The best known insect viruses with double-stranded DNA genome are baculoviruses (Baculoviridae) of which ~600 members are described that mainly infect Lepidoptera, but also Hymenoptera and Diptera (Hemiou et al., 2005). The prototype baculovirus species is Autographa californica multiple nucleopolyhedrovirus (AcMNPV), of which the infection cycle has been studied in great detail, including transcriptome analysis (Chen et al., 2013; Chen Y.-R. et al., 2014). Despite being a DNA virus, however, a large number of viral siRNAs (vsRNAs), with Dicer-processing signature, can be identified during AcMNPV infection (Mehrabadi et al., 2015). Further evidence, that RNAi is involved in the innate immune response against AcMNPV infection, was the identification of a VSR gene, p35, in its genome. Interestingly, p35 was first identified as an inhibitor of apoptosis, but its anti-apoptotic activity was not linked to its capacity to inhibit RNAi (Mehrabadi et al., 2015). A similar occurrence of vsRNAs located to hot spots in the genome was also observed during Helicoverpa armigera single NPV (HaSNPV) infections (Jayachandran et al., 2012). Silencing of Dicer-2 (siRNA pathway), but not Dicer-1 (miRNA pathway) resulted in higher transcript accumulation and higher viral replication levels. On the other hand, an RNAi strategy was successfully employed to inhibit Bombyx mori NPV (BmNPV) replication in transgenic silkworm cells (Kanginakudru et al., 2007; Subbaiah et al., 2013; Zhang P. et al., 2014), indicating that high levels of dsRNAs targeting different essential viral genes could overcome the counter defense mechanisms of the virus.

While vsRNAs are generated during baculovirus infection, presumably after annealing of overlapping complementary transcripts (Jayachandran et al., 2012), several other studies have indicated a role for miRNAs to regulate the baculoviral infection cycle (Kolliopoulou and Swevers, 2014 and references therein). Several miRNA precursors are predicted in the BmNPV genome, of which bmnpv-miR-1 targets the 3′-UTR of the mRNA of the Ran GTP-binding protein, a regulator of the nucleocytoplasmic transport of small RNAs (Singh et al., 2010, 2012). This results in depression of the cellular host miRNA population, including bmo-miR-8, which has multiple binding sites on the mRNA of the viral gene ie-1, a master regulator of the viral infection cycle (Singh et al., 2012). While this can be viewed as a clear defense mechanism, baculoviruses can also escape the immune response in more subtle ways by auto-regulating gene expression during the infection cycle. Another miRNA produced by BmNPV, bmnpv-miR-3, was shown to negatively auto-regulate expression of viral late genes to decrease the viral load and the detection by the immune response (Singh et al., 2014). Down-regulation of ODV-E25 mRNA expression by AcMNPV-miR-1 is proposed to accelerate the formation of occlusion-derived virions during the infection cycle (Zhu et al., 2013). Also host cell miRNAs can play a role in the regulation of the baculoviral infection cycle or antiviral defense. During AcMNPV infection, most miRNAs from the host are down-regulated (Mehrabadi et al., 2013), although exceptions exist, such as bantam, which is upregulated and negatively regulates viral gene expression and replication (Shi et al., 2016). Finally, it was shown that the miRNA pathway could be harnessed to control BmNPV infection using expression constructs of artificial miRNA precursors (Zhang J. et al., 2014).

Similar interactions with the RNAi machinery were reported for other viruses with large DNA genomes such as ascoviruses and nudiviruses. The infection cycle of Heliothis virescens ascovirus (HvAV) is regulated by both cellular and viral miRNAs that target the viral DNA polymerase (Hussain et al., 2008; Hussain and Asgari, 2010). HvAV also encodes an RNase III gene that functions as an inhibitor of gene silencing (Hussain et al., 2010). The miRNA pathway was also shown to be involved in the establishment of latent infections of Heliothis zea nudivirus 1 (HzNV-1) (Wu et al., 2011). Another virus with a large DNA genome is the Invertebrate iridescent virus 6 (IIV-6; Iridoviridae) which has a broad host range. IIV-6 is a target of the RNAi machinery in Drosophila (Bronkhorst et al., 2012) and produces a dsRNA-binding protein that acts as a VSR during infection (Bronkhorst et al., 2014).

Another group of DNA viruses studied in some detail in insects consists of the densonucleosis viruses, or densoviruses (DNVs). Densoviruses belong to the subfamily Densovirinae within the family Parvoviridae and are characterized by a linear, single-stranded DNA genome of 4–6 kb that is packaged in icosahedral, non-enveloped particles (Bergoin and Tijssen, 2000). Small RNAs corresponding to densoviruses were reported in small RNA deep sequencing data of mosquitoes, but it is unclear whether these are generated by a canonical RNAi pathway (Ma et al., 2011). In DNA viruses, complementary transcripts can anneal to form dsRNA (Bronkhorst et al., 2012), but there do not seem to be overlapping complementary transcripts in densoviruses (Carlson et al., 2006).

The plant-infecting Geminiviridae have single-stranded circular DNA genomes, but have been shown to be susceptible to RNAi in transgenic plants, at least for the genus Begomovirus (Bonfim et al., 2007). Begomoviruses are transmitted by whiteflies in a persistent, circulative manner although it is not clear whether they replicate in insect tissues (Rosen et al., 2015). Reverse genetics systems for begomoviruses are available in plants (Tuttle et al., 2008), thus offering possibilities of engineering for dsRNA delivery in insects, analogous to the engineering of plant RNA virus vectors (see also on Section Plant Viruses).

Viral siRNAs are also produced during infection of Drosophila cells with vaccinia virus, a mammalian dsDNA virus that replicates in the cytoplasm (Poxviridae) (Sabin et al., 2013). The most abundant siRNAs are produced from long, structured RNAs derived from the terminal repeats. Interestingly, vaccinia virus infection in Drosophila cells results in poly-adenylation and degradation of cellular miRNAs by the viral poly(A) polymerase (Backes et al., 2012). Degradation of miRNAs during infection with entomopoxvirus is also observed in lepidopteran cells (Backes et al., 2012).

Insects can also be infected with a wide range of RNA viruses that possess positive-strand single-stranded RNA [(+)ssRNA], negative-strand ssRNA [(−)ssRNA] or double-stranded RNA (dsRNA) genomes. Moreover, the genomes can consist of one or multiple segments (Karlikow et al., 2014; Xu and Cherry, 2014). Many plant RNA viruses rely on insects (mainly Hemiptera) for transmission (Fereres and Raccah, 2015) while mosquitoes are the most important vectors for transmission of arthropod-borne viruses (arboviruses) that cause devastating diseases in humans and livestock (Weaver and Reisen, 2010). Both plant viruses and arboviruses can be highly pathogenic to their respective plant and vertebrate hosts. By contrast, arboviruses are known to establish persistent, non-pathogenic infections in the insect vectors to ensure maximal transmissibility (Umbach and Cullen, 2009). For plant viruses, the situation is more complex since plant viruses can be transmitted in both a non-persistent and persistent manner by insects (Dietzgen et al., 2016; explained in more detail in Section Plant Viruses). While plant viruses that replicate in insect hosts are relatively rare, they receive considerable attention in this review just because of this property: they can function as silencing vectors in insects during persistent infections. On the other hand, insect-specific viruses, with only insects as hosts, are often pathogenic although recent deep sequencing projects established the occurrence of many persistent virus infections in insects, without any apparent harmful effects on host physiology (Vasilakis and Tesh, 2015; Cory, 2016). Because of this distinction, insect-specific RNA viruses, plant viruses and arboviruses will be discussed separately.

Viruses are well-known for causing disease (pathogenic infections) and the occurrence of persistent viral infections only recently has caught attention after the accumulation of data from deep sequencing projects. As indicated already above, arbovirus infections in mosquitoes and ticks, and some plant virus infections in hemipteran vectors can be considered as persistent infections because the viruses became adapted to the insect host physiology to ensure efficient transmission to the respective vertebrate and plant hosts. However, deep sequencing projects have revealed the presence of an enormous multitude of viral sequences in insects (Li C.-X. et al., 2015; Liu et al., 2015; Webster et al., 2015), that suggests that persistent viral infections are very common.

Such persistent infections can arise from pathogenic infections that were not cleared effectively and to which insect hosts became adapted (Goic and Saleh, 2012; Moreno-Garcia et al., 2014). Persistent infections can be beneficial to the hosts since they can provide protection against pathogenic infections by related viruses (Villarreal, 2009). Both horizontal and vertical transmissions can occur (Cory, 2016) and different viruses can interact during co-infections, resulting in shifts in infections in insect populations (Cornman et al., 2013). Deep sequencing data therefore have revealed the complexities of virus communities in natural populations of which the ecology is barely or not understood.

An interesting question is how the status of persistence is established and maintained. In both the Drosophila S2 cell line and mosquitoes, virus-derived DNA molecules are produced through reverse transcriptase activity of endogenous retrotransposons in the infected cells (Goic et al., 2013, 2016). The (episomic or genome-integrated) viral DNA forms subsequently produce transcripts that are processed by the RNAi machinery, resulting in control of virus replication. In several other instances, sequences of non-retro RNA viruses were found inserted into insect genomes (Crochu et al., 2004; Maori et al., 2007; Cui and Holmes, 2012; Cheng et al., 2014) and it was suggested that genomic integration could provide a defense mechanism using the RNAi mechanism, for instance through the production of piRNAs as is observed for the control of retrotransposons (Brennecke et al., 2007).

During arbovirus and plant virus infections of insects, viral siRNAs are produced as in the case of pathogenic virus infections. As a correlation exists between viral siRNA production and viral replication levels (Lan et al., 2015; Zografidis et al., 2015), it can be assumed that lower levels of siRNAs are produced during persistent infection than during pathogenic infection. Interestingly, not all viral siRNAs, which accumulate during persistent infection, appear to be functional. In S2 cells persistently infected with FHV, viral siRNAs are not associated with Ago-2-containing RNA-induced silencing complexes (RISCs) and are incapable of silencing complementary reporter constructs (Flynt et al., 2009). Silencing of RNAi components revealed also a more important role for Dcr-2 than Ago-2, indicating that Dcr-2 controls viral replication directly by its nuclease activity rather than as initiator of RNAi. Similar data were obtained in lepidopteran cell lines, where infections of FHV and Macula-like virus (derived from plant viruses of the Tymoviridae family; Katsuma et al., 2005) did not interfere with plasmid-based-expression of their capsid proteins (Swevers et al., 2016). An RNA hairpin construct against the VSR gene B2 of FHV could clear the viral infection efficiently, confirming an intact antiviral RNAi mechanism in the lepidopteran cells. Persistent FHV genomes of lepidopteran cell lines encode an intact VSR gene that can inhibit RNAi potently in a plant-based assay (Swevers et al., 2014), indicating that protection against the host RNAi machinery remains essential. Because the VSR protein B2 interacts with replication complexes of FHV in mitochondrial membranes (Aliyari et al., 2008), it likely acts locally during persistent infections and does not affect the RNAi machinery in the rest of the cytoplasm. Also a cell line from the gypsy moth, Lymantria dispar (Lepidoptera), was found to be persistently infected with iflavirus (Carrillo-Tripp et al., 2014). Transfection of an RNAi inhibitor in the cell line resulted in higher virus genome levels, suggesting an involvement of RNAi. Persistent infection of cell lines with RNA viruses seems to be a rather common phenomenon and therefore could function as a model to study the mechanism of RNA virus persistence (Li et al., 2007; Ma et al., 2014; Carrillo-Tripp et al., 2016; Swevers et al., 2016).

Finally, it can be noted that in some instances no activation of RNAi was observed during persistent RNA virus infections (Carpenter et al., 2009; Habayeb et al., 2009). In those cases, persistent virus infections are controlled by other innate immune pathways (e.g. Shelly et al., 2009; Nakamoto et al., 2012; Paradkar et al., 2012; Cao et al., 2016; review by Merkling and van Rij, 2013). Thus, control of virus infection is not always associated with RNAi mechanisms. Because of the general focus on the importance of the RNAi response to control viral infections, the contribution of alternative pathways may be underestimated from the literature.

Productive virus infections in insects are considered to be highly host-specific. This has been documented most extensively with baculoviruses that are used in integrated pest management as specific biological pesticides. However, the blockage of viral infection can occur at different levels of the infection process and virions may actually invade the host without generation of viral progeny (Haas-Stapleton et al., 2003; Simon et al., 2004; Wu et al., 2016). In addition, there are viruses with a wide host range, such as FHV and CrPV (Reinganum, 1975; Dasgupta et al., 2003). Illustrating further the susceptibility of insects to infection by a variety of different viruses, it has been observed that Drosophila can be experimentally infected by Sindbis virus, Dengue virus, West Nile virus, and Rift Valley Fever virus, that naturally infect mosquitoes and vertebrates, and recently this has emerged as a model to study arbovirus infections (Xu and Cherry, 2014). These examples make it clear that in many cases the virus-host specificity is not well-defined and that it is possible for viral particles to enter non-target hosts with unknown consequences. With the identification of many new viral species through next-generation sequencing techniques (Liu S. et al., 2011; Liu et al., 2015; Li C.-X. et al., 2015; Webster et al., 2015), it remains to be determined how specific the newly virus infections will be and how efficiently they can cross other species barriers.

During the last decades, the baculovirus expression system, based on AcMNPV, has evolved into a major platform for production of recombinant proteins (van Oers et al., 2015). Important applications of the baculovirus expression system include the production of vaccines and gene therapy vectors based on adeno-associated virus (Felberbaum, 2015). The availability of commercial platforms for production of recombinant AcMNPV also allows the rapid generation of baculoviruses expressing RNA hairpins. In insect culture cells, baculoviruses expressing dsGFP were reported to silence GFP fluorescence (Huang et al., 2007). Silencing effects were also observed with antisense constructs (Hajos et al., 1999). However, baculoviruses generally are pathogenic and effects of gene silencing can be masked by non-specific toxic effects. When recombinant BmNPV (also generated by bacmid technology; Motohashi et al., 2005) was used to deliver dsRNA targeting a juvenile hormone esterase-related (JHER) gene in the corn borer, Sesamia nonagrioides, many non-specific phenotypic effects associated with disrupted metamorphosis were observed (as also observed in BmNPV expressing control dsRNA; Kontogiannatos et al., 2013), despite the restricted host specificity of BmNPV (Maeda et al., 1993). Only through the analysis of a large number of animals could specific phenotypes be identified that were associated with knockdown of JHER (Kontogiannatos et al., 2013). A solution for this drawback could be the generation of incapacitated baculoviruses, i.e., baculoviruses that are deficient for an essential gene in the infection cycle, such as ie-1 or lef-8 (Efrose et al., 2010; Ioannidis et al., 2016). Incapacitated baculoviruses are predicted to enter target cells and initiate early gene expression without progressing to the late phase of the infection cycle and cell lysis. For gene transduction purposes, incapacitated baculoviruses should therefore be engineered with expression cassettes with constitutive cellular or early viral promoters instead of the very late promoters used for protein expression. However, the production of virions that contain deficient genomes in rescue cell lines, which provide the missing functions as transgenes, has turned out to be challenging and only low titers of incapacitated baculoviruses could be produced (Efrose et al., 2010; Ioannidis et al., 2016). The use of baculoviruses for transduction of dsRNA in insects therefore requires the isolation of variants/mutants with low virulence that can be easily manipulated genetically and produced in cell lines.

Plasmids that carry densovirus genomes have been used to produce infectious particles after injection into susceptible insect hosts or after transfection of cell lines (Afanasiev et al., 1999; Ren et al., 2008; Suzuki et al., 2014). Interestingly, plasmids containing densovirus sequences can integrate into the cellular genome and copy numbers increase upon deletion of the non-structural genes (Bossin et al., 2003).

Recombinant densovirus vectors (Brevidensovirus genus) were constructed for delivery of short RNA hairpins in mosquitoes (Gu et al., 2011). Aedes albopictus C6/36 cells were used for production of recombinant virions which required co-transfection with plasmids containing intact densovirus genomes (resulting in production of helper viruses). In recombinant virus genomes, an intron was inserted between the ORF of the non-structural gene 1 (NS1) and a GFP cassette, that contained a RNA polymerase III-driven short hairpin RNA (shRNA) cassette (Figure 1). Transcription by the viral promoter results in the production of a pre-mRNA which is spliced to generate mRNA for functional NS1-GFP proteins (and green fluorescence) while the intronic expression cassette functions as a source of siRNA. When Ae. albopictus mosquito larvae were exposed to mixtures of helper and recombinant densovirus with the shRNA cassette targeting V-ATPase, silencing of the targeted gene was observed together with increased mortality. Monitoring of GFP fluorescence after infection demonstrated that densovirus could infect many different mosquito tissues (Gu et al., 2011).

Densoviruses as a group have a wide distribution among insect species and were reported to occur in representatives of many insect orders (Liu K. et al., 2011; Han et al., 2013). However, mosquito densoviruses are highly specific to mosquitoes (Suzuki et al., 2015). Specificity and virulence toward different hosts seem to be determined by residues in the capsid sequences (Multeau et al., 2012). Despite their sometimes high virulence in insects, densoviruses are not known to infect mammals (Carlson et al., 2006; Jiang et al., 2007). On the other hand, densoviruses can also occur as persistent infections (and being vertically transmitted) and can make insect hosts refractory to other pathogens (Xu et al., 2014). In summary, densoviruses represent a group of insect DNA viruses which have a limited host range, are capable of different infection states and can be easily manipulated and produced in cell lines. These properties will drive the research on densoviruses for applications in environmentally safe pest control, and as gene transduction and RNAi delivery vectors.

Although tedious, production of recombinant DNA virus is straightforward if an in vitro culture system is available. This is illustrated by the generation of recombinant iridovirus in a coleopteran cell line (Ozgen et al., 2014). After identification of a locus in the viral genome of which the inactivation would not interfere with the infection cycle, a transfer vector is created consisting of a plasmid that contains the homologous viral regions flanking the insertion site of the transgene. For applications as RNAi delivery vector, the transgene consists of an RNA hairpin expression cassette, perhaps in combination with a fluorescence reporter cassette to facilitate selection of recombinant virus. After transfection of the transfer vector in virus-infected cells, fluorescent viruses are generated by homologous recombination between the transfer vector and the wild-type DNA virus. Recombinant viruses are subsequently purified by plaque purification (O'Reilly et al., 1992).

As is the case for DNA viruses, production of recombinant RNA viruses typically requires the cloning of the genome which then can be manipulated and expressed using a suitable promoter. However, efficient amplification and packaging of linear viral RNA genomes in capsids, and eventually envelopes, to produce infectious virions typically requires precise 5′- and 3′-ends in the viral genomic RNAs. In the case of FHV (Nodaviridae), 5′-extensions are minimized by placing the promoter such that transcripts are initiated at or very close to the authentic 5′-end of the viral genome, while precise 3′-ends are generated by cleavage of appropriately positioned hepatitis delta virus ribozyme (Ball and Johnson, 1999; van Cleef et al., 2014). In addition, the ends of linear RNA genomes often form strong secondary structures which complicate the isolation of 5′- and 3′-ends (Carrillo-Tripp et al., 2015). Picorna-like viruses, such as iflaviruses and dicistroviruses, often occur as quasi-species in which multiple viral forms complement each other to support infection that prevents the isolation of efficient individual infectious sequences (Ojosnegros et al., 2011) (see also further below). Such features make the construction of infectious cDNAs of RNA viruses challenging, even if it concerns small genome sizes. However, a population cloning strategy in which multiple independent clones of the 5′ part of the genome are ligated to multiple clones representing the 3′ part may make it possible to recover infectious clones with different properties. For the plant potexvirus Alternanthera mosaic virus (AltMV), this strategy allowed the isolation of multiple virus clones that differ in replication and silencing suppression (Lim et al., 2010b). Different RNA silencing suppression efficacy in the clones could be correlated with their suitability for virus-induced silencing (VIGS) or protein expression applications (Lim et al., 2010a).

Another obstacle that prevents the establishment of plasmid-based production systems of RNA viruses is the availability of suitable insect cell lines. The vast majority of cell lines is derived from lepidopteran and dipteran insects (Lynn, 2001), while more recently cell lines have been reported from coleopterans and hymenopterans (Smagghe et al., 2009; Goodman et al., 2012; Goblirsch et al., 2013). Cell lines derived from insects of other orders are very rare (Boyapalle et al., 2008). Viruses may be restricted to particular cellular environments (from single or related host species) for robust replication and virion reproduction. In case of adaption to more widely used cell lines, danger exists for decrease in infectivity of the original insect host. It is also observed that species-specific viruses have great difficulties to replicate in cell lines derived from the host. For instance, an in vitro cell culture system for CPV (Reoviridae) in insects has not been reported despite the availability of many host cell lines (Hill et al., 1999; Swevers et al., 2014). This contrasts with the existence of reverse genetic systems for reoviruses using mammalian cell lines (Boyce et al., 2008). For viruses that are difficult to grow in cell lines, tissue culture has been used as an alternative (Arif and Pavlik, 2013).

Because of its importance with respect to human and animal health, considerable effort has been paid to the development of infectious clones and reverse genetics systems for arboviruses that cause disease in humans and livestock (Aubry et al., 2015). Such systems were developed some time ago based on mammalian cell lines for alphaviruses (Sindbis virus; Bredenbeek et al., 1993), flaviviruses (Yellow Fever virus, Dengue virus; Ruggli and Rice, 1999), rhabdoviruses (Vesicular Stomatitis Virus; Lawson et al., 1995), bunyaviruses (model Bunyamwera virus; Bridgen and Elliott, 1996), and also more recently reoviruses (Blue Tongue virus; Boyce et al., 2008). Recombinant arboviruses generated from infectious clones were also used to study infection in mosquito cell lines and mosquitoes (e.g. Siu et al., 2011; Slonchak et al., 2014). In particular, a Sindbis virus expression system was developed, in which engineered Sindbis viruses contain a second subgenomic RNA promoter element that can be used for protein or antisense RNA expression (Johnson et al., 1999; Foy et al., 2004). Engineered Sindbis viruses were used to deliver RNAi in mosquitoes and the silkworm Bombyx mori (Adelman et al., 2001; Shiao et al., 2001; Uhlirova et al., 2003). Because of potential health hazard, however, manipulation of arboviral genomes and production of infectious virions require containment procedures of biosafety level 2 or higher. Safety issues will also prevent the use of arbovirus vectors as RNAi delivery vehicles in insect (mosquito) populations in nature.

Similar safety issues apply for the use of pseudo-typed mammalian retroviruses that were used in the past as gene transduction vectors in Drosophila and mosquitoes (Matsubara et al., 1996; Jordan et al., 1998; Teysset et al., 1998). After infection, retroviral genomes undergo reverse transcription into a DNA form which subsequently integrates in the cellular genome as a provirus. Proviruses are transcribed as cellular genes, and retroviral vectors engineered with RNA hairpin cassettes therefore could be applied as a permanent means of dsRNA production in infected cells. Interest for pseudo-typed “pantropic” retroviruses declined because of difficulties in production of virions at high titers. More recently, there exists renewed interest for pseudo-typed mammalian retroviruses as a transgenesis system in worms (Mann et al., 2014; Hagen et al., 2015). Retrovirus-based vectors have been used extensively in mammalian systems and were applied to perform large scale screens to analyze basic physiological processes and investigate disease mechanisms (Root et al., 2006; Manjunath et al., 2009) (Figure 2).

Plasmid-based systems were developed for FHV (Nodaviridae) that permit robust replication in a variety of organisms including insects, yeast and plants (Selling et al., 1990; Price et al., 1996; van Cleef et al., 2014). The (+)ssRNA genome of FHV consists of two segments that encode the RNA-dependent polymerase and the capsid protein, respectively. During replication, subgenomic RNA3 is generated that produces the B2 RNAi inhibitor (Venter and Schneemann, 2008). An expression cassette that transcribes RNA1 with precise 5′- and 3′-ends, realized by positioning of the promoter sequence and self-cleaving ribozyme, respectively, can initiate high levels of FHV replication (Figure 3). In the presence of RNA2, the replication system will generate functional virions. The FHV replication/production system has been modified to drive GFP expression in infected cells, either through insertion in the 3′-part of RNA1 (Price et al., 2000) or through modified (defective) RNA2 molecules (Dasgupta et al., 2003). FHV infection results in production of viral siRNAs (Gammon and Mello, 2015) and insertion of foreign sequences in the FHV genome could therefore be employed to deliver specific RNAi effects in infected cells.

Two other important (+)ssRNA virus families that are known for their ubiquitous infections of insects are dicistroviruses and iflaviruses (Picornavirales order) (Ryabov, 2016). Both types of virus have a (+)ssRNA genome of 9–10 kb containing an attached viral protein (VPg) at the 5′-end and a poly(A)-tail at the 3′-end. Viral proteins are expressed after proteolytical cleavage of polyproteins encoded by one (iflaviruses) or two (dicistroviruses) long ORFs. Their occurrence in bees has been well studied because of possible links with colony-collapse disorder in the honeybee and declining health of pollinators (Niu et al., 2014; McMenamin and Genersch, 2015; Gisder and Genersch, 2016). Full-length clones were reported for the genomes of dicistroviruses infecting the Bird cherry-oat aphid Rhopalosiphum padi (RhPV) and the honeybee (Black queen-cell virus, BQCV), but their capacity to produce infectious virions was not established conclusively (Benjeddou et al., 2002; Boyapalle et al., 2008; Pal et al., 2014; Carrillo-Tripp et al., 2015). Only very recently infectious genomic RNA was produced by T7 RNA polymerase transcription of cloned sequences of the model dicistrovirus CrPV (Kerr et al., 2015). As is the case for FHV, insertion sites for foreign sequences that do not disrupt the viral function need to be identified for the generation of recombinant reporter viruses or RNAi delivery vectors.

An important question is to what extent siRNAs are produced after infection of insect-specific RNA viruses. RNA viruses encode VSRs that prevent the activation of the RNAi mechanism (Ding and Voinnet, 2007; Vijayendran et al., 2013; Gammon and Mello, 2015; Olson and Blair, 2015). As for baculoviruses, general pathogenic effects may mask the RNAi response. During persistent infections, on the other hand, Dicer-produced siRNAs may not become loaded in RISC complexes, limiting the extent of silencing the target genes (Flynt et al., 2009; Swevers et al., 2016). SiRNAs may not be produced evenly across the length of the viral genomes (e.g. for FHV Aliyari et al., 2008; Flynt et al., 2009) and insertion in particular “hotspot” regions may trigger a more effective silencing mechanism.

Studies of the effect of VSRs on VIGS during virus infection in insects are rare. One example is the incorporation of the B2 RNAi suppressor of FHV in the genome of Sindbis virus, an arbovirus without known RNAi suppression activity, which resulted in high mortality and suppression of accumulation of vsRNAs (Myles et al., 2008). Differences in pathogenicity between the dicistroviruses DCV and CrPV were also attributed to the strength of the encoded VSRs (Nayak et al., 2010). The balance between VSR potency and efficiency of viral replication kinetics is therefore considered as a major determinant for the outcome (persistent or pathogenic) of a viral infection (O'Neal et al., 2014). On the other hand, more data are available on the effect of VSR on VIGS in plants. For example, synergism between Tobacco etch virus (TEV) and Potato X virus (PXV) is caused by the efficient VSR activity of the TEV-encoded helper component proteinase (HC-Pro) which acts to protect and enhance replication of co-infecting PVX (Shi et al., 1997). Mutations in HC-Pro that are correlated with decreased VSR activity are also responsible for decreases in viral pathogenicity in other potyviruses (Shiboleth et al., 2007; Seo et al., 2011; Li M.-J. et al., 2014). As already mentioned above, variability in efficacy of the VSR affects the applicability of AltMV for protein expression vs. VIGS in plant tissue (Lim et al., 2010a,b). While parallels may exist between insect and plant infections, differences in the RNAi machinery between the two need to be considered, such as the existence of multiple dicer enzymes and the amplification mechanism by RNA-dependent RNA polymerase in plants (Siomi and Siomi, 2009), which likely make the RNAi response more powerful and therefore more subjected to control mechanisms.

Finally, more sophisticated approaches may be needed that go beyond simply inserting a target sequence in the viral genomes. One example is the insertion of a functional miRNA-expressing cassette in an intron of one of the segments of influenza viral genome which was used to silence a GFP reporter in mammalian cells (Varble et al., 2010). Influenza virus (Orthomyxoviridae) however replicates in the nucleus and therefore will provide a more natural environment for processing miRNA-like genes when compared to RNA viruses that replicate in the cytoplasm (Asgari, 2015). All these considerations underline the immaturity of the field and major research efforts are required to investigate the suitability of the approach which may depend on the type of virus and the type of infection.

Plant viruses are discussed separately because robust production systems are typically available using plants. Agro-infiltration is being used for high efficiency delivery of recombinant plant virus genomes in plant tissue (Mallory et al., 2002; Marillonnet et al., 2005; Lindbo, 2007). Besides for protein production (Gleba et al., 2007), viral expression systems, e.g. based on tobacco mosaic virus (TMV) and potato virus X (PVX), were also used to study phenotypic effects of specific gene silencing in plants (virus-induced gene silencing, VIGS; Kumagai et al., 1995; Ruiz et al., 1998; Purkayastha and Dasgupta, 2009; Lange et al., 2013). Because the recombinant viruses carry an insert corresponding to a cellular gene, the RNAi response will silence both viral genes and the host gene targeted by the insert. A more recent application is the use of plant viral expression vectors as viral dsRNA production systems (VDPS; Kumar et al., 2012). In this case, additional dsRNA molecules produced during viral replication are not targeted to endogenous plant genes but to genes of herbivore nematodes and insects to provide protection against feeding (Dubreuil et al., 2009; Kumar et al., 2012). Vectors based on tobacco rattle virus (Figure 4) are preferred because mild disease symptoms are induced and expression is both more intense and more persistent (Ratcliff et al., 2001). However, different plant virus expression systems may need to be tested for different host plant-insect pest systems to achieve optimal performance (Wuriyanghan and Falk, 2013). For instance, viruses that replicate preferentially in the phloem, such as the citrus tristeza virus (CTV), were engineered in similar fashion to provide protection against the Asian citrus psyllid Diaphorina citri, which feeds on the phloem (Hajeri et al., 2014). Several more examples of the application of plant viruses to affect mortality or fecundity of insect pests include the use of recombinant TMV in the caterpillar Mythimna separata and the citrus mealybug Planococcus citri (Khan et al., 2013; Bao et al., 2016) and of PVX in the solenopsis mealybug, Phenacoccus solenopsis (Khan et al., 2015). A high-throughput Gateway-enhanced AltMV system was also described to test RNAi constructs from a cDNA library for effects on growth and fecundity in the whitefly, Bemisia tabaci (Ko et al., 2015). Because the transgenic approach of plant-mediated RNAi requires a long term of development, the use of plant viral vectors provides a fast-track approach for testing the efficiency of silencing of candidate genes and induction of associated toxic effects.

However, the VDPS systems mentioned above cannot be viewed as strict viral delivery systems of RNAi. The plant viruses are rather used as tools to produce dsRNA in plant cells at sufficient levels to cause silencing effects in the insect after feeding on the plant tissue. Viral delivery is here viewed as the use of viruses to transmit the RNAi signal, which is typically realized after infection of insect tissues by viral particles and when viral replication occurs in insect cells. Since most plant viruses do not replicate in insect tissues, they cannot be regarded as strict viral delivery systems but merely as dsRNA production systems in plant tissue. Only plant viruses that can penetrate the gut barrier are therefore considered as engaged in “viral delivery of dsRNA.”

As discussed already above, persistent circulative plant viruses are capable to transverse the gut-hemocoel barrier. Of these, non-propagative viruses transverse different tissues without replication and it is not clear whether active dismantling of virions occurs during passage. This more passive behavior resembles that of “virus-like particles” which will be discussed in the next section. A few reports however indicated the occurrence of replication in insect vectors of viruses commonly considered to be transmitted in a non-propagative manner (Rivera and Gamez, 1986; Pakkianathan et al., 2015). During passage through the insect body, it can be expected that cells become infected, virions are dismantled and moderate viral replication may be initiated. Such considerations invite research that (re-)addresses the capacity of plant viruses to replicate in insects and can lead to surprising findings, such as the recent report of replication and virion production of tobacco ringspot virus (Picornavirales) in honeybees and their Varroa mite parasites (Li J. L. et al., 2014); but which is contested (Miller et al., 2014). As previously mentioned, B. mori-derived BmN cells and other lepidopteran cell lines can be persistently infected by Macula-like virus that is related to plant Tymoviruses and apparently has crossed the plant-insect species barrier (Katsuma et al., 2005). However, virus replication may not at all be necessary for “survival” of the virus during the passage of the insect body. It has also been reported some time ago that specific interactions between the luteovirus potato leafroll virus and GroEL, a chaperone-like protein produced in the aphid haemolymph by endosymbiotic Buchnera bacteria, protect virus particles from degradation in the hemocoel of the aphid when it is transported to the salivary glands of the aphid (Hogenhout et al., 1998).

Plant viruses that are transmitted by insect vectors in a persistent circulative and propagative manner are the most interesting candidates since active viral replication, and consequently amplification of the RNAi response, will occur in the insect hosts. Circulative and propagative viruses belong to the families Bunyaviridae, Reoviridae, and Rhabdoviridae, and have a complex (−)ssRNA, segmented ambisense RNA or segmented dsRNA genome structure which has made the generation of infectious clones and the development of reverse genetics systems challenging. Only very recently a reverse genetics system was developed for Sonchus yellow net rhabdovirus (SYNV; Wang et al., 2015). The procedure involves agro-infiltration of tobacco plants of a construct expressing the negative strand of the viral genome together with co-expression of nucleoprotein, phosphoprotein, core polymerase and suppressors of RNAi (Figure 4). Recombinant SYNV expressing GFP was generated and used to investigate the functions of the movement protein and the glycoprotein (Wang et al., 2015). It will be interesting to see how the reporter SYNV infects the insect vector and whether delivery of RNAi triggers is possible. A shortcoming for this research also is that the production of rhabdoviral virions occurs in a plant system. There is a need for the development of insect cell lines that are capable to support replication and virion production of plant viruses to study their infection cycle in the insect vectors (Ma et al., 2013). This is feasible since bunyaviruses, reoviruses, and rhabdoviruses occur also as insect-specific viruses and arboviruses, and therefore likely originated in their insect hosts.

Interestingly, plants are capable of producing insect viruses with a broad host range, such as FHV (Nodaviridae). High titers of FHV could be generated in transgenic tobacco plants that express movement proteins of plant viruses (Dasgupta et al., 2001). It is speculated that FHV originated from recombination events between a plant virus and a virus from a plant-feeding insect, which would explain its wide host range. Although it replicates in insects (Dasgupta et al., 2007), FHV is not known to be transmitted by them, while infection by mechanical disruption is possible.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.