tenOever LaboratoryVirology · Host defense · RNA biology
Home / Publication
Publication

RNase III nucleases from diverse kingdoms serve as antiviral effectors

lab-led

RNase III nucleases, including human Drosha and homologues from bacteria, archaea, yeast and a urochordate, restrict positive-strand RNA viruses by recognizing unbranched RNA stem loops and impairing the viral polymerase, separably from microRNA biogenesis, catalysis and interferon.

Lauren C. Aguado; Sonja Schmid; Jared May; Leah R. Sabin; Maryline Panis; Daniel Blanco-Melo; Jaehee V. Shim; David Sachs; Sara Cherry; Anne E. Simon; Jean-Pierre Levraud; Benjamin R. tenOever

2017 · Nature · primary research

DOIPubMedOpen-access full textPDF · PubMed Central
Senior authors
Benjamin R. tenOever
Correspondence
Benjamin R. tenOever

Research areas & themes

Small RNA Biology and the Limits of Antiviral SilencingRNase III Enzymes as Interferon-Independent Antiviral EffectorsThe Evolutionary Logic of Antiviral Defense

Citation

Aguado LC, Schmid S, May J, Sabin LR, Panis M, Blanco-Melo D, Shim JV, Sachs D, Cherry S, Simon AE, Levraud JP, tenOever BR. RNase III nucleases from diverse kingdoms serve as antiviral effectors. Nature. 2017. 547(7661), 114-117.

DOI 10.1038/nature22990. PMID 28658212. PMCID PMC5846625.

Lauren C. Aguado and Sonja Schmid are marked in the paper as having contributed equally.

One-sentence contribution

RNase III nucleases, including human Drosha and homologues from bacteria, archaea, yeast and a urochordate, restrict positive-strand RNA viruses by recognizing unbranched RNA stem loops and impairing the viral polymerase, separably from microRNA biogenesis, catalysis and interferon.

Executive summary

Drosha and Dicer are the two catalytically active RNase III enzymes in humans and are usually considered in terms of microRNA biogenesis. Earlier work had reported that Drosha leaves the nucleus after virus infection, which raised the question of what it does there. Disrupting Drosha in an existing Dicer-deficient human cell line produced cells with no mature microRNAs at all, and in those cells Sindbis virus replicated markedly better. The effect was specific to positive-strand RNA viruses, since Ross River virus and Langat virus were also enhanced while influenza A virus and Sendai virus were not, and it was reproduced in primary conditional Drosha mouse fibroblasts. Structure and function mapping showed that a Drosha variant that is catalytically inactive, cannot process primary microRNAs, and does not bind DGCR8 still restricted the virus, while Dicer did not, so the activity does not run through microRNA production. SELEX with the RNA-binding mutant enriched unbranched hairpins, and the same protein bound a hairpin in the first 200 nucleotides of the Sindbis genome. A replicon system showed increased genomic RNA, subgenomic RNA, antigenome, and luciferase in knockout cells without detectable changes in RNA decay or translation, and an in vitro minus-strand synthesis assay showed Drosha reducing polymerase output by nearly half. Cytoplasmic processing of virus-encoded microRNA hairpins in zebrafish embryos and Arabidopsis protoplasts, and antiviral activity of RNase III proteins from three domains of life, extend the phenomenon beyond mammals.

Scientific context

The introduction frames antiviral defense in evolutionary terms. Prokaryotes contend mostly with DNA viruses and use systems such as CRISPR, while eukaryotic compartmentalization allowed RNA viruses to expand, creating pressure for RNA-targeting defense, met in plants and invertebrates by antiviral RNA interference and in vertebrates largely by the interferon system. Drosha and Dicer are presented as retained components of that older machinery, with prior work cited indicating that Drosha is more closely related than Dicer to the ancestral bacterial RNase III and that in Drosophila the antiviral Dicer-2 has diverged sharply from the microRNA-dedicated Dicer-1. Three prior reports, two from this laboratory, had found that Drosha translocates to the cytoplasm after infection, and a 2014 paper from the laboratory had already described Drosha as an interferon-independent antiviral factor. What was missing was an account of what cytoplasmic Drosha recognizes and how that recognition restricts a virus, separable from its role in the microprocessor.

Central question

Does Drosha possess an antiviral activity that is independent of microRNA biogenesis and of the interferon system, and if so, what does it recognize in viral RNA and by what means does that recognition limit replication, and is the activity restricted to Drosha or shared across the RNase III family?

Experimental strategy

The genetic starting point removes the usual confound. Working in a published Dicer-deficient human line in which RNA virus replication was unaltered, the authors disrupted Drosha as well, so that the resulting cells lack mature microRNAs entirely and any antiviral phenotype cannot be attributed to loss of a particular microRNA. Virus panel selection then tests polarity as a variable, with two alphaviruses and a flavivirus against an orthomyxovirus and a paramyxovirus. Separation of function is pursued through mutants rather than inference, using a panel of six Drosha variants all incapable of processing primary microRNAs, including a catalytic mutant and an RNA-binding mutant that fails to associate with DGCR8, and using a phosphorylation variant that sits exclusively in the cytoplasm to ask whether the cytoplasmic pool alone suffices. To find the ligand without assuming it, SELEX is run against the RNA-binding mutant and the enriched structures are then sought in the viral genome and tested directly by mobility shift. To decide among stability, translation, and polymerase inhibition as mechanisms, each is tested separately, with a temperature-sensitive polymerase mutant for decay, an in vitro luciferase construct for translation, and a reconstituted minus-strand replicase assay for polymerase output. Finally, conservation is examined across taxa by expressing RNase III proteins from bacteria, archaea, yeast and a urochordate, and by using virus-encoded microRNA hairpins as reporters of cytoplasmic RNase III activity in zebrafish and plants.

Key findings

  1. Cells lacking both Dicer and Drosha have no detectable mature microRNAs with or without Sindbis infection, and restoring miR-93 requires both proteins (Fig. 1a and Extended Data Fig. 1b). This establishes the genetic background for everything that follows.
  2. Sindbis virus capsid production and kinetics were strongly increased in the double knockout cells (Fig. 1b). Ross River virus and Langat virus were likewise enhanced, while influenza A virus and Sendai virus were not (Fig. 1c to 1f), so enhancement tracked with positive genome polarity.
  3. Baseline transcriptome comparison showed induction of primary microRNAs and of DGCR8, both expected consequences of losing Drosha, and reconstitution of Drosha restored the transcriptome (Fig. 1g and Extended Data Fig. 1c and 1d). The double-stranded RNA response showed little differential expression (Fig. 1h) and no defects in the type I interferon system were observed (Extended Data Fig. 2a and 2b), which the authors use to exclude an indirect interferon explanation.
  4. The phenotype was reproduced in primary conditional Drosha mouse lung fibroblasts (Extended Data Fig. 2c to 2e), showing it is not an artifact of one engineered human line.
  5. A green fluorescent protein-tagged Drosha carrying serine to alanine substitutions at positions 300 and 302 localized exclusively to the cytoplasm, retained enzymatic activity, and reduced Sindbis titers by roughly one log (Extended Data Fig. 3). The authors read this as implicating the cytoplasmic pool of Drosha directly.
  6. Among six Drosha variants unable to process primary microRNAs, wild-type Drosha reconstituted antiviral activity while Dicer did not, and the RNA-binding mutant that is catalytically inactive and fails to associate with DGCR8 still suppressed the virus (Fig. 2b to 2f). The antiviral activity therefore requires RNA binding and does not require catalysis, DGCR8, or microRNA processing.
  7. Five rounds of SELEX with the RNA-binding mutant enriched RNAs with no conserved sequence that fold into unbranched, microRNA-like hairpins (Fig. 3a), and a recombinant Drosha RNA-binding fragment bound an enriched hairpin by mobility shift with no other host factors present (Fig. 3c and 3d). Recognition is thus structural rather than sequence-specific in this assay.
  8. In Sindbis-infected cells, immunoprecipitated Drosha RNA-binding mutant complexes were enriched about one log for viral RNA relative to a Sendai nucleoprotein control (Extended Data Fig. 4a and 4b), and mobility shift confirmed binding to a hairpin within the first 200 nucleotides of the genome (Extended Data Fig. 4c and 4d).
  9. A Sindbis replicon produced more genomic RNA, subgenomic RNA, antigenome, and luciferase in the knockout cells (Fig. 3e to 3g). Among the candidate explanations, RNA decay was excluded using a temperature-sensitive polymerase at the nonpermissive temperature (Extended Data Fig. 5c) and translation was excluded using an in vitro luciferase-encoding construct (Extended Data Fig. 5d and 5e).
  10. In a vaccinia-based reconstitution of minus-strand replicase complexes, fractions containing both Sindbis replicase and Drosha showed polymerase output reduced by almost 50 percent (Fig. 3h and 3i). This is the direct biochemical support for polymerase impairment.
  11. RNase III proteins from bacteria, archaea, yeast, and Ciona intestinalis conferred antiviral activity against positive-strand but not negative-strand viruses when expressed in the knockout cells (Fig. 4c and 4d and Extended Data Fig. 7a and 7b), indicating the activity is a general property of the domain rather than of Drosha specifically.
  12. Cytoplasmic RNase III activity outside mammals was inferred from processing of virus-delivered microRNA hairpins, Drosha-dependent processing of a Sindbis-encoded artificial microRNA in zebrafish embryos (Fig. 4a) and processing of pri-miR-124 delivered by turnip crinkle virus in Arabidopsis protoplasts (Fig. 4b).
  13. In Drosophila DL1 cells, knockdown of Drosha increased Sindbis capsid expression and increased Drosophila C virus replication (Extended Data Fig. 7c, Fig. 4e and 4f). Small RNA sequencing showed more virus-derived small RNAs in Drosha-depleted cells but indistinguishable alignment contours, which the authors use to argue against an RNA interference explanation.

Mechanistic model

The model proposed is that cytoplasmic Drosha, and by extension other RNase III proteins, binds unbranched stem loops in the genomes of positive-strand RNA viruses and acts as a clamp, physically obstructing the viral RNA-dependent RNA polymerase rather than cleaving the RNA. Several components are demonstrated. Binding is to structure and requires only the RNA-binding region, catalysis is dispensable, decay and translation are excluded as the affected step in the systems tested, and polymerase output falls by nearly half in vitro when Drosha is present in the replicase-containing fraction. The clamp language is the authors' interpretation of that set of results and is not itself observed, since no structural or single-molecule evidence for steric occlusion is presented, and the study does not establish which hairpin engagement events in an infected cell produce the replication defect. The restriction of the phenotype to positive-strand viruses is consistent with the model, given that these genomes serve directly as polymerase templates in the cytoplasm, but the paper does not test why negative-strand genomes, which are encapsidated, escape. The evolutionary account in the discussion, in which an ancestral RNase III clamp preceded RNA interference and later gave rise to the sensors of the interferon system, is explicitly offered as speculation by the authors and should be read as category 2 at best.

Conceptual or technical advance

The work separates an antiviral function of Drosha from everything Drosha is normally studied for. Because the experiments are done in cells with no microRNAs at all and with mutants that cannot process, cannot cleave, and cannot bind DGCR8, the activity cannot be explained by loss of a regulatory microRNA, and the absence of interferon defects removes the other common indirect route. That leaves direct RNA binding as the operative property, and SELEX supplies an unbiased description of what is bound, namely unbranched stem loops of the kind that positive-strand genomes require and cannot easily discard. The demonstration that RNase III proteins from bacteria through yeast to a urochordate share the activity converts a Drosha observation into a statement about a protein domain, and it makes testable the idea that structured RNA recognition by an ancient nuclease fold can function as defense without any downstream silencing pathway.

Relationship to the broader research program

This is the mechanistic extension of a line the laboratory had been building for several years. Prior laboratory work had shown noncanonical cytoplasmic processing of viral microRNAs, presented evidence for a cytoplasmic microprocessor, and identified Drosha as an interferon-independent antiviral factor, all cited here. The engineering of viruses to carry microRNA hairpins, developed in the laboratory for delivery and for targeting, is used here in a different role, as a reporter that reveals cytoplasmic RNase III activity in a zebrafish embryo or a plant protoplast. The evolutionary framing connects to the senior author's review on the evolution of antiviral defense systems, also cited. The recurring question across these papers, namely what the relationship is between the small RNA machinery and antiviral defense in vertebrates, is addressed here by decoupling the two, and that answer sits directly against the position taken in the 2013 Minireview in this corpus, where the somatic-cell case for antiviral RNA interference was judged unproven. Reading those two together is category 3 synthesis, and it belongs to central assembly rather than to either record alone.

  • Shapiro et al. 2014, Drosha as an interferon-independent antiviral factor. Predecessor from the tenOever laboratory, the direct antecedent of the antiviral activity characterized here.
  • Shapiro et al. 2012, evidence for a cytoplasmic microprocessor of pri-miRNAs. Predecessor from the same laboratory, source of the cytoplasmic translocation premise.
  • Shapiro et al. 2010, noncanonical cytoplasmic processing of viral microRNAs. Methodological foundation, source of the virus-encoded hairpin reporters used in zebrafish and plants.
  • tenOever 2016, the evolution of antiviral defense systems. Review or synthesis from the same laboratory, the framework the discussion argues within.
  • tenOever 2013, RNA viruses and the host microRNA machinery. Review or synthesis from the same laboratory, cited for the relationship between microRNA machinery and RNA interference.
  • Bogerd et al. 2014, derivation and characterization of Dicer- and microRNA-deficient human cells. Methodological foundation from the Cullen group, the parental cell line used here.
  • Sabin et al. 2013, Dicer-2 processes diverse viral RNA species. Methodological foundation from the Cherry group, the source of the Drosophila knockdown and small RNA sequencing approach, with the same first author contributing the fly data here.
  • Cullen, Cherry and tenOever 2013, is RNA interference a physiologically relevant innate antiviral immune response in mammals. Conceptual extension, two of its authors appear here and the present work bears on the question that review left open.

Limitations and boundaries

The enhancement phenotype is established for a small panel of viruses, and the positive-strand specificity rests on two alphaviruses and one flavivirus against one orthomyxovirus and one paramyxovirus, all in cultured cells. Most of the mammalian mechanism is worked out in one engineered HEK293T-derived line and in transfection or reconstitution settings, with the primary mouse fibroblast experiment serving as the only independent mammalian confirmation, and no whole-animal mammalian infection is reported. The polymerase inhibition result is a roughly 50 percent reduction in an in vitro assay using fractions that contain both replicase and Drosha, which shows interference with output but does not show direct contact with the polymerase or exclude an intermediary in the fraction. The clamp model is an interpretation and no structural data accompany it. Cross-species evidence outside Drosophila is indirect, since zebrafish and plant experiments measure processing of an engineered hairpin as a proxy for cytoplasmic RNase III activity rather than measuring virus restriction in those hosts. Heterologous expression of bacterial, archaeal, and yeast RNase III proteins in human cells shows what those domains can do in that setting and not what they do in their own organisms. The evolutionary narrative in the discussion is labeled speculative by the authors and is not supported by experiment.

Audience summaries

25 words

Drosha restricts positive-strand RNA viruses by gripping structured RNA rather than cutting it, an activity shared by RNase III enzymes from bacteria, archaea, yeast, and animals.

75 words

Drosha is known for making microRNAs, but it also leaves the nucleus during infection. Removing it from cells that already lacked microRNAs let positive-strand RNA viruses grow much better, and a version of Drosha that cannot cut RNA or make microRNAs still blocked them. The protein binds hairpin structures in viral genomes and cuts polymerase output roughly in half. Related enzymes from bacteria through yeast did the same, suggesting a very old defensive capability.

150 words

Using human cells engineered to lack both Dicer and Drosha, and therefore all mature microRNAs, this work shows that Drosha restricts positive-strand RNA viruses through a route independent of microRNA biogenesis, of catalysis, of its partner DGCR8, and of interferon. Sindbis, Ross River, and Langat viruses replicated better without Drosha while influenza and Sendai viruses did not. SELEX against the RNA-binding mutant enriched unbranched stem loops carrying no shared sequence, and the same protein bound a hairpin within the first 200 nucleotides of the Sindbis genome. Replicon experiments localized the defect to RNA synthesis rather than decay or translation, and a reconstituted minus-strand assay showed Drosha reducing polymerase output by nearly half, which the authors interpret as steric hindrance by an antiviral clamp. RNase III proteins from bacteria, archaea, yeast, and a urochordate reproduced the activity, and cytoplasmic RNase III function was detected in Drosophila, zebrafish, and Arabidopsis.

Discoveries supported by this paper

Documented publication relationships

Pathogens

Influenza A virusSindbis virusSendai virusDrosophila C virusLangat virusRoss River virusTurnip crinkle virus

Technologies

Bulk RNA sequencingSmall RNA deep sequencingImmunoblottingNorthern blotElectrophoretic mobility shift assayIn vitro reconstitution with purified componentsCoimmunoprecipitationCRISPR-Cas9 knockout cell linesMinigenome, minireplicon and replicon reportersMorpholino knockdownSELEX selection of bound RNA