Small RNA Biology and the Limits of Antiviral Silencing

Do mammals use small RNAs against viruses, and if not, why not?

In one paragraph

Beginning from the technical question of whether an RNA virus can be made to produce a microRNA, this body of work became a sustained argument about whether vertebrates use small RNA silencing against viruses at all. The laboratory built engineered viruses that produce functional microRNAs from the nucleus and from the cytoplasm, corrected its own initial account of how the cytoplasmic products are made, and then used one of those tools to ask whether a virus armed with the ability to destroy host small RNAs gains fitness. It does not. The laboratory then rebuilt the missing system artificially and showed that it protects mice from lethal influenza challenge with no contribution from type I interferon signalling, which removed the simplest explanation for why vertebrates gave the system up. Two further lines complicate the picture. Drosha does restrict positive-strand RNA viruses, but by binding structured RNA rather than by producing small interfering RNAs, and microRNAs do act during infection, but on cytokine output rather than on the intrinsic antiviral response.

The defining scientific question

Plants, nematodes and arthropods process viral double-stranded RNA into small interfering RNAs that guide cleavage of viral transcripts, and viruses of those hosts encode dedicated suppressors of the pathway. Vertebrates detect the same double-stranded RNA and answer it transcriptionally, retaining most of the silencing machinery but apparently using it for microRNA regulation instead. Whether that appearance is correct was a live dispute, reopened in 2013 by reports from the Voinnet and Ding groups arguing for antiviral RNA interference in mammalian cells, and this laboratory took a public position in it. The empirical half of the question asks whether vertebrate somatic cells use small RNAs against viruses. The explanatory half asks why not, and requires either that such a defence cannot work in a mammalian cell or that something else displaced it.

Origins

The line begins with engineering rather than with immunology. Varble 2010 removed an assumed prohibition, that an RNA virus could not encode a microRNA because RNase III excision of the hairpin would fragment the genome and the genome would itself be a perfect target for the product. Splitting the overlapping NS1 and NEP/NS2 reading frames of influenza A virus segment 8 created an intron, and a cellular pri-microRNA placed there produced mature miR-124 within four hours of infection at levels comparable to abundant endogenous microRNAs, with multicycle growth matching wild type and no detectable cleavage of the genome. Target sites on the genomic strand were not silenced while identical sites in messenger RNA were, attributed by the authors to nuclear ribonucleoprotein organisation as an inference rather than a demonstration.

Shapiro 2010 in RNA then asked the same question of a virus that never enters the nucleus. Sindbis virus carrying the pri-miR-124-2 locus produced abundant mature miR-124 in a Dicer-dependent but DGCR8-, Exportin-5- and interferon-independent manner, silenced a reporter, and attenuated itself by roughly two logs at low multiplicity. Langlois 2012 extended this to a negative-sense cytoplasmic virus and into animals, answering a published objection that cytoplasmic processing was an artifact of rapidly dividing transformed cells. These three papers supply the tools the rest of the area depends on.

Major findings

Cytoplasmic pri-microRNA processing requires Drosha, and infection relocalises Drosha to the cytoplasm. Shapiro 2012 in RNA established the requirement by conditional deletion, with Dicer needed only at the second cleavage, and showed endogenous Drosha redistributing on infection with parental as well as microRNA-expressing virus. The cytoplasmic cleaving complex itself was never identified.

Destroying host small RNAs gives a virus no advantage. Backes 2014 armed vesicular stomatitis virus with vaccinia VP55, the poly(A) polymerase subunit that Backes 2012 had identified as tailing and degrading RISC-loaded small RNAs. The armed virus destroyed microRNAs efficiently in fibroblasts, primary macrophages and mouse lung, gained no fitness in any of them, and was attenuated by about a log in wild-type mice through derepression of interferon-stimulated transcripts normally held down by microRNAs. In mice lacking both type I and type III interferon receptors all viruses reached comparable titres, which removes the objection that a silencing contribution might be hidden beneath interferon. The authors call this a strong argument rather than a demonstration, since absence of an activity is difficult to prove.

A reconstructed small RNA defence protects without interferon. Benitez 2015 inserted perfectly complementary target sites for host microRNAs into influenza A virus, and separately had the virus encode a small interfering RNA against its own nucleoprotein segment. A five-site virus grew in eggs but was undetectable in mammalian cells and caused no morbidity in mice at 25,000 plaque-forming units, including in animals lacking the type I interferon receptor. The authors frame this as a possibility claim, that chordates could have used RNA interference in place of interferon.

Drosha is antiviral without making small interfering RNAs. Shapiro 2014 showed that deleting Drosha, but not Dicer, raised Sindbis and vesicular stomatitis virus titres in primary fibroblasts, that infection and double-stranded RNA drive Drosha out of the nucleus by CRM1-dependent export without requiring RIG-I, TBK1 or IFNAR1, and that no virus-derived small interfering RNA signature accompanies the restriction. Aguado 2017, in cells lacking all mature microRNAs, found the enhancement specific to positive-strand RNA viruses, showed that an RNA-binding Drosha mutant which cannot cleave, cannot process and cannot bind DGCR8 still suppresses the virus, identified unbranched hairpins as the recognised structure by SELEX, and reproduced the activity with RNase III proteins from bacteria, archaea, yeast and a urochordate. The clamp model is interpretation, unaccompanied by structural evidence.

MicroRNA function during infection is confined to cytokine control. Aguado 2015 delivered VP55 from an inert adenovirus vector, removing microRNAs from primary cells within a day without inducing interferon-stimulated genes. Only 12 of 1,548 genes induced by double-stranded RNA and 12 of 179 induced by interferon beta changed, whereas nine days of depletion changed more than 1,700 transcripts, dominated by chemokines and cytokines, while leaving core interferon machinery untouched.

Escape from an engineered silencing pressure runs through the virus in one case and the host in another. Aguado 2018 found negative-strand viruses cleared by more than five logs while positive-strand viruses recovered by precise excision, with a recombination-defective poliovirus unable to escape, which makes template switching rather than polarity itself the requirement. Uhl 2023 then found the route available to negative-strand viruses, and it is not a viral adaptation. Host ADAR1 edits adenosines within the target sites, its knockout abolishes escape in all 96 wells and reconstitution restores it, and human ADAR1 suppresses endogenous silencing in Nicotiana benthamiana.

How the work evolved

Two revisions matter more than the accumulation of results.

The first is internal. Shapiro 2010 reported cytoplasmic microRNA processing as microprocessor independent and named the products virtrons. Shapiro 2012 in RNA, using conditional deletion rather than inference, found Drosha absolutely required and said so, revising the earlier reading. That correction redirected the programme toward Shapiro 2014 and Aguado 2017, because a nuclear RNase III enzyme in the cytoplasm during infection is a more interesting object than an unexplained processing activity.

The second is the trajectory of the position itself. tenOever 2013 argues from quantitative thresholds that chordate microRNAs cannot be antiviral. Cullen 2013, a Minireview led by Bryan Cullen with Sara Cherry and tenOever and classified collaborative, is more cautious than this laboratory's own review of the same year. It accepts the mouse embryonic stem cell evidence, declines to settle the somatic case, and specifies the criteria that would settle it, noting that the viral proteins on which the positive claims rest also antagonise interferon so that the two explanations have not been separated. Backes 2014 then supplies the direct test with a negative result, and Benitez 2015 the complementary positive result, that the system mammals do not use would have worked. Neither half of that pair makes sense alone.

The reading of VP55 also shifted with the vehicle carrying it. In Backes 2012 it was evidence that host microRNAs impose a cost on a virus. Delivered from a replicating virus in Backes 2014 it showed no fitness gain, but the inflammatory environment of that vector confounded any separation of microRNA effects on cytokines from secondary interferon-stimulated gene induction. Aguado 2015 is presented by its own authors as the correction of that confound.

Some threads were not returned to. The enzyme that cleaves a cytoplasmic primary microRNA transcript was never identified. The virtron concept, proposed in Shapiro 2010 with the authors' own caution that self-targeting makes natural virtrons unlikely to be genuine viral products, does not recur as a subject of investigation. Morales 2017 extends small viral RNAs to coronaviruses, but was conceived and executed by the Enjuanes and Sola group in Madrid, with the tenOever contribution recorded as reagents, conceptual advice and manuscript writing. Oishi 2023, assigned here for its cross-kingdom logic, belongs primarily to influenza genome regulation and carries a declared conflict of interest tied to commercialisation of L7Ae.

Principal publications

Shapiro 2012 in RNA and Shapiro 2014 for the reassignment of Drosha. Backes 2014 for the negative answer on mammalian antiviral silencing, and Benitez 2015 for the reconstruction and the interferon-independent protection. Aguado 2017 for the microRNA-independent antiviral activity of the RNase III domain across three domains of life, and Aguado 2015 for the positive account of what microRNAs do during infection. Aguado 2018 and Uhl 2023 for the two escape routes, and tenOever 2016 for the evolutionary framing.

Connections to other areas

The engineered viruses that begin this area are the working tools of programmable virology, where the vectors of Varble 2010 and Langlois 2012 serve delivery rather than biogenesis questions. The cytokine target set of Aguado 2015 and the microRNA restraint on interferon-stimulated transcripts reported in Backes 2014 connect to innate immune signalling and to how the magnitude of the interferon response is set. Benitez 2015, Aguado 2018 and Uhl 2023 are also assigned to viral populations and evolution under recombination and escape from selective pressure. Morales 2017 and Oishi 2023 reach into influenza genome regulation.

Current implications

Because vertebrate RNA viruses have no reason to antagonise the small RNA machinery, it remains available as an engineering handle, which is the premise of the attenuation and biocontainment work elsewhere in the corpus. Because escape from microRNA-based targeting occurs by excision in positive-strand viruses and by host editing in at least one negative-strand virus, designs relying on such targeting will erode in predictable ways, a point Uhl 2023 makes directly for vector design. And because the RNase III fold restricts positive-strand RNA viruses by binding structure rather than by cutting, structured RNA recognition is a candidate antiviral mechanism outside the interferon axis, with the caveat that no mammalian animal infection was performed in either paper establishing it.

Open questions

The activity that converts a cytoplasmic primary microRNA transcript into a precursor remains unidentified, as do the phosphatase implied by the serine 300 and 302 data in Shapiro 2014, the sensing route that triggers Drosha export, and how cytoplasmic Drosha distinguishes viral from host structured RNA. Whether the clamp model of Aguado 2017 describes a real steric event, and why encapsidated negative-strand genomes escape it, are untested. Whether ADAR1 is recruited by the silencing complex or acts on duplex viral RNA is unresolved in Uhl 2023, which also reports that a five-target influenza A virus under the same regime showed neither editing nor escape. The evolutionary explanation remains a hypothesis by the explicit statement of tenOever 2016, and the phylogenetic support in Uhl 2023 is called limited in scope and correlative by its own authors. The possibility that residual antiviral silencing is confined to pluripotent cells was not examined experimentally here, and Backes 2014 lists the absence of stem cell work among its own limitations.

Publications referenced

• Noncanonical cytoplasmic processing of viral microRNAs
• Engineered RNA viral synthesis of microRNAs
• Degradation of Host MicroRNAs by Poxvirus Poly(A) Polymerase Reveals Terminal RNA Methylation as a Protective Antiviral Mechanism
• In Vivo Delivery of Cytoplasmic RNA Virus-derived miRNAs
• Evidence for a cytoplasmic microprocessor of pri-miRNAs
• Is RNA Interference a Physiologically Relevant Innate Antiviral Immune Response in Mammals?
• RNA viruses and the host microRNA machinery
• The Mammalian Response to Virus Infection Is Independent of Small RNA Silencing
• Drosha as an interferon-independent antiviral factor
• microRNA Function Is Limited to Cytokine Control in the Acute Response to Virus Infection
• Engineered Mammalian RNAi Can Elicit Antiviral Protection that Negates the Requirement for the Interferon Response
• The Evolution of Antiviral Defense Systems
• RNase III nucleases from diverse kingdoms serve as antiviral effectors
• SARS-CoV-Encoded Small RNAs Contribute to Infection-Associated Lung Pathology
• Homologous recombination is an intrinsic defense against antiviral RNA interference
• Archaeal Kink-Turn Binding Protein Mediates Inhibition of Orthomyxovirus Splicing Biology
• ADAR1 Biology Can Hinder Effective Antiviral RNA Interference

Explore the themes

Connected discoveries

Publications in this area