RNase L reprograms translation by widespread mRNA turnover escaped by antiviral mRNAs

SUMMARY

In response to foreign and endogenous double-stranded RNA (dsRNA), protein kinase R (PKR) and ribonuclease L (RNase L) reprogram translation in mammalian cells. PKR inhibits translation initiation through eIF2α phosphorylation, which triggers stress granule (SG) formation and promotes translation of stress responsive mRNAs. The mechanisms of RNase L-driven translation repression, its contribution to SG assembly, and its regulation of dsRNA stress-induced mRNAs are unknown. We demonstrate that RNase L drives translational shut-off in response to dsRNA by promoting widespread turnover of mRNAs. This alters stress granule assembly and reprograms translation by allowing translation of mRNAs resistant to RNase L degradation, including numerous antiviral mRNAs such as IFN-β. Individual cells differentially activate dsRNA responses revealing variation that can affect cellular outcomes. This identifies bulk mRNA degradation and the resistance of antiviral mRNAs as the mechanism by which RNase L reprograms translation in response to dsRNA.

Graphical abstract

eTOC

Burke et al. demonstrate that RNase L reduces bulk translation and stress granule assembly in response to dsRNA via rapid and widespread decay of basal mRNAs. In contrast to basal mRNAs, antiviral mRNAs escape RNase L-mediated mRNA decay, which permits their translation during RNase L-mediated host shut-off.

INTRODUCTION

Double-stranded RNA (dsRNA) is a viral pathogen-associated molecular pattern (PAMP) that can initiate the innate immune response (Jensen and Thomsen, 2012). Endogenous “self’ dsRNAs can also initiate the innate immune response, and dysregulation of cellular pathways that reduce self-dsRNA causes human diseases, such as Aicardi-Goutières syndrome (AGS) (Pestal et al., 2015, Liddicoat et al., 2015, Li et al., 2017). Elevated levels of endogenous dsRNAs may also contribute to neurodegenerative diseases, such as ALS (Saldi et al., 2014, Krug et al. 2017), and can contribute to chronic inflammation associated with cancers and autoimmune disorders (Grivennikov et al., 2010, Waldner 2009).

Several pattern recognition receptors (PRRs) recognize dsRNA in mammalian cells (Jensen and Thomsen, 2012). Recognition of dsRNA by RIG-I, MDA-5, or TLR3 activates the transcription factors IRF3 and/or IRF7 thereby inducing type-1 interferons (IFNs) and inflammatory cytokines, which prime the antiviral state of cells (Ivashki et al., 2014). Concurrent with the induction of antiviral genes, global translation is reduced by protein kinase R (PKR) and ribonuclease L (RNase L) to promote an antiviral cellular state that limits viral gene expression and transforms the functional cellular transcriptome, a process termed host shutoff (Iordanov et al., 2000).

Activation of PKR by dsRNA results in phosphorylation of eIF2α on serine 51, which reduces canonical translation initiation and promotes the translation of stress response mRNAs that use non-canonical translation initiation (Dalet et al., 2015). This also triggers the formation of stress granules (SGs), conserved RNA-protein complexes that contain non-translating mRNAs, RNA-binding proteins – G3BP1, PABPC1, TIA1 – and several key antiviral PRRs – OAS/RNase L, PKR, MDA-5, and RIG-I (García et al., 2006, Onomoto et al., 2012, Reineke et al., 2012, Yoo et al., 2014). Many viruses inhibit SG assembly, suggesting that SGs serve as antiviral signaling hubs and/or reduce viral replication through the sequestration of viral mRNAs/proteins (Lloyd, 2013). However, the disassembly of SGs via dephosphorylation of p-eIF2α by GADD34, which is induced by IRF3, has been proposed to promote translation of stress-induced antiviral mRNAs that are sequestered to SGs (Dalet et al., 2017). Thus, the mechanisms and functions of SG assembly/disassembly during the dsRNA/antiviral response remain unclear.

RNase L is an endonuclease activated by oligo(2’-5’A), which is produced when OAS proteins bind to dsRNA. RNase L cleaves ssRNA regions at UN^N sites in several types of RNAs, including small non-coding RNAs (Y-RNAs and tRNAs), rRNAs, and host/viral mRNAs (Andersen et al., 2009, Chakrabarti et al., 2011, Brennan-Laun et al., 2014, Donovan et al., 2017). These activities of RNase L reduce viral gene expression and replication, arrest global translational, and promote apoptosis (Zhou et al., 1997). RNase L is proposed to arrest translation by either cleavage of rRNA or production of RNA cleavage fragments that signal for translational arrest (Wreschner et al., 1981; Donovan et al., 2017). However, because these modes of translational arrest are presumably non-specific, a mystery in the field is how dsRNA-induced antiviral mRNAs would be translated during RNase L-driven translational arrest.

We present data demonstrating that RNase L promotes rapid and widespread degradation of cellular mRNAs in response to dsRNA, thus inhibiting SG assembly and decreasing translation. However, mRNAs encoding key antiviral proteins and cytokines, such as the IFN-β and IL-6 mRNAs, escape RNase-L mediated mRNA turnover, which permits their translation during host shut-off when bulk mRNA turnover is the primary driver of global translation repression.

RESULTS

RNase L catalytic activity alters SG assembly and reduces SG-associated RNAs

RNase L represses translation and accumulates in stress granules (Onomoto et al., 2012, Reineke et al., 2015). Thus, we examined if RNase L activity affected stress granule assembly. We generated RNase L knockout (RL-KO) A549 and U-2 OS cell lines using CRISPR-Cas9, and then reconstituted expression of either RNase L or catalytically inactive RNase L-R667A in the RL-KO cells via lentiviral transduction or transient transfection (Figure S1A,B,C). Cells were transfected with high molecular weight poly(I:C), a viral dsRNA mimic that induces PKR-dependent SG assembly and activates the OAS/RNase L pathway. SG assembly was assessed by immunofluorescence assay (IF) for SG-associated proteins G3BP1 and PABPC1.

In comparison to the parental (WT) cell lines, we observed two distinct phenotypes in RL-KO cell lines that were rescued by expression of RNase L, but not RNase L-R667A. First, SGs in the RL-KO cells were canonical in morphology (large and irregular in shape), whereas cytoplasmic puncta of G3BP1 and PABPC1 observed in the WT cells were invariably small and punctate (Figures 1A,B,C and S1D,E). Second, a substantial fraction of PABPC1 translocated from the cytosol to the nucleus in WT cells, whereas PABPC1 remained localized in the cytosol and SGs in RL-KO cells. The RNase L-dependent reduction in SG size was specific to the dsRNA stress response, as sodium arsenite treatment induced canonical SGs in both WT and RL-KO cells (Figure S1F). Moreover, the canonical dsRNA-induced SGs in RL-KO cells require PKR, whereas the small punctate dsRNA-induced SGs in WT cells are independent of PKR, but require RNase L catalytic activity (Figures 1D and S1G,H,I,J). We refer to these small punctate SGs in WT cells as RLBs (RNase L-dependent Bodies) and use RLBs and nuclear PABP accumulation as markers for RNase L activation in subsequent single-cell analyses. These data indicate that RNase L activation alters SG assembly and instead leads to the assembly of RLBs.

Figure 1. RNase L catalytic activity alters SG assembly and reduces SG-associated RNAs.

(A) IF for SG-associated proteins G3BP1 and PABPC1 in WT and RL-KO U-2 OS cells. (B) G3BP1-positive foci from greater than thirty WT and RL-KO U-2 OS cells binned by volume. (C) IF for G3BP1 and PABPC1 in parental RL-KO A549 cells stably expressing either RNase L (RL) or RNase L-R667A (RL-CM) eight hours post-poly(I:C). Images for G3BP1 and PABPC1 staining are shown in Figure S1E. (D) G3BP1 and PABPC1 IF in PKR-KO and PKR and RNase L double KO (PKR/RL-KO) A549 cells rescued with RNase L (RL) or RNase L-R667A (RL-CM) six hours post-poly(I:C). (E) smFISH for AHNAK mRNA in WT and RL-KO U-2 OS cells +/− poly(I:C) with G3BP1 as a RLBs/SG marker. N.r. indicates non-responsive cells with respect to RLB/SG assembly. (F) Quantification of AHNAK mRNA smFISH from (E), with analysis of 17-30 RLB/SG cells in three fields of view. (G) RT-qPCR analysis of AHNAK mRNA in WT and RL-KO A549 cells at zero- and six-hours post-poly(I:C) transfection.

Since mRNAs are an integral component of SGs (Van Treeck et al., 2018), and PABPC1 translocation to the nucleus occurs upon mRNA turnover (Lee and Glaunsinger, 2009; Kumar and Glaunsinger, 2010), we hypothesized that RNase L limits SG formation by degrading SG mRNAs. Single-molecule fluorescent in situ hybridization (smFISH) analysis of the SG-enriched AHNAK mRNA (Khong et al., 2017) revealed that this mRNA was strongly reduced in the cytosol and did not localize to RLBs in WT cells post- poly(I:C) (Figure 1E,F). In contrast, AHNAK mRNA remained abundant and localized to SGs in RL-KO cells. qRT-PCR analysis confirmed these results, showing that AHNAK mRNA levels significantly decreased in WT, but not RL-KO cells, post-poly(I:C) (Figure 1G). Analysis of the SG-enriched NORAD long non-coding RNA (lncRNA) yielded similar results (Figure S2A,B,C). These results demonstrate that RNase L alters SG assembly by degrading SG-associated RNAs.

RNase L initiates rapid and widespread turnover of mRNAs

We next examined whether RNase L affects mRNAs not enriched in SGs. The GAPDH mRNA, which is abundant and depleted from SGs (Khong et al., 2017), was reduced by at least 90% in the majority of WT A549 and U2-OS cells containing RLBs by two and six hours post-poly(I:C) (Figures 2A,B and S2D,E,F). Near complete reduction of GAPDH mRNA was observed in a fraction of A549 cells containing RLBs as early as one-hour post poly(I:C) (Figure S2G). In non-responsive neighboring cells (n.r.) that lack RLBs, GAPDH mRNA levels remained abundant and comparable to mock-treated cells. GAPDH mRNA levels were unchanged in RL-KO cells post-poly(I:C) (Figure 2A,B), and this was rescued by RNase L but not RNase L-R667A (Figure 2C,D), indicating that RNase L catalytic activity promotes GAPDH mRNA degradation. qRT-PCR analysis confirmed these observations, revealing that GAPDH mRNA levels, as well as Actin B and Tubulin A mRNAs, decreased (greater than 75%) in WT but not RL-KO cells post-poly(I:C) (Figure 2E). Moreover, FISH for poly(A)+ RNAs revealed at least a 70% RNase L-dependent decrease in cytosolic poly(A)+ RNAs by two hours post-poly(I:C) in cells containing RLBs (Figure 2F,G). These results suggest RNase L is degrading the majority of cytoplasmic mRNAs in response to dsRNA.

Figure 2. RNase L promotes widespread turnover of mRNAs.

(A) smFISH for GAPDH mRNA (red) in WT and RL-KO A549 cells. (B) Quantification of GAPDH mRNA smFISH represented in (A). (C) smFISH/IF for GAPDH mRNA and G3BP1 in RL-KO cells rescued with RNase L (RL) or RNase L-R667A (RL-CM). (D) Quantification of GAPDH mRNA smFISH represented in (C). (E) qRT-PCR quantification in WT and RL-KO A549 cells transfected with or without (Mock) poly(I:C) for six hours. Bars represent the average C_t value differential +/− S.E.M. from at least five independent experiments. (F) Oligo(dT) (red) and G3BP1 (green) staining of WT and RL-KO A549 cells. (G) Quantification of mean oligo(dT) FISH signal in RLB/SG-positive cells as represented in (F). Greater than twenty cells were analyzed at the indicated times post-poly(I:C).

IFN-β and IL-6 mRNAs escape RNase L-mediated mRNA turnover

The degradation of host mRNAs by RNase L creates a problem for how cells undergoing the antiviral/dsRNA response produce proteins from IRF3-induced genes, such as IFN-β. Our smFISH/IF analyses revealed that not all cells activate RNase L-mediated mRNA decay (Figure 2 and S2F). Thus, one possibility is that antiviral mRNAs are degraded in cells that activate RNase L-mediated mRNA decay but are increased in cells not undergoing RNase L responses. Alternatively, IRF3-induced mRNAs may escape RNase L-dependent turnover. To test the latter hypothesis, we performed smFISH for the IFN-β mRNA simultaneously with the GAPDH mRNA to determine if IFN-β mRNAs were present in cells that degraded the GAPDH mRNA. Strikingly, in cells wherein GAPDH mRNA had been reduced by RNase L in response to poly(I:C), cytosolic IFN-β mRNA was abundant (Figures 3A and S3A,B). Quantification of IFN-β mRNA by both smFISH and RT-qPCR revealed that it was induced to comparable levels between WT and RL-KO cells post-poly(I:C) (Figure 3B,C). These data indicate that the IFN-β mRNA is resistant to RNase L-mediated mRNA degradation.

Figure 3. IFN-β mRNA escapes RNase L-mediated mRNA turnover.

(A) smFISH/IF for IFN-β mRNA GAPDH mRNA and IF for G3BP1 in WT and RL-KO A549 cells. Cells that contain RLBs/SGs and IFN-β mRNA indicated by white arrows. Cells with RLBs/SGs without IFN-β mRNA are indicated by yellow arrows. Cells with IFN-β mRNA without RLBs/SGs are indicated by orange arrows. Cells lacking both RLBs/SGs and IFN-β mRNA indicated by blue arrows. (B) Quantification of IFN-β mRNA smFISH in WT and RL-KO A549 cells. (C) RT-qPCR analysis of IFN-β, IL-6, and GADD34 mRNA expression in WT and RL-KO A549 cells six-hours post-poly(I:C). Bars represent the average C_t value +/− the S.E.M from greater than 5 independent experiments.

We observed a similar phenomenon assaying the IL-6 mRNA via smFISH, although there were more IL-6 mRNAs present in the RL-KO cells (Figure S3C,D), which suggests the IL-6 mRNA is only partially resistant to RNase L degradation. Consistent with this, qRT-PCR revealed that IL-6 mRNA induction was less in WT cells as compared to RL-KO cells (Figure 3C). Similarly, the GADD34 mRNA is induced by poly(I:C) but is reduced in an RNase L-dependent manner as assessed by qRT-PCR. Thus, there are mRNA specific differences in resistance to RNase L degradation, even among mRNAs induced by IRF3/7.

Differential PRR activation in individual cells in response to dsRNA

Our single-cell analyses of mRNAs, RLBs, and SGs revealed an unexpected cell-to-cell variability in the response to dsRNA. For example, while the six WT cells treated with poly(I:C) in Figure 3A all showed RNase L activation, as determined by the formation of RLBs and degradation of GAPDH mRNA, only three cells strongly induced the IFN-β mRNA (Figure 3A, white arrows). We also observed cells that induced the IFN-β mRNA without activation of RNase L (Figure S3B). Similarly, in RL-KO cells, we can observe individual cells that activate PKR (as assessed by SG formation), with or without induction of the IFN-β mRNA (Figure 3, white and yellow arrows). We also observed RL-KO cells that induced the IFN-β mRNA without activation of PKR (Figure 3A, orange arrow). Interestingly, WT cells only generate RLBs and not SGs (Figure 1), indicating that PKR activation does not commonly occur without co-activation of RNase L in individual WT cells. These differences are not due to failures in transfections since cells are activating one portion of the response to dsRNA, but not other pathways. This demonstrates that individual cells respond differentially to dsRNA, which can allow multiple cellular outcomes during the dsRNA/antiviral response (see discussion).

Genome-wide analysis of RNase L-mediated mRNA degradation

To identify the RNase L sensitive and RNase L resistant mRNAs on a comprehensive scale, we performed high-throughput RNA sequencing on WT and RL-KO A549 cells +/− poly(I:C) transfection. Standard differential expression analysis showed that few RNAs were significantly up- or down-regulated between untreated WT and RL-KO cell lines (Figure S4A). In response to poly(I:C), a substantial number of RNAs significantly decreased or increased in WT cells (Figure 4A). In contrast, only a small number of RNAs significantly increased in RL-KO cells. We then normalized RNA levels to ERCC spike-in RNAs to control for the expected gross reductions in mRNA. This was a necessary step since the number of upregulated and downregulated RNAs in WT cells post-poly(I:C), as well as the magnitude by which they changed, was overestimated and underestimated, respectively, by standard differential expression analyses (Figures S4B,C,D,E). We then calculated the differential in ERCC-normalized RNA levels between WT and RL-KO cells in response to poly(I:C) and plotted as a color gradient on the scatterplot of WT cells treated with or without poly(I:C) (Figure 4B and Data S1). These analyses revealed several important observations.

Figure 4. RNase L regulates global mRNA levels during dsRNA stress.

(A) Differential expression analysis without normalization from total RNA isolated from WT or RNase L-KO A549 cells before and six hours after poly(I:C) transfection. (Upregulated transcripts = Fold Change > 2, p-value < 0.05; downregulated transcripts = Fold Change < 0.5, p-value < 0.05). (B) Differential expression plot of mRNAs in WT cells treated with or without poly(I:C). The linear regression of the ERCC spike-in control RNAs is represented by dashed line. The color of the dots indicates their ERCC-normalized differential expression relative to RL-KO cells transfected +/− poly(I:C). Orange = RNase L-dependent decrease, blue = RNase L-dependent increase. (C) Scatterplot depicting half-life vs. expression levels (FPKM). Each transcript is color-coded by its ERCC-normalized differential expression between WT and RL-KO cells transfected +/− poly(I:C). (D) GO analysis of mRNAs that are induced in both WT and RL-KO cells post-poly(I:C) (greater than 4-fold, FPKM > 1 in either WT or RL-KO cells with poly(I:C) treatment) (Data S4).

First, we observed that poly(I:C) transfection in WT cells led to a striking reduction in essentially all abundant mRNAs, with RNA transcripts from 6,310 genes being reduced by 2-fold or more (Figure 4B and Data S2). This was most notable for abundant mRNAs with long steady-state half-lives (Figures 4C and S4F,G,H,I), which is consistent with our qRT-PCR analysis of GAPDH, Actin B, and Tubulin A mRNAs (Figure 2E). Importantly, the decrease in these mRNAs is due to RNase L since the vast majority of abundant mRNAs did not substantially change in RL-KO cells post-poly(I:C) (Figures 4A and S4C,E). Mitochondrial mRNAs (i.e. MT-ND4 and MT-ND4L) were unaffected, and in fact slightly increased in WT cells relative to RL-KO cells post-poly(I:C), effectively serving as an internal control that suggests our analyses may underestimate the level of RNase L-dependent reductions in cytosolic mRNAs (Figures 4B and S4D,E). These results confirm that RNase L promotes the degradation of the majority of abundant and stable cytoplasmic mRNAs.

Second, we identified 5101 RNA transcripts whose levels increase at least 2-fold in WT cells relative to RL-KO cells in response to post-poly(I:C) (Figure 4B, blue dots and Data S3). The majority of these RNAs are lower in abundance and have a short half-life in untreated cells (Figures 4C and S4F,G,H,I), which may be due to several explanations. First, RNase L may not increase the overall decay rate of these mRNAs since their decay rate is already high. Therefore, despite our use of the ERCC spike-in control, the increase in these RNAs may be a technical artifact, whereby their relative abundance is artificially inflated due to the reduction in the majority of abundant mRNAs. Second, RNase L activation may stabilize a subset of mRNAs with short half-lives. Finally, RNase L activation may promote the transcriptional induction of numerous genes.

Third, we identified constitutively expressed mRNAs (i.e. MYC, SRRM2, STAT1) that do not decrease in WT cells post-poly(I:C). Interestingly, STAT1 mediates interferon signaling for ISG production, and MYC is a negative regulator of IRF7 (Kim et al., 2016). We note that STAT1 has been shown to be induced by type I and type II interferons in some studies (Rusinova et al., 2013), and thus we cannot rule out that it is sensitive to RNase L, but also transcriptionally induced. Nevertheless, these data suggest that a portion of constitutively expressed mRNAs, some of which regulate the antiviral response, also escape RNase L-mediated mRNA turnover.

Fourth, we identified a population of ~100 mRNAs that substantially increases (greater than 4-fold) in WT and/or RL-KO cells in response to poly(I:C) (Figures 4B and Data S4). This population of mRNAs, which includes IFN-β, IFIT2, OAS2, IFIH1 (MDA5), DDx58 (RIG-I) and IL-6, is highly enriched for IRF3-induced mRNAs and interferon-stimulated genes (ISGs) (Figure 4D). Several of these mRNAs, such as IFN-β, are induced equally between WT and RL-KO cells (Figure 4B and Data S4), which is consistent with our smFISH and RT-qPCR analyses and suggests that these dsRNA-induced antiviral mRNAs escape RNase L-mediated mRNA turnover (Figure 3). Some mRNAs, such as the IL-6 and RAET1L mRNAs, are induced to higher levels in the RL-KO cell line suggesting they are partially sensitive to RNase L. These data suggest there is a range of RNase L resistance amongst dsRNA-induced mRNAs.

The dsRNA-induced mRNAs resistant to RNase L do not have substantial differences in GC-content, 5’ UTR length, 3’ UTR length, and total transcript length of in comparison to RNase L sensitive mRNAs (Data SI), suggesting that these features may not contribute to RNase L resistance. MEME analysis of their 3’ UTRs did not identify a notable sequence motif common to all of these mRNAs, though many of these mRNAs (31%) contain AU-rich elements (AREs) (Data S4), a common motif in mRNAs encoding cytokines (Savan, 2014). Meta-analysis of CLIP studies for RBPs did not reveal a significant enrichment in association with RBPs and these mRNAs (data not shown). However, we note that many of these mRNAs are expressed at very low levels without induction by dsRNA, and the CLIP studies were not performed during dsRNA stress.

These data demonstrate that RNase L activation promotes the turnover of abundant mRNAs in response to poly(I:C), whereas highly-induced antiviral mRNAs are partially or wholly resistant to this process.

RNase L drives translational repression via bulk mRNA turnover.

The RNase L-dependent turnover of mRNAs led us to examine if this contributes to dsRNA-induced translational repression. Currently, RNase L is thought to repress translation by degrading rRNA (Wreschner et al., 1981), or by RNase L-cleaved RNAs inhibiting translation (Donovan et al., 2017), possibly by triggering phosphorylation of eIF2 α by PKR, which is the primary eIF2 α kinase activated by dsRNA (Dalet et al., 2015). To resolve the relative contributions of these mechanisms to RNase L-driven translational repression, we evaluated dsRNA-induced translational repression in cell lines with or without RNase L and/or PKR.

Our data argue that RNase L activity drives rapid translation repression independently of PKR in the majority of A549 cells in response to dsRNA. The key observation is that metabolic labeling of nascent proteins with either ³⁵S-labeled methionine and cysteine or puromycin revealed that both WT and PKR-KO cells strongly repressed translation by two-hours post-poly(I:C), whereas RL-KO cells maintained translation (Figures 5A,B and S5A,B), consistent with previous analyses in A549 cells (Donovan et al., 2017).

Figure 5. RNase L drives translational shut-off.

(A) S-35 metabolic labeling of newly synthesized proteins in WT, RL-KO, and PKR-KO A549 cells post-poly(I:C). (B) Quantification of experiments represented in (A). Bars represent the average signal from at least two independent replicates +/− S.D. normalized to time zero. (C) Representative images from SUnSET puromycin-labeling (green) analysis of indicated cells lines four hours post-poly(I:C) transfection. Dapi-stained nuclei (blue). Scale bar is 25 μm. (D) Similar to (C) but enlarged. (E) Quantification of the mean intensity normalized to mock-treated cells of puromycin staining in cells (between 51-176 cells analyzed for each cell line) that contain RLBs/SGs four hours post-poly(I:C). (F) Quantification of the percentage of cells translating (puromycin signal greater than 80% of average signal from untreated cells) four hours post-poly(I:C). Bars represent the average +/− S.D. normalized to untreated cells from at least three independent replicates in which cells from at least five fields of view were analyzed. (G) Representative image of simultaneous SUnSET puromycin-labeling (green), PABP IF (red), and GAPDH smFISH (white) in A549 cells post-poly(I:C). Cells that do not contain RLBs or do not display PABP translocation are labeled non-responsive cells (n.r.).

The lack of translational repression in the RL-KO cells was surprising since some cells generate PKR-dependent SGs (Figure 1), which we hypothesized might be due to variation in the response of individual cells. Therefore, we examined translation in individual cells by IF for puromycin-labeled proteins (Schmidt et al., 2009). This analysis revealed that the RL-KO cells containing SGs in fact repressed bulk translation in response to poly(I:C), which was dependent on PKR since PKR/RL-dKO cells did not repress translation in response to poly(I:C) (Figure 5C,D,E,F). However, less than 30% of the RL-KO cells repressed translation in response to poly(I:C) (Figure 5F). In contrast, greater than 80% of the WT and PKR-KO cells underwent translational repression and showed RLB body formation with nuclear translocation of PABPC1. These data demonstrate that RNase L represses translation in the majority of A549 cells in response to dsRNA, whereas PKR represses translation in a minority of cells lacking RNase L, providing additional evidence for variation in how individual cells respond to dsRNA. Importantly, reduced puromycin-labeling was only observed in WT cells that underwent RNase L-dependent mRNA degradation (Figure 5G). This indicates that bulk mRNA degradation may significantly contribute to the RNase L-mediated translational arrest.

In addition to degrading mRNAs, RNase L could repress translation by degrading rRNA or by activating signaling pathways that trigger eIF2a phosphorylation independently of PKR. Examination of rRNA revealed it was largely intact at two hours post-poly(I:C) when RNase L-mediated translation repression and mRNA degradation is robust (Figures 6A and S5C). Furthermore, a parallel study to ours demonstrated that cleavage of rRNA by RNase L does not reduce the ability of ribosome to translate in extracts (Rath et al., 2019). These observations argue that rRNA degradation does not drive RNase L-driven translational repression at early times post-dsRNA, which is also supported by our analysis of IFN-β translation (see below).

Figure 6. RNase L-driven translational repression is independent of rRNA degradation and p-eIF2α.

(A) TapeStation analysis of the 28S and 18S rRNAs. (B) Immunoblot analysis of eIF2α, eIF2α-P51S (p-eIF2α), GADD34, and GAPDH in indicated A549 cell lines. (C) Quantification of the p-eIF2α:eIF2α ratio as represented in (B). Bars represent the average ratio +/− SEM from independent replicates (n=5-9). (D) S-35 metabolic in A549 cells transfected with poly(I:C) or treated with 250 uM sodium arsenite then treated with or without ISRIB (50 nM). (E) Quantification of IFN-β secretion from WT and RL-KO A549 via ELISA. Limit of quantification was 50 pg/ml. Bars represent Average +/− S.D. from three independent experiments. *** indicates p-value <0.001 as determined by student’s t-test. (F) Diagram of the eGFP expression vector in which eGFP ORF containing the IFN-β UTRs is driven by the IFN-β promoter. (G) smFISH for GAPDH and eGFP mRNAs in A549-WT cells with the IFN-β promoter-eGFP vector stably incorporated. GFP fluorescence is shown. (H) Similar to (G) but in A549- RL-KO cells. (I) Quantification of mean eGFP intensity from at least fifteen WT and RL-KO cells as represented in (G) and (H).

Examination of eIF2α phosphorylation revealed that RNase L promotes phosphorylation of eIF2α, as p-eIF2α levels were lower (2-3 fold) in cells lacking RNase L activity post-poly(I:C) (Figures 6B,C and S5D,E). This phosphorylation of eIF2α is partly independent of PKR, since PKR-KO cells also showed eIF2α phosphorylation post-poly(I:C), albeit less than in WT cells (Figures 6B,C and 5SF). Furthermore, this alternative mode of eIF2α phosphorylation is dependent on RNase L since eIF2α phosphorylation was abolished to background levels in PKR/RL-dKO cell lines. This suggests that RNase L activates an additional eIF2α kinase. RNase L also inhibits the induction of the GADD34 eIF2α phosphatase post-poly(I:C) (Figures 3C, 6B and S5G), which likely promotes elevated p-eIF2α levels. Thus, RNase L activation promotes phosphorylation of eIF2α.

To determine whether RNase L-promoted phosphorylation of eIF2α contributes to RNase L-driven translational shut-off, we quantified translational activity in WT cells following transfection of poly(I:C) in the presence or absence of ISRIB, which bypasses the inhibitory effect of p-eIF2α on translation (Sidrauski et al., 2015). Strikingly, poly(I:C)-induced translational arrest was unaffected by ISRIB treatment (Figure 6D). In contrast, ISRIB de-repressed sodium arsenite-induced translation arrest, which occurs through eIF2α phosphorylation. This demonstrates that phosphorylation of eIF2α is not the primary driver of rapid RNase L-mediated translational repression.

RNase L resistant mRNAs continue to translate during a dsRNA response

The observations above argue that RNase L-mediated mRNA turnover accounts for RNase L-dependent decrease in bulk translation. Nevertheless, it remained possible that translation is repressed by an unknown RNase L-dependent mechanism. We reasoned that if translational repression is simply due to mRNA degradation, then mRNAs that escape RNase L degradation, such as IFN-β, should still produce protein. Alternatively, if global translation is repressed by an unknown signaling mechanism or degradation of tRNAs, then IFN-β protein would not be produced despite the presence of the IFN-β mRNA.

An important observation is that WT and RL-KO cells secreted equivalent levels of IFN-β at six hours post-poly(I:C) as measured by ELISA (Figure 6E), indicating that RNase L-driven translational repression at early times post-poly(I:C) does not negatively affect IFN-β translation and secretion. These results demonstrate that RNase L-resistant mRNAs are able to be translated at early times post-dsRNA. This argues that the translation machinery is functional and translation is simply limited by the presence of mRNAs. At twelve hours post-poly(I:C), secreted IFN-β from WT cells was ~10-fold lower than from RL-KO cells (Figure 6E). This is consistent with previous findings that RNase L activity limits IFN-β secretion at later times, perhaps due to promotion of eIF2α phosphorylation, complete degradation of the translational machinery, and/or the cells entering the apoptotic pathway (Banerjee et al., 2014).

To directly demonstrate that mRNAs could be translated in cells where RNase L had degraded most cytoplasmic mRNAs, we stably transfected A549 cells with an IFN-β gene where the coding region was replaced with the eGFP ORF, referred to as I₅,G_orfI₃, mRNA (Figure 6F). In untreated cells, the I₅,G_orfI₃, mRNA and eGFP protein was essentially undetectable by smFISH and fluorescence microscopy (Figure 6G). However, transfection of poly(I:C) induced the I₅,G_orfI₃, mRNA in WT cells in which the majority of GAPDH mRNA was degraded, and those cells produced comparable, albeit slightly lower (24%), levels of GFP protein as compared to RL-KO cells eight hours post-poly(I:C) (Figure 6G,H,I). These data directly demonstrate that a cell with widespread mRNA degradation can still support translation from a transcriptionally induced mRNA that is of sufficient abundance.

ORF sequences can modulate RNase L-mediated decay

In the above experiment, the I₅,G_ORFI₃, mRNA levels were sufficient to produce eGFP protein during RNase L-mediated mRNA decay/translational repression (Figure 6I). However, the I₅,G_orfI₃, mRNA levels were 2-fold lower in WT cells in comparison to RL-KO cells by at by eight hours post-poly(I:C), whereas endogenous IFN-β mRNA levels were equivalent (Figure 7A,B,C and Figure S6A). This observation suggested that the I₅,G_ORFI₃, transcript is more sensitive to RNase L-mediated mRNA decay than the endogenous IFN-β mRNA even when expressed from the IFN-β promoter. In principle, the increased RNase L sensitivity of the I₅,G_ORFI₃, mRNA could be because the IFN-β mRNA contains a protective element within its ORF, or because the eGFP ORF is more sensitive to RNase L cleavage.

Figure 7. IFN-β mRNA escapes RNase L-mediated decay via resistance to RNase L and RNase L-promoted transcriptional induction.

(A) Diagrams of endogenous IFN-β gene and eGFP constructs driven by the IFN-β promoter. (B) smFISH for IFN-β (top images) and eGFP (lower images) mRNAs from expression constructs depicted directly to the left in (A) eighth hours post-poly(I:C). IFN-β smFISH was performed in WT and RL-KO A549 cells with the IFN-β_pI₅,G_ORFI₃ construct stably incorporated. (C) Quantification of smFISH foci per cell as represented in the images in (B) directly to the left. (D) Diagram of CMV promoter-driven eGFP. (E) smFISH for eGFP mRNA driven by the CMV promoter in WT cells with or without poly(I:C) transfection. smFISH for GAPDH mRNA and GFP fluorescence in WT and RL-KO cells is shown in Figure S7A,B. (F) Quantification of eGFP smFISH represented in (E). (G) Immunoblot analysis of p-IRF3 (rep 1). Normalized p-IRF3 band intensity from two experiments are shown below. (H) IFN-β smFISH from non-deconvolved images. Arrows mark high-intensity RNase A-and Actinomycin D-sensitive foci consistent with nascent transcripts at IFN-β loci (Figure S7D,E,F). (I) Quantification of the relative intensity of IFN-β TSS in WT and RL-KO cells represented in (H). Between 111-138 foci were analyzed from three independent experiments. (J) Model of RNase L-mediated regulation of translation via mRNA degradation escaped by antiviral mRNAs.

To distinguish between these two models, we inserted the eGFP ORF into the IFN-β mRNA (I₅,I+G_ORFI₃,) and determined how this chimeric mRNA was affected by RNase L activation. We observed that the I₅,I+G_ORFI₃, was expressed at ~4-fold lower levels in WT cells as compared to RNase L-KO cells (Figure 7A,B,C), indicating that addition of the eGFP ORF to the IFN-β mRNA sensitized it to RNase L. This demonstrates that the eGFP ORF contains elements that confer RNase L sensitivity to the IFN-β mRNA. This observation also argues that the IFN-β mRNA does not contain a dominant protective element that limits RNase L-mediated degradation.

Since the ORFs of eGFP and IFN-β affected RNase L degradation differentially, we examined if ribosome association with these mRNAs modulated RNase L-mediated mRNA degradation by treating cells with puromycin, which will release ribosomes from mRNAs, or with emetine or cycloheximide, which trap elongating ribosomes on mRNAs. We observed that the RNase L resistance of IFN-β mRNA, as well as the RNase L sensitivity of GAPDH, I₅,G_ORFI₃, and I₅,I+G_ORFI₃, mRNAs, were unaffected by the chemical perturbation of translation (Figure S6A-H). We interpret this data to suggest that the translation status of an mRNA is not a major determinant of its sensitivity or resistance to RNase L.

The partial escape of the I₅,G_ORFI₃, mRNAs from RNase L led us to assess the contribution of the IFN-β UTRs in protecting the eGFP ORF from RNase L. To test this, we swapped the IFN-β 5’ and 3’ UTRs in the I₅,G_ORFI₃’ for the 5’ and 3’ UTRs of the pcDNA3.1 plasmid (termed V₅,G_ORFV₃,) and examined expression in WT and RL-KO eight hours post-poly(I:C). We observed that the V₅,G_ORFV₃, and I₅,G_ORFI₃, mRNAs were expressed to similar levels, and both were reduced to a similar degree in WT cells as compared to RL-KO cells (Figure 7A,B,C). This argues that, at least in this context, the 5’ and 3’ UTRs of the IFN-β mRNA do not confer RNase L resistance to the eGFP mRNA.

RNase L-promoted transcription of IFN-β promotes escape from RNase L-mediated mRNA decay

The partial escape of the I₅,G_ORFI₃, and, V₅,G_ORFV₃, mRNAs driven by the IFN-β promoter from RNase L led us to investigate if transcriptional induction could contribute to mRNAs escaping RNase L degradation. To examine this, we compared the RNase L sensitivity of the V₅,G_orfV₃, mRNA constitutively expressed from the CMV promoter (Figure 7D), which is not induced by dsRNA, to the same mRNA expressed from the IFN-β promoter. We observed that CMV-driven the V₅,G_orfV₃, mRNA significantly decreased in WT but not RL-KO cells post-poly(I:C) (Figures 7E,F and S7A,B,C), indicating that the V₅,G_orfV₃, mRNA is sensitive to RNase L to a degree similar to GAPDH mRNA. In contrast, the eGFP mRNAs driven by the IFN-β promoter increased in abundance post-poly(I:C). Furthermore, the relative RNase L-dependent reduction of the V₅,G_orfV₃ mRNA was twofold less than when driven from the IFN-β promoter in comparison to the CMV promoter. These observations suggest that transcriptional activation from the IFN-β promoter counter-acts RNase L-mediated decay of the eGFP mRNAs.

One possibility is that RNase L activation promotes induction of the IFN-β promoter to off-set increased mRNA decay (Malathi et al., 2007). Indeed, we observed higher (30%) p-IRF3 levels in WT cells in comparison to RL-KO cells (Figure 7G) and the average intensity of nascent transcripts at the IFN-β transcriptional start site (TSS) was higher (24%) in WT cells in comparison to RL-KO cells as assessed by smFISH (Figures 7H,I and S7D,E,F). These data indicate that RNase L promotes transcription of antiviral mRNAs and are consistent with mathematical modeling suggesting that transcriptional induction with a stable positive feedback activator can promote escape from RNase L activation (Rath et al., 2019). Thus, these data combined support that RNase L-promoted transcriptional induction coupled with elements/structures of the IFN-β mRNA that resist RNase L cleavage are sufficient for the IFN-β mRNA to effectively escape RNase L-mediated mRNA decay (Figure 7J).

DISCUSSION

Our studies demonstrate that RNase L leads to rapid and widespread degradation of the majority of abundant cytoplasmic mRNAs in response to dsRNA (Figures 1E, 2, and 4A,B,C). Consistent with these findings, a related study demonstrated that RNase L increases decay rates of large numbers of mRNAs in response to dsRNA (Rath et al., 2019). However, despite the promiscuous nature of RNase L degradation, some mRNAs fully, or partially, escape RNase L-mediated mRNA degradation. For example, the IFN-β mRNA is expressed at similar levels in WT and RL-KO cells post-poly(I:C) by both qRT-PCR and smFISH (Figure 3), indicating that RNase L does not affect the levels of this mRNA. Indeed, from our RNA-Seq analyses, we identify numerous mRNAs that fully, or partially, escape RNase L-mediated mRNA degradation (Figure 4B). Importantly, many of these mRNAs encode for dsRNA-induced antiviral proteins and cytokines (Figure 4D).

Multiple mechanisms presumably contribute to the observation that mRNAs are differentially affected by RNase L. First, the magnitude of an mRNA being reduced by RNase L depends on its normal mRNA decay rate relative to the rate of RNase L-mediated decay. This in part explains the observation that stable mRNAs are more strongly reduced by RNase L in comparison to unstable mRNAs (Figures 4C and 4F,G,H,I). Since many RNase L-resistant antiviral mRNAs contain destabilizing elements, such as AREs (Data S4), RNase L-mediated mRNA decay would be expected to have less of an impact on their overall decay rate.

Second, some mRNAs may have evolved sequence and/or structural compositions that resist RNase L cleavage or specific elements that directly modulate RNase L activity. Notably, a conserved RNA structure in poliovirus inhibits RNase L activation (Han et al., 2007; Townsend et al., 2008), and selective pressure by RNase L has been proposed to account for the reduced frequency of UA and UU dinucleotides in HCV RNA (Washenberger et al., 2007). Similar to the latter mechanism, we observed that the RNase L-sensitive eGFP ORF is sufficient to sensitize the IFN-β mRNA to RNase L (Figures 7A,B,C,D,E,F and 7SA,B,C). This effect appears to be independent of ribosome association and translational status since we did not observe differences in RNase L sensitivity of GAPDH, IFN-β, I₅,I+G_orfI₃, and I₅,G_orfI₃ mRNAs upon chemical perturbation of translation (S6A-H). These observations suggest that RNase L targeting of mRNA is not strongly associated with ribosomes, which leads to two important implications. First, bulk mRNAs can be degraded by RNase L in cells undergoing translational repression via PKR-mediated phosphorylation of eIF2α. Second, antiviral mRNAs that are resistant to RNase L cleavage would be permitted to be translated without being targeted for degradation. Combined, these observations argue that the IFN-β mRNA may have evolved in sequence composition/structure to reduce cleavage by RNase L. Indeed, inherent differences in RNase L-mediated mRNA degradation rates are observed in cell free systems (Rath et al., 2015). Since mRNAs typically fold well in vitro in a manner that reflects the in vivo structure (Beaudoin et al., 2018), this suggests that mRNAs can be differentially degraded by RNase L simply based on their structure and the accessibility of RNase L cleavage sites.

Finally, transcriptional rates of IRF3-induced mRNAs may out-pace RNase L-mediated mRNA decay rates of these mRNAs. Indeed, expression of RNase L-sensitive eGFP mRNA from the IFN-β promoter is sufficient to allow partial escape from RNase L (Figures 7A-G and S7A,B,C). Furthermore, our studies and previous studies support that RNase L promotes IRF3 phosphorylation and transcriptional induction of IFN-β (Figures 7G,H,I and S7D,E,F) (Malathi et al., 2007), which could compensate for increased mRNA decay. Consistent with this, kinetic analyses of the dsRNA response and RNase L-mediated mRNA decay links increased stability of antiviral mRNAs to transcriptional positive feedback of the interferon response, whereby RNase L-resistant molecules (i.e. IFN proteins or p-STAT protein) protect non-resistant molecules (i.e. mRNAs) from RNase L (Rath et al., 2019). In this view, a combination of inherent mRNA instability, structural resistance to RNase L cleavage, and transcriptional induction promoted by RNase L and/or IFN signaling likely contributes to the ability of antiviral mRNAs to escape RNase L-mediated mRNA decay.

Several of our observations argue that one consequence of RNase L-mediated mRNA degradation is host shut-off of translation. First, we observed that RNase L catalytic activity represses translation by two hours post-dsRNA (Figures 5A,B and S5A,B), when rRNA is largely intact (Figures 6A and S5C), but bulk mRNA levels are substantially reduced (Figures 2 and S2G). Second, RNase L-dependent bulk translational repression is unaffected by ISRIB or knockout of PKR (Figures 5A,B, 6D and S5A.B), indicating RNase L-driven translational repression occurs independently of eIF2α phosphorylation. Third, reduced puromycin labeling of cells correlates with RNase L-dependent loss of GAPDH mRNA and PABP translocation (Figure 5G). Finally, because the IFN-β mRNA escapes RNase L-driven mRNA turnover and IFN-β protein is produced at similar levels in WT and RL-KO cells prior to robust rRNA degradation (Figures 3 and 6E), this demonstrates that the presence of a stable mRNA is sufficient to produce protein, and therefore the translation machinery is functional. Consistent with this latter finding, a related study demonstrated that antiviral cytokines are produced during RNase L-mediated translational arrest (Chitrakar et al., 2019). Combined, these observations reveal a new mode of RNase L-driven translational repression that occurs via bulk mRNA turnover and permits the translation of key RNase L-resistant antiviral mRNAs, such as the IFN-β mRNA (Figure 7J).

A second consequence of robust mRNA degradation by RNase L is inhibition of SG assembly concurrent with the assembly of small punctuate SG-like bodies (RLBs) and PABPC1 translocation to the nucleus (Figure 1). Interestingly, PABPC1 translocation and small punctate SGs that resemble RLBs, which are both hallmarks of RNase L activation revealed in our studies, have been observed during infection of numerous and diverse viruses (Lee and Glaunsinger, 2009, Kumar and Glaunsinger, 2010, Gray et al., 2015, Montero et al., 2008, Dobrikova et al., 2010, Borah et al., 2011, Montero et al., 2008, Fros et al., 2012, Khaperskyy et al., 2014). However, in contrast to previous studies (Dalet et al., 2017), we did not observe strong enrichment of antiviral mRNAs in SGs or RLBs (Figures 3 and S3), and the presence or absence of SGs and RLBs did not appear to markedly affect IFN-β production at early times post-poly(I:C) (Figure 6E). Therefore, an important focus of future work will be to determine the functions of RNase L-mediated inhibition of SGs, RLBs, and RNase L-mediated PABP translocation during the dsRNA/antiviral response.

We suggest that the activation of RNase L and widespread degradation of host mRNAs acts in concert with other aspects of the dsRNA response. Specifically, the sensing of dsRNA by PRRs leads to: 1) the activation of IRF3, which promotes the transcription of antiviral genes such as type I interferons; 2) the activation of RNase L, which initiates widespread degradation of both host and viral RNAs; 3) PKR-and RNase L-mediated promotion of eIF2α phosphorylation changing the specificity of translation initiation; 4) Inhibition of canonical translation re-initiation via RNase L-mediated downregulation of the GADD34 p-eIF2α phosphatase (Figures 6B and 5SG). In the context of these four outcomes, the majority of IRF3 induced genes appear to be resistant to RNase L-mediated mRNA turnover, and this allows for their selective translation. Thus, RNase L-mediated degradation of cellular mRNAs leads to reprograming of the produced proteins to only those mRNAs resistant to RNase L activity.

Strikingly, by examining single cells, we observe that different aspects of the cellular response to dsRNA can be triggered independently in individual cells. We observe numerous instances where RNase L activation and PKR activation are uncoupled from induction of IRF3 targets, such as the IFN-β mRNA (Figures 3A and S3A,B). The ability to independently activate different arms of the cellular response to dsRNA allows cells to modulate their transcriptional and translational output, as well as to determine whether to enter apoptosis. An important aspect of future work will be to determine the additional inputs that affect the independent activation of these pathways and how those affect the cellular response to viral infection.

STAR METHODS

LEAD CONTACT AND METHODS AVAILABILITY

Information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Roy Parker (roy.parker@colorado.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

A549, U-2 OS, and A549-PKR-KO cell lines were provided by Dr. Christopher Sullivan, Dr. Nancy Kedersha, and Dr. Susan Weiss, respectively. Cells were maintained at 5% CO₂ and 37 degrees Celsius in Dulbecco’s modified eagle’ medium (DMEM) supplemented with fetal bovine serum (FBS; 10% v/v) and penicillin/streptomycin (1% v/v). Cells were routinely tested for mycoplasma contamination by the cell culture core facility and were negative for mycoplasma contamination throughout the study. Cells were transfected with poly(I:C) HMW (InvivoGen: tlrl-pic) using 3-ul of lipofectamine 2000 (Thermo Fisher Scientific) per 1-ug or poly(I:C).

METHODS DETAILS

Plasmids

To generate the 3xflag-tagged RNase L expression vectors used for transient transfections, the RNase L and RNase L-R667 coding sequences were tagged with 3x-flag encoding sequence via PCR amplification from the pLenti-CMV-RNase L and pLenti-RNase L-R667A vectors using the RL-3xFlag-Kpn1/Xho1 primers (Data S5) and Phusion polymerase (New England BioLabs). The PCR amplicons were digested with Kpn1 and Xho1 and ligated into the Kpn1/Xho1 sites of pcDNA3.1-puro using T4 DNA ligase (New England BioLabs). To generate pLenti-EF1-Blast lentiviral genomes encoding RNase L and RNase L-R667A, the RNase L and RNase L-R667A coding sequences were N-terminally-tagged with 3x-flag tag via PCR amplification expression vectors using RL-3x-flag-fusion primers (RL_sen_3x_FLAG, RL_sen_kpn1_kozak_flag, and RL_anti_EcoN1) (Data S5) and inserted into the Xho1/EcoN1 sites of the pLenti-EF1-Blast vector using In-Fusion (Clontech). The pVSV-G, pRSV-Rev, and pMDLg/pRRE plasmids were a gift from Dr. Sabrina Spencer. The px458 Cas9 vector (Addgene: 48138) was used to generate knockout cell lines. The CRISPR/Cas9 guide RNAs were designed using the Integrated DNA Technologies (IDT) CRISPR guide target design tool. Overlapping oligos (RL sgRNA 1 sense an RL sgRNA 1 antisense (Data S5) were annealed in T4 DNA ligase buffer and ligated into the Bbs1 sites in px458 using T4 DNA ligase. To generate the pIFNB1-eGFP chimera plasmids, the pcDNA3.1-puromycin plasmid was digested with BglII and Xba1, which removed the CMV promoter. The eGFP ORF and IFNB-GFP ORF chimeras were generated by a gblock (IDT, Data S5) encoding the IFN-B promoter, either the eGFP ORF or IFNB1-eGFP chimeric ORF, and the IFNB UTRs. These were fused into the pcDNA3.1-puromycin vector at the bglII and xba1 sites using in-fusion cloning (Clontech). A xho1 site was included between the eGFP ORF and IFNB1-3’UTR to allow for manipulation of the vector as needed. To generate the pIFNB1-eGFP-Vector UTR chimera, a gblock encoding the IFNB1 promoter with the pcDNA3.1+ vector 5’ UTR that overlapped with the 5’end of the eGFP ORF (Data S5), a PCR amplicon from the pLMJ1-eGFP plasmid of the eGFP ORF overlapping with the 5’-UTR and 3’-UTRs of pcDNA3.1 at the 5’-and 3’ ends, respectively, was fused into the pcDNA3.1-puromycin vector at the bglII and xba1 sites using in-fusion cloning (Clontech). To generate the CMV-driven eGFP construct, the eGFP ORF PCR amplicon was fused into the Nhe1/Xho1sites of pcDNA3.1. Plasmids were sequence verified via sanger sequencing (Quintarabio or Genewiz).

Antibodies

Mouse anti-RNase L Antibody 2E9 (Novus Biologicals: NB100-351) was use at 1:1500 for immunoblot analyses. Rabbit anti-IRF3 (Abcam: ab25950) was used at 1:1000 for IB analysis. Rabbit anti-phospho-IRF3 (Ser396) (D601M) (Cell signaling Technology: 29047S) was used at 1:1000 for IB analysis. Mouse monoclonal anti-G3BP antibody (Abcam: ab56574) was used at 1:500 for IFA and 1:1000 for IB analyses. Rabbit anti-EIF2S1 (Phospho S51- eIF2α) monoclonal antibody (Abcam: ab32157) was used at 1:500 for IB analysis. Rabbit anti-eIF2α (Cell Signaling Technology: CST9722S) was used at 1:1000 for IB analysis. Mouse monoclonal anti-FLAG (Sigma Aldrich: F1804) was used at 1:1000 for IB analysis. Rabbit anti-GAPDH (Cell Signaling Technology: 2118L) was used at 1:2000 for IB analysis. Rabbit anti-PKR (Cell Signaling Technology: 12297S) was used at 1:1000 for IB analysis. Rabbit polyclonal anti-PABP antibody (Abcam: ab21060) was used at 1:500 for IFA. Goat anti-mouse IgG H&L FITC (Abcam: ab97022) was used at 1:1000 for IFA. Goat anti-rabbit IgG H&L Alexa Fluor 647 (Abcam: ab150079) was used at 1:1000 for IFA. Anti-rabbit IgG, HRP-linked antibody (Cell Signaling Technology: 7074S) was used at 1:3000 for IB analysis. Anti-mouse IgG, HRP-linked antibody (Cell Signaling Technology: 7076S) was used at 1:10,000 for IB analysis. Anti-puromycin (MilliporeSigma: MABE343) was used at 1:1000 for SUnSET analyses.

Generation of knockout cell lines

To generate RNase L knockout A549 and U-2 OS lines, cells (T-25 flask; 70% confluent) were co-transfected with 2-ug of px458-RL and 200-ng of pcDNA3.1-puro using 6-ul of Lipofectamine 2000 (Thermo Fisher Scientific) according to manufacturer’s instructions. Twenty-four hours post-transfection after Cas9-GFP expression was observed via fluorescent microscopy, the medium was replaced with medium containing 2ug/ml of puromycin. Selective medium was replaced 3 days post-transfection. Five days post-transfection, selective growth medium was replaced with normal growth medium. When cells became 80% confluent, cells were serial diluted and plated on 15-cm dishes. Individual colonies were isolated, propagated, and screened via immunoblot analysis.

Generation of Lentiviral particles

To generate the RNase L and RNase L-R667A lentiviral particles, HEK293T cells (15-cm dish; 80% confluent) were co-transfected with either 11.7-ug of pLenti-EF1-RNase L-blast and pLenti-EF1-3xflag-RNase L-R667A-blast, 3.5-ug of pVSV-G, 2.9-ug of pRSV-Rev, and 5.6-ug of pMDLg-pRRE using 100-ul of lipofectamine 2000. Medium was replaced 6 hours post-transfection. Medium was collected at twenty-four and forty-eight hours post-transfection and filter-sterilized with a 0.45-um filter.

Generation of stable cell lines

To reconstitute A549-RL-KO cells with either 3xflag-RNase L or 3xflag-RNase L-R667A, A549-RL-KO cells were seeded in T-25 flask. When 80% confluent, cells were incubated for 1 hour with 1-ml of either 3xflag-RNase L or 3xflag-RNase L-R667A lentiviral particles containing 10-ug of polybrene with periodic rocking. Normal medium was then added to the flask and incubated for twenty-four hours. Medium was removed 24 hours post-transduction and replaced with selective growth medium containing 10-ug/ml of Blasticidine S hydrochloride (Sigma-Aldrich). Selective medium was changed every three days. After one-week, selective medium was replaced with normal growth medium. Expression of RNase L was confirmed via immunoblot analysis. To generate the stable eGFP/IFNB chimera lines, 1-ug of each plasmid was linearized with pvuII, purified, and then transfected into cells using lipofectamine 2000. Forty-eight hours later, cells were passaged into medium containing 2-ug/ml of puromycin, and cells were maintained in selective medium. To generate cell lines with stable incorporation of the IFN-β /eGFP chimeric constructs, plasmids were linearized with Pvu1, purified, and 1-ug was transfected into cells (T-25 Flask; 50% confluent) using lipofectamine 2000. Forty-eight hours later, cells were passage into selective medium containing 2-ug/ml of puromycin. Cells were maintained in selective media.

Immunoblot analyses

To screen and confirm for knockout or reconstitution of proteins, cells were lysed in SDS solution (1% SDS, 2% β-mercaptoethanol) by boiling for 10 min followed by 1 min of vortexing. Equal volumes of lysates were fractionated on 4-12% Bis-Tris Protein gels (Thermo Fisher Scientific) in MOPS buffer and transferred to nitrocellulose membrane (GE Healthcare). Membranes were blocked in 5% BSA in TBST. Membranes were then incubated with primary antibodies for 2 hours at room temperature or overnight at 4 degrees Celsius. After washing, membranes were incubated with HRP-linked anti-rabbit IgG or anti-mouse IgG secondary antibodies for 1 hour at room temperature. After washing, membranes were incubated with ECL substrates (Thermo Fisher Scientific: 32106) for 1-5 minutes. Membranes were then stripped using Restore western blot stripping buffer (Thermo Fisher Scientific: 21059) and re-blocked with 5% BSA in TBST. Membranes were then incubated with anti-GAPDH, washed, incubated with HRP-linked anti-mouse antibody, washed, and incubated in ECL substrate. For quantitation of p-eIF2α, cells were seeded in 12-well format. Cells were transfected with 250-ng of poly(I:C) and cell lysates were collected at indicated time post-transfection. A fifth of the lysate (12-ul/60-ul) was fractionated on 4-12% Bis-Tris Protein gels in MOPS buffer and transferred to nitrocellulose membrane. Membranes were blocked in 5% MILK in TBST. Membranes were then incubated with rabbit anti-EIF2S1 (Phospho S51- eIF2α) antibodies overnight at 4 degrees Celsius. After washing, membranes were incubated with HRP-linked anti-rabbit antibody for 1 hour at room temperature. After washing, membranes were incubated with SuperSignal West Femto Maximum Sensitivity ECL substrate (Thermo Fisher Scientific: 34095) for 1-5 minutes. Membranes were stripped and re-probed using with rabbit anti-eIF2α, washed, incubated with HRP-linked anti-rabbit antibody for 1 hour at room temperature, washed, incubated with SuperSignal West Femto Maximum Sensitivity ECL substrate for 1-5 minutes. Photographs of membranes were taken using ImageQuant LAS 4000 (GE Healthcare) and analyzed using ImageJ with Fiji plug-in. For quantification of western blots, average band signal was determined using the measure function in imageJ.

Sequential immunofluorescence and single molecule FISH

Sequential immunofluorescence and smFISH was performed following manufacturer’s protocol (https://biosearchassets.blob.core.windows.net/assets/bti_custom_stellaris_immunofluorescence_seq_protocol.pdf). Ship ready GAPDH smFISH probes labeled with Quasar 570 Dye were purchased from Stellaris. Custom AHNAK and NORAD smFISH probes labeled with Quasar 670 dye were designed and purchased from Stellaris and are described in (Khong et al., 2017). Custom IFNB1, eGFP, and IL-6 smFISH probes (Data S6) were designed using Stellaris smFISH probe designer (Biosearch Technologies) available online at http://www.biosearchtech.com/stellaris-designer. Reverse complement DNA oligos were purchased from IDT. The IFNB1 smFISH probes were labeled with Atto-633 using ddUTP-Atto633 (Axxora: JBS-NU-1619-633) and terminal deoxynucleotidyl transferase (Thermo Fisher Scientific: EP0161) as described in (Gaspar et al., 2017). Oligo d(T)30-Cy3 were purchased from IDT.

Microscopy and Image Analysis

Immunofluorescence and smFISH with DAPI staining were imaged using a wide field DeltaVision Elite microscope with a 100X objective using a PCO Edge sCMOS camera. For IFA, 10 Z sections at 0.3 um/section were taken for each image. For IFA/smFISH, 15 Z planes at 0.2 um/section were taken for each image. Deconvoluted images were processed using ImageJ with FIJI plugin. Z-planes were stacked and minimum and maximum display values were set in ImageJ for each channel to properly view fluorescence. Quantification SGs was determined using Imaris Image Analysis Software (Bitplane) (University of Colorado-Boulder, BioFrontiers Advanced Light Microscopy Core). Live cell imaging was performed using a Nikon Spinning Disk Confocal microscope outfitted with an environmental chamber with O₂, CO₂, temperature, and humidity control (University of Colorado-Boulder, BioFrontiers Advanced Light Microscopy Core). All images were acquired using a 2x Andor Ultra 888 EMCCD camera.

RT-qPCR

WT and RL-KO A549 cells (12-well; 60% confluent) were transfected with or without poly(I:C). Six hours post-transfection, RNA was extracted, treated with DNase I (NEB) for 15 minutes, and re-purified via ethanol (75%) sodium acetate (0.3M) precipitation, and re-suspended in 15-ul of water. Equal volumes (400-ng of WT RNA) were then reverse transcribed using super script III reverse transcriptase (Thermo Fisher Scientific) and polydT(₂0) primer (Integrated DNA Technologies). cDNA was diluted to 100-ul. 2-ul of cDNA was added to qPCR reaction containing iQ SYBR green master mix (Bio-Rad) and 10-pmol of gene-specific primers, which are listed in Data S5. Reactions were run in duplicate or triplicate (technical replicates) on CFX96 qPCR machine (Bio-Rad) using standard two-step cycle. PCR fragment sizes were confirmed by ethidium bromide staining and RT- controls were included to demonstrate that amplification was from cDNA and not gDNA.

Metabolic labeling of newly synthesized proteins

Wild-type, RNase L knockout or PKR knockout A549 cells were transfected with poly(I:C) as described above. S35 metabolic labeling of nascent proteins was performed as described in Moon and Parker, 2018. Briefly, cells were incubated with ³⁵S labeled met and cys (EXPRE35S35S Protein Labeling Mix, PerkinElmer) in labeling medium (DMEM lacking met and cys (Sigma Aldrich) supplemented with 10% dialyzed FBS (Sigma Aldrich), glutaMAX (Gibco) and 1% streptomycin/penicillin) for 30 minutes at two- and four- hours post-transfection following a 30 minute incubation in labeling medium to deplete intracellular amino acid stores. Cells were harvested in NP-40 lysis buffer with protease and phosphatase inhibitors (Cell Signaling Technologies), lysed and equal volumes of lysate run on NuPAGE 4-12% Bis-Tris protein gels (Thermo Fisher Scientific). Gels were exposed to phosphor screens and imaged on a typhoon FLA 9500 phosphorimager. The average relative translation activity was determined using ImageJ (Fiji) (Schindelin et al., 2012) to quantify total lane intensity for each sample from 2-3 independent experiments.

SUnSET was performed as described by Schmidt et al., 2009. Briefly, puromycin (10 ug/ml) was added to cells 10 minutes prior to fixing or harvesting lysates. For IF of puro-labeled cells, cells were imaged on 60X objective. Between 5 and 10 fields of view were imaged. The fluorescent intensity of puromycin labeling of individual cells was measured using fiji software.

IFN-β ELISA

WT and RL-KO A549 cells (6-well format, 1 ml or medium, 70% confluency) were transfected with poly(I:C). At six- and twelve- hours post-poly(I:C), 50-ul of medium was removed from well and immediately assayed via ELISA using IFN beta human ELISA kit (Thermo Fisher Scientific; 414101) using manufacturer’s instructions. Time zero was taken by removing 50-ul of medium prior to poly(I:C) transfection.

High-throughput RNA sequencing

WT and RL-KO A549 cells (100-mm dish, 60% confluent) were transfected with or without 500-ng/ml of poly(I:C). Three independent replicates were performed. Total RNA was isolated six hours post-transfection. The ERCC spike-in was added according to manufacturer’s instructions and RNA libraries were prepared by the BioFrontiers Sequencing Core at CU Boulder. Libraries were sequenced on Illumina NextSeq 500. Read quality was assessed using fastqc. Illumina adaptors were trimmed using Trimmomatic 0.32 in paired and (PE) mode (Bolger et al., 2014). An index genome was acquired from GENCODE (Release 19 GRCh37.p13). Reads were aligned using Tophat (version 2.0.6) and Bowtie2 (version 2.0.2) and the following parameters:–b2 –fast –microexon-search (Kim et al., 2013; Langmead and Salzberg, 2012). Differential expression analysis was performed using Cuffdiff (version 2.2.1) with the default parameters (Trapnell et al., 2013). Gene ontology biological processes were derived from Gene Ontology Consortium enrichment analysis (Mi et al., 2016; http://www.geneontology.org/). MEME analyses were performed using MEME suite (Bailey et al, 2009). Meta-analysis of CLIP sites were derived from Yang et al., 2015. Ensembl Biomart was used for retrieval of transcript characterization (Kinsella et al., 2011). Au-rich elements were predicted using ARED (https://brp.kfshrc.edu.sa/ared/Home/BasicSearch) (Bakheet et al., 2018).

QUANTIFICATION AND STATISTICAL ANALYSIS

Student’s t-test was used to determine statistical significance. The exact value of n, which either represents independent experiments or individual cells, for each analysis can be found in the figure legends. On graphs, * indicates p-value <0.05, ** indicates p-value < 0.005, and *** indicates p-value <0.001 as determined by student’s t-test.

DATA AND CODE AVAILABILITY

Raw sequencing data are available for download GEO: GSE124144. Unprocessed images for microscopy and western blot images are deposited in the Mendeley database (http://dx.doi.org/10.17632/khkbgmkt4d.1)

Supplementary Material

1

Data S1. Table of processed RNA-seq data, related to Figures 4 and S4.

2

Data S2. Table of RNA transcripts reduced 2-fold or greater by RNase L, related to Figure 4.

3

Data S3. Table of RNA transcripts increased 2-fold or greater by RNase L, related to Figure 4.

4

Data S4. Table of mRNAs induced in response to poly(I:C), related to Figure 4.

5

Data S5. Table of oligonucleotides, related to Figures 1, 6, 7, S1, and S7.

6

Data S6. Table of custom-designed smFISH probes, related to Figures 3, 6, 7, S3, S6, and S7.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-RNase L Antibody 2E9	Novus Biologicals	NB100-351
Rabbit anti-IRF3	Abcam	ab25950
Rabbit anti-phospho-IRF3 (Ser396) (D601M)	Cell signaling Technology	29047S
Mouse monoclonal anti-G3BP antibody	Abcam	ab56574
Rabbit anti-EIF2S1 (Phospho S51- eIF2α)	Abcam	ab32157
Rabbit anti-eIF2α	Cell Signaling Technology	CST9722S
Mouse monoclonal anti-FLAG	Sigma Aldrich	F1804
Rabbit anti-GAPDH	Cell Signaling Technology	2118L
Rabbit anti-PKR	Cell Signaling Technology	12297S
Rabbit polyclonal anti-PABP	Abcam	ab21060
Goat anti-mouse IgG H&L FITC	Abcam	ab97022
Goat anti-rabbit IgG H&L Alexa Fluor 647	Abcam	ab150079
Anti-rabbit IgG, HRP-linked antibody	Cell Signaling Technology	7074S
Anti-mouse IgG, HRP-linked antibody	Cell Signaling Technology	7076S
Anti-puromycin	MilliporeSigma	MABE343
Chemicals, Peptides, and Recombinant Proteins
Sodium (meta)arsenite	Sigma-Aldrich	S7400
Poly(I:C) HMW	InvivoGen	tlrl-pic
Lipofectamine 2000	Thermo Fisher Scientific	11668027
Trizol LS reagent	Thermo Fisher Scientific	10296-028
Blasticidine S hydrochloride	Sigma-Aldrich	15205
ddUTP-Atto633	Axxora	JBS-NU-1619-633
terminal deoxynucleotidyl transferase	Thermo Fisher Scientific	EP0161
Polybrene	Millipore	TR-1003-G
iQ SYBR green master mix	Bio-Rad	1708880
Super script III reverse transcriptase	Thermo Fisher Scientific	18080085
Critical Commercial Assays
IFN beta human ELISA kit	Thermo Fisher Scientific	414101
Deposited Data
Raw and processed sequencing files	This study	GEO: GSE124144
Unprocessed image files, Mendeley Dataset DOI	This study	http://dx.doi.org/10.17632/khkbgmkt4d.1
Experimental Models: Cell Lines
A549 cells	Burke et al., 2016
U-2 OS cells	Kedersha et al., 2014
A549-PKR-KO cells	Li et al., 2017
HEK 293T cells	Burke et al., 2016
Oligonucleotides
Stellaris FISH Probes Quasar 570 Dye GAPDH	Biosearch Technologies	SMF-2026-1
Custom IFNB1, IL-6, eGFP Stellaris FISH Probes	Biosearch Technologies	Data S6
Oligo d(T)30-Cy3	IDT	N/A
qRT-PCR primers	IDT	Data S6
NORAD and AHANK smFISH probes	Biosearch Technologies	Khong et al., 2017
Recombinant DNA
px458 Cas9 vector	Addgene	48138
Software and Algorithms
Trimmomatic 0.32	Bolger et al., 2014	http://www.usadellab.org/cms/?page=trimmomatic
HTSeq	Anders et al., 2015	https://htseq.readthedocs.io/en/release_0.9.1/
Bowtie2 2.0.2	Langmead and Salzberg, 2012	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Tophat 2.0.6	Kim et al., 2013	http://ccb.jhu.edu/software/tophat/index.shtml
Cuffdiff 2.2.1	Trapnell et al., 2013	http://cole-trapnell-lab.github.io/cufflinks/cuffdiff/
Imaris Image Analysis Software 8.4.1	Bitplane	http://www.bitplane.com/
ARED	Bakheet et al., 2018	https://brp.kfshrc.edu.sa/ared/Home/BasicSearch
Gene Ontology Consortium enrichment analysis	Mi et al., 2016	http://www.geneontology.org/

Highlights.

RNase L promotes rapid and widespread decay of basal mRNAs.

RNase L limits SG assembly via turnover of SG-associated mRNAs.

Degradation of mRNAs accounts for RNase L-mediated shut-off of translation.

Antiviral mRNAs escape RNase L, permitting their translation during host shut-off.