Influenza A virus-generated small RNAs regulate the switch from transcription to replication

Small RNA Deep Sequencing from Influenza A Virus-Infected Cells.

Because many viruses are known to produce small RNAs that aid in aspects of their viral life cycle (31, 32), we speculated that influenza A virus also may generate such products. Given the need for a double-stranded RNA promoter in the context of a single-stranded genome, a small RNA could serve to reconstitute promoter functionality. In contrast, a small RNA could be incorporated into the RdRp to modify its replicase activity, similar to many cellular RNP complexes.

To investigate if influenza A virus produced such small RNAs, we performed deep sequencing on influenza A/PR/8/34 (H1N1) virus-infected lung epithelial cells. Twelve hours postinfection (hpi), total RNA was resolved by polyacrylamide gel electrophoresis (PAGE), and RNA (10–40 nt in length) was isolated and processed using SOLiD (Applied Biosystems) deep sequencing methods (33). Upon analyzing more than 4 million unique reads, we found that more than 90% of the sequences were composed of cellular microRNAs (miRNAs) (34). However, a small percentage of the total captured sequences in infected cells derived directly from influenza A virus (Table S1). Although most of these small RNAs comprise viral breakdown products, which are distributed equally across the viral genome in both polarities, 30% of the influenza A virus sequences captured in this size fraction were significantly enriched for the 5′ end of the vRNA (Fig. 1 and Table S2). Influenza A virus-derived small RNAs (svRNAs) ranged from 22–27 nt in length, were unique to each of the eight segments at positions 14–16 and beyond the 21st base, and terminated three to four bases from the polyU tract (Table S3). Moreover, because these RNA species likely contain a 5′ triphosphate, their inefficient ligation during sample preparation would result in a gross underestimate of the percentage of svRNA in the total RNA captured during sequencing (35). Overall, the predominant species lengths were 25 nt (segments 2, 3, 6, 7, and 8) and 27 nt (segments 1, 4, and 5) (Table S3).

Characterization of svRNA Expression.

To ascertain whether svRNA was produced at significant levels for detection by Northern blot, and to determine the kinetics of svRNA synthesis, we performed time-course studies in infected lung epithelial cells. To monitor total svRNA produced during the course of infection, a pan-specific probe, capable of hybridizing to each of the eight svRNAs, was generated. Total svRNA expression demonstrated accumulation at ≈12 hpi and increased in concentration for the duration of the experiment (Fig. 2A). Although viral RNA breakdown products, which also were reflected in the deep sequencing results, were evident by Northern blot, the predominant accumulation of small RNA products between 22 and 25 nts in length corroborated the abundance of this 5′ small RNA. To ascertain how the kinetics of svRNA synthesis correlated to viral transcription, the nonstructural 1 (NS1) protein levels were determined by Western blot (Fig. 2B). These levels demonstrated that protein expression, visible within 8 hpi, preceded the detection of svRNA. Furthermore, quantitative PCR (qPCR) using a reverse transcription strategy selective for genomic viral RNA demonstrated a dramatic switch to viral replication at 12 hpi, concurrent with the appearance of svRNA (Fig. 2C). To ensure that svRNA was specific to influenza A virus and was not a virus- or IFN-inducible miRNA, we analyzed total RNA from lung epithelial cells treated with vesicular stomatitis virus (VSV) or type I IFN (IFN-I). Northern blot analysis demonstrated the appearance of influenza A virus-specific svRNAs of 22 and 25 nts that were absent in both VSV- and IFN-I–treated samples (Fig. 2D). To ensure that svRNA was not specific to a particular influenza A virus subtype, we isolated total RNA from 10-day-old eggs inoculated with H1N1 (A/PR/8/34), H3N2 (A/Panama/2007/99), and H5N1 (A/Vietnam/1203/04) and analyzed them for the generation of svRNA. Analysis of total RNA demonstrated that svRNA was produced by all three influenza A virus subtypes in the context of an in ovo infection (Fig. 2E); these overall levels correlated with the extent of virus replication (Fig. S1A). Furthermore, H1N1 infection of human, canine, and murine fibroblasts also resulted in the generation of svRNA (Fig. 2F). Collectively, these data indicate that svRNA is an abundant virus-specific RNA species, it is produced by a broad range of influenza A virus subtypes, its production is not cell- or species-specific, and generation corresponds to a shift from viral transcription to replication.

svRNA Production in Trans Requires RdRp Activity, NP, and Nuclear Export Protein NS2.

To investigate how svRNA species are generated, we determined what viral components were required for its production. To this end, bidirectional plasmids representing each of the eight viral segments were transfected into fibroblasts (Fig. S1B). Because each segment-specific plasmid produces an RNA polymerase I-dependent vRNA (3′ to 5′ RNA of negative polarity with a 5′ triphosphate and no polyA tail) and a polymerase II-dependent mRNA (5′ to 3′ RNA of positive polarity with both a 5′ cap and a 3′ polyA) tail, transfection serves both as a template source of vRNA and to express segment-specific proteins (36). To determine whether svRNA could be generated in the absence of virus infection and to ascertain which viral proteins were required for its processing, each of the eight bidirectional segment-specific plasmids was transfected into fibroblasts, and total RNA was analyzed by Northern blot (Fig. 3A). As compared with mock treatment, transfection of each of the influenza A virus segments induced robust expression of svRNA. In contrast, removal of segments 1 (PB2), 2 (PB1), 3 (PA), or 8 [which encodes NS1 as well as the nuclear export protein (NEP)/nonstructural protein 2 (NS2)] resulted in a complete loss of svRNA. Removal of segments 4 and 6, encoding HA and neuraminidase (NA), resulted in no significant changes. Loss of segment 7, encoding the matrix proteins M1 and M2, although demonstrating a decrease in svRNA production, did not have the dramatic loss observed for segments 1–3 and 8. Interestingly, loss of segment 5, which encodes NP, resulted in the loss of svRNA but the production of two fragments of ~32 and ~42 nt. To ensure that the bidirectional plasmids produced each of their respective transcripts in trans, qPCR was performed on NP, M, and PB2 (Fig. S1 C–E). In each case, loss of PB2 and PB1 expression caused a dramatic decrease in NP and M expression levels, indicating that the majority of mRNA from these transcripts was produced from the RdRp rather than directly from the transfected bidirectional plasmid. These data confirm that influenza A virus produces a specific small RNA during infection in an RdRp-dependent manner, suggests a possible structural role for NP, and implicates an unknown requirement for segment 8 vRNA, NS1, and/or NEP/NS2.

To elucidate further the role for segment 8 products in the generation of svRNA, segments 1–7 were coexpressed with plasmids expressing either NS1 or NEP/NS2 mRNA (Fig. 3B). The expression of NS1 failed to rescue svRNA production, but expression of NEP/NS2 demonstrated a partial rescue. Taken together, these data suggest that NEP/NS2 provides an essential component of the viral machinery responsible for the generation of svRNA. Furthermore, given the recent implication of NEP/NS2 in genome replication (37, 38), the total requirement for the RdRp, NP, and NEP/NS2 strongly suggests that a functional polymerase and vRNA template are the minimal components for the production of svRNA.

svRNA Physically Associates with RdRp.

To determine whether any of the known RNA-binding protein components physically associated with svRNA, Flag-tagged GFP and the influenza A virus proteins PB2, PB1, PA, RdRp (PB1/PB2, and PA), NP, and NS1 were cotransfected with a nonstructural (NS) svRNA mimetic containing a 5′ triphosphate (pNS) (Fig. 3C, Figs. S1F and S2). Flag immunoprecipitation, followed by small RNA Northern blot assay, demonstrated that svRNA associated only with the complete RdRp (PB1, PB2, and PA) and not with any individual polymerase component. Furthermore, bound svRNA contained both a 5′ triphosphate and a less dominant species lacking such a motif, as corroborated through phosphatase treatment (Fig. 3D).

svRNA Does Not Induce the Cell's Autonomous Antiviral Defenses.

Because the pattern recognition receptor retinoic acid-inducible gene 1 has been found to associate with 5′ triphosphate-containing RNAs, we evaluated whether svRNAs, which likely contain a 5′ triphosphate, act as pathogen-associated molecular patterns. To determine the cellular response to svRNA, we compared the pNS svRNA mimetic with unphosphorylated NS protein svRNA, scrambled (Scbl) RNA controls, or phosphorylated scrambled (pScbl) (Fig. S2). To ensure that these oligonucleotides were comparable to endogenous svRNA, RNA from infected cell extract was compared with extract from fibroblasts transfected with each of the four RNA oligonucleotides and evaluated with an svRNA-specific probe (Fig. S3A). Northern blot analysis confirmed the delivery of NS and pNS and demonstrated their similarity to endogenous svRNA, migrating at the higher 25-nt position of svRNA instead of the 22-nt predominant svRNA visualized by the pan-specific probe (Fig. S3A). Despite the robust levels achievable by transfection and the presence of an exposed 5′ triphosphate, svRNA failed to induce phosphorylation of IFN regulatory factor 3 (IRF3) (Fig. S3B). In contrast, either transfection of the dsRNA mimetic, polyinosinic:polycytidylic acid (polyIC), or infection with an influenza A virus encoding an NS1 RNA-binding mutation (mFlu) (39) induced strong IRF3 phosphorylation, induction of IFN-β, and the subsequent regulation of known IFN- and virus-stimulated genes (Fig. S3B) (40). These results are consistent with the model that cellular antiviral helicases require an exposed 5′ triphosphate, with additional secondary structure and length requirements (41–43).

Anti-svRNA Induces Loss of vRNA and Inhibits Viral Spread.

Because svRNA does not appear to be a by-product of the cell's antiviral defenses, we investigated whether it was a functional component of the viral life cycle. Because inhibiting mRNA synthesis would be detrimental to overall viral replication, we reasoned that svRNA would bias cRNA/vRNA synthesis only moderately, but its inhibition would result in a dramatic loss of genomic RNA. To analyze the impact of svRNA inhibition on RNA replication, we synthesized a locked nucleic acid (LNA) anti-svRNA, complementary to segment 4 svRNA (anti-HA) (44). We transfected Scbl LNA or anti-HA and performed a de novo infection with influenza A/PR/8/34 (H1N1) virus and analyzed the levels of HA mRNA, cRNA, and vRNA by primer extension (Fig. 4A). These results demonstrated that anti-HA had low impact on both HA mRNA and cRNA but induced a significant reduction in HA vRNA.

To ascertain whether loss of vRNA in the absence of soluble svRNA is observed during a single-cycle infection, we transfected Scbl LNA or anti-HA and monitored protein production directly following the inoculation of influenza A/PR/8/34 virus. Treatment with anti-HA had no impact on NP or NS1 protein levels following transfection but induced a modest decrease in HA protein, likely the result of loss of segment 4 vRNA (Fig. 4B). However, the supernatant from these initial infections demonstrated a dramatic decrease in viral progeny when applied to target Madin-Darby canine kidney (MDCK) cells (Fig. 4C). Because efficient influenza A virus egress requires RNP association (45), the near total loss of HA, NP, and NS1 in the target MDCK cells suggests a defect in packaging, possibly because of the loss of HA vRNA. This possibility is supported further by the fact that HA vRNA packaging sequences are essential components required for viral egress (46).

svRNA Function Is Segment-Specific.

To determine whether blocking svRNA resulted in the inhibition of vRNA synthesis in a segment-specific manner, anti-NA and -NS svRNA LNAs were synthesized and evaluated in the context of a de novo infection by primer extension (Fig. 5A). These results demonstrated that only loss of HA svRNA impacted HA transcripts. Although a modest decrease of mRNA was observed in the presence of anti-HA, a robust loss of vRNA was evident. In contrast, anti-NA or -NS svRNA LNAs had no significant impact on any HA transcripts.

To corroborate svRNA segment specificity further, we determined whether prolonged treatment with anti-HA would result in only a specific loss of HA vRNA and the consequential loss of HA mRNA and protein. In comparison with no treatment or with delivery of Scbl LNA, anti-HA induced a dramatic loss of HA expression but had no significant impact on NP or NS1 protein levels (Fig. 5B). Furthermore, viral growth curves in the presence of either Scbl or HA-specific anti-svRNA corroborated the loss of viral protein production, because titers were reduced by 80% (Fig. 5C). Taken together, the data presented support a model in which svRNA expression impacts the transition from transcriptase to replicase activity and further suggest that it may function in a segment-specific manner.

Cell Culture and Viral Infections.

Cell collections and growth conditions can be found in the SI Materials and Methods.

Deep Sequencing.

Deep sequencing was performed on human alveolar basal epithelial cells (A549) untreated or infected with influenza A/PR/8/34 at a multiplicity of infection (MOI) of 1.0 for 12 h. RNA smaller than 40 nt was isolated from total RNA samples using a FlashPAGE fractionator (Ambion), and SOLiD small RNA libraries were prepared using the SOLiD Small RNA Expression Kit (Ambion). Complete materials and methods are given in SI Materials and Methods.

svRNA Detection.

Detection of svRNA was performed by Northern blot in a manner similar to that previously described for miRNAs (49). Briefly, RNA was resolved on a 15% denaturing polyacrylamide gel containing 7.5 M urea and 20 mM 3-(N-morpholino)propanesulfonic acid-NaOH (pH 7) and transferred to Hybond NX membrane. Chemical crosslinking was performed per the above-mentioned protocol for 1 h at 60 °C, blocked for I h at 65 °C in 6× SSC and 7% SDS, and subsequently probed with radiolabeled oligodeoxyribonucleotides. Methods used in probe design and labeling can be found in SI Materials and Methods .

Plasmid-Based svRNA Production.

HEK293 cells were transfected with bidirectional plasmids encoding each of the eight viral segments of A/PR/8/34 (pDZ1-8, a gift from P. Palese, Mount Sinai School of Medicine, New York). Exogenous expression of Flag-tagged NEP/NS2 and NS1 (pCAGGs Flag-NEP and NS1, gifts from A. García-Sastre and R. Albrecht, Mount Sinai School of Medicine, New York) were performed in a similar manner. Transfections were performed in duplicate for total protein and RNA isolation at 24 h posttransfection.

Synthetic svRNA and Anti-svRNA Oligonucleotides.

Details regarding synthesis, transfections, immunoprecipitations, and treatments can be found in SI Materials and Methods.

Primer Extension, Western Blot, and Quantitative PCR Analysis.

For analysis of HA mRNA, cRNA, and vRNA, primer extension was performed on HEK293 cells that were mock-transfected or transfected with 25 nM Scbl LNA or HA-, NA-, or NS-anti-svRNA. Twelve hours posttransfection, cells were infected with A/PR/8/34 at an MOI of 10 in Opti-MEM, and total RNA was harvested at the indicated times and subjected to RT-PCR using previously published primers (50). The resulting cDNA was resolved by 6% denaturing gel electrophoresis, transferred to Hybond-N nylon membrane (GE Healthcare) at 350 mA for 60 min, UV crosslinked at 200,000 μJ/cm², and viewed by autoradiogram. Details regarding methods for Western blot analysis and qPCR can be found in SI Materials and Methods.