Do the enzymes of the silencing pathway have an antiviral role that does not involve silencing?
The debate over mammalian antiviral RNA interference had been conducted almost entirely in terms of products. Either virus-derived small interfering RNAs of the right size and Dicer dependence could be found in infected cells or they could not. That framing assumes the only way the silencing machinery can be antiviral is by making small RNAs. Two observations from this laboratory's own earlier work suggested otherwise. Shapiro 2012 in RNA had found that Drosha, a nominally nuclear enzyme, redistributes to the cytoplasm after infection with either parental or microRNA-expressing Sindbis virus, so relocalisation responds to infection rather than to a substrate. A nuclear RNase III enzyme found in the cytoplasm during infection raises a question the small interfering RNA framing cannot answer, which is what it is doing there.
Shapiro 2014 separated the two RNase III enzymes genetically. In primary fibroblasts with conditional deletion by Cre-expressing adenoviral vectors, loss of Drosha raised Sindbis virus and vesicular stomatitis virus titres and capsid protein across multicycle growth, while loss of Dicer had no significant effect. Infection with positive-sense, negative-sense and segmented nuclear RNA viruses, and transfection of double-stranded RNA alone, all drove Drosha into the cytoplasm within six hours. Export required CRM1, proceeded without new protein synthesis, and occurred in cells lacking RIG-I, TBK1 or the type I interferon receptor, which is the basis for calling the activity interferon independent. That claim rests on three specific genetic lesions and does not exclude other sensors. Localisation and activity were then coupled by mutation. A Drosha variant with serines 300 and 302 replaced by alanine was constitutively cytoplasmic and restricted Sindbis virus most strongly, by more than two logs against about one log for the wild-type and phosphomimetic forms. The authors infer a virus-inducible phosphatase acting on those residues and do not identify one.
The decisive negative result is in the same paper. Small RNA sequencing recovered more than 675,000 reads across the Sindbis virus genome with no enrichment of the 21 nucleotide species characteristic of virus-derived small interfering RNAs, and loss of Drosha raised total viral small RNA reads without changing their profile or genomic distribution. The same held in Drosophila cells. Purified Drosha cleaved Sindbis genomic RNA in vitro and Drosha shaped the infected host transcriptome, so two candidate activities remained, but neither is canonical RNA interference. The authors describe their mechanistic account as speculation.
Aguado 2017 in Nature resolved the mechanism far enough to make a general claim. Disrupting Drosha in an existing Dicer-deficient human line produced cells with no mature microRNAs at all, which removes the possibility of attributing any phenotype to loss of a particular microRNA. Sindbis virus, Ross River virus and Langat virus all replicated better in those cells while influenza A virus and Sendai virus did not, so enhancement tracked with positive genome polarity, and the phenotype was reproduced in primary conditional Drosha mouse lung fibroblasts. Separation of function came from mutants rather than inference. Among six Drosha variants unable to process primary microRNAs, an RNA-binding mutant that is catalytically inactive and does not associate with DGCR8 still suppressed the virus, while Dicer did not. The activity therefore requires RNA binding and requires neither catalysis, nor DGCR8, nor microRNA processing. SELEX against that mutant enriched RNAs with no conserved sequence that fold into unbranched hairpins, and the same protein bound a hairpin within the first 200 nucleotides of the Sindbis genome. Replicon experiments localised the defect to RNA synthesis, with decay excluded using a temperature-sensitive polymerase and translation excluded in vitro, and a reconstituted minus-strand replicase assay showed polymerase output reduced by nearly half in fractions containing Drosha. The clamp language the authors use for this is their interpretation. No structural or single-molecule evidence for steric occlusion is presented, and the study does not test why encapsidated negative-strand genomes escape.
The broadest result is taxonomic. RNase III proteins from bacteria, archaea, yeast and the urochordate Ciona intestinalis conferred antiviral activity against positive-strand but not negative-strand viruses when expressed in the double knockout cells, which converts a Drosha observation into a statement about a protein domain. Cytoplasmic RNase III activity outside mammals was inferred indirectly, from Drosha-dependent processing of a Sindbis-encoded artificial microRNA in zebrafish embryos and of pri-miR-124 delivered by turnip crinkle virus in Arabidopsis protoplasts, which measures processing of an engineered hairpin as a proxy rather than virus restriction in those hosts. In Drosophila DL1 cells, Drosha knockdown increased Sindbis capsid and Drosophila C virus replication, and small RNA sequencing again showed more viral small RNAs with indistinguishable alignment contours.
The line begins as a byproduct. Engineering viruses to make microRNAs required explaining how a cytoplasmic hairpin is cleaved, and the answer, cytoplasmic Drosha, turned out to be more interesting than the tool that revealed it. Shapiro 2014 established that the enzyme matters for replication and that the effect is not RNA interference. Aguado 2017 established what the enzyme recognises, that binding suffices without cutting, and that the capacity is shared across three domains of life. The trajectory is a narrowing of mechanism accompanied by a widening of scope, and it lands on a position that neither side of the 2013 antiviral RNA interference dispute had occupied. A component of the silencing machinery does contribute to antiviral defence in mammalian somatic cells, and it does so without the small interfering RNA products the dispute was about.
Both papers are bounded to cultured cells. No mammalian animal infection is reported in either, the positive-strand specificity rests on three viruses against two, and most of the mechanism in Aguado 2017 is worked out in one engineered human line with primary mouse fibroblasts as the single independent mammalian confirmation.
Aguado 2017 is also assigned to the evolutionary theme, where its cross-kingdom result supports the argument about ancestral defence designs.
Shapiro 2012 in the noncanonical biogenesis theme supplies the founding observation of cytoplasmic Drosha. The negative small RNA result here is the specific reason the position of Cullen 2013 and Backes 2014, that silencing is not the mammalian antiviral system, can stand alongside a real antiviral role for a silencing enzyme. The cross-kingdom RNase III result connects to tenOever 2016 and to Oishi 2023, where proteins from other domains of life are again used against a vertebrate virus.
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.
Loss of the nuclear RNase III enzyme Drosha, but not of Dicer, increases RNA virus replication in mammalian fibroblasts, and diverse RNA viruses drive Drosha into the cytoplasm by CRM1-dependent export in a manner that does not require new protein synthesis, RIG-I, TBK1 or type I interferon signaling.