The Evolving World of Small RNAs from RNA Viruses

Abstract

RNA virus infection in plants and invertebrates can produce virus-derived small RNAs. These RNAs share features with host endogenous small interfering RNAs (siRNAs). They can potentially mediate RNA interference (RNAi) and related RNA silencing pathways, resulting in specific antiviral defense. Although most RNA silencing components such as Dicer, Ago2, and RISC are conserved among eukaryotic hosts, whether RNA virus infection in mammals can generate functional small RNAs that act in antiviral defense remains under discussion. Here, we review recent studies on the molecular and biochemical features of viral siRNAs and other virus-derived small RNAs from infected plants, arthropods, nematodes, and vertebrates and discuss the genetic pathways for their biogenesis and their roles in antiviral activity.

INTRODUCTION

RNA-mediated gene silencing is an evolutionarily conserved mechanism in which the presence of intracellular double-stranded (ds) RNA triggers a pathway that leads to the production of small RNAs that mediate regulation of gene expression. Small RNAs, ~ 20–30 nucleotides (nt) in length, are non-coding and regulate gene expression by base pairing with their RNA targets to effect posttranscriptional and/or transcriptional silencing. In the past decade, virus-derived small RNAs have been discovered in plants, arthropods, nematodes, and mammals. These small RNAs share features with host endogenous small interfering RNAs (siRNAs) and can potentially mediate RNA silencing pathways against virus infection.

In infected cells, replication of the virus genome generates dsRNAs. Viral dsRNA molecules are produced in cells infected with diverse types of RNA viruses: (i) dsRNA viruses such as Drosophila X virus (DXV) ¹; (ii) single-stranded RNA viruses that produce dsRNA as the result of the formation of secondary structures, such as Sindbis virus ^{2, 3}, vesicular stomatitis virus ⁴, and enterovirus 71 (EV71) ⁵, and/or dsRNA replication intermediates such as Drosophila C virus (DCV) and Semliki Forest virus (SFV) ^{6, 7}.

Long viral dsRNA molecules trigger the antiviral siRNA pathway ⁸. The dsRNAs are processed by the dsRNA-specific endoribonuclease (RNase) Dicer into ~22 bp dsRNAs. One strand of the RNA duplex is loaded into the RNA-induced silencing complex (RISC). ^{9, 10}. The loaded viral siRNA binds a viral RNA (genome or transcript) by sequence complementarity leading to Ago-dependent degradation of the targeted RNA and restriction of virus replication ^{9, 10}. However, viruses have evolved numerous evasion strategies, including RNAi suppressors, to counteract the host’s antiviral immunity. Recent studies indicate that in addition to siRNA, other related small RNAs, such as microRNAs (miRNAs) and Piwi-interacting RNAs (piRNAs), contribute to antiviral responses as well. This has led to a better understanding of small RNA-based antiviral immunity and its contribution to the control of viral spread and pathogenesis ^11–15. In this review, we summarize our current understanding of the biogenesis of RNA virus-derived small RNAs (vsRNAs) and their roles in antiviral defense in plants, arthropods, nematodes, and vertebrates. We will also discuss the controversy as to whether or not vsRNAs elicit an RNAi-type antiviral response in vertebrates. Finally, we will survey ways in which viruses short circuit the RNAi pathway.

RNA SILENCING PATHWAYS

Synthesis of siRNAs from long viral dsRNA and their mechanisms of action in posttranscriptional gene silencing are well characterized. In Drosophila for example, Dicer-2 cleaves long dsRNAs into a pool of 21-nt siRNA duplexes (Fig. 1A) ¹⁶. Loquacious isoform PD (Loqs-PD), a dsRNA binding protein, and Arsenite resistant protein 2 (Ars2) increase the affinity of Dicer-2 for its substrates ¹⁷. R2D2, another dsRBP, and Armitage, an RNA helicase, participate in siRNA loading with Ago2 and facilitate assembly of RISC ^{18, 19}. C3PO, an endoribonuclease, aids the cleavage of one strand (passenger strand) of the siRNA duplex by Ago2 ²⁰. The remaining strand (guide strand) is subsequently methylated at its 3′ end by Hen-1 (an S-adenosylmethionine methyltransferase) to facilitate the maturation of RISC ²¹. RISC binds its RNA target by base pairing via the guide strand RNA. Ago2-mediated degradation of the target RNA ensues. siRNA synthesis and function is similar in plants (Fig. 1B). However, plants present the ability to produce secondary siRNAs via action of cellular RNA-dependent RNA polymerases (RDRP).

Fig 1. General antiviral RNAi pathways and virus-encoded RNAi suppressors.

RNA virus-encoded RNAi suppressors target steps in the arthropod siRNA pathway (A) and the plant siRNA pathway (B). During viral replication, double-stranded RNAs (dsRNAs) are targeted by Dicer-2 (DCR2) in arthropod cells and Dicer-like protein (DCL) in plant cells and are cleaved into small fragments of dsRNA. The 3′-end of a siRNA is methylated (m) at the indicated steps in arthropod and plant cells. One strand of the small dsRNA is loaded into Argonaute protein (Ago) complexes and combined with other proteins to form the RNA-induced silencing complex (RISC). The guide siRNA within RISC forms base pairs with viral RNA to elicit RNA silencing via RNA degradation. The plant host can amplify siRNAs by the activity of cellular RNA-dependent RNA polymerase (RDRP). Viruses and their respective RNAi suppressor proteins (red ovals) target the indicated steps within the RNAi pathways depicted. For arthropods, these include the following viruses: cricket paralysis virus (CrPV), Culex Y virus (CYV), Drosophila C virus (DCV), dengue virus (DENV), flock house virus (FHV), Nodamura virus (NoV), sfRNA from flavivirus infection, and Wuhan nodavirus (WhNV). For plants, these include the following viruses: cucumber mosaic virus (CMV), polerovirus, sweet potato mild mottle virus (SPMMV), tombusvirus, tospovirus, and turnip crinkle virus (TCV).

ANTIVIRAL FUNCTION OF siRNA

Plants

RNA silencing is a major antiviral mechanism in plants. Several lines of evidence underlie the antiviral role of siRNAs in virus-infected plants. (i) vsRNAs were discovered for the first time in potato virus X (PVX)-infected plants in 1999 ²². Abundant antisense small RNAs complementary to viral RNA were detected in PVX-infected plants. These RNAs were about 25 nt in length and were derived from viral RNA, as their accumulation required virus replication. These antisense small RNAs exhibited antiviral activity as they base pair with the viral RNA to induce posttranscriptional gene silencing and inhibit virus replication. Subsequently, antiviral siRNAs derived from cucumber mosaic virus (CMV), cymbidium ringspot virus, tombusvirus, tobacco rattle virus, turnip crinkle virus, turnip mosaic virus, sugarcane mosaic virus, and Chinese wheat mosaic virus were discovered ^23–30. (ii) Many plant viruses encode suppressors of RNA silencing to overcome the host’s RNAi-based defense system (discussed later). Plants infected by suppressor-deficient viruses rapidly recover from virus-induced disease ^{11, 31–34}. (iii) Plants harboring mutant components of the siRNA pathway are more susceptible to virus infection ^{31, 35}.

Virus-derived siRNAs can also enter the plant RDRP-mediated amplification cycle to enhance the antiviral silencing response. For example, Wang et al. ²³ and Garcia-Ruiz et al. ²⁸ demonstrated that antiviral RNA silencing within Arabidopsis thaliana cells infected with cucumber mosaic virus (CMV) or turnip mosaic virus required the host’s RDRP. RDRP serves to amplify the numbers of antiviral siRNA molecules initially generated during virus replication.

The diversity of Ago proteins in Arabidopsis thaliana, for example, lends flexibility to the siRNA-mediated antiviral response in this organism ^{36, 37}. A. thaliana has ten Ago genes, of which Ago1, Ago2, and Ago7 have been implicated in antiviral defense against various viruses. Ago3 and Ago5 also associate with virus-derived siRNAs and play roles in antiviral responses ^{38, 39}. Genetic analyses of A. thaliana revealed that Ago 1, Ago 2, and Ago10 enhance antiviral defense against turnip mosaic virus in various organs ³⁹.

Arthropod invertebrates

The siRNA/RNAi pathway mediates the antiviral response in invertebrates, as indicated by the effects of mutations within the RNAi pathway on virus replication. Li et al. ⁴⁰ showed that infection of Drosophila S2 cells with flock house virus (FHV) resulted in rapid accumulation of FHV-specific siRNAs of both positive- and negative-strand polarities. Accumulation of the siRNAs decreased the abundance of FHV RNAs at later stages of virus infection. Clearance of FHV genomic and subgenomic RNA from infected cells required Ago2, suggesting an antiviral, RNA silencing role for the siRNAs. Moreover, flies deficient for Dicer or Ago2 were unable to control virus replication, and as a consequence, were hypersensitive to FHV infection. This study provided the first experimental evidence that RNA silencing acts as an antiviral defense mechanism in animal cells.

An abundance of other discoveries followed regarding additional RNAi factors and viruses. For example, Ago2 is essential in order to block two other insect positive-strand RNA viruses, Drosophila C virus (DCV) and cricket paralysis virus (CrPV) ^{41, 42}. Vesicular stomatitis virus (VSV), a negative-sense, single-stranded RNA virus, replicates more efficiently in Dicer, Ago2, or R2D2 mutant flies ⁴³. Dicer protects flies against infection by Sindbis virus (SINV), an enveloped, positive-sense, single-stranded RNA virus ⁴⁴. Additionally, depletion of Ago2, R2D2, VIG, Armitage, Rem 62, or Ars2 in flies increases susceptibility to a diverse set of RNA viruses including Drosophila X virus (DXV), FHV, DCV, SINV, and VSV ^{45, 46}.

Further studies reported discovery of positive-, negative-, and double-stranded vsRNAs and their involvement in antiviral defense in D. melanogaster, mosquitoes, silkworms, and C. elegans ^{11, 14, 35}. An analysis of small RNAs produced during VSV infection of Drosophila cells revealed that VSV-derived siRNAs were approximately evenly distributed between the positive and negative strands; the majority of these siRNAs clustered in a 1.6 kb region at the 5′ end of the genome ⁴. Rift Valley Fever virus (RVFV)-derived siRNAs were present in approximately equal ratios between the positive and negative strands, and were distributed evenly across the genome. Likewise, analyses of vsRNAs from Himetobi P virus (HiPV)-infected small brown planthoppers revealed virus-derived siRNAs 21- and 22-nt in length ⁴⁷. Nearly two-thirds of the siRNAs were derived from the HiPV genomic RNA strand and were evenly distributed across the genome. Silencing Ago2 enhanced HiPV accumulation, providing genetic evidence for siRNA-mediated defenses against HiPV in this insect.

Mosquitoes encode homologues of all the RISC components described in Drosophila. An abundance of evidence indicates that siRNA/RNAi provides a critical antiviral mechanism in mosquitoes. For example, replication of o’nyong-nyong virus (ONNV) in mosquitoes is inhibited upon introduction of dsRNAs that target the viral genome ⁴⁸. This was the first demonstration that siRNA acts as an anti-arboviral response in mosquitoes. Studies from Blair and colleagues demonstrated that replication of Sindbis virus (SINV) and dengue type 2 virus (DENV-2) in cultured mosquito cell lines and adult mosquitoes was inhibited by transient expression of, or cytoplasmic introduction of, a long dsRNA derived from the virus genome. vsRNAs were detected, suggesting that siRNAs target viruses in mosquitoes ^49–52. Aedes aegypti employs RNAi against Sindbis virus infection, and depletion of Ago2, Dicer, or R2D2 in adult mosquitoes increases virus load in cells infected with ONNV or DENV-2 ^{48, 49, 53}.

Expression by viruses of proteins or RNAs that suppress RNAi in mosquitoes provide additional evidence for RNAi-based antiviral immunity. For example, suppressor protein B2 inhibits siRNA function during ONNV and SINV infection ². Recently, van Cleef et al. ⁵⁴ demonstrated that, while the mosquito RNA machinery processes Culex Y virus (CYV) dsRNAs into siRNAs, the CYV VP3 protein promotes suppression of RNA silencing. These observations strongly suggest that mosquitoes use siRNA/RNAi as a major antiviral mechanism against arboviruses. Clearly, viruses that encode siRNA/RNAi suppressors will replicate and disseminate more effectively in the host. (A later section will provide more discussion of virus-encoded RNAi suppressors.)

We should note that the antiviral response in mosquitoes and other insects is a critical factor for viral transmission and dissemination ^{55, 56}. For example, arboviruses in the Flaviviridae, Bunyaviridae, Togaviridae, and Reoviridae families are transmitted to vertebrate hosts via arthropod vectors such as mosquitoes, ticks, and midges. Several mosquito-borne arboviruses, such as DENV and chikungunya viruses, cause worldwide epidemics and high mortality or morbidity rates in humans. The mosquito vectors of arboviruses must control viral infections to minimize pathology in the mosquito, and thereby allow persistent infection and lifelong virus transmission.

The antiviral siRNA response in midges resembles that of other insect vectors of arboviruses. Midges transmit bluetongue virus (BTV) (Reoviridae), Oropouche virus (Bunyaviridae), and Schmallenberg virus (Bunyaviridae) to humans and livestock. Using reporter gene-based assays, Schnettler et al. ⁵⁷ demonstrated functional antiviral siRNAs in BTV-infected Culicoides sonorensis-derived KC cells. These siRNAs inhibited BTV replication. Sequencing of small RNAs from infected cells revealed 21-nt, virus-derived siRNAs, similar to the virus-derived siRNAs produced during arbovirus infections of mosquitoes. In addition, the BTV dsRNA genome is accessible to a Dicer-like nuclease.

Unlike mosquitoes and midges, tick antiviral siRNAs are less understood and present distinct differences. Recently, Schnettler et al. ⁵⁸ reported identification of vsRNAs in Langat virus (flavivirus)-infected tick cells. The vsRNAs appeared to be a key factor in the antiviral response. However, the small RNAs were longer (22 nt) than those from other arbovirus vectors and were mainly derived from the termini of the viral genome. Apparently, the subgenomic RNA of tick-borne flavivirus interferes with the tick’s siRNA-dependent, antiviral immunity. This study highlights an important difference in the antiviral RNAi response between two major classes of arbovirus vectors.

Non-arthropod invertebrates

The siRNA/RNAi pathway also operates in non-insect invertebrates such as Caenorhabditis elegans. Virus-derived siRNAs protect C. elegans against virus infection. C. elegans cells defective for RDE-1 (Ago2 homolog), DRE-4 (R2D2 homolog), or RRF-1 (RDRP, required for siRNA amplification) are more susceptible to VSV infection ^{59, 60}. In addition, C. elegans lacking RDE-1 is more sensitive to Orsay virus (a positive-strand RNA virus, related to nodavirus) infection ⁶¹.

In C. elegans, Dicer partners with the dsRNA binding protein RDE-4 to process virus dsRNA into siRNAs. Full antiviral activity also requires a number of worm-specific genes such as rsd-2 (RNAi spreading defective 2) ⁶². The rsd-2 null mutants permit higher levels of viral RNA accumulation. Ectopic expression of RSD-2 in rsd-2 mutants reverses the effect. RSD-2 functions in a mechanism that is independent of RDE-4, but it requires the RDRP, rrf-1. Thus, it is involved in the amplification of siRNAs for transduction of siRNA signals throughout the organism. As such, antiviral defense in C. elegans depends on the collective actions of RDE-4-dependent and RDE-4-independent pathways to generate siRNAs ⁶³. We also note that the viral small RNA-mediated silencing effects can be transmitted to multiple generations. The virus-derived small RNAs are transmitted in a template-independent manner and silence viral genomes present in worms that are unable to produce their own virus-derived small RNAs. These inherited RNA molecules protect ensuing generations from the virus by silencing the viral genome ⁶⁴.

Deep sequencing of viral small RNAs elucidated the Lepidopteran antiviral response against infection by cricket paralysis virus (CrPV; Reoviridae), a segmented dsRNA virus ⁶⁵. The analysis showed that the silkworm Bombyx mori generates virus-derived siRNAs when infected by Bombyx cytoplasmic polyhedrosis virus (BmCPV). Moreover, in addition to cleavage of viral dsRNAs by Dicer, an uncharacterized RNase cleaves the viral RNA to generate vsRNAs. This RNase activity may be an important early factor for the host’s antiviral defense in Lepidoptera.

Vertebrates

As discussed above, it is well established in plants and invertebrates that for antiviral RNAi, Dicer processes virus-derived dsRNA into siRNAs that guide Ago proteins to silence complementary viral RNA. As a counter measure, viruses encode suppressors that block RNA silencing ³⁵. However, whether mammals restrict virus infections by virus-derived siRNAs and RNAi is still controversial. Cullen et al. ⁶⁶ and Weng et al. ¹² have recently discussed findings that favor or challenge RNAi as an antiviral response in mammals. Below, we update and summarize recent reports on this topic.

1. Evidence against antiviral RNAi in mammals

Although several studies have provided evidence suggesting vsRNAs, including siRNAs, promote an antiviral response in infected mammals, other research groups have challenged the effectiveness of vsRNAs in antiviral defense. One rationale is simply that the abundance of vsRNAs generated in infected mammalian cells may not be sufficient to initiate an antiviral response. Estimates are that at least 100 copies per cell of a small RNA are required for it to be functional ⁶⁷. Parameswaran et al. ⁶⁸ examined vsRNAs from mammalian host cells infected with five vertebrate viruses: hepatitis C virus (HCV), polio, DENV, VSV, and West Nile virus (WNV). In some cases they detected only “vanishingly rare” levels (i.e., < 0.1% of total cellular small RNAs) of certain vsRNAs in infected cells. However, other groups readily detected vsRNAs with Northern blots ^{12, 69–71}. These observations might be explained by a bias in vsRNA measurement by deep sequencing and Northern blotting, such that levels of vsRNA determined by deep sequencing might be underestimated.

Another issue is whether vsRNAs in infected mammalian cells arise from Dicer cleavage or non-specific RNA degradation. Dicer produces siRNAs that are typically 22 ± 2 nt with two unpaired nucleotides at the 3′ termini of either strand, and derive in roughly equal proportions between the plus and minus strand of virus dsRNA ^{68, 72}. By contrast, vsRNAs from mammalian cells typically present a random size-distribution and strong strand bias; this suggests they result from non-specific degradation of viral genome RNA ^{5, 68}. A recent study by Bogerd er al. ⁷² using deep sequencing demonstrated that infection of human cells by DENV or WNV did not result in production of any virus-derived siRNA or miRNA. The reads from DENV-infected cells showed no length preference between 18–29 nt; most of the reads from WNV-infected cells were < 20 nt. Additionally, they used gene editing to generate human cell lines lacking Dicer activity (NoDice) and thus unable to produce small RNAs. Infection of these cells with a wide range of viruses, including DENV, WNV, yellow fever virus, Sindbis virus, Venezuelan equine encephalitis virus, measles virus, influenza A virus, reovirus, VSV, human immunodeficiency virus type 1 (HIV-1), or herpes simplex virus 1 (HSV-1) did not enhance virus replication. This overall uncertainty has led to the speculation that the type I interferon (IFN-I) response is sufficient to suppress virus replication ^{66, 73, 74}. In mammalian cells, the IFN-I response is an antiviral defense that recognizes viral dsRNA and triggers IFN-stimulated gene (ISG) expression to suppress virus replication ^{73, 74}.

Work from Backes et al. ⁷⁵ examined the antiviral contributions of both silencing by small RNAs and the IFN-I system in mammalian cells. Deep sequencing results showed that the two most abundant VSV-derived small RNAs were Dicer independent. To assess the contribution of small RNAs and IFN to the mammalian response to virus infection, they engineered mutant VSV strains that were able to disrupt small RNA-mediated silencing, or to disrupt IFN activity, or to disrupt both activities. They found that a lack of small RNAs did not affect virus titers in vitro or in vivo. By contrast, a virus strain with the capacity to inhibit the IFN system increased virus titers. This study indicated that small RNA-mediated silencing is not a physiological contributor to the IFN-triggered cellular response to RNA virus infection in mammalian cells. Their findings are consistent with an earlier report of Seo et al. ⁷⁶, which showed that IFN signaling stimulated by virus infection inhibited mammalian RNAi by inducing poly-ADP-ribosylation of RISC. These results suggest that in uninfected cells, ISG expression is suppressed by miRNAs. Virus infection leads to inhibition of miRNA activity, thus releasing the suppression of ISG expression. Taken together, these data suggest that mammalian RNAi and miRNAs may not serve as antiviral effectors in mammalian cells.

tenOever and colleagues ⁷⁷ have hypothesized that even though the IFN-I response provides the major antiviral defense in mammalian cells (and chordates in general), the RNAi machinery might nonetheless be sufficiently robust to provide antiviral immunity under certain conditions. To examine their hypothesis, they engineered an RNAi-like response in mammalian cells and assessed its efficacy against infection by Influenza A virus (IAV) in the presence or absence of the IFN-I response. Their strategy was to produce virus strains that would be targeted by species-specific miRNA (acting as a siRNA) or that would be self-targeted by virus-derived small RNA. The major finding from their study was that an RNAi response elicited by the host- or virus-derived small RNAs was highly effective at attenuating IAV in mammalian cells, even in the absence of the IFN-I response. Thus, RNAi can be engineered to act as an efficient antiviral defense in mammals. However, as they also concluded, an evolutionary event may have resulted in the transition from an RNAi-based to an IFN-I–based antiviral system since both systems can function in mammals.

2. Evidence for antiviral RNAi in mammals

By contrast, several studies suggest the existence of vsRNAs with antiviral activity in infected mammalian cells. Mammals have the necessary protein components such as Dicer, Ago2, and RISC to generate viral-derived siRNAs and load them into enzymatically active complexes. As well, even though the study by Parameswaran et. al. ⁶⁸, noted above, found vanishingly rare quantities of some vsRNAs, others were found to be highly abundant, i.e., comparable in levels to abundant miRNAs.

Another issue noted above is whether vsRNAs in infected mammalian cells arise from Dicer cleavage or non-specific RNA degradation. Maillard et al. ⁷⁸ reported that encephalomyocarditis virus (EMCV) or Nodamura virus (NoV) infection of undifferentiated mouse cells led to accumulation of ~22-nt RNAs with all the signature features of siRNAs, i.e., they were derived from viral dsRNA replication intermediates, associated with Ago2, were not produced in Dicer-knockout cells, and decreased in abundance upon cell differentiation. The study by Parameswaran et al. ⁶⁸ also examined several characteristics of vsRNAs that would distinguish them from non-specific RNA degradation products. (i) For HCV, polio, VSV, and WNV infected cells, the positive-to-negative strand ratios among vsRNAs were nearly 1:1 as would be expected for processing by Dicer or a different, dsRNA-specific endoribonuclease. (ii) There was evidence of two unpaired nucleotides at the 3′ termini of either strand within the total pool of vsRNAs. (iii) There was evidence that vsRNAs associated with the four Ago proteins in RISC and only one strand of the duplex was incorporated into these complexes.

Additional evidence consistent with RNAi-mediated antiviral immunity in mammals is that a number of mammalian viruses encode potent viral suppressors of RNA silencing (VSR). For example, the nucleocapsid N protein from mouse hepatitis virus (MHV, Coronaviridae) suppresses RNAi activity in infected cells and this requires its dsRNA-binding activity ⁷⁹. Knockdown of Dicer-1 or Ago2 enhances virus replication in mouse cells, as is expected for an RNA-mediated response. Likewise, genetic ablation of a NoV-encoded VSR, which antagonizes Dicer during infections, inhibits virus replication; viral replication is normal in RNAi-deficient mouse cells ⁷⁸. Finally, infection of hamster cells and suckling mice by NoV requires suppression of RNAi by the NoV B2 protein ⁸⁰. Loss of B2 activity leads to accumulation of viral siRNAs and rapid clearance of mutant viruses in mice. These studies provide compelling evidence for an antiviral RNAi response in mammals.

Recent studies from our groups showed that vsRNAs might also act through mechanisms beyond RNAi. Substantial levels of vsRNAs are generated in enterovirus 71 (EV71)-infected mammalian cells, and vsRNA accumulation is Dicer-dependent ⁵. Overexpression of vsRNA inhibits internal ribosomal entry site (IRES)-dependent virus translation in infected cells. By contrast, blocking vsRNA function with antimers enhances virus yield and virus protein synthesis. At least a subset of vsRNAs blocks viral IRES activity by targeting stem-loop II within the viral IRES. Whether EV71 vsRNAs play an antiviral role in viral infection awaits further investigation. Subsequent studies found that vsRNA can enhance association of host regulatory RNA-binding proteins with the EV71 IRES ⁸¹. The proteins identified so far include AUF1, HuR, KSRP/FBP2, and Ago2. Given the roles of these host proteins in translation and mRNA degradation, we speculate that vsRNAs may control virus translation, and replication, by affecting interaction of these regulatory proteins with the IRES ⁸¹.

In conclusion, one possible contributor to the controversy over an RNAi-mediated antiviral response in mammalian cells is that studies have employed different cell lines, viruses, and VSRs, yielding conflicting results. This suggests that certain cell types may support vsRNA generation and function, whereas others do not. It is perhaps noteworthy that ESCs and possibly other undifferentiated cell types appear to utilize RNAi for antiviral defense, but these are cells that do not mount an IFN-I response ^{66, 78}. This has led to the speculation that RNAi- and interferon-mediated antiviral responses play complementary roles ^{66, 77}. Verifying the specific cell niches that support vsRNA production and identifying potential roles of vsRNAs in mammalian hosts are essential. Clearly, additional work is needed to ascertain whether vsRNAs do significantly contribute to the mammalian antiviral response, particularly somatic cells.

VIRUS-ENCODED RNAi SUPPRESSORS

Viruses are both inducers and targets of gene silencing ^{82, 83}. Virus-induced gene silencing is elicited by dsRNA intermediates of replicating viruses or structured regions of viral RNAs ⁸⁴. The RNA silencing, antiviral defense mechanism is thus a major obstacle towards virus replication. As a counter defense, viruses encode protein suppressors of RNA silencing, or VSRs, that permit virus replication. Discovery of diverse viral proteins as suppressors of RNA silencing provides strong evidence for the effectiveness of vsRNAs for antiviral defense ^{11, 13}.

Diverse RNA viruses encode protein suppressors of RNA silencing. In general, RNA silencing suppressors from different virus families that target the same step in the RNAi pathway do not share sequence identity or structural conservation. It is believed that these proteins have evolved independently in different viruses. Below, we provide a brief survey of virus-encoded suppressors to provide a sampling of the different steps and trans-acting factors within the RNAi pathway that viruses target (Fig. 1).

Suppressors targeting Ago proteins

Some suppressors target the Ago proteins and act via multiple mechanisms. The first suppressor shown to bind an Ago protein in vivo was the 2b protein of cucumber mosaic virus (CMV), a positive-strand RNA virus ⁸⁵. Interaction of the viral 2b protein with Ago1 in CMV-infected A. thaliana plants inhibits Ago1-mediated slicing of the target RNA in vitro. The 2b protein also interacts with long dsRNA and siRNA duplexes in vitro ⁸⁶ and can inhibit host RDRP-dependent amplification of siRNAs ⁸⁷. Polerovirus encodes the F-box–containing P0 protein, which promotes degradation of Ago1 in tobacco plants ^{88, 89}. The binding of P0 to a component of RISC prevents siRNA-RISC assembly. Ago1 molecules not assembled into RISC are readily degraded. This occurs by P0 targeting Ago1 for ubiquitination and subsequent degradation by proteasomes.

Cellular proteins possessing glycine/tryptophan (GW/WG) repeats are important regulators of RNAi through their interactions with Ago proteins. For example, turnip crinkle virus (TCV) encodes the P38 protein, which contains two GW/WG repeats. P38 directly interacts with Ago1 and inhibits Ago1 activity ⁹⁰. Sweet potato mild mottle virus (SPMMV) encodes P1, which contains three GW/WG repeats. P1 binds to Ago1 and inhibits the silencing activity of RISC in two ways: by outcompeting an essential, endogenous GW/WG-containing component of RISC, or preventing recognition of target RNA by complementary siRNA loaded in RISC ⁹¹.

Some VSRs inhibit the slicing activity of Ago proteins. For example, cricket paralysis virus (CrPV) encodes protein 1A, which suppresses Ago2-mediated antiviral defense in Drosophila S2 cells ⁴². 1A directly interacts with Ago2 and inhibits RNA silencing ⁹². Likewise, Nora virus of Drosophila melanogaster (DmelNV) counteracts antiviral RNA silencing activity by encoding a host-specific VP1 protein. VP1 interacts with Ago2 and inhibits its slicer activity ⁹³.

Suppressors targeting dsRNAs and/or siRNAs

VSRs also target the RNA components of RNA silencing. Functionally, these suppressors either bind long viral dsRNA to prevent processing into siRNAs or they bind siRNAs to block their activity. For example, the B2 protein encoded by FHV contains a dsRNA-binding domain. In FHV-infected Drosophila cells, B2 interacts with viral dsRNA intermediates and inhibits processing of long dsRNA into siRNAs ⁹⁴. The B2 protein encoded by NoV, like FHV, a member of the Alphanodavirus genus, inhibits RNA silencing by interacting with viral dsRNA and siRNA duplexes in mammalian cells ⁹⁵. Notably, B2-deficient NoV replicates slower than the wild-type virus in BHK-21 (baby hamster kidney) cells, suggesting B2-dependent resistance to the antiviral vsRNA mechanism in mammalian cells ⁸⁰. The DCV 1A proteins bind and shield long dsRNA from Dicer cleavage ⁹⁶. P19 encoded by tombusviruses promotes systemic infection by selectively binding to siRNA duplexes and preventing the siRNAs from being assembled into RISC ^{97, 98}. HC-Pro encoded by turnip mosaic virus is required to establish infection and systemic spread. HC-Pro acts by sequestering virus-derived siRNAs. HC-Pro may also function through its interaction with transcription factor RAV2, translation initiation factor eIF4E, Hsp90, and/or via downstream effects on silencing factors ³⁹. The mosquito Culex Y virus (CYV) VP3 protein binds dsRNA and siRNA and interferes with Dicer cleavage. This was the first virus-encoded suppressor reported in infected mosquito cells ⁵⁴. The tospovirus nonstructural NS protein sequesters siRNAs and prevents them from being loaded into RISC in infected plants ^{99, 100}. Moreover, as described above, MHV-encoded nucleocapsid N protein suppresses antiviral RNA silencing through its dsRNA-binding activity in mammalian cells ⁷⁹.

RNase-like suppressors

Some RNA viruses that infect plants, insects, fish, or other animals encode class I RNase III-like proteins. Sweet potato chlorotic stunt virus (SPCSV) encodes an RNase III that contains an RNase domain and a dsRNA-binding domain. The viral RNase III cleaves long dsRNA, and 21–24–bp siRNA duplexes, into 14-bp fragments resulting in suppression of RNA silencing in SPCSV-infected plants ¹⁰¹. Suppression requires the catalytic activity of the RNase III. Interestingly, even though most of the viral siRNA suppressor proteins identified so far do not share any conserved amino acid motifs or structural features, the RNase III enzymes are an exception, as they are homologous in unrelated RNA viruses ¹⁰².

Suppressors targeting Dicer

Flaviviruses generate a small, structured, non-coding RNA from the viral 3′UTR, called sfRNA. Recently, sfRNAs of WNV, DENV, Kunjin virus, tick-borne Langat virus, and tick-borne encephalitis virus (TBEV) were shown to suppress RNA silencing in insect and mammalian cells ^{58, 103–105}. sfRNAs co-immunoprecipitate with Dicer and Ago2 in infected cells. sfRNAs compete with Dicer substrates resulting in decreased Dicer activity. Furthermore, DENV non-structural protein 4B (NS4B) inhibits Dicer activity through an unknown mechanism ¹⁰⁶. Interestingly, in addition to binding dsRNAs and siRNAs, the FHV B2 protein can also interact with Dicer and inhibit its activity ¹⁰⁷. The Wuhan nodavirus (WhNV) B2 protein suppresses RNA silencing in Drosophila by interacting with Dicer-2 and sequestering dsRNAs and siRNAs ¹⁰⁸. B2 directly binds to the RNase III and PAZ domains of Dicer-2 and blocks processing of dsRNAs by Dicer-2 and loading of siRNAs into RISC. RNA binding promotes the interaction of B2 and Dicer-2.

CONCLUSIONS AND PERSPECTIVES

RNA silencing is a primary defense mechanism against virus infection in plants, invertebrates, and possibly vertebrates. Viruses have evolved counter-defensive measures through the activity of virus-encoded, silencing-suppressor proteins. Understanding how antiviral silencing is controlled, and how suppressor proteins function, is essential for understanding how virus and host co-evolve, why some viruses are highly virulent in different hosts, and how sustainable antiviral strategies can be developed. Key challenges in the field are to characterize the molecular and biochemical features of vsRNAs, to identify the genetic pathway(s) for their biogenesis, and to determine their antiviral mechanisms. For example, what are the molecular/structural requirements for Dicer to recognize the substrates and initiate the siRNA pathway in infected cells? How do vsRNAs guide Ago2 and by what mechanisms? Do the vsRNAs in Ago-RISC turn over during virus infection? What downstream events are required for the most effective antiviral RNA silencing? As virus-derived small RNA research progresses further, it will surely advance our understanding of RNA silencing as an antiviral response.

Table 1.

siRNAs or virus-derived small RNAs from virus-infected hosts.

Host species	Virus family	Viral genome	Virus (abbr.)	Experimental system	reference
Plant	Alphaflexiviridae	ssRNA(+)	potato virus X (PVX)	in vivo: Nicotiana benthamiana	22
			cymbidium ringspot virus (CymRSV)	In vivo: Nicotiana benthamiana; Nicotiana clevelandii	24, 25
	Bromoviridae	ssRNA(+)	cucumber mosaic virus (CMV)	In vivo: Arabidopsis thaliana	23
	Tombusviridae	ssRNA(+)	turnip crinkle virus (TCV)	In vivo: Arabidopsis thaliana	27
	Virgaviridae	ssRNA(+)	tobacco rattle virus (TRV)	In vivo: Arabidopsis thaliana	26, 27
			Chinese wheat mosaic virus (CWMV)	In vivo: wheat cv. Yannong 15	30
	Potyviridae	ssRNA(+)	turnip mosaic virus (TuMV)	In vivo: Arabidopsis thaliana	28
			sugarcane mosaic virus (SCMV)	In vivo: Maize (Zea mays L.)	29
Arthropod	Nodaviridae	ssRNA(+)	flock house virus (FHV)	Cell: DrosophilaS2	40
	Dicistroviridae	ssRNA(+)	Himetobi P virus (HiPV)	In vivo: small brown planthopper (Laodelphax striatellus)	47
			cricket paralysis virus (CrPV)	Cell: DrosophilaS2 in vivo: Drosophila melanogaster	41, 42
	Togaviridae	ssRNA(+)	o’nyong-nyong virus (ONNV)	In vivo: Anopheles gambiae	48
	Togaviridae	ssRNA(+)	Sindbis virus (SINV)	in vivo: Aedes aegypti	53
			Semliki Forest virus (SFV)	Cell: Aedes albopictus U4.4 and Aedes aegypti Aag2 cells	6
	Flaviviridae	ssRNA(+)	West Nile virus (WNV)	Cell: Aedes albopictus C6/36 cells	71
			dengue virus (DENV)	Cell: Aedes albopictus U4.4	76
			Langat virus (LGTV)	Cell: Ixodes scapularis IDE8 cells and BHK-21 cells	58
	Rhabdoviridae	ssRNA(−)	vesicular stomatitis virus (VSV)	Cell: Drosophila DL1and Aag2 cells	4
	Bunyaviridae	ssRNA(−)	The Rift Valley Fever virus (RVFV)	Cell: Aag2, U4.4 and C6/36 cells	57
	Birnaviridae	dsRNA	Culex Y virus (CYV)	Cell: Culex quinquefasciatus Hsu and C. tarsalis CT cells	58
	Reoviridae	dsRNA	Midges transmit bluetongue virus (BTV)	Cell: Culicoides sonorensis- KC cells	59
Vertebrate	Picornaviridae	ssRNA(+)	Poliovirus	Cell: HeLa and MEFs In vivo: mouse	71
			encephalomyocarditis virus (EMCV)	Cell: mESCs	80
			enterovirus 71 (EV71)	Cell: RD and SF268 cells	5
	Flaviviridae	ssRNA(+)	Dengue (DNEV)	Cell: Huh7, HuH and Vero cells	71, 84
			West Nile viruses (WNV)	Cell: MEFs and cortical neurons In vivo: mouse	71
			Hepatitis C virus	Cell: Huh7.5	71
	Nodaviridae	ssRNA(+)	Nodamura virus (NoV)	Cell: BHK-21 in vivo: Suckling mouse	80, 81
	Rhabdoviridae	ssRNA(−)	Vesicular stomatitis virus (VSV)	Cell: HeLa, BHK-21; MEFs	71