RNA regulatory mechanisms that control antiviral innate immunity

Abstract

From the initial sensing of viral nucleotides by pattern recognition receptors, through the induction of type I and III interferons (IFN), upregulation of antiviral effector proteins, and resolution of the inflammatory response, each step of innate immune signaling is under tight control. Though innate immunity is often associated with broad regulation at the level of gene transcription, RNA-centric post-transcriptional processes have emerged as critical mechanisms for ensuring a proper antiviral response. Here, we explore the diverse RNA regulatory mechanisms that modulate the innate antiviral immune response, with a focus on RNA sensing by RIG-I-like receptors (RLR), interferon (IFN) and IFN signaling pathways, viral pathogenesis, and host genetic variation that contributes to these processes. We address the post-transcriptional interactions with RNA-binding proteins, non-coding RNAs, transcript elements, and modifications that control mRNA stability, as well as alternative splicing events that modulate the innate immune antiviral response.

1 |. INTRODUCTION

Innate immune pathways require stringent regulation to promote a robust response to infection while preventing detrimental excess inflammation. The sensing of viral infection activates diverse host antiviral defense programs, but viruses actively counteract these measures to promote their own replication. Broad transcriptional and metabolic changes are key features of both cellular antiviral programs and viral replication. During the rapid timescale of the innate immune response, post-transcriptional regulation of RNA serves as potent mechanism for modulating host antiviral processes.

The function and expression of RNA can be tuned at every stage of its life cycle, beginning with co-transcriptional control of capping, splicing, and polyadenylation; continuing with nuclear export and translation; and terminating with RNA degradation. RNA-binding proteins (RBPs) are inextricably involved in every facet of RNA metabolism. RNA molecules are decorated with RBPs from their inception, and exist exclusively as ribonucleoproteins (RNPs). The critical importance of RNPs is highlighted by the fact that approximately 10% of the human proteome is comprised of RBPs. This percentage continues to grow as additional cellular factors are found to contribute to RNA regulation. Several distinct RBPs bound to a single RNA molecule can act in concert with or against one another. Such interactions between RBPs and RNA lead to transcript-specific molecular outcomes that ultimately shape the cellular transcriptome.

RNA-RBP pairing is mediated by RNA-binding domains within RBPs that interact with specific sequences and structures within RNA molecules.¹ For example, the RBP HuR, which contains three globular RNA recognition motifs, interacts with U-rich and AU-rich sequences within introns and 3′ UTRs, influencing the splicing of target mRNAs and promoting RNA stability.^2,3 On the other hand, the RBP Staufen1 interacts with structured RNA helices through its double-stranded (dsRNA)-binding domains in a largely sequence-independent fashion.^4,5 RBPs may also contain helicase domains that bind highly structured RNAs and unwind them, thereby potentiating additional RNA-RBP interactions.⁶ In addition to such ordered RNA-binding domains, intrinsically disordered domains can also mediate specific and non-specific RNA-protein interactions and promote phase transition of RNPs into membrane-less organelles where many RNA regulatory processes occur.^7–10

Chemical RNA base modifications also engage RBPs to alter RNA fate and function. A 7-methylguanosine (m⁷G) RNA cap protects the 5′ terminus of mRNAs from exoribonucleolytic activity. This cap is also required for ribosome engagement and translation initiation through the cap-binding protein eIF4E.¹¹ Recently, internal modifications in mRNAs have been discovered and recognized as an additional “epitranscriptomic” regulatory layer of gene regulation.¹² For example, N6-methyladenosine (m⁶A) recruits “reader” RBPs, which contain domains that specifically interact with the methyl group of the modified adenosine.^13–18 m⁶A can also modulate local RNA structure and alter the binding of structure-dependent RBPs such as hnRNPC, hnRNPG, and ADAR1.^19–21 Other RNA modifications including pseudouridine, 5-methylcytidine, and N4-acetylcytidine may also modulate RNA function through similar mechanisms.^12,22–25

Non-coding RNAs play key roles in RNA regulatory processes through RNA protein as well as RNA-RNA interactions. Although only 2% of the human genome encodes proteins, more than 85% is transcribed into non-coding RNAs, including microRNAs (miRNAs) and long non-coding RNAs (lncRNAs).^26,27 miRNAs are single-stranded RNAs that are 21–23 nucleotides in length. After extensive processing of primary miRNA transcripts, miRNAs are loaded into the RNA-induced silencing complex (RISC), where they bind to the 3′ UTRs of target mRNAs through their “seed sequence,” which spans nucleotides 2–8, causing degradation or translational suppression of target mRNAs.^28–30 At least 30% of human mRNAs have the potential for regulation via miRNA binding.^28–30 LncRNAs, thousands of which are encoded in the human genome, are defined as non-coding RNAs greater than 200 nucleotides in length and are often differentially expressed during viral infection and immune stimulation.^31,32 Until recently, lncRNAs were considered transcriptional noise due to their often-low levels of expression. However, lncRNAs are increasingly implicated as diverse regulators of transcription, protein modulation, and RNA activity, most often serving as molecular decoys, guides, or scaffolds.^33–37 The biological roles of the vast majority of lncRNAs are still undetermined and there is great need for further investigation of their functions.

Given their importance in controlling biological processes, it is not surprising that RNA-centric mechanisms regulate the immune response to viral infection. Antiviral immune pathways are activated through cellular sensing of viral nucleic acid motifs by a group of sentinel proteins called pattern recognition receptors (PRRs). Viral RNA motifs are sensed by the RIG-I-Like receptors (RLRs) RIG-I and MDA5, as well as by the Toll-Like Receptors (TLRs) TLR3, TLR7, and TLR8.^38,39 Viral DNA is detected by TLR9, cGAS, and IFI16.⁴⁰ Although these sensors reside in diverse subcellular compartments and signal through various adaptors, their detection of viral nucleic acids culminates in the same downstream activation of the interferon regulatory factor (IRF) family of transcription factors and the subsequent production of type I and III interferons (IFNs). IFNs are cytokines that act in both autocrine and paracrine fashions to induce hundreds of interferon-stimulated genes (ISGs), which are potent cellular defense factors that establish an antiviral state and counteract infection.⁴¹

In this review, we will highlight examples of RNA-centric processes that modulate the landscape of gene expression during the innate immune response to viruses, thereby shaping the outcome of infection. Because of their critical function in host protection from RNA virus infection, we will focus specifically on the RNA regulatory controls of RLR signaling pathways, type I and type III IFNs.

2 |. RNA REGULATION OF VIRAL SENSING BY RLRS

When the RLRs RIG-I or MDA5 interact with viral “non-self” or endogenous “self” RNA motifs, they undergo conformational changes, oligomerize through their caspase activation and recruitment domains (CARDs), and translocate to the vicinity of the adaptor protein MAVS.³⁸ MAVS is anchored to both mitochondria and contact sites between mitochondria and ER, as well as to peroxisomes. MAVS itself encodes a CARD, and as such, activated RIG-I or MDA5 oligomers trigger the aggregation of MAVS through CARD-CARD interactions, building a platform for downstream signaling. While a diverse array of proteins is involved in MAVS signalosome function, the kinases IKKε and TBK1, which phosphorylate the transcription factors IRF3/IRF7, are especially critical.^38,42 Phosphorylated homodimers of IRF3 or IRF7 translocate into the nucleus and induce type I and III IFNs. RLRs and the MAVS signalosome also activate NF-κB, augmenting robust IFN induction. In this section, we highlight RNA-centric mechanisms that control RLR signaling (Figure 1).

FIGURE 1.

Summary of RNA regulatory mechanisms that control RIG-I-like receptor (RLR) activation. Sensing of viral (“non-self”) or cellular (“self”) RNA motifs by the RLRs RIG-I or MDA5 triggers a signaling cascade centered around the adaptor protein MAVS, ultimately leading to the phosphorylation and nucleation of IRF3 or IRF7, and the induction of type I and III IFNs. These pathways can be regulated by diverse RNA regulatory mechanisms. RNA-binding proteins (RBPs) or alternatively spliced (AS) proteins, lncRNAs, and miRNAs that modulate this pathway are labeled in black, red, and blue, respectively

2.1 |. “Self” and “Non-Self” RNA sensing by RLRs

The composition and structure of “non-self” viral RNA determine how it is sensed by RLRs. Both RIG-I and MDA5 can recognize double-stranded RNA (dsRNA). MDA5 senses long dsRNA intermediates of viral replication during infection by enteroviruses and coronaviruses.^43–47 On the other hand, RIG-I is activated by short dsRNA containing either a 5′ triphosphate (5′-ppp) or diphosphate (5′-pp), but lacking an m⁷G cap.^48–51 RIG-I (and potentially MDA5) can also recognize RNA lacking 2′O-methylation at its 5′ end.^52–54 Both the m⁷G cap and terminal 2′O-methylation, found on “self” mRNAs, prevent RIG-I from recognizing endogenous mRNAs and aberrantly activating an IFN response. The antiviral effector IFIT1, an ISG upregulated downstream of RIG-I and MDA5 signaling, senses the lack of 2′O-methylation on viral transcripts and acts to limit the translation of viral RNA.⁵⁵ Emulating the 5′ cap structure is an efficient strategy for viruses to avoid detection by RIG-I and to engage cap-dependent translation machinery. NS5, the RNA-dependent RNA polymerase of flaviviruses, which are positive-sense, single-stranded RNA viruses, encodes both m⁷G- and 2′O-methyltransferase activity.^56–58 Mutation of NS5 2′O-methyltransferase activity in yellow fever virus and West Nile virus (WNV) results in increased IFN production, demonstrating that flaviviruses use this strategy to evade recognition by RIG-I.^52,59 Similarly, the NSP16 protein of coronaviruses has cap 2′O-methyltransferase activity.^54,60,61 Ablation of NSP16 cap 2′O-methyltransferase activity attenuates the diverse coronavirus species in a sensing- and IFN-dependent manner. In addition to the 5′ cap structure, several families of viruses contain m⁶A and other RNA modifications within their transcripts.⁶² In vitro transcribed RNAs containing modified nucleobases like m⁶A can suppress detection by RLRs and TLRs.^63,64 Inhibition of RLR activation through m⁶A modifications has been demonstrated to occur during hepatitis C virus (HCV), hepatitis B virus, and human metapneumovirus infection.^65,66 Thus, viruses may co-opt cellular RNA modification processes to appear more like endogenous transcripts, thereby shielding their genomes from detection by PRRs.

RIG-I-like receptors can also sense “self” RNAs which become unmasked during infection or accumulate due to loss of proper RNA metabolism. Sensing of such transcripts may enhance antiviral signaling. During infection by the DNA viruses herpes simplex virus (HSV)-1 and Epstein-Barr virus (EBV), as well as by the negative-stranded RNA virus influenza A virus (IAV), RIG-I is activated by the mis-localized host 5S ribosomal RNA pseudogene transcript RNA5SP141, which accumulates aberrantly in the cytosol. RNA5SP141 contains dsRNA elements plus a 5′-ppp moiety and is therefore a potent RIG-I substrate. Recognition of RNA5SP141 by RIG-I has been shown to be critical for the induction of IFN and restriction of viral replication during HSV-1 infection.⁶⁷ Similarly, the reactivation of the DNA virus Kaposi’s sarcoma herpes virus (KSHV) leads to the accumulation of 5′-ppp containing vault RNAs, a class of poorly understood small RNAs, which act as RIG-I agonists. At homeostasis, 5′-ppp on vault RNA is reduced by the cellular triphosphatase DUSP11; however, KSHV reactivation inhibits the transcription of DUSP11, increasing the immunostimulatory potential of these small RNA species.⁶⁸ During viral infection, the activation of the OAS-RNase L antiviral system also generates “self” ligands that trigger RLR activation. The 2′−5′ oligoadenylate synthetase (OAS) family of ISGs sense viral dsRNA and catalyze the production of the second messenger 2′−5′ adenylate (2′−5′A). 2′−5′A then activates the latent endoribonucleolytic activity of RNase L, which cleaves both viral and host single-stranded RNA.^69,70 Although RNase L-mediated RNA cleavage primarily denies a permissive cellular environment for viral replication, RNase L cleavage products can also form small duplex RNA ligands that activate both RIG-I and MDA5 to amplify the antiviral response.⁷⁰

“Self” RLR activation can promote antiviral immunity through counteracting viral mechanisms that hide viral RNA from cellular pathogen sensors. However, RLR signaling in the absence of infection is detrimental and may lead to inflammatory and autoimmune diseases.^71,72 As such, the host employs multiple strategies to restrict sterile activation of RLR pathways during homeostasis. As mentioned, the triphosphatase DUSP11 restricts RIG-I-activating 5′-ppp moieties in certain “self” non-coding RNAs.^68,73 Keeping aberrant MDA5 activation in check, however, involves the suppression of endogenous dsRNA. For example, an RNA degradosome comprised of the helicase SUV3 and the polynucleotide phosphorylase PNPT1 is responsible for rapid turnover of dsRNA intermediates that result from mitochondrial transcription. Inhibition of SUV3 and PNPT1 leads to accumulation of mitochondrial dsRNA, which enters the cytoplasm and engages MDA5.⁷⁴ Pairs of Alu elements, ~300 nt long retrotransposons abundantly dispersed throughout primate genomes, can form long dsRNA regions in cellular RNAs and also act as a potent substrate for MDA5 activation.^75–78 However, ADAR edits structured Alu elements, changing adenosine bases to inosine, thereby reducing base-pairing potential and inhibiting MDA5 filament assembly plus downstream signaling.^77,78 Gain-of-function mutations in MDA5 that promote MDA5-RNA binding and loss-of-function mutations in ADAR1 that inhibit RNA editing both cause autoimmune interferonopathies such as Aicardi-Guitières syndrome.^77,79–81

Differential splicing of cellular RNAs may alter RLR signaling. This facet of RNA regulation is best exemplified in tumor cell suppression of IFN activation. Transcriptomic analyses of tumor samples reveal dramatic alterations to RNA splicing.⁸² The splicing factor hnRNPC, upregulated in many cancers, suppresses the retention of Alu element-containing introns during mRNA processing.^83–85 Depletion of hnRNPC in breast cancer cell lines causes tumor suppression through heightened sensing of Alu elements in endogenous dsRNA ligands, which subsequently induces type I IFN.⁸⁶ Furthermore, spliceosome-targeted therapies (STTs) which aim to repress key splicing factors, have shown efficacy in cancer models. In a recent report, STTs resulted in the production of dsRNA species via intron retention, thus activating the RLR pathway.⁸⁷ In murine breast cancer models, STTs that activated the RLR pathway resulted in increased antiviral immune signaling, tumor cell death, and increased adaptive immune responses. Similarly, breast cancer patients with increased intron retention exhibited improved disease-free survival compared to those with lower intron retention.

Sensing of immunostimulatory RNA danger-associated molecular patterns (DAMPs) in cancer cells may also lead to increased resistance to radiotherapy or chemotherapy. In breast cancer models, stromal fibroblasts co-cultured with tumor cells produce RNA containing exosomes, which activate STAT1 in a RIG-I-dependent manner. Activation of STAT1 and ISG induction via exosome RNA (exoRNA) and NOTCH3 signaling in breast cancer cells cooperatively render these cancer cells refractory to therapy, likely by promoting DNA damage resistance.⁸⁸ Interestingly, these exoRNAs consist mainly of 7SL RNA transcripts such as RN7SL1, which normally nucleate the signal recognition particle (SRP), a highly conserved RNP essential for protein membrane localization and secretion.^89,90 RN7SL1 contains a terminal triphosphate moiety within its 5′ Alu-like RNA sequence. In the cytoplasm of homeostatic cells, the heterodimeric SRP proteins SRP9 and SRP14 interact with the Alu domain of RN7SL1 to shield the 5′-ppp from RIG-I. However, reduced SRP9/14 incorporation into exosomes exposes the 5′-ppp of RN7SL1, which can then be sensed by RIG-I in tumor cells.⁸⁹ Taken together, these findings reveal that both “self” and “non-self” RNAs are important in regulating IFN production through the RLR pathway. While RLR recognition of “non-self” RNA is largely understood, future studies with undoubtedly reveal further cellular factors and mechanisms that either promote or prevent the detection of “self” RNA motifs to control IFN activation.

2.2 |. Post-transcriptional regulation of the RLR pathway

2.2.1 |. Alternative splicing and open reading frames

Nearly 85% of all transcripts undergo alternative splicing.⁹¹ Through alternative exon usage, intron retention, or splice site selection, alternative splicing serves to diversify the proteome and generate proteins with distinct properties. Further, alternative splicing can also contribute to post-transcriptional mRNA regulation through nonsense-mediated decay, nuclear retention of transcripts, or modulation of translation efficiency.⁹² All cellular factors involved in antiviral signaling have the potential to be affected by splicing, and indeed, modulation of both host- and pathogen-driven splicing is abundant during viral infection.^91,92 In fact, RIG-I itself undergoes IFN-inducible alternative splicing, generating a truncated isoform termed RIG-I SV. RIG-I SV cannot trigger downstream signaling as it lacks the ability to interact with TRIM25, an E3 ubiquitin ligase required for activating RIG-I through K63-linked ubiquitination. Furthermore, RIG-I SV serves as a dominant-negative signaling inhibitor by forming a complex with full-length RIG-I and preventing its interaction with MAVS.⁹³ An alternate splicing isoform of TBK1 also reduces antiviral signaling. Sendai virus infection or IFN-β treatment induces a truncated isoform TBK1s that lacks the kinase domain required to phosphorylate the transcription factor IRF3. Instead, TBK1s binds directly to RIG-I and disrupts RIG-I/MAVS interactions.⁹⁴ These findings highlight the potential for alternative splicing to generate dominant-negative isoforms of antiviral signaling molecules as a feedback mechanism to downregulate antiviral signaling.

Conversely, splicing regulatory mechanisms may also promote antiviral signaling, as seen in the case of Irf7, a key transcription factor of IFN in myeloid cells. The fourth intron of Irf7 contains a weak 5′ splice junction site that hampers efficient splicing and limits IRF7 expression. However, the splicing factor BUD13 binds this weak splice site and promotes more efficient splicing. Depletion of BUD13 results in increased retention of intron 4, decreased expression of IRF7 and subsequently diminished type I IFN, causing increased susceptibility to Indiana vesiculovirus (VSV) infection.⁹⁵ “Bottleneck introns” like intron 4 of Irf7 may constitute a widespread mechanism for proper temporal control of immune-related transcripts.⁹⁶ More broadly, specific splicing factors have been shown to control the proper splicing of immunity-related regulons.^97,98 However, the precise roles of various RBPs in alternative splicing of antiviral signaling elements remains an open area of investigation.

Alternative translation initiation is yet another mechanism that may control RLR signaling. Proper translation initiation at methionine codons depends on their being situated in the context of a strong Kozak nucleotide sequence.⁹⁹ Leaky ribosome scanning across initiation sites with weaker Kozak context can cause protein translation to begin from downstream methionine codons. For example, the MAVS transcript translation may initiate from at least two distinct methionine codons.¹⁰⁰ Translation initiation from the canonical start codon leads to the production of full-length, signaling-competent MAVS protein. However, upstream open reading frames in the MAVS 5′ UTR sometimes mask the canonical start site.^100,101 In such scenarios, translation will initiate from a downstream, in-frame methionine codon at amino acid 142, producing a truncated protein called miniMAVS. As miniMAVS lacks the N-terminal CARD, it cannot oligomerize to form a signaling complex and therefore cannot transduce RLR signaling. MiniMAVS negatively regulates IFN production, likely by competing with full-length MAVS for binding of other key signaling platform components like TRAF2 and TRAF6. Recently, other truncated variants of MAVS protein that prevent full-length MAVS oligomerization have been identified.¹⁰² These variants are smaller than miniMAVS and are likely translated from methionine codons even further toward the 3′ end of MAVS mRNA than Met142. Ribosome profiling experiments in a human monocytic cell line indicated that mRNAs of other RLR pathway components like MDA5 and TRIM25, as well as downstream antiviral effectors like MX2 and IFITM2, may also be regulated by alternative translation initiation, but this remains to be validated. Much like alternative splicing, alternative translation initiation and cis-regulatory upstream ORFs have profound effects on antiviral signaling pathways. Better understanding the mechanisms by which they regulate RLRs and the IFN response will yield further implications for infection, autoimmunity, and cancer.

2.2.2 |. RNA-binding proteins and RNA modifications

In addition to splicing, RBPs regulate antiviral sensing and signaling via control of nuclear export of mRNA. DEAD-box (DDX) RNA helicases, encoded in domains of both RIG-I and MDA5, have been shown to play diverse roles in antiviral immunity, primarily through modulating export of host mRNA transcripts central to viral RNA sensing pathways.¹⁰³ For example, DDX46 negatively regulates the production of type I IFN during viral infection by retaining MAVS, TRAF3, and TRAF6 mRNAs in the nucleus.¹⁰⁴ These transcripts all contain the RNA modification m⁶A, which is known to promote nuclear export.¹⁰⁵ However, during viral infection, DDX46 recruits the m⁶A “eraser” ALKBH5 to demethylate these transcripts, resulting in nuclear retention and reduced protein expression. The splicing factor DDX39A also inhibits nuclear export of MAVS, TRAF3, and TRAF6 during viral infection.¹⁰⁶ The addition of the small ubiquitin-like modifier protein SUMO to DDX39A inhibits its ability to bind RNA. However, viral infection reduces SUMOylation of DDX39A by downregulating its E3-ligase, RanBP2. In the absence of SUMOylation, DDX39A binds MAVS, TRAF3, and TRAF6 mRNAs and sequesters them in the nucleus. Whether DDX39A and DDX46 co-operate in nuclear retention of the same set of transcripts is unknown.

2.2.3 |. miRNAs

Several miRNAs are known to modulate antiviral immune responses through interactions with various components and regulators of the RLR signaling pathway. miR-485, induced after viral sensing, targets the 3′ UTR of DDX58, which encodes RIG-I. Ectopic expression of miR-485 is sufficient to decrease RIG-I levels, inhibit type I and III IFN production, and promote viral infection. Interestingly, miR-485 also targets the PB1 RNA segment of IAV H5N1. During high dose infection with IAV H5N1, ectopically expressed miR-485 targets PB1 significantly more than RIG-I and thus decreases IAV replication.¹⁰⁷ Such examples of bimodal regulation of both host and viral gene expression add complexity to the role of miRNAs in antiviral immunity. In this context, miR-485 negatively regulates the antiviral immune response, but switches to a viral target and protects the host when exposed to high levels of IAV viral RNA. miR-526a, also induced during viral infection in an IRF3- and IRF7-dependent manner, promotes the antiviral response by targeting CYLD, a negative regulator of RIG-I signaling.¹⁰⁸ CYLD downregulates RIG-I signaling by removing K63-linked ubiquitin modifications required for critical conformational changes that occur during RIG-I recognition of viral RNA motifs. By downregulating CYLD expression, miR-526a enhances RIG-I ubiquitination and upregulates IFN production. Accordingly, inhibition of miR-526a results in stabilized CYLD, decreased immune responses to virus, and higher viral burden.

Both miR-155 and miR-146 were initially discovered as miRNAs differentially regulated in response to diverse immune stimuli including TLR signaling, and type I IFN, and viral infection,.^109–112 Intriguingly, miR-155 and its complementary miR-155* (also called miR-155–3p), encoded from the same primary miRNA locus, affect IFN production in opposite ways. In plasmacytoid dendritic cells (pDCs), major producers of type I IFN, TLR7 activation induces miR-155* at early stages and miR-155 at later times after stimulation, likely due to differences in pri-miR155 processing mediated by the RBP KSRP. miR-155* augments type I IFN production by inhibiting the negative regulator IRAK3, while miR-155 restricts type I IFN by targeting the positive regulator TAB2. Thus, miR-155 and miR-155* help to ensure robust, early induction of IFN and inhibit prolonged, deleterious expression of this inflammatory cytokine.¹¹²

miR-146a, in contrast with miR-155, consistently downregulates type I IFN production and allows for increased viral replication during infection. In myeloid cells, VSV infection induces miR-146a in a RIG-I-dependent manner. Treatment with miR-146a mimics quenches IFNB1 expression, while miR-146a inhibitors enhance IFN production; as would be expected, these treatments promote and restrict VSV infection, respectively. Mechanistically, miR-146 targets IRAK1, IRAK2, and TRAF6 mRNAs, all of which encode key signaling molecules positioned at the nexus of IRF and NF-kB activation.^109,113 During infection with enterovirus 71, miR-146a blocks TRAF6 and IRAK4 signaling, limiting type I IFN induction.¹¹⁴ Interestingly, like miR-146a, miR-144 also targets TRAF6 to suppress type I IFN production, exemplifying redundant miRNA regulation of a single transcript.¹¹⁵ In human PBMCs, miR-146a overexpression significantly dampens type I IFN and ISG expression, potentially through modulating IRF5 and STAT1 expression.¹¹⁶ This axis of modulation was discovered upon observing lower miR-146a expression in immune cells of individuals with systemic lupus erythematous (SLE), an autoimmune disease characterized by aberrantly high levels of type I IFN production. Though it may seem odd that such a highly induced miRNA actively inhibits inflammation in context of viral infection, the downregulation of miR-146a seen in SLE patients frames this miRNA as an important factor in resolving the type I IFN response to avoid excess inflammatory pathology following viral infection.

2.2.4 |. LncRNAs

Host long lncRNAs have emerged as regulators of several biological processes, including innate immunity to RNA viruses. Viral infection and sterile immune stimulation cause transcriptional changes in hundreds of lncRNAs; however, very few have been functionally dissected and most play unknown roles in the innate immune response to viruses.^31,32 The highly abundant nuclear transcript NEAT1 enhances the innate immune response to Hantaan virus. NEAT1 preferentially interacts with the RNA/DNA-binding protein SFPQ and sequesters it in paraspeckles, thereby inhibiting its function as a transcriptional repressor of RIG-I and DDX60.¹¹⁷ NEAT1 also induces expression of the pro-inflammatory cytokine IL-8 through a similar mechanism, releasing transcriptional repression mediated by SFPQ.¹¹⁸ Lastly, NEAT1 is implicated in IFN production following foreign DNA sensing by cGAS through interactions with cGAS, DNA-PK, and the RBP HEXIM1.¹¹⁹ These studies demonstrate pleiotropic roles for NEAT1 in antiviral immune responses.

Cytoplasmic-localized lncRNAs regulate antiviral immune signaling through interactions with diverse proteins in the RLR pathway. Infection-induced LncITPRIP-1 boosts IFN production and restricts HCV replication by acting as a cofactor for MDA5, binding its C-terminus and promoting its oligomerization around target RNAs.¹²⁰ In mouse models, Lnczc3h7a interacts with TRIM25, an E3 ubiquitin ligase required for robust RIG-I signaling. Deletion of Lnczc3h7a decreases production of type I IFN and increases VSV infection in vitro and in vivo. Lnczc3h7a is proposed to act as a scaffold for TRIM25 and RIG-I, promoting K63-linked ubiquitination of RIG-I and augmenting IFN production.¹²¹ Though no human paralog of Lnczc3h7a is known, whether an alternative human lncRNA functions similarly remains an interesting question. Recently, mechanistic investigation of LncRNA-GM, which promotes type I IFN production, revealed a hitherto unknown post-translational modification that inhibits TBK1 function: S-glutathionylation. LncRNA-GM sequesters the glutathione-S-transferase GSTM1 away from TBK1, thus promoting sustained IFN production.¹²² Intriguingly, infection by several viruses reduces LncRNA-GM expression, enhancing viral replication. This suggests that modulation of lncRNA expression is a useful strategy for viral evasion of host immunity. LncRNAs are not always antiviral and can actively hinder viral RNA sensing. In infected mice, the cytosolic lnc-Lsm3b directly inhibits antiviral signaling by interacting with the sensor RIG-I and competing with its binding of viral RNA.¹²³ lnc-Lsm3b interacts with stem-loop structures to stabilize the CARD-helicase interaction and prevent the release of RIG-I autoinhibition. Consistent with this finding, loss of lnc-Lsm3b both in vitro and in vivo results in heightened antiviral signaling and reduced VSV viral burden.

Highly modulated, location-specific expression of lncRNAs during infection implicates these biomolecules as important for fine tuning the innate immune response to viruses. With a greater understanding of the molecular functions of lncRNAs in immunity, these molecules may emerge as viable therapeutic targets for infection or autoimmune disorders driven by antiviral signaling.

3 |. RNA REGULATION OF INTERFERON AND INTERFERON-STIMULATED GENES

Interferons are indispensable for a robust antiviral response. Type I IFNs, comprised of the fourteen subtypes of IFNα plus the singular IFN-β, are arguably the most critical molecules that drive antiviral immunity. Once secreted from infected cells, type I IFNs bind a heterodimeric receptor composed of IFNAR1 and IFNAR2 (referred to as IFNAR) on the surface of the same, neighboring, or distal cells. Signaling downstream of IFNAR proceeds through the JAK and STAT families of signal transducers.¹²⁴ The four subtypes of type III IFNs act in a similar fashion at epithelial surfaces, albeit through IFNLR1/IL10R2 receptor complex. Through these receptors, both type I and III IFNs activate a common transcription factor complex of composed of STAT1, STAT2, and IRF9 which induces hundreds of ISGs.^124,125 Many ISGs code for direct antiviral effectors, as well as proteins involved in feedback loops that amplify or inhibit antiviral pathways.^41,126 In this section, we will discuss RNA regulatory mechanisms that directly control IFN and ISG transcripts (Figure 2).

FIGURE 2.

Summary of RNA regulatory mechanisms that control IFN and ISGs expression. Type I and III IFN signaling leads to the activation of the ISGF3 transcription factor complex and the expression of antiviral ISGs. RNA-binding proteins (RBPs) or alternatively spliced (AS) proteins, lncRNAs, and miRNAs that modulate this pathway are labeled in black, red, and blue, respectively

3.1 |. mRNA stability elements and RBPs that control IFNs

While the sensing of viral infection leads to a robust transcriptional induction of antiviral IFNs, type I IFN mRNAs are stringently regulated through several post-transcriptional mechanisms. The 3′ UTRs of all IFN mRNAs, like other transcripts encoding pro-inflammatory cytokines, contain AU-rich elements (AREs).^127,128 AREs are cis-regulatory features composed of one or more repeats of the sequence “UUAUUUAUU” or a more minimal sequence of “AUUUA”.^129–131 3′ UTR AREs tend to destabilize transcripts through specific AU-rich element-binding proteins (ARE-BPs) which recruit the deadenylase complex to shorten the polyA tail, ultimately leading to mRNA decay.^132,133 Beyond controlling RNA stability, 3′ UTR AREs can also positively or negatively influence translation of mRNAs depending on which ARE-BPs they recruit.^134,135

The ARE-BP KSRP has been shown to inhibit type I IFN production. Deletion of KSRP in murine cells leads to increased RNA and protein expression of IFN-β and IFN-α4 following treatment with a TLR3 agonist. KSRP interacts directly with Ifnb1 and Ifna4 transcripts and reduces the half-life of these transcripts.¹³⁶ However, KSRP deletion fails to enhance the basal expression of these type I IFN transcripts in the absence of immune stimulation. It is possible that other ARE-BPs like TTP (encoded by ZPF36) play redundant roles and destabilize these mRNAs, limiting their expression.¹³⁷ TTP is responsible for the degradation of other pro-inflammatory cytokine transcripts in an ARE-dependent manner, including TNF, IL2, and IFNG.^128,138,139 To date, however, a direct role for TTP in the control of type I or III IFN transcripts has not been described. On the other hand, ARE-mediated destabilization of transcripts can be countered by HuR (encoded by ELAVL1) which typically competes with negative regulators like TTP or KSRP for ARE-binding.^2,140–143 Consistent with this, HuR interaction with AREs in the 3′ UTR promotes the stability of IFNB1. Both depletion and inhibition of dimerization, required for HuR function, result in a significant reduction of IFNB1 mRNA levels, promoting viral replication.¹⁴⁴ Furthermore, HuR was recently shown to have an altered binding pattern across the transcriptome during innate immune stimulus by a dsRNA analog.¹⁴⁵ Activating the RLR pathway decreased HuR binding in intronic sequences and increased its interaction within 3′ UTRs. In this context, HuR bound to the 3′ UTRs of several ISGs or ISG regulators and promoted their mRNA stability. Thus, it is clear that AREs and ARE-BPs like HuR, KSRP, and TTP work in concert and antagonistically to modulate antiviral and inflammatory responses.

Recently, our laboratory has discovered that IFN-inducible alternative polyadenylation of the antiviral RBP ZAP (encoded by ZC3HAV1) functionalizes ZAP to act through AREs for regulation of IFN production.¹⁴⁶ The alternative polyadenylation of ZAP is driven by the cleavage factor CSTF2 which leads to the expression of two distinct isoforms. The large isoform ZAP-L contains a C-terminal CaaX (cysteine-aliphatic-aliphatic-any residue) amino acid motif. This CaaX motif directs the post-translational prenylation of ZAP-L and targets this isoform to endosomal and lysosomal membranes, where it binds to and degrades viral RNA localized in these compartments. On the other hand, ZAP-S, preferentially induced in response to IFN, does not encode a terminal CaaX motif, and is thus localized diffusely in the cytosol. ZAP-S, but not ZAP-L, binds type I and III IFN mRNAs and destabilizes them. Thus, ZAP-S acts as part of a negative feedback mechanism to prevent prolonged IFN production. AREs in IFN transcripts appear to aid in ZAP binding as mutating AREs in the IFNL3 3′ UTR reduces the interaction of ZAP-S with this transcript.¹⁴⁶ However, it is unlikely that ZAP can directly interact with AREs. Although ZAP can bind UA dinucleotide-rich artificial RNAs, it optimally interacts with CG-dinucleotide containing RNA motifs in viral RNA.^147–149 Thus, how AREs mediate ZAP binding and control of IFN mRNAs still requires further investigation.

While alternative polyadenylation and divergent subcellular localization direct the ability of ZAP isoforms to interact with different RNA species, what regulates altered RBP interactions with specific mRNAs during immune activation is incompletely understood. miRNA binding to 3′ UTRs has the potential to influence local RBP interactions and vice versa.¹⁵⁰ Post-translational modifications are also likely to influence the RBP RNA-binding process.¹⁵¹ Phosphorylation and SUMOylation are post-transcriptional modifications emerging as key mechanisms that regulate the ability of RBPs to bind RNA.^98,106 RBP specificity and function are therefore important considerations when studying the immune response to viral infection and thus warrant further investigation.

3.2 |. RNA modifications in the IFN response

The RNA modification m⁶A affects the host response to viral infection in a virus- and cell-type dependent fashion.⁶² During infection of fibroblasts by the DNA virus human cytomegalovirus (HCMV), enhanced expression of the m⁶A methyltransferase METTL3 induces m⁶A deposition in the 3′ UTR of the IFNB1 transcript near the stop codon. m⁶A in IFNB1 transcript reduces its half-life, inhibits IFN-β expression, and thus promotes viral infection.^152,153 This mechanism of regulation is likely cell-type specific, as neither m⁶A modification of IFNB1 nor a role for the cellular m⁶A machinery in IFNB1 expression is observed in hepatocyte or lung epithelial cell lines.^154–156 However, during infection of hepatocytes by viruses in the Flaviviridae family, activation of RLR signaling and ER stress pathways alters m⁶A modification on both pro- and antiviral transcripts.¹⁵⁴ Many ISGs are also sensitive to m⁶A-mediated regulation; indeed, the transcripts of >50 ISGs are modified by m⁶A. m⁶A promotes the translation of a subset of these ISGs by facilitating transcript interaction with the m⁶A-binding protein YTHDF1.¹⁵⁵ Currently, the role of m⁶A in other m⁶A-modified ISG transcripts remains unknown, but may involve modulation of mRNA processing or half-life.

3.3 |. miRNAs in the IFN response

Beyond regulating RLRs and sensing pathways that lead to IFN induction, miRNAs can directly degrade type I and III IFN transcripts. Several miRNAs target the 3′ UTR of IFNB1, including Let7b, miR-26a, miR-34a, and miR145.¹⁵⁷ These miRNAs are induced during both viral infection and type I IFN treatment and inhibit the expression of IFNB1 in human and non-human primates, serving as a miRNA-regulated negative feedback mechanism for type I IFN production. Work from our laboratory has shown that Hepatitis C virus (HCV) can induce the expression of the miRNAs miR-208b and miR-499 in human hepatocytes.^158,159 These miRNAs are encoded within the introns of the myosin genes MYH7 and MYH7B. These genes and miRNAs are typically restricted in expression to cardiac and skeletal muscle cells and are therefore called myomiRs. Only HCV infection–not infection by other Flaviviridae family members dengue virus (DENV) or WNV nor stimulus with synthetic immunostimulatory ligands–upregulates myomiRs in hepatocytes, suggesting an HCV-specific mechanism of induction. Further, though type I IFN amplifies the expression of these myomiRs during HCV infection, type III IFN does not. These myomiRs target the 3′ UTRs of the type III IFN mRNAs IFNL2 and IFNL3 as well as the 3′ UTR of IFNAR1, leading to destabilization and downregulation. Our studies demonstrated a mechanism by which HCV can suppress host IFN responses by using host RNA machinery to shut down expression of antiviral mediators.

As demonstrated with myomiRs that target IFNAR1, miRNAs may regulate type I IFN signaling at or downstream of the IFN receptor. Ligation of IFNAR1/2 by IFN-α/β induces the phosphorylation of STAT1 and STAT2, which then associate with IRF9 to form the ISGF3 complex. This transcription factor complex then recognizes IFN-sensitive response elements (ISREs) in the genome and induces expression of hundreds of ISGs. miR-155, discussed previously, upregulates IFN signaling downstream of IFNAR through repression of SOCS1, an ISG that limits IFNAR signal transduction. In mice, miR-155 targets the Socs1 3′ UTR, enhancing STAT1 phosphorylation and ISG expression while reducing viral replication.¹¹¹ Alternatively, miR-221 and miR-222 negatively regulate IFN signaling. Depleting miR-221/222 increases expression and phosphorylation of STAT1 and STAT2 and heightens induction of ISGs. However, as miR-221/222 do not share seed-sequence complementarity with STAT1 and STAT2 transcripts, it is likely that they influence IFN signaling through an indirect mechanism. With the examples of myomiRs, mir-155 and miR-221/22, and several others, it is evident that miRNAs modulate the expression of type I and III IFNs as well their downstream signaling pathways, thereby fine tuning the immune response to viral infection.

3.4 |. LncRNAs in the IFN response

Antiviral signaling leads to widespread changes in lncRNA expression. For example, our laboratory has characterized differentially regulated lncRNAs in plasmacytoid dendritic cells (pDCs), a major source of type I IFN during viral infections. 500 lncRNAs were differentially expressed in pDCs following treatment with either a TLR7/8 agonist which induces IFN or with IFN-β itself.^32,160 Several lncRNAs differentially expressed in pDCs during TLR7/8 ligation were also found to be similarly up- or downregulated during IAV infection indicating that these genes also respond to innate immune activation during infection.¹⁶¹ Further, a large proportion of these differentially expressed lncRNA loci were found in proximity on the chromosome to ISGs or other genes involved in immune processes.³² Given that lncRNAs can affect transcription in both cis and trans, lncRNA expression may serve as a means of positive or negative feedback for the cellular antiviral state during infection.^31,162,163

Correspondingly, several lncRNAs are implicated in the regulation of the IFN response and viral infection.¹⁶⁴ For example, transcription of IFNB1 is inhibited by lnc-MxA, an antisense strand non-coding RNA that originates from the ISG locus MX1.¹⁶⁵ When induced via IFN signaling, lnc-MxA interacts directly with the IFNB1 promoter to form an RNA:DNA triplex that inhibits the binding of the transcription factors IRF3 and NF-κB LUCAT1 inhibits ISG expression more broadly through interacting with the transcription factor STAT1.¹⁶⁶ The lncRNA NRAV suppresses the expression of several ISGs by altering histone modification and transcriptional availability of their promoters. Specifically, NRAV appears to promote the deposition of the repressive epigenetic mark histone 3 lysine 27 trimethylation (H3K27me3) and impair decoration with the activating mark histone 3 lysine 4 trimethylation (H3K4me3). Intriguingly, NRAV is rapidly degraded during infection through unknown mechanisms, allowing for ISG expression during the host antiviral response.¹⁶⁷ NRIR (also called lncRNA-CMPK2) also acts to suppress the expression of proximal and distal ISG loci through epigenetic mechanisms. Unlike NRAV, which is downregulated upon viral infection, NRIR is a bonafide ISG induced by IFN treatment and is thus involved in a negative feedback loop during the IFN response.¹⁶⁸ Conversely, the lncRNA IVRPIE promotes ISG expression. Infection-inducible IVRPIE promotes the deposition of the activating histone mark H3K4me3 and blocks that of repressive mark H3K27me3 at IFNB1 and other key ISG loci via interactions with the RBP hnRNPU.¹⁶⁹ hnRNPU also stabilizes LUARIS (also called lncRNA#32), which directs the transcription factor ATF2 and promotes transcription of several ISGs.¹⁷⁰ While the aforementioned lncRNAs all modulate a broad swathe of antiviral genes, some directly regulate ISGs. For example, Lnc-ISG20 is transcribed in the sense orientation from the ISG20 locus. Despite sharing a significant overlap with the mRNA that encodes the antiviral protein ISG20, Lnc-ISG20 cannot be translated. Instead, this non-coding transcript acts as a sponge for miR-326, which normally targets and degrades ISG20, thereby increasing ISG20 protein levels by incapacitating a direct ISG20 inhibitor.¹⁷¹ Further investigation of lncRNAs differentially expressed during infection or immune stimulus will certainly uncover other roles for these versatile molecules in the antiviral immune response.

4 |. HOST GENETIC VARIATION IN THE RNA REGULATION OF THE ANTIVIRAL RESPONSE

Genetic variation allows for disparate outcomes from infection with the same pathogen in different individuals. Single nucleotide polymorphisms (SNPs) in key genes or regulatory elements may significantly influence the strength of an individual’s antiviral immunity. Genome-wide association studies (GWAS), expression quantitative trait locus (eQTL) mapping, and candidate gene approaches have revealed numerous associations between various polymorphisms and immune processes.^172,173,174 However, assigning causality to a particular SNP from a group of polymorphisms in linkage disequilibrium and then following this single SNP’s functional burden proves challenging.¹⁷⁵ Despite the difficulty, several single SNP genetic variants have been demonstrated to impact RNA regulatory processes during antiviral immunity.

Splicing regulatory polymorphisms in the IRF5 gene can contribute to systemic lupus erythematosus (SLE), an autoimmune disease characterized by excess production of type I IFN.¹⁷⁶ IRF5 is an important component of immune activation downstream of TLR7 and TLR8 signaling.^177–179 It directly promotes the transcription of type I IFN, promotes inflammatory pathology, and is also a key regulator of B cell differentiation and function.^180–182 Distinct promoters and transcription start sites delineate four potential first exons for the IRF5 gene (exons 1A-D). Exon 1B initiates a constitutively expressed IRF5 transcript. However, this isoform can only be spliced correctly when the T allele at rs2004640 generates a strong 5′ splice donor after exon 1B. Thus, rs2004640T enhances the expression of functional IRF5 transcripts and promotes dysregulated IFN pathways in SLE.^183–185 An additional SNP rs10954213A in the 3′ UTR of IRF5 generates a strong polyadenylation sequence (AAT(A/G)A) proximal to the gene’s stop codon and therefore produces IRF5 transcripts with a shorter 3′ UTR.¹⁸⁵ This shorter 3′ UTR is associated with extended half-life and higher expression of IRF5, consistent with findings that longer 3′ UTRs promote RNA degradation via facilitating interactions with RBPs and microRNAs.^186–189 By influencing differential splicing and polyadenylation of a central player in antiviral pathways, genetic polymorphisms in IRF5 promote type I interferonopathy; how these SNPs affect the outcome of viral infection is still largely unclear.

In addition to functionalizing polyadenylation sequences, polymorphisms in the 3′ UTR of mRNAs can alter sequence or structure-dependent RBP and miRNA interactions, thus impacting transcript metabolism.^190,191 One such variant exists in the 3′ UTR of the IFNL3 gene which encodes IFNλ3, a type III IFN crucial for balanced antiviral responses.^192,193 Several SNPs at the IFNL3 locus are associated with both response to IFN therapies and with natural clearance of hepatitis C virus (HCV).^194–198 We and others have identified the functional mechanism of one SNP in the IFNL3 3′ UTR, rs4803217T/G.^158,199 The unfavorable T allele of rs4803217, associated with HCV persistence, inhibits compact RNA structure in the IFNL3 3′ UTR, reducing the translation efficiency of this mRNA.¹⁹⁹ The T allele also enhances seed-sequence pairing of the microRNAs miR-208b and miR-499a-5p at the polymorphic site. These microRNAs, together with neighboring AREs, destabilize IFNL3 mRNA harboring the T allele, thereby downregulating cytokine expression and worsening infection outcome.¹⁵⁸ As described in the previous section, the myomiRs induced by HCV infection also downregulate IFNL2 and the type I IFN receptor IFNAR1, pointing to a viral strategy that invokes host miRNAs for evasion of type I and type III IFN signaling.^158,159

Intriguingly, the unfavorable T allele in IFNL3 described above is strongly linked with the ΔG allele of rs368234815TT/ΔG in IFNL4, a type III IFN gene upstream of the IFNL3 locus.^199–201 While the TT allele introduces a frameshift to pseudogenize IFNL4, the ΔG allele can, in principle, lead to functional IFNλ4 expression; thus, IFNλ4 expression may be linked to adverse outcomes of HCV and other viral infections.²⁰¹ However, our laboratory has shown that while IFNλ4 retains bioactivity comparable to IFNλ3, several RNA regulatory mechanisms suppress IFNλ4 expression.²⁰⁰ IFNL4 mRNA can be alternatively spliced to produce three protein-coding isoforms, but only the largest, IFNλ4p179 (named for the number of amino acids), is secreted, despite each isoform containing identical N-terminal secretion signals. Additional isoforms of IFNL4 harbor retained introns and are thus more likely to be retained in the nucleus or subject to NMD. All IFNL4 transcripts poorly associate with polysomes, indicating inefficient translation. Thus, alternative splicing of IFNL4 may serve to limit functional IFNλ4 protein expression. Lastly, the 3′ UTR of IFNL4 does not encode a canonical polyadenylation sequence, further inhibiting the proper processing and maturation of transcripts from this locus.²⁰⁰ Thus, HCV persistence observed with the rs368234815ΔG phenotype may stem from the inability of functional IFNλ4 to compensate for reduced IFNλ3 expression.

miRNA-mediated mechanisms similar to those described in regulation of IFNL3 during HCV infection may also explain the effect of several polymorphisms in the 3′ UTRs of IFNAR1 (the type I IFN receptor) and JAK1 (a crucial kinase required for signaling downstream of IFNAR1). Polymorphisms in these genes are associated with hepatocellular carcinoma (HCC), a disease strongly linked with viral hepatitis. rs17875871 is a 4 bp indel in the IFNAR1 3′ UTR. Perfect seed-sequence pairing between the deletion allele in IFNAR1 and miR-1231 is disrupted by the 4-nucleotide insertion. Consequently, the hetero- or homozygous deletion allele is associated with reduced IFNAR1 expression and elevated risk of HCC, likely through persistence of viral hepatitis.²⁰² Similarly, rs112395617, a 4 bp indel polymorphism in the 3′ UTR of JAK1 (a kinase crucial for type I and III IFN signaling) alters the binding of miR-431–5p and impacts JAK1 expression.²⁰³ While intriguing, more precise functional dissection of these polymorphisms and their interplay with microRNAs is required for full understanding of their effect on antiviral immunity.

Single nucleotide polymorphisms within splice donor and acceptor sites or within cis-regulatory sequences can lead to novel gain-of-function or loss-of-function isoforms of antiviral proteins. These isoforms may contribute to differential control of viral infection and can also lead to autoimmunity. One splicing regulatory polymorphism is found in the OAS1 gene, which encodes an ISG that senses dsRNA and catalyzes the production of the second messenger 2′−5′A. 2′−5′A binding dimerizes and activates the latent endoribonuclease activity of RNase L for degradation of viral RNA.⁶⁹ The C-terminal of human OAS1 is alternatively spliced to produce several potential isoforms including OAS1 p42, p44, p46, p48, and p52. Of these isoforms, only p42 or p46 are expressed robustly at the protein level due to the instability of the other protein variants and of the RNA isoforms encoding them.^203–205 The expression of OAS1 p42 or p46 is controlled by the SNP rs10774671A/G in the splice acceptor site of exon 6. The G allele of this SNP shifts the splice acceptor site by one nucleotide and thus generates p46 with a novel C-terminal amino acid sequence that includes a terminal CaaX motif, subject to post-translational prenylation. Our laboratory has found that prenylation of p46 targets this OAS1 isoform to intracellular organelle membranes such as the Golgi apparatus, whereas unprenylated p42 is localized diffusely in the cytosol. Membrane-localized OAS1 p46 demonstrates enhanced antiviral potential over OAS1 p42 against multiple RNA viruses from distinct families including enteroviruses, flaviviruses, and coronaviruses, all of which use cellular membranes to establish their replication compartments.²⁰⁴ Consequently, by altering the splicing of OAS1, rs10774671G is associated with reduced clinical susceptibility to HCV, WNV, and SARS-CoV-2.^{204,207–210}

As with protein-coding genes, lncRNAs are also subject to regulation by genetic variants which impact viral infection. For example, the SNPs rs1015164 and rs2027820 are polymorphisms in linkage disequilibrium within the CCR5 locus, which encodes a key chemokine and the co-receptor for human immunodeficiency virus-1 (HIV-1).^211,212 rs1015164-G and rs2027820-A are associated with decreased expression of an antisense lncRNA CCR5AS from this locus. CCR5AS promotes the expression of CCR5 by sequestering the RBP RALY and inhibiting RALY-mediated destabilization of the CCR5 transcript.²¹² Therefore, these polymorphisms, which reduce the expression of CCR5AS, also inhibit expression of CCR5 protein levels and therefore promote resistance to HIV-1. However, given that CCR5 deficiency is detrimental to resolution of influenza A virus (IAV) and West Nile virus (WNV) infection, how CCR5AS and its associated genetic variation influences other infections remains in question.^213,214

These examples serve to highlight the breadth of mechanisms by which genetic variation can influence the RNA biology that underlies antiviral innate immunity. While functional characterization of most human polymorphisms remains incomplete, most well-validated variants involve changes in protein sequence. However, the vast majority of all pathogenic SNPs identified through GWAS are in non-coding regions of the genome.^46,47 Up to 15% of all pathogenic SNPs are intronic, perhaps disrupting cis-regulatory RNA elements or affecting splicing.²¹⁵ A significant fraction of SNPs (~3.5%) are in untranslated regions.⁴⁸ Many lie within or regulate lncRNAs.²¹⁶ Mechanistic investigation of pathogenic SNPs in antiviral genes will enhance our understanding of immune pathways and reveal novel RNA regulatory mechanisms in viral infection.

5 |. RNA REGULATION OF VIRAL PATHOGENESIS AND EVASION OF HOST IMMUNITY

Given the importance of RNA regulation in antiviral immunity, it is not surprising that viruses have evolved mechanisms to co-opt or subvert host RNA metabolism. Viral RNAs themselves may also promote pathogenesis. Viruses often target host splicing to evade the immune response and to enhance their own virluence.⁹² The ongoing COVID-19 pandemic has led to rapid advances in our knowledge of how SARS-CoV-2 antagonizes host immune responses. It was recently shown that SARS-CoV-2 non-structural protein NSP16 can interfere with cellular mRNA splicing.²¹⁷ NSP16 interacts with the small nuclear RNAs (snRNAs) U1 and U2, required for canonical intron splicing, at their mRNA splice recognition sites. Infection with SARS-CoV-2 or ectopic expression of NSP16 suppresses global splicing of host mRNAs, including that of antiviral ISGs like RIG-I. DENV disrupts the efficient splicing of cellular genes including RIG-I in a similar manner, as its RNA-dependent RNA polymerase NS5 inhibits host U5 snRNP, another component of the canonical spliceosome.²¹⁸ While viral RNA from coronaviruses and flaviviruses is not spliced, some RNA viruses like IAV require efficient splicing of their RNA for productive infection.¹⁸⁹ During IAV infection, the host splicing factor hnRNPK accumulates in nuclear speckles where viral RNA is spliced, leading to perturbations in hnRNPK-dependent host cellular splicing events.^219–221 Reversal of splicing disruptions induced by IAV attenuates viral infection, demonstrating that impeding host splicing is a convergent strategy used by several viruses to generate a permissive environment for their replication.

As obligate intracellular parasites, viruses require the host translational machinery for the production of their proteins. Consequently, they have evolved numerous mechanisms to subvert or shut off host translation while promoting the translation of viral RNA. Conversely, restricting translation is also a host strategy for combatting viral replication. Thus, access to efficient translation is a key zone of evolutionary conflict between virus and host. Translation initiation in eukaryotes typically starts in a cap-dependent manner, where initiation factors recognize the 5′ m⁷G mRNA cap and recruit the 40S ribosomal subunit.²²² However, many viruses including poliovirus and HCV translate their RNA through cap-independent mechanisms involving exquisitely structured RNA sequences known as internal ribosomal entry sites (IRES), which engage translation initiation factors or the 40S ribosomal subunit directly.²²³ The presence of an IRES in poliovirus RNA allows for its translation despite viral cleavage of the initiation factor eIF4G, which is required for the cap-dependent translation of cellular RNAs.²²⁴ On the other hand, the HCV IRES reduces the sensitivity of viral RNA translation to eiF2α phosphorylation. Phosphorylation of the initiation factor eiF2α by kinases such as the dsRNA sensor PKR is a cellular strategy to limit the translation of viral and cellular RNA during infection and stress responses. As the HCV IRES directly recruits the 40S subunit and bypasses eiF2α, eiF2α phosphorylation only cripples antiviral ISG protein production.^222,225 5′ terminal RNA structures can also rescue viral RNA from translation inhibition by the antiviral effector IFIT1, which binds to viral RNAs lacking 2′O-methylation in the cap structure. In the case of alphaviruses, defined stem-loop structures in the 5′ UTR prevent the ability of IFIT1 to bind to the unmethylated cap of these viruses and thus allow for efficient translation of viral RNA.^55,226

SARS-CoV-2 blocks cellular translation through its NSP1 protein, which interacts with 18S ribosomal RNA and impedes the mRNA entry channel of the ribosome.²¹⁷ Importantly, the stem-loop 1 (SL1) RNA structure in the viral 5′ UTR allows SARS-CoV-2 mRNAs to bypass this blockage and initiate translation. Given that the proximity of SL1 to the 5′ end of viral RNA is critical in avoiding translational suppression, it is thought that NSP1 likely recognizes SL1 and consequently dissociates from the ribosome, allowing for translation of viral mRNA. NSP1 involvement in host translational shutdown is further supported by cryo-EM structure analysis showing NSP1 interaction with the 40S ribosomal subunit and 18S ribosomal RNA.^227,228 In addition to disrupting host translation, SARS-CoV-2 NSP1 can induce degradation of host mRNA transcripts, similar to NSP1 from SARS-CoV.^229,230 The precise interplay between cellular translational inhibition and host mRNA degradation mediated by NSP1 during SARS-CoV-2 infection remains unknown.

In addition to viral proteins, viral RNA also allows for evasion of host responses and may enhance mechanisms of viral pathogenesis. For instance, flaviviruses subvert host responses through the production of a non-coding RNA species called subgenomic flaviviral RNAs (sfRNAs). sfRNAs are produced by all flaviviruses and contribute to viral pathogenesis.²³¹ sfRNAs are generated from the incomplete digestion of the flavivirus genome by the host 5′−3′ exonuclease XRN1.²³² XRN1 colocalizes with the 3′ UTR of WNV in processing bodies, subcellular sites of RNA degradation, and depletion of XRN1 decreases sfRNA generation. Mechanistically, highly structured, pseudoknot reinforced RNA stem loops in the 3′ UTRs called XRN1-resistant RNAs (xrRNAs) work to stall processive XRN1 degradation of the viral genome.^233–235 xrRNAs also contribute to sfRNA generation in mosquito cells, as XRN1 is a highly conserved mechanism of RNA decay across species. sfRNAs in insect cells are important determinants of flavivirus transmission.^231,236,237 Accordingly, mutations in xrRNAs and associated defects in sfRNA production generate bottlenecks in transmission between insect cells and mammalian cells.^238,239

sfRNA-driven flavivirus pathogenesis in human cells occurs through multiple mechanisms. By stalling and sequestering XRN1 at viral RNAs, xrRNA structures and sfRNAs can dysregulate normal cellular mRNA decay, leading to detrimental effects in the host as well as increased viral replication.²⁴⁰ sfRNAs can also dampen the type I IFN response by modulating the function of TRIM25 in RLR signaling.²⁴¹ TRIM25, an E3 ubiquitin ligase that adds activating K63-linked ubiquitin chains to RIG-I during viral sensing, needs to be deubiquitinated itself to function in this capacity. DENV sfRNAs bind to TRIM25 and inhibit its deubiquitination, thus restricting TRIM25 activation of RIG-I. In line with these findings, infection by DENV strain PR-2B, which maintains a higher ratio of sfRNA to genomic RNA than the strain PR-1, demonstrates reduced TRIM25 deubiquitination and decreased IFN production when compared with PR-1. DENV sfRNAs also disrupt the function of the cellular RBPs G3BP1, G3BP2, and CAPRIN1, which are required for the loading of key antiviral ISG transcripts onto polyribosomes for efficient translation. This body of literature provides strong evidence for the role of sfRNAs in promoting viral replication, antagonizing the translation of ISGs, and impairing the antiviral immune response.

While miRNAs influence the antiviral response in various ways, miR-122 plays a uniquely important role in HCV infection. Expressed abundantly in the liver, two miR-122 transcripts may interact simultaneously with the 5′ UTR of HCV.²⁴² miR-122 is essential for HCV replication; inactivation of miR-122 or mutation of the miR-122 binding site in the HCV 5′ UTR drastically reduces viral RNA and protein levels. miR-122 likely promotes HCV infection through several interconnected mechanisms. First, miR-122 bound to AGO2 serves as an RNA “cap” for the 5′ end of HCV RNA, which lacks a typical m⁷G cap, thereby increasing the stability of viral RNA.^243–245 Second, miR-122 binding to the 5′ UTR promotes the translation of HCV RNA by altering the IRES structure.^246–249 Third, miR-122 affects the ratio of viral RNA involved in translation versus replication and thus promotes the synthesis of viral RNA.²⁵⁰ Finally, when miR-122 is bound to viral RNA, it is sequestered from its cellular host targets, the consequential upregulation of which may establish a more permissive cellular environment for HCV replication and persistence.²⁵¹ miR-122 is so essential to the HCV life cycle that it is considered the primary determinant of the virus’s liver-specific tropism.^242,252 Therefore, targeting miR-122 may be a viable host-directed therapeutic strategy to combat HCV infection and is currently being explored in clinical trials.²⁵³ Together, these examples clearly indicate that host RNA processes are a key target that viruses subvert or co-opt for their advantage.

6 |. CONCLUDING REMARKS

The examples discussed in this review highlight the diverse roles of RNA regulatory mechanisms in controlling antiviral innate immunity. However, much remains to be discovered. A systematic dissection of post-translational modifications and how they modulate RBP function may shed light on how antiviral signaling events are rapidly integrated into post-transcriptional regulatory circuits. The mechanistic investigation of the thousands of lncRNAs with unknown functions will undoubtedly reveal their widespread involvement in the immune system. We will also likely gain a greater appreciation of RNA-nucleated biomolecular condensates in organizing cellular immune signaling processes. These advances, and others like them, will likely be driven by recently developed technologies. For example, the widespread use of several crosslinking immunoprecipitation techniques as well as affinity purification strategies coupled with mass spectrometry has greatly enhanced our ability to deconvolve RNA-RBP interactomes. Similarly, the advent of long-read sequencing has allowed for the easy identification of alternative splice isoforms with minimal algorithmic processing as compared with traditional short-read sequencing methods. The application of such state-of-the-art techniques to immunological problems will greatly further our understanding of RNA-centric regulation of the immune response to viral pathogens. With RNA molecules increasingly proved as viable druggable targets, manipulating the RNA-associated processes that control antiviral innate immune signaling pathways may be an efficient therapeutic strategy to combat viral infection and immune dysregulation.