Negative regulators of the RIG-I-like receptor signaling pathway

SUMMARY

Upon recognition of specific molecular patterns on viruses, bacteria and fungi, host cells trigger an innate immune response, which culminates in the production of type I interferons (IFN), pro-inflammatory cytokines and chemokines, and restricts pathogen replication and spread within the host. At each stage of the immune response, there are stimulatory and inhibitory signals that regulate the magnitude, quality, and character of the response. Positive regulation promotes an antiviral state to control and eventually clear infection whereas negative regulation dampens inflammation and prevents immune-mediated tissue damage. An over-exuberant innate immune response can lead to the destruction of cells and tissues, and the development of spontaneous autoimmunity. The RIG-I-like receptors (RLRs) retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated gene 5 (MDA5) belong to a family of cytosolic host RNA helicases that recognize distinct non-self RNA signatures and trigger innate immune responses against several RNA virus infections. The RLR signaling pathway is tightly regulated to achieve a well-orchestrated response aimed at maximizing antiviral immunity and minimizing immune-mediated pathology. This review highlights contemporary findings on negative regulators of the RLR signaling pathway, with specific focus on the proteins and biological processes that directly regulate RIG-I, MDA5 and MAVS function.

INTRODUCTION

Recognition of viral pathogen associated molecular patterns (PAMPs) triggers production of type I interferons (IFN), and pro-inflammatory cytokines and chemokines. This response is one of the first lines of defense during virus infection and primes activation of innate immune cells (dendritic cells (DCs), macrophages, monocytes, innate lymphoid cells (ILCs), and γ/δ T cells) and adaptive immune responses (B and T cells). Each aspect of the immune response is regulated by stimulatory and inhibitory signals that modulate the strength and nature of the response. Positive regulators enhance the antiviral immune response to control and clear viral infection. Conversely, negative regulators dampen inflammatory responses to prevent immune-mediated tissue damage and spontaneous autoimmunity [1–5].

The RIG-I-like receptors (RLRs) RIG-I and MDA5 are important initiators of the innate immune response to RNA virus infection [6]. The RLRs are expressed within the cytoplasm of nearly every mammalian cell. RIG-I and MDA5 both contain two N-terminal caspase activation and recruitment domains (CARDs) that act as signaling domains, a central DExD/H RNA helicase domain that facilitates ATP hydrolysis and RNA binding, and a C-terminal domain (CTD) that aids in RNA ligand recognition and binding specificity (reviewed in detail in [7]).

RIG-I and MDA5 are normally maintained in an inactive state. This is, in part, due to an auto-regulatory function of the CTD, also called the regulatory domain (RD), which interacts with the CARD regions to prevent unwarranted interaction with downstream factors [8]. Phosphorylation of the CARDs, by PKCα/β_II in the case of RIG-I, also helps maintain an inactive state [9, 10]. Following binding to non-self RNAs, the interaction between the RD and the CARDs is disrupted and the RLRs undergo post-translational modification to reach an activated state. PP1α/γ dephosphorylates the CARDs [11], and the E3 ubiquitin ligase TRIM25 promotes K63-polyubiquitionation of the second CARD of RIG-I, while TRIM65 K63-polyubiquitinates the helicase domain of MDA5 [12, 13]. Additionally, Riplet, another E3 ubiquitin ligase, adds K63-linked polyubiquitin to the CTD of RIG-I [14] (Figure 1).

Figure 1. Activation of the RLR signaling pathway.

RIG-I and MDA5 bind distinct moieties on non-self RNAs, which triggers post-translational modifications, oligomerization (not depicted in this schematic), and translocation to mitochondria and mitochondrial-associated membranes (not depicted). Here, activated RIG-I and MDA5 interact with MAVS through CARD-CARD interactions. This triggers the formation of a signaling synapse that includes the recruitment of adaptor proteins (e.g., TRAF3, TRAF6), kinases (e.g., TBK-1, IKK-α, β, ε), and transcription factors (e.g., IRF-3, NF-κB). This results in the formation of the IFN-β enhanceosome complex that promotes IFN-β, pro-inflammatory cytokine, and antiviral effector gene transcription.

Upon activation, the RLRs translocate to mitochondria and mitochondrial-associated membranes where they interact with the essential adaptor protein mitochondrial antiviral signaling [15, 16] (MAVS; also known as IPS-1, VISA, Cardif) via binding of the RLR CARDs with a CARD region on MAVS (Figure 1). MAVS must also be post-translationally modified to attain an activated state – it is phosphorylated by TBK1 [17]. Binding of RIG-I or MDA5 to MAVS triggers formation of a signaling synapse that recruits adaptor proteins, kinases, and transcription factors [15, 16] resulting in the formation of the canonical IFN-β enhanceosome complex that promotes IFN-β transcription (Figure 1). Subsequent type I IFN signaling through the JAK-STAT pathway rapidly induces expression of hundreds of IFN stimulated genes (ISGs), which restrict virus infection and replication through direct effector functions (reviewed in [18]).

Like many pathways, RLR signaling is tightly regulated to achieve an orchestrated response aimed at maximizing antiviral immunity and minimizing immune- or non-immune-mediated collateral damage. To achieve an appropriately balanced response, down-regulation of antiviral signaling is equally important to its activation. In recent years, there are several biological processes that have been identified as playing crucial roles in the negative regulation of RLR signaling. This review highlights these recent findings on the negative regulatory processes and proteins that directly inhibit RIG-I, MDA5 and MAVS function during virus infection.

I. LGP2 regulates RIG-I and MDA5

Laboratory of Genetics and Physiology 2 (LGP2) is the third member of the RLR family and is a known regulator of both RIG-I and MDA5. LGP2 lacks the CARDs required for signaling, but shares homology with RIG-I and MDA5 within its DExD/H helicase domain and CTD [19–21]. Similar to the other RLRs, the DExD/H helicase domain of LGP2 is capable of ATP hydrolysis and RNA binding [22–24], whereas the CTD confers specificity to RNA ligand recognition and binding [25, 26]. LGP2 can bind a range of RNA ligands, including short, double-stranded RNAs (dsRNA; <2kb) and 5′-triphosphate single-stranded RNA (5′-ppp-ssRNA), which serve as RIG-I ligands [19, 27–29], and long dsRNAs (>2kb), which serve as MDA5 ligands [24, 28].

LGP2 initially was believed to negatively regulate RIG-I and MDA5 [20, 30] (Figure 2). However, these studies were performed prior to the discovery that RIG-I and MDA5 recognize and respond to different viruses (reviewed in [6]). The results from these initial studies showed that LGP2 can inhibit Sendai virus (SeV) and Newcastle disease virus (NDV)-induced IFN-β promoter activity. Both SeV and NDV are now known to be recognized exclusively by RIG-I; so while these early studies suggest that LGP2 negatively regulates RIG-I-mediated signaling, the same may not be true for MDA5-dependent signaling. Indeed, LGP2 was subsequently shown to potentiate MDA5 signaling by binding MDA5-targeted RNA ligands. This increases MDA5 affinity for stimulatory RNA and subsequently enhances innate antiviral signaling [24, 31, 32]. In this context, the ability of LGP2 to modulate MDA5 function appears to depend on intracellular expression levels – low levels potentiate MDA5 activation whereas high levels inhibit MDA5 activation [24, 31]. However, these findings have only been demonstrated through ectopic expression of LGP2 and have not been corroborated with endogenous LGP2 during virus infection.

Figure 2. Negative regulators of RIG-I and MDA5.

Negative regulators act at multiple points along the RLR signaling pathway to inhibit activation of RIG-I and MDA5, or to prevent their interaction with MAVS. ARL16, ARL5B, and possibly LGP2 prevent RIG-I and MDA5 from recognizing and binding RNA ligands. A20, NOD2, and LGP2 may inhibit initial activation of RIG-I, whereas TRIM13 and TRIM59 may inhibit MDA5 activation. FAT10 stops activated RIG-I from translocating to MAVS, and Atg5-Atg12, SEC14L1, and USP15 all prevent proper CARD-CARD interaction between RIG-I and MAVS. De-ubiquitinating enzymes, such as USP15, USP3, USP25, and USP21, remove the K63-linked polyubiquitin chains required for RIG-I and MDA5 activation, thus de-activating the RLRs. In contrast, E3 ubiquitin ligases, such as RNF125, RNF122, and c-Cbl K48-ubiquitinate RIG-I and MDA5, targeting them for degradation by the proteasome.

The role of LGP2 in regulating RIG-I-mediated signaling is less well understood, and there are studies supporting both positive [33–35] and negative regulatory functions. For instance, one study found that, in the absence of LGP2, IFN-β production by bone marrow-derived dendritic cells (BMDCs) was diminished upon infection with MDA5-targeted viruses (e.g., encephalomyocarditis (EMCV) and Mengo viruses) and RIG-I-targeted viruses (e.g., SeV and vesicular stomatitis virus (VSV)), suggesting a positive regulatory role for LGP2 [33]. This study also showed a negligible difference in IFN-β production by Lgp2^−/− BMDCs following influenza virus (PR8 ΔNS1) infection, another RIG-I-targeted virus. However, another group observed that infection of Lgp2^−/− MEFs with a human seasonal H3N2 influenza virus resulted in enhanced IFN-β transcription compared to WT MEFs [36], suggesting negative regulation by LGP2. Complementary to this finding, Pothlichet and colleagues demonstrated a decrease in type I IFN and pro-inflammatory mediators during infection with H3N2 in mice overexpressing human LGP2 [37]. Additionally, infection with seasonal H1N1 viruses that activate IRF3, but not those that avoid activation of IRF3, also resulted in increased IFN-β transcription [36]. H1N1 viruses that do not activate IRF3 showed no difference in induction of IFN-β mRNA in WT and Lgp2^−/− MEFs.

Several other studies support LGP2 as a negative regulator of RIG-I signaling. Three distinct mechanisms have been proposed for how LGP2 inhibits RIG-I signaling: (1) LGP2 binds to and sequesters RIG-I ligands. In this model, LGP2 would compete for RIG-I stimulatory RNAs leading to reduced RIG-I activation [20, 30]. While plausible, recent evidence suggests that LGP2 inhibits RIG-I signaling independent of its ability to bind RNA, as mutations within LGP2 that ablate RNA binding and ATP hydrolysis had no effect on its capacity to inhibit RIG-I [23]. (2) The CTD of LGP2 blocks RIG-I activation. In this model, it is believed that the CTD of LGP2 interacts with the CARD regions of RIG-I to inhibit downstream signaling [8]. This hypothesis is based on amino acid sequence homology between the LGP2 and RIG-I CTDs, as the CTD of RIG-I has an autoregulatory function through its interaction with the N-terminal CARDs [8]. In support of this idea, the LGP2 CTD associates with RIG-I in Huh7 but not Huh7.5 cells, the latter of which contain a point mutation (T55I) in the RIG-I CARD [38]. Nonetheless, other studies have shown that the helicase domain of LGP2 also may be necessary for its full regulatory activity [26], and that LGP2 may not even interact directly with RIG-I [30]. (3) LGP2 binds to an activation region within MAVS. Horvath and colleagues found that LGP2 competes for binding to MAVS with IKKε, a key kinase for activating NF-κB [39] (Figure 3). However, subsequent studies have questioned this model as IKKε may not directly bind MAVS, but rather associate indirectly through NEMO or TANK [40, 41]. Furthermore, LGP2 is now known to differentially regulate RIG-I and MDA5 [35], and inhibiting signaling at the level of MAVS might be expected to have a similar effect on both RIG-I- and MDA5-mediated signaling. Thus, despite significant effort from multiple groups, the molecular mechanism underlying the negative regulation of RIG-I by LGP2 still remains uncertain.

Figure 3. Negative regulators of MAVS.

Additional negative regulators prevent or diminish the activity of MAVS, which blocks downstream formation of the signaling synapse and transcription factors. A20 may prevent the initial activation of MAVS, whereas Atg5-Atg12, USP15, and EZH2 prevent CARD-CARD interactions between MAVS and the RLRs by binding the MAVS CARD. TSPAN6, PLK1, and possibly LGP2 inhibit interactions between MAVS and downstream adaptor proteins. PPMA1 de-activates MAVS by de-phosphorylation. PSAM7 and E3 ubiquitin ligases, such as MARCH5, RNF125, SMURF2, and AIP4 target MAVS for degradation by the proteasome. NLRX1 also targets MAVS for degradation, but through the autophagy pathway.

II. Negative Regulation of RIG-I

Ubiquitination

E3 ligases

Ubiquitination, and the opposing process of de-ubiquitination, produce key post-translational modifications that regulate innate immune signaling (reviewed in [35]). Ubiquitination occurs through the activation, conjugation and ligation of ubiquitin to target proteins. E3 ubiquitin ligases recognize and bind a target protein to catalyze the transfer of ubiquitin to a lysine residue on the target protein. Commonly, this process occurs iteratively until a polyubiquitin chain is formed. K63-linked polyubiquitin chains (ubiquitin proteins connected by their lysine 63 residues) generally serve as modifications that enable activation of protein effector functions (e.g. TRIM25 K63-ubiquitination of RIG-I) whereas K48-linked polyubiquitin chains (ubiquitin proteins connected by their lysine 48 residues) often tag target proteins for degradation by the proteasome. In this context, several E3 ligases have been implicated in inhibiting RIG-I signaling following virus infection – RING finger protein 125 (RNF125; also known as TRAC1), RING finger protein 122 (RNF122), Casitas B-lineage lymphoma (c-Cbl) proto-oncogene, and A20 (Figure 2).

RNF125 and RNF122 interact with and ubiquitinate the N-terminal CARDs of RIG-I, targeting RIG-I for proteasomal degradation [42, 43]. It is believed that RNF125 promotes K48-linked ubiquitination of RIG-I, although this has not been demonstrated directly. In comparison, RNF122 was shown to specifically promote K48-ubiquitination of RIG-I lysine residues 115 and 146. After VSV infection, Rnf122^−/− mice produced higher levels of IFN-α and IFN-β which coincided with more efficient control of virus replication than wild-type (WT) congenic mice [43]. Alternatively, c-Cbl is recruited to RIG-I by the lectin SIGLEC-G and the tyrosine phosphatase SHP2, whereby it K48-ubiquitinates the CTD of RIG-I at position K813, also targeting RIG-I for proteasomal degradation [44]. RIG-I protein levels were reduced in the presence of exogenous c-Cbl, but expression was restored upon addition of the proteasome inhibitor MG132.

The protein A20 is an ubiquitin-editing protein with both ubiquitin ligase and de-ubiquitinase activities [45]. However, mechanistically, the negative regulatory function of A20 depends on its ubiquitin ligase, but not its de-ubiquitinase activity, though the nature of ubiquitination (K48 versus K63) has not been identified [46]. Ectopic expression of A20 inhibited RIG-I-mediated antiviral signaling, whereas silencing of A20 gene expression resulted in enhanced ISRE promoter activity. Contrary to expectations, A20 did not directly associate with RIG-I or promote its degradation, and thus the E3 ligase-dependent mode of RIG-I regulation is not understood.

De-ubiquitination

Ubiquitin-specific proteases

In opposition to the process of ubiquitination are de-ubiquitinating enzymes (DUBs), which bind target proteins and cleave the bonds between the ubiquitin proteins and the target. The removal of K48-polyubiquitin chains by de-ubiquitinating enzymes can serve to activate protein function, as in the case of TRIM25 [47] (Figure 1). Alternatively, removal of activating K63-linked polyubiquitin chains is a mechanism of negative regulation. Several DUBs, including a group of ubiquitin-specific proteases (USPs), negatively regulate RLR signaling by targeting the K63 polyubiquitin chains on the N-terminal CARDs that are necessary for RLR activity. In particular, USP3, USP21, USP25 and USP15 have all been shown to directly inhibit RIG-I.

USP3 binds RIG-I and removes the K63-polyubiquitin chains on the N-terminal CARDs [48]. Ectopic expression of USP3 inhibited IFN-β promoter activity in cells treated with low molecular weight (LMW) poly(I:C), and in cells infected with VSV. Gene silencing of USP3 enhanced IFN-β protein expression under similar treatments. Notably, USP3 inhibited signaling by the constitutively active, truncated form of RIG-I (N-RIG-I), but did not inhibit MAVS signaling. Consistent with this, USP3 was shown to directly bind RIG-I CARDs, but not MAVS. Overexpression of USP3 specifically inhibited K63-ubiquitination of RIG-I, while silencing of the USP3 gene resulted in greater K63-ubquitination. Finally, the inhibitory function of USP3 is dependent on its DUB activity. This constitutes a thorough study supporting USP3 as a negative regulator of RIG-I.

Endogenous USP21 was found to associate with RIG-I [49], although a direct interaction has not been established. Ectopic expression of USP21 resulted in reduced ubiquitination of the constitutively active N-RIG-I. Usp21^−/− mice exhibited enhanced antiviral responses to VSV and SeV infection [49]. However, it remains unclear whether USP21 acts exclusively on RIG-I or whether it also de-ubiquitinates and inhibits MDA5 signaling.

USP25 has the ability to remove both K63- and K48-linked polyubiquitin chains from target proteins [50]. Overexpression of USP25 resulted in reduced IFN-β promoter activity and IFN-β expression in SeV-infected or poly(I:C)-treated cells, whereas gene silencing of USP25 enhanced IFN-β promoter activity [50]. These results suggest that USP25 may cleave the N-terminal K63-ubiquitin chains of RIG-I, although again, direct interaction with RIG-I has not been established. Similar to USP21, it will be important to explore whether USP25 also impacts MDA5 signaling in addition to negatively regulating RIG-I.

USP15 was originally identified as a DUB enzyme that interacts with TRIM25, serving to cleave the K48-ubiquitin chain that maintains TRIM25 in an inactive state; thus, de-ubiquitination of TRIM25 by USP15 potentiates TRIM25-dependent activation of RIG-I [47]. However, a more recent study found that USP15 also acts directly on RIG-I to cleave K63-polyubiquitin chains and negatively regulate signaling [51]. In addition, DUB catalytically inactive mutants of USP15 were still capable of inhibiting both SeV- and N-RIG-I-induced signaling, suggesting a secondary mechanism to inhibit RIG-I. Indeed, DUB-defective USP15 was found to directly interact with the N-terminal CARDs of RIG-I and subsequently prevent the interaction between RIG-I and MAVS [51].

ADP-ribosylation factor-like proteins

ARL16

The ADP-ribosylation factors (Arf) comprise a family of small GTP-binding proteins involved in regulating many aspects of membrane trafficking and cytoskeletal reorganization. The Arf-like (Arl) proteins are also members of the Arf family. More than 20 Arl proteins have been identified in humans to date [52] and appear to broadly regulate many biological processes [53]. In particular, ARL16 has been shown to play a role in inhibiting the early stages of RIG-I activation. Chen and colleagues performed a yeast-two hybrid screen with RIG-I as bait and identified an association with ARL16 [54]. Ectopically expressed ARL16 interacted with the CTD of RIG-I (positions 792–925) and negatively regulated downstream activation of the IFN-β promoter. Molecular studies revealed that ARL16 functions in a GTP-dependent manner and likely inhibits signaling by decreasing RIG-I binding to stimulatory RNA ligands.

Autophagy proteins

Atg5-Atg12

Autophagy is a catabolic process that maintains cellular homeostasis through the turnover and recycling of unwanted and damaged cellular material in a lysosomal-dependent manner. In the context of RLR signaling, autophagy is critical for regulating inflammation and maintaining mitochondrial homeostasis [55]. In particular, Okuda and colleagues showed that the Atg5-Atg12 conjugate also regulates RLR signaling. VSV, but not herpes simplex virus (HSV), replication was strongly inhibited in Atg5-deficient cells. VSV infection of Atg5-deficient cells led to enhanced IRF-3 activation, transcription of IFN-β, and restriction by ISGs. [56]. An absence of Atg5 also resulted in accumulation of dysfunctional mitochondria and an increase of reactive oxygen species (ROS) in mitochondria [57]. More recently, ROS production during virus infection was shown to be required for efficient activation of RIG-I signaling and subsequent type I IFN production [58]. In response to dsRNA, Atg5, Atg12 and Atg7 are required for negative regulation of RLR signaling. Molecular studies revealed that Atg5-Atg12 conjugates, rather than their monomeric forms, were required for dampening RLR signaling [59]. Immunoprecipitation studies revealed that Atg5-Atg12 conjugates interact with the CARDs of RIG-I and MAVS. Based on these findings, a model was suggested in which Atg5-Atg12 does not prevent CARD-CARD interaction between MAVS and RIG-I but rather intercalates between MAVS and RIG-I to inhibit RLR-dependent antiviral signaling.

Nod-like proteins

NOD2

The nucleotide binding domain (NBD) and leucine-rich-repeat-containing (NLR) proteins are a family of pattern recognition receptors found within the cytoplasm and are involved in production of IL1β, induction of pyroptosis, and augmentation of type I IFN signaling during virus infection [60]. Following activation, NLRs regulate inflammation during virus infection through the formation of multi-protein complexes, such as the inflammasome complex, to induce inflammation and cell death, or negatively regulate NF-κB or MAPK signaling pathways (reviewed in detail in [61]). NOD2 is an NLR that is predominantly expressed in the cytosol and on the plasma membrane. NOD2 was originally implicated in sensing bacterial peptidoglycan, a major component of the cell wall of Gram-positive bacteria [62]. Most recently, NOD2 has been implicated in regulating RIG-I signaling during virus infection. Bose and colleagues found that NOD2 is required for potentiating RLR signaling, as NOD2 expression enhanced IFN-β promoter activity and IRF-3 activation following virus infection [63]. NOD2 was found to interact with single-stranded viral RNA, as demonstrated through immunoprecipitation studies using ectopically-expressed NOD2 during respiratory syncytial virus (RSV) infection, and in a cell-free system with purified NOD2 and RSV RNA. Furthermore, endogenous NOD2 associated with MAVS, through an interaction involving the leucine-rich-repeat and nucleotide-binding domain of NOD2, during virus infection. The authors proposed a model, wherein, NOD2 serves as a PRR in a similar manner to RIG-I and MDA5 and promotes activation of IRF-3 and NF-κB through its interaction with MAVS. However, these findings are not without controversy as a recent study found that NOD2 negatively regulates RIG-I signaling. In this study, Coyne and colleagues found that NOD2 localizes to cellular ruffles and cell-cell junctions under homeostatic conditions in cultured human intestinal epithelial cells and embryonic kidney cells [64]. Through ectopic expression and in vitro binding assays, RIG-I and NOD2 were found to interact, and this interaction was mapped to the second CARD region of RIG-I and multiple regions within NOD2. NOD2 was found to inhibit N-RIG-I mediated IFN-β and NF-κB promoter activity. In support of this data, NOD expression also inhibited ISG56-luc and NF-κB promoter activity during SeV infection. However, it is still unclear as to how NOD2 functions as both a positive and a negative regulator of RLR signaling during virus infection.

Protein-protein interactions

FAT10

The small ubiquitin-like modifier (UBL) HLA-F adjacent transcription 10 (FAT10) has been described as a signal for proteasome-mediated degradation when conjugated to target proteins, similar to K48-linked ubiquitin [65]. Yoo and colleagues described a conjugation-independent mechanism for FAT10 in negatively regulating RIG-I signaling [66]. Ectopically-expressed FAT10 inhibited RIG-I-mediated signaling in poly(I:C)- and 5’ppp-dsRNA-treated cells and in influenza A virus-infected cells, as well as in the context of constitutively active N-RIG-I. Additionally, exogenous WT and E3 ligase activity-defective FAT10 associated with RIG-I via the T55 residue in the RIG-I N-terminal CARDs. FAT10 appears to aggregate RIG-I, which prevents its translocation to MAVS on the mitochondrial membrane and thus inhibits downstream signaling. This aggregation also appears to prevent the formation of RIG-I-G3BP-containing antiviral stress granules [66].

SEC14L1

SEC14L1, a member of the SEC14 family, appears to negatively regulate RIG-I signaling by associating specifically with the RIG-I N-terminal CARDs [67]. By interacting with the CARDs, SEC14L1 directly competes with MAVS for binding to RIG-I. Co-expression of SEC14L1 with RIG-I or N-RIG-I resulted in decreased RIG-I signaling, whereas gene silencing of SEC14L1 resulted in enhanced signaling. Co-expression of SEC14L1 prevented the interaction of RIG-I and MAVS in a dose-dependent manner, thus inhibiting downstream signaling. Similar phenotypes were observed after infection with SeV and NDV. This study provides a strong basis for further exploration of SEC14L1 and related proteins as regulators of RLR signaling.

III. Negative Regulation of MDA5

Ubiquitination

E3 ligases

RNF125 was found to associate with and ubiquitinate MDA5 [68]. Although the nature of the polyubiquitin chains (K63 versus K48) was not determined, MDA5 protein levels were diminished in the presence of exogenous RNF125, suggesting K48-ubiquitination, which would target MDA5 for proteasomal degradation. Co-expression of MDA5 and RNF125 resulted in a reduction of IFN-β promoter activity, whereas gene silencing of RNF125 increased IFN-β promoter activity [68].

De-ubiquitination

Ubiquitin-specific proteases

In addition to RIG-I, USP3 also binds MDA5 to remove K63-polyubiquitin chains on the N-terminal CARDs [48]. Ectopic expression of USP3 inhibited IFN-β promoter activity in cells treated with high molecular weight (HMW) poly(I:C), whereas gene silencing of USP3 enhanced IFN-β protein expression under similar treatment. Notably, USP3 was shown to directly bind the MDA5 CARDs and inhibited signaling by the constitutively active, truncated form of MDA5 (N-MDA5). Overexpression of USP3 specifically inhibited K63-ubiquitination of MDA5, while silencing of the USP3 gene resulted in greater K63-ubquitination. Finally, mutagenesis studies suggest the inhibitory function of USP3 is dependent on its DUB activity. This constitutes a thorough study supporting USP3 as a negative regulator of MDA5.

ADP-ribosylation factor-like proteins

ARL5B

Similar to the negative regulation of RIG-I by ARL16, ARL5B has been shown to play a role in inhibiting the early stages of MDA5 activation. Through an ectopic expression screen of 18 Arl family members, Kawai and colleagues identified ARL5B, which when co-expressed with MDA5, reduced IFN-β promoter activity [69]. ARL5B is an ISG that associates with and inhibits MDA5-dependent signaling. Of note, ARL5B represses MDA5 signaling independent of its GTP/GDP binding activity. Mechanistically, ARL5B binds to MDA5 and prevents its interaction with dsRNA [69].

Protein-protein interactions

TRIM13

The E3 ubiquitin ligase TRIM13 interacts with and negatively regulates MDA5 function [70]. TRIM13 is expressed in multiple innate immune cell types including murine bone marrow derived macrophages (BMDMs), where its expression increases upon poly(I:C) stimulation and infection with SeV, but not EMCV. In a human cell line, ectopically expressed TRIM13 inhibited MDA5 signaling. Trim13^−/− mice had enhanced antiviral responses and increased survival compared to WT mice after EMCV infection, an MDA5-targeted virus, whereas survival was unchanged in mice infected with VSV, a RIG-I-targeted virus. Although these data collectively suggest that TRIM13 negatively regulates MDA5, the mechanism remains to be elucidated. For example, it is not known whether MDA5 and TRIM13 interact directly or whether the E3 ubiquitin ligase activity of TRIM13 is required for inhibition. Of note, the effects of TRIM13 on RIG-I signaling paradoxically showed that TRIM13 may positively regulate RIG-I function [70].

TRIM59

Kang and colleagues investigated the role of TRIM59 in regulating RLRs and found evidence for negative regulation of MDA5 signaling [70]. TRIM59 was expressed basally in unstimulated BMDMs, but this expression decreased upon stimulation of cells either by poly(I:C) or after infection with SeV or EMCV. Ectopically-expressed TRIM59 inhibited MDA5-induced IFN-β promoter activation. However, as only limited characterization of TRIM59 function was assessed, more work is needed to establish conclusively that TRIM59 negatively regulates MDA5 signaling.

IV. Negative Regulation of MAVS

Ubiquitination

E3 ligases

RNF125, A20, MARCH5 (also referred to as MITOL or RNF153), Smad ubiquitin regulatory factor 2 (SMURF2), and atrophin-1-interacting protein 4 (AIP4) are E3 ubiquitin ligases implicated in inhibiting RLR signaling following virus infection by targeting MAVS (Figure 3).

RNF125 associates with and ubiquitinates MAVS [68] resulting in a reduction of IFN-β promoter activity when RNF125 is co-expressed with MAVS. Gene silencing of RNF125 increased IFN-β promoter activity. Ectopic expression of RNF125 led to a decrease in MAVS protein levels, suggesting K48-ubiquitination by RNF125, however, this was not demonstrated directly. Similarly, the negative regulatory function of A20 is dependent on its ubiquitin ligase activity [45]. Though A20 was shown to inhibit MAVS signaling, MAVS was not targeted for degradation. Thus, at present, the mechanisms by which A20 negatively regulates RLR signaling though MAVS are still poorly understood.

MARCH5, SMURF2 and AIP4 inhibit RLR signaling by K48-ubiquitinating MAVS and promoting its degradation by the proteasome. MARCH5 localizes to the mitochondrial membrane where it interacts with the MAVS CARD and transmembrane domains, and K48-ubiquitinates MAVS on lysine residues K7 and K500 [71]. March5^+/− mice (March5^−/− is embryonic lethal) infected with VSV exhibit enhanced survival and less viral replication compared to their WT counterparts. SMURF2 also directly interacts with MAVS leading to K48-ubiquitination [72]. Overexpression of SMURF2 inhibits SeV-stimulated IFN-β promoter activity at the level of MAVS, whereas gene silencing of SMURF2 resulted in enhanced expression of type I IFN and ISGs. Finally, AIP4 associates with MAVS through the adaptor poly(rC) binding protein 2 (PCBP2) to promote K48-ubiquitination of MAVS, though the exact lysine residue was not identified [73, 74]. However, whereas proteasomal degradation of MAVS was demonstrated in the presence of exogenous PCBP2, this has not yet been directly demonstrated for AIP4.

Autophagy proteins

Atg5-Atg12

As mentioned previously, Atg5-Atg12 conjugates interact with the CARD regions of RIG-I and MAVS [59]. The proposed model suggests Atg5-Atg12 does not prevent CARD-CARD interaction between MAVS and RIG-I but rather intercalates between MAVS and RIG-I to inhibit RLR-dependent antiviral signaling. Recently, it was found that TUFM, an NLRX1-interacting partner, interacts with the autophagy proteins Atg5-Atg12 and Atg16L1, which in turn promotes autophagy and reduces RLR-mediated type I IFN induction [75].

Nod-like proteins

NLRX1

In addition to NOD2, NLRX1 is another NLR protein that has been implicated in regulating RLR signaling during virus infection. NLRX1 is an outer mitochondrial membrane-associated NLR that was simultaneously identified as an inducer of ROS production, which may help potentiate innate immune signaling [76], and a negative regulator of RIG-I signaling [77, 78]. Through both ectopic expression studies and analysis of the endogenous proteins, Ting and colleagues found that the NBD region of NLRX1 interacts with the CARD of MAVS during virus infection to inhibit RLR-dependent signaling [77]. The absence of NLRX1 in mice resulted in enhanced inflammation (as marked by increased IFN-β and IL-6 production) and increased lung pathology (enhanced airway epithelial injury and reduced airway size) during influenza virus infection. This resulted in reduced virus replication, more rapid kinetics of clearance and reduced weight loss in Nlrx1^−/− mice following influenza virus infection [78]. Mechanistically, NLRX1 is believed to interact with TUFM (Tu translation elongation factor, mitochondrial), a mitochondria-associated protein that bridges the autophagy pathway machinery to NLRs by interactions with the Atg12-Atg5 complex and Atg16L1 [75]. However, it remains unclear as to how the direct interaction between NLRX1 and MAVS functions to inhibit RLR signaling [77].

Protein-protein interactions

EZH2

The enhancer of zeste homolog 2 (EZH2), a histone methyltransferase, was proposed to negatively regulate RLR signaling in a methyltransferase-independent manner by interacting with the CARD region of MAVS and preventing MAVS-RIG-I interaction [79]. Overexpression of EZH2, as well as a catalytically inactive mutant, inhibited IFN-β promoter activity induced by overexpression of MAVS and RIG-I, and dampened antiviral responses to influenza A virus infection. Correspondingly, gene silencing of EZH2 resulted in increased MAVS signaling and protection against influenza A virus infection. EZH2 directly interacts with the MAVS CARD, which disrupted the CARD-CARD interaction of RIG-I and MAVS. Also demonstrated in this study, exogenous EZH2 was able to inhibit TLR signaling, endowing it with a broad functional activity.

PLK1

Polo-like kinase (PLK1) binds to the CTD of MAVS and prevents interaction of the downstream factor TRAF3, thus inhibiting TRAF3-mediated activation of IRF-3 [74, 80]. Ectopically-expressed PLK1 inhibited MAVS-mediated IFN-β promoter activity after SeV infection, whereas gene silencing of endogenous PLK1 resulted in enhanced MAVS activity. PLK1 appears to interact with MAVS at two locations: the first, along the N-terminus (between amino acids 180–280), is dependent on phosphorylation of MAVS in this region, and the second, at the CTD (between amino acids 364–470), is phosphorylation-independent. Both interactions require the Polo-box domain (PBD) of PLK1 and contribute to the negative regulation of MAVS. However, only the second interaction in the CTD was found to disrupt the association between MAVS and TRAF3. Further studies are needed to determine the role of the N-terminal PLK1-MAVS interaction and to corroborate these findings in a more biologically relevant context.

TSPAN6

Tetraspanin 6 (TSPAN6) is another MAVS-associated protein that interferes with formation of the MAVS signalosome and impedes downstream signaling [81]. TSPAN6 inhibits RIG-I-, MDA5- and MAVS-dependent IFN-β promoter activity, as well as MAVS-dependent NF-κB and ISRE promoter activity, indicating that TSPAN likely acts at the level of MAVS. Consistent with this hypothesis, gene silencing of TSPAN6 resulted in enhanced RLR signaling after stimulation with poly(I:C), influenza A viral RNA, or infection with SeV. TSPAN6 associated with MAVS, and this association was enhanced after SeV infection, though direct interaction was not confirmed. In the presence of ectopically expressed TSPAN6, TRAF3 did not associate with MAVS, leading to inhibition of downstream signaling. While a promising mechanistic foundation, these functions should be confirmed with endogenous TSPAN6 protein.

PPMA1

Protein phosphatase magnesium-dependent 1A (PPM1A; also referred to as PP2Cα) is a Ser/Thr phosphatase that dephosphorylates MAVS, which negatively regulates its signaling potential [82]. Ectopic expression of PPM1A inhibited MAVS-dependent IFN-β promoter activity, whereas gene silencing of PPM1A resulted in increased promoter activity. Similarly, Ppm1a^−/− BMDMs expressed higher levels of antiviral genes than WT controls after SeV infection, and Ppm1a^−/− mice had enhanced antiviral responses and survival after VSV infection. MAVS is normally phosphorylated by TBK1, which promotes its activation and signaling [17]. PPM1A was found to directly dephosphorylate MAVS both in vitro and in cell culture [82].

PSMA7

The proteasomal subunit PSMA7 was found to negatively regulate RLR signaling by interacting with MAVS and promoting its degradation by the proteasome [83]. PSMA7 is an α-type subunit of the 20S core complex of the proteasome and has been shown to bind target proteins for proteasomal degradation [84]. PSMA7 interacts with MAVS in vitro and in the context of ectopically or endogenously expressed protein in cells. Both the N-terminal CARD and transmembrane (TM) domain of MAVS were required for this interaction.

CONCLUSION

Negative regulation of antiviral signaling is critical for reducing inflammation, preventing immune-mediated pathology, avoiding persistent inflammation, and returning to a homeostatic state following viral infection. These counter-regulatory cellular processes are especially important for reducing inflammation in immune-tolerant organs, including the heart, central nervous system, eyes, and testis. The RLRs were discovered over a decade ago, and since then a multitude of positive and negative regulators of this signaling pathway have been proposed. Positive regulators are crucial for the activation of RLR signaling, an integral part of the innate antiviral response responsible for controlling and clearing viral infections. Negative regulation of this response is equally important to prevent immune-mediated tissue damage and spontaneous autoimmunity. Indeed, studies have shown that hyper-responsive or constitutive RLR signaling can lead to autoimmune diseases such as Aicardi-Goutières syndrome (AGS), systemic lupus erythematosus (SLE), and Singleton–Merten syndrome (SMS) (reviewed in [85]). Further study of specific negative regulators is required to determine whether each is capable of directly preventing such conditions through their regulation of the RLR pathway.

We highlighted key negative regulators that directly impact RIG-I, MDA5, and MAVS-dependent signaling. Through genetic screens, protein-protein interaction studies, and viral infection of gene-targeted mice and cells, several pathways and biological processes have been linked to controlling RLR signaling, including ubiquitination/de-ubiquitination, protein trafficking, autophagy, inflammasome-related proteins and competition for binding downstream signaling factors within cells. While these studies have been critical for understanding the regulation of RLR antiviral immune signaling, several unanswered questions remain: (1) Do negative regulators function in a cell type-specific manner? Many of these genes have been studied using global knockout mice. The use of conditionally gene-targeted mice will enhance our understanding of the regulation of the RLR pathway within key innate immune cells, including dendritic cells, monocytes and macrophages as well as within specific organs; (2) What role do negative regulators play in mediating non-canonical signaling functions of RLRs? Our group reported that LGP2 and MAVS function within CD8⁺ T cells to promote proliferation and effector function. Additionally, LGP2 has an important role in promoting CD8⁺ T cell survival during virus infection [34], suggesting the RLRs likely possess functions apart from their defined roles in promoting antiviral responses. It could be informative to study the negative regulators with as yet undefined, or poorly understood mechanisms in the context of non-canonical RLR signaling to determine whether their functions are more apparent in these pathways; (3) How do host genetics impact the functions of RLR negative regulators? Human studies have revealed single nucleotide polymorphisms (SNPs) within innate immune genes that can influence susceptibility to viral infection or autoimmunity (reviewed in [86]). For instance, AGS has been associated with increased production of type I IFN, otherwise known as interferonopathies. Mutations in genes encoding nucleic acid metabolism or sensing (TREX1, RNASEH2A, RNASEH2C, SAMHD1 and ADAR) were found to associate with a type I IFN signature in AGS patients [4]. However, few studies have evaluated the presence of SNPs within negative regulators or their impact on RLR regulation and protein function. Several mutations within NOD2 have been identified as risk factors for developing Crohn’s disease [3, 5]. More recently, a frameshift mutation within NOD2 corresponded to an increase in the negative regulatory activity of NOD2 and subsequent dampening of RIG-I antiviral signaling [64]. Genetic studies with collaborative cross (CC) animals might rapidly identify novel regulators of RLR signaling and polymorphisms within known regulators that affect protein function [87–89].

Table 1.

Negative regulators of RIG-I, MDA5 and MAVS function

Protein	Target	Mechanism	Ref.
LGP2	RIG-I MAVS	Binds and sequesters RIG-I ligands CTD blocks RIG-I activation Binds to an activation region within MAVS	[8, 20, 30, 39]
RNF125	RIG-I MDA5 MAVS	Interacts with and ubiquitinates RIG-I, MDA5 and MAVS Proteasome-mediated degradation May lead to K48-linked ubiquitination	[42, 68]
RNF122	RIG-I	Interacts with the N-terminal CARD region of RIG-I Promotes K48-ubiquitination of RIG-I K115 and K146 Proteasome-mediated degradation	[43]
c-Cbl	RIG-I	Recruited to RIG-I by SIGLEC-G and SHP2 K48-ubiquitinates C-terminal domain of RIG-I at position K813 Proteasome-mediated degradation	[44]
A20	RIG-I MAVS	E3 ligase activity dependent/DUB-independent RIG-I and MAVS not degraded Mechanism unknown	[46]
USP3	RIG-I MDA5	Binds to RIG-I and MDA5 Removes K63-linked polyubiquitin chains on CARDs	[48]
USP21	RIG-I	Associates with RIG-I Removes ubiquitin from RIG-I CARD region	[49]
USP25	RIG-I	Removes ubiquitin from RIG-I	[50]
USP15	RIG-I	Removes K63-polyubiqutin chain of RIG-I Interacts with RIG-I CARDs to prevent RIG-I-MAVS association	[51]
ARL16	RIG-I	Interacts with C-terminal region of RIG-I (aa 792–925) Functions in a GTP-independent manner May reduce RIG-I binding to ligands	[54]
ATG5-ATG12	RIG-I MAVS	Intercalates between MAVS and RIG-I CARDs to inhibit signaling Interacts with TUFM to promote autophagy	[59, 75]
NOD2	RIG-I	Interacts with RIG-I through the 2nd N-terminal CARD region Mechanism unknown	[64]
FAT10	RIG-I	Independent of E3 ligase activity Associates with RIG-I Prevent translocation to mitochondrial membrane Inhibits formation of RIG-I-G3BP-containing antiviral stress granules	[65, 66]
SEC14L1	RIG-I	Associates with RIG-I CARDs Competes with MAVS for binding to RIG-I	[67]
ARL5B	MDA5	Associates with MDA5 Reduces MDA5 binding to dsRNA	[69]
TRIM13	MDA5	Mechanism unknown	[70]
TRIM59	MDA5	Mechanism unknown	[70]
MARCH5	MAVS	Interacts with MAVS (CARDs and transmembrane domain) Promotes K48-ubiquitination of MAVS K7 K500 Proteasome-mediated degradation	[71]
SMURF2	MAVS	Interacts with MAVS Promotes K48-ubiquitination of MAVS Proteasome-mediated degradation	[72]
AIP4	MAVS	Associates with MAVS through the adaptor PCPB2 Promotes K48-ubiquitination of MAVS (site unknown) Proteasome-mediated degradation	[73, 74]
NLRX1	MAVS	Nucleotide binding region interacts with MAVS CARDs Interacts with TUFM, ATG5-ATG12 and ATG16L1 May promote autophagy	[75, 77, 78]
EZH2	MAVS	Associates with MAVS CARDs Prevents RIG-I-MAVS interaction	[79]
PLK1	MAVS	Interacts with MAVS N-terminus (aa 180–280) Associates with MAVS CTD (aa 364–470) Prevents interaction of TRAF3 with MAVS	[74, 80]
TSPAN6	MAVS	Associates with MAVS Possibly disrupts TRAF3-MAVS interaction (?)	[81]
PPMA1	MAVS	Dephosphorylates MAVS	[82]
PSAM7	MAVS	Interacts with MAVS (through CARD and transmembrane regions) Promotes degradation	[83]