Ribonuclease L and metal-ion–independent endoribonuclease cleavage sites in host and viral RNAs

Ribonuclease L (RNase L) is a metal-ion–independent endoribonuclease associated with antiviral and antibacterial defense, cancer and lifespan. Despite the biological significance of RNase L, the RNAs cleaved by this enzyme are poorly defined. In this study, we used deep sequencing methods to reveal the frequency and location of RNase L cleavage sites within host and viral RNAs. To make cDNA libraries, we exploited the 2′, 3′-cyclic phosphate at the end of RNA fragments produced by RNase L and other metal-ion–independent endoribonucleases. We optimized and validated 2′, 3′-cyclic phosphate cDNA synthesis and Illumina sequencing methods using viral RNAs cleaved with purified RNase L, viral RNAs cleaved with purified RNase A and RNA from uninfected and poliovirus-infected HeLa cells. Using these methods, we identified (i) discrete regions of hepatitis C virus and poliovirus RNA genomes that were profoundly susceptible to RNase L and other single-strand specific endoribonucleases, (ii) RNase L-dependent and RNase L-independent cleavage sites within ribosomal RNAs (rRNAs) and (iii) 2′, 3′-cyclic phosphates at the ends of 5S rRNA and U6 snRNA. Monitoring the frequency and location of metal-ion–independent endoribonuclease cleavage sites within host and viral RNAs reveals, in part, how these enzymes contribute to health and disease.


INTRODUCTION
Endoribonucleases include metal-ion-dependent enzymes and metal-ion-independent enzymes (1). Metal-iondependent enzymes (e.g. RNase P and RNase MRP) cleave single-and double-stranded RNA to produce RNA fragments with 5 0 -phosphate and 3 0 -hydroxyl termini. In contrast, metal-ion-independent enzymes (e.g. RNase L, RNase A and IRE1) target singlestranded RNA exclusively, producing RNA fragments with 5 0 -hydroxyl and 2 0 , 3 0 -cyclic phosphate termini ( Table 1). The specificity of metal-ion-independent endoribonucleases for single-stranded sites in RNA is due in large part to the mechanism of RNA cleavage. A structurally flexible 2 0 -OH, in the context of single-stranded RNA, functions as the nucleophile, resulting in the generation of the 2 0 , 3 0 -cyclic phosphate at the site of RNA cleavage by metal-ion-independent endoribonucleases.
Hepatitis C virus (HCV), which is restricted by the 2-5A/ RNase L pathway in tissue culture cells (39), is able to overcome both innate and acquired immune responses in patients to establish persistent infections. Nonetheless, RNase L is predicted to influence the outcomes of HCV infections (40), and RNase L cleavage sites in HCV RNA have been mapped by primer extension (27).
To detect and quantify RNase L cleavage sites in host and viral RNAs, we optimized methods using Arabidopsis thaliana tRNA ligase, an enzyme that specifically recognizes RNA fragments with 2 0 , 3 0 -cyclic phosphate termini (41). We validated the methods using defined viral RNAs cleaved with purified RNase L and RNase A, as well as RNA from uninfected-and poliovirus (PV)-infected HeLa cells under conditions where RNase L was activated during the course of the virus infection (30). As described herein, we discovered discrete regions in HCV and PV RNA genomes that were profoundly susceptible to single-strand-specific endoribonucleases, as well as regions that were largely resistant to cleavage by singlestrand-specific endoribonucleases. We identified a constellation of RNase L-dependent and RNase L-independent cleavage sites in rRNAs and we mapped these cleavage sites onto the secondary and tertiary structures of the human 80S ribosome (42). In addition, we detected 2 0 , 3 0 -cyclic phosphates at the end of U6 snRNA, consistent with the enzymatic activity of C16orf57 (6,7). Lastly, and unexpectedly, our results suggest that 2 0 , 3 0 -cyclic phosphates are present at the end of 5S rRNA.

Viral RNAs cleaved by RNase L and ribonuclease A
Recombinant human RNase L expressed in insect cells and trimer 2-5A [p 3 5'A(2'p5'A) 2 ] were purified as previously described (30,45,46). Bovine pancreatic RNase A was obtained from Ambion. Viral RNAs (100 nM) were incubated at 30 C in reactions (50 ml of volume) containing cleavage buffer (25 mM Tris-HCl [pH 7.4], 100 mM KCl, 10 mM MgCl 2 , 40 mM ATP, 7 mM bME), RNase A (0.5 nM), or RNase L (25 nM), and 2-5A (25 nM). These reaction conditions were optimized empirically to give limited and increasing amounts of viral RNA cleavage over time. RNase A, even at relatively low concentrations, Cells were infected with 10 plaque forming units (PFU) per cell of PV diluted in phosphate buffered saline ($1.2 Â 10 6 cells per 35 mm well). Following 1 h of virus adsorption at room temperature, the inoculum was removed and replaced with 2 ml of DMEM containing fetal bovine serum, pen-strep and G418. Mock infections were performed using phosphate buffered saline without virus. Mock-infected and PV-infected cells were incubated at 37 C. At the indicated times (0, 2, 4, 6 and 8 h after adsorption), RNA was isolated from the cells using guanidine thiocyanate disruption (4 M guanidine thiocyanate, 25 mM sodium citrate, 0.5% N-laurylsarcosine and 0.1 M bME), P:C:I extraction and ethanol precipitation. Host and viral RNAs in each sample were separated by agarose gel electrophoresis and visualized using ethidium bromide and UV light. PV was isolated from a parallel set of cells and quantified by plaque assays to monitor the magnitude and kinetics of virus replication (30).  (20 mM), in excess to RNA fragments with 2 0 , 3 0 -cyclic phosphates, maximized ligation efficiencies (data not shown). After incubation at 37 C for 1 h, RNAs were purified using P:C:I extraction and ethanol precipitation. GlycoBlue (Ambion) was included in all ethanol precipitation steps associated with the preparation of cDNA libraries. The RNAs were fragmented for 10 min at 70 C using 1X Fragmentation Reagents (Ambion), followed by an ethanol precipitation. Because a 2 0 -phosphate persists at the junction between RNA fragments and RNA linkers, fragmented RNAs were dephosphorylated at 30 C for 1 h in reactions (20 ml of volume) containing Tpt1 buffer (20 mM Tris-HCl [pH 7.5], 5 mM MgCl 2 , 0.1 mM DTT and 0.4% Triton X-100), NAD+ (10 mM) and Tpt1 (4.5 mM). These RNAs were P:C:I extracted and ethanol precipitated.
Following the ligation and Tpt1 reactions, RNAs were resuspended in denaturing sample buffer (95% formamide, 18 mM EDTA, 0.025% SDS and 0.025% each bromophenol blue and xylene cyanol), heat denatured (10 min at 65 C) and fractioned on a 6% urea-polyacrylamide gel. RNAs in the gel were visualized using SYBR-Gold (Invitrogen) and blue light transillumination. RNAs 100-500 nt in length were excised from the gel, crushed into a slurry with 0.3 M sodium acetate and incubated at 42 C for 2 h. RNA was purified from polyacrylamide by running samples through DTR cartridges (EdgeBio) followed by ethanol precipitation.
Superscript III (Invitrogen) and a DNA primer (5 0 -AAT GAT ACG GCG ACC ACC GAG ATC TAC ACT CTT TCC CTA CAC GAC GCT C-3 0 ) were incubated with the gel-purified RNAs to make cDNA (per the manufacturer's instructions). Exonuclease I (0.5 U) and recombinant shrimp alkaline phosphatase (0.25 U) (USB/Affymetrix) were added to each RT reaction, followed by additional incubation at 37 C for 30 min, to remove excess DNA primer and dNTPs. RT reactions were then heated to 95 C to inactivate the enzymes. RNAs were hydrolyzed using 1 N NaOH treatment at 100 C for 10 min and neutralized with 1 N HCl. cDNA was P:C:I extracted and ethanol precipitated.
A DNA linker was attached to the 3 0 -end of the cDNA in reactions (10 ml of volume) containing 1 mM miRNA cloning linker 1 (5 0 -rApp-CTG TAG GCA CCA TCA AT-ddC-3 0 [Integrated DNA Technologies]), 10 units of T4 RNA ligase I (New England Biolabs) and T4 RNA ligase buffer (50 mM Tris-HCl [pH 7.5], 10 mM MgCl 2 and 1 mM DTT). Linked cDNAs were P:C:I extracted and ethanol precipitated. Linked cDNAs were resuspended in formamide sample buffer and fractioned on a 6% urea-polyacrylamide gel. cDNAs >150 bases long were gel purified and resuspended in 30 ml of water.
Polymerase chain reaction (PCR) was used to amplify cDNA libraries before Illumina sequencing. PCR products 250-1000 bp in length were gel purified from a 6% nondenaturing polyacrylamide gel and resuspended in 10 mM Tris-HCl [pH 8.5] with 0.1% Tween-20. The amounts of DNA in each library were determined using the Qubit fluorometer (Invitrogen). Five to seven DNA libraries were combined for multiplexed sequencing for a combined total of 10 nM of DNA for each Illumina sequencing run. A multiplexing index read primer (5 0 -CTG TAG GCA CCA TCA ATG AAC TCC AGT CAC-3 0 ) was used to identify the library after sequencing on the MiSeq or GAIIx (Illumina, Inc).
To determine the position and frequency of endonuclease cleavage sites in host and viral RNAs, the 3 0 -end of each cDNA read was plotted against the nucleotide position of each genome using R (50). For RNase L-cleaved HCV, and RNase A-cleaved HCV and PV RNA data, the signal from the 'no 2-5A' and 'no RNase A' was subtracted from the 0, 2.5, 5, 10 and 20 min data sets. The 3 0 -dinucleotides of aligned reads were quantified using the UMIs and the sum of each of the 16 possible dinucleotides was divided by the total number of UMI-corrected reads for each RNA of interest and multiplied by 100 to get a percentage.

NCBI GEO
The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (Cooper et al., 2014) and are accessible through GEO Series accession number GSE52489 (51).

RESULTS
2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods RNase L, RNase A and other metal-ion-independent endoribonucleases target single-stranded regions of RNA, leaving 2 0 , 3 0 -cyclic phosphates at the end of RNA fragments (Table 1). We exploited the 2 0 , 3 0 -cyclic phosphates at RNA cleavage sites to make cDNA libraries suitable for Illumina sequencing (Supplementary Figure S1).

Viral RNAs cleaved with purified RNase L and RNase A
Initially, we used viral RNAs and purified endoribonucleases to optimize and validate the 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods. HCV and PV RNAs were incubated for 0-20 min in reactions containing RNase L and 2-5A or RNase A, followed by agarose gel electrophoresis to characterize the RNA fragments ( Figure 1). RNase L and RNase A generated viral RNA fragments ranging from 100 to several thousand bases in length. Notably, HCV and PV RNA fragments with distinct sizes were evident, consistent with nonrandom cleavage of the viral RNAs ( Figure 1).

Frequency, location and dinucleotide specificity of endoribonuclease cleavage sites in viral RNAs
The viral RNAs from each time point shown in Figure 1 were analyzed by 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing (Supplemantary Tables S1 and  S2 Figure S4). Note that UU 4465 and UU 4466 are adjacent cleavage sites, where cleavage at UU 4465 would preclude detectable cDNA reads at UU 4466 . These data suggest that viral RNAs were partially digested, even after 20 min of incubation with RNase L, and that prominent cleavage sites detected at early times did not disappear after more prolonged periods of incubation.

RNA from uninfected and PV-infected HeLa cells
After optimizing and validating the 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods using purified endoribonucleases, we used the methods to characterize RNA from uninfected and PV-infected HeLa cells ( Figure 3). W12 HeLa cells express wild-type RNase L, whereas M25 HeLa cells express a dominant-negative mutant form of RNase L (R667A mutation). We previously established that RNase L activity, which requires 2-5A from dsRNA-activated oligoadenylate synthetase, is provoked during the course of PV infection in W12 HeLa cells, whereas RNase L activity remains undetectable during the course of PV infection in M25 HeLa cells (30). The kinetics and magnitudes of PV replication in W12 and M25 HeLa cells are similar ( Figure 3A). When RNAs from the PV-infected cells were analyzed by agarose gel electrophoresis, the accumulation of viral RNA was evident at 4-8 h post adsorption (hpa) ( Figure 3B). rRNA fragments characteristic of RNase L activity were evident in RNAs from PV-infected W12 HeLa cells at 6 and 8 hpa, but these rRNA fragments were not detected in PV-infected M25 HeLa cells ( Figure 3B, asterisks indicate the location of rRNA fragments characteristic of RNase L activity).
cDNA libraries were prepared and sequenced using the RNAs from HeLa cells (Figure 3C and D and  Supplementary Table S3). cDNA reads corresponding to PV RNA increased from undetectable levels at early times after infection to 5% of the cDNA at 6 hpa in W12 HeLa cells and 26.1% of the cDNA at 6 hpa in M25 HeLa cells (Figure 3C and D and Supplementary Table S3). Abundant amounts of cDNA in each sample aligned to rRNAs (28S/18S/5.8S/5S rRNAs) and U6 snRNA ( Figure 3C and D and Supplemantary Table S3). cDNA reads corresponding to host mRNAs and other nonribosomal RNAs were present in each sample, ranging from 2.6% of all cDNA from PV-infected W12 HeLa cells at 2 hpa to 16.3% of cDNA in mock-infected W12 HeLa cells at 8 hpa.

Endoribonuclease cleavage sites in PV RNAs isolated from HeLa cells
Discrete regions of PV RNA were targeted by endoribonucleases both in vitro ( Figure 2B and Supplementary Figure S3) and in vivo (Figure 4). One prominent RNase L-dependent cleavage site at UA 1715 was evident in PV RNA isolated at 6 and 8 hpa from W12 HeLa cells ( Figure 4A). A second potential RNase L-dependent cleavage site at UU 3102 was evident in PV RNA isolated at 6 and 8 hpa from W12 HeLa cells ( Figure 4A); however, there was substantial RNase L-independent cleavage at UU 3102 in M25 HeLa cells ( Figure 4C), making it uncertain as to the enzyme(s) responsible for cleavage at this site in cells. UA 1715 and UU 3102 are prominent RNase L cleavage sites identified using purified RNase L ( Figure 2B and Supplementary Figure S5). Surprisingly, there were no other obvious RNase L-dependent cleavage sites in PV RNA isolated from W12 HeLa cells. The increased frequency of UA and UU dinucleotides in cDNAs from W12 HeLa cells at 6 and 8 hpa was consistent with the activation of RNase L at these times (compare Figure  4B and D). RNase L-independent cleavage sites in PV RNA from HeLa cells mapped to the same regions of PV RNA as the cleavage sites found using purified RNase L and RNase A, with notable peaks of cleavage in the 5 0 -half of the open-reading frame (compare Figure 2B with Figure 4A and C). Potential reasons for these hypersensitive regions in PV RNA are considered in the discussion. Remarkably, PV cDNA reads from HeLa cells corresponded to cleavage sites exclusively within positive-strand RNA; no cDNA reads were found corresponding to cleavage sites within PV negative-strand RNA.

Endoribonuclease cleavage sites in rRNAs isolated from HeLa cells
The location and frequency of endoribonuclease cleavage sites in rRNAs was largely consistent across all of the RNA samples from uninfected and PV-infected HeLa cells ( Figures 5 and 6 Figure S11).

3' 5'
A striking distribution of RNase L-independent cleavage sites, with the majority of signal at the 3 0 -end of mature 5S rRNA ( Figures 5D and 6D and Supplementary Figure  S12 Table S3).
The location of these RNase L-dependent and RNase L-independent endoribonuclease cleavage sites were mapped onto the secondary structures of rRNAs (Supplementary Figures S13 and S14), as well as the tertiary structure of 80S ribosomes (Figure 7) (27). RNase L-dependent cleavage sites within 18S rRNA map to single-stranded regions of RNA on the surface of the 40S subunit ( Figure 7A and B). Likewise, the 3 0 -ends of 5.8S and 5S rRNAs are located on the surface of the 60S subunit ( Figure 7C). RNase L-independent cleavage sites mapped to various locations on the surface of 40S subunits ( Figure  7D). Cleavage sites were less frequently present at the surface of 60S subunits ( Figure 7D).

DISCUSSION
RNase L, like other metal-ion-independent endonucleases, is expected to produce RNA fragments with 2 0 , 3 0cyclic phosphate termini (Table 1). Our data confirm that RNase L, like RNase A, produces RNA fragments with 2 0 , 3 0 -cyclic phosphate termini. We used 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods to identify and quantify metal-ion-independent endoribonuclease cleavage sites within host and viral RNAs. We optimized and validated the methods using viral RNAs cleaved with purified RNase L, viral RNAs cleaved with purified RNase A and RNA from uninfected and PV-infected HeLa cells (Figures 1-7 and Supplementary Tables S1-S6). Overall, these methods reliably and reproducibly reveal the location and relative frequency of 2 0 , 3 0 -cyclic phosphates in host and viral RNAs. These cDNA synthesis and Illumina sequencing methods did not detect RNA fragments corresponding to RNase P or RNase MRP activity, consistent with the specificity of the method for RNA fragments with 2 0 , 3 0 -cyclic phosphates. Several lines of evidence suggest that cyclic phosphates within host and viral RNAs were due to metal-ionindependent endonucleolytic cleavage of RNA within cells rather than cleavage during RNA processing or within linker reactions. Cyclic phosphates were not frequent in viral RNA transcripts isolated from control reactions lacking endonucleases, confirming the absence of  Figure S15), a cellular enzyme known to leave cyclic phosphates at the end of mature U6 snRNA in cells (6,7). Nonetheless, it is important to note that divalent metal ions can promote RNA cleavage at specific sites, leaving cyclic phosphates (53). Thus, in addition to metal-ion-independent endoribonucleases, some of the RNA cleavage sites revealed by 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing could be due to metal RNA binding pockets within RNAs. In fact, Pb 2+ -dependent cleavage of 28S rRNA at AG 409 has been reported (54). This is a prominent and reproducible cleavage site within our data sets ( Figures 5A  and 6A and Supplementary Figures S7A and S8A). Thus, a metal binding pocket at this site in 28S rRNA, under physiologic conditions within cells, may lead to rRNA cleavage. Carefully controlled experimental variables are needed to attribute specific cleavage sites to particular endonucleases or to metal binding pockets.

Endoribonuclease cleavage sites in HCV RNA
Our original interest in 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods was driven by previous studies of RNase L and HCV RNA. Primer extension reveals RNase L cleavage sites in HCV RNA in vitro (27)   results are consistent with those obtained using primer extension (27); however, the quantitative data from 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing is more reliable than that obtained using primer extension. Quantitative analyses of RNase L cleavage sites in HCV RNA by primer extension were confounded by multiple factors: the use of multiple primers (30 primers across the HCV RNA genome), variable hybridization efficiencies for each primer and decreased sensitivity of cleavage site detection as the distance increased from the site of hybridization. In contrast, quantitative analyses of RNase L cleavage sites in HCV RNA using 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing are extremely precise. Because HCV RNAs from distinct genotypes have variably reduced frequencies of UA and UU dinucleotides (40), the potential role(s) of RNase L in HCV infections and interferon-based antiviral therapies deserves further examination now that a method has been developed to reliably detect and quantify RNase L cleavage sites in host and viral RNAs.

Endoribonuclease cleavage sites in PV RNA
Endoribonuclease cleavage sites across PV RNA genomes have not been previously described. The striking distribution of RNase L and RNase A cleavage sites in PV RNA ( Figure 2B), with prominent amounts of cleavage in discrete regions of the genome, was not expected. The same discrete regions of PV RNA were targeted by endoribonucleases both in vitro ( Figure 2B) and in vivo ( Figure 4). Furthermore, one RNase L-specific cleavage site was reliably detected both in vitro and in vivo ( Figures 2B and 4, UA 1715 in PV RNA). We entertained two hypotheses as we considered these data. First, PV RNA may assume a secondary and tertiary structure that renders some portions of the genome particularly susceptible to single-strand-specific endoribonucleases while rendering other regions resistant (Supplementary Figure S5). Second, antigenic variation in the capsid proteins, as a consequence of antibody selection, could render portions of the capsid genes more sensitive to cleavage by single-strand-specific endoribonucleases, due to the counter-selective forces of antigenic variation and resistance to endoribonucleases (Supplementary Figure S6). The predicted secondary structures of HCV and PV RNAs do explain much of the data (55), as metal-ion-independent endoribonucleases do not cleave dsRNA portions of viral RNA (Supplementary Figures  S4 and S5). In addition, atomic force microscopy of HCV and PV RNA genomes reveals globular structures, where some regions of the viral RNA genomes might be more exposed to endoribonucleases than others (56,57). Intriguingly, antibody neutralization escape mutations are found within and near endonuclease susceptible regions of PV RNA (58,59); however, the peaks of endonucleolytic cleavage in the PV capsid genes do not overlap exclusively with the location of RNA changes associated with antigenic variation (Supplementary Figure S6). A complete understanding of the discrete distribution of endoribonuclease cleavage sites in PV RNA awaits further investigation.

Endoribonuclease cleavage sites in rRNA
A constellation of RNase L-dependent and RNase L-independent cleavage sites in rRNAs was revealed by 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing (Figures 5 and 6; Supplementary Figures  S7-S14 and Supplementary Tables S4-S6). Unlike many studies, which actively exclude rRNAs from deep sequencing experiments to enrich for other RNA populations, we intentionally retained rRNAs in our study to identify RNase L-dependent and RNase L-independent cleavage sites in rRNAs. Although we did not detect RNase L-dependent cleavage sites in 28S rRNA under the conditions of our experiment, others have reported RNase L-dependent cleavage of 28S rRNA (60,61). We did identify two RNase L-dependent cleavage sites in 18S rRNA, UU 541 and UU 743 ( Figures 5B and 7A and B). Cleavage of 18S rRNA (1869 bases long) at these sites would produce rRNA fragments of 541, 743, 1126 and 1328 bases long, consistent with the characteristic rRNA fragments observed by agarose gel electrophoresis in this report ( Figure 3B), and in previous studies by our lab and others (30,37). As expected for RNase L, the 18S rRNA cleavage sites, UU 541 and UU 743 , are associated with single-stranded regions of 18S rRNA, on the surface of 80S ribosomes (Figure 7). A number of RNase L-independent cleavage sites were identified consistently in rRNAs (Figures 5-7;  Supplementary Figures S7-S14 and Supplementary  Tables S4-S6). Each of the sites mapped to singlestranded regions of rRNA (Supplementary Figures S13 and S14), and most were located at the surface of 40S and 60S ribosomal subunits ( Figure 7D). The RNase L-independent cleavage sites of 28S rRNA occurred in core segments of the 28S rRNA, while little cleavage occurred in the GC-rich expansion segments ( Figures 5A  and 6A), despite the location of expansion segments on the surface of ribosomes (Figure 7). The paucity of cDNA reads in 28S rRNA expansion segments is likely due to the dsRNA-rich characteristics of expansion segments (42) (Supplementary Figure S13/dsRNA is resistant to metalion-independent endoribonucleases) and to the reduced frequency of Illumina cDNA reads in GC-rich regions of RNA (62). The paucity of cDNA reads in rRNA expansion segments evident in our data ( Figures 5 and 6) is similar to data from others (63).
The functional impact of rRNA cleavage at particular sites is yet to be determined. rRNA cleavage has been associated with RNase L-dependent (37) and RNase L-independent antiviral responses (38), ER-stress (64), apoptosis (65), oxidative stress (66) and the inhibition of translation during spermatogenesis (63). The RNase L cleavage sites identified in this investigation in 18S rRNA (UU 541 and UU 743 ) are distant from the mRNA decoding channel (Figure 7). It is unclear whether cleavage of 18S rRNA at these sites would affect translation. AG 409 in 28S rRNA, one of the most prominent RNase L-independent cleavage sites in both uninfected and PV-infected HeLa cells (Figures 5A and 6A;  Supplementary Figures S7 and S8 and Supplementary  Table S4), is analogous to an oxidative stress-induced rRNA cleavage site identified in yeast (66). Yet again, there are no definitive data to indicate how cleavage of 28S rRNA at this site would affect translation. Nonetheless, because decreased translation is a common aspect of antiviral responses, ER-stress, apoptosis and oxidative stress, it is reasonable to speculate that cleavage of rRNAs at these sites might inhibit the translation activity of ribosomes. It is also important to consider the potential role of endonucleases in rRNA decay (67,68). RNase T2 has been implicated in rRNA decay pathways in plants (69) and vertebrates (70).
The 2 0 , 3 0 -cyclic phosphates detected at the ends of 5.8S and 5S rRNAs were unexpected. Cyclic phosphates have been described at the 3 0 -end of U6, 7SK and MRP RNAs, but not at the 3 0 -end of 5S rRNA (71). Nonetheless, 5S rRNA was among the most common sequences detected in 2 0 , 3 0 -cyclic phosphate cDNA libraries ( Figure 3C and D and Supplementary Table  S3), and almost all of the 5S rRNA signal was associated with the 3 0 -end of mature 5S rRNA ( Figures 5D and 6D and Supplementary Figure S12). The 3 0 -ends of 5.8S and 5S rRNAs are on the surface of 60S ribosomal subunits ( Figure 7C), consistent with the possibility that ribosomes contain RNAs with these terminal modifications. Alternatively, the mature 3 0 -end of 5S rRNA, in the context of Mdmx complexes, could be modified with 2 0 , 3 0 -cyclic phosphates (72). Rex1p, a 3 0 to 5 0 exonuclease in the RNase D family, is involved with modifying the 3 0ends of both 5S and 5.8S rRNA in yeast; however, RNase D family members are metal-ion-dependent, and thus it is unlikely that a human homolog of Rex1p is responsible for the 2 0 , 3 0 -cyclic phosphate found at the end of these rRNAs (73,74).
While the identity of endonucleases responsible for many of the RNase L-independent cleavage sites in rRNA remains to be determined, the cyclic phosphates found at the ends of 5.8S and 5S rRNA may be consistent with the enzymatic activity of the enzyme associated with U6 snRNA maturation (6,7). The cyclic phosphate at the 3 0 -end of U6 snRNA facilitates interactions with Lsm proteins (75), which promote U6 snRNP maturation (8). Lsm proteins are also implicated in rRNA processing and maturation (76). Therefore, the cyclic phosphates found at the 3 0 -ends of 5.8S and 5S rRNAs may contribute to their interaction with Lsm proteins and RNP maturation.
Endoribonuclease cleavage sites in U6 snRNA U6 snRNA is a Pol III transcript modified posttranscriptionally by an exonuclease that leaves 2 0 , 3 0 -cyclic phosphates at the end of mature U6 snRNAs (6,7). The abundant amounts of cDNA in our libraries corresponding to this particular modification of U6 snRNA further validate the reliability of the 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing methods ( Figure 3C and D and Supplementary Figure S15).

Summary
We used 2 0 , 3 0 -cyclic phosphate cDNA synthesis and Illumina sequencing to identify and quantify metal-ionindependent endoribonuclease cleavage sites in host and viral RNAs. With these methods, we identified discrete regions of HCV and PV RNA genomes that were profoundly susceptible to RNase L and other singlestrand specific endoribonucleases, RNase L-dependent and RNase L-independent cleavage sites within rRNAs, and 2 0 , 3 0 -cyclic phosphates at the ends of 5S rRNA and U6 snRNA. We expect these methods will prove useful in the characterization of other metal-ion-independent endoribonucleases, like the NendoU of arteriviruses and coronaviruses whose RNA substrates remain unknown (17), and the T2 endonucleases that target undefined sites in rRNAs (70).