Ubiquitin in the activation and attenuation of innate antiviral immunity

RIG-I and melanoma differentiation-associated gene 5 receptor (MDA5), together with the regulatory homologue laboratory of genetics and physiology 2 (LGP2), form the apex of the RLR signaling cascade. All three are expressed ubiquitously, but only RIG-I and MDA5 possess N-terminal caspase activation and recruitment domains (CARDs) that are capable of downstream immune signal transduction.

Based on their distinctive C-terminal domains that sense different types of invasive RNA, the RLRs are activated by different viruses. RNA lacking a 5′-7-methylguanosine cap, a feature of mature eukaryotic RNA, potently activates MDA5-mediated signaling (Züst et al., 2011). Except during viral infection, the mammalian cytosol is normally vacant of immature 5′-triphosphorylated RNA, certain kinds of which activate RIG-I (Schlee et al., 2009).

RLR activation up-regulates expression of two IFN-I isotypes, IFN-α and IFN-β, which regulate transcription of hundreds of ISGs during infection (Fig. 2, middle). IFN regulatory factors (IRFs), including IRF3 and IRF7, dimerize and translocate into the nucleus to drive transcription of various IFN-α/β subtypes upon phosphorylation by TNF receptor–associated factor (TRAF) family member–associated NF-κB activator (TANK)–binding kinase 1 (TBK1). TBK1 localization changes from the cytosol to distinct subcellular compartments depending on upstream signaling events (Goncalves et al., 2011). In promoting IFN-I signaling, TBK1 associates with RIG-I as well as key adaptor proteins, including TANK and NF-κB essential modulator (NEMO; Guo and Cheng, 2007; Zhao et al., 2007; Wang et al., 2012a). This facilitates interactions between TBK1, inhibitor of NF-κB kinase subunit ε (IKKε) and TRAFs, particularly TRAF3. Upon RLR activation, these proteins are colocalized at the cytosolic surface of the mitochondrial outer membrane (Parvatiyar et al., 2010; van Zuylen et al., 2012), coordinated by the mitochondrial antiviral-signaling protein (MAVS; also termed VISA/Cardif/IPS-1).

At rest, the RLR cascade is maintained in an inactive but tensioned state through an intricate negative feedback system involving protein expression levels, conformational changes, compartmentalization, and PTMs. Part of this system operates at the receptor through conformational auto-inhibition of the RIG-I CARDs. The RIG-I C-terminal repressor domain (RD) audits the cytosol for viral RNA, binding of which induces a major structural rearrangement in the RD and CARD (Saito et al., 2007). Conversely, MDA5 oligomerizes along the length of RNA ligands, forming immunogenic filaments that are potentiated by ATP hydrolysis and interaction with LGP2 (Peisley et al., 2012; Bruns et al., 2014).

The unfurled RIG-I CARDs undergo tetramerization upon K63-linked polyubiquitination or unanchored polyubiquitin chain association (Peisley et al., 2014). These modifications drive mitochondrial accumulation of RIG-I, promoting CARD–CARD interactions with MAVS and inducing its oligomerization and filamentation. The RIG-I CARDs contain a high proportion of hydrophobic residues and are prone to aggregation, thus oligomerization and polyubiquitination may stabilize the activated CARDs or elicit a separate mitochondrial targeting signal. Conversely, ubiquitination has no known role in MDA5 or LGP2 activation.

The first virus-triggered RIG-I ubiquitination site described, K172, depends on the E3 activity of tripartite motif protein 25 (TRIM25; Gack et al., 2007). Plausibly as a means of restricting escape mutant selection, this activation mechanism now appears to have evolved with partial redundancy using alternate E3s. TRIM4 was recently described to modify this same site in addition to two other CARD residues: K154 and K164 (Yan et al., 2014). Furthermore, these same three residues are reportedly ubiquitinated by really interesting new gene (RING) finger protein-135 (RNF135; also termed Riplet/REUL; Gao et al., 2009), although this is controversial (Fig. 3; Oshiumi et al., 2010). Underscoring the importance of these modifications, ubiquitin-specific protease 3 (USP3) and ubiquitin C-terminal hydrolase are DUbs that inhibit IFN-I production by removing such chains from RIG-I (Friedman et al., 2008; Cui et al., 2014).

Both TRIM25 and RNF135 are targets of the influenza A virus (IAV) nonstructural protein 1 (NS1), which blocks their E3 activity and ubiquitin-dependent RIG-I activation (Fig. 3; Gack et al., 2009; Rajsbaum et al., 2012). IAV-NS1 binds the central coiled-coil domain (CCD) of TRIM25 and is postulated to prevent CCD-mediated homo-oligomerization. Although the NS1-binding site on RNF135 is unknown, RNF135 and TRIM25 share a similar RING-CCD-B30.2/SPRY (sp1A and ryanodine receptors) domain organization. However, the CCDs differ in size and sequence markedly, suggesting that IAV-NS1 may bind multiple sites on TRIM25 and RNF135. Alternatively, given that CCDs often mediate protein–protein interactions, IAV-NS1 may sense and subvert CCD-interacting domains more broadly. Notably, the IAV-NS1:RNF135 interactions observed by Rajsbaum et al. (2012) were strain dependent.

RNF135 enables RIG-I CARD activation by TRIM25 upon ubiquitinating RD residues K849 and K851. RNF135 knockdown inhibits interaction between RIG-I:TRIM25 and eliminates TBK1 recruitment (Oshiumi et al., 2009), revealing an ordered functional interplay between ubiquitination and phosphorylation in coordinating RIG-I activation. Hepatitis C virus (HCV) NS3-4A protease exploits this concept by targeting RNF135 for proteolytic cleavage (Oshiumi et al., 2013). Furthermore, numerous herpesviruses encode their own DUbs that inhibit IFN-I expression by stripping ubiquitin modifications from activated RIG-I (Inn et al., 2011b). Accordingly, HCV and herpesvirus infections are treatable with IFN (Oberman and Panet, 1988; Nguyen et al., 2014), although this can carry significant side effects. Endogenous IFN-I expression and self-regulation may be restored by defeating such mechanisms of viral antagonism.

RLR signaling is also counterbalanced and diminished through ubiquitin modification as the antiviral posture becomes unnecessary. RNF125 forms part of this process, ligating K48-linked polyubiquitin chains to the activated CARD of RIG-I and MDA5, leading to proteasome-mediated degradation of both receptors and diminished IFN-I signaling. USP4 is a DUb that sustains RLR signaling by specifically removing such chains (Wang et al., 2013a). In the same way, RNF125 ubiquitinates and degrades activated MAVS (Arimoto et al., 2007), suggesting that RNF125 is an E3 that destabilizes proteins containing activated CARDs. Given how commonly CARD-containing proteins and their homotypic interactions feature in immune signaling pathways (Bouchier-Hayes and Martin, 2002), RNF125 may represent a general immune signaling antagonist. Conversely, the 52-kD repressor of the inhibitor of the protein kinase (p52rIPK) binds and enhances the stability of RIG-I by blocking its ubiquitin-mediated degradation (Now and Yoo, 2011). Accordingly, the properties of p52rIPK or RNF125 may be exploitable in the treatment of viral infections or autoimmune disorders.

The linear ubiquitin assembly complex (LUBAC), containing SHANK-associated RH domain–interacting protein (SHARPIN), heme-oxidized IRP2 ubiquitin ligase 1L (HOIL-1L), and HOIL-1–interacting protein (HOIP), was proposed to negatively regulate RLR-mediated IFN-I expression via two independent mechanisms (Inn et al., 2011a). First, HOIL-1L competes with TRIM25 for RIG-I CARD binding, abrogating the RIG-I:MAVS interaction. Second, HOIP promotes M1- and K48-linked polyubiquitination of TRIM25 and induces its proteasomal degradation, thereby decreasing TRIM25-mediated activation of RIG-I. If LUBAC were capable of ligating K48-linked polyubiquitin chains to substrates, TRIM25 would be the first example to our knowledge.

Another route of RLR inhibition involves tetraspanin-6 ubiquitination by an unknown E3. During RLR activation, polyubiquitinated tetraspanin-6 is recruited to MAVS and blocks the RLR:MAVS interaction, thereby impeding recruitment of the downstream signaling apparatus (Wang et al., 2012b).

Mitochondria, peroxisomes, and endoplasmic reticulum function as immune signaling platforms linking viral pattern recognition with effector molecule production (Fig. 2, middle). Although this process remains poorly characterized, the nature and context of viral ligands detected by PRRs drives accumulation of the downstream signaling apparatus to these platforms. Adaptor proteins mediate this accumulation: mitochondria and peroxisomes by MAVS (Dixit et al., 2010) and endoplasmic reticulum by stimulator of IFN genes (STING; Ishikawa and Barber, 2008). Cyclic GMP-AMP synthase (cGAS) and AIM2-like receptors (ALRs), which include AIM2 and human IFN-inducible protein 16 (IFI16), have also emerged as important DNA virus PRRs. cGAS signals via STING and AIM2 generates inflammasome oligomers, whereas IFI16 can stimulate both signaling mechanisms (Diner et al., 2015).

RNA-activated RIG-I and MDA5 colocalize with MAVS, inducing its filamentation. It remains unclear why these filaments are potent inducers of downstream signaling; however, RLR cascade proteins including NEMO, IKKε, and various TRAFs possess MAVS-targeting signals (Paz et al., 2011). Furthermore, TBK1 and other key RLR cascade proteins interact with these proteins but are activated only upon oligomerization. Thus, steady state isolation of MAVS may represent a spatiotemporal barrier that restrains innate immune signaling, overcome through coordinating these proteins into signaling complexes upon MAVS multimerization. In this way, it is conceivable how immunomodulating E3s/DUbs may be compartmentalized together with their substrates.

The central position that MAVS occupies within the RLR cascade is commensurate with the many PTMs that modulate its role. To our knowledge, MAVS ubiquitination has not been observed in resting cells using a variety of proteomic and biochemical approaches, indicating that MAVS ubiquitination occurs specifically during viral infection. At least seven E3s ubiquitinate MAVS, leading to MAVS degradation in almost every case, as described later in this section (Fig. 3). At least five of these modify other substrates within the same cascade, highlighting MAVS as a crucial locus of RLR regulation. Accordingly, MAVS is targeted by numerous viruses in a variety of ways; however, with the exception of HBV and severe acute respiratory syndrome–coronavirus (SARS-CoV; Fig. 3; Wei et al., 2010; Shi et al., 2014), this is usually achieved by means other than manipulating MAVS ubiquitination, likely given the extensive ubiquitin-mediated negative regulatory systems already in place.

MAVS aggregation is a key feature of RLR cascade activation, but how these aggregates are resolved during deactivation is only beginning to be clarified. In addition to ubiquitinating RIG-I and enhancing its association with MAVS, TRIM25 ubiquitinates MAVS at Lys7 and Lys10 and induces its partial proteolysis (Castanier et al., 2012). This was proposed as a means of discharging the activated RLR signalosome from the mitochondrial recruitment platform and would begin to address how IRF3 and other RLR signalosome components traffic correctly after activation. More recently, Lys7 and Lys500 were shown to be polyubiquitinated by membrane-associated RING finger protein 5 (MARCH5), a mitochondrial membrane-bound E3 that effectively dissolves MAVS aggregates by specifically targeting them for degradation. MARCH5 is an important regulator of mitochondrial fission and fusion whose expression is up-regulated during infection (Yoo et al., 2015). These mechanisms of MAVS aggregate resolution may be nonredundant, with the TRIM25 mechanism occurring throughout the immune response and the MARCH5 mechanism amplifying gradually in an IFN-I negative feedback loop.

Complete MAVS degradation is independently promoted by the E3s RNF5, RNF125, atrophin-1–interacting protein 4 (AIP4; also termed ITCH), SMAD ubiquitination regulatory factor 1 (Smurf1), and Smurf2 (Fig. 3). RNF5 polyubiquitinates MAVS at Lys362 and Lys461, whereas the adjacent residues Lys371 and Lys420 are polyubiquitinated by AIP4 upon recruitment by poly(rC)-binding protein 1 (PCBP1) or PCBP2 (You et al., 2009; Zhong et al., 2010; Zhou et al., 2012). AIP4 additionally inhibits IFN-I as well as NF-κB activation by ubiquitinating the inhibitor of apoptosis protein 1 (cIAP1), targeting it for lysosomal degradation (Tigno-Aranjuez et al., 2013). cIAP1 is an E3 that activates TRAF3/6 during viral infection (Mao et al., 2010), revealing that AIP4 broadly and multiply inhibits NLR-, RLR-, and TLR-mediated immune signaling. The acceptor site or sites for RNF125-induced MAVS ubiquitination are unknown; however, given that RNF125 also ubiquitinates the activated CARDs of RIG-I and MDA5 (Arimoto et al., 2007), the MAVS CARD appears a likely candidate. NEDD4 family–interacting protein 1 (Ndfip1) binds MAVS and recruits Smurf1 and possibly Smurf2, facilitating ubiquitination of unknown sites within MAVS (Wang et al., 2012c; Pan et al., 2014). Moreover, numerous TRAFs, including TRAF3 and TRAF6, interact with MAVS, and Smurf1 also targets these for degradation (Li et al., 2010a). Finally, Lys500 was reported as a single site of IFN-I–activating polyubiquitination by an unknown E3, inhibiting NF-κB activation by sequestering IKKε (Paz et al., 2009).

The TRAFs are six multifunctional adaptor proteins that regulate both NF-κB activation and IFN-I expression via the RLR, NLR, and TLR protein signaling cascades (Fig. 2). TRAF-mediated signaling outcomes are augmented by ubiquitin, and, excepting TRAF1, all TRAFs possess a RING finger domain and multiple zinc coordination sites, features typical of ubiquitin E3s. K63-linked autoubiquitination at Lys124 is a key activation mechanism of TRAF6 (Lamothe et al., 2007; Jiao et al., 2015) and possibly TRAF2 (Habelhah et al., 2004), TRAF4 (Marinis et al., 2012), and TRAF5 (Zhong et al., 2012). In vitro TRAF3 ubiquitination assays and analysis of recombinant TRAF3ΔRING isolated from mammalian cell lysates are also consistent with an autoubiquitination activation mechanism for TRAF3 (Kayagaki et al., 2007; Zeng et al., 2009).

TRAF3 and TRAF6 are among the first molecules activated by MAVS in the RLR pathway (Fig. 2, middle). Furthermore, there is increasing evidence of ubiquitin-mediated cross-talk between TRAFs. TRAF3 promotes IFN-I expression by activating TBK1/IRF3 (Parvatiyar et al., 2010), whereas TRAF6 activates mitogen-activated protein kinase kinase kinase 1 (MEKK1) to activate NF-κB, which also enhances IFN-I expression (Yoshida et al., 2008). Simultaneously, TRAF3 suppresses NF-κB by inhibiting IKK activation upon binding TRAF2 (Zarnegar et al., 2008), likely as a mechanism to skew innate immune effector molecule expression as required. Inversely, the E3 cIAP2, after itself being ubiquitinated by TRAF6, promotes TRAF3 degradation by ligating K48-linked polyubiquitin chains to TRAF3 at residues K107 and K156, thereby restoring NF-κB activation (Tseng et al., 2010). However, as well as degrading TRAF3, cIAP1/2 can also activate TRAF3 by catalyzing its K63-linked polyubiquitination (Fig. 3; Mao et al., 2010). This suggests that the context-dependent ubiquitination state of cIAP1/2 determines its effect on TRAF3. The E3 RNF166 was recently reported to ubiquitinate and activate both TRAF3 and TRAF6 (Chen et al., 2015). Finally, the RIG-I–activating E3 TRIM25 was reported to enhance MDA5-mediated NF-κB activation at the level of TRAF6 (Lee et al., 2015), although mechanistic details remain unclear.

TRAF-mediated signaling is also terminated by ubiquitin in numerous ways. HSV encodes the DUb UL36USP, which strips K63-linked polyubiquitin chains from TRAF3 to prevent downstream protein recruitment (Fig. 3; Wang et al., 2013b), possibly antagonizing cIAP1/2-mediated ubiquitination. TRAF3 and TRAF6 are both deactivated by the DUbs otubain 1 (OTUB1) and OTUB2, which remove K63-linked polyubiquitin chains (Li et al., 2010b). TRAF3 is further deactivated by the deubiquitinase DUBA, which removes K63-linked polyubiquitin chains from TRAF3 (Kayagaki et al., 2007). Furthermore, the E3 Triad3A redirects TRAF3 to the proteasome by ligating K48-linked polyubiquitin chains (Nakhaei et al., 2009). Altogether, this constitutes a ubiquitin-dependent feedback mechanism that enables TRAFs to dictate the direction of immune signal transmission in a context-dependent manner.

In contrast to the three RLR receptors, the 22 NLRs have diverse expression patterns and largely under-characterized functions. The NLRs are well recognized for their roles in regulating NF-κB activation and antibacterial immunity; however, at least five members have emerging roles in antiviral immune signaling: NOD1, NOD2, NLRC5 (NLR family CARD domain–containing protein 5), NLRP4 (NACHT, LRR, and PYD domain–containing protein 4), and NLRX1 (NLR family member X1; Fig. 2, right). Although NLRs recruit E3s and modulate the ubiquitination of other proteins, including several in the RLR cascade, the role of PTMs in NLR regulation remains under-defined.

The PRRs NOD1 and NOD2 are the best characterized NLRs. NOD1 is expressed ubiquitously, whereas NOD2 is expressed mainly in cells of myeloid and lymphoid origin and is up-regulated during bacterial and viral infection. The classic NOD2-activating ligand is bacterial muramyl dipeptide (MDP), which promotes NF-κB activation. However, NOD2 also promotes IFN-I expression during infection by numerous RNA viruses, in part through recognizing single-stranded RNA (ssRNA) and interacting with MAVS. NOD2 may also promote IFN-I expression during infection by particular DNA viruses by an undefined mechanism (Sabbah et al., 2009; Kapoor et al., 2014). Accordingly, NOD2 dysfunction leads to inefficient innate and adaptive immune responses to viral infection (Lupfer et al., 2014).

NOD2 features regularly in the immune signaling landscape, yet mechanisms of NOD2 regulation and cross-talk are only beginning to be revealed. Upon activation by MDP, NOD2 is ubiquitinated by TRIM27, leading to NOD2 degradation and NF-κB inhibition (Zurek et al., 2012). NOD2 signaling is further suppressed by AIP4, which ubiquitinates Lys209 of receptor-interacting serine/threonine protein kinase 2 (RIPK2), the immediate downstream interacting partner of NOD2 (Fig. 2, right; Tao et al., 2009). Conversely, NOD2-driven NF-κB activation is enhanced by LUBAC, a negative regulator of RLR signaling, as well as X-linked IAP (XIAP), which respectively ligate M1- and K63-linked polyubiquitin chains to NOD2 and RIPK2 (Damgaard et al., 2012). These activating ubiquitin chains may be antagonized by the ubiquitin-editing enzyme A20 (Hitotsumatsu et al., 2008), which also disrupts ubiquitin-mediated TBK1 activation in the RLR signaling cascade as well as ubiquitin-mediated TRAF6 activation in the TLR cascade (Turer et al., 2008; Parvatiyar et al., 2010).

Another mitochondrial link between the NLR and RLR signaling pathways is the ubiquitously expressed NLRX1 (Fig. 2, right), whose role in IFN-I regulation is controversial. NLRX1 is localized to the mitochondrial outer membrane and was reported to inhibit MAVS-dependent IFN-I signaling by blocking the interaction between activated RIG-I/MDA5 and MAVS, although viral replication experiments using gene knockout cells have produced conflicting results (Soares et al., 2013). NLRX1 also potentiates NF-κB signaling by promoting reactive oxygen species production during bacterial infection, linking the mitochondrial immune signaling platform with proinflammatory cytokine generation (Tattoli et al., 2008).

NLRC5 was initially described to enhance IFN-γ and IFN-α expression and inhibit NF-κB and IFN-β, in the latter case through sequestering the activated effector domains of RIG-I and MDA5 (Cui et al., 2010; Kuenzel et al., 2010). NLRC5 has also been shown to bind and inhibit TBK1-mediated IFN-β induction in HEK-293T cells, although NLRC5^−/− mice show relatively normal cytokine responses upon exposure to RLR-, TLR-, and NLR-activating stimuli (Kumar et al., 2011). Still other findings indicate that the RIG-I:NLRC5 interaction also positively regulates IFN-β expression, and this interaction is targeted by the IAV-NS1 protein (Fig. 2, right; Neerincx et al., 2010; Ranjan et al., 2015). These disparate conclusions may reflect cell-specific differences given that NLRC5 is predominantly expressed in hematopoietic cells or differences between mouse and human signaling pathways, suggesting that the NLRC5 regulatory framework is complex. Adding to this, ubiquitination plays an uncharacterized role in regulating NLRC5 upon LPS stimulation and may be induced by NLRC5 overexpression (Cui et al., 2010; Kuenzel et al., 2010). Given the diversity of interactions that NLRC5 takes part in, it is likely that further PTMs will be shown to regulate NLRC5 during viral infection.

NLRP4 has gained prominence as another negative regulator of multiple immune signaling pathways that is more widely expressed than NLRC5. NLRP4 was initially described to inhibit IKKα-mediated NF-κB activation (Fiorentino et al., 2002). Upon RLR cascade activation, NLRP4 also inhibits IRF3 activation by recruiting the E3 deltex-4 (DTX4) to ubiquitinate and degrade TBK1 (Cui et al., 2012), revealing yet another route for RLR/NLR cross-talk (Fig. 2, right).

TLRs are differentially expressed in a wide range of cell populations. TLR3, TLR7, TLR8, and TLR9 are expressed in endosomal vesicles, whereas TLR2 and TLR4 are expressed on the cell surface. TLR3 recognizes double-stranded RNA, activating NF-κB–mediated proinflammatory cytokine production and strongly up-regulating TBK1/IRF3-dependent IFN-I expression. TLR7 and TLR8 recognize ssRNA, up-regulating IFN-α and proinflammatory cytokine production. TLR9 recognizes unmethylated cytosine-phosphate-guanine (CpG) DNA, a common feature of nonmammalian genomes, and stimulates IFN-α production. TLR2 and TLR4 are activated by a variety of microbial ligands, including specific viral proteins, resulting in proinflammatory cytokine expression (Fig. 2, left).

TLR2, TLR4, TLR7, TLR8, and TLR9 signaling is mediated through the adaptor protein myeloid differentiation primary response gene 88 (MyD88; Fig. 2, left). MyD88 recruits NF-κB and IFN-I signaling components, including interleukin-1 receptor–associated kinase 1 (IRAK1), IRAK4, TRAF6, and IRF7. Activated TRAF6 ubiquitinates IRF7, leading to IFN-α expression (Kawai et al., 2004). TRAF6 also promotes K63-linked polyubiquitination of NEMO, enabling recruitment of the TGF-β–activated kinase (TAB)–TAK1 kinase complex (Tseng et al., 2010). Subsequent association between NEMO and M1-polyubiquitin chains induces TAK1-mediated phosphorylation of IKKα and IKKβ, priming them for full transactivation through autophosphorylation (Zhang et al., 2014). Activated IKKα phosphorylates the IκBα subunit, leading to its ubiquitination and proteasomal degradation and releasing NF-κB for nuclear translocation. Furthermore, MyD88, IRAK1/4, and TAB2/3 are also modified and activated by K63-linked polyubiquitin chains. Such chains were recently described as substrates for M1-polyubiquitination by HOIP, resulting in hybrid chains that may connect the MyD88/IRAK and TAK1/IKK signaling apparatus (Emmerich et al., 2013). TLR3 signaling is mediated by TIR domain–containing adaptor-inducing IFN-β (TRIF). TRIF activates TRAF3, which promotes IRF3/IRF7 activation (Tseng et al., 2010), and also TRAF6, which promotes IKK activation (Fig. 2, left; Jiang et al., 2004).

Although TLRs undergo extensive PTM, ubiquitination performs no known role in regulating TLRs directly. Instead, ubiquitination modulates their downstream signaling targets, particularly MyD88, TRIF, and TRAF6, and it is often here that viruses terminate TLR-mediated immunity.

Nrdp1 is an E3 that promotes IFN-I expression at the expense of proinflammatory cytokines. TBK1 polyubiquitination by Nrdp1 activates TBK1 in TRIF-mediated IFN-I expression, which simultaneously K48-polyubiquitinates and down-regulates MyD88-mediated NF-κB activation (Wang et al., 2009). Conversely, the ubiquitin-editing enzyme A20 inhibits ubiquitin-mediated activation of TRAF6, inhibiting NF-κB activation via the TLR and NLR cascades, as well as TBK1, inhibiting IFN-I expression (Hitotsumatsu et al., 2008; Turer et al., 2008; Parvatiyar et al., 2010). Additional avenues of signaling cross-talk include Smurf1 and Smurf2, which degrade MAVS and inhibit IFN-I activation. Smurf1 and Smurf2 also degrade MyD88, inhibiting TLR-mediated NF-κB activation (Lee et al., 2011).

cIAP2 is an E3 that, after itself being ubiquitinated by TRAF6, targets TRAF3 for degradation (Tseng et al., 2010). This is important in promoting TLR4-mediated signaling and cytokine production at the expense of type I IFN production (Zhong et al., 2013). Furthermore, TRAF6 itself promotes proinflammatory cytokine production at the expense of IFN-I. TRAF6 is activated by trans-autoubiquitination at K124, abolition of which eliminates NEMO ubiquitination and TAK1 activation (Lamothe et al., 2007). This mechanism is exploited by HSV, which uses the virally encoded E3 infected cell polypeptide 0 (ICP0) to recruit USP7 to deubiquitinate NEMO and TRAF6 (Daubeuf et al., 2009). In addition, ICP0 directly catalyzes ubiquitination and degradation of MyD88 and TIRAP (Toll/interleukin-1 receptor domain–containing adaptor protein; van Lint et al., 2010). Kaposi’s sarcoma–associated herpesvirus (KSHV) encodes replication and transcription factor (RTA), an E3 that activates latent virus. This activation process involves suppression of antiviral cytokines, partly involving RTA-catalyzed ubiquitination and degradation of MyD88 and TRIF (Ahmad et al., 2011; Zhao et al., 2015), thereby impairing all TLR-mediated immune signaling pathways.

TLR, NLR, and RLR IFN-I signaling converges at IRF activation, the penultimate step toward IFN-I transcription. IRF3 is constitutively expressed in most cell types, residing inactive in the cytosol until phosphorylation by TBK1/IKKε within two activation clusters (Ser385/Ser386 and Ser396/Ser398/Ser402/Thr404/Ser405), resulting in homodimerization, nuclear accumulation, DNA binding, and participation in IFNB gene transcription (Lin et al., 1998). IFN-β acts in an autocrine and paracrine manner upon its cognate receptor, IFN-α/β receptor (IFNAR), thereby activating JAK/STAT signaling and ISG expression. IRF7 expression is up-regulated in this way, which in turn activates IFNA transcription and additional ISG expression by a similar mechanism.

The IRFs are among the most tightly controlled IFN-I signaling proteins through an interplay of PTMs, including phosphorylation, ubiquitination, and ubiquitin-like modifications. Phosphorylation of IRF7 at Ser477 and Ser479 by TBK1/IKKε is required for its activation (tenOever et al., 2004). However, ubiquitination by TRAF6 at nearby residues Lys444, Lys446, and Lys452 appears to be a prerequisite to this and serves as a link between the NF-κB and IFN-I activation pathways (Ning et al., 2008). Furthermore, TRIM28 binds active IRF7 and ligates small ubiquitin-like modifier (SUMO) at two of these ubiquitination sites, Lys444 and Lys446, negatively regulating virus-triggered IFN-α production (Liang et al., 2011), indicating that as yet unidentified DUbs or deSUMOylating enzymes participate in regulating IRF7.

Reminiscent of IRF7, IRF3 residues Lys70 and Lys87 accept both polyubiquitin chains and SUMO, and competition between these modifications can determine the fate of IFN-I signal transduction. At steady state, the SUMO-conjugating enzyme ubiquitin carrier protein 9 (Ubc9) protects IRF3 from ubiquitin-mediated degradation by occupying these sites with SUMO. Alternatively, the deSUMOylating enzyme sentrin-specific protease 2 (SENP2) removes SUMO from IRF3, enabling its K48-linked polyubiquitination (Ran et al., 2011). Subsequent work identified TRIM26 as an E3 that conjugates K48-linked polyubiquitin chains to these same sites (Fig. 3; Wang et al., 2015), triggering degradation of the active, nuclear-localized form of IRF3. Furthermore, activated IRF3 undergoes phosphorylation at Ser339. This promotes interaction with peptidyl-prolyl cis/trans isomerase NIMA-interacting 1 (Pin1), a nuclear-localized protein that promotes IRF3 degradation (Saitoh et al., 2006). The E3 recruited by Pin1 for this purpose is unknown; however, TRIM26 is also localized to the nucleus and seems a strong candidate.

IRF3 degradation is undesirable at early stages of the innate immune response and is limited in several ways. The IRF3-Pin1 interaction is inhibited by the HECT (homologous to the E6-AP C terminus) domain and RCC1-like domain–containing protein 5 (HERC5), which ligates another ubiquitin-like protein, ISG15, onto IRF3 at Lys193, Lys360, and Lys366, thereby sustaining IRF3 activation (Shi et al., 2010). TRIM21 is a ubiquitin E3 described to both inhibit the IRF3-Pin1 interaction and target IRF3 for proteasomal degradation (Higgs et al., 2008; Yang et al., 2009). TRIM21 reportedly also targets IRF7 for degradation upon TLR7 or TLR9 activation (Higgs et al., 2010), although the TRIM21-dependent IRF3/IFR7 ubiquitin acceptor sites remain undefined (Fig. 3).

Several additional E3s regulate IRF abundance. The Skp-Cullin–F-box (SCF)–containing complex, of which cullin1 (Cul1) is a core component, catalyzes IRF3 degradation as well as IκB degradation, promoting NF-κB activation (Fig. 3; Bibeau-Poirier et al., 2006). RanBP-type and C3HC4-type zinc finger–containing protein 1 (RBCK1) catalyzes K48-linked polyubiquitination and degradation of IRF3 during viral infection (Fig. 3; Zhang et al., 2008). Finally, the forkhead box protein O1 (FoxO1), a regulator of insulin signaling, binds IRF3 and promotes its degradation by recruiting an unknown E3. FoxO1 also negatively regulates IRF7 transcription (Lei et al., 2013), altogether implying a link between metabolism and innate immune induction. Expression of Cul1 and the E3s RBCK1, TRIM21, TRIM26, and HERC5 is IFN-I inducible (Henig et al., 2013), constituting a multiply redundant negative feedback web in which IFN-I expression is self-restraining.

IRFs are a significant target of viral disruption, usually resulting in their proteasome-mediated degradation. Rotavirus (RoV) nonstructural protein 1 blocks NF-κB signaling and usurps the ubiquitin modification system to redirect IRF3/5/7/9 to the proteasome in a strain-specific manner (Fig. 3; Morelli et al., 2015). Cells produce trace quantities of IFN-I at rest through basal activation of endogenous IRF3/IRF7, the intracellular concentration of which are regulated by the E3 RTA-associated ubiquitin ligase (RAUL). KSHV exploits this mechanism to diminish immune signaling, recruiting USP7 to deubiquitinate RAUL and thereby maintain RAUL-mediated IRF3/IRF7 degradation (Yu and Hayward, 2010). The RAUL-dependent ubiquitin acceptor sites on IRF3/IRF7 remain unknown (Fig. 3), but better characterization of the RAUL-IRF interaction may have implications for antiviral and autoimmunity treatments. Furthermore, the KSHV RTA protein catalyzes polyubiquitination and degradation of IRF7 and MyD88 (Yu et al., 2005). Thus, KSHV effectively terminates several signaling pathways at multiple stages.

Similar to KSHV, HIV infection fails to stimulate activation of IRF3, endogenous levels of which are quickly reduced upon infection. Underscoring the importance for HIV to disrupt early IFN-I–mediated immunity, IRF3 degradation is independently promoted by two viral accessory proteins, viral infectivity factor (Vif) and viral protein R (Vpr). The E3s hijacked for this purpose are unknown (Fig. 3), although Vif and Vpr recruit SCF-related components to degrade other antiviral proteins (Okumura et al., 2008).

The innate immune signaling architecture is complex and has coevolved with the pathogens it guards against, meanwhile restraining autoimmunity through an elaborate negative feedback scheme. A cornerstone of this dynamic regulatory framework is the ubiquitin modification system, which is manipulated by viruses relevant to human disease. Going forward in understanding mechanisms of infection and autoimmunity, we must address significant gaps in knowledge regarding the specificity and context-dependent regulation of E3s and DUbs and the consequences of ubiquitin modification. This will expose further cross-talk between the immune signaling cascades, revealing a functional and self-regulating whole. In the search for a new generation of antiviral and autoimmune treatments, we continue to learn from the pathogens that have long adapted to exploit this ready-made system of functional regulation; humans possess hundreds of specific- and general-effect E3s and DUbs, many of which could be harnessed for therapeutic use.

We are grateful to Dr. Sarah Atkinson for critical review of the manuscript.

N.A. Borg is an Australian Research Council (ARC) Future Fellow (FT110100223).

The authors declare no competing financial interests.