Drosha as an interferon-independent antiviral factor

Significance

Virus infections must be combated at a cellular level. The strategies used to inhibit virus differ dramatically when comparing plants and insects to mammals. Here, we identify an evolutionary conserved antiviral response that is independent of these known defenses. We demonstrate that an RNA nuclease called Drosha is repurposed during virus infection to cleave viral RNA and modulate the cellular environment as a means of inhibiting virus replication.

Abstract

Utilization of antiviral small interfering RNAs is thought to be largely restricted to plants, nematodes, and arthropods. In an effort to determine whether a physiological interplay exists between the host small RNA machinery and the cellular response to virus infection in mammals, we evaluated antiviral activity in the presence and absence of Dicer or Drosha, the RNase III nucleases responsible for generating small RNAs. Although loss of Dicer did not compromise the cellular response to virus infection, Drosha deletion resulted in a significant increase in virus levels. Here, we demonstrate that diverse RNA viruses trigger exportin 1 (XPO1/CRM1)-dependent Drosha translocation into the cytoplasm in a manner independent of de novo protein synthesis or the canonical type I IFN system. Additionally, increased virus infection in the absence of Drosha was not due to a loss of viral small RNAs but, instead, correlated with cleavage of viral genomic RNA and modulation of the host transcriptome. Taken together, we propose that Drosha represents a unique and conserved arm of the cellular defenses used to combat virus infection.

In plants, nematodes, and arthropods, a major response to virus infection is Dicer-dependent generation of virus-derived small interfering RNAs (vsiRNAs) (1, 2). vsiRNAs associate with the RNA-induced silencing complex (RISC) and mediate cleavage of homologous viral RNA, attenuating virus replication in a process termed antiviral RNA interference (RNAi) (3, 4). Although many components of antiviral RNAi are conserved in chordates, the small RNA-mediated response to virus infection has largely been replaced with the protein-based type I IFN (IFN-I) response although evidence for mammalian RNAi has recently been reported in some cell types against particular viruses (5–7). Indeed, ectopic expression of siRNAs directed against viral genomes in diverse cell types potently inhibits virus replication of a wide range of viruses (8–15). However, vsiRNAs have been difficult to detect in IFN-I–sensitive cells (2, 16, 17). These data suggest that, whereas chordates may not produce robust levels of vsiRNAs, they are capable of harnessing the small RNA machinery in an antiviral capacity when presented with the proper substrate. These same data also suggest that mammalian RNA viruses have not incurred any clear selective pressure to inhibit small RNA-mediated signaling, in contrast to the IFN-I induction pathway where virus antagonism is common (18).

In mammals, RNA virus infection is recognized in response to replication, as this process generates a diverse array of pathogen associated molecular patterns (PAMPs). PAMPs include double stranded RNA (dsRNA), RNA with an exposed 5′ triphosphate, or RNA lacking a 2′ O-methyl–containing cap (18). In the vast majority of cells, PAMPs are detected by one of the two PAMP recognition receptors (PRRs): RIG-I (Encoded by the Ddx58 gene) and MDA5 (18). PRR detection culminates in a signal transduction event that includes activation of the IFN regulatory factors (IRFs) by tank-binding kinase 1 (TBK1) (19). Kinase activation results in assembly of a multisubunit enhancer that promotes transcription of the IFN beta gene, a member of the IFN-I family. IFN-I production subsequently results in the up-regulation of hundreds of IFN-stimulated genes (ISGs) through a ubiquitous IFN-I receptor (encoded by the Ifnar1 gene) (18).

Despite the lack of robust vsiRNA production, chordates have retained genome-encoded microRNAs (miRNAs). These noncoding RNAs are transcribed by RNA polymerase II and processed in a stepwise fashion by two RNase III enzymes: first, Drosha in the nucleus; and second, Dicer in the cytoplasm (20–26). Similar to vsiRNAs, miRNAs are also capable of exerting RNAi although they more commonly act to fine-tune host gene expression through translational repression and/or mRNA deadenylation and are thought to contribute to cellular fitness (27–32). Given the modest repression of miRNAs on their targets, a property that results from imperfect binding complementarity, they are unlikely to serve as direct inhibitors of viral transcripts (33). However, viruses can be engineered to encode perfect complementary target sites for endogenous miRNAs as an effective mechanism to attenuate virus replication (34–41).

Despite the apparent evolutionary loss of vsiRNAs as an antiviral defense in chordates, there are many overlaps between the RNAi and IFN-I pathways, most notable being that both IFN-I and RNAi can be triggered by the presence of dsRNA (42, 43). Furthermore, a number of proteins involved in miRNA production have also been implicated in the IFN-I response. For example, the dsRNA-binding proteins TRBP and PACT, which aid in precursor-miRNA dicing, RISC maturation, and target silencing, have also been reported to inhibit and activate effectors of the IFN-I pathway, respectively (44, 45). In addition, both the ubiquitous and IFN-I–inducible isoform of ADAR1 can function to alter miRNA expression (46) and associate with Dicer to enhance enzyme activity (47). Conversely, many viruses interact with Drosha and Dicer for the production of viral miRNAs or to regulate the levels of viral transcripts (48–51). The range in interplay between virus and the mammalian miRNA pathway demonstrates the capacity for cross-talk between these two systems, but the physiological relevance of this cross-talk remains poorly understood. Supporting data for direct RNAi against viral RNAs in mammalian cells includes evidence for RNase III-like activity in the restriction of retrotransposons and two RNA virus infection models (5, 6, 52). Given these findings and associations, we sought to determine whether Dicer or Drosha, the only mammalian RNase III nucleases, contributed to the mammalian response to virus infection in somatic cells, which are the major targets of viral infection.

Results

Drosha Translocation Is a General Response to RNA Virus Infections.

Recent evidence has demonstrated the capacity to engineer cytoplasmic viruses to produce miRNAs (53–57). Subsequently, we found that cytoplasmic miRNA synthesis was dependent on a Drosha translocation event to process the miRNA from Sindbis virus (SINV) (58). Given the recent findings relating to the ability of the miRNA machinery to naturally exert an antiviral response in mammalian fibroblasts (6), we sought to investigate whether the SINV-induced translocation of Drosha into the cytoplasm represents a broad antiviral response. Therefore, we investigated Drosha localization in response to infection with a positive sense virus (SINV), a negative sense virus [vesicular stomatitis virus (VSV)], and a nuclear, segmented RNA virus [mutated influenza A virus (mIAV)], which lacks its main antagonist of the antiviral response [nonstructural protein (NS1), described in ref. 59], and in response to treatment with the canonical viral PAMP, dsRNA (Fig. 1A). Interestingly, we found that, despite exclusive expression of Drosha in the nucleus in mock-treated cells, there was robust translocation to the cytoplasm in response to SINV, VSV, mIAV, or dsRNA (Fig. 1A). Furthermore, cytoplasmic Drosha (cDrosha) was evident during the early hours of infection and dsRNA treatment (Fig. 1A). These data suggest that detection of a broad range of viral PAMPs results in the accumulation of Drosha in the cytoplasm.

Fig. 1.

Broad accumulation of cDrosha in virus-infected cells. (A) Immunohistochemistry of murine fibroblasts infected with SINV (MOI = 3), VSV (MOI = 1), or mIAV lacking the main viral IFN antagonist nonstructural protein NS1 (MOI = 5), or transfected with poly(I:C) for 6 h and stained for nuclei, viral proteins, or Drosha. (B) Small RNA Northern blot of Drosophila cells (DL1) mock-treated or infected (MOI = 1) with SINV WT or with SINV124 for the indicated times. (C) Small RNA Northern blot of DL1 cells treated with dsRNA against β-galactosidase (bgal) or Drosha and subsequently infected with SINV124.

Drosha-Dependent Cytoplasmic miRNA Processing Is Conserved in Arthropods.

Given the generality of virus-induced Drosha translocation, we next assessed whether insects also display cDrosha activity by assaying miRNA production from a Drosha-dependent, cytoplasmic RNA transcript. Drosophila melanogaster cells are permissive hosts of many alphaviruses and, as in mammalian cells, support a cytoplasmic SINV replication cycle (60). As such, Drosophila cells (DL1) were infected with a recombinant SINV encoding primary (pri)-miR-124 (SINV124) (53), and miR-124 synthesis from the cytoplasmic transcript was monitored. Similar to mammalian infections, SINV124 resulted in miRNA biogenesis throughout the course of the Drosophila cell infection (Fig. 1B). Next, to determine whether this activity was also Drosha-dependent, we depleted cells of this nuclease and assessed cytoplasmic (c)-pri-miR-124 processing (Fig. 1C and Fig. S1). Drosha depletion abrogated the ability of SINV124 to produce a mature miRNA and resulted in an enhancement in the level of unprocessed virus-derived pri-miR-124 and viral RNA (Fig. 1C and Fig. S1). Taken together, these data implicate a possible role for Drosha in the cleavage of the cytoplasmic virus-derived RNA transcripts in Drosophila and suggest increased virus replication in the absence of the nuclease.

Sindbis Virus Is Susceptible to an RNAi-Mediated Antiviral Response.

Given the broad accumulation of cDrosha in response to a panel of RNA viruses, we hypothesized that Drosha may play a role in an RNAi-like response perhaps related to the recent findings in mammalian cells (5–7). We hypothesized that, if Drosha had a role in antiviral RNAi, then both Drosha and Dicer would have antiviral properties where products from Drosha would be fed to Dicer, which would then produce substrates for RISC that would silence viral RNAs. To determine whether these two RNase III nucleases, Drosha and Dicer, contribute to the cellular response to virus infection, we measured the impact of virus replication in the presence or absence of each nuclease. We chose to further study SINV as it is a virus model that has previously been shown to be capable of generating miRNAs, suggesting that it does not disrupt the host machinery responsible for small RNA biogenesis (53). However, to first ensure that the virus lacked a suppressor of RNA silencing (SRS), we assessed whether small RNAs could be harnessed to inhibit SINV replication by engineering the virus with a scrambled RNA (scbl) in its 3′ UTR or two or four miR-124 target sites in the same location (2 × 124T, or 4 × 124T, respectively). Because miR-124 is restricted to neurons, infection of SINV scbl, 2 × 124T, or 4 × 124T in fibroblasts resulted in equal levels of SINV replication as measured by capsid protein synthesis (Fig. 2A). In contrast, exogenous expression of miR-124 resulted in a complete loss of capsid expression in 2 × 124T or 4 × 124T virus while having no impact on SINV scbl or on host protein disulfide isomerase (PDI) (Fig. 2A). Taken together, these results suggest that SINV is capable of being targeted by RNAi during infection.

Fig. 2.

Drosha restricts virus replication. (A) 293T cells transfected with empty or miR-124 producing vector for 36 h and subsequently infected with SINV expressing a scrambled sequence (scbl) or two or four miR-124 target sites (2 × 124T or 4 × 124T, respectively) in the 3′ UTR at an MOI of 1 for 24 h. (Top and Middle) Western blot for SINV capsid or PDI. (Bottom) Small RNA Northern blot of ectopically expressed miR-124. (B) Small RNA Northern blot of Rnasen^f/f or Dicer1^f/f fibroblasts treated with replication-incompetent Adeno-based vectors (AdV) expressing GFP or a GFP-Cre fusion protein (Cre) for 5 d and probed for miR-93 (Upper) or U6 (Lower). (C) Plaque assay of Rnasen^f/f fibroblasts treated with AdV-GFP or AdV-Cre and subsequently infected with recombinant SINV (MOI = 0.1). (D) Western blot for same conditions as in C. (E) Plaque assay of Dicer1^f/f fibroblasts treated with AdV-GFP or AdV-Cre and subsequently infected with recombinant SINV (MOI = 0.1). (F) Western blot for same conditions as in E. Data in C and E are represented as the mean ± SEM for n = 3. *Significant P value of <0.05, using a two-tailed, unpaired Student’s t test.

Loss of Drosha Results in an Increase in RNA Virus Replication.

Given the lack of a SINV SRS, we next used conditional knockout fibroblasts for Drosha or Dicer (Rnasen^f/f and Dicer1^f/f, respectively) and disrupted each gene using replication-incompetent Adeno-based vectors (AdV) expressing GFP or a GFP-Cre fusion protein (Fig. 2B). Cells were incubated for 5 d to allow for efficient clearance of both the vector and targeted host protein. Small RNA Northern blot analysis confirmed loss of endogenous, Drosha- and Dicer-dependent miR-93, demonstrating loss-of-functional enzymatic activity of both genes (Fig. 2B). In contrast, U6, a small nuclear RNA that does not depend upon Drosha or Dicer, was not impacted by these treatments (Fig. 2B). Drosha-depleted cells were subsequently infected with SINV at a multiplicity of infection (MOI) of 0.1 for a multicycle growth curve and compared against control cells. Interestingly, SINV titers reached significantly higher levels in the absence of Drosha throughout the course of infection (Fig. 2C). SINV capsid protein also accumulated to higher quantities in Drosha-depleted cells, consistent with elevated levels of virus replication (Fig. 2D). In contrast, depletion of Dicer did not result in a significant alteration in the SINV titers over the course of infection compared with control cells (Fig. 2E). Western blot also revealed unaltered virus levels between Dicer-deficient and control cells (Fig. 2F). Furthermore, VSV also displayed enhanced replication in the absence of Drosha but not Dicer (Fig. S2 A–D). These data demonstrate that Drosha restricts RNA virus replication whereas loss of Dicer (and consequently miRNAs) does not significantly contribute to the mammalian response to virus infection, at least in the context of primary fibroblasts.

Virus Infection Results in Nuclear Export of Drosha.

We next sought to define the mechanism responsible for the accumulation of Drosha in the cytoplasm upon SINV infection because cDrosha may account for the antiviral activity. To discern whether cDrosha was the product of active nuclear export or was the result of newly synthesized protein retention, we investigated the requirement for the nuclear export protein CRM1 in cytoplasmic processing of primary miRNAs (pri-miRNAs). Subsequent to RNAi-mediated depletion of CRM1, cells were infected with SINV124, and subcellular fractionation was performed (Fig. 3 A and B). Infection with SINV124 resulted in accumulation of Drosha in the cytoplasm in a CRM1-dependent fashion (Fig. 3B and Fig. S3A). Excitingly, CRM1-depleted cells were no longer capable of supporting miRNA biogenesis from the cytoplasmic SINV-derived pri-miRNA whereas endogenous miR-93 was not impacted (Fig. 3C). These data demonstrate that RNA virus infection results in the active transport of Drosha from the nucleus into the cytoplasm in a CRM1-dependent fashion.

Fig. 3.

Host requirements for Drosha export. (A) Western blot of 293Ts at 48 hpt with 50 nM control (ctrl) pool of siRNA or siRNA pool directed against Crm1. (B) Western blot of subcellular fractionation of conditions in A additionally mock-treated or infected with SINV124 (MOI = 3, 8 hpi). (C) Small RNA Northern blot of conditions in A additionally mock-treated or infected with SINV124 (MOI = 3, 16 hpi). (D) MEFs mock-treated or treated with CHX for 2 h and subsequently mock-treated or infected with VSV124 (MOI = 1, 8 hpi).

Next, we determined whether de novo translation is required for the accumulation of cDrosha. To this end, we inhibited translation with cyclohexamide (CHX) and subsequently infected with VSV encoding miRNA-124 (VSV124), which has also been demonstrated to generate a functional miRNA from the cytoplasm (55). VSV124, rather than SINV124, was chosen for this assay as SINV requires host translation to generate virus transcripts. In contrast, VSV packages its own RNA-dependent RNA polymerase, and, therefore, CHX treatment does not inhibit pri-miR-124 synthesis although viral protein production is lost (61). We found that the production of VSV-derived miR-124 was not dependent on translation as mature miR-124 accumulated to higher levels compared with mock-treated cells infected with VSV124, despite a complete loss of VSV G protein (Fig. 3D). Taken together, these data demonstrate that RNA virus infection results in the active transport of Drosha from the nucleus into the cytoplasm in a CRM1-dependent fashion.

Given our observation that Drosha relocalizes upon infection with diverse viruses and dsRNAs, which are canonical PAMPs that are sensed by RIG-I, we set out to determine whether the relocalization of Drosha was dependent on RIG-I or classical IFN signaling. We chose to analyze SINV124 miRNA processing in the presence and absence of RIG-I as cytoplasmic pri-miRNA (c-pri-miRNA) processing is dependent on cDrosha (Fig. 3C). Surprisingly miR-124 accumulated to WT levels in the absence of RIG-I (Ddx58^−/−), demonstrating that this dsRNA sensor is not required for cDrosha activity (Fig. S3B). Moreover, cDrosha activity was retained in the absence of TBK1 (Tbk1^−/−) and IFN-I signaling (Ifnar1^−/−) (Fig. S3 C and D). These data suggest that the relocalization of Drosha represents a unique, IFN-I independent arm of the cellular antiviral response.

Cytoplasmic Translocation Is Necessary for Drosha’s Antiviral Activity.

It has previously been reported that translocation of Drosha into the nucleus is dependent upon phosphorylation of serines 300 and 302 (62). To investigate whether these same residues play a role in virus-induced cDrosha, we used GFP-tagged Drosha constructs expressing Alanines (A) or phosphomimetic residues (Aspartate, D, or Glutamate, E) at these positions. Consistent with previous reports, transfection of GFP-Drosha (WT), S300/302A (2A), or S300E/S302D (ED) demonstrated that these residues are critical components in defining the cellular localization of Drosha: WT and ED were nuclear whereas 2A was predominantly cytoplasmic (Fig. 4A). Thirty-six hours posttransfection (hpt) of the Drosha constructs, SINV infections were performed at an MOI of five, and GFP_NLS and GFP-Drosha constructs were visualized 4 h postinfection (hpi). Although SINV infection had no impact on the cellular localization of the control protein, GFP_NLS, infection resulted in cytoplasmic accumulation of GFP-Drosha-WT (Fig. 4A), as observed with endogenous Drosha (Fig. 1A) (58). Furthermore, virus infection did not impact the constitutive cytoplasmic localization of the 2A mutant. In contrast, GFP-Drosha-ED failed to translocate to the cytoplasm in response to virus infection (Fig. 4A). These results suggest that serine 300 and 302 must be dephosphorylated in virus-infected cells to allow for the cytoplasmic accumulation of Drosha.

Fig. 4.

Phosphorylation requirements for antiviral Drosha activity. (A) Immunohistochemistry of 293T cells transfected with the indicated plasmids and subsequently mock-treated or infected with SINV (MOI = 5, 4 hpi) at 36 hpt. Cells were stained for SINV capsid, and GFP was imaged. (B) Plaque assay of supernatants from BHK cells, cotransfected with the indicated plasmids and with SINV genomic RNA (gRNA) for 24 h. Data are represented as the mean ± SD for n = 3. *Significant P value of <0.05, using a two-tailed, unpaired Student’s t test; ns, nonsignificant P value of >0.05. (C) Western blot for the same conditions as B.

Next, we determined whether cytoplasmic localization was required for Drosha’s antiviral activity. We expressed GFP or the GFP-Drosha constructs (wt, 2A, or ED) and transfected infectious SINV genomic RNA (gRNA) (Fig. 4B). A plaque assay was used to assess virus replication. These studies found that the overexpression of WT or ED forms of Drosha led to a one log attenuation of infection whereas Drosha 2A potently inhibited SINV replication by more than two logs compared with GFP (Fig. 4B). These results were also observed at the protein level as measured by immunoblot (Fig. 4C). These data suggest that Drosha’s antiviral activity is potently exerted in the cytoplasm and that Drosha levels are limiting because ectopic expression is restrictive. Furthermore, these data suggest that there is a virus-inducible phosphatase that drives the relocalization of Drosha during virus infection and that this relocalization is required for Drosha to elicit its full antiviral activity.

Drosha-mediated cleavage of viral RNA.

To parse out Drosha’s nuclear and cytoplasmic mechanisms of antiviral activity, we investigated whether cDrosha was capable of processing the cytoplasmic SINV genome to generate 21-nt vsiRNAs recently described in fibroblasts (5–7). To this end, we performed an in vitro cleavage assay on SINV gRNA derived from SINV124 as previously described (63). Purified Flag-tagged GFP or Drosha was used to perform in vitro cleavage assays that were subsequently analyzed by small RNA Northern blot. As a control, we assessed Drosha’s ability to cleave the pri-miR-124 embedded in the SINV gRNA. Not surprisingly, we found that Drosha is able to specifically cleave the viral gRNA to produce an ∼55-nt preliminary (pre)-miR-124 whereas purified GFP did not process this miRNA (Fig. 5A). We next probed for small RNAs specific to SINV gRNA using a probe that bound to a sequence in the ORF of the nonstructural polyprotein of SINV. Indeed, we found that Drosha mediated cleavage of the viral gRNA to yield an ∼55-nt product (Fig. 5A). Furthermore, we observed an increase in RNAs of ∼20–25 nt. These data suggest that Drosha is capable of cleaving the viral gRNA, likely at secondary structures, which may then result in RNA degradation because the 5′ and 3′ would not be protected via a cap or poly(A).

Fig. 5.

Drosha impacts viral and host RNAs during infection. (A) In vitro cleavage reaction of SINV124 genome incubated with purified Flag-tagged GFP or Drosha and probed for miR-124 (124, Left) or SINV gRNA (SINV, Right). (B) Small RNA deep sequencing of Rnasen^f/f fibroblasts treated with AdV-GFP or AdV-Cre for 5 d and subsequently infected with SINV (MOI = 3, 8 hpi). (C) Small RNA deep sequencing of Drosophila DL1 cells treated with indicated dsRNA for 3 d and subsequently infected with SINV (MOI = 20, 96 hpi). (D) Heat map of RNA-seq reads for Rnasen^f/f fibroblasts treated with AdV-GFP or AdV-Cre for 5 d and subsequently mock-treated or infected (SINV, MOI = 3, 8 hpi). Data represent the relative fold gene expression of AdV-Cre–treated, SINV-infected cells over AdV-GFP–treated, SINV-infected cells and is graphed as fold change over mock-treated cells. Only mRNA transcripts with a greater than fivefold increase or decrease are represented.

Given the ability of Drosha to cleave structures embedded in the viral gRNA (Fig. 5A), we investigated whether Drosha impacted the small RNA profile during virus infection. Small RNA deep sequencing of SINV-infected fibroblasts captured over 675,000 reads with coverage over the complete SINV genome (Fig. S4A and Dataset S1). However, in contrast to what was reported for a mutant Nodavirus (5, 6), we found no enrichment in 21-nt RNAs characteristic of vsiRNAs (Fig. S4B). This is in agreement with other recent studies performed in fibroblasts (64). Lack of canonical vsiRNA production prompted us to investigate the possibility that virus stem loops are targets for the dsRNA nuclease activity of Drosha in its capacity to limit virus replication in the cytoplasm. To this end, we cloned small RNAs (19-25 nt) from wild-type and Drosha-deleted cells infected with SINV (Fig. 5B and Dataset S2). Although loss of Drosha resulted in a significant enhancement in overall small RNA reads, no change in the profile or genomic positioning of RNAs could be detected. Furthermore, we analyzed the small RNA profiles in Drosophila cells infected with SINV in the presence or absence of Drosha. Again, we observed that Drosha depletion increased the number of virus-specific reads without changing the overall profile or genomic locations (Fig. 5C and Dataset S3). This observation is in agreement with independent published fly data that found that the loss of Drosha resulted in elevated levels of vaccinia virus, Drosophila C virus, and VSV RNA (2). These data suggest that cDrosha does not mediate its antiviral activity through generation of vsiRNAs.

Drosha modulates the transcriptome during virus infection.

In addition to cleavage of viral RNA, we additionally assessed the Drosha-dependent changes in gene expression upon virus infection. We performed RNA-Seq analysis, which demonstrated that Drosha influenced the mRNA profile of cells in the absence of infection in agreement with recent publications (65–68) (Dataset S4). Moreover, RNA-Seq revealed greater than 25 transcripts that were highly induced (greater than fivefold) in response to virus infection in the absence of Drosha (Fig. 5D, group 1). In addition to these up-regulated transcripts, a comparable number of mRNAs were down-regulated in virus-infected cells lacking Drosha (group 2) whereas others remained unchanged (a small subset of which are depicted as group 3) (Fig. 5D). We also identified more than 25 noncoding RNAs (ncRNAs) of unknown function, which were significantly up-regulated in virus-infected, Drosha-deficient cells (Fig. S4C). We verified that Drosha was lost upon transduction (Fig. S4D) and that Itaga2 was induced by infection and dependent on Drosha whereas Hspa1a was induced and repressed by Drosha using independent quantitative real-timePCR (Fig. S4 E and F). Taken together, these data demonstrate that Drosha modulates the host transcriptome during virus infection. Whether this modulation occurs in the nucleus or cytoplasm is unknown. Although future work will be needed to fully characterize the functional network of Drosha targets in infected cells, these exciting findings suggest a novel interplay between a miRNA biogenesis factor and the restriction of virus infection.

Discussion

Following the recent evidence for RNAi in mammals and the discovery that RNA virus infection causes a dramatic change in the subcellular localization of Drosha, we set out to determine the range, requirements, and physiological function of this activity. We began our studies by screening a panel of diverse viruses and PAMPs, which we found all capable of inducing the relocalization of Drosha to the cytoplasm (Fig. 1). We next investigated the impact of the small RNA biogenesis factors, Drosha and Dicer, on SINV and VSV replication. Although both encoding an RNA genome, their mechanisms of replication are very different. Multicycle growth curves in conditional knockout cells identified Drosha as a restriction factor in virus replication, independent of Dicer and miRNAs (Fig. 2). In an attempt to elucidate the mechanism for Drosha’s antiviral activity, we studied the biology of cDrosha. Depletion of the nuclear export protein CRM1 resulted in a loss of Drosha-dependent processing of virus-derived pri-miRNAs (Fig. 3). This activity was unaffected by the disruption of Ddx58, Tbk1, or Ifnar1 (Fig. S3). Furthermore, we implicated serine 300 and 302 in the virus-inducible translocation of Drosha and showed that cytoplasmic localization is required to confer the full antiviral activity of Drosha (Fig. 4). Finally, we investigated whether Drosha was capable of generating the small RNAs recently described in fibroblasts (64). In vitro analyses in conjunction with deep sequencing suggested that Drosha is capable of cleaving viral RNA (Fig. 5). Small RNA data from both mammalian and insect virus infections showed that loss of Drosha did not impact the small RNA profile of specific viral RNAs, but, rather, loss of Drosha enhanced the presence of viral RNA, presumably as a result of increased replication (Fig. 5). In addition, mRNA seq data from infected cells with or without Drosha demonstrated Drosha-dependent changes in the host transcriptome that, in concert with the cleavage of viral RNA, likely contribute to its antiviral property.

Recent data suggest that mammalian cells retain some aspect of the antiviral RNAi response (5, 6). We indeed identified small RNA fragments from in vitro Drosha cleavage assays, which may have correlates with vsiRNAs in fibroblasts. However, the size distribution of these small RNAs was not representative of canonical siRNAs. Alternatively, the previously described mammalian Dicer-dependent siRNAs may be produced downstream of Drosha processing. Because we found no role for Dicer in our assays, the role for Drosha in antiviral defense described here is distinct from previous observations. Indeed, we speculate that the antiviral activity imposed by Drosha is rather a consequence of (i) the nuclease’s altered cellular localization, (ii) cleavage of novel substrates that include stem loops in the virus genome, and (iii) cleavage of substrates in host RNAs. Given the lack of a phenotype in Dicer-deficient cells, and that our Drosha cleavage products are not the obvious substrates for Dicer, we do not suggest that these cleavage products are funneled into the canonical miRNA pathway to directly target viral RNAs. One outstanding question is how Drosha recognizes its target in the cytoplasm. It is possible that the nuclease differentiates between host and viral RNA based on adenine/uracil content. Conserved sequence motifs have recently been identified in pri-miRNAs that are believed to identify hairpins as Drosha targets and potentially stabilize the reaction (69). It would be interesting to determine whether these cis-acting elements required for Drosha processing are avoided in viral genomes as a mechanism to circumvent Drosha-mediated attenuation.

In summary, this research ascribes a previously unidentified function for Drosha in the restriction of RNA virus replication and suggests that the enzyme acts independent of its canonical role in miRNA biogenesis during virus infection. Taken together, we speculate that Drosha orchestrates an antiviral response that is conserved from arthropods to chordates and is aimed at disrupting both viral and host RNA profiles to attenuate virus replication.

Materials and Methods

Cell Culture, Transfections, and Viruses.

For cell culture, transfections, and viruses, see SI Materials and Methods.

Drosha in Vitro Cleavage Assay.

Drosha in vitro assays were performed as previously described (63).

Protein Analysis and Statistics.

For protein analysis and statistics, see SI Materials and Methods.

Small RNA Northern Blots.

Small RNA Northern blots and probe labeling were performed as previously described (70). Probes used are provided in Dataset S5. All data are representative of at least three independent experiments.

Small RNA and mRNA Deep Sequencing.

For small RNA and mRNA deep sequencing, see SI Materials and Methods.