RNA regulatory processes in RNA virus biology.

Numerous post-transcriptional RNA processes play a major role in regulating the quantity, quality and diversity of gene expression in the cell. These include RNA processing events such as capping, splicing, polyadenylation and modification, but also aspects such as RNA localization, decay, translation, and non-coding RNA-associated regulation. The interface between the transcripts of RNA viruses and the various RNA regulatory processes in the cell, therefore, has high potential to significantly impact virus gene expression, regulation, cytopathology and pathogenesis. Furthermore, understanding RNA biology from the perspective of an RNA virus can shed considerable light on the broad impact of these post-transcriptional processes in cell biology. Thus the goal of this article is to provide an overview of the richness of cellular RNA biology and how RNA viruses use, usurp and/or avoid the associated machinery to impact the outcome of infection. This article is categorized under: RNA in Disease and Development > RNA in Disease.

OVERVIEW OF RNA BIOLOGY AND RNA VIRUSES

Post‐transcriptional processes play a major role in gene expression. No RNA transcript is synthesized in a “ready‐to‐use” fashion by a DNA‐dependent RNA polymerase in the cell. All primary transcripts undergo some form of processing in their maturation. Messenger RNAs made by RNA polymerase II, for example, must be capped at their 5′ end, spliced to remove introns, and cleaved/polyadenylated at their 3′ ends to serve as effective mRNAs. Many transcripts also undergo RNA editing or base methylations to further fine tune their RNA sequence/base composition/structure. Furthermore, RNAs must be properly localized in the cell for effective gene expression. RNA degradation rates play a major role in regulating the levels of specific transcripts. Translation rates are also regulated. To add to this growing complexity, there is a large number of small and long non‐coding RNAs in the cell that undoubtedly also influence aspects of gene expression in a very significant way. Thus to truly understand how gene expression is regulated, one needs to fully consider all aspects of a transcript's post‐transcriptional fate.

Transcripts made by RNA viruses can be significantly impacted by these various aspects of RNA biology that are occurring in the cell. Understanding how viral RNAs interface with and/or avoid these cellular processes can shed considerable light on both virus biology as well as cell biology. Thus it is the goal of this article to provide an overview of how RNA viruses are influenced by RNA regulatory processes.

RNA SPLICING AND RNA VIRUSES

Transcribed pre‐mRNAs require at least three post‐transcriptional modifications to become mature mRNA transcripts capable of being exported from the nucleus and translated into proteins in the cytoplasm. These post‐transcriptional processing events include 5′‐capping, splicing, and 3′‐polyadenylation. Cellular pre‐mRNAs typically contain multiple exons and introns. Splicing is a molecular process whereby introns are removed from the pre‐mRNA, leaving the exons ligated together to form the mature mRNA transcript. To generate diversity in eukaryotes, various exons within the pre‐mRNA can be spliced together in unique combinations, a process termed alternative splicing. Alternative splicing is a general phenomenon in eukaryotic genomes, and ~95% of genes possessing multiple exons in humans are alternatively spliced, creating numerous isoforms of a given protein (Pan, Shai, Lee, Frey, & Blencowe, 2008). Alternative splicing of pre‐mRNA is an important regulatory pathway that contributes to the spatiotemporal control of gene expression and generates functional diversification of proteins (Lopez, 1998).

Alternative RNA splicing and RNA viruses

Unlike most cellular pre‐mRNAs which contain one or more introns, most transcripts generated by RNA viruses possess no introns. Influenza viruses and Borna disease virus are among the rare classes of RNA viruses that generate pre‐mRNAs that undergo splicing (Fournier et al., 2014). RNA viruses that utilize splicing as a component of their strategy of gene expression all have a nuclear aspect to their life cycle as their transcripts must be present in this cellular compartment to usurp the cellular RNA splicing machinery (Chua, Schmid, Perez, Langlois, & Tenoever, 2013; Dubois, Terrier, & Rosa‐Calatrava, 2014; Tsai et al., 2013). The genome of influenza A virus is comprised of eight single‐stranded negative RNA segments that encode 10 main viral proteins and numerous other auxiliary proteins. The two smallest viral RNAs, M and NS, undergo splicing. The M segment can generate the unspliced M1 mRNA transcript and the spliced M2 mRNA transcript (Figure 1a). M1 encodes the M1 matrix protein and M2 an ion channel protein. Alternatively spliced transcripts of the M pre‐mRNA encode a small polypeptide of unknown function (Shih, Suen, Chen, & Chang, 1998), and for some viral strains alternative splicing of M generates variant M42 of the M2 ion channel (Wise et al., 2012). Likewise, the NS segment can generate the unspliced NS1 mRNA transcript and the spliced NS2 mRNA. These transcripts encode two multifunctional proteins. NS1 encodes the nonstructural NS1 protein, which acts (among other things) as an interferon antagonist, countering the cellular antiviral responses. NS2 encodes the NS2/NEP protein, which has dual roles both in viral genome replication and nuclear export of newly synthetized viral RNPs (Fournier et al., 2014). Alternative splicing of the NS RNA generates a transcript encoding a truncated version of NS1 termed NS3. This alternatively spliced transcript has been reported to exist for some viral strains (Selman, Dankar, Forbes, Jia, & Brown, 2012).

Figure 1

RNA viruses can use and/or usurp cellular RNA splicing during infection. Panel a. Two segments of influenza A viral RNA are processed by the cellular RNA splicing machinery to increase the coding capacity of viral genome. Illustrated here is segment 7 which can generate mRNAs that remain unspliced to provide a transcript to generate the viral M1 protein or can be alternatively spliced to yield several other products, including mRNAs to generate the M2 and M42 proteins as illustrated here. Panel b. While no transcripts of the cytoplasmic reoviruses are directly spliced by the cellular machinery, the reovirus T1L strain does generate the μ2 protein which sequesters the alternative splicing factor SRSF2 in nuclear speckles and causes dysregulation of cellular mRNA splicing

The influence of cytoplasmic RNA viruses on nuclear RNA splicing

Mammalian reoviruses are double‐stranded RNA (dsRNA) nonenveloped viruses that replicate in membrane‐associated cytoplasmic viral factories (de Fernández et al., 2014). Reovirus strain Type 1 Lang (T1L) has been shown to repress interferon (IFN)‐β signaling while Type 3 Dearing (T3D) does not (Zurney, Kobayashi, Holm, Dermody, & Sherry, 2009). The different influence of the two strains on IFN‐ β signaling is determined by the M1 gene encoding the reovirus μ2 protein (Irvin et al., 2012). The μ2 protein from the T3D strain contains a serine at position 208 while the T1L strain contains a proline at this position. A minor capsid protein, μ2 protein is highly expressed in infected cells and has RNA binding (Wise et al., 2012) and NTPase activities (Kim, Parker, Murray, & Nibert, 2004; Kobayashi, Ooms, Chappell, & Dermody, 2009). It also determines virus strain‐specific differences in the morphology of viral factories via its ability to bind and stabilize microtubules (Parker, Broering, Kim, Higgins, & Nibert, 2002). μ2 localizes mainly to viral factories upon infection (Broering, Parker, Joyce, Kim, & Nibert, 2002), but it can also be observed diffusely throughout the cytoplasm and nucleus (Mbisa, Becker, Zou, Dermody, & Brown, 2000). Recently it has been shown that μ2 from T1L and reovirus recombinants encoding the T1L μ2 amino acid polymorphism form a complex with the pre‐mRNA splicing factor SRSF2 in nuclear speckles, interchromatin domains that are highly enriched in splicing factors (Figure 1b) (Rivera‐Serrano, Fritch, Scholl, & Sherry, 2017). T1L thus has demonstrated an ability to alter splicing of cellular transcripts. Depletion of SRSF2 also leads to enhanced reovirus replication and cytopathic effects, indicating that T1L μ2 influences splicing of cellular mRNAs in a manner benefitting the virus. This suggests that antagonism of the splicing factor SRSF2 by a cytoplasmic RNA virus has global consequences within the cell.

It is highly possible that other cytoplasmic viruses may influence nuclear RNA splicing events. Alphaviruses, for example, sequester the cellular HuR protein and relocalize it the cytoplasm during infection (Barnhart, Moon, Emch, Wilusz, & Wilusz, 2013; Sokoloski et al., 2010). This results in a disruption of alternative splicing patterns that are normally regulated by the HuR protein (Sokoloski et al., 2010). Thus the impact of RNA viruses on cellular RNA splicing patterns is an interesting and likely highly fruitful area for future study.

POLYADENYLATION AND RNA VIRUSES

Polyadenylation is a two‐step process that occurs co‐transcriptionally on all nonhistone mRNAs in the cell. First, the pre‐mRNA undergoes endonucleolytic cleavage at a site downstream of an AAUAAA hexanucleotide (Shi & Manley, 2015). After cleavage, poly(A) polymerase (PAP) adds adenosine residues to the newly formed 3′ terminus. This addition of a poly(A) tail is facilitated by poly(A) binding protein 2 (PABP2), which increases the affinity between PAP and the RNA, resulting in increased processivity in the enzymatic process (Eckmann, Rammelt, & Wahle, 2011). Global analyses of gene expression have documented the large extent of alternative polyadenylation and its potential role as a major regulator of gene expression (Tian & Manley, 2017). The poly(A) tail functions in RNA export, stability and translation initiation (Curinha, Oliveira Braz, Pereira‐Castro, Cruz, & Moreira, 2014; Fuke & Ohno, 2008). Thus, many RNA viruses have evolved a strategy to put a poly(A) tail on the 3′ end of their transcripts. Key questions surrounding polyadenylation of RNA viruses include: What is the mechanism of polyadenylation, particularly since many RNA viruses are cytoplasmic and cannot effectively usurp the standard nuclear cellular poly(A) machinery; and how do RNA viruses that do not polyadenylate their 3′ ends compensate for the lack of a poly(A) tail?

Generation of poly(a) tails on cytoplasmic viruses

Generally speaking, RNA viruses add a poly(A) tail to their transcripts using their own encoded RNA dependent RNA polymerase (RdRp) by one of two mechanisms. As seen in picornaviruses, viral poly(A) tails can be synthesized using a poly(U) stretch on the 5’end of the template strand (Figure 2a) (Kempf & Barton, 2015; Steil, Kempf, & Barton, 2010). Long poly(U) stretches can also be found at the ends of the template for the synthesis of plus strands in alphaviruses and coronaviruses as well. Alternatively, poly(A) tails can be made by the viral RdRp stuttering over a short U tract (Figure 2b). Good examples of this stuttering mechanism can be seen throughout the negative sense RNA viruses, including rhabdoviruses (Barr, Whelan, & Wertz, 2002), filoviruses (Volchkova, Vorac, Repiquet‐Paire, Lawrence, & Volchkov, 2015), paramyxoviruses (Hausmann, Garcin, Delenda, & Kolakofsky, 1999), and orthomyxoviruses (Poon, Pritlove, Fodor, & Brownlee, 1999). Interestingly, there have been several reports that indicate the polyadenylation event in RNA viruses may be regulated. The usage of the RNA editing site in the GP gene of Ebola viruses appears to be regulated by alternative polyadenylation (Volchkova et al., 2015). Mutational analyses suggest that there is a functional link between 5′ capping and 3′ polyadenylation by the vesicular stomatitis virus (VSV) L protein (Li, Rahmeh, Brusic, & Whelan, 2009). There is also variation in the size of poly(A) tails that could have biological relevance. Among the picornaviruses, encephalomyocarditis virus (EMCV) transcripts have a relatively short ~20 base poly(A) tail as opposed to polioviruses/rhinoviruses which have longer (~60 base) poly(A) tails (Kempf & Barton, 2015). Coronavirus poly(A) tails have also been demonstrated to show small variations in length throughout infection (Wu, Ke, Liao, & Chang, 2013). These differences in the size of the poly(A) tail may reflect differences in viral mRNA translatability or relative stability. Finally, several RNA viruses have been shown to be able to use cellular poly(A) polymerase activities to repair/generate poly(A) tails (Liu et al., 2008).

Figure 2

Three major approaches used by RNA viruses to generate the 3′ ends of viral mRNAs. Panel a. Template‐dependent poly(A) formation. Some viruses (e.g., picornaviruses) possess a poly(U) stretch at the 5′ end of their negative strand RNA template that can be directly copied by the viral RdRp to generate a poly(A) tail on the viral mRNAs. Panel b. Reliance on a stuttering RdRp to generate a poly(A) tail. Some viruses (e.g., rhabdoviruses) contain short U‐rich tracts near the 5′ end of genes on the negative sense strand that serves a slippery site for the viral RdRp to stutter upon, thus generating a poly(A) tail on the positive‐sense mRNA product. Panel c. Termination with a large structure at the 3′ end precludes the need for a poly(A) tail. Finally, other viruses (e.g., flaviviruses) have a large hairpin structure at their 3′ end which serves the purpose of a poly(A) tail to protect the RNAs from 3′‐5′ exonucleases as well as promote translation

Nonpolyadenylated cytoplasmic RNA viruses—alternative strategies for 3′ end formation

RNA viruses that do not generate polyadenylated transcripts still have to ensure that their RNAs are protected from exonucleolytic digestion and function effectively in translation. Host cell histone mRNAs are naturally nonpolyadenylated, thus this is clearly achievable (Romeo & Schümperli, 2016). RNA viruses with nonpolyadenylated 3′ ends (e.g., flaviviruses, bunyaviruses, etc.) all contain large 3′ terminal structures that likely serve as effective structural impediments to 3′‐5′ exonucleases (Figure 2c) (Ibrahim, Wilusz, & Wilusz, 2008). Thus these terminal structures make a significant contribution to viral RNA stability—but how do they also compensate for the lack of poly(A) binding protein on their 3′ ends? In order to mimic the role of the poly(A) tail in translation initiation, the reovirus NSP3 protein has a higher affinity for eIFG4 than cellular PABP2 (Gratia et al., 2015). Transcripts from another family of nonpolyadenylated RNA virus, flaviviruses, contain 3’ UTR elements that coordinate with the 5′ sequences to mediate efficient translation as well as bind PABP2 in a poly(A)‐independent fashion (Chiu, Kinney, & Dreher, 2005; Polacek, Friebe, & Harries, 2009).

The influence of RNA viruses on cellular Polyadenylation and 3′ end processing

Finally, RNA viruses can also have a dramatic impact on cellular polyadenylation. Influenza virus, for example, inhibits pre‐mRNA cleavage and uses its NS1 protein to target the cellular PABP2 factor (Chen, Li, & Krug, 1999; Shimizu, Iguchi, Gomyou, & Ono, 1998). Alphavirus RNAs use a high affinity binding site in their 3’ UTR to bind and cause the relocalization of the predominantly nuclear HuR protein (Barnhart et al., 2013; Dickson et al., 2012). This results in the dysregulation of polyadenylation at sites normally regulated by HuR. Finally, it has been recently demonstrated that VSV infection causes dramatic changes in alternative polyadenylation site usage on cellular mRNAs (Jia et al., 2017). This may have implications regarding cellular innate immune responses to the virus.

CAPPING AND RNA VIRUSES

Given the critical nature of mRNA capping for translation and stability, many RNA viruses have also adapted this RNA modification to enhance their gene expression. Viral RNA caps must be as “host‐like” as possible to avoid degradation of the viral transcripts as well as avert viral clearance from host cells by the immune system due to signaling of the innate immune system by improperly capped transcripts. Interestingly, the routes by which the capping of viral transcripts are achieved are rather dissimilar across all RNA viruses. For example, some viruses have encoded specific enzymes that mimic host capping mechanisms (Ogino & Ogino, 2017), while others have evolved ways to steal caps from cellular mRNAs (Hopkins et al., 2013). It is important to understand the mechanisms/dynamics associated with viral RNA capping as it is an attractive target for development of novel antiviral therapeutics.

Capping of cellular mRNAs occurs co‐transcriptionally on short nascent RNAs in the nucleus, although there have been reports of cytoplasmic recapping of RNAs (Kiss et al., 2016; Moteki & Price, 2002). Transcripts made by RNA Polymerase II are initially generated with a 5′ triphosphate (pppRNA). An RNA triphosphatase activity in the complex of capping enzymes then cleaves the γ‐phosphate on the nascent RNA transcript, leaving a diphosphate intermediate (ppRNA). Following the cleavage, RNA guanylyltransferase (GTase) activity transfers a GMP to the ppRNA to form a G cap. The cap is matured by methylation on the seventh position, forming the m7G “cap 0” structure (Moteki & Price, 2002) and the transcript is subsequently methylated at the 2’ O position of the ribose to form a cap 1 structure.

Capping enzymes of RNA viruses

While most capping enzymes encoded by RNA viruses follow the overall enzymatic blueprint set out by the cellular capping machinery, there is one interesting nuance in some viral enzymes in terms of the source of the phosphates in the final 7mGpppN cap structure. VSV encodes for an enzymatic domain in the multifunctional L protein known as GDP polyribonucleotidyltransferase (PRNTase) which is responsible for generating a fully methylated cap on viral mRNA. In short, PRNTase forms a covalent bond with the viral mRNA via a monophosphate (p) on the 5′ end (as opposed to a ppN 5′ end like the cellular enzyme). This PRNTase‐pRNA intermediate is transferred to GDP (rather than GMP like cellular transcripts). The GDP for capping is generated by the L proteins GTPase activity on GTP. The resulting GpppRNA is further modified, becoming first an m7G cap 0, followed by 2’‐O methylation of the first base (cap 1). Through the use of GDP analogues, the PRNTase domain was shown to recognize the C2‐amino group and either the 2′ or 3′ hydroxyl group of GDP to facilitate this novel mechanism of cap formation (Li, Wang, & Whelan, 2006; Ogino & Ogino, 2017; Ogino, Yadav, & Banerjee, 2010).

In addition to the enzymatic pathway difference noted above, viral capping enzymes in general are structurally different enough from their host cell counterparts that they are an attractive candidate for antiviral development (Kiss et al., 2016). One example of this is with the flavivirus NS5 capping enzyme (Bullard et al., 2015; El Sahili & Lescar, 2017; Han & Lee, 2017). Small molecule inhibitors targeting multiple areas of the NS5 methyltransferase (MTase), including the SAM/SAH binding pocket, cap binding pocket and allosteric regulatory domains have been described [(Benmansour et al., 2017; Brecher et al., 2015; Coutard et al., 2014; Idrus, Tambunan, & Zubaidi, 2012; Lim et al., 2011; Milani et al., 2009); reviewed in (Decroly & Canard, 2017)]. Inhibition of the ZIKV MTase has also been proposed as an antiviral strategy for that emerging pathogen (Coutard et al., 2017). Interestingly, inhibiting viral MTases has shown favorable results in attenuating multiple viruses that have the potential to be used as vaccines (Li et al., 2013; Zhang, Wei, Zhang, Cai, & Niewiesk, 2014).

Viral cap snatching

Rather than encoding for enzymes that build caps for their RNA, some viruses steal a cap from host transcripts in a process known as cap snatching. This strategy is used by several families of RNA viruses that contain segmented genomes, including the Bunyaviridae (e.g., Hantavirus), Arenaviridae (e.g., Lassa fever virus) and the Orthomyxoviridae (e.g., influenza viruses) (Reguera et al., 2016; Rosenthal et al., 2017; Sikora, Rocheleau, Brown, & Pelchat, 2017). The virus encodes for a protein(s) that has an endonuclease domain, such as the L protein of Bunyaviridae or the heterotrimer of PA, PB1, and PB2 of influenza virus (with the endonuclease activity in the PA subunit). This viral endoribonuclease cuts close to the 5′ cap of host RNA transcripts, most commonly occurring ~15 nucleotides downstream of the cap, but the cut site can vary by 10 to 20 nucleotides. This capped RNA fragment then acts as a primer for transcription (Cheng & Mir, 2012; Dias et al., 2009; Reguera, Weber, & Cusack, 2010).

As opposed to most RNA viruses, influenza A virus (IAV) replicates in the nucleus. Interestingly, high throughput sequencing of the 5′ end of IAV transcripts has revealed that IAV preferentially steals caps from noncoding RNAs, specifically small nuclear RNAs (snRNA) and the promoter associated capped small RNAs (csRNA) (Gu et al., 2015). In addition to preferentially cap snatching from snRNAs, there was a bias towards stealing caps from the splicing‐associated U1 and U2 snRNAs (Gu et al., 2015). It will be interesting to see in the future how modulation of the levels of these noncoding RNAs plays a role in IAV infection.

Bunyaviruses replicate in the cytoplasm and must actively compete with host mRNA decapping machinery to find and steal 5′ caps from cellular mRNAs (Cheng & Mir, 2012; Reguera et al., 2010). These viruses use the viral nucleocapsid (N) protein to find and direct capped host transcripts into cellular P bodies until used by the viral RNA dependent RNA polymerase (RdRp). The preferential mRNAs that are targeted by bunyaviruses appear to be cell cycle‐associated mRNAs that have been marked for destruction by the host decapping enzyme Dcp2. Knock down or over‐expression of the cellular Dcp2 enzyme leads to either increased or decreased bunyavirus replication respectively (Hopkins et al., 2013). Phosphorylation of Dcp2 can also upregulate enzymatic activity and further inhibit bunyavirus replication. These studies clearly indicate that capping/decapping is an important front in the molecular arms race for defense against certain viral infections.

Host recognition of viral caps/5′ ends

Nuances in the 5′ ends/caps of viral RNAs are also used by cellular innate immune mechanisms to target and eliminate virally‐derived nucleic acid in the cell. While the presence of a 5′ cap effectively masks an RNA from recognition as foreign, viral capping may not be 100% efficient and viral transcripts often have structured ends which can be detected by the cellular nucleic acid surveillance machinery. One of the best characterized of these pathways involves RIG‐I, a protein which uses its C terminal domain to recognize viral RNA via either the presence of 5′‐triphosphate on nascent transcripts or blunt‐ended dsRNA (Domain et al., 2010; Wang et al., 2010). Upon recognition of viral 5′ ends with “noncellular” structural configurations, RIG‐I becomes activated via a conformational change and signals the cell to begin interferon production. A good example of how the terminal structural features of viral RNAs can influence RIG‐I can be found in influenza viruses where the 5′ and 3′ ends of individual transcripts form a partial duplex (known as a panhandle) with each other. RIG‐I has a very high affinity for this structural feature and thus the panhandle effectively stimulates interferon production even though influenza virus does not have a 5’‐PPP attached to the blunt‐ended dsRNA (Lee, Kim, et al., 2016).

It was previously believed that to protect themselves from recognition by RIG‐I, RNA viruses could merely attach a m7G “cap 0” structure to their RNAs. However, recent research has illustrated that this cap 0 structure alone does not prevent recognition by the cellular innate immune system. Surprisingly, RNAs with a cap 0 structure have a similar ability to activate RIG‐I as does an RNA containing a 5′ triphosphate. However, RNAs that contain a cap 1 structure, one in which the first base of the transcript also contains a 2’‐O ribose methylation, do not activate RIG‐I (Devarkar et al., 2016). The presence or absence of this 2’‐O‐methylation cap 1 on transcripts is also monitored by the MDA5 protein which also triggers an interferon response—primarily mediated by IFIT proteins, when this RNA modification is missing (Daffis et al., 2010; Lee, Kim, et al., 2016; Schlee et al., 2009; Züst et al., 2011). MDA5 activation involves a conformational change and induction of tetramer formation, similar to RIG‐I. To avoid detection by this cellular surveillance mechanisms, viruses encode specific 2’‐O‐methyltransferases (MTases) that are conserved in virus families, such as flaviviruses and recently roniviruses (Egloff, Benarroch, Selisko, Romette, & Canard, 2002; Zeng et al., 2016). Further research into the structural features of these MTases may reveal characteristics about the enzyme domains allowing for novel antiviral therapies that can target multiple viruses in a single family (Coutard et al., 2017).

RNA METHYLATION AND RNA VIRUSES

N6‐methyladenosine (m6A) methylation is one of the most abundant modifications that occurs to RNA in the cell (Fu, Dominissini, Rechavi, & He, 2014). m6A methylation has effects on both structural and functional roles of RNA. The methylation of RNA is a dynamic process that consists of proteins known as “writers,” “erasers,” and “readers.” Writers are responsible for methylation, erasers are responsible for de‐methylation, and readers serve as effector proteins in recognizing the m6A methylation for biological function. Mapping of the transcriptome for m6A methylation has demonstrated that the modification often occurs at preferential locations that can also be well‐conserved. Thus the importance of this RNA modification for the regulation of cellular gene expression is becoming clearer (Dominissini et al., 2012; Schwartz et al., 2012). Interestingly, m6A methylation has recently been shown to occur on viral transcripts and has the potential to induce antiviral or proviral effects on viral replication. Furthermore, the studies of m6A writer, eraser, and reader proteins on cytoplasmic RNA viruses have helped to confirm the presence of these proteins in the cytoplasm (Gokhale et al., 2016), which was suggested by previous work (Chen et al., 2015; Lin, Choe, Du, Triboulet, & Gregory, 2016).

Antiviral effects of m6A methylation

Knock down of the m6A writer proteins (methyltransferases) METTL3 and METTL 14 increases the replication of hepatitis C virus (HCV) (Gokhale et al., 2016). On the other hand, when the eraser protein (de‐methyltransferase) FTO was knocked down, viral replication decreased. Knock down of m6A reader proteins, the YTH domain family proteins (YTHDF), increased viral replication late in infection. The YTHDF proteins were also shown to have competitive, antagonistic effects against the HCV core protein, which is responsible for viral packaging (Gokhale et al., 2016). Similar antiviral effects of m6A modification on viral RNAs were recently duplicated in another member of the Flaviviridae, Zika virus (ZIKV) (Lichinchi et al., 2016). These results strongly suggest that tagging of viral RNAs with an m6A modification may be part of the cellular arsenal of strategies to minimize the impact of viral infection.

Proviral effects of m6A methylation

The role of m6A modifications in RNA virus infections, however, does not appear to be straightforward. Mapping of m6A modification sites across HCV and ZIKV genomes showed highly conserved targeted sequences (Gokhale et al., 2016; Lichinchi et al., 2016). Given the capability of RNA viruses to evolve quickly under selective pressures, it seems unusual that they would maintain these m6A methylation targets if the modification solely led to antiviral effects. RNA viruses may maintain these regions to prevent activation of the immune system via toll‐like receptors and RIG‐I (Durbin, Wang, Marcotrigiano, & Gehrke, 2016; Karikó, Buckstein, Ni, & Weissman, 2005). It should also be noted that host cells appear to be able to sense and react to viral infection by changing m6A methylation status of cellular transcripts (Lichinchi et al., 2016). In the case of HCV, m6A methylations may actually increase viral fitness by slowing replication and facilitating the characteristic slow, persistent infection strategy favored by the virus (Gokhale et al., 2016). Finally, the patterns of m6A vary between different ZIKV lineages (Lichinchi et al., 2016). As different lineages provide varying levels of pathogenicity (Weaver et al., 2016), regulation of m6A could theoretically be contributing to these differences. Clearly, further research into the role of m6A in cellular and viral gene regulation is needed to define this interesting new aspect of host‐virus RNA interplay.

RNA EDITING ON RNA VIRUS TRANSCRIPTS

Viral genomes display some of the highest genomic mutational frequencies known. Mutational patterns in viral genomes do not occur randomly, rather certain nucleotide mutations always occur more frequently that others, a phenomenon known as directional mutational pressure. The source of such unequal viral mutation rates is largely due to error‐prone polymerases. RNA editing can be simply defined as the addition or substitution of RNA bases that were not originally encoded by the genome. RNA editing in paramyxoviruses, for example, occurs co‐transcriptionally when the RNA‐dependent‐RNA polymerase (RdRp) interacts with a cis‐acting sequence element (3’‐UAAUUUUUUCCC) in the genome that causes a stuttering of the polymerase (Figure 3a). This “stuttering” mechanism is responsible for editing the RNA via the addition of up to 6 G's at a single RNA editing site (Iseni et al., 2002). Interestingly, RNA editing could also be done via cellular enzymes in a completely post‐transcriptional process.

Figure 3

RNA editing of RNA virus transcripts. Panel a. Select mRNAs generated by paramyxoviruses and filoviruses can contain an additional one to ~8 G residues due to slippage/stuttering of the viral RdRp in certain regions during transcription. Panel b. The cell possesses inducible RNA deaminases such as ADAR that can deaminate adenosine residues to inosines. These enzymes can increase the mutation rate of RNA viruses and thus can affect viral fitness and evolution

One cellular means of carrying out RNA editing is via the double‐stranded RNA‐specific adenosine deaminase (ADAR) enzymes (Figure 3b) (George, John, & Samuel, 2014). Humans possess three major ADAR isoforms: ADAR1‐p150, ADAR1‐p110, and ADAR2. ADAR enzymes facilitate adenosine deamination to inosine, which is biologically relevant since inosine behaves like guanosine and can base‐pair with cysteine. During viral replication, the cysteine‐inosine base‐pairing will lead to guanosine being incorporated in place of inosine, ultimately leading to adenosine to guanosine transition mutations. The expression level of ADAR often increases as the cell detects increasing levels of foreign RNA. Interestingly, several viruses of the family Flaviviridae, including Hepatitis C virus, Bovine viral diarrhea virus, and Dengue virus, can stimulate ADAR expression, while many other virus families do not (Khrustalev, Khrustaleva, Sharma, & Giri, 2017). The consequence of ADAR editing of viral RNA is an increase in A to G transition rates (Liu et al., 2015). Studies on human metapneumovirus suggest that high levels of viral RNA editing may lead to the production of increased levels of defective interfering particles (van den Hoogen et al., 2014). The increased proportion of G's that results from ADAR editing might also have a positive effect on virus biology, enabling, for example, viral RNAs to form stronger or novel secondary structures given the higher stability of G‐C base pairing. Finally, RNA editing strategies like ADAR enzymes do not act uniformly along the length of viral RNA genome, thus directional mutational pressure and subsequently genomic mutational rates are not constant. An emerging hypothesis is that stalling of either the viral RdRp or cellular ribosomes on viral RNAs is associated with mutational pressure, perhaps by enabling RNA editing enzymes like ADAR to act on viral RNAs (Khrustalev et al., 2017).

RNA NUCLEAR EXPORT AND RNA VIRUSES

Export of RNA from the nucleus is a key step in mRNA synthesis. In order to be exported, mRNA generally must be fully processed and packaged it into a messenger ribonucleoprotein (mRNPs) (Bindereif & Green, 1986). The majority of mRNA export is mediated by the nonkaryopherin heterodimer Nxf1 and Nxt1, although some are exported by the karyopherin chromosome region maintenance 1 (CRM1) (Aibara, Katahira, Valkov, & Stewart, 2015; Brownawell & Macara, 2002; Delaleau & Borden, 2015). The mRNA is trafficked through the nuclear pore complexes (NPCs) to the cytoplasm (Allen, Cronshaw, Bagley, & Goldberg, 2000). For the small number of RNA viruses that transcribe in the nucleus, viral mRNA export is best characterized for influenza virus.

Influenza virus transcripts use both the Nxf1‐Nxt1 heterodimer and CRM1 pathways to move from the nucleus into the cytoplasm. Interestingly, the virus appears to have multiple, perhaps even redundant strategies to ensure the effective export of its transcripts to the cytoplasm. Influenza virus manipulates the host Nxf1 factor to transport a defined subset of its mRNAs to the cytoplasm (Larsen et al., 2014). NP, NA and HA mRNAs use Nxf1, but PA, PB1 and PB2 mRNAs do not. Late in infection, influenza virus activates caspase activity which causes enlargement of nuclear pores. This permits the passive transport of the viral ribonucleoprotein (vRNP) via the CRM1 export pathway (Mühlbauer et al., 2015). The influenza NS2 protein also has been reported to interact with CRM1, perhaps further enhancing nuclear export of viral mRNAs (Elton et al., 2001; Wang, Zhou, & Du, 2014). In addition, the viral MI protein has also been implicated in nuclear export of vRNPs through CRM1 (Brunotte et al., 2014). Recently, the MI and PB2 viral proteins have been implicated in the subnuclear movement of viral mRNAs, perhaps as a precursor to effective viral RNA export (Ando et al., 2016). Nucleolin also influences the subnuclear trafficking of vRNPs (Terrier et al., 2016). Influenza virus also recruits/hijacks the DExD/H‐box RNA helicase DDX19, that remodel viral mRNA and perhaps also contribute to efficient export (Diot et al., 2016; Fuller‐Pace, 2006).

Disruption of nuclear RNA export by cytoplasmic RNA viruses

Cytoplasmic RNA viruses can also take steps to block the export of cellular mRNAs from the nucleus. The nonstructural protein 5 (NS5) of dengue virus and other flaviviruses is shuttled between the cytoplasm and nucleus where it may influence CRM‐mediated nuclear export (Rawlinson, Pryor, Wright, & Jans, 2009). Porcine reproductive and respiratory syndrome virus (PRRSV) uses its nsp1β protein to induce accumulation of host mRNAs in the nucleus during infection (Han, Ke, Zhang, & Yoo, 2017). Expression of the NSS gene of the bunyavirus Rift Valley Fever Virus also induces nuclear accumulation of cellular mRNAs (Copeland, Van Deusen, & Schmaljohn, 2015). 2A protease encoded by enteroviruses can cleave components of the nuclear pore complex, altering RNA export kinetics in infected cells (Park, Schweers, & Gustin, 2015).

Given the importance of viral RNP nuclear export, this process has recently received attention as a possible target for antiviral therapies (Kakisaka et al., 2015). While targeting host proteins is in some ways attractive since there is less probability of viral resistance, this therapeutic avenue should be pursued with caution due to potential consequences on normal host cell functions.

TRANSLATION AND RNA VIRUSES

Translation initiation in eukaryotic cells is a highly regulated process (Aitken & Lorsch, 2012). It involves 12 or more initiation factor proteins to direct the ribosomal subunits and Met‐tRNAi Met over the AUG start codon. Key initial steps include the formation of a preinitiation complex (PIC) combined with the formation of the ternary complex (TC) (eIF2, GTP, and Met‐tRNAi Met). The PIC then interacts with the 40S ribosomal subunit and scans the mRNA for the AUG start codon. Once the AUG is found, several initiation factors are released, allowing for the binding of the 60S ribosomal subunit, forming the 80S complex that initiates translation. Not surprisingly, RNA viruses have evolved ways to manipulate this process to favor translation of their mRNAs.

Viral IRES elements

Some viral genomes encode an alternative translation initiation mechanism known as internal ribosome entry sites (IRES) that bypass the need for 5′ cap recognition by translation factors. These elements were found in the 1980's, first in picornaviruses followed by Hepatitis C Virus (HCV) (Pelletier & Sonenberg, 1988; Pestova, Shatsky, Fletcher, Jackson, & Hellen, 1998). IRES elements are located in the 5’ UTRs of transcripts and form extensive secondary and tertiary structures that recruit the host translational machinery. IRES elements can be classified into three types based upon their secondary structure and initiation factors that they selectively recruit (Pacheco, Serrano, & Fernandez, 2008).

We'll focus on the type III IRES located in the HCV positive strand RNA to illustrate the state of the art in our mechanistic understanding of how these elements function (Jaafar, Oguro, Nakamura, & Kieft, 2016). As outlined in Figure 4, the HCV IRES recruits an intermediate of the PIC that includes eIF1, eIF1A, and eIF3 and forms in the absence of the ternary complex (TC). Binding to the IRES displaces eIF1 from the PIC and leads to one of two pathways to initiate translation depending on eIF2 availability in the cell. If eIF2 is in an active state, the TC is transferred to the IRES‐PIC complex, causing the IRES to undergo a conformational change. eIF2 would then hydrolyze GTP and leave, followed by binding of eIF5B to the complex. In an eIF2 inactive state, it is proposed that Met‐tRNAi Met can bind either directly without the need of a delivery factor and then is stabilized in the complex by eIF1A and eIF5B, or is delivered to the complex by either eIF1A or eIF5B. After Met‐tRNAi Met binding occurs in either of the two pathways, eIF1A and eIF5B proceed to then assist in the final codon‐anti‐codon verification and ribosomal subunit joining steps (Jaafar et al., 2016).

Figure 4

An overview of HCV IRES‐dependent translation. The HCV IRES is a classified as “Type III” and recruits a preinitiation complex (PIC) that includes eIF1, eIF1A, and eIF3. Binding to the HCV IRES causes displacement of eIF1. Translation initiation then proceeds by one of two pathways depending on the state (active or inactive) of eIF2 in the cell. For more information, see the accompanying text in section 8.1

IRES elements and associated structures have the capability of being targeted by antivirals. A recent study illustrated the potential of siRNA as an antiviral therapeutic against HCV. The capacity of the HCV IRES to form secondary and tertiary structures can naturally preclude siRNAs from binding to many target sequences. Using an siRNA tiling technique, a handful of siRNAs were found to be effective in disrupting the HCV IRES and HCV replication (Moon, Lee, Kim, Cho, & Lee, 2016).

NonIRES‐mediated cap independent mechanisms of viral translation

Curiously, some viruses are translated in a cap‐independent manner that does not involve an IRES element. Notably, members of the norovirus family have a cap‐like protein known as VPg that interacts with the HEAT‐1 domain of the initiation factor eIF4G via its C‐terminal region. VPg, however, is also still capable of binding to eIF4E and it is proposed that this interaction potentially regulates host response to the viral infection (Chung et al., 2014; Leen, Sorgeloos, Correia, & Chaudhry, 2016). Norovirus‐encoded protease targets several translation factors for decay, thus tipping the balance of translation initiation to favor viral mRNAs (Emmott, Sorgeloos, Caddy, & Heesom, 2017). Because of the conserved nature of using the HEAT‐1 domain for binding by VPg in the norovirus family (Leen et al., 2016), small molecules could be selected to disrupt the VPg‐eIF4G complex as possible antivirals.

Virus manipulation of stress granules

In an effort to reduce viral translation, the cell induces an innate immunity pathway triggered by dsRNA leading to the formation of stress granules (SG). SG are membrane‐less aggregates of stalled translation preinitiation complexes that form in the cytoplasm (Protter & Parker, 2016). To maintain the highest fidelity of viral translation, viruses have developed methods to interfere with this pathway. Some viruses, such as Ebola virus, have encoded proteins (VP35) to interfere and inhibit the function of G3BP1, a host protein that helps drive SG formation (Le Sage et al., 2017). Other viruses, such as Semliki Forest virus, sequester G3BP1 via nsP3 to its replication complex, preventing SG formation and enhancing viral replication (Panas et al., 2012). Additional viruses, such as flaviviruses, have been shown to modulate the translational landscape of the cell and prevent SG formation even though the exact mechanism has yet to be revealed (Haneke, Lohmann, Bartenschlager, & Fackler, 2017). The commonality of SG disruption in RNA virus infections, as well as the plethora of mechanisms through which SG formation is modulated, indicate its strategic importance for maintaining efficient viral replication and translation. Interestingly, not all RNA viruses target SGs for disruption. Rabies virus does not have mechanisms to modulate SG assembly and possibly uses them to maintain a healthy equilibrium of mRNA/translation for viral replication (Blondel, Nikolic, & Civas, 2016).

miRNA regulation of RNA virus translation

MicroRNAs (miRNAs) are 21–23 nucleotide long RNAs that have the capacity to control many cellular processes and are commonly produced in many eukaryotic organisms. The function of miRNAs relies on their ability to either degrade mRNA and/or inhibit translation. Interestingly, cellular miRNAs can sometimes effectively target viral RNA genomes. Thus several RNA viruses have evolved synonymous mutations (Enterovirus 71) (Zheng et al., 2013) or deletions (neurotropic flaviviruses) (Heiss, Maximova, Thach, Speicher, & Pletnev, 2012) to avoid recognition and repression of viral gene expression by host cell miRNA. In addition to blocking translation, miRNAs can at times also promote gene expression (Vasudevan, 2012). It has been shown that miR‐122 is responsible for stabilization of HCV RNA as well as influencing the balance between how much RNA is translated and how much is involved in replication synthesis (Masaki et al., 2015; Mengardi et al., 2017). MicroRNAs let‐7 and miR‐17 are also responsible for enhanced translation in bovine viral diarrhea virus (BVDV) (Scheel et al., 2016).

RNA STABILITY/DEGRADATION AND RNA VIRUSES

RNA degradation is a tightly controlled and regulated process that, in coordination with transcription, is essential for maintaining the homeostasis of cellular gene expression (Schoenberg & Maquat, 2012). Messenger RNA decay is largely initiated by removal of the poly(A) tail (deadenylation) followed by degradation of the body of the transcript by one of two exonucleolytic pathways. The 5′ to 3’ RNA degradation pathway is initiated via decapping proteins responsible for removing the 5′ cap structure on mRNA transcripts (Hsu & Stevens, 1993). The RNA exosome is a cellular machine that is largely responsible for the 3′ to 5′ decay pathway (Chekanova et al., 2007). Distinct forms of the RNA exosome exist in the cytoplasm and nucleus, associating with specific protein cofactors unique to each subcellular compartment (Staals et al., 2010). Numerous RNA binding proteins have been implicated in regulating RNA decay, including HuR (Grammatikakis, Abdelmohsen, & Gorospe, 2017), TTP (Wells, Perera, & Blackshear, 2017), AUF1 (White, Matsangos, & Wilson, 2017), and others. These proteins can target mRNAs to multiple degradation pathways. Finally, inducible ribonucleases and the RNAi machinery all feed their products into cellular RNA decay pathways (Drappier & Michiels, 2015).

RNA decay plays a major role in controlling both the quantity and quality of RNA transcripts in the cytoplasm. Thus it is not surprising that RNA viruses have developed ways to successfully interface with the cellular RNA decay machinery (Figure 5). The goal of this section is to provide an overview of some this interplay.

Figure 5

Major approaches used by RNA viruses to confound the cellular RNA decay machinery. RNA viruses effectively subvert the cellular RNA decay machinery by a variety of mechanisms. Top Panel: RNA‐mediated mechanisms include the use of high affinity binding sites to sequester and usurp cellular RNA stabilizing factors (e.g., alphaviruses) as well as the use of complex RNA structures to prevent binding or repress RNA exonucleases (e.g., flaviviruses). Bottom Panel: RNA viruses (e.g., polioviruses) can also encode proteases that can selectively cleave cellular factors involved in RNA decay. Alternatively RNA viruses like coronaviruses encode nonstructural proteins that facilitate massive degradation of cellular RNA that somehow gives viral RNAs a competitive advantage

Many viruses target the cellular regulators of RNA decay to successfully navigate around the cellular RNA decay machinery. All alphaviruses contain a high affinity binding site for the cellular HuR protein in their 3’ UTR (Barnhart et al., 2013; Sokoloski et al., 2010). HuR binds and stabilizes viral transcripts—and the sequestration of a large amount of HuR by alphavirus RNAs causes a dysregulation of mRNAs normally influenced by binding of this RNA stability factor (Sokoloski et al., 2010). Polioviruses target and cleave the cellular RNA regulatory factor AUF‐1 (Ullmer & Semler, 2016).

RNA viruses must also successfully evade specialized RNA quality control pathways. UPF1, Smg5 and Smg 7, key components of the nonsense‐mediated RNA decay pathway, are clear restriction factors for numerous RNA viruses (Balistreri, Bognanni, & Mühlemann, 2017). Finally, bringing in their own exonuclease is another strategy employed by RNA viruses. Coronaviruses encode the nsp1 protein and a highly active endoribonuclease (Endo U; nsp15) that essentially reprograms RNA decay in the cytoplasm of infected cells to allow for downregulation of cellular defense mechanisms and effective viral replication (Kindler et al., 2017). This might simply shut down quality control pathways in the cell due to the plethora of new decay products that overwhelms the machinery.

Direct targeting of the major RNA decay machinery by RNA viruses

Several RNA viruses are capable of stalling and sequestering the 5′ to 3′ exonuclease XRN1. Flaviviruses are single‐stranded positive‐sense RNA viruses that are responsible for a number of serious human diseases (Moon et al., 2012; Moon et al., 2015). A unique feature of all flaviviruses studied thus far is the formation of small subgenomic flavivirus RNA (sfRNA) representing the 3’ UTR of the viral genomic RNA by stalling of the cytoplasmic 5′‐3′ exoribonuclease XRN1 (Pijlman et al., 2008). XRN1 becomes stalled and ultimately sequestered on flavivirus UTRs when it encounters a knot‐like three helix junction structure (Akiyama et al., 2016; Chapman et al., 2014; Chapman, Moon, Wilusz, & Kieft, 2014; Funk et al., 2010; Silva, Pereira, Dalebout, Spaan, & Bredenbeek, 2010). Stalling of XRN1 on viral UTRs leads to the functional repression of the enzyme—which surprisingly also downregulates decapping and deadenylation activities in infected cells (Moon et al., 2012; Moon, Blackinton, et al., 2015). This results in the up‐regulation of numerous short‐lived cellular mRNAs which may have a significant impact on viral cytopathology and pathogenesis (Moon et al., 2012; Moon, Blackinton, et al., 2015). Other virus families may also target XRN1 and other components of the 5′‐3′ pathway to confound the RNA decay machinery. XRN1 and DCP1a/2, for example, are degraded during a poliovirus infection (Dougherty, White, & Lloyd, 2011).

A variety of cellular factors, including ZAP (Guo, Carroll, Macdonald, Goff, & Gao, 2004), may recruit the exosome to RNA viruses. The Trf–Air–Mtr4 polyadenylation (TRAMP) complex consists of a set of cofactors that aid in degrading targeted RNA substrates via the nuclear RNA exosome (LaCava et al., 2005). It was recently shown that the TRAMP complex participates in viral defense within the cell's cytoplasm (Molleston et al., 2016). The underlying mechanisms of TRAMP complex component export to the cytoplasm and subsequent targeting of viral 3′ UTRs have yet to be elucidated.

CYTOPLASMIC RNA GRANULES AND RNA VIRUS INFECTION

Under conditions of stress host cells will increase the formation two types of cytoplasmic granules: processing bodies (P bodies) and stress granules. These are membrane‐less, dynamic structures that may be the result of liquid phase transitions induced by disordered regions of RNA binding proteins (Protter & Parker, 2016). P bodies are cytoplasm foci that accumulate translationally repressed mRNPs along with multiple proteins/enzymes involved in mRNA decay (Emara & Brinton, 2007; Shah, Zhang, Ramachandran, & Herman, 2013; Ward et al., 2011; Yang, Yu, Gulick, Bloch, & Bloch, 2006). While they may not necessarily be sites of active RNA decay (Hu, Sweet, Chamnongpol, Baker, & Coller, 2009), P bodies are associated with a role in mRNA surveillance and decay (Lavut & Raveh, 2012). Stress granules are most likely involved in storing mRNAs that were stalled in translation initiation and it is suggested that this aids in protection of genomic information during stress (Frydryskova et al., 2016; Shah et al., 2013; Ward et al., 2011). As outlined below, viral infections can significantly influence P body and stress granule formation.

The role of cellular P body formation and function has been explored for a number of RNA viruses. In general, it appears that RNA viruses may initially promote/induce the formation of P bodies to benefit viral gene expression and then repress P body formation at later times post infection either directly or indirectly. This duality of effects of viral infections on P bodies can make elucidating their overall impact on virus biology difficult. P body formation is significantly affected in HCV infection, however the impact of P bodies on HCV replication efficiency is not clear (Pérez‐Vilaró, Scheller, Saludes, & Díez, 2012). Depletion of select P‐body proteins decreases HCV gene expression (Pager, Schütz, Abraham, Luo, & Sarnow, 2013), and West Nile virus (WNV) and dengue virus (DENV) flaviviruses have been shown to enhance their replication by interacting with P bodies (Chahar, Chen, & Manjunath, 2013; Ward et al., 2011). However, late in infection, flaviviruses appear to inhibit P body assembly (Emara & Brinton, 2007). Influenza A virus NS1 protein interacts with RAP55‐containing P bodies, restricting the accumulation of NP proteins to P bodies and promotes viral replication (Mok et al., 2012; Yang et al., 2006). Polioviruses use proteases and other virally‐encoded proteins to disperse P bodies late in infection (Dougherty, Tsai, & Lloyd, 2015). P bodies are also dispersed during infection by the dsRNA‐containing rotaviruses (Bhowmick, Mukherjee, Patra, & Chawla‐Sarkar, 2015).

Two major stress granule proteins, TIA‐1 and G3BP1, are major targets of RNA viruses during infection. West Nile and dengue viral RNAs sequester TIA‐1 and interfere with stress granule formation during infection (Mok et al., 2012). RNAs from another flavivirus, tick‐borne encephalitis virus, also binds TIA‐1 to regulate translation from viral RNAs with minimal effects on overall G3BP1‐containing stress granule formation (Albornoz, Carletti, Corazza, & Marcello, 2014). Rabies virus infections induce stress granule formation near viral factories in Negri bodies which appears to be have a weak antiviral effect as TIA‐1 can block viral translation (Blondel et al., 2016). Overall, G3BP1‐enriched stress granules are induced during some RNA viral infections with both positive and negative effects on viral replication (Albornoz et al., 2014; Amorim, Temzi, Griffin, & Mouland, 2017; Bhowmick et al., 2015; Chahar et al., 2013; Courtney, Scherbik, Stockman, & Brinton, 2012; Dougherty et al., 2015; Frydryskova et al., 2016; Hou et al., 2017; Hu et al., 2009; Kim et al., 2016; Lavut & Raveh, 2012; Mok et al., 2012; Nelson et al., 2016; Pager et al., 2013; Panas et al., 2012; Pérez‐Vilaró et al., 2012; Rabouw et al., 2016; Sage et al., 2016; Scholte et al., 2015; Xia, Chen, Xu, et al., 2015; Zhou et al., 2017). Taken together these data show the need to effectively address the knowledge gap of the impact of stress granules on RNA virus infections.

NON‐CODING RNA REGULATION AND RNA VIRUSES

Non‐coding RNAs (ncRNAs) are a large class of transcripts that are not used as templates for productive translation and represent the vast majority of RNAs in cells (Fortes & Morris, 2016). RNAs such as ribosomal RNAs (rRNA), transfer RNAs (tRNA), microRNAs (miRNA), small nuclear RNAs (snRNA), small nucleolar RNAs (snoRNA) and long ncRNA (lncRNA) are among the members of this large and diverse family of non‐coding transcripts. They have a plethora of roles in cell biology, however much remains to be discovered about these transcripts, particularly in terms of lncRNA function. LncRNAs are classified as non‐coding transcripts longer than 200 nucleotides and they often interact with proteins, DNA and/or other RNAs to aid in the regulation of gene expression (Fortes & Morris, 2016).

RNA interference pathways and RNA viruses

Small RNAs such as miRNAs (~22 nucleotides) or PIWI‐associated RNAs (piRNAs) (~24–30 bases) associate with proteins of the Argonaute family and play important roles in RNA silencing and post‐transcriptional regulation (Ghosh, Mallick, & Chakrabarti, 2009). The ability of the RNA interference pathway, including the generation of small interfering (si)RNAs and microRNAs, to silence viral RNA is a major part of antiviral innate defense mechanisms in plants and insects (Blair & Olson, 2015; Szittya & Burgyán, 2013). The relevance of RNAi to antiviral responses in mammals is unclear (Cullen, Cherry, & tenOever, 2013; Li et al., 2016). Many viruses generate suppressors of the cellular RNAi machinery which function through interactions with RNAi components or by sequestering viral double‐stranded RNA (Figure 6) (Bivalkar‐Mehla et al., 2011). Interestingly, flaviviruses produce subgenomic flavivirus (sf)RNAs that can interact with cellular RNAi machinery of mosquitoes (Goertz et al., 2016; Gokhale & Horner, 2017; Moon et al., 2015; Schnettler et al., 2012). The strategy of using non‐coding RNAs to evade the RNAi response is not limited to flaviviruses. Semliki Forest virus appears to produce a viral RNA decoy to aid in evading the RNAi response (Siu et al., 2011). Viral proteins are also generated that interfere with the RNAi machinery. The NS4B and NS3 proteins of dengue virus interfere with RNAi pathways in insect cells (Kakumani et al., 2013; Kakumani et al., 2015). Several plant and animal viruses generate an RNAse III‐like enzyme that inhibits both the siRNA and miRNA pathways (Weinheimer et al., 2015). The NS1 protein of influenza virus can suppress the cellular RNAi machinery (Li, Basavappa, et al., 2016). Additionally, Ebola virus proteins VP30 and VP35 interfere with RNAi pathways (Fabozzi, Nabel, Dolan, & Sullivan, 2011; Pleet, DeMarino, Lepene, Aman, & Kashanchi, 2017). Clearly the interaction of RNA viruses with the cellular RNAi pathways represents an excellent example of the molecular arms race that has evolved between host and pathogen.

Figure 6

Examples of mechanisms RNA viruses use to suppress RNA interference. RNA viruses target various components of the host cell RNAi machinery as part of the molecular arms race to effectively establish infection. Double‐stranded RNA (dsRNA) is sequestered (e.g., by Ebola VP35 or Influenza A virus NS1) or degraded (e.g., by virally‐encoded RNase III‐like enzymes) to inhibit the RNAi machinery from detecting this key substrate. Cellular enzymes like dicer or argonaute are bound and repressed by viral proteins (e.g., Ebola VP30 or Dengue virus NS4B) or viral RNAs (e.g., flavivirus sfRNA). Finally, DENV NS3 protein interacts with hsc70, impacting the assembly of the RISC effector complex

Long non‐coding RNAs and RNA viruses

Our understanding of the generation and role of small and long‐non‐coding RNAs in RNA virus infections is currently rapidly developing (Li, Weng, Shih, & Brewer, 2016). Changes in the expression of cellular lncRNAs, for example, have been associated with flavivirus infections. DENV infection of mosquitoes causes an increase in cellular lncRNA production, some of which are associated with suppression of viral replication (Etebari, Asad, Zhang, & Asgari, 2016). Interestingly, RNA viruses can also manipulate the expression of cellular lncRNAs to favor their replication. For example, HCV infections induce production of eosinophil granule ontogeny transcript (EGOT) and other cellular lncRNAs that function as negative regulators of the interferon response (Carnero et al., 2016; Nishitsuji et al., 2016). In addition, influenza A virus can also induce the host VIN lncRNA that enhances viral replication (Winterling et al., 2014). Infection of mice with a variety of RNA viruses, including Japanese encephalitis virus, rabies virus and SARS‐CoV also results in induction of lncRNA, although the impact of these lncRNA on the viral infection remains unexplored (Peng et al., 2010). Clearly the interface of RNA viruses with cellular lncRNAs is a fertile area for future research.

RNA viruses also generate their own lncRNAs that assist in replication and virus‐host interactions. Antigenome replicative intermediates that are generated as part of the replication pathway of RNA viruses are an excellent example of a non‐coding transcript that plays a pivotal role in viral infections. As discussed above, arthropod‐borne flaviviruses generate a ~400–500 base non‐coding RNA from their 3′ untranslated regions that plays a role in RNAi suppression as well as modulating the cellular RNA decay machinery (Filomatori, Carballeda, Aguirre, Pallarés, et al., 2017; Roby, Pijlman, Wilusz, & Khromykh, 2014). Future endeavors to further determine the role of non‐coding RNAs in virus biology should be a very fruitful area of investigation.

RNA “SPONGING” AND RNA VIRUSES

Noncoding RNAs, such as long non‐coding RNA (lncRNA), have shown the ability to regulate the availability/function of gene products through sequestration of other RNA or proteins in a process known as “sponging.” In eukaryotic cells, a class of competing endogenous RNAs (ceRNA) have been characterized as being able to interfere with miRNA regulation (Lin et al., 2017; Zhang, Guo, Ma, Ma, & Xue, 2017). Additionally, cases of host RNA sequestering proteins have been illustrated that can lead to disease, such as the case of myotonic dystrophy, where excessive CUG triplet repeats in the 3’ UTR of the DMPK mRNA bind and sequester muscleblind like 1 (MBNL1) (Lee, Li, et al., 2013). Given the nature of viruses to adapt and evolve to mimic host‐like pathways, they likely use protein or RNA sponging as a major strategy in host‐virus interactions. Below are a few examples of where this is occurring.

RNA–RNA sponging

The sponging of cellular miRNAs has been well‐described for the transcripts of several DNA viruses (Cazalla, Yario, & Steitz, 2010; Liu et al., 2013). To date, the best example of miRNA sponging in RNA virus biology is associated with HCV where an interaction with cellular miR‐122 is essential for genome stability (Figure 7a) (Wilson & Huys, 2013). However, the sequestration of miR‐122 by the large number of HCV plus stranded RNAs has impact on the expression of cellular genes that are normally regulated by miR‐122. In some cases, these alterations in gene expression can promote viral replication. An up‐regulation of STAT3 in HCV infection, for example, represses an interferon (IFN) type I response (Xiong et al., 2015). The global effects of de‐repression of miR‐122 target genes in HCV have been documented via high‐throughput sequencing and cross‐linking immunoprecipitation (HITS‐CLIP) procedures (Luna et al., 2015). Other viruses besides HCV also disrupt cellular miRNA regulatory pathways. The sequestration of miR‐17 by the 3’ UTR of numerous pestivirus RNAs induced a global de‐repression of miR‐17 gene targets (Figure 7a) (Scheel et al., 2016). It will be interesting to examine the capacity of other viral genomic RNAs to act as sponges for select cellular miRNAs or other transcripts.

Figure 7

Sponging of cellular factors by RNA virus transcripts. Panel a: Sponging of cellular miRNAs by members of the Flaviviridae. HCV uses sequences in its 5’ UTR to sponge miR‐122 while bovine viral diarrhea virus and other pestiviruses use 3’ UTR sequences to sponge miR‐17. In both cases, miR sponging increases viral RNA stability while dysregulating aspects of cellular gene expression normally influenced by the small RNAs. Panel b: Insect‐borne flaviviruses contain a knot‐like structure in the 3’ UTR region of their mRNA that generates a stable RNA decay intermediate called sfRNA. The sfRNA not only effectively represses the exoribonuclease that generates it due to slow release of the stalled enzyme, but also sponges key proteins involved in the RNA interference and interferon pathways, resulting in the repression of these key aspects of cellular anti‐viral defense

RNA‐protein sponging

Viral RNAs can also effectively sponge host proteins, leading to benefits for viral replication and possibly contributing to host cell pathogenesis. Subgenomic flavivirus RNA (sfRNA) is a ~400–500 base noncoding RNA derived from the 3’ UTR region of insect‐borne flaviviruses, such as Dengue Virus (DENV) and Zika Virus (ZIKV) (Roby et al., 2014). The production of sfRNA is not due to an internal promoter, but rather the RNA represents a stable decay intermediate that is generated by the cellular 5′‐3′ exoribonuclease XRN1 stalling at a knot‐like three‐helix junction RNA structure (Chapman, Costantino, et al., 2014). DENV sfRNAs have been shown to interact with almost 200 cellular proteins (Figure 7b)—including three that are essential for production of interferon stimulated proteins (G3BP1, G3BP2, and CAPRIN1) (Garneau et al., 2008) as well as one that blocks RIG‐I activation (TRIM25) (Manokaran et al., 2016). ZIKV sfRNA was also shown to interfere with RIG‐I mediated type I interferon responses, perhaps through a similar mechanism (Donald et al., 2016). Flavivirus sfRNAs have also been shown to interact with Dicer and Argonaute proteins and interfere with RNAi responses (Moon, Dodd, et al., 2015; Schnettler et al., 2012).

Another conserved example of protein sponging by a group of RNA viruses can be found in the alphaviruses of the Togaviridae family. The 3’ UTR region of Sindbis virus (SINV) contains a U‐rich region that interacts with the cellular protein HuR with high affinity (Barnhart et al., 2013). The binding of HuR protein by viral RNA leads to the stabilization of SINV RNAs by preventing deadenylation and decay (Garneau et al., 2008; Sokoloski et al., 2010). The sequestration or sponging of large amounts of HuR protein by alphavirus 3’ UTRs causes the protein to relocalize from the nucleus to the cytoplasm in infected cells (Sokoloski et al., 2010). Besides adding stability to the viral RNA, sponging of HuR leads to disruption of post‐transcription regulation of host genes via destabilization of transcripts and dysregulation of splicing and polyadenylation (Barnhart et al., 2013; Dickson et al., 2012). The dysregulation of host homeostasis perhaps creates a new cellular landscape driven by the virus to favor its replication and contributes to virus‐induced cytopathology.

Antiviral treatments involving RNA sponging

There is growing evidence that the sponging strategy can be engineered to be used against viruses to develop novel antiviral techniques. HCV replication, for example, can be inhibited by RNA aptamers that bind to HCV NS5B RNA replicase protein with high affinity. This aptamer can also decrease HCV replication in vivo without off‐target effects or activation of the innate immune system (Lee, Lee, et al., 2013). Finally, efforts are also well under way to develop antiviral RNA sponges. A miR‐122 antagonist effectively inhibits HCV replication, and is currently being evaluated as a novel antiviral therapeutic (Elmén, Lindow, et al., 2008; Lindow, Silahtaroglu, et al., 2008; Ottosen et al., 2015).

SUMMARY/FUTURE DIRECTIONS

The interplay between RNA viruses and post‐transcriptional processes is clearly extensive and highly relevant to virus biology and pathogenesis. Three fundamental themes emerge from the examples we have outlined above. First, the interplay of viruses and cellular RNA biology is clearly part of the molecular arms race between the host cell and the invading pathogen. By impacting fundamental aspects of gene expression through altered post‐transcriptional pathways, invading RNA viruses significantly decrease the ability of the cell to alter its gene expression to respond to the pathogen. RNA viruses can target a wide range of aspects of post‐transcriptional including nuclear pre‐mRNA processing, modulation of cellular capping or subcellular localization of cellular mRNAs, and/or usurping translation and mRNA decay to favor virus gene expression. Cellular efforts to disfavor the use of these post‐transcriptional pathways by viruses—many of which undoubtedly are yet to be discovered—are part of the innate defense mechanism of the cell to infection.

The second theme is the simple but fundamental notion that RNA viruses must rely on host cell functions to further their own replication. The relatively small size of the RNA virus genome precludes encoding many proteins that are needed to regulate viral gene expression, including the post‐transcriptional processing and fate of viral transcripts. As we increase our appreciation of the mechanistic depth of impact of RNA biology on virus replication, we gain new insights into not only virus biology but also potential avenues for the targeting of antiviral therapeutics.

The final theme is one of the excitement that lies in future discoveries in this area. There is, for example, significant insights to be gained into novel aspects of cellular RNA biology through the study of RNA virus‐cell interactions. Over their evolution, viruses and viral RNAs have experimented and developed highly effective ways to usurp and/or avoid aspects of cellular post‐transcriptional regulation. Understanding this interplay from a detailed mechanistic perspective is thus likely to be very valuable for the field. In addition, challenging questions in RNA biology, such as what are the components and structure of a bona fide messenger RNP, may be best answered by studying abundant viral mRNAs and then extending these discoveries to cellular transcripts. Thus the interdisciplinary collaboration between cell biologists and virologists should continue to prove very fruitful in the near future.

CONFLICT OF INTEREST

The authors have declared no conflicts of interest for this article.

RELATED WIREs ARTICLES

https://doi.org/10.1002/wrna.1498

https://doi.org/10.1002/wrna.1458

https://doi.org/10.1002/wrna.1351