of nsp15.P refers to the RNA purification products of reaction samples.The gel bands above the full-length substrate gel band are marked by black arrows.( C ) Schematic representation of the 40LR −15 substrate and its variants in which consecutive Us were only present in one strand within the indicated AU-rich area.( D ) Cleavage of the 40LR −15 substrate and its variants shown in (C) by nsp15.The concentration of nsp15 or H234A used here was 10 nM.( E ) Schematic representation of the mL100R80 substrate and its nick-containing variant.The nick is marked with a vertical blue line.( F ) Cleavage of the mL100R80 substrate and its nick-containing variant by various concentrations (0, 2.5, 5 and 10 nM) of nsp1 5. ( G ) Sc hematic representation of the RNA-DNA hybrid substrates related to the +F-L and −F-L substrates.The DNA strands are shown in purple.( H ) Cleavage of the +F-L, +F-L#, −F-L and −F-L# substrates by various concentrations (0, 5, 10 and 20 nM) of nsp15.In (A and H), samples were analyzed by 20% TBE-urea PAGE (denaturing gel).In (B, D and F), samples w ere analyz ed b y 12% TBE PAGE (nativ e gel).

Figure 6 .
Figure 6.Coronavirus nsp15 degrades viral replication dsRNA intermediates.( A ) Pie diagram grouping the coronavirus genomes based on their AU content.The genomes of 41 known coronavirus species from NCBI were included.Detailed information is listed in Supplementary Table S4 .( B ) Clea v age of the dsRNA substrates related to the SARS-CoV-2 mini-genome (O967) or SARS-CoV-2 S protein gene ORF (dsRNA1 and dsRNA2) into small fragments by nsp15.Samples were analyzed by 2.5% TBE AGE (native gel).The black dashed line indicates the position of the 50-bp DNA marker.( C ) Schematic representation of the 39-bp 6-FAM-labeled dsRNA substrate mimicking the dsRNA replication intermediate of the 3 -polyA of the SARS-CoV-2 genome.The ssRNA substrate corresponding to the negative-sense polyU-containing strand is marked by a grey dashed box.( D ) Cleavage of the polyU-containing ssRNA F-TH39 − and dsRNA −F-TH39 substrates by various concentrations (0, 2.5, 5 and 10 nM) of nsp15.Samples were analyzed by 20% TBE-urea PAGE (denaturing gel).The alkaline hydrolysis products of the F-TH39 − substrate were used as an RNA size ladder.( E ) Model depicting ho w corona virus nsp15 mediates e v asion of host cell dsRNA sensors.Nsp15 preferentially clea v es consecutiv e Us in AU-rich areas of dsRNA in a nicking manner.The AU-rich areas are widely distributed in coronaviral genome replication dsRNA intermediates.Nsp15 efficiently cleaves dsRNA ≥40 bp, and thus is able to clea v e the large dsRNA intermediates into fragments shorter than 50 bp, which may evade cytosolic dsRNA sensors, such as Mda5, PKR and OA S , e v en if the y are released from the DMV.
Figure 6.Coronavirus nsp15 degrades viral replication dsRNA intermediates.( A ) Pie diagram grouping the coronavirus genomes based on their AU content.The genomes of 41 known coronavirus species from NCBI were included.Detailed information is listed in Supplementary Table S4 .( B ) Clea v age of the dsRNA substrates related to the SARS-CoV-2 mini-genome (O967) or SARS-CoV-2 S protein gene ORF (dsRNA1 and dsRNA2) into small fragments by nsp15.Samples were analyzed by 2.5% TBE AGE (native gel).The black dashed line indicates the position of the 50-bp DNA marker.( C ) Schematic representation of the 39-bp 6-FAM-labeled dsRNA substrate mimicking the dsRNA replication intermediate of the 3 -polyA of the SARS-CoV-2 genome.The ssRNA substrate corresponding to the negative-sense polyU-containing strand is marked by a grey dashed box.( D ) Cleavage of the polyU-containing ssRNA F-TH39 − and dsRNA −F-TH39 substrates by various concentrations (0, 2.5, 5 and 10 nM) of nsp15.Samples were analyzed by 20% TBE-urea PAGE (denaturing gel).The alkaline hydrolysis products of the F-TH39 − substrate were used as an RNA size ladder.( E ) Model depicting ho w corona virus nsp15 mediates e v asion of host cell dsRNA sensors.Nsp15 preferentially clea v es consecutiv e Us in AU-rich areas of dsRNA in a nicking manner.The AU-rich areas are widely distributed in coronaviral genome replication dsRNA intermediates.Nsp15 efficiently cleaves dsRNA ≥40 bp, and thus is able to clea v e the large dsRNA intermediates into fragments shorter than 50 bp, which may evade cytosolic dsRNA sensors, such as Mda5, PKR and OA S , e v en if the y are released from the DMV.
Figure 6.Coronavirus nsp15 degrades viral replication dsRNA intermediates.( A ) Pie diagram grouping the coronavirus genomes based on their AU content.The genomes of 41 known coronavirus species from NCBI were included.Detailed information is listed in Supplementary Table S4 .( B ) Clea v age of the dsRNA substrates related to the SARS-CoV-2 mini-genome (O967) or SARS-CoV-2 S protein gene ORF (dsRNA1 and dsRNA2) into small fragments by nsp15.Samples were analyzed by 2.5% TBE AGE (native gel).The black dashed line indicates the position of the 50-bp DNA marker.( C ) Schematic representation of the 39-bp 6-FAM-labeled dsRNA substrate mimicking the dsRNA replication intermediate of the 3 -polyA of the SARS-CoV-2 genome.The ssRNA substrate corresponding to the negative-sense polyU-containing strand is marked by a grey dashed box.( D ) Cleavage of the polyU-containing ssRNA F-TH39 − and dsRNA −F-TH39 substrates by various concentrations (0, 2.5, 5 and 10 nM) of nsp15.Samples were analyzed by 20% TBE-urea PAGE (denaturing gel).The alkaline hydrolysis products of the F-TH39 − substrate were used as an RNA size ladder.( E ) Model depicting ho w corona virus nsp15 mediates e v asion of host cell dsRNA sensors.Nsp15 preferentially clea v es consecutiv e Us in AU-rich areas of dsRNA in a nicking manner.The AU-rich areas are widely distributed in coronaviral genome replication dsRNA intermediates.Nsp15 efficiently cleaves dsRNA ≥40 bp, and thus is able to clea v e the large dsRNA intermediates into fragments shorter than 50 bp, which may evade cytosolic dsRNA sensors, such as Mda5, PKR and OA S , e v en if the y are released from the DMV.