Endogenous mRNAs produced in the nucleus of eukaryotic cells are subsequently exported to the cytoplasm for protein synthesis. There they are exposed to a complex RNA regulatory network. This network is able to adjust the amount of individual mRNAs within the total RNA population and to keep translationally active and inactive mRNAs in a delicate balance. Regulation of RNA allocation and turnover concerns not only mRNAs but also other RNA types in the cells including intruding viral RNAs (vRNAs). vRNAs possess features of aberrant RNAs which are often recognized as targets of the RNA decay machinery. Typically virus infection is initiated by the production of viral replication proteins and establishment of the replication complex in which progeny genomes and subgenomic RNAs (sgRNAs) are transcribed. Newly synthesized vRNAs and sgRNAs are translated into various viral proteins which carry out all viral functions together with compatible host factors. They also fight against the host's defense pathways. Therefore, to achieve a robust and productive infection, the vRNA molecules need to be able to cope with the cellular degradation and translation machineries. The eukaryotic RNA regulatory network including antiviral defense involves several complex pathways that have been shown to interact, compete and overlap on many levels (Figure 1) and altogether have a central role in maintaining cellular homeostasis. This review aims to illuminate the role of the RNA regulatory network in plant virus infection from different angles. We will overview the essential processes, cover the sites for these processes, the RNA granules, and highlight the ways some plant viruses modify and take advantage of the RNA regulatory network and its components.

More than 70 different proteins have been identified in the SGs and over 20 in the PBs of eukaryotic cells. In addition many proteins are shared between different RNA granule types in yeast and mammals (Poblete-Durán et al., 2016). Although it is evident that plant, yeast and animal PBs and SGs have similar functions (Xu et al., 2006; Weber et al., 2008; Xu and Chua, 2009; Sorenson and Bailey-Serres, 2014), the composition of RNA granules is currently less studied in plants. Nevertheless, several hallmark-proteins have been identified from plant SGs, including UBP1, eukaryotic initiation factor 4E (eIF4E), poly-A binding protein (PABP), and PBs, including DCP1, DCP2, VCS, and AGO1 (Xu et al., 2006; Weber et al., 2008; Sorenson and Bailey-Serres, 2014).

Multifunctional G3BPs participate in SG assembly upstream of eIF2α phosphorylation. In animal cells stress-induced dephosphorylation of S149 in G3BP causes a conformational change enabling homo- and heteromultimerization with G3BP2 (Tourriere et al., 2003; Matsuki et al., 2013). Oligomerization of G3BPs causes the nucleation of SGs even in the absence of stress suggesting it might be a prerequisite for the eventual assembly of other SG components (Matsuki et al., 2013). SGs are regarded as antiviral compartments and many viruses have developed means to disrupt SG assembly via G3BP interactions. Some viruses, such as vaccinia virus and hepatitis C virus, have co-opted G3BP to serve as a potential component of viral factories and replication complexes (Yi et al., 2006; Katsafanas and Moss, 2007) or an assisting factor for virion assembly (Garaigorta et al., 2012). Polioviral 3C protease cleaves G3BP leading to the formation of SGs without G3BP (White et al., 2007; Piotrowska et al., 2010). A G3BP-like protein localizing to plant SGs was recently discovered in Arabidopsis. AtG3BP was identified through its interaction with a viral protein, the nuclear shuttle protein 3 (NSP3) of the abutilon mosaic virus (genus Begomovirus), in SGs (Krapp et al., 2017). The interaction between G3BP and viral proteins was mapped to a conserved FGDF-type motif. Also in animal cells G3BP has been shown to associate with the C-terminal FGDF motifs of ns3 protein of the alphaviruses Semliki Forest virus and Chikungunya virus (Panas et al., 2014; Scholte et al., 2015). The FGDF motif is relevant for G3BP-interactions of the host proteins as well (Panas et al., 2015). Interestingly, FGDF motifs are present in the proteases of potyviruses, waikaviruses, and closteroviruses suggesting that SG regulation via G3BP interactions may have greater influence in plant virus infections than currently known.

Animal RNA viruses make use of TIA1, the homolog of UBP1. A recent report showed that the knockdown of TIA-1 impaired the accumulation of Newcastle disease virus (family Paramyxovirdae) in HeLa cells (Sun et al., 2017). Other viruses such as the Dengue virus (genus Flavivirus) and Poliovirus (PV; genus Enterovirus) have evolved a different strategy: viral proteins sequester TIA1 so that it cannot form functional SGs. The Dengue virus interferes with SG formation by using nonstructural protein NS1 to sequester TIA1 to the perinuclear space (Xia et al., 2015). Likewise the poliovirus disrupts host defense by aggregating TIA1 into nonfunctional foci (White and Lloyd, 2011; Fitzgerald and Semler, 2013).

Nicotiana benthamiana UBP1 was identified as a component of PVA-induced RNA granules (Hafrén et al., 2015). Although these granules do not represent canonical SGs, UBP1 is important for the formation of PVA-induced granules and seems to be one of the many host factors PVA employs to ensure a successful infection. Potyviral genome-linked protein VPg specifically promotes PVA translation (Eskelin et al., 2011; Hafrén et al., 2013). Interestingly, the granulation-prone UBP1 does not promote PVA translation together with VPg, whereas many other components co-localizing to PVA-induced granules, e.g., VCS and eIF(iso)4E, do. The role of these granule proteins in PVA translation in turn propose a link between the granules and viral translation. Supporting this, VPg is able to disrupt the PVA-induced granules while promoting translation (Hafrén et al., 2015). Additional evidence suggested a functional link between granule formation and RNA silencing suppression. PVA-induced granules were neither formed in the absence of potyviral silencing suppressor HCPro nor by HCPro silencing suppression-deficient mutants suggesting that PVA-induced granules could be the sites for HCPro-guided silencing-suppression function (Hafrén et al., 2015). Similar granule-like foci, observed in the vicinity of the ER and the microtubule cytoskeleton, were induced by the HCPro of potato virus Y in Nicotiana benthamiana (del Toro et al., 2014).

Cysteine3Histidine (CCCH)-type zinc finger proteins and tandem zinc finger proteins (TZF) comprise a large protein family well conserved across eukaryotes. The PB and SG component Tristetraprolin (TTP) belongs to this protein family. It has a critical role in mRNA metabolism in recruiting enzymes involved in deadenylation, decapping, and exonucleolytic activities to specific mRNAs related to innate immunity and inflammatory responses in animals proposing a role in virus infection (Hamid and Makeyev, 2016). In plants the CCCH-motif is preceded by an arginine-rich (RR)-motif and therefore these proteins are called RR-TZF proteins (reviewed in Bogamuwa and Jang, 2014). Although plant RR-TZF are involved in biotic stress responses potentially by binding to specific elements of the corresponding mRNAs and recruiting the assembled mRNP complexes into PBs and SGs, no report has linked them to regulation of virus infection in plant cells so far.

Although PBs have housekeeping functions and are always present in the cells, formation of visible PBs is activated for example as a consequence of RNA silencing leading to an increase in translationally repressed mRNAs in the cell (Eulalio et al., 2007b). PB assembly upon induction of gene silencing was shown in plants using an RNA hairpin targeting one specific transcript, the Sulphur mRNA (Meteignier et al., 2016) and in connection to virus recovery (Ma et al., 2015). Viruses can affect PBs either by dispersing PB structures and components or by interfering with the mechanism of PB formation. For example poliovirus and human Coxsackie virus, both belonging to family Picornaviridae, interfere with PB formation during their replication cycle (reviewed in Malinowska et al., 2016). In addition, polioviral 3C protease processes XRN1 and DCP1 (Dougherty et al., 2011). Some viruses exploit PB components for their benefit. The LSm1-7 complex, which is an activator of decapping in the 5′–3′ exoribonucleolytic pathway (Tharun et al., 2000) and a component of PBs (Ingelfinger et al., 2002), is a central regulator of brome mosaic virus (genus Bromovirus) translation and replication (Galao et al., 2010). Studies in yeast revealed that the cis-acting RNA replication signals within RNA 2 and RNA3 were found to direct these brome mosaic virus RNAs into PBs (Beckham et al., 2007). PBs create an environment which lacks the components of the translation machinery and subsequently may facilitate replication complex assembly. On this basis, a role for PBs in the transition of brome mosaic virus RNAs from translation to replication was suggested in Galao et al. (2010; Figure 2C).

Plant antiviral defense based on the PB component AGO1 and its slicer activity can be vulnerable to viral silencing suppressors. For example Arabidopsis AGO1 has been shown to be a target of the 2b silencing suppressor of the cucumber mosaic virus (genus Cucumovirus). Cucumber mosaic virus 2b co-localizes with AGO1 to PBs and directly blocks AGO1's slicer activity both in vitro and in vivo (Zhang et al., 2006). AGO1 inactivation is mediated also by the turnip crinkle virus capsid P38 which can mimic AGO-binding GW-repeat proteins (Azevedo et al., 2010) and sweet potato mild mottle virus (genus Ipomovirus) P1 protein which blocks target RNA binding to preassembled AGO1 (Kenesi et al., 2017). AGO1 degradation has been reported to be mediated by PVX P25 (Chiu et al., 2010), beet western yellows virus (genus Polerovirus) P0 (Baumberger et al., 2007; Bortolamiol et al., 2007) and tomato ringspot virus (genus Nepovirus) coat protein (Karran and Sanfacon, 2014). In spite of the important role of AGO2 in antiviral RNA silencing in planta (reviewed in Carbonell and Carrington, 2015), it is not known whether it associates e.g., with PBs or SGs. Different types of stresses direct human AGO2 to SGs (Detzer et al., 2011) and it is a component of mammalian PBs (Sen and Blau, 2005; Hubstenberger et al., 2017).

Some recent studies link the antiviral defense response both to translational repression and PBs. Recovery is a phenomenon during which host plants recover from viral symptoms. The amount of polysome-associated viral transcripts is reduced and the amount of PBs is increased in plants recovered from tobacco rattle virus (genus Tobravirus) infection proposing an important role for PBs in the elimination of vRNAs (Ma et al., 2015). Translational repression is involved also in recovery from tomato ringspot virus infection (Ghoshal and Sanfaçon, 2014). Activation of a plant nucleotide-binding site leucine-rich repeat (NB-LRR) protein by the tobacco mosaic virus p50 replicase protein fragment can induce an antiviral response against PVX and turnip crinkle virus in N. benthamiana (Bhattacharjee et al., 2009). During this antiviral response PVX transcripts are translationally repressed and the number of PBs increases significantly in the affected cells (Meteignier et al., 2016). Although the NB-LRR response, UV stress and RNAi against certain transcripts all induce translational repression and PB formation in plants, the mechanisms leading to enhanced PB assembly are distinct. While induction of RNAi increases the number of PBs and the presence of silencing suppressors like P19 inhibits PB induction, PB formation during the NB-LRR response against PVX is not dependent on RNA silencing (Meteignier et al., 2016). Regardless of the mechanism that promotes PB formation, their increased assembly most likely reflects the high concentration of translationally repressed mRNAs.

siRNA bodies are RNA granules associated with the synthesis of tasiRNA and the amplification of the RNA silencing signal. Plant proteins found in siRNA bodies are listed in Table 3. Many plant viruses interfere with siRNA body functions in order to suppress their contribution to siRNA production. Some viruses such as rock bream iridovirus (genus Megalocytivirus), a fish DNA virus, (Zenke and Kim, 2008) and sweet potato chlorotic stunt virus (genus Closterovirus; Kreuze et al., 2005), encode class I RNase III enzymes which have RNA silencing suppression capacity. Class 1 RNase III endoribonuclease (RNase3) of sweet potato chlorotic stunt virus contains a dsRNA binding domain and a Mg²⁺-dependent catalytic domain for dsRNA cleavage (Kreuze et al., 2005). Its expression in sweet potato as a transgene promotes the accumulation of sweet potato feathery mottle virus (genus Potyvirus; Cuellar et al., 2009). Sweet potato chlorotic stunt virus RNase3 co-localizes to siRNA bodies via interaction with SGS3 (Weinheimer et al., 2016). It is likely that the virus uses SGS3 to target its RNase3 to dsRNA synthesized by RDR6 to cleave them into sRNAs which are too small to be used in RNAi. The potyviral protein VPg is, among its other functions, a suppressor of sense-mediated silencing that localizes to siRNA bodies and interacts with SGS3 to benefit virus infection (Rajamäki et al., 2014; Cheng and Wang, 2017). The turnip mosaic virus VPg carries out its silencing suppression function by targeting SGS3 and its binding partner RDR6 directly to degradation via the 20S proteasome (Cheng and Wang, 2017). As a result the production of vsiRNA in the siRNA-bodies will be inhibited. SGS3 functions are disrupted also by other viruses. For example, tomato yellow leaf curl virus (genus Begomovirus) V2 protein interacts directly with SGS3. Mutational analysis showed that this interaction is required for V2 silencing suppression function (Glick et al., 2008). Plantago asiatica mosaic virus (genus Potexvirus) TGBp1 (P25) co-localizes to siRNA bodies and inhibits dsRNA synthesis by the SGS3/RDR6 complex (Okano et al., 2014). Furthermore, yeast two-hybrid and bimolecular fluorescence assays revealed that the p2 silencing suppressor protein of the rice stripe virus (genus Tenuivirus) interacts with SGS3 suggesting that it could act through inactivation of the RDR6/SGS3 pathway (Du et al., 2011).

Some viruses interfere with the RNAi amplification step by inhibiting RDR6 functions. For example, the βC1 protein of tomato yellow leaf curl China betasatellite (family Geminiviridae) up-regulates the expression of N. benthamiana calmodulin-like protein REGULATOR OF RNA SILENCING (Nbrgs-CaM) which is an endogenous silencing suppressor. In turn, up-regulation of Nbrgs-CaM leads to the suppression of RDR6 gene expression resulting in both PTGS suppression and symptom induction (Li et al., 2014). Downregulation of RDR6 mRNA accumulation is also achieved by the silencing suppressor HCPro of sugarcane mosaic virus (genus Potyvirus; Zhang X. et al., 2008). The P6 protein of rice yellow stunt rhabdovirus (genus Nucleorhabdovirus) interacts with rice and Arabidopsis RDR6 in vivo and affects the production of secondary siRNAs by RDR6 (Guo et al., 2013). In conclusion, interference with siRNA body functions to inhibit the RNAi amplification phase seems to be widely used by both RNA and DNA viruses.

The nucleocytoplasmic compartmentalization of eukaryotic cells paved the way to the development of an extensive post-transcriptional regulatory network at the level of RNA decay. In addition to the core mRNA decay machinery, regulation is provided by several sRNA pathways (siRNAs, miRNAs), NMD and mechanisms targeting RNAs displaying specific secondary structures. An interesting hypothesis of the exaptive origins of regulated mRNA decay in eukaryotes has been set forward by Hamid and Makeyev (2016). They propose that the emergence of RNAi, NMD, and mRNA decay based on the recognition of specific RNA structural elements to aid in non-self RNA recognition coincided with the expansion of RNA viruses. Only later these mechanisms adapted to regulate cellular mRNAs. Out of the currently recognized plant virus genera roughly 70 % represent (+)RNA viruses, 9% (-)RNA and 9% dsRNA viruses (Dolja and Koonin, 2011). RNA viruses are overrepresented in plants as only about 12% of plant virus genera are DNA viruses and retroelements and viroids still add to the number of pathogenic RNA molecules within plant cells. Taking into account the vast number of RNA viruses in plants it is likely that more discoveries linking plant virus infection, antiviral defense, innate immune signaling and cellular RNA biology will be made in the future. This is exemplified by a recent finding of a novel type of RNA-targeting antiviral defense by Drosha-type Class 2 RNase III nuclease (Aguado et al., 2017). Drosha, which has evolved to function in the biogenesis of miRNAs in eukaryotic organisms, and related RNase III enzymes are translocated into the cytoplasm upon RNA virus infection. There they elicit unique RNA-targeting antiviral activity against viruses belonging to families Togaviridae and Flaviviridae via binding to certain stem-loop structures on vRNA thus hindering the activity of the RNA-dependent RNA polymerases (Aguado et al., 2017). When a pre-miRNA sequence is nested into the flavivirus-like turnip crinkle virus RNA it also becomes processed by an RNase III activity indicating that this enzyme is translocated from the nucleus to the cytoplasm upon virus infection also in higher plants. The authors propose that this particular type of RNase III-mediated antiviral defense conserved in mammals, arthropods, fish and plants is independent of the other known eukaryotic antiviral defense mechanisms and may have preceded them. Furthermore, Martinez-Perez et al. (2017) recently discovered a novel mechanism that regulates plant viral infection via N⁶-methyladenosine (m⁶A) modifications of vRNA. Human immunodeficiency virus (genus Lentivirus) and hepatitis C virus (genus Hepacivirus) use this mechanism to regulate viral gene expression and infection (Tirumuru et al., 2016) and infectious particle production (Gokhale et al., 2016), respectively. The m⁶A abundance on alfalfa mosaic virus (genus Alfamovirus) genomic RNA is regulated by Arabidopsis AtALKBH9B, which carries m⁶A demethylase activity (Martinez-Perez et al., 2017). Its silencing restricts the alfalfa mosaic virus infection. In vivo AtALKBH9B co-localizes with siRNA body component SGS3 and the NMD component UPF1 and shows a partial association with the PB component DCP1 both in healthy and infected N. benthamiana tissues. An NMD-mediated regulation of alfalfa mosaic virus infection was proposed by the authors. Infection by another bromovirus, the cucumber mosaic virus, was not affected by the knockdown of AtALKBH9B. Interestingly, AtALKBH9B interacts with alfalfa mosaic virus coat protein whereas there is no interaction with the cucumber mosaic virus coat protein. Therefore, the authors found it possible that AtALKBH9B exerts its regulatory function on virus infection via coat protein interaction.

While many investigations have paid attention to the amplification of vRNA in replication complexes and post-replication fate of vRNA in relation to gene silencing, the virus-host interactions regulating the replicated vRNA through the rest of the cellular RNA regulatory network is a research area that only recently gained attention in plant biology. Viruses have evolved to regulate host RNA metabolism and SGs, PBs, siRNA bodies and other subcellular structures associated with it in order to maintain the cellular environment optimal for robust viral gene expression and avoid the elimination of vRNA. Apart from the core proteins, the protein composition of RNA granules formed by different plant viruses likely varies. Methods based on the enrichment of the RNA granules via centrifugation and subsequent affinity purification via antibodies or specific affinity tags have been developed for plant (Hafrén et al., 2015), animal and yeast (Wheeler et al., 2017) RNA granules. A recent study in which PBs were purified by fluorescence-activated particle sorting from human epithelial cells followed by characterization of the associated mRNAs revealed that PBs mainly accumulate translationally repressed but not decayed mRNAs (Hubstenberger et al., 2017), suggesting a more pronounced role for PBs in accumulating translationally repressed mRNAs and less so in mRNA decay than previously anticipated. This study shows the importance of biochemical purification of the RNA granules in obtaining more detailed information about their protein and RNA composition as well as functions, giving one possible direction were to go to understand also the pro- and antiviral activities associated with RNA granules of plant origin.

Even with the best studied animal viruses there are still substantial knowledge gaps in understanding the viral interactions with RNA granule components as discussed by Lloyd (2016). For example the molecular mechanisms how enteroviral proteases control the translational apparatus, SGs, PBs and the activation of innate immune response by cleaving their key regulatory components have been worked out extensively but many open questions related to the mechanistic roles of these key components in RNA granule assembly and mRNP recruitment as well as in innate immune activation still remain. While many SG, PB, and siRNA body proteins of plant origin are known and some plant viral interactions with them have been described there is much to discover in this insufficiently studied area. Similarly the ways how plant viruses utilize RNA granule components for their benefit, how plants defend themselves against viruses via RNA granules and how viruses counter the RNA granule-associated antiviral actions are likely versatile. As an example of these complex interactions, a recent study indicated the presence of an autophagy-associated cargo adaptor protein NBR1 in potyvirus-induced RNA granules in turnip mosaic virus infection (Hafrén et al., 2017). Bulk autophagy was found essential for plant fitness and virus accumulation, whereas an antiviral role was identified for selective NBR1-mediated autophagy targeting the RNA silencing suppressor HCPro. NBR1 co-localizes together with HCPro and AGO1 to granule structures which proposes that HCPro is likely targeted to the RNA granule context. Interestingly, a specific form of autophagy called granulophagy is targeting both SGs and PBs in yeast (Buchan et al., 2013). Potyvirus-induced granules, which contain both SG and BP markers (Hafrén et al., 2015), were proposed to be targeted by granulophagy (Hafrén et al., 2017). Potyviral proteins VPg and to a lesser extent 6K2 were found to resist autophagy-related HCPro turnover. The authors propose that these complex interactions between potyvirus-induced granules, potyviral proteins and autophagic degradation form another layer to the co-evolutionary arms race between potyviruses and their host plants. This example together with the others discussed in this review show that beyond RNA silencing and its counter defense there is a plethora of interactions connected to post-replication vRNA regulation in plant cells to be discovered during the coming years.

KM: Writing the text; AL: Writing the text, drawing the figures; MP: Writing the text, Compiling the tables.