Main

Life, as far back as self-replicating genetic material, demanded the capacity to generate and maintain sufficient diversity to allow for adaptation and evolution. In the approximately 4 billion years that followed, self-replicating genetic entities, such as viruses, co-evolved with early cellular life, creating an unremitting host–pathogen arms race2. From prokaryotic systems such as clustered regularly interspaced short palindromic repeats (CRISPR) to the RNA interference (RNAi) systems of plants and invertebrates, many diverse antiviral strategies have proved successful in combating viral pathogens. As vertebrates transitioned from RNAi to interferon (IFN), remnants of this past system remained, including the two catalytically active RNase III members, Drosha and Dicer, both of which are critical for the biogenesis of microRNAs (miRNAs), a regulatory system designed on the same principles and machinery as antiviral RNAi3.

As miRNA biology stemmed from the development of RNAi4, but was not subjected to the selective pressures imposed by the biological arms race between host and pathogen, these components probably reflect the early eukaryotic RNAi machinery. This idea is supported computationally in model systems such as Drosophila where the antiviral Dicer (Dicer-2) underwent significant evolutionary changes as opposed to its miRNA counterpart (Dicer-1)5. Interestingly, of the two human RNase III nucleases, Drosha has greater homology than Dicer to the ancestral founder of this ancient domain4. The close relation of Drosha to its antiviral counterparts in invertebrates is particularly noteworthy given that this nuclease has recently been found to translocate to the cytoplasm after infection6,7,8.

To better understand Drosha biology as it relates to the cellular response to virus infection, we disrupted the gene encoding Drosha in a previously characterized human Dicer-deficient cell line (NoDice cells) that did not alter RNA virus replication9 (Extended Data Fig. 1a). We characterized these cells (herein referred to as RNaseIII−/− cells) by small RNA northern blot, probing for endogenous miR-93. As expected, mature miRNA levels in the parental NoDice cells could be restored by transfection of Dicer, whereas miR-93 expression in RNaseIII−/− cells required exogenous expression of both Dicer and Drosha (Fig. 1a). Small RNA sequencing from RNaseIII−/− cells confirmed a complete loss of all miRNAs both in the presence and in the absence of Sindbis virus (SINV) infection (Extended Data Fig. 1b).

Figure 1: Drosha mediates miRNA-independent antiviral activity.
figure 1

a, Northern blot of RNA from NoDice and RNaseIII−/− cells reconstituted with indicated plasmids. Blots probed for miR-93 and U6. b, Western blot of whole-cell extract from NoDice and RNaseIII−/− cells infected with SINV (multiplicity of infection (MOI) = 0.01) at 4, 8, and 12 hours post-infection (h.p.i.). Blot probed for SINV Capsid (SIN-C) and pan-Actin (Actin). c–f, Western blots of Ross River virus (RRV) (MOI = 0.1) (c), Langat virus (LGTV) (MOI = 0.1) (d), influenza virus (IAV) (MOI = 1) (e), and GFP-encoding Sendai virus (SeV) (MOI = 1) (f). Protein levels were assessed at 24 h.p.i. using virus-specific or GFP antibodies as indicated. g, RNA sequencing correlation analyses of NoDice and RNaseIII−/− cells at baseline. h, As described in g, except cells were treated with dsRNA for 8 h. Graphs in g and h depict data from biological replicates.

PowerPoint slide

Full size image

SINV replication was dramatically elevated in RNaseIII−/− cells, as determined by the production and kinetics of capsid (SINV-C) protein (Fig. 1b). The specificity of this activity was assessed by infecting cells with positive- versus negative-stranded RNA viruses including Ross River virus (Fig. 1c) and Langat virus (Fig. 1d), as well as influenza A virus (Fig. 1e) and Sendai virus (Fig. 1f), respectively. These data demonstrated that only viruses with positive polarity had enhanced protein production in the absence of Drosha. Transcriptome profiling of NoDice versus RNaseIII−/− cells at baseline revealed the induction of primary miRNAs and DGCR8, a known target for Drosha-mediated cleavage10,11 (Fig. 1g, Supplementary Table 1 and Extended Data Fig. 1c). Reconstitution of Drosha restored the transcriptome to that of NoDice cells (Extended Data Fig. 1d). In response to double-stranded RNA (dsRNA), NoDice versus RNaseIII−/− also showed little differential expression (Fig. 1h). Furthermore, we observed no defects relating to the IFN-I system (Extended Data Fig. 2a, b). This phenotype was corroborated in primary Drosha (Rnasenf/f) conditional mouse fibroblasts (Extended Data Fig. 2c–e).

We next sought to identify the molecular basis of the impact of Drosha on positive-stranded RNA viruses. To ensure that Drosha’s antiviral activity was not due to nuclear sequestration of a host RNA, we first generated a cell line expressing endogenous Drosha and a recombinant green fluorescent protein (GFP)–Drosha S300A/S302A (herein referred to as Drosha-2A), which localized exclusively to the cytoplasm12 (Extended Data Fig. 3a). Drosha-2A maintained enzymatic activity (Extended Data Fig. 3b) and, as predicted by our knockout studies, demonstrated an approximate 1-log decrease in SINV titres (Extended Data Fig. 3c). These data implicate cytoplasmic Drosha directly in the observed antiviral activity. Next, to map the domains required for this activity, we generated six deletion- and/or site-specific Drosha mutants (Fig. 2a), all unable to process pri-miRNAs, including the catalytic mutant E1045Q/E1222Q10 (herein referred to as RIIIDmut) (Fig. 2b). To test for antiviral activity of these mutants, we exogenously expressed them together with replication-competent genomic SINV RNA. While Dicer exhibited no suppression of SINV, Drosha-WT efficiently reconstituted antiviral activity (Fig. 2c, d). This viral inhibition was also evident with a catalytically inactive RBmut that failed to associate with its canonical RNA binding partner, DGCR8 (Fig. 2e, f).

Figure 2: The RNA binding domain of Drosha is essential for virus inhibition.
figure 2

a, Core domains and variants of Drosha including proline-rich (P-rich), arginine–serine-rich (RS-rich), conserved central domain (CED), RNaseIII domain (RIIID), and double-stranded RNA binding domain (dsRBD). Black boxes indicate point mutations (mut). b, Northern (NB) and western (WB) blots of RNaseIII−/− cells transfected with the indicated Drosha variants and cytoplasmic miR-124. Western blot depicts Drosha, Dicer, and pan-Actin. c, d, Western blot (c) and titres (d) from RNaseIII−/− cells expressing the indicated transcripts and SINV RNA 24 hours post-transfection (h.p.t.). Graph denotes average titres obtained; p.f.u., plaque-forming units; error bars, s.d. from three independent experiments. NS, not significant; all other conditions had P < 0.05 by Student’s t-test. e, Western blot as described in c with Drosha-RBmut. f, Western blot of input and Flag-immunoprecipitated (IP) fractions derived from HEK293T cells expressing indicated transcripts and Flag–DGCR8.

PowerPoint slide

Full size image

RNase III nucleases are known to recognize RNA stem-loop structures in all three domains of life13. To identify RNA that associated with the RNase III domain of Drosha, we performed systemic evolution of ligands by exponential enrichment (SELEX) with RBmut. Five rounds of enrichment identified a subset of unique RNAs that lacked conserved sequences but folded into miRNA-like hairpin structures (Fig. 3a). GFP pull-downs yielded two dominant RNA species with complex predicted secondary structures that were evident in all samples, which were subsequently used as controls for specificity (Fig. 3b). Using a recombinant DroshaRB fragment, we corroborated specific association between Drosha’s RNase III domain and an RNA hairpin identified by SELEX via electrophoretic mobility shift assay (EMSA), suggesting that no other host factors were required for recognition of the RNA in vitro (Fig. 3c). These data were further corroborated using immunoprecipitated proteins from whole-cell extract (Fig. 3d).

Figure 3: Cytoplasmic Drosha binds stem-loop structures in viral RNA to inhibit RdRp activity.
figure 3

a, RNA hairpins (HP-1/2) enriched by Drosha-RBmut-based SELEX as predicted by RNAfold. b, Same as a, using GFP as control (Ctl-1/2) bait. c, RNA-based EMSA of hairpins in a, and (b) using the recombinant Drosha RNA binding domain. BSA, bovine serum albumin. d, EMSA as in c, performed with immunoprecipitated GFP, Sendai nucleoprotein (SeV-NP), or Drosha RNA binding domain mutant (Drosha-RBmut). e, Northern blot of RNA from NoDice and RNaseIII−/− transfected with a SINV-based replicon denoting genomic (g) and subgenomic (Sub-g) SINV RNA. f, g, Luciferase (f) and antigenome expression (g) of replicon as described in e. Data are representative of independent experiments where each condition was done in triplicate. Error bars, s.d. h, Western blot of cytoplasmic membrane fractions from control or Drosha-2A cells expressing SINV replicase-components. i, In vitro minus-strand RNA synthesis assay using membrane fractions from h.

PowerPoint slide

Full size image

Given that the genomes of positive-stranded RNA viruses frequently utilize stem loops similar to those identified by SELEX14,15, we next investigated whether this RNase III domain could engage SINV RNA. Compared with Flag-tagged SeV-NP, immunoprecipitated RBmut protein–RNA complexes in SINV-infected cells showed a 1-log enrichment of viral RNA (Extended Data Fig. 4a, b). Moreover, RNA-EMSA confirmed that RBmut interacted with a specific hairpin at the 5′ end of the genome (Extended Data Fig. 4c, d).

To better understand how Drosha engagement of these structures impedes replication, we used a Gaussia luciferase-encoding Sindbis replicon system (Extended Data Fig. 5a). RNaseIII−/− cells produced significantly more genomic RNA, sub-gRNA, luciferase activity, and antigenome than the NoDice parental cells (Fig. 3e–g). These results suggest that Drosha is affecting RNA stability or translation, and/or directly blocking RNA-dependent RNA polymerase (RdRp) processivity. To assess the RNA stability hypothesis, we used a temperature-sensitive RdRp (RdRpts) SINV mutant that was inactive at 40 °C (ref. 16) (Extended Data Fig. 5b). At this non-permissive temperature, no significant differences in genomic RNA decay in infected NoDice and RNaseIII−/− were observed (Extended Data Fig. 5c). To determine whether Drosha impacted translation, we measured the activity of a firefly luciferase-encoding Sindbis construct in vitro (SIN-nsP3Luc) and found Drosha presence to be inconsequential (Extended Data Fig. 5d, e). Lastly, we assessed whether RdRp itself was directly impeded. To this end, we used a vaccinia virus-based system to generate functional minus-strand-specific replicase complexes, as previously described17. Cellular fractions containing both SINV replicase and Drosha were isolated and assayed in vitro, demonstrating that Drosha could reduce RdRp output by almost 50% (Fig. 3h, i).

As RNA hairpins, often essential for virus replication, are structurally conserved and evolutionarily constrained, they represent optimal targets for an antiviral system18. To determine whether Drosha from other phyla can recognize this putative pathogen-associated molecular pattern as demonstrated both in mammals and in arthropods19, we used miRNA-containing viruses to determine whether cytoplasmic processing, as an indicator of virus-induced translocation, could be observed. In zebrafish (Danio rerio), we demonstrated Drosha-dependent processing of an SINV-derived artificial miRNA in inoculated embryos (Fig. 4a and Extended Data Fig. 6a, b). We next examined cytoplasmic processing activity in plants. To accomplish this, we cloned pri-miR-124 into turnip crinkle virus (TCV) and infected Arabidopsis protoplast cultures, which, like zebrafish, supported cytoplasmic miRNA processing (Fig. 4b and Extended Data Fig. 6c, d).

Figure 4: RNase III inhibition of virus replication is highly conserved.
figure 4

a, Northern blot of RNA from zebrafish embryos treated with indicated morpholinos, and inoculated 2 days later with wild-type SINV (WT) or a strain encoding an artificial miRNA (amiRNA). Blot depicts amiRNA and U6 at 40 h.p.i. b, Northern blot of RNA from Arabidopsis thaliana protoplasts treated with a TCV containing a scrambled sequence (Scbl) or miR-124 (124). c, d, Western blot from RNaseIII−/− cells, co-transfected with the indicated plasmids and either in vitro transcribed SINV gRNA or Langat virus rescue plasmid. e, f, Small RNA sequencing of DL1 cells treated with indicated dsRNA for 3 days and subsequently infected with SINV (MOI = 1) (e) or Drosophila C virus (MOI = 7) (f). Graphs depict the number of reads mapping to indicated positions along the viral genomes.

PowerPoint slide

Full size image

Given the retention of this ancient activity, we turned to RNase III members from all domains of life (Extended Data Fig. 7a). Expression of each of these diverse nucleases, from bacteria, to archaea, to yeast, demonstrated robust antiviral activity against positive- but not negative- stranded viruses (Fig. 4c, d and Extended Data Fig. 7b). We next sought to demonstrate the antiviral activity of endogenous Drosha in an invertebrate. SINV-infected Drosophila melanogaster cells were pre-treated with dsRNA to knockdown Drosha as previously described20. Consistent with the results from mammalian cultures, loss of Drosha enhanced capsid expression (Extended Data Fig. 7c). Small RNA sequencing demonstrated that while more viral-derived RNAs were present in Drosha-depleted cells, the contours of mapped alignments were indistinguishable, ruling out a role for RNAi (Fig. 4e). This same phenotype was observed with Drosophila C virus in D. melanogaster, and SINV in human cells (Fig. 4f and Extended Data Fig. 7d).

These results suggest that the RNase III family of nucleases, used in all domains of life for the maturation of ribosomal and other structural RNAs, may represent an ancient RNA recognition platform that helped eukaryotes combat viruses. Interestingly, of the two RNase III proteins encoded in the human genome, Drosha is more related to bacterial RNase III than is Dicer4. In contrast, Dicer has been suggested to be the ancestor that gave rise to MDA5, which later duplicated and generated RIG-I21. Given these evolutionary connections, it is tempting to speculate that the RNase III domain of prokaryotes may have provided an early defence system for eukaryotes. In its most simple form, a Drosha-like homologue would recognize and engage RNA as an antiviral clamp but perhaps later evolve to utilize its enzymatic activity to give rise to RNAi. As the arms race continued, duplication, domain swapping, and time provided chordates with the opportunity to use RNase III domains to create new RNA sensors and the eventual development of the IFN-I system, which is believed to be fundamentally incompatible with RNAi21,22,23. As RNAi was functionally replaced with IFN-I during vertebrate development, the roles of Dicer and Drosha became largely dedicated to post-transcriptional regulation. Despite all the changes that have accrued over this expansive time, it would seem that fossils of this evolutionary arms race may remain in the basic biology of our miRNA machinery.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Viruses

Recombinant SINV, SINV124, and SINVamiRNA were previously described7,24. To generate SINV-GFP, the open reading frame of enhanced GFP was cloned under control of the additional subgenomic promoter of the SINV TE12Q construct7,25, replacing miR-124. Virus infections in cell culture were performed using the specified MOI in DMEM without FBS for 1 h. Generation of a pre-miR-124 hairpin-encoding TCV was done as previously described26. Briefly, we inserted a gBlock (Integrated DNA Technology) containing the 3′ end of the virus genome with the hairpin immediately downstream of the stop codon and flanked by BssHII and MluI into the MscI and SpeI sites of the pTCV66 construct. A control sequence was subsequently inserted by replacing miR-124 in the BssHII and MluI sites. A/Puerto Rico/8/34 (H1N1) influenza virus was rescued as previously described27. Ross River virus, strain T-48, was purchased from the American Type Culture Collection (ATCC) (VR-373) and amplified on BHK21 cells. Langat (strain TP21) and Sendai-GFP viruses were gifts from J. Lim and B. Lee, respectively. Vaccinia viruses encoding T7 DNA-dependent RNA polymerase, ubiquitin-tagged SINV nsP4, and the SINV polyprotein 123C>S, containing a mutation rendering nsP2 protease activity non-functional, were gifts from R. W. Hardy. Adenoviruses expressing GFP or Cre were purchased from Vector Biolabs (catalogue numbers 1768 and 1779).

Cell culture and reagents

Drosophila cells (DL1) were grown, maintained, and treated with dsRNA as previously described20,28. All mammalian cells were cultured in DMEM supplemented with 10% FBS and 1× penicillin/streptomycin. HEK293T cells were purchased from ATCC (CRL-3216). NoDice HEK293T cells were a gift from B. R. Cullen9. BSR-T7 cells were provided by K.-K. Conzelmann. Rnasenf/f primary lung fibroblasts were a gift from D. Littman and are described elsewhere29. HEK293T cells stably expressing GFP-tagged Drosha-WT or Drosha-2A were generated using a lentiviral expression vector and selected with puromycin (1 μg ml−1). To generate RNaseIII−/− cells, the CRISPR/Cas9 system was used to disrupt Drosha in NoDice cells. A custom-made plasmid encoding for two guide RNAs targeting the human RNASEN gene, nickase Cas9, and GFP (NickaseNinja) was purchased from DNA2.0. NoDice cells were transfected with this plasmid and GFP-expressing cells were single-cell sorted into 96-well plates. To identify clones lacking a functional Drosha protein, we PCR amplified and sequence verified the targeted genomic region. Co-transfection of plasmids and SINV in vitro transcribed genomic RNA was performed using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s instructions. In brief, 1.75 × 106 cells were transfected using 4 μg of plasmid DNA, 0.5 μg of RNA and 6 μl of transfection reagent. Transfection with in vitro transcribed SINV minigenome RNA was performed using Transmessenger Reagent (Qiagen) according to the manufacturer’s instructions. In brief, 1.5 × 105 cells were transfected with 0.8 μg RNA and 4 μl transfection reagent. All cell lines used in these studies were routinely screened for mycoplasma.

Plasmids

The CMV promoter-driven Langat virus rescue plasmid was a gift from A. G. Pletnev15. Sendai virus rescue plasmids encoding the SeV-GFP genome, SeV-N, SeV-P, and SeV-L were a gift from B. Lee. The pMini(+) SINV minigenome plasmid was a gift from R. W. Hardy17. GFP-tagged Drosha mutants were generated starting from pEGFP–Drosha-WT12. The two point mutations in the RIIIDs (E1045Q/E1222Q) were introduced by site-directed mutagenesis. To generate GFP-tagged deletion mutants, the desired fragment of Drosha was PCR amplified and cloned into pEGFP–Drosha-WT using the In-Fusion HD Cloning Kit (Clontech). GFP-tagged Ciona intestinalis and Streptococcus pyogenes constructs were generated using synthesized gBlock fragments of the carboxy (C)-terminal 425 amino acids of C. intestinalis or full-length S. pombe Pac1, and M. maripaludis or S. pyogenes RNase IIIs into the HindIII and KpnI sites of pEGFP–C1. A construct containing two tandem codon optimized FAK NES230 sequences was additionally inserted into the HindIII site of the pEGFP–C. intestinalis–Drosha construct using an In-Fusion HD Cloning Kit (Clontech). To generate a plasmid encoding for the SINV replicon expressing Gaussia luciferase, we PCR amplified the open reading frame of Gaussia luciferase and cloned it into the SINV TE12Q construct25 via NarI and ApaI using the In-Fusion HD Cloning Kit (Clontech). The resulting SINV replicon encodes for Gaussia luciferase fused to 76 nucleotides of the SINV capsid protein. To generate SINV expressing a temperature-sensitive RdRp (SIN-RdRpts), we substituted glycine 153 of non-structural protein four (ns4) to glutamic acid as previously described16. G153E was generated using a synthesized gBlock fragment cloned into SpeI and HpaI of the SINV TE12Q construct25.

Northern blot analysis

Total RNA was isolated using TRIzol (Thermo Fisher) according to the manufacturer’s instructions. Small RNA northern blotting was performed as previously described27,31. For large RNA northern analysis, 5 μg of total RNA in loading buffer (30% (v/v) formamide, 3% (v/v) formaldehyde, 5% (v/v) glycerol, 1× MOPS, and 0.01% (w/v) bromophenol blue) was separated in a 1% agarose gel in 1× MOPS buffer. RNA was transferred by capillary action in 20× SSC buffer (3 M NaCl, 0.3 M sodium citrate, pH 7.0) onto Hybond-NX nylon membrane (GE Healthcare). RNA was then ultraviolet cross-linked to the membrane (0.2 mJ cm−2) and incubated in blocking buffer (6× SSC and 7% SDS) for 1 h at 65 °C. Next, the membrane was incubated with a radio-labelled DNA oligonucleotide overnight at 42 °C in 6× SSC and 7% SDS, washed three times for 10 min in 3× SSC and 0.1% SDS, and exposed to film. Sequences of probes are listed in Supplementary Table 2.

Quantitative PCR

Total RNA was isolated with TRIzol (Thermo Fisher) and reversed transcribed using SuperScript II (Thermo Fisher) according to the manufacturer’s instructions. Complementary DNA was amplified with gene-specific primers using a KAPA SYBR Green Master Mix (KAPA Biosystems) and analysed on a LightCycler 480 (Roche). Sequences of gene-specific primers are listed in Supplementary Table 2.

RNA sequencing

For messenger RNA (mRNA) sequencing, libraries from two biological replicates per condition were prepared with a TruSeq RNA Library Preparation Kit version 2 (Illumina) according to the manufacturer’s instructions. In brief, 1 μg of total RNA was enriched for mRNA using oligo-dT beads. Purified RNA was fragmented and reverse transcribed, followed by second-strand synthesis, end repair, A-tailing, adaptor ligation, and PCR amplification. Prepared libraries were then quantified using a universal complete KAPA Library Quantification Kit (KAPA Biosystems). Libraries were sequenced on an Illumina MiSeq platform with a MiSeq Reagent Kit v3 or on a HiSeq 2500. Sequence reads were analysed with the RNA Express application (Illumina, BaseSpace) or custom pipelines. Alignment of reads to viral genomes was performed using Bowtie and subsequently visualized using the Integrative Genomics Viewer32. The correlation analyses were performed by comparing the natural logarithm of the normalized read counts (reads per million) between two experimental settings. A Pearson correlation coefficient (r) was calculated using the cor() function in R with default parameters. For each experimental setting the average normalized read count was calculated first using two biological replicates. Small RNA sequencing of infected DL1 cells was performed as previously described20. Briefly, cells were treated with indicated dsRNA for 3 days and subsequently infected with SINV (MOI = 1) or Drosophila C virus (MOI = 7) for 96 h before RNA extraction. Total RNA (40 μg) was separated on a 15% TBE-urea gel. The ~15 to 29 nucleotide fraction was extracted, eluted, and ethanol precipitated. Libraries were prepared using a Small RNA Sample Prep Kit, version 1.5 (Illumina, San Diego, California, USA) according to the manufacturer’s instructions and sequenced on an Illumina Genome Analyzer II. Libraries for small RNA sequencing of HEK293T and RNaseIII−/− cells were prepared using 1 μg of total RNA and a Small RNA Sample Prep Kit, version 1.5 (Illumina) according to the manufacturer’s instructions. The entire small RNA fractions (~15 to 200 nucleotides) were gel extracted and barcoded samples were PCR amplified for an additional 25 cycles before sequencing on an Illumina MiSeq platform with a MiSeq Reagent Kit version 3. Adapters were trimmed and analysis was performed using the Small RNA Sequencing App from Illumina. Raw and processed data have been deposited in the Gene Expression Omnibus under accession numbers GSE86610, GSE89790, GSE96889, GSE43031, and GSE98135.

Antiviral assays

RNaseIII−/− cells were co-transfected with Drosha, Drosha variants, or RNase III plasmids, and either in vitro transcribed SINV RNA or Langat rescue plasmid. At 24 h.p.t. for SINV, 30 h.p.t. for Langat, cells were harvested and virus replication was assessed by western blot. BSR-T7 cells were co-transfected with indicated RNase III constructs and a pool of Sendai virus rescue plasmids (described above). BSR-T7 samples were harvested at 48 h.p.t. for assessment of RNase III expression and at 72 h.p.t. for virus replication.

Immunoprecipitations

For immunoprecipitations of Flag–GFP and –DGCR8, 2 × 106 HEK293T cells were transfected with 2 μg each of the indicated plasmids. At 24 h.p.t., cells were lysed in 50 mM Tris pH = 7.5, 10% glycerol, 1 mM EDTA, 30 mM NaF, 1% NP-40, 150 mM NaCl, 1× PMSF, and 1× protease inhibitor cocktail and sonicated once for 15 s. Lysates were rotated overnight at 4 °C with 5 μl anti-Flag M2 antibody before addition of 25 μl protein G agarose beads. Samples were rotated for 1.5 h, washed in the same buffer used for lysis and then analysed by western blot.

RNA immunoprecipitation was performed as previously described33. Briefly, 36 h.p.t. of RNaseIII−/− cells with the indicated plasmids, cells were infected with SINV at an MOI of 3. At 7 h.p.i., cell lysates were prepared in lysis buffer (15 mM Tris/HCl pH 7.5, 150 mM NaCl, 15 mM MgCl2, and 1% (v/v) Triton X-100) supplemented with 10 mM of Ribonucleoside Vanadyl complex (NEB S1402S) and 1× protease inhibitor cocktail (Roche). Cell lysates were incubated with anti-Flag affinity resin (Sigma A2220) at 4 °C for 2 h followed by four washes of 5 min each with lysis buffer. Protein–RNA complexes were eluted with 300 μg ml−1 Flag-peptide (Sigma F3290) and purified RNA was extracted using TRIzol.

SELEX

SELEX was performed as previously described34. Briefly, a T7 promoter-driven library of 92,918 scrambled 110-base oligonucleotides was synthesized by CustomArray. The DNA oligonucleotide library was PCR amplified, followed by in vitro transcription using a MEGAscript T7 Transcription Kit (Thermo Fisher) according to the manufacturer’s instructions. RNA was heated at 65 °C for 5 min and immediately put on ice before the selection process. Flag-tagged proteins were purified from RNaseIII−/− cells at 24 h.p.t. in lysis buffer (15 mM Tris/HCl pH 7.5, 150 mM NaCl, 15 mM MgCl2, and 1% (v/v) Triton X-100) supplemented with 1× protease inhibitor cocktail (Roche). Cell lysates were incubated with anti-Flag affinity resin (Sigma A2220) at 4 °C overnight, followed by four washes of 5 min each with lysis buffer. One microgram of in vitro transcribed RNA in lysis buffer was added to the beads and incubated at 4 °C for 2 h, followed by four washes of 5 min each with lysis buffer. Protein–RNA complexes were eluted with 300 μg ml−1 Flag-peptide (Sigma F3290) and purified RNA was extracted using TRIzol. Purified RNA was reverse transcribed, PCR amplified, and in vitro transcribed to repeat the selection process. A total of five rounds of selection were performed.

EMSA

EMSAs were performed as previously described35. Flag-tagged proteins were purified from RNaseIII−/− cells at 24 h.p.t. in lysis buffer (15 mM Tris/HCl pH 7.5, 150 mM NaCl, 15 mM MgCl2, and 1% (v/v) Triton X-100) supplemented with 1× protease inhibitor cocktail (Roche). Cell lysates were incubated with anti-Flag affinity resin (Sigma A2220) at 4 °C overnight, followed by four washes of 5 min each with lysis buffer. Bound protein was eluted with 300 μg ml−1 Flag-peptide (Sigma F3290) in 20 mM Tris/HCl pH 8.0, 100 mM KCl, and 0.2 mM EDTA. Alternatively, a fragment of recombinant human Drosha protein was purchased from Abcam (ab94010) and 0.25–1 μg of this purified protein was used for EMSAs. Radioactively labelled RNA was in vitro transcribed from a PCR-amplified template using a MAXIscript Kit (Ambion). For complex formation of protein and RNA, protein eluate or recombinant protein was incubated with radio-labelled RNA (~100,000 counts per minute per reaction) for 30 min at room temperature in 20 mM Tris/HCl pH 8.0, 100 mM KCl, and 0.2 mM EDTA, supplemented with 0.1 μg μl−1 BSA and 1 mM DTT. Reactions were run on a 6% polyacrylamide gel.

In vitro cleavage assay

In vitro cleavage assays with Drosha were performed as previously described36.

In vitro minus-strand RNA synthesis assay

In vitro minus-strand assays were performed as previously described37. Briefly, HEK293T or HEK293T Drosha-2A cells were infected for 6 h with VacV-T7, -nsP4-ub, and - nsP123C>S at an MOI = 5 for each virus. Cells were resuspended in hypotonic buffer (10 mM Tris-HCl (pH 7.4) and 10 mM NaCl) and allowed to swell on ice for 15 min then lysed by 50 strokes through a 25-gauge 5/8-inch needle. Nuclei were removed by centrifugation (900g for 5 min at 4 °C). Membranes were isolated from the remaining supernatant by centrifugation at 15,000g for 20 min at 4 °C. Pellets were resuspended in hypotonic buffer plus 15% glycerol, aliquoted, and stored at −80 °C. Reactions were performed in a total volume of 50 μl in a mixture containing 50 mM Tris-HCl (pH7.4), 50 mM KCl, 3.5 mM MgCl2, 10 mM DTT, 10 μg actinomycin D per millilitre, 1 mM each of ATP, GTP, and UTP, 40 μM CTP, 1 μl CTP, α-32P-labelled 800 Ci mmol−1 10 mCi ml−1 EasyTide, 250 μCi per reaction, 1 μl RNaseOUT (Thermo Fisher, 10777019), 0.8 μg pMini(+) in vitro transcribed RNA template, approximately 25 μg (total protein input adjusted to ensure equal nsP4 levels across cell types, quantified from western blots using ImageJ software) membrane fractions, and H2O for 1.5 h at 30 °C before 5 U of alkaline phosphatase was added and the incubation continued for another 20 min. The reactions were terminated by the addition of SDS to 2.5% and proteinase K at a final concentration of 100 μg ml−1. RNA was isolated with TRIzol, equal volumes were run on a 1.25% agarose gel, transferred onto Hybond-NX nylon membrane (GE Healthcare), and exposed to film.

In vitro translation assay

In vitro translation assays were performed using a Rabbit Reticulocyte Lysate System (Promega) according to the manufacturer’s instructions. Membrane fractions (10 μg) from uninfected HEK293T and HEK293T Drosha-2A cells and 2 μg of SINV-nsP3Luc RNA were added to the reaction components and incubated for 2 h at 30 °C before analysis for firefly luciferase production.

Western blot analysis

Whole-cell extracts were separated on Mini Protean TGX gels and transferred onto nitrocellulose membrane (BioRad). SINV-specific ascites fluid (ATCC-VR-1248AF) and antibodies specific to actin (Thermo Fisher, MS-1295-P), GAPDH (Sigma, G9545), Flag (Sigma, F1804), NS1 (Center for Therapeutic Antibody Development, Icahn School of Medicine at Mount Sinai), STAT1a (Santa Cruz, c-111), and RIG-I (provided by A. Garcia-Sastre) were used at a dilution of 1:1,000 in 5% milk. Russian spring–summer encephalitis virus ascites fluid (ATCC- V558701562) was used for the detection of Langat virus at a dilution of 1:500 in 5% milk. Antibodies specific to GFP (Santa Cruz, sc-8334) and Ross River virus-specific ascites fluid (ATCC-VR-1246AF) were used at a dilution of 1:200 in 5% milk. The SINV nsP4 antibody was a gift from R. W. Hardy and was used at a 1:5,000 dilution in 5% milk in 25 mM Tris-HCl (pH 7.4), 137 mM NaCl, 2.7 mM KCl, and 0.1% Tween20. Mouse and rabbit HRP-conjugated secondary antibodies (GE Healthcare) were used at a 1:2,000 dilution in 5% milk. For antibody detection, Immobilon Western Chemoluminescent HRP (Millipore) and SuperSignal West Femto Maximum Sensitivity (Thermo Fisher) substrates were used according to the manufacturer’s instructions.

Luciferase assays

To assess translational inhibition by Drosha viral RNA binding, SINV replicon constructs (described above) were in vitro transcribed using an SP6 mMessage Machine Kit (Ambion) and co-transfected with a control Gaussia luciferase RNA as described for hairpin constructs. Luciferase expression was analysed at the indicated time points after transfection using a Promega Dual-Luciferase Reporter Assay System. In all experiments, Gaussia luciferase expression was normalized to firefly luciferase.

In vivo infections

Zebrafish embryos (AB strain) were randomly pooled in groups and injected with 8 ng of control (GAAAGCATGGCATCTGGATCATCGA) or of Drosha translation blocking (AGTGAAACTGAAGCAAACCTGTCCA) morpholinos at the one-cell stage (sex was not determined). At 2 days after fertilization, embryos were inoculated intravenously and intracranially, receiving a total of ~5,000 plaque-forming units of SINVwt or SINVamiRNA. RNA was extracted from surviving larvae at 40 h.p.i. Each sample was a pool of 5–11 larvae. Zebrafish experiments were performed in accordance with the Institut Pasteur Animal Care and Use Committee (reference CEEA 89).

Protoplasts were prepared from callus cultures of A. thaliana (Col-0-strain). Transfections were performed using polyethylene glycol as previously described38. Briefly, 15 μg of T7-transcribed TCV genomic RNA was transfected into 7 × 106 cells in a 60 mm Petri dish and incubated in the dark at 22 °C for 40 h. Samples were collected at 13,000 r.p.m. Total RNA was isolated using RNA extraction buffer (50 mM Tris-HCl pH 7.5, 5 mM EDTA pH 8.0, 100 mM NaCl, 1% SDS), followed by phenol–chloroform extraction and ethanol precipitation.

Fluorescence in situ hybridization

Fluorescence in situ hybridization (FISH) of A. thaliana protoplasts was performed following an established protocol39. Briefly, mock and TCV-infected protoplasts were fixed with methanol and hybridized with 4 μM Cy3-labelled probe complementary to bases 1210–1259 of the TCV genome (5′-Cy3-GCCTTCACGAATGTTTTGAGTTCTGCGTCCTTCCGGGATACCGGTCTGTC-3′). Cells were collected and washed with 2× SSC + 10% formamide before mounting with ProLong Gold + 4′,6-diamidino-2-phenylindole (DAPI) (Invitrogen). Cells were imaged using a laser scanning LSM 710 AxioObserver confocal microscope (Carl Zeiss). Images were acquired using a 63×/1.40 oil differential interference contrast (DIC) objective with 405 nm and 561 nm filters for DAPI and Cy3 respectively.

Zebrafish larvae were infected as described above and fixed at 24 .p.i. Whole-mount FISH was performed using a FITC-labelled SINV capsid probe amplified with anti-FITC-AF488. Hoechst dye was used for detection of nuclei. Single confocal plains from an infected region in the tail were imaged using a 40× objective.

Data availability

Deep sequencing data have been deposited in the Gene Expression Omnibus under accession numbers GSE86610, GSE89790, GSE96889, GSE43031, and GSE98135. All other data are available from the corresponding author upon reasonable request.