RNase L-induced bodies sequester subgenomic flavivirus RNAs to promote viral RNA decay

SUMMARY

Subgenomic flavivirus RNAs (sfRNAs) are structured RNAs encoded by flaviviruses that promote viral infection by inhibiting cellular RNA decay machinery. Herein, we analyze sfRNA production and localization using single-molecule RNA fluorescence in situ hybridization (smRNA-FISH) throughout West Nile virus, Zika virus, or dengue virus serotype 2 infection. We observe that sfRNAs are generated during the RNA replication phase of viral infection in the cytosol and accumulate in processing bodies (P-bodies), which contain RNA decay machinery such as XRN1 and Dcp1b. However, upon activation of the host antiviral endoribonuclease, ribonuclease L (RNase L), sfRNAs re-localize to ribonucleoprotein complexes known as RNase L-induced bodies (RLBs). RLB-mediated sequestration of sfRNAs reduces sfRNA association with RNA decay machinery in P-bodies, which coincides with increased viral RNA decay. These findings establish a functional role for RLBs in enhancing the cell-mediated decay of viral RNA by sequestering functional viral RNA decay products.

Graphical abstract

In brief

Watkins and Burke show that the assembly of antiviral biomolecular condensates known as RNase L-induced bodies sequester subgenomic flavivirus RNA (sfRNA) during West Nile virus, dengue virus, or Zika virus infection. This reduces sfRNA association with mRNA decay machinery in processing bodies, which promotes the degradation of viral RNAs.

INTRODUCTION

Mosquito-borne flaviviruses, such as dengue virus (DENV), West Nile virus (WNV), and Zika virus (ZIKV), cause severe and fatal diseases in humans, including hemorrhagic fever, microcephaly, and encephalitis.^1–3 Due to their expanding territory^4–7 and lack of vaccines approved for humans,⁸ understanding fundamental host-pathogen interactions between mammalian cells and flaviviruses will be critical for developing treatments for flavivirus-associated pathogenesis.

Flavivirus genomes are positive-sense, single-stranded RNAs (+ssRNA) ~11 kb in length that encode 5′ and 3′ untranslated regions (UTRs) and an open reading frame encoding a polyprotein.^9–15 During infection in both mosquitoes and mammalian cells, the structured 3′ UTR is generated as an independent fragment termed subgenomic flavivirus RNA (sfRNA).¹⁶ The sfRNA structures contribute to the cytopathic effect in cultured mammalian cells and pathogenicity in mouse models.¹⁶ Currently, sfRNAs are widely thought to promote viral replication by sequestering cellular mRNA decay machinery to reduce degradation of viral RNA. This is based on studies showing that the conserved three-helix junction knot stalls XRN1 during 5′→3′ decay of the viral genome.^17–20 Because XRN1 is the primary 5′→3′ cytoplasmic RNA decay exoribonuclease,^21–26 sfRNA-mediated stalling of XRN1 reduces the ability of XRN1 to degrade RNA in vitro and leads to a reduction in bulk cellular mRNA decay during infection.¹⁸ Moreover, sfRNAs interact with additional components involved in mRNA decay, including DDX6 and EDC3.²⁷ Lastly, sfRNAs have been shown to localize to processing bodies (P-bodies), which are cytosolic ribonucleoprotein complexes proposed as sites of mRNA decay.^28,29

Ribonuclease L (RNase L) is a latent endoribonuclease in mammalian cells that is activated upon viral infection and degrades nearly all cellular mRNAs.^30–35 Concurrent to the initiation of cellular mRNA decay, cells assemble cytoplasmic biomolecular condensates known as RNase L-induced bodies (RLBs).³⁴ RLBs concentrate poly(A)+ RNA and several RNA-binding proteins, including poly(A) binding protein C1 (PABPC1), G3BP1, and UBAP2L.^36,37 Notably, many of these RNA-binding proteins can also localize to stress granules (SGs), which are larger ribonucleoprotein complexes composed of non-translating mRNAs that form when translation initiation is repressed by the phosphorylation of eIF2α.^38,39 However, RLBs are distinct from SGs based on several criteria. First, unlike SGs, RLBs do not require protein kinase R (PKR)-mediated phosphorylation of eIF2α or G3BP1-mediated RNA condensation for their assembly in response to double-stranded RNA (dsRNA).^36,40 Second, RLBs have a distinct proteome in comparison to SGs. Third, RLBs stably associate with P-bodies,³⁶ whereas SG association with P-bodies is transient.⁴¹ Lastly, RLBs typically lack intact mRNAs but concentrate poly(A)+ RNA, suggesting that they enrich for 3′ end decay fragments of mRNAs.³⁶ In contrast, SGs enrich for long untranslating mRNAs.⁴² Although RLBs have been hypothesized to play a role in mRNA decay,³⁷ their composition and functions during viral infections are still unclear.

Herein, we analyze sfRNA production and localization during DENV-2, ZIKV, and WNV infection using single-molecule super-resolution confocal microscopy. We observe that sfRNAs and longer viral 3′ end decay intermediates localize to RLBs. RLB assembly reduces sfRNA localization with mRNA decay machinery in P-bodies, and this coincides with robust XRN1-mediated decay activity and degradation of viral RNAs. These findings establish a function for RLBs, whereby RLBs promote the degradation of viral RNA genomes by sequestering viral RNA products that antagonize cellular RNA decay machinery.

RESULTS

Single-molecule RNA visualization of sfRNAs

To examine sfRNA production and localization during flavivirus infection, we generated single-molecule RNA fluorescence in situ hybridization (smRNA-FISH) probe sets that target the sfRNA-encoding regions of DENV-2, WNV, or ZIKV (Figure 1A). To differentiate sfRNAs from the full-length genomes or longer 3′-degradation fragments, we generated probe sets specific to the first 1 kb (5′-coding sequence [CDS]) or the last 1 kb (3′-CDS) of the coding sequence for each virus. The 5′-CDS, 3′-CDS, and sfRNA probes were labeled with ATTO-488, ATTO-550, and ATTO-633, respectively, to allow for simultaneous imaging of viral RNA species.⁴³

Figure 1. sfRNAs diffusely localize in the cytosol during early infection.

(A) Schematic showing regions of the flavivirus genome and target locations of smRNA-FISH probes (not to scale) and examples of different RNA decay species.

(B) Co-staining smRNA-FISH for WNV 5′-CDS, 3′-CDS, and sfRNAs at indicated time points of A549 cells. White arrows represent smRNA puncta that co-stain with the 5′-CDS, 3′-CDS, and sfRNA probes (full-length genomes). Cyan arrows indicate smRNA puncta that co-stain with 3′-CDS and sfRNA probes (3′-degradation intermediates). Yellow arrows indicate smRNA puncta that only stain for sfRNA (independent sfRNA fragment). Green arrows represent smRNA puncta that only stain with 5′-CDS probes (5′ end degradation intermediate).Scale bar, 10 μm. Inset scale bar, 2 μm.

(C) smRNA-FISH for WNV sfRNAs. Yellow arrows point to an individual sfRNA molecule. White arrows point to larger cytoplasmic structures composed of multiple sfRNA molecules. Nuclei (blue) are stained with 4′,6-diamidino-2-phenylindole (DAPI). Scale bar, 10 μm. Inset scale bar, 1 μm.

We first tested if each probe set (5′-CDS, 3′-CDS, sfRNA) detected viral RNA by individually staining A549 cells with each probe set 24 h post-infection (p.i.) (Figure S1A). While none of the probe sets stained mock-infected A549 cells (Figure S1A), each probe set intensely stained a region proximal to the nucleus 24 h p.i. with DENV-2, ZIKV, or WNV (Figure S1A). This structure is consistent with the morphology of the viral replication organelle (RO) and would be expected to concentrate full-length genomes and viral dsRNA, which we confirmed by immunofluorescence (IF) assays for dsRNA and smRNA-FISH for the 5′-CDS (Figure S1B). These data demonstrate that the 5′-CDS, 3′-CDS, and sfRNA probes detect the specific region of interest in each viral RNA.

We next examined viral RNA production during the early phase of WNV infection to determine if the probe sets detect single viral RNAs. At 4 h p.i., we typically observed 10–50 WNV RNAs in the cytosol (Figure 1B). Most of these viral RNA puncta stained with all three probe sets, indicating that they are full-length genomes (Figure 1B). We observed an increase in viral RNAs at both 8 and 12 h p.i. (Figure 1B). At these times p.i., most viral RNA puncta were full-length genomes, as they stained for all three probe sets. However, a fraction of viral RNA puncta lacked 5′-CDS staining but stained for both 3′-CDS and sfRNA, indicative of long 3′-degradation intermediates (Figure 1B).

Importantly, we observed viral RNA puncta that stained for sfRNA but neither the 5′-CDS nor the 3′-CDS8 and 12 h p.i., indicating that independent sfRNA fragments are generated early during infection (Figure 1B). By 18 h p.i., we detected between 500 and 1,000 sfRNA puncta per cell during WNV or ZIKV infection (Figures S2A–S2D). These data demonstrate that sfRNAs are produced during the RNA replication phase of infection (4–18 h p.i.) and diffusely localize throughout the cytosol.

sfRNAs accumulate in cytoplasmic structures over the course of infection

By 12 h p.i., a small number of WNV-infected cells contained larger cytoplasmic structures that stained for multiple sfRNAs based on their larger size and increased intensity relative to single sfRNA molecules (Figure 1C). At both 18 and 24 h p.i., these larger structures containing sfRNAs became more predominant, and most sfRNAs localized in these larger structures (Figure 1C). We refer to these larger cytoplasmic sfRNA structures as sfRNA granules in the following sections.

The majority of A549 cells infected with ZIKV, WNV, or DENV-2 contained sfRNA granules by 24 h p.i. (Figures 2A and 2B). Notably, the sfRNA granules did not stain for the 5′-CDS, indicating that they lack full-length genomes (Figure 2A). Co-staining ZIKV-infected cells with all three probe sets revealed that some sfRNA granules stained for the 3′-CDS but none stained for the 5′-CDS (Figures 2C, inset A, and S3B). All three probe sets equally stained the RO (Figure 2C, inset B), indicating that the lack of 5′-CDS staining is specific to sfRNA granules. Based on the intensity of individual smRNA-FISH puncta vs. the intensity of sfRNA granules for each probe set, we estimate that each sfRNA granule contained an average of 50–75 copies of sfRNAs, 5–15 copies of longer 3′ end degradation intermediates, and less than 10 RNAs containing the 5′-CDS (Figure 2D). These data indicate that sfRNA granules primarily contain sfRNAs and longer 3′ end fragments of flavivirus genomes but lack full-length genomes or 5′ end decay products. We typically observed between 20 and 40 sfRNA granules per cell (Figure 2E). We estimate that between 1,000 and 2,000 sfRNA molecules accumulate in sfRNA granules per cell by 24 h p.i. (Figure S3C), which accounts for 60%–90% of total sfRNA molecules in the cell (Figure 2F).

Figure 2. sfRNAs re-localize to cytoplasmic granules distinct from P-bodies.

(A) Representative image of infected cells co-stained for 5′-CDS and sfRNAs 24 h post-infection (p.i.) with WNV and 36 h p.i. with ZIKV. White arrows correspond to cells containing sfRNA granules, while yellow arrows correspond to cells lacking these granules.Scale bar, 20 μm.

(B) Percentage of infected cells with sfRNA granules 24 h p.i. n = 3 biological replicates.

(C) Left: co-staining smFISH for ZIKV 5′-CDS, 3′-CDS, and sfRNAs 24 h p.i. of A549 cells with ZIKV (MOI = 10). Right: plotted profiles of lines in the insets. Scale bar, 10 μm.

(D) Estimated RNA molecules of each stain per sfRNA granule 24 h p.i. Error bars represent mean with SD. n = 5 cells.

(E) Number of sfRNA granules per cell 24 h p.i. Error bars represent mean with SD. n = 10 cells.

(F) Percentage of total sfRNA (excluding the replication organelle) localized in granules at 24 h p.i. Error bars represent mean with SD. n = 5 cells.

(G) IF for DCP1b and smFISH for WNV 5′-CDS and sfRNAs 24 h p.i. of A549 cells with WNV (MOI = 10). Right of the inset, plotted profile shows raw intensity of indicated channels. Scale bar, 20 μm.

(H) Super-resolution microscopy of P-bodies (DCP1b) docked to WNV sfRNA granules. Nuclei (blue) are stained with DAPI. Scale bar, 5 μm. Inset scale bar, 1 μm.

sfRNA granules are distinct from P-bodies

We hypothesized that sfRNA granules are P-bodies based on the rationale that sfRNAs associate with mRNA decay machinery that enrich in P-bodies.^{16–18,44–46} To test this, we performed IF assays for DCP1b, a P-body marker,^47,48 and co-stained for sfRNAs. Several observations indicate that sfRNA granules are not P-bodies. First, WNV-infected cells contained 2-fold more sfRNA granules (median: 30) than P-bodies (median:14) (Figures 2G, S4A, and S4B). Second, only a small percentage (<20%) of sfRNA granules co-localized with P-bodies (Figures 2G and S4C). Third, only 30% of P-bodies stained for sfRNAs (Figures 2G and S4C). Similar results were obtained in U-2 OS cells that stably express RFP-Dcp1a (Figure S4D), as well as during ZIKV infection (Figure S4E). Lastly, while we observed instances in which P-bodies and sfRNA granules co-localized (Figure 2G), super-resolution confocal microscopy revealed that the sfRNA granules were often docked with P-bodies as opposed to being localized within P-bodies, with single sfRNA granules sometimes docking with multiple P-bodies (Figure 2H). These data indicate that sfRNAs concentrate in cytoplasmic structures that are distinct from, yet interact with, P-bodies.

sfRNAs localize to RLBs

We next considered the possibility that sfRNA granules are either SGs or RLBs. This is based on the rationale that both SGs and RLBs are ribonucleoprotein complexes that can assemble in the cytoplasm in response to viral infection and can interact with P-bodies.^36,37 Additionally, flaviviruses are known to activate both PKR and RNase L,^49,50 which trigger the assembly of SGs or RLBs, respectively.³⁴

SGs and RLBs concentrate similar RNA-binding proteins.³⁴ However, their assembly is mutually exclusive. This is because the activation of RNase L, which is required for RLB assembly, inhibits SG assembly due to the degradation of SG-associated mRNAs.³⁴ For example, lipofection of poly(I:C) (an immunogenic viral dsRNA mimic) results in G3BP1 localization to RLBs in wild-type (WT) A549 cells (Figure 3A). However, in RNase L-knockout (KO) cells, G3BP1 localizes to SGs assembled in response to poly(I:C) (Figure 3A). Notably, SGs in RNase L-KO (RL-KO) cells are more irregular in their morphology and larger (5 μm diameter) than RLBs in WT cells, which are smaller (1 μm diameter) and more spherical (Figure 3A). Additional key differences between SGs and RLBs are that SGs require PKR and G3BP1 for their assembly, whereas RLBs do not, and that RLBs do not enrich some SG-associated RBPs, such as TIA-1.³⁶

Figure 3. sfRNA localize to RNase L-induced bodies.

(A) IF for G3BP1 in WT and RNase L-KO (RL-KO) cells treated with poly(I:C) demonstrating phenotypic differences between RLBs and SGs. Scale bar, 20 μm.

(B) IF for G3BP1 and smFISH for ZIKV sfRNAs in A549 cells 24 h p.i. (MOI = 10) with ZIKV. Scale bar, 20 μm.

(C) Similar to (B) but staining IF for PABPC1 or UBAP2L or RNA FISH for poly(A)+ RNA. Panels displaying individual stains are shown in Figures S5A–S5C. Scale bar, 20 μm.

(D) IF for PABPC1 and smFISH for ZIKV sfRNAs in parental (WT) A549 and RNase L, PKR, and/or G3BP1/2-KO A549 cells 24 h p.i. with ZIKV (MOI = 10). Scale bar, 10 μm.

(E and F) PABPC1 IF and smFISH for ZIKV 5′ and sfRNAs in WT (E) and RL-KO (F) A549 cells. The line graph on to the right of the images displays the intensity of PABPC1, sfRNA, and 5′-CDS staining from the line trace is displayed in the merged image. Scale bar, 10 μm.

(G) smFISH staining of ZIKV sfRNA in WT vs. RL-KO cells. Yellow arrows indicate cells with sfRNA puncta, and white arrows indicate cells with no sfRNA puncta. Scale bar, 20 μm.

(H) IF for PABPC1 and smFISH for ZIKV sfRNAs in PKR/RL-KO A549 cells rescued with either parental RL or a catalytically inactive version (R667A) 24 h p.i. with ZIKV (MOI = 10). Nuclei (blue) are stained with DAPI. Scale bar, 10 μm.

To determine if sfRNAs concentrate in either SGs or RLBs, we performed IF assays for G3BP1 (RLB/SG marker) 24 h p.i. with ZIKV. Importantly, we observed that nearly all ZIKV sfRNA granules co-stained for G3BP1 (Figure 3B). Moreover, we observed that additional RLB/SG markers—PABPC1, UBAP2L, poly(A)+ RNA—co-localize with sfRNA granules (Figures 3C and S5A–S5C). PABPC1 and G3BP1 also co-localized with sfRNA granules during DENV-2 and WNV infection (Figures S5D and S5E). These data indicate that sfRNA granules are either RLBs or SGs.

Several observations indicate that the sfRNA granules are RLBs as opposed to SGs. First, the small and spherical morphology of sfRNA-G3BP1 granules in ZIKV-infected cells more closely resembled RLBs formed in WT cells as opposed to SGs assembled in RL-KO cells in response to poly(I:C) (Figures 3A and 3B). Second, TIA-1 did not strongly enrich in sfRNA granules (Figure S5F). Third, sfRNA/PABPC1 granules formed in PKR-KO and G3BP1/2-KO cell lines infected with ZIKV (Figure 3D), whereas SGs that assemble in RL-KO cells in response to ZIKV infection were completely abolished upon KO of PKR or G3BP1/2 (Figure S5G). Fourth, in A549 cells infected with ZIKV, sfRNA staining intensity in the PABPC1 puncta was 10-fold higher than the signal in the viral RO (Figure 3E). In contrast, sfRNA enrichment in SGs in RL-KO cells was much lower in comparison to the RO (Figure 3F). Similar results were obtained for DENV-2 and WNV (Figures S6A and S6B).

A key observation supporting that sfRNA granules are RLBs is that they are dependent on RNase L for their formation, as they do not form in RL-KO A549 cells infected with ZIKV, WNV, or DENV-2 (Figures 3G, S6A, and S6B). For example, when ~50% of ZIKV-infected WT cells contained sfRNA granules, no ZIKV-infected RL-KO cells contained sfRNA granule structures (Figure 3G). Instead, sfRNA probes strongly stained the viral RO or diffusely stained the cytoplasm (Figure 3G). Importantly, the formation of sfRNA-PABPC1 granules in RNase L-null A549 cells infected with ZIKV was rescued upon stable expression of RNase L but not RNase L-R667A (catalytically inactive mutant) (Figure 3H). These data indicate that the formation of sfRNA-PABPC1 granules in mammalian cells requires expression of catalytic competent RNase L, which is the key factor required for RLB assembly.³⁴

Consistent with RNase L being required for sfRNA granule assembly, we did not observe sfRNA granules in WNV-infected c6/36 mosquito cells (Figure S7A), which do not encode for RNase L and thus would not be expected to assemble RLBs. However, we did observe ZIKV sfRNAs localize to RLBs in primary human pulmonary artery endothelial cells (1° hPAECs) (Figure S7B), which have a robust RNase L response,⁵¹ demonstrating that sfRNAs localize to RLBs in non-cancer human cells.

Combined, these data argue that sfRNAs encoded by DENV-2, ZIKV, and WNV localize to RLBs during viral infection in mammalian cells.

sfRNAs re-localize to RLBs upon activation of RNase L-mediated RNA decay

We next sought to understand how sfRNAs localize to RLBs. RNase L is a latent endoribonuclease that is activated upon detection of viral dsRNA by oligoadenylate synthetase (OAS) proteins.⁵² Following activation, RNase L rapidly degrades cellular RNA.^34,53 As cellular mRNAs degrade, RLBs assemble.³⁴ The observation that not all A549 cells display sfRNA granules led to the hypothesis that these cells had not activated RNase L and thus did not contain RLBs, whereas the cells with sfRNA granules activated RNase L and thus contained RLBs. To test this, we stained WT and RL-KO A549 cells infected with ZIKV for human GAPDH mRNA, which is rapidly degraded upon the activation of RNase L.³⁴

Consistent with the hypothesis that sfRNAs concentrate at RLBs that form upon the activation of RNase L-mediated RNA decay, we only observed sfRNA granules (RLBs) in ZIKV-infected A549 cells that had degraded GAPDH mRNA (Figure 4A, white arrows). In contrast, sfRNAs in ZIKV-infected A549 WT cells containing abundant GAPDH mRNA were diffusely distributed in the cytosol and did not contain sfRNA granules (RLBs) (Figures 4A and 4B). The degradation of GAPDH mRNA in response to ZIKV infection was dependent on RNase L because GAPDH mRNA levels were not reduced in ZIKV-infected RL-KO A549 cells (Figures 4A and 4B). We obtained similar results during DENV-2 and WNV infection (Figures S8A and S8B). These data indicate that sfRNAs concentrate at RLBs following the activation of RNase L-mediated RNA decay and subsequent assembly of RLBs.

Figure 4. sfRNAs localize to RLBs following the initiation of RNase L-mediated mRNA decay.

(A) smFISH for human GAPDH mRNA and ZIKV sfRNAs in mock or 24 h p.i. with ZIKV (MOI = 10) in parental (WT) and RL-KO A549 cells. White arrows indicate cells with active RNase L, while yellow arrows point to cells without RNase L activated. Scale bar, 20 μm.

(B) Quantification of GAPDH mRNA smFISH in mock-infected WT cells, WT cells infected with ZIKV that did or did not display sfRNA granules, and RL-KO cells, as represented in (A). n = 24 cells. ***p < 0.001. Box plots represent median, inner quartile, and range.

(C) smFISH for WNV 5′-CDS (yellow) and sfRNA (magenta) 12 h p.i. (MOI = 10). Individual magenta foci represent sfRNAs localized to the cytoplasm, whereas white dots (co-localized magenta and yellow foci) represent full-length genomes containing the 5′-CDS and sfRNA. The viral replication organelle (RO) is indicated, which concentrates full-length genomes. Scale bar, 20 μm.

(D) Quantitation of sfRNA count by smFISH in WT vs. RL-KO A549 cells at 12 h p.i. n = 19 cells.

(E) Northern blot analysis for WNV sfRNAs 24 h p.i. with WNV (MOI = 10).

(F) Schematic for generating sfRNA in vitro and transfecting into cells.

(G) IF for G3BP1 and smFISH for ZIKV sfRNAs co-transfected with or without in-vitro-transcribed (IVT)-sfRNA (ZIKV) with poly(I:C) (RNase L activator) in WT and RL-KO cells with and without SGs. Gray arrows represent the line used for the line trace (below). Nuclei (blue) are stained with DAPI. Scale bar, 10 μm.

We considered two non-mutually exclusive possibilities by which sfRNAs localize to RLBs, whereby sfRNAs are generated either by RNase L-mediated cleavage of the viral RNA shuttle to RLBs or independently of RNase L but localize to RLBs upon their assembly. To test these possibilities, we first examined if RNase L promotes sfRNA generation. Because RNase L has been proposed to specifically cleave viral RNA during the early phase of infection,⁵⁴ we first asked if RNase L could specifically cleave viral RNAs to generate sfRNAs, which then localize to RLBs upon their assembly. One prediction of this would be that sfRNA levels would be lower in RL-KO cells. However, smRNA-FISH for WNV sfRNAs revealed that comparable numbers of sfRNA molecules are generated in WT and RL-KO A549 cells 12 h p.i. (Figures 4C and 4D). Thus, RNase L does not promote sfRNA generation by specifically cleaving viral genomes early during infection prior to widespread RNase L activation.

We next examined if widespread activation of RNase L, which occurs by 24 h p.i (Figure 4A), increases WNV sfRNA levels. Because viral RNAs are too abundant to reliably differentiate independent sfRNAs from full-length genomes at this time via smRNA-FISH (Figure S2A), we performed northern blot analyses for WNV sfRNA. WNV sfRNA-1 (the largest sfRNA species) was comparable between WT and RL-KO cells (Figure 4E), indicating that sfRNA-1 can be generated independently of RNase L. However, RNase L increased production of the smaller WNV-encoded sfRNA species (sfRNA-2 and sfRNA-3), presumably by cleaving sfRNA-1(Figure 4E). This effect was rescued by stable expression of RNase L in RL-KO cells but not by the R667A catalytic mutant, indicating these lower bands depend on RNase L-mediated RNA cleavage. Quantification of total sfRNA species between WT and RL-KO cells showed that RNase L increases sfRNA levels in cells by ~2-fold (Figure S8C). Similar results were observed in ZIKV infection (Figure S8D).

Because sfRNAs are able to be generated independently of RNase L prior to RLB assembly, this suggests that pre-formed sfRNAs could re-localize to RLBs. To test this, we generated ZIKV sfRNA-1 using T7 polymerase (T7-sfRNAs). Following purification, we co-transfected ZIKV T7-sfRNA-1 into WT or RL-KO A549 cells with poly(I:C) to activate RNase L (Figure 4F). In WT cells that activated RNase L and assembled RLBs (G3BP1 marker), sfRNAs were mostly localized to RLBs, whereas in RL-KO cells, which do not form RLBs, the T7-sfRNAs were widely distributed in the cytosol (Figure 4G). We note that T7-sfRNAs also localized to the SGs in RL-KO cells that generated SGs, though the enrichment of T7-sfRNA in SGs was ~10-fold less than that observed in RLBs (Figure 4G).

Taken together, these data indicate that both pre-formed and RNase L-generated sfRNAs re-localize to RLBs upon RLB assembly following RNase L activation.

RLBs sequester sfRNAs away from mRNA decay machinery and P-bodies to promote viral RNA decay

We next wanted to understand how RLB-mediated sequestration of sfRNAs alters the cellular response to flavivirus infection. A primary function of sfRNAs is the inhibition of cellular mRNA decay. This function is supported by studies showing that sfRNAs interact with mRNA decay machinery^18,55 and can localize to P-bodies, which enrich for mRNA decay machinery.¹⁶ Moreover, sfRNAs can limit the XRN1-mediated decay of cellular and viral RNAs. Thus, we hypothesized that RLBs could sequester sfRNAs away from cellular RNA decay machinery in the cytoplasm and/or P-bodies, and this in turn would promote the decay of cellular and viral RNA. The following observations support this hypothesis.

RLB assembly reduces sfRNA association with P-bodies

We considered the possibility that sfRNA sequestration by RLBs could reduce sfRNA association with mRNA decay machinery. One prediction of this hypothesis is that in the absence of RLBs, sfRNAs would enrich in P-bodies. However, upon RNase L activation and RLB assembly, sfRNA association with P-bodies would decrease. To test if RLBs could reduce sfRNA interactions with mRNA decay machinery in P-bodies, we stained WNV-infected WT and RL-KO cells for G3BP1 (RLB marker), DCP1b (P-body marker), and sfRNAs.

In RL-KO cells, which do not assemble RLBs based on the lack of G3BP1 puncta, we observed that WNV sfRNAs strongly enriched in all P-bodies at both 24 and 48 h p.i. (Figures 5A, 5B, and S9A). Importantly, in WT A549 cells that assembled RLBs, sfRNA intensity in P-bodies was ~3.5-fold lower than that observed in P-bodies in RL-KO cells (Figure 5A). Moreover, sfRNA intensity was 3.5-fold higher in RLBs than P-bodies in these WT cells (Figure 5A). We obtained similar results by performing IF for XRN1, whereby sfRNAs co-localized with XRN1 in P-bodies in RL-KO cells (Figure 5B). However, in WT cells containing RLBs, sfRNAs did not co-localize with XRN1 and instead localized to RLBs, which did not contain XRN1 (Figure 5B). We obtained similar results during ZIKV infection, though sfRNA abundance in P-bodies was notably lower in comparison to WNV (Figure S9B).

Figure 5. sfRNAs re-localize from P-bodies to RLBs upon RNase L activation.

(A) IF of G3BP1, DCP1B, and smFISH of sfRNA in RL-KO and WT A549 cells. Gray lines indicate the plot profile (bottom). Scale bar, 20 μm.

(B) IF of G3BP1, XRN1, and smFISH of sfRNA in RL-KO and WT A549 cells. Gray lines indicate the plot profile (bottom). Scale bar, 10 μm.

(C) Mean intensity of sfRNAs in P-bodies in binned cell types. n = 23 granules. ***p < 0.001. Box plots represent median, inner quartile, and range.

(D) Super-resolution of IF for G3BP1 and Dcp1b co-stained with sfRNA. Inset and line trace show an example of a docked P-body. Scale bar, 5 μm.

(E) Plotted profiles of gray line indicated in (D).

Quantification of sfRNA enrichment in P-bodies in RL-KO cells or in WT cells that either contained (RLB+) or did not contain (RLB−) RLBs in several cells confirmed these observations. In comparison to WT cells that lacked RLBs or RL-KO cells, sfRNA enrichment in P-bodies was significantly reduced in WT cells that contained RLBs (Figure 5C). Importantly, we observed a corresponding increase in WNV sfRNAs enriched in RLBs (Figure 5C), indicating that sfRNAs re-localize from P-bodies to RLBs following RLB assembly.

P-bodies and RLBs are known to stably interact,³⁶ suggesting that sfRNAs could be directly transferred between P-bodies and RLBs. Indeed, we observed docking between sfRNA-positive P-bodies and RLBs (Figure 5D). This suggests that sfRNAs could be directly transferred from P-bodies to RLBs. However, we did not further examine if docking of P-bodies to RLBs is required for sfRNA accumulation in RLBs or whether sfRNAs can accumulate in RLBs independently of P-bodies.

XRN1-mediated decay of cellular mRNAs is robust in flavivirus-infected cells containing RLBs

sfRNAs are known to inhibit the XRN1-mediated decay of cellular mRNAs.^18,27 However, two observations, in addition to the reduced co-localization with XRN1 in P-bodies shown above, suggest that flavivirus-infected cells that contain RLBs have robust cellular mRNA decay potential. First, smRNA-FISH analyses showed robust decay of GAPDH mRNA upon the activation of RNase L during DENV-2, ZIKV, or WNV infection (Figures 4A and S7). Consistent with these data, total poly(A)+ RNA was significantly reduced in an RNase L-dependent manner in ZIKV- and WNV-infected cells containing RLBs (Figures 6A, 6B, and S10A–S10C). These data demonstrate that upon RNase L activation, flavivirus-infected cells have robust cellular mRNA decay potential.

Figure 6. RLBs sequester sfRNAs to re-establish XRN1-mediated decay of viral RNA.

(A) FISH for poly(A) RNA and smFISH for ZIKV sfRNAs at 24 h p.i. Scale bar, 10 μm.

(B) Quantification of the cytosolic levels of poly(A) RNA in listed cell types during ZIKV infection. n = 20 cells. ***p < 0.001. Box plots represent median, inner quartile, and range.

(C) IF of PABPC1 and smFISH of sfRNA in WT and RL-KO A549 cells. White arrows point to cells with PABC1 translocation into the nucleus, and yellow arrows point to cells lacking nuclear PABC1. Scale bar, 25 μm.

(D) Quantification of images from (C). n = 3 fields of view.

(E) Live-cell imaging of A549 cell expressing mRuby2-PABPC1 infected with WNV virus. Scale bar, 10 μm.

(F) Single-cell quantification of PABP translocation rate during poly(I:C), ZIKV, and WNV. Zero time point set to first RLB formation in each cell. n = 9 cells. Error bars represent mean with SD.

(G) Co-staining of smFISH of sfRNA, 5′-CDS, and IF for PABPC1 24 h p.i. Pink outlines indicate RNase L activation, while yellow outlines indicate inactive RNase L. Scale bar, 10 μm.

(H) Quantification of intensity and size of RO in cells with or without RNase L active from (G). n = 25 cells.

(I) Proposed model of sfRNA function. Nuclei (blue) are stained with DAPI.

Second, when RNase L cleaves cellular mRNAs and XRN1 degrades the 3′ fragment, PABPC1 is released from poly(A) tails and translocates to the nucleus.³⁴ Importantly, we observed RNase L-dependent PABPC1 translocation to the nucleus in WNV- and ZIKV-infected cells that contained RLBs (Figures 6C and S11A) and that the initiation of PABPC1 translocation to the nucleus over the course of infection closely coincided with sfRNA-RLB formation (Figures 6D, S11B, and S11C).

We then performed live-cell imaging of PABPC1 translocation during WNV or ZIKV infection or following poly(I:C) lipofection to determine the rate of PABPC1 translocation (Figures 6E, S12A, and S12B). These data demonstrated that the rate of PABPC1 translocation following RLB formation between ZIKV-infected cells and poly(I:C)-lipofected cells was comparable (Figures 6F and S12A–S12C). WNV-infected cells displayed a slightly slower rate of PABPC1 translocation compared to ZIKV and poly(I:C) (Figures 6E and S12C), though it did fully translocate within 4 h following RLB assembly. Notably, WNV sfRNAs more highly enrich in P-bodies prior to RNase L activation in comparison to ZIKV sfRNAs (Figures S12D and S12E). We interpret these data to indicate that WNV sfRNAs more strongly antagonize mRNA decay machinery, thus resulting in a reduction in mRNA decay kinetics until RLBs can sequester WNV sfRNAs from P-bodies.

We confirmed that PABPC1 translocation to the nucleus during viral infection is XRN1 dependent by showing that knockdown of XRN1 reduced the rate of PABPC1 translocation during ZIKV infection (Figures S13A–S13C). These data demonstrate that XRN1 is active in cells that have activated RNase L-mediated assembly of RLBs.

RLB assembly correlates with reduced viral genomic RNA

The observation that XRN1 is active in cells that have activated RNase L and formed RLBs led us to ask whether RLBs could promote viral RNA decay. To test this, we measured 5′-CDS staining (full-length genome) in A549 cells infected with WNV 24 h p.i. We stained for PABPC1 and sfRNAs to identify cells that activated RNase L and thus contained sfRNA+ RLBs and nuclear PABPC1. We then quantified 5′-CDS in cells that contained RLBs (RLB+) or did not contain RLBs (RLB−).

Importantly, we observed a significant reduction (~5-fold) in the full-length genome (5′-CDS staining) in WNV-infected WT A549 cells that contained RLBs in comparison to cells that lacked RLBs (Figures 6G and 6H). Moreover, we observed a significant reduction (3-fold) in the area of the RO (Figures 6G and 6H). Similar results were obtained with ZIKV (Figures S13D and S13E). These data support that RLB-mediated sequestration of sfRNA can promote the decay of viral mRNAs/genomes by cellular RNA decay machinery (Figure 6I).

DISCUSSION

In this article, we analyzed the production and localization of sfRNAs by smRNA-FISH during DENV-2, WNV, or ZIKV infection (Figures 1 and S1–S3). We observed that the vast majority (~90%) of sfRNAs generated by DENV-2, WNV, and ZIKV re-localize from the cytosol to cytoplasmic ribonucleoprotein complexes termed RLBs (Figures 1 and 2), which form upon the activation of the RNase L-mediated RNA decay pathway (Figures 3 and 4). In cells that assemble RLBs, we observed reduced localization of sfRNAs in P-bodies containing XRN1 (Figure 5), and this coincided with an increase in the decay of viral RNA (Figure 6). Based on these observations and previous literature supporting that sfRNAs inhibit cellular RNA decay,^18,27,55 we propose that the RLBs sequester sfRNAs away from RNA decay machinery localized in the cytosol and P-bodies, thus leading to an increase in the capacity for cellular RNA decay pathways to degrade viral RNA (Figure 6I).

The identification of sfRNAs localizing to RLBs is important because specific cellular RNAs had yet to be detected in RLBs, and viral RNAs were not known to localize to RLBs. While previous studies failed to identify full-length cellular mRNAs in RLBs,^34,36,37 RLBs strongly enrich for cellular poly(A)+ RNA.^34,36 This has led to the hypothesis that RLBs concentrate the degradation fragments of mRNAs containing poly(A)+ tails following the activation of RNase L-mediated mRNA decay.³⁷ Thus, RLBs concentrating subgenomic viral RNAs (sfRNAs and longer 3′ end fragments), but not full-length viral genomes, supports the notion that RLBs are composed of 3′ end fragments of RNAs generated via cellular RNA degradation pathways following RNase L activation. Based on these findings, RLBs appear to be involved in the turnover of host/viral mRNAs and/or are sites to compartmentalize host/viral mRNAs that cannot be fully degraded by typical RNA decay pathways.

Our data indicate that two non-mutually exclusive mechanisms potentially contribute to sfRNA localization to RLBs. The first potential mechanism is that RNA fragments generated by RNase L-mediated cleavage specifically shuttle to RLBs. While RNase L is not required for sfRNA-1 generation based on our northern blot analyses 24 h p.i. with either WNV or ZIKV (Figures 4E, S8C, and S8D), RNase L was required for the biogenesis of downstream sfRNA fragments, sfRNA-2 and sfRNA-3 (Figures 4E, S8C, and S8D). Thus, it is possible that the cleavage of sfRNA-1 into sfRNA-2 and sfRNA-3 by RNase L results in their incorporation into RLBs.

A second potential mechanism by which sfRNA localize to RLBs is through sfRNA interactions with RNA-binding proteins that concentrate in RLBs, which would result in sfRNA re-localization from the cytosol to RLBs upon RLB assembly. Consistent with this, we observed sfRNAs widely distributed in the cytosol early during infection prior to RNase L activation (Figure 1C) in cells with inactive RNase L (Figure 4A). However, in cells with RLBs, nearly all sfRNAs localize to RLBs, suggesting that sfRNAs re-localize to RLBs after RLB assembly. Consistent with this, pre-formed T7-generated sfRNAs localized to RLBs (Figure 4G). This suggests that sfRNAs interact with an RNA-binding protein that concentrates in RLBs. Notably, we observed a low level of sfRNAs enrich in SGs that formed in RL-KO cells in response to viral infection or T7-sfRNA lipofection (Figures 3F and 4G). While sfRNAs did not greatly enrich in SGs, which is consistent with the fact that short RNAs do not typically concentrate in SGs,^56–58 this suggests that an RNA-binding protein common to RLBs and SGs may be responsible for sfRNA localization to RLBs. Although sfRNAs have been shown to bind G3BP1,⁵⁹ G3BP1 is not the RNA-binding protein responsible for this based on the observation that sfRNAs localize to RLBs in G3BP1/2-KO cells (Figure 3D). Studies are underway to further define the molecular basis responsible for sfRNA sequestration in RLBs.

Previous studies have shown that sfRNAs interact with cellular RNA decay factors (XRN1, EDC3, DDX6) that enrich in P-bodies.^16,27,44,60 Moreover, FISH analyses of sfRNAs showed that sfRNAs primarily co-localize with XRN1 in P-bodies in A549 cells.¹⁶ Thus, a consensus model for the function of sfRNA is that they antagonize cellular RNA decay pathways to increase the stability of viral RNA by inhibiting mRNA decay machinery that enrich in P-bodies. Our data are consistent with these previous findings, as we observed that sfRNAs localize to P-bodies in cells with inactive RNase L (Figure 5A). However, our data add to these studies by showing that most sfRNAs re-localize to RLBs in cells that activated RNase L and assembled RLBs, with RLBs sequestering most of the sfRNAs that would otherwise associate with mRNA decay machinery in P-bodies (Figure 5). The interactions between RLBs and P-bodies suggest that sfRNAs might be directly transferred from P-bodies to RLBs (Figure 5C), although re-partitioning of sfRNAs from P-bodies to the cytosol and then to RLBs could also lead to a reduction of sfRNAs in P-bodies over time (Figures 5A and 5B). Regardless of the mechanism, the reduction of sfRNA localization to P-bodies in cells with RLBs suggests that sfRNAs are no longer associating with RNA decay machinery, which would be expected to promote RNA decay machinery. It should be noted that it is unsettled whether P-bodies are sites of RNA decay and may be context dependent.^28,29,61 Nevertheless, co-localization of sfRNA and P-bodies is likely reflective of sfRNA interactions with cellular RNA decay machinery in the cytosol.

Three observations indicate that RNA decay machinery is active in mammalian cells that contain RLBs despite the presence of sfRNAs. First, GAPDH mRNAs and total poly(A)+ RNA decreased in cells that have activated RNase L and assembled RLBs (Figures 4A and 6A). Second, PABPC1 rapidly translocates from the cytosol to the nucleus in an XRN1-dependent manner following the activation of RNase L in cells infected with DENV-2, ZIKV, or WNV (Figures 6C, S10, and S11).⁶² Because PABPC1 translocation from the nucleus to the cytoplasm is determined by cytosolic RNA levels,^63,64 this observation suggests that cytosolic RNAs are undergoing rapid decay. Lastly, we observed an enhanced decay of viral genomes/mRNAs in ZIKV- and WNV-infected cells that contained RLBs (Figures 6G, 6H, and S13), indicating that cellular RNA decay machinery is functional in cells in which sfRNAs have been sequestered. A limitation of these data is that RNase L can cleave viral RNAs and thus could account for this reduction in overall viral RNA levels. However, we would expect that RNase L-mediated cleavage would not result in a decrease in total fluorescence of viral RNAs without further decay by XRN1. Thus, we interpret our data to suggest that XRN1 is functional in degrading viral genomes following RLB assembly, and that both RNase L-mediated cleavage of viral genomes and RLB-mediated sequestration of sfRNA promote XRN1-mediated decay of viral genomes. Current studies are underway to identify factors other than RNase L that are required for RLB assembly, which will allow for further dissection of the function of RLBs.

In conclusion, our data strongly argue that the OAS/RNase L antiviral pathway regulates sfRNA biology in mammalian cells, both by altering sfRNA biogenesis and by reducing the interactions of sfRNAs with RNA decay machinery in P-bodies via sequestration to RLBs. While our data are consistent with RLB-mediated sequestration of sfRNAs promoting the decay of viral RNA, we do not rule out the possibility that sfRNAs could inhibit a yet-to-be identified function of RLBs or that RLBs could inhibit additional functions of sfRNA. Ongoing studies aim to further understand the molecular basis and functional consequences of sfRNA and RLB interactions. Lastly, an investigation of RLB interactions with diverse viruses will be important to further understand how RLBs impact the innate antiviral response.

Limitations of the study

1. We cannot separate the formation of RLBs from the catalytic activity of RNase L. This limits our ability to directly understand the relative impact of RLBs and RNase L on these processes.

2. While we can observe that sfRNAs initially localize to P-bodies and then later re-localize to RLBs, we did not test whether sfRNAs directly transfer between the P-bodies and RLBs. We were unable to attain live-cell smRNA imaging that would allow us to directly observe the transfer. The mechanism of this transfer could be mediated by direct interactions of P-bodies and RLBs or occur through re-partitioning indirectly through exchange with the cytosol.

3. Our single-cell assay for the detection of XRN1 activity relies on an indirect measurement of decay of poly(A) tails. While PABPC1 translocation to the nucleus is largely dependent on XRN1 based on small interfering RNA (siRNA)-mediated knockdown of XRN1, we cannot rule out that other nucleases contribute to poly(A)+ RNA degradation.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, James M. Burke (james.burke@ufl.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• Original western blot, northern blot, and microscopy images have been deposited at Mendeley and are publicly available as of the date of publication. DOIs are listed in the key resources table.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
DCP1B (D2P9W) Rabbit mAb	Cell Signaling Technology Inc	13233S
XRN1 Polyclonal Antibody	Life Technologies Corporation	A300443A
Anti-PABP antibody	Abcam	ab21060; RRID:AB_777008
Anti-UBAP2L antibody	Abcam	ab70319; RRID:AB_1271381
G3BP1 Polyclonal Antibody	Life Technologies Corporation	PA529455; RRID:AB_2546931
dsRNA (K1) Mouse mAb	Cell Signaling Technology Inc	28764
Goat pAb to Rb IgG Alexa Fluor 647	Abcam	Ab150079;RRID:AB_2722623
Goat Anti-Rabbit IgG H&L (FITC)	Abcam	AB6717; RRID:AB_955238
Goat pAb to Ms IgG (Alexa Fluor 555)	Abcam	AB150114; RRID:AB_2687594
Goat Anti-Mouse IgG H&L (FITC)	Abcam	AB6785; RRID:AB_955241
Bacterial and virus strains
Dengue Virus serotype 2, NGC	ATCC	VR-1584
West Nile Virus lineage 1 NY99	Laboratory of Hyerun Choe	N/A
ZIKV strain PB 81	Laboratory of Hyerun Choe	N/A
Chemicals, peptides, and recombinant proteins
Poly(I:C) HMW	InvivoGen	tlrl-pic
Lipofectamine 2000	Thermo Fisher Scientific	11668027
Trizol LS reagent	Thermo Fisher Scientific	10296–028
5-Propargylamino-ddUTP-ATTO-488	Axxora	JBS-NU-1619-488
5-Propargylamino-ddUTP - ATTO-550	Axxora	JBS-NU-1619-633
5-Propargylamino-ddUTP-Atto633	Axxora	JBS-NU-1619-633
Deposited data
Unprocessed image files, Mendeley Dataset DOI	This study	Mendeley Data: http://www.doi.org/10.17632/9kwwrcbhbr.1
Experimental models: Cell lines
A549 cells	Burke et al., 2016	N/A
U-2 OS cells	Kedersha et al., 2016	N/A
A549-PKR-KO cells	Burke et al., 2020	N/A
A549-RL-KO cells	Burke et al., 2019	N/A
A549-G3BP1/2-KO cells	Burke et al., 2024	N/A
A549-G3BP1/2-KO-RL-KO cells	Burke et al., 2024	N/A
A549-PKR-KO-RL-KO cells	Burke et al., 2019	N/A
A549-RL-KO-RL cells	Burke et al., 2019	N/A
A549-RL-KO-CM cells	Burke et al., 2019	N/A
A549-PKR-KO-RL-KO-RL-rec cells	Burke et al., 2020	N/A
A549-PKR-KO-RL-KO-CM-rec cells	Burke et al., 2020	N/A
A549-mRUBY2-PABP	Burke et al., 2020	N/A
U2OS GFP-G3BP1 RFP-DCP1B	Kedersha et al., 2005	N/A
Oligonucleotides
smFISH pobes	This study, Table S1	N/A
PCR primers	This study, Table S1	N/A
T7 Oligonucleotides	This study, Table S1	N/A
Software and algorithms
Stellaris probe Designer	Stellaris	https://www.biosearchtech.com/support/tools/design-software/stellaris-probe-designer
FIJI	ImageJ	https://imagej.net/software/fiji/downloads
Biorender	Biorender	Biorender.com
Excel	Microsoft	N/A

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cell culture

A549 and U2OS cells were cultured in DMEM (Corning, cat#10-013-CV) +10% HI FBS(Sigma Aldrich, cat#F4135-500ML) + penicillin/streptomycin (Sigma, cat #P4333). Passaging used 0.25% trypsin (Corning, cat#45001-082). C6/36 and Vero cell lines were cultured in EMEM (cat #30–2003) + 10% FBS + P/S. CCL125 cells were cultured in EMEM +20% FBS + P/S. C6/36 and CCL125 cells were cultured at 28°C with 5% CO2. A549, Vero, and U2OS cells were cultured at 37°C with 5% CO2. HPAEC cells were grown in EGM-2 Bulletkit Medium (Lonza, #CC-3162) at 37C with 5% CO2.

METHOD DETAILS

Generation and quantification of viral stock

Viral stocks were produced in Vero cells. Once the cells were >80% confluent, cells were infected at 0.1 MOI. Supernatants were harvested at either 48 or 72 h. After collecting the supernatant, dead cells were pelleted by centrifugation at 500g for 5 min. The remaining media was syringe filtered using a 0.45 μM filter.

Viral infections

Cells were plated one day before infection and grown to 60–80% confluency. Viruses were added in serum-free media at indicated MOIs. After 1 h, media was removed, cells were washed 1x with Dulbecco’s PBS (VWR, cat#02-0119-1000), and fresh media was added.

smFISH probe labeling

The smFISH probes were labeled with Atto-488, Atto-550, or Atto-633 using 5-Propargylamino-ddUTP-ATTO-488 (Axxora: JBS-NU-1619-488), 5-Propargylamino-ddUTP - ATTO-550 (Axxora: JBS-NU-1619-550), or 5-Propargylamino-ddUTP-Atto633 (Axxora: JBS-NU-1619-633) and). Oligos were labeled by adding 16uM oligos, 2υL of ddNTPs, 1x TDT buffer, and 1.5 υL of terminal deoxynucleotidyl transferase (Thermo Fisher Scientific: EP0161) in a 50uL total reaction volume. Oligos were then incubated overnight at 37°C and precipitated with 400 μL ethanol and 50 μL 3M Sodium acetate, pH 3.2. Oligos were then incubated at −20°C overnight and centrifuged at 20,000xg for 30 min. The supernatant was then aspirated, and the pellet was washed twice with 75% ethanol. Pellet was allowed to dry, then resuspended into 80 μL RNase free water. Oligo d(T)30-Cy3 were purchased from IDT.

Microscopy

Cells were plated onto glass coverslips (Fisher Scientific Co LLC:NC1418755) in 12-well plates(Greiner Bio-One:82050-930), then virally infected. After the indicated time of infection, the media was removed, cells were washed 1x in DPBS, and then fixed in 500uL 4% paraformaldehyde for 12 min. To permeabilize the cells, pfa was then removed, washed with PBS, and then 1mL 75% ethanol was added. Cells were then stored at 4°C for at least 2 h before beginning staining protocol.

For dual staining of immunofluorescence and smFISH, cells were washed 2x in PBS, then incubated in 500uL primary antibody in PBS for 4 h at 4°C. Cells were washed 2x in PBS, then secondary antibodies were added for 2 h at 4°C. Cells were washed 2x, then fixed in 4% pfa (Fisher Scientific Co LLC: 50980495) for 10 min. Cells were washed 3x in PBS, then washed in buffer A (filter-sterilized 2x SSC with 10% formamide) for 5 min smFISH probes were then added to a hybridization chamber (square Petri dish cat #) containing a wet paper towel with parafilm on top. 50uL of smFISH probes diluted 1:100 in hybridization buffer (0.45um filtered(Fisher Scientific Co LLC: 09-719D), 10% dextran sulfate(Fisher Scientific Co LLC: S4030), 10% formamide (Fisher Scientific Co LLC: BP227500), 1x nuclease-free SSC(Life Technologies Corporation: 15557044), diluted in nuclease-free water(Fisher Scientific Co LLC: 10977023)) were dropped onto the parafilm. Glass slips were then flipped onto the smFISH probes. Hybridization chambers were sealed with parafilm and incubated overnight at 37C. The next day, slips were washed 2x in Buffer A, then once in 2x SSC. Slips were then mounted on slides (cat #) with Vectashield (Vector Laboratories: 101098-044) and dried. Images were taken on a Nikon Eclipse Ti2 with a CFI60 Plan Apochromat Lambda D 100x Oil Immersion Objective Lens, N.A. 1.45, W.D. 0.13mm, F.O.V. 25mm, DIC, Spring Loaded. The filter set included: C-FL DAPI Filter Set, High-Signal-Noise, Semrock Brightline, Excitation: 356/30nm (341-371nm). Super-resolution microscopy was imaged using GATACA Systems Live-SR.

Post-processing and analysis of images was done using FIJI 2.14.0.

Live cell imaging

Cells were seeded on a 4-chamber 25-mm glass bottom dish (Cellvis, #D35C4-20-1.5N) at 80% confluency. A TOKAI Thermobox – Enclosure/Stage top incubator (Nikon, # 77025305) was used to maintain cells at 37C with 5% CO2.

Post-processing and analysis of images was done using FIJI 2.14.0.

RNA extraction

After infection, 1mL Trizol (Life Technologies: 15596018) was used to resuspend each well of a 6 well plate. All remaining steps were at 4C or on ice. 200uL chloroform (Fisher Scientific: BPC298500Z) was added, then tube was mixed by inversion. Tubes were centrifuged for 15 min at 13,000 rcf. Aqueous phase was slowly removed and added into a new tube. 400uL of 2-propanol (Fisher Scientific: BPA451SK1) was added. Tubes were mixed, then spun for 15 min at 13,000 rcf. Liquid was aspirated from pellet, then pellet was washed in 500uL 75% ethanol (Fisher Scientific: B09SBN6QQJ). Samples were centrifuged 13,000 rcf for 5 min, then ethanol was removed. Any remaining ethanol was evaporated off on a heat block at 65°C. Pellets were then resuspended in 50uL nuclease-free water.

Northern blot protocol

2-10ug of RNA extractions were loaded into a 5% acrylamide (Bio-Rad Laboratories: 1610156), 0.5x TBE (Bio-Rad: 1610733) gel. Samples were run at 150V for 1 h. Images were taken with ethidium bromide for visualization of RNAs. Gel was then transferred to BrightStar-Plus Positively Charged Nylon Membrane (Life Technologies: AM10102) at 30V for 90 min. Membrane was dried, then hybridized using an HL-2000 HYBRIDIZATION OVEN (Fisher: UVP95003101), and put in NorthernMax Prehybridization/Hybridization Buffer (Life Technologies: AM8677) for 1 h. Probes were labeled using ATP[γ-32P]-6000 Ci/mmol 10 mCi/ml EasyTide Lead, 250 μCi (PerkinElmer Health Sciences: NEG502Z250UC) and T4 PNK (NEB: M0201S) for 1 h in a 37°C water bath. Probes were added to blot in hybridization buffer overnight. Blots were then washed in 2x SCC, 0.1% SDS buffer 3x, then exposed to a phosphor screen (ThermoFisher, #NC2086042) overnight.

siRNA transfection

siRNAs were transfected using a reverse transfection according to the manufacturer’s protocol for lipofectamine 2000 (ThermoFisher, # 11668027). Cells recovered for 48 h, then were transfected with siRNAs a second time using a reverse transfection. Cells were then seeded for viral infection or transfection 48 h later.

T7 Transcription and transfection

T7 constructs were ordered from IDT and synthesized using HiScribe T7 High Yield RNA Synthesis Kit (NEB, #E2040S). RNA was purified via RNA extraction, as previously described, and quantitated via nanodrop. Transfection then followed protocols for lipofectamine 2000 (Invitrogen: 11668019), using 1ug/well poly(I:C) and 500ng/well synthetic RNA in 12-well plates.

Quantification of sfRNA granules

Quantification was performed using NIS-Elements AR. Percentage of cells containing sfRNA granules were manually counted. To measure the total percentage of sfRNA that colocalized to PABP granules, individual PABP puncta were circled as regions of interest, then the total fluorescence of those puncta was measured. Total sfRNA was measured by circling entire cells, then subtracting the fluorescence in the replication organelle, which was also manually circled.

smFISH probes

GAPDH smFISH probes (Biosearch Technologies Inc.: SMF-2026-1). All other probes were generated using Stellaris probe design tool.

FISH probes

Poly(A) RNA was detected using oligo(dT)18 Cy5, ordered from IDT.

Figure generation

Models were created using BioRender.com. Line tracings and graphs were created using Graphpad Prism and Microsoft Excel.

QUANTIFICATION AND STATISTICAL ANALYSIS

P-values were derived by student’s t-test (Excel). The specific test used for each figure is specified in each figure legend. *p < 0.05, **p > 0.01, ***p > 0.001 unless otherwise noted in the figure legend.

Error bars represent standard deviation, calculated using Microsoft Excel.

Highlights.

• Single-molecule imaging of sfRNA production and localization

• sfRNAs localize to RNase L-induced bodies

• RNase L-induced bodies sequester sfRNAs away from P-bodies

• Sequestration of sfRNAs by RNase L-induced bodies enhances decay of viral genomes