Small RNAs tackle large viruses: RNA interference-based antiviral defense against DNA viruses in insects

Introduction

Viruses can infect all living organisms and are a major cause of infectious diseases. Survival of an infected host depends on an effective immune response that limits viral replication and minimizes virus-induced damage.¹ A critical attribute of the innate immune system is its ability to recognize ongoing virus infection and to distinguish non-self molecular patterns from those of the host (self). An effective mechanism to sense virus infection is by the specific recognition of viral nucleic acids based on non-self signatures or on their subcellular localization. Double-stranded RNA (dsRNA) is a strong non-self pattern; it is present at high levels in virus-infected cells, whereas healthy cells normally do not produce detectable amounts of dsRNA.²

In mammals, viral nucleic acids are sensed by innate pattern recognition receptors, such as Toll-like receptors (TLRs), retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs) and nucleotide oligomerization domain (NOD)-like receptors (NLRs). For example, the transmembrane receptors TLR3 and TLR7/8 mediate the recognition of double-stranded (ds) RNA and single-stranded (ss) RNA in endosomal and lysosomal compartments, whereas endolysosomal TLR9 detects virus derived DNA.³ The RLR family members RIG-I and melanoma differentiation-associated gene 5 (MDA5) sense replicating RNA viruses in the cytoplasm based on non-self RNA signatures. MDA5 detects dsRNA replication intermediates that are generated during the viral life cycle of positive strand RNA viruses,^4,5 and the protein is also believed to mediate the recognition of viral genomic RNA of dsRNA viruses.⁶ RIG-I recognizes short dsRNA as well as RNAs with a 5′ triphosphate moiety that is present on RNAs of specific virus families, but absent from highly abundant cellular RNAs, such as mRNAs and tRNAs.⁶ Activation of intracellular signaling cascades by the TLR and RLR receptor families leads to the induction of the transcription factors nuclear factor-kB (NF-kB) and interferon-regulatory factor 3 (IRF3) and/or 7 (IRF7), which induce expression of type I interferon (IFN) and other inflammatory cytokines that in turn establish an antiviral state of the cell.³

DNA viruses are also known to induce production of type I IFNs. Viral dsDNA can be sensed directly by a variety of cytosolic DNA-binding proteins that activate downstream signaling cascades.^7,8 Interestingly, two recent reports show that the cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) synthase enzyme binds viral cytoplasmic DNA and induces the production of the second messenger molecule cGAMP, which subsequently triggers STING (stimulator of interferon genes)-dependent IFN signaling.^9,10 More indirect detection of cytoplasmic DNA may occur via host DNA-dependent RNA polymerase III that transcribes viral cytoplasmic AT-rich dsDNA into 5′ triphosphate containing dsRNA, which is subsequently detected by RIG-I.¹¹

Lacking the adaptive and interferon-based innate responses of vertebrates, insects rely on other mechanisms for antiviral defense. Over the last few years it has become increasingly clear that RNA interference (RNAi) is a major mechanism for the defense against RNA viruses.^12-14 This gene-silencing pathway is primed by dsRNA, thus providing a strong basis for self-nonself discrimination. However, it was not clear how insects defend themselves from infections by DNA viruses. In a recent publication, we demonstrated that RNAi is also involved in the defense against dsDNA virus infection in flies.¹⁵ Here, we discuss our results in light of other recent observations in Drosophila and other invertebrate model systems.

The Antiviral RNA Interference Pathway in Insects

In plants, fungi, nematodes, and arthropods, viral dsRNA triggers an antiviral RNAi response (Fig. 1).^12,13 Several studies in Drosophila melanogaster and mosquitoes demonstrated that viral small RNAs are produced during infections with different classes of RNA viruses.^14,16,17 These virus-derived small interfering RNAs (vsiRNAs) are RNA duplexes of 21 nt with 2-nt 3′ overhangs that are generated by cleavage of viral dsRNA by the RNase III enzyme Dicer-2 (Dcr-2). After incorporation of vsiRNA duplexes into an Argonaute-2 (AGO2) containing RNA-induced silencing complex (RISC), one of the strands (the passenger strand) is excluded from RISC. The guide strand remains associated with the mature RISC and mediates recognition of complementary viral target RNAs. Upon recognition of a fully complementary sequence, AGO2 cleaves the viral target RNA. Thus, the RNAi pathway may exert its antiviral activity at two different levels: by Dicer-mediated cleavage of essential dsRNA replication intermediates or viral dsRNA genomes and by AGO2-mediated cleavage (slicing) of viral genomes and transcripts (Fig. 1).

Fly mutants with defects in the core RNAi genes Dcr-2 and AGO2 are highly susceptible to RNA virus infection.^18-22 The inability of these RNAi-deficient flies to control RNA virus replication is also evident from higher viral RNA copy numbers and higher viral titers in RNAi mutants. The hypersensitivity to virus infection of R2D2 and AGO2 mutants,^19,20,22 which have defects in RISC loading and RISC activity but are fully competent in Dcr-2 activity, implies that slicing of viral target RNAs contributes to the antiviral potential of the RNAi machinery. This notion is further supported by the observation that 2 unrelated RNA viruses encode RNAi antagonists that inhibit the catalytic activity of the RISC complex.^23,24

The Antiviral RNAi Pathway Controls DNA Virus Infection in Drosophila

We used Invertebrate iridescent virus 6 (IIV-6), a member of the Iridoviridae family, as a model to study antiviral immunity against DNA viruses in Drosophila.¹⁵ Although IIV-6 is not a natural pathogen of Drosophila, it is known to experimentally infect a broad range of Dipteran species including Drosophila.^25-27 IIV-6 replicated efficiently in wild-type flies, with a rapid increase in viral titers over the first 6 d after inoculation and stable titers thereafter. Despite the stable high titers, infected flies survived for prolonged periods of time (> 30 d), suggesting that IIV-6 establishes a productive non-lethal infection. To monitor whether the RNAi pathway controls DNA virus infection in vivo, we next analyzed survival of Dcr-2 and AGO2 mutants upon IIV-6 infection. IIV-6 infection dramatically decreased survival of Dcr-2 null mutant flies compared with wild-type controls and mock-infected flies. Similarly, flies that do not encode a functional AGO2 protein were also more susceptible to IIV-6 infection. Dcr-2 mutants seem to be more severely affected than AGO2 mutants. However, the reduced stress-resistance and shorter life span of non-infected Dcr-2 mutants²⁸ complicates the interpretation of this apparent difference. Our results were further supported by a recent study of Kemp et al. who show that Dcr-2, R2D2 and AGO2 null mutants die more rapidly upon IIV-6 challenge than wild-type controls.²⁹ Moreover, Kemp et al. observed that the increased lethality in RNAi mutant flies correlates well with an increase in viral load at 10 d post infection.²⁹ We, however, only observed mild differences in viral titers in AGO2 and Dcr-2 mutant flies early in infection, whereas viral load was not affected at a later time point (12 d post infection). The reason for this difference between the 2 studies remains unclear.

Dcr-2 Dependent vsiRNA are Produced from Overlapping Transcripts

Having determined that the RNAi pathway controls DNA virus infection in vivo, we next analyzed whether vsiRNAs are produced during IIV-6 infection. Using small RNA cloning and next-generation sequencing technology, we readily detected viral small RNAs in wild-type and AGO2 mutant flies, the majority of which were 21 nt in size. A strong decrease in normalized levels of 21-nt viral small RNAs in Dcr-2 mutant flies indicates that these RNAs are indeed Dcr-2 dependent vsiRNAs. Using a similar approach, Kemp and colleagues also demonstrated Dcr-2-dependent vsiRNA biogenesis in IIV-6 infection.²⁹

Viral dsRNA replication intermediates or structured elements in single-stranded viral RNAs are potential targets for Dcr-2 dependent vsiRNA production. Several studies reported similar ratios of positive strand (+) over negative strand (-) vsiRNAs in invertebrate hosts infected with dsRNA viruses or positive-strand RNA viruses.^30,31 In addition, (+) and (-) vsiRNAs are present in approximately equal ratios in flies infected with the (-) RNA virus vesicular stomatitis virus (VSV).^22,32 Moreover, viral siRNAs from these different classes of RNA viruses generally cover the entire viral genome. This is further exemplified by the finding that vsiRNAs can be used to deduce up to full-length genomes of known and novel (+) RNA and dsRNA viruses.^24,30,33 Together, these results imply that viral replication intermediates or viral dsRNA genomes serve as the Dcr-2 substrates for vsiRNA biogenesis in RNA virus infection.^13,31

Nevertheless, for several RNA viruses structured RNA elements seem to be preferentially processed by Dcr-2, indicating that vsiRNA biogenesis may be more complex. For example, in small RNA profiles of 2 (+) RNA viruses, Drosophila C virus (DCV) and Flock House virus (FHV), there is an over-representation of (+) strand vsiRNAs.^32,33 This bias reflects the higher abundance of viral genomic (+) RNA strands over antigenomic (-) RNA strands that is generally observed in (+) RNA virus infection. It was thus suggested that structured elements in viral genomic RNA become the predominant target for Dcr-2 when dsRNA replication intermediates are shielded from Dcr-2 cleavage by a viral RNAi suppressor.^32,34 This hypothesis was supported by the observation that vsiRNA reads are heavily biased toward the (+) strand in wild-type FHV infections, whereas similar numbers of (+) and (-) vsiRNAs were derived from infections with RNAi suppressor-deficient virus.^34,35 DCV encodes an RNAi antagonist that binds dsRNA²⁰ and may similarly skew vsiRNAs toward (+) viral strands.³²

DNA viruses do not replicate via a dsRNA replication intermediate and other dsRNA sources must therefore be processed by Dcr-2 for vsiRNA biogenesis. In mammalian cells, dsRNA can be detected upon dsDNA virus infection,² which is probably generated by base pairing of convergent overlapping transcripts of both strands of the viral genome or derives from secondary structures in viral transcripts.³⁶

To identify the dsRNA substrate for vsiRNA biogenesis during IIV-6 replication, we analyzed the distribution of vsiRNAs across the IIV-6 genome. Viral open reading frames (ORFs) in IIV-6 are oriented in both directions in the viral genome. Nevertheless, irrespective of the orientation of the viral ORFs, vsiRNAs mapped in similar proportions to the upper and lower strand of the viral genome (Fig. 2A). These data suggest that vsiRNAs are derived from overlapping sense and anti-sense transcripts. Indeed, sense and anti-sense transcripts that cover highly targeted ORFs (i.e., ORFs with high density of vsiRNA reads) could readily be detected by strand-specific RT-PCR assays and northern blot analyses. These bidirectional transcripts have the potential to base pair and form long dsRNA molecules that trigger the antiviral RNAi machinery via Dcr-2 dependent vsiRNA biogenesis (Fig. 3A).

Uneven Distribution of vsiRNAs Across the IIV-6 Genome

IIV-6 derived vsiRNAs display a highly uneven distribution along the viral genome, with some regions in the genome producing high numbers of vsiRNAs (hotspots) and other regions only producing few vsiRNAs. The uneven coverage of vsiRNAs across the genome was very similar between wild-type and AGO2 mutant flies and between 2 independent data sets from our study. Strikingly, the distribution of vsiRNAs along the viral genome in our study is highly similar to that of Kemp et al. (Fig. 2A and B). To further substantiate this observation, we divided the IIV-6 genome in non-overlapping 500-bp bins and counted the number of vsiRNAs mapping to individual bins in both data sets. Indeed, we observed a strong correlation between vsiRNA mapping between the 2 studies (Fig. 2C, r = 0.894, p < 0.001). Nevertheless, when zooming in on highly targeted regions, it becomes apparent that vsiRNA profiles are not fully identical in the 2 studies, with some highly targeted sequences from one study being markedly less abundant in the other study (Fig. 2B). It has been reported that vsiRNA cloning procedures may induce cloning biases. Such biases are introduced during the ligation steps that are used to link the small RNAs to 3′ end 5′ adaptor sequences, with small RNA sequences that more stably anneal to the adapters producing higher read numbers.³⁷ As the 2 studies used different adaptor sequences, slight differences in vsiRNA profiles, such as observed in Figure 2B, may be due to cloning/sequencing biases.

How do the major vsiRNA hotspot regions in the viral genome arise? Since the 2 independent studies used different sequencing platforms and adaptor sequences but display highly similar profiles, we deem it unlikely that cloning biases explain the observed IIV-6 vsiRNA hotspots. Another putative source for biases could be differential stability of RISC-loaded small RNAs, as siRNAs that possess a cytidine at their 5′ termini seem to be preferentially loaded into AGO2.^38,39 IIV-6 derived siRNA profiles are highly similar between wild-type and AGO2 mutant flies that lack a functional RISC.¹⁵ Thus, the observed uneven distribution of vsiRNA profiles cannot be explained by differences in stability within RISC.

Viral siRNA hotspots often map to regions in the genome that correspond to predicted ORFs (Fig. 2B). It is likely that vsiRNA hotspots derive from regions in the genome in which there is high bidirectional transcriptional activity. We thus propose a model in which highly abundant transcripts base pair with low-level antisense transcripts to generate dsRNA substrates for Dcr-2. Alternatively, there may be particularly high level of antisense transcription in hotspot regions. At this point, the transcriptional landscape of IIV-6 remains poorly defined; thus, the temporal regulation of transcription, transcription start and stop sites, as well as the extent of antisense transcription throughout the viral genome remains unclear. Future studies using strand-specific next generation cDNA sequencing (RNA-seq) may provide a high-resolution map of transcriptional activity across the IIV-6 genome. Once these data become available, our model can be more robustly tested.

DNA Virus-Derived vsiRNA Profiles in Other Model Systems

A number of recent publications have now analyzed vsiRNA profiles in DNA virus infections in different invertebrate model systems. These studies further underscore that DNA viruses are targets of an antiviral RNAi response, but that the mechanism for vsiRNA biogenesis may differ from that in IIV-6 infected flies. In a recent study, Dcr-2 dependent vsiRNAs were recovered from a Drosophila cell line infected with Vaccinia virus (VACV), a dsDNA virus of the poxvirus family.³² Abundant vsiRNAs were particularly recovered from tandem repeats located at the covalently closed genomic termini. Computational predictions indicated that transcripts of these repetitive sequences have the ability to fold into hairpin RNA structures (Fig. 3B). Indeed, knockdown experiments indicated that these RNA structures are targets of Dcr-2, and to a lesser extent of the microRNA processing enzyme Dcr-1. Base paired bidirectional transcripts or structured regions of single-stranded transcripts were proposed to be Dcr-2 substrates in other regions of the VACV genome, but these hypotheses remain to be experimentally tested. Moreover, Drosophila cells do not support a full VACV replication cycle, and the full extent of vsiRNA production in poxvirus infection remains to be established.

Structured RNA elements were also implied as Dcr-2 substrates in Culex tritaeniorhynchus densovirus, a novel, single-stranded DNA virus of the densovirus genus that was recently identified in wild-caught mosquitoes.⁴⁰ Two inverted repeat sequences at the non-coding genomic termini, which were previously not known to be transcribed, gave rise to abundant viral small RNAs. This observation suggests that these termini are transcribed and fold into dsRNA structures that are processed into small RNAs. Indeed, in silico predictions suggest that putative transcripts of these regions have the potential to fold into Y-shaped RNA structures with extensive perfectly base paired stems of 35 and 68 bp for the 5′ terminal inverted repeat 1 (IR1, Figure 3C) and the 3′ terminal inverted repeat 2 (IR2, not shown), respectively. To explore whether these structures are putative Dicer substrates, we analyzed the published small RNA sequences to determine the size profiles of the viral small RNAs. 74.3% and 79.8% of the small RNAs that map to IR1 and IR2, respectively, were 21 or 22 nt in size. Moreover, these small RNAs mapped to the stem, but not to the 2 arms of the predicted Y-shaped RNA structures (data not shown). These results suggest that IR-derived small RNAs are bona fide Dicer products, but this remains to be experimentally demonstrated. Densoviral terminal stem-loop structures likely play a role in genome replication and/or packaging.⁴¹ It remains to be investigated how transcription of the inverted repeats and Dicer-mediated processing of the resulting transcripts affect these processes. Another source of viral small RNAs in this study were the protein coding transcripts of the viral genome.⁴⁰ These small RNAs predominantly derive from the sense transcripts of the viral genome and have a broad distribution of sizes, ranging from 18 to 30 nt, with only a minor enrichment at 21 or 22 nt (32.3%, data not shown). The majority of these small RNAs are therefore unlikely to be generated in a Dicer-dependent manner.

Like our observations in IIV-6 infected flies, convergent transcripts also seem to the source of vsiRNAs in other insect species infected with complex dsDNA viruses. In a recent study, the antiviral RNAi response to the baculovirus Helicoverpa armigera single nucleopolyhedrovirus (HaSNPV) was analyzed in larvae of the cotton bollworm moth (Helicoverpa armigera).⁴² Similar to vsiRNAs in IIV-6 infected flies, small RNAs displayed an uneven distribution across the viral genome, with some hotspot regions generating abundant amounts of small RNAs, and other regions generating few small RNAs. Moreover, in most hotspots, vsiRNAs map to both strands of the viral genome, consistent with base paired overlapping transcripts as a Dicer substrate. For a number of highly targeted ORFs, small RNAs were enriched for those derived from only 1 strand of the viral genome. The substrates for the biogenesis of these small RNAs remain unclear. When Dcr-2 was depleted in Helicoverpa zea derived fat body cells, higher expression levels were detected for 2 highly targeted ORFS, but not for 2 ORFs that did not give rise to vsiRNAs. Moreover, a modest increase in viral DNA replication was observed upon Dcr-2 knockdown. It was thus proposed that RNAi-mediated silencing of essential viral genes restricts virus replication.⁴²

The antiviral potential of RNAi to control DNA virus infection was recently also confirmed in another invertebrate system. Knockdown of Dcr-2 expression in white spot syndrome virus (WSSV)-infected shrimp resulted in an increase in virus replication, suggesting that this circular dsDNA virus is a target of the invertebrate RNAi machinery.⁴³ Unfortunately, the authors only monitored the production of a single viral siRNA by northern blot. This particular siRNA was indeed produced in a Dcr-2 dependent manner. Future studies using small RNA deep sequencing may provide more detailed insights into the Dcr-2 substrates in this virus and into the mechanism of vsiRNA biogenesis.

Seminal studies have established that RNA viruses are processed into vsiRNAs in plants.^12,44 Interestingly, DNA viruses are also targets of the RNAi machinery in plants. Like DNA viruses in insects, vsiRNA may arise from processing of structured RNA elements⁴⁵ or from dsRNA generated by base-pairing of sense and antisense transcripts.^46,47 Indeed, vsiRNAs of both sense and antisense polarity are generated upon infection with Gemini viruses (ssDNA virus) and the pararetrovirus Cauliflower mosaic virus (circular dsDNA virus).^46-49

Concluding Remarks

Together, these studies demonstrate that an antiviral RNAi response is mounted against distinct DNA viruses in different invertebrate hosts. Our study implies that base paired, overlapping sense and antisense transcripts are the major Dcr-2 substrates in IIV-6 infection; in other viruses (VACV and densovirus) structured RNA elements may be additional substrates for Dcr-2. Important questions remain. For example, are vsiRNAs incorporated in RISC and functional in directing target RNA cleavage? Using sensor assays, we demonstrated that some IIV-6-derived vsiRNAs are indeed functional, but that the efficiency of silencing is modest. Moreover, for 5 selected ORFs that contained a high density of vsiRNA reads, we did not observe a general increase in viral transcript levels in AGO2 mutant flies. Further genome-wide transcriptome analyses are required to establish if and to what extent RNAi regulates IIV-6 viral gene expression. Strikingly, in HaSNPV infection an increase in transcript levels upon Dcr-2 knockdown was observed, suggesting that vsiRNAs are RISC-associated and functional in silencing viral gene expression.⁴²

Other questions pertain to the function of vsiRNAs in DNA virus infection. Is dsRNA produced as an inevitable by-product of a densely packed genome that is transcribed in 2 directions? Does the RNAi machinery exploit this feature for antiviral defense? Alternatively, the virus may exploit the antiviral RNAi machinery to regulate expression of viral genes. RNAi-mediated regulation of expression of essential viral genes may contribute to the fine-tuning of viral replication and pathogenesis. In this respect, it would be of interest to see whether antisense transcription (and thus the potential for dsRNA formation) is dynamically regulated. For some ORFs of HaSNPV, sense transcripts seem to be the major origin of small RNAs.⁴² These small RNAs by definition cannot target coding transcript as they have the same orientation. Perhaps these highly abundant transcripts are decoy Dcr-2 substrates that saturate the RNAi machinery, thereby preventing other essential viral genes from being targeted. Dedicated non-coding transcripts may have a similar function in large complex DNA viruses. Finally, as was reported for the plant DNA virus cauliflower mosaic virus,⁴⁵ viral small RNAs may regulate the expression of cellular genes, in functional analogy to virus-encoded miRNAs.⁵⁰ A striking similarity between VACV and Culex tritaeniorhynchus densovirus is that both viruses produce small RNAs from putative RNA structures encoded by non-coding genomic termini. How these small RNAs affect the viral life cycle is an open question for future investigation.

In several organisms, small RNAs mediate gene silencing at the transcriptional level. For example, small RNAs guide RNA-mediated DNA methylation in plants (Arabidopsis thaliana) or induce heterochromatin formation by guiding the deposition of repressive marks on histone proteins in yeast (S. pombe) and in D. melanogaster).^51,52 During nuclear replication of geminiviruses in plants, viral dsDNA intermediates associate with cellular histone proteins, thus forming viral minichromosomes. Interestingly, it was proposed that viral small RNAs guide de novo cytosine methylation of the geminivirus genome as an antiviral defense strategy.⁵³ DNA viruses in mammals also form chromatin-like structures that are subject to epigenetic regulation.^54,55 Whether vsiRNAs guide transcriptional silencing of DNA viruses in insects remains an open question.

A final important question is whether there are alternative mechanisms for antiviral defense against DNA viruses. Several cytoplasmic DNA sensors have been identified in mammals over the last few years. While insects lack the interferon response of mammals that are induced by these sensors, direct cytoplasmic DNA sensing could also play a role in activating immune or stress responses in insects.

Altogether, recent studies reveal that DNA viruses produce bona fide Dcr-2 substrates that elicit antiviral RNAi responses in invertebrates; nevertheless, the identification of several atypical Dcr-2 substrates suggests that the interaction between DNA viruses and the host RNAi machinery may be highly complex.

Bronkhorst AW, van Cleef KW, Vodovar N, Ince IA, Blanc H, Vlak JM, Saleh MC, van Rij RP. The DNA virus Invertebrate iridescent virus 6 is a target of the Drosophila RNAi machinery. Proc Natl Acad Sci U S A. 2012;109:E3604–13. doi: 10.1073/pnas.1207213109.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.