TRAF molecules in cell signaling and in human diseases

The tumor necrosis factor receptor (TNF-R)-associated factor (TRAF) family of intracellular proteins were originally identified as signaling adaptors that bind directly to the cytoplasmic regions of receptors of the TNF-R superfamily. The past decade has witnessed rapid expansion of receptor families identified to employ TRAFs for signaling. These include Toll-like receptors (TLRs), NOD-like receptors (NLRs), RIG-I-like receptors (RLRs), T cell receptor, IL-1 receptor family, IL-17 receptors, IFN receptors and TGFβ receptors. In addition to their role as adaptor proteins, most TRAFs also act as E3 ubiquitin ligases to activate downstream signaling events. TRAF-dependent signaling pathways typically lead to the activation of nuclear factor-κBs (NF-κBs), mitogen-activated protein kinases (MAPKs), or interferon-regulatory factors (IRFs). Compelling evidence obtained from germ-line and cell-specific TRAF-deficient mice demonstrates that each TRAF plays indispensable and non-redundant physiological roles, regulating innate and adaptive immunity, embryonic development, tissue homeostasis, stress response, and bone metabolism. Notably, mounting evidence implicates TRAFs in the pathogenesis of human diseases such as cancers and autoimmune diseases, which has sparked new appreciation and interest in TRAF research. This review presents an overview of the current knowledge of TRAFs, with an emphasis on recent findings concerning TRAF molecules in signaling and in human diseases.


Background
The tumor necrosis factor receptor (TNF-R)-associated factor (TRAF) family of intracellular proteins were originally identified as signaling adaptors that bind directly to the cytoplasmic regions of receptors of the TNF-R superfamily [1][2][3]. There are six known members of the TRAF family (TRAF1 to 6) in mammals. Although a novel protein was named TRAF7 [4], this claim is controversial as the protein does not have the TRAF homology domain that defines the TRAF family ( Figure 1). The distinctive feature of all TRAF proteins is a C-terminal TRAF domain, which is composed of an N-terminal coiled-coil region (TRAF-N) and a C-terminal β-sandwich (TRAF-C) [2,3,5]. The TRAF domain mediates protein-protein interactions, including TRAF oligomerization as well as interactions with upstream regulators and downstream effectors [2,3,5]. For example, the eight-stranded β-sandwich structure of the TRAF-C domain mediates the interaction with receptors, and the minor structural differences in this domain among TRAFs (as revealed by X-ray crystallography) define the specificity of each TRAF binding to various receptors [6,7]. Therefore, one important role of TRAFs is to serve as adaptor proteins in the assembly of receptor-associated signaling complexes, linking upstream receptors to downstream effector enzymes. Most TRAFs, with the exception of TRAF1, contain an N-terminal RING finger domain, followed by a variable number of zinc fingers [2,3,8]. The RING finger is found in many E3 ubiquitin ligases and comprises the core of the ubiquitin ligase catalytic domain. Indeed, increasing evidence indicates that in addition to their role as adaptor proteins, TRAFs (including TRAF2, 3, 5 and 6) also act as E3 ubiquitin ligases [3,8,9]. Thus, TRAFs function as both adaptor proteins and E3 ubiquitin ligases to regulate signaling.
Despite the similarities in the signaling pathways activated by different TRAF proteins, each TRAF appears to play obligatory and non-redundant physiological roles. Germ-line and conditional knockout mice have been instrumental in revealing the overlapping yet distinct roles of different TRAFs in whole animals. Compelling evidence from these studies demonstrates that TRAFs critically regulate a plethora of physiological processes, including innate and adaptive immunity, embryonic development, tissue homeostasis, stress response, and bone metabolism [3,18,19]. The pivotal roles of TRAFs in host immunity are further highlighted by the discoveries that pathogens adopt deliberate strategies to subvert TRAF functions [20,21]. An emerging paradigm of TRAF functions is that alterations in TRAFs may contribute to the pathogenesis of important human diseases, including cancers, autoimmune diseases and immunodeficiencies [18,22,23]. This has sparked new appreciation and interest in TRAF research during the past few years. Here I attempt to provide an overview of the current knowledge of TRAFs, with an emphasis on recent advances in understanding TRAFs in receptor signaling and in human diseases as well as recent insights into the regulatory mechanisms of TRAF ubiquitination.

TRAFs in signaling by the TNF-R superfamily
Receptors of the TNF-R superfamily have wide tissue distribution and regulate diverse biological functions, including immune responses, inflammation, lymphoid organ and brain development, osteoclastogenesis, and tissue homeostasis [3,[24][25][26]. Structurally, these receptors are characterized by the presence of conserved cysteine-rich domains (CRDs) in their extracellular region that are responsible for the binding of their ligands of the TNF superfamily. Based on the intracellular structures, the TNF-R superfamily is categorized into two main groups. The first group of receptors, termed death receptors, contain a death domain in the intracellular region. The second group, also the majority of the TNF-R superfamily, do not have a death domain but contain TRAF-interacting motifs (TIMs) in their intracellular region [3,24,26]. TRAF2, 3 and 5 usually have overlapping binding motifs, whereas TRAF6 has a distinct interacting motif on these receptors [3,27].
Receptors of this family do not have kinase activity and depend on the binding of adaptor proteins to assemble signaling complexes to activate downstream pathways [3,24,26]. Signaling by death receptors mainly relies on adaptor proteins containing a death domain, such as TRADD or FADD, thereby culminating in caspase activation and cell apoptosis. In contrast, signaling by the TIM-containing receptors is mediated primarily, albeit not exclusively, via TRAFs [3,24,26]. These include TRAFs that can interact with the receptors either directly through TIMs or indirectly through other TRAFs or adaptor proteins ( Table 1). Binding of TRAFs to TNF-Rs typically induces signaling cascades leading to the activation of NF-κB and MAPKs, including ERK, p38 and JNK, and ultimately regulates cell survival or functionality depending on the cell type and the context [3,24,26]. Notably, TRAF2 and TRAF5 can also modulate signaling by death receptors through association with TRADD, FADD or RIP1 (Table 1). Most TRAFdependent receptors of this family trigger the canonical NF-κB pathway (RelA/p50, NF-κB1). In contrast, the alternative NF-κB pathway (RelB/p52, NF-κB2) is activated by a subset of TNF-Rs, including CD40, BAFF-R, the lymphotoxin-β receptor (LTβR), 4-1BB, and Fn14 [28][29][30][31][32]. Interestingly, however, unlike CD40 or BAFF-R, TWEAKinduced Fn14 signaling promotes NF-κB2 activation through a distinct mechanism that induces lysosomal degradation of cIAP1-TRAF2 in a cIAP1-dependent manner
Using CD40 and BAFF-R as examples, here I briefly summarize recent advances in understanding how TRAFs regulate the two NF-κB pathways and activation of MAPKs ( Figure 2). In the absence of stimulation, TRAF3 constitutively binds to NIK (the upstream kinase of the NF-κB2 pathway) and TRAF2 (which associates with cIAP1/2). In this complex, cIAP1/2 induces K48-linked polyubiquitination of NIK, and thus targets NIK for proteasomal degradation and inhibits NF-κB2 activation [37,[68][69][70][71]. Following BAFF or CD154 stimulation, trimerized BAFF-R or CD40 recruits TRAF3, TRAF2, cIAP1/2 and MALT1 to membrane signaling rafts, releasing NIK from the TRAF3/TRAF2/cIAP1/2 complex [37,[72][73][74]. NIK protein is accumulated in the cytoplasm, induces the activation of IKKα and NF-κB2, and eventually up-regulates the expression of anti-apoptotic proteins of the Bcl-2 family (such as Bcl-2, Bcl-xL, and Mcl-1) to induce cell survival [28]. In the receptor signaling complex, TRAF2 induces K63-linked polyubiquitination of cIAP1/2, which is subsequently activated to catalyze K48linked polyubiquitination and degradation of TRAF3 and TRAF2 [37,72,74,75]. Following CD40 activation, many other signaling proteins (including TRAF5, TRAF6, TRAF1, Ubc13, MEKK1, TAK1 and NEMO) are also recruited to the cytoplasmic domain of the receptor, and the K63-specific ubiquitin ligase activity of TRAF2 and TRAF6 is rapidly stimulated [27,72,75]. These proteins form several separate multiprotein signaling complexes, which result in the phosphorylation and activation of MEKK1 and TAK1. Activated MEKK1 and TAK1 and their associated protein complexes are subsequently released from the receptor into the cytoplasm to activate MAPKs and NF-κB1, which eventually mediate the effector functions of CD40 [35,72]. Interestingly, the releasing step of MEKK1 and TAK1 is inhibited by TRAF3 via a yet unknown mechanism, but promoted by cIAP1/2catalyzed K48-linked polyubiquitination and proteasomal degradation of TRAF3 [72,74,75]. In response to BAFF stimulation, a signaling pathway of c-Raf-MEK-ERKdependent phosphorylation and down-regulation of the pro-apoptotic protein Bim also contributes to B cell survival [76]. In light of the evidence that TRAF1 mediates 4-1BB-induced ERK-dependent phosphorylation and down-regulation of Bim to promote T cell survival [77][78][79], it would be interesting to investigate the role of TRAF1 in BAFF-induced Bim down-regulation in B cells. Collectively, the above evidence indicates that TRAFs are critical regulators of signaling by the TNF-R superfamily.
TRAF3 is required for both MyD88-dependent and TRIF-dependent activation of IRF3 and IRF7, and thus production of type I IFNs [95,96], a class of cytokines with potent antiviral and antibacterial activities. In MyD88-dependent signaling downstream of TLR7 and TLR9, TRAF3 is recruited to MyD88 and IRAK1. Activated TRAF3 catalyzes its K63-linked autoubiquitination, and assembles a signaling complex with MyD88, IRAK4, IRAK1, IKKα and IRF7. Within this complex, IRF7 is phosphorylated and activated by IRAK1 and IKKα to induce the production of type I IFNs [8,82,86]. In TRIF-dependent signaling downstream of TLR3 and TLR4, TRAF3 interacts with oligomerized TRIF, and activated TRAF3 recruits TBK1 and IKKε through NAP1 and TANK. In this signaling complex, TRAF3, in cooperation with Ubc13 and/or Ubc5, catalyzes K63-polyubiquitination of TRAF3 itself, TBK1 and IKKε, which facilitates the phosphorylation of IRF3 and IRF7. The phosphorylated IRF3 and IRF7, in turn, form homodimers or heterodimers, translocate into the nucleus and induce the expression of type I IFNs as well as IFN-inducible gene [8,82,86] (Figure 3).
Interestingly, TRAF1 was also identified as a TRIFinteracting protein in yeast two-hybrid screens. Overexpression of TRAF1 inhibits TRIF-and TLR3mediated activation of NF-κB and expression of IFN-β, suggesting that TRAF1 inhibits TRIF-dependent signaling [83,97,98]. Similarly, TRAF4 physically interacts with and functionally counteracts TRAF6 and TRIF in TLR signaling [99]. Taken together, recent advances indicate that TRAF6, TRAF3, TRAF1 and TRAF4 play critical and largely distinct roles in signaling by TLRs.

TRAFs in NLR signaling
NOD-like receptors (NLRs) are a family of cytosolic sensors of PAMPs and DAMPs, and are functionally analogous to TLRs [100][101][102]. Each NLR appears to be activated by multiple agonists. However, in many cases, evidence of direct interaction between NLRs and PAMPs/DAMPs is lacking [103,104]. Effector functions of NLRs include secretion of pro-inflammatory cytokines, chemokines, antimicrobial peptides and type I IFNs, generation of ROS, autophagy, antigen processing, and expression of MHC class II on antigen presenting cells. These responses induce innate immune clearance of the pathogen, and also tailor the adaptive immune system to fight the infection [100][101][102]. NLRs are characterized by a central NOD domain that mediates nucleotide-binding and oligomerization, and the C-terminal LRRs that possibly mediate ligand detection. In addition, they contain Nterminal effector domains, such as caspase recruitment domains (CARD), pyrin domains (PYD), baculovirus inhibitor of apoptosis repeat (BIR) domains, or an acidic transactivation domain, which recruit downstream signal transduction molecules after ligand sensing [100,101,104]. One well-studied pathway of several NLRs, including NLRP3, NLRP1, NLRP6, and NLRC4, is the assembly of multi-protein complexes called 'inflammasomes' , which contain caspase-1 and apoptosis-associated speck-like protein containing a CARD (ASC). Inflammasomes induce proteolytic processing of pro-IL-1β and pro-IL-18 into secretable IL-1β and IL-18, as well as caspase 1-dependent apoptosis termed 'pyroptosis' [15,100,101]. The role of TRAF2 in inflammasome signaling has recently been explored, but the published data are contradictory. Labbe et al.
reported that depletion of TRAF2 by siRNA inhibits inflammasome signaling in HEK293T cells [105]. However, Vince et al. found that inflammasome activation is normal in TRAF2 −/− bone marrow-derived macrophages (BMDMs) [71]. Potential involvement of other TRAFs in inflammasome signaling remains to be elucidated. TRAF2, TRAF5, and TRAF6 are required for NF-κB and MAPK activation induced by NOD1 and NOD2 (Figure 4), the founding members of the NLR family [15,102,106]. Upon detection of meso-diaminopimelic acid (DAP) by NOD1 or muramyl dipeptide (MDP) by NOD2 at the vicinity of plasma membranes, oligomerization of NOD1 or NOD2 recruits the dual specificity kinase RIP2 (also called RICK) via a homotypic CARD-CARD interactions [101][102][103]. Activated RIP2 induces the formation of the signaling complex containing TRAF2, TRAF5, TRAF6, TRAF4, CARD9, cIAP1/2, and Ubc13/Uev1A. In this complex, cIAP1/2, in coordination with Ubc13/Uev1A, catalyze K63-linked polyubiquitination of RIP2, which further recruits TAB2/3-TAB1-TAK1 and NEMO-IKKα/β, leading to NF-κB activation [15,100,[107][108][109]. Interestingly, a recent study by Damgaard et al. demonstrated that XIAP is also recruited to the NOD2 signaling complex, in which XIAP primarily conjugates ubiquitin chains on RIP2 that are linked through lysine residues other than K63 and K48 [110]. Thus, XIAP, together with cIAP1/2, constitutes the major ubiquitin ligase activity that ubiquitinates RIP2 in NOD2 signaling, and cIAP1/2 appear to be rate limiting only when XIAP is not present [110]. It has been shown that TRAF2 and TRAF5 are required for NOD-induced NF-κB activation, while TRAF6, CARD9, and ITCH are important for p38 and JNK activation in NOD signaling [15,111,112]. However, the exact mechanism of how these occur is still unknown. Interestingly, TRAF4 is identified as a key negative regulator of NOD2 signaling. TRAF4 binds directly to NOD2 in an agonistdependent manner, and inhibits NOD2-induced NF-κB activation and bacterial killing [109]. This inhibitory effect of TRAF4 requires its phosphorylation at Ser426 by IKKα, which is also recruited to the NOD2 signaling complex [113].
TRAF3 mediates type I IFN production induced by NOD1 [114], and presumably also that induced by NOD2 ( Figure 4). NOD1 and NOD2 induce type I IFN production through distinct mechanisms. Upon sensing DAP, oligomerization of NOD1 recruits TRAF3 via RIP2. TRAF3 in turn activates TBK1 and IKKε, leading to subsequent activation of IRF7 and type I IFN production in epithelial cells [100,102,114]. In contrast, NOD2 induces type I IFN production only in response to viral ssRNA, but not in response to MDP, via a RIP2independent pathway [102,115]. Following the detection of viral ssRNA, NOD2 engages a signaling complex containing MAVS on mitochondria, which induces IRF3 activation and type I IFN production [115]. TRAF3 has been shown to directly interact with MAVS to mediate RLR-induced type I IFN production [116]. It is thus speculated that TRAF3 may similarly activate TBK1 and IKKε in NOD2-MAVS signaling, but this awaits experimental investigation.
Depletion of either TRAF2 or TRAF5 leads to reduced IRF3 and NF-κB activation upon RIG-I stimulation, and TRAF2 and TAK1 are important for p38 activation [125,128,139,140]. Biochemical studies revealed TRAF2 and TRAF5 interaction motifs in the C-terminal region Type I IFNs Figure 5 TRAFs in signaling by RIG-I. Upon ligand binding, RIG-I recruits TRAF3, TRAF6, TRAF2 and TRAF5 to mitochondria via MAVS. TRAF3 mediates RIG-I-induced IRF3 but not NF-κB1 activation. TRAF6 mediates RIG-I-induced IRF7 activation and also contributes to activation of NF-κB1, JNK, and p38. TRAF2 is important for p38 activation, and both TRAF2 and TRAF5 also contribute to activation of IRF3 and NF-κB1 in RIG-I signaling.
of MAVS. Upon RIG-I signaling, interaction of TRAF5 with MAVS induces K63-linked TRAF5 autoubiquitination and subsequent NEMO-dependent activation of IRF3 and NF-κB [139]. Similarly, activation of p38 by RIG-I proceeds via a TRAF2-TAK1-dependent pathway. The p38 activation in turn stimulates the production of IFNs and IL-12 [125]. Nonetheless, details of TRAF2-or TRAF5-signaling pathways downstream of MAVS remain to be elucidated. TRAF6 is required for RLR-induced IRF7 activation and also contributes to activation of NF-κB, JNK, and p38 by directly interacting with MAVS, which has two TRAF6interacting motifs [127,141]. Activation of IRF7 after viral infection resembles IRF3 activation, and involves the direct phosphorylation of IRF7 by TBK1 and IKKε. However, activation of IRF7 but not IRF3 is impaired in TRAF6 −/− fibroblasts, and TRAF6 mediates IRF7 ubiquitination [141,142]. Thus, MAVS-induced IRF7 activation is transduced through a unique TRAF6-dependent pathway. Uncoupling IRF3 from the IRF7 activation pathway might be a way of avoiding their simultaneous inhibition by virusencoded inhibitory proteins [128]. TRAF6 and MEKK1 are also important for RLR-induced activation of NF-κB and MAPKs [127,141]. Interestingly, RIG-I-MAVS-TRAF6 signaling leads to IKKβ-mediated phosphorylation of p65 at ser536, which is under the control of the NADPH oxidase NOX2 [143].
Notably, cIAP1 and cIAP2 are also recruited to MAVS, and mediate K48-and K63-linked polyubiquitination of TRAF3 and TRAF6 in response to viral infections [144]. However, the kinetics of these two types of ubiquitination on TRAF3 and TRAF6 is still unclear. Interestingly, viruses also induce IRF3-dependent apoptosis in infected cells, which require the presence of RIG-I, MAVS, TRAF3, TRAF2, TRAF6 and TBK1, as demonstrated by studies using genetically defective mouse and human cell lines [140]. Apoptosis is triggered by direct interaction of IRF3, through a newly identified BH3 domain, with the proapoptotic protein Bax. Co-translocation of IRF3 and Bax to mitochondria results in the induction of mitochondriadependent apoptosis, and transcriptionally inactive IRF3 mutants could efficiently mediate apoptosis [140]. Although why TRAF3, TRAF2 and TRAF6 are all required for IRF3-induced apoptosis awaits further clarification, it appears that these TRAF molecules cooperate in this process.

TRAFs in cytokine receptor signaling
It was initially recognized that TRAF6 is utilized for signaling by the IL-1R family (IL-1R, IL-18R and IL-33R), which also contain TIR domains found in TLRs [23,145,146]. However, recent evidence indicates that TRAFs also directly regulate signaling by a variety of other cytokine receptors, including receptors for the proinflammatory IL-17 family, anti-viral IFNs, antiinflammatory TGFβ, and the T cell cytokine IL-2.
Interestingly, TRAF2 and TRAF5 transduce the IL-17 signals to stabilize mRNA transcripts of chemokines (such as CXCL1) and cytokines (such as IL-6) by recruiting the splicing factor SF2 (also known as alternative splicing factor, ASF) into the IL-17R-Act1 signaling complex [151,153,161]. The IL-17R-Act1-TRAF2-TRAF5 complex also induces the activation of MAPKs, which further enhance mRNA stability. Notably, formation of this complex requires IKKε, an inducible IKK that mediates Act1 phosphorylation at Ser311, adjacent to a putative TRAFbinding motif. Substitution of Ser311 of Act1 with alanine impairs the IL-17-induced Act1-TRAF2-TRAF5 interaction and inflammatory gene expression [161,162]. In contrast, TRAF3 and TRAF4 are negative regulators of IL-17R signaling [12,163,164]. Upon IL-17 stimulation, IL-17RA and IL-17RC directly recruit TRAF3 via a distal C-terminal TRAF3-binding site. The binding of TRAF3 to IL-17Rs interferes with the formation of the activation signaling complex of IL-17R-Act1-TRAF6, resulting in suppression of downstream signaling, including NF-κB and MAPK activation, and production of inflammatory cytokines and chemokines [12,163]. TRAF4 exerts its negative regulation on IL-17 signaling by competing with TRAF6 for the interaction with Act1, as TRAF4 and TRAF6 use the same TIMs on Act1. Indeed, primary epithelial cells derived from TRAF4 −/− mice display markedly enhanced IL-17 signaling [164]. Thus, both TRAF3 and TRAF4 restrict IL-17 signaling at receptor proximal steps ( Figure 6).

IFN receptors
Interferons induce the synthesis of a variety of antiviral proteins that mediate swift innate immune responses to control virus replication and spread, and also shape the adaptive immune response by acting directly on T and B cells [116]. TRAF2 and TRAF6 are recognized as direct signal transducers of IFN receptors. Upon IFN engagement, TRAF2 directly binds to the membrane proximal half of the signal-transducing subunit of the IFN receptor, IFNAR1, and is required for IFN-induced NF-κB2 activation and anti-viral responses [13,165]. Similarly, direct interaction of TRAF6 with the intracellular domain of IFNλR1 regulates NF-κB activation and IFNλR1 stability in response to type III IFNs (IFNλ1, IFNλ2, and IFNλ3) [166]. Whether other TRAFs contribute to the regulation of IFN signaling remains to be determined.

TGFβ receptors
The anti-inflammatory cytokine TGFβ binds to type II and type I serine/threonine kinase receptors (TβRII and TβRI). TRAF6 interacts with a consensus TIM present in TβRI [14,167,168]. The TβRI-TRAF6 interaction induces auto-ubiquitination of TRAF6. TβRI kinase activity is required for activation of the canonical Smad pathway, whereas TRAF6 regulates the activation of TAK1 in a receptor kinase-independent manner. Activated TRAF6 mediates K63-linked polyubiquitylation of TAK1 at Lys34 and Lys158, and results in subsequent activation of p38 and JNK, leading to cell apoptosis [14,167,168]. Thus, TRAF6 is specifically required for the Smad-independent activation of JNK and p38 in response to TGFβ. However, in cancer cells, TRAF6mediated K63-linked polyubiquitination of TβRI also promotes cleavage of TβRI by TNFα converting enzyme (TACE) in a PKCζ-dependent manner. The liberated intracellular domain of TβRI associates with the transcriptional regulator p300 to activate genes involved in tumor invasiveness, such as Snail and MMP2 [169]. In this case, TRAF6 is critical for TGFβ-induced invasion of cancer cells. Additionally, TRAF6 mediates the suppressive effect of IL-1β or LPS on TGFβ-induced signaling through interaction with the type III TGF-β receptor (TβRIII), an accessory receptor that presents the TGFβ ligand to TβRII. Co-treatment with TGFβ and IL-1β or LPS promotes the interaction between phosphorylated TβRIII and ubiquitinated TRAF6, and thereby sequesters TβRIII from the TβRII/TβRI complex, resulting in inhibition of Smad2/3 activation [170]. Taken together, TRAF6 plays multiple roles in signaling by TGFβ receptors. Interestingly, TGFβ also induces the posttranslational loss of TRAF1, whereas IL-7 restores TRAF1 levels in T cells [171]. No evidence is available about the participation of other TRAFs in TGFβ signaling.

IL-2 receptor
The binding of TRAF6 to the TIM of the IL-2R β-chain negatively regulates IL-2-induced Jak1 activation in CD4 T cells, which is likely involved in the proper regulation of T cell activation and development [172].
TRAFs in other signaling pathways T cell receptor TRAF1, TRAF3, and TRAF6 are able to regulate signaling by the T cell receptor (TCR). TRAF1 inhibits CD3induced NF-κB2 activation and proliferation in T cells [31,173]. TRAF3 is recruited to the signaling rafts, and mediates the synergistic activation of ERK, LAT, PLCγ1 and ZAP70 as well as cytokine production and proliferation in T cells following co-stimulation with TCR and CD28 [11]. TRAF6 is also recruited to the TCR signaling rafts containing CARMA1-MALT1-Bcl10-PKCθ-IKK-Caspase 8 via interaction with the paracaspase MALT1, and contributes to the induction of NF-κB activation and IL-2 production in T cells [174,175]. Interestingly, a recent study by Xie et al has shown a distinct mechanism of TRAF6 in TCR signaling, in which TRAF6 is recruited to the TCR/CD28 signaling complex by LAT and promotes the ubiquitination and phosphorylation of LAT as well as the activation of NF-AT in T cells [176].

C-type lectin receptors
Using macrophages derived from TRAF6 −/− mice, it has been shown that TRAF6 is required for NF-κB and JNK activation, and expression of proinflammatory cytokines in response to engagement of C-type lectin receptors during fungal infection [177]. This will elicit further studies of other TRAFs in signaling by C-type lectin receptors.

DNA damage response
TRAF6 is essential for DNA damage-induced NF-κB activation. In this process, TRAF6 is activated by the kinase ataxia telangiectasia mutated (ATM), which is a DNA strand break sensor. Following DNA damage, ATM translocates in a calcium-dependent manner to cytosol and membrane fractions, and interacts with TRAF6 via a TIM, resulting in K63-linked polyubiquitination of TRAF6 and recruitment of cIAP1 [178]. The ATM-TRAF6-cIAP1 module stimulates TAB2-dependent TAK1 phosphorylation, and cIAP1 catalyzes monoubiquitination of NEMO at Lys285. NEMO monoubiquitination is a prerequisite for genotoxic NF-κB activation and DNA damage response [178]. Potential involvement of other TRAFs in this response awaits further investigation.

Substrates, E3 ligases and deubiquitinases of TRAFs
Ubiquitination has emerged as a key regulatory mechanism of TRAFs in signaling. As mentioned above in receptor signaling sections, E3 ligase activity has been demonstrated for TRAF2, TRAF3, TRAF5 and TRAF6, which catalyze non-degradative K63-linked polyubiquitination of their substrates. This is mediated in cooperation with the E2 ubiquitin-conjugating enzymes Ubc13-Uev1A or UbcH5c. It is believed that K63-linked polyubiquitin chains serve as docking sites for formation of signaling complexes, facilitate the recruitment and activation of effector kinases, and thus enable the propagation of signals [9,89,179]. The substrates of TRAFs include TRAF themselves, receptors, kinases, adaptor proteins, transcription factors, E3 ubiquitin ligases, and other functional proteins involved in autophagy or ROS production (Table 2). However, in many cases, substrates of TRAFs (especially those of TRAF2, TRAF3 and TRAF5) have not been unequivocally demonstrated by in vitro ubiquitination assays using purified proteins. Interestingly, a recent study has shown that TRAF2 becomes a highly active K63-specific ubiquitin ligase when bound to sphingosine-1-phosphate (S1P), which appears to be a cofactor for TRAF2 E3 ligase activity [180]. This suggests that addition of S1P may improve the efficiency of in vitro ubiquitination assays for TRAF2. Future studies need to determine whether similar cofactors exist for TRAF3 and TRAF5. Interestingly, however, the crystal structure of the RING domain of TRAF2 [181] and the phenotype of the ΔRING TRAF2 mutant [61,182] suggest that TRAF2 may not function as an E3 ligase at all. The controversy about whether TRAF2 is actually a RING E3 ligase is described in detail in an excellent review by Silke [183].
Interestingly, A20, an unusual enzyme that contains both ubiquitinating and deubiquitinating activities, negatively regulates inflammation by inhibiting NF-κB activation in TNF-R and TLR signaling. A20 is a target gene of NF-κB, and able to remove K63-linked polyubiquitin chains from TRAF6 to turn off NF-κB activation. A20 also inhibits the E3 ligase activities of TRAF6, TRAF2, and cIAP1 by promoting K48-linked polyubiquitination and degradation of the E2 enzymes Ubc13 and UbcH5c [8,128,217]. Furthermore, A20 is capable of targeting an associated signaling molecule such as TRAF2 to the lysosomes for degradation, a process that does not require A20 ubiquitin modifying activity [218]. Notably, A20 −/− and MCPIP1 −/− mice spontaneously develop severe inflammatory syndrome [219,220], while CYLD −/− and Usp25 −/− mice are more susceptible to inflammation [204,221]. Thus, negative regulation of TRAF signaling is necessary to prevent harmful immune responses and inflammatory diseases.
In addition to ubiquitination, other post-translational modifications, including phosphorylation and glutathionylation, are also reported to regulate TRAFs in signaling. Phosphorylation of TRAF1 (at Ser 139 in mouse and Ser 146 in human by PKN1) inhibits TNF-R2-dependent tonic NF-κB and JNK signaling in HeLa cells [233], and also has a negative impact on the recruitment of TBK1 to the 4-1BB signaling complex and the subsequent NF-κB activation in T cells [234]. Phosphorylation of TRAF2 (at Ser11 and Ser55 by PKCζ or IKKε, and at Thr117 by PKCδ and PKCε), which promotes K63-linked ubiquitination of TRAF2 and NF-κB activation, has been demonstrated in TNFα signaling or in transformed cells [235][236][237][238]. Following NOD2 activation, phosphorylation  [213] of TRAF4 (at Ser 426 by IKKα) negatively regulates NOD2 signaling in macrophages, including NF-κB activation, cytokine production and antibacterial activity [113]. Tyrosine phosphorylation of TRAF6 by Fyn and c-Src has been shown following LPS stimulation [191]. Interestingly, a recent study reported that TRAF6 is S-glutathionylated under normal conditions. Upon IL-1 stimulation, TRAF6 undergoes deglutathionylation catalyzed by glutaredoxin-1 (GRX-1), a process that is essential for TRAF6 auto-ubiquitination and subsequent NF-κB activation [239]. These findings suggest that different post-translational modifications of TRAF proteins coordinate to regulate the activity of TRAFs in signaling in a dynamic manner.

Viral proteins that target or hijack TRAFs
TRAFs are critical players in host immunity, as demonstrated by their shared usage by both innate immune receptors (such as TLRs, NLRs, RLRs, and cytokine receptors) and adaptive immune receptors (such as CD40, BAFF-R, OX40, 4-1BB, and TCR). Interestingly, viruses and bacteria have developed a variety of strategies to target or hijack TRAFs to evade host immune responses and to promote their own propagation or persistence (  [242,243]. Similarly, A52R and K7 proteins of vaccinia virus disrupt signaling complexes containing TRAF6 and IRAK2 to block NF-κB activation and antiviral defense [20,244]. (3) Some viral proteins usurp TRAFs for viral signaling to promote their own propagation or persistence. The best example of this group is latent membrane protein 1 (LMP1) of Epstein-Barr virus, which sequesters most cellular TRAF3, and hijacks TRAF1, 2, 3, 5 and 6 to mimic constitutively activated CD40 signaling [245][246][247][248][249]. (4) The v-FLIP member MC159 of the human molluscum contagiosum virus mediates the recruitment of both TRAF2 and TRAF3 into the Fas death inducing signaling complex to modulate Fas signaling, and powerfully inhibits both caspase-dependent and caspaseindependent cell death induced by Fas [250]. (5) Some viruses up-regulate the expression of specific miRNAs to target TRAFs. For example, the Tat protein of HIV-1 and VSV infection up-regulate miR-32 and miR-146a, which directly target the protein expression of TRAF3 and TRAF6, respectively [251][252][253]. Together, the above evidence further highlights the crucial importance of TRAFs in host immunity against pathogens.

In vivo functions of TRAFs in mice
The in vivo functions of TRAFs in whole animals have been explored by gene targeting in mice. Mice genetically deficient in individual TRAFs have been generated. Among these knockout mice, only TRAF1 −/− , TRAF5 −/− , and 67% of TRAF4 −/− mice could survive to adulthood. In contrast, mice deficient in TRAF2, 3, or 6 exhibit perinatal death with multiple organ abnormalities, indicating that TRAF2, 3, and 6 are indispensable in early development. Although viable, mice deficient in TRAF1, 4 or 5 exhibit distinct phenotypes (Table 6). For example, skin of TRAF1 −/− mice is hypersensitive to TNF-induced necrosis [173], and these mice are resistant to allergic lung inflammation in an experimental model of asthma [257]. TRAF4 −/− mice suffer respiratory disorder and wheezing caused by tracheal ring disruption, and exhibit numerous developmental abnormalities, including defects in the development of the axial skeleton and in the closure of the neural tube as well as myelin perturbation [258][259][260]. For mice with early lethality, the causes of death appear to be different. TRAF2 −/− mice succumb to severe colitis that result from apoptosis of colonic epithelial cells and accumulation of IL-10-secreting neutrophils, which can be ameliorated by deletion of TNFR1 or combined treatment with neutralizing antibodies against TNFα and IL-10 [261,262]. The early lethality of TRAF3 −/− mice is rescued by compound loss of the NF-κB2 gene, suggesting that constitutive NF-κB2 activation leads to the lethal phenotype of TRAF3 −/− mice [263]. In contrast, TRAF6 −/− mice die of severe osteopetrosis, splenomegaly, and thymic atrophy [264,265]. Taken together, these findings demonstrate that although TRAFs have overlapping functions, each TRAF molecule also plays unique and distinct roles that could not be compensated or substituted by other TRAFs in whole animals.
During the past few years, different laboratories have employed the conditional gene targeting strategy to circumvent the early lethality of TRAF −/− mice. These new mouse models allow more detailed analyses and direct comparison of specific functions of TRAFs in different cell types of whole animals ( Table 6).

B lymphocytes
TRAF2, 3, 5 and 6 are important in the survival, development, and activation of B cells. In the absence of either TRAF2 or TRAF3, B cells exhibit remarkably prolonged survival independent of BAFF, which result from the constitutive NF-κB2 activation [36,270,271].

Hepatocytes and skeletal muscles
Hepatocyte-specific TRAF2 −/− mice exhibit significantly decreased blood glucose levels under high-fat diet conditions. Although these mice show normal insulin signaling and the hypoglycemic response to insulin, they have severely impaired glucagon signaling and the hyperglycemic response to glucagon. In addition, TRAF2 overexpression significantly increases the ability of glucagon or a cAMP analog to stimulate CREB phosphorylation, gluconeogenic gene expression, and hepatic glucose production in primary hepatocytes. Thus, hepatic cell TRAF2 autonomously promotes hepatic gluconeogenesis, and contributes to hyperglycemia in obesity [268]. Interestingly, skeletal muscle-specific depletion of TRAF6 in mice improves satellite cell activation and skeletal muscle regeneration through up-regulation of Notch signaling and reducing the inflammatory repertoire [287]. TRAF6 deficiency inhibits the induction of atrophy program in response to starvation, denervation, or cancer cachexia by suppressing the expression of key regulators of atrophy, including MAFBx, MuRF1, p62, LC3B, Beclin1, Atg12, and Fn14 [60,285,286]. Ablation of TRAF6 also improves the phosphorylation of Akt and FoxO3a and inhibits the activation of 5′ AMP-activated protein kinase in skeletal muscle in response to starvation. Moreover, K63-linked autoubiquitination of TRAF6 regulates ER stress and unfolding protein response pathways in starvationinduced muscle atrophy [288]. It remains to be elucidated whether other TRAFs regulate hepatocyte and skeletal muscle functions.

Atherosclerosis
Experiments of mouse models of atherosclerosis have provided evidence that TRAF1, 5 and 6 regulate the pathogenesis of this disease. Missiou et al. reported that TRAF1 deficiency attenuates atherosclerosis in low-density lipoprotein receptor (LDLR) −/− mice by impairing monocyte recruitment to the vessel wall [291]. Deletion of TRAF1 inhibits adhesion of inflammatory cells to the endothelium, reduces the expression of CD29 in macrophages, and decreases the expression of the adhesion molecules ICAM-1 and VCAM-1 in endothelial cells [291]. In contrast, TRAF5 deficiency accelerates atherogenesis in LDLR −/− mice. Deletion of TRAF5 in endothelial cells or in leukocytes enhances adhesion of inflammatory cells to the endothelium, thus facilitating inflammatory cell recruitment to the atherosclerotic plaques. In addition, TRAF5 deficiency increases the expression of adhesion molecules and chemokines, and potentiates macrophage lipid uptake and foam cell formation [292]. Interestingly, endothelial and myeloid cell TRAF6 proteins have opposite roles in atherosclerosis in ApoE −/− mice. Endothelial TRAF6 deficiency inhibits atherosclerosis by reducing proinflammatory gene expression and monocyte adhesion to endothelial cells. In contrast, myeloid cell-specific TRAF6 deletion exacerbates atherosclerosis. TRAF6 −/− macrophages exhibit impaired expression of the atheroprotective cytokine IL-10, elevated ER stress, increased sensitivity to oxidized LDL-induced apoptosis, and reduced capacity to clear apoptotic cells [293]. Similar mouse models of TRAF2, 3 and 4 need to be generated and characterized in future studies.

TRAFs in human diseases
Findings obtained from TRAF-deficient mouse models have laid the basis to understand the roles of TRAFs in the pathogenesis of human diseases. Given their importance in regulating the development, survival and activation of various cell types, it would be expected that aberrant functions of TRAFs may contribute to different diseases. However, the roles of TRAFs in human diseases are just beginning to be revealed. Available evidence implicates TRAFs in the pathogenesis of cancers, autoimmune diseases, immunodeficiencies, and neurodegenerative diseases (Table 7).

B cell malignancies
Growing literature documents the prominent relevance of TRAF3, TRAF2 and TRAF1 in B cell malignancies.

Carcinomas
Overexpression and gene amplification of TRAF4 and TRAF6 have been reported in human carcinomas. TRAF4 is overexpressed in breast and lung carcinomas [304,318,319]. TRAF4 protein overexpression is limited to cancer cells and the subcellular localization is consistently cytoplasmic in a large majority of cases. Increased TRAF4 gene copy number is one major mechanism responsible for TRAF4 protein overexpression in human cancers. Indeed, TRAF4 is located at chromosome 17q11.2 in a region of amplification devoid of other known oncogenes [304,318,319]. Intriguingly, TRAF4 is a target gene of the p53 family of transcription factors, including p63, p73 and p53, in squamous cell carcinoma of the head and neck (SCCHN). TRAF4 locates in the nucleus in normal oral epithelium and highly/moderately differentiated cells, but is localized in the cytoplasm in poorly differentiated SCCHN. Overexpression of TRAF4 in SCCHN induces apoptosis and suppresses colony formation [320][321][322]. Thus, TRAF4 overexpression has different outcomes in different carcinomas. Notably, TRAF6 gene is located in another frequently amplified region at chromosome 11p13. TRAF6 exhibits overexpression and gene amplification in lung cancer and osteosarcoma cells [305,306,323]. Downregulation of TRAF6 in human lung cancer and osteosarcoma cells suppresses NF-κB activation, cell survival and proliferation, and tumor formation and invasion. These observations suggest that TRAF6 overexpression may promote the tumorigenesis and invasion of lung cancer and osteosarcoma cells [305,306,323].

Autoimmune diseases
Single nucleotide polymorphisms (SNPs) in TRAFs have been linked to autoimmune diseases such as systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA). SNPs of TRAF6 are associated with both SLE and RA [22]. Similarly, SNPs at the TRAF1/C5 locus are associated with both SLE and RA [307,308,310,311]. A single SNP (rs7514863), mapping upstream of the TRAF5 gene and affecting a putative transcription factor binding site, demonstrates a significant association with RA [309]. In addition, decreased expression of TRAF2 has been detected in peripheral blood mononuclear cells of SLE patients [324]. However, further association and functional studies are required to determine whether these TRAFs play causal roles in increasing susceptibility to SLE or RA.

Immunodeficiencies
An autosomal dominant mutation of TRAF3 has been reported in a young adult with a history of herpes simplex virus-1 (HSV-1) encephalitis in childhood [312]. The TRAF3 mutant allele is loss-of-expression, lossof-function, dominant-negative, and associated with impaired responses upon stimulation of both TNF-Rs and TLRs. The recurrent HSV-1 infection and encephalitis result from the impairment of TLR3-induced type I IFN production [312].

Hypohidrotic ectodermal dysplasia
A heterozygous mutation of TRAF6 has recently been identified in a patient with hypohidrotic ectodermal dysplasia (HED) [325]. The mutant TRAF6 protein is capable of forming a complex with TAK1 and TAB2, but cannot bind to the receptor XEDAR. Furthermore, the mutant TRAF6 protein potently inhibits the interaction between wild type TRAF6 and XEDAR, and suppresses the XEDAR-mediated NF-κB activation. Thus, this mutant TRAF6 protein acts in a dominant-negative manner to affect the XEDAR-mediated NF-κB activation during the development of ectoderm-derived organs, leading to HED phenotype [313].

Neurodegenerative diseases
Interestingly

Chronic inflammation and infection
In light of their crucial importance in inflammatory and immune responses, it would be predicted that TRAF molecules may also contribute to chronic inflammation and infection. Although no genetic association of TRAFs and chronic inflammation or infection has been identified, recent evidence of alterations of TRAF protein levels supports this possibility. Notably, TRAF2 and TRAF3 are often degraded in response to signaling by the TNF-R superfamily [3,27,32,328]. In contrast, TRAF1 expression is up-regulated by NF-κB activation in response to signaling by a variety of receptors, including TNF-R superfamily and cytokine receptors, etc. [329][330][331]. The dynamic change of the stoichiometry of different TRAF molecules inside the cell impacts subsequent cellular responses to inflammatory or infectious stimuli. For example, the presence of TRAF1 stabilizes TRAF2, which plays a role in promoting proinflammtory responses in HeLa cells [332,333]. More direct evidence was provided by a recent study demonstrating that TRAF1 is specifically lost from virus-specific CD8 T cells during the chronic phase of infection with HIV in humans [171]. This area warrants further investigation.

Conclusions
Since the first TRAFs were cloned in the mid 1990s, we have witnessed a remarkable progress in understanding the functions of TRAFs in signaling. TRAFs are now recognized as signal transducers of a wide variety of receptors, including the TNF-R superfamily, TLRs, NLRs, RLRs, IL-1R family, IL-17Rs, IFN receptors, TGFβ receptors, IL-2R, TCR, and C-type lectin receptors. Although initially defined as adaptor proteins, most TRAFs also function as E3 ubiquitin ligases through their RING finger domain. Furthermore, activation of TRAFs is exquisitely regulated by post-translational modifications, especially ubiquitination, which has become the subject of intense investigations during the past few years. Termination of TRAF activation could be achieved through either K48linked polyubiquitination followed by proteosomal degradation or removal of K63-linked polyubiquitin chains catalyzed by deubiquitinases. Acting alone or in combination, TRAF-dependent signaling pathways regulate the activation of NF-κBs, MAPKs, or IRFs to control diverse cellular processes. Accumulating evidence obtained from TRAF-deficient mice demonstrates that each TRAF plays obligatory and distinct roles critical for innate immunity, adaptive immunity, embryonic development, and tissue homeostasis. The pivotal roles of TRAFs in host immunity are further highlighted by the finding that targeting TRAFs appears to be a common mechanism employed by pathogenic proteins of viruses and bacteria. Furthermore, the interest in TRAFs is also driven by recent discoveries that link TRAF genetic variations to human diseases such as cancers, autoimmune diseases, and immunodeficiencies. In conclusion, TRAFs are versatile and indispensable regulators of signal transduction and immune responses, and aberrant functions of TRAFs contribute to the pathogenesis of human diseases.

Perspectives
Despite the wealth of current knowledge about TRAFs, many key questions remain, which will drive the next stage of research in this important area. (1) What is the stoichiometric composition of TRAFs and other signaling proteins in each signaling complex? What are the dynamic kinetics of activation and spatial regulation of each TRAF molecule in response to each specific stimulus? Cutting-edge biochemical, proteomic, and imaging technologies will be needed to uncover these details. (2) How is the E3 ligase activity of each TRAF regulated precisely? What are the substrates of the E3 ligase activities of TRAF2, 3 and 5? Are there additional E3 ligases, deubiquitinases, kinases, and phosphatases that target different TRAFs? Are the enzymes targeting TRAFs regulated by TRAF-dependent signaling pathways? In vitro reconstitution experiments and ligase activity assays, high throughput screens for substrates and enzymes, and systems biology approaches will be needed to address these issues. (3) What are the molecular structures of each TRAF in complex with its specific signaling partner, substrate, or enzyme? This requires access to cocrystals containing TRAFs and their interacting partners, and the crystal structure of full-length TRAFs/substrates remains a challenge. (4) Are there additional pathogenic factors of invading microorganisms that target TRAFs during infections? If so, by what precise mechanisms? Yeast 2-hybrid screen, bioinformatic studies and proteomic approaches may be applied toward this end. (5) During pathogen infections, multiple TRAF-dependent signaling pathways are triggered either sequentially or simultaneously, including innate immune receptors (such as TLRs, NLRs and RLRs), adaptive immune receptors (such as TCR, CD40, OX-40 and 4-1BB), and cytokine receptors (such as IL-1R, IL-17R, IFN-Rs, and TβRs). How does each TRAF act in such complex and concerted signaling pathways in different cellular context during infection? Whether and how does each TRAF regulate the crosstalk between different immune signaling pathways? Responses of TRAF −/− mice, and especially cell type-specific TRAF −/− mice, to infections will be instrumental in addressing these questions. Sequential or simultaneous co-engagement of different immune receptors also needs to be investigated thoroughly in cultured cells. (6) What are the cell type-specific factors that dictate cell type-specific TRAF functions? For example, TRAF2 or TRAF3 deficiency leads to prolonged survival in B cells, but not in T cells or DCs. Genetic and systems biology approaches will be required for such studies. (7) Are there additional TRAF genetic alterations and SNPs associated with human diseases? Do epigenetic modifications of TRAFs contribute to disease conditions? How? Systematic and comprehensive analyses employing genetic, bioinformatic, and deep sequencing approaches will facilitate such investigation. Generation and examination of TRAF −/− or TRAFtransgenic mouse models of human diseases are also required to decipher the underlying mechanisms. Together, these future studies will undoubtedly yield valuable information to advance our understanding of TRAFs.
Given the importance of TRAFs in host immunity and in human diseases, the above future studies will also provide a platform for the development of therapeutic intervention of TRAF-mediated human diseases. For example, insights gained into the structures of each TRAF in complex with its specific signaling partner, substrate, or enzyme will guide the development of structure-based therapeutics. Small agonists and antagonists of TRAFs may be devised to enhance beneficial signaling pathways and to interfere with harmful ones, respectively. In this regard, cell-permeable TRAF6 decoy peptides potently inhibit TRAF6 signaling in cultured cells, and their therapeutic potential in disease settings are currently under investigation [191,334]. A chemical compound 5-(4-methoxyarylimino)-2-N-(3,4-dichlorophenyl)-3-oxo-1,2,4-thiadiazolidine (P(3)-25), which possesses anti-bacterial and anti-fungal activities, specifically inhibits TRAF2-mediated NF-κB activation while enhancing TRAF2-mediated AP-1 activation [335]. However, the diverse and cell type-specific functions of TRAFs may prevent systemic administration of therapeutic agents that directly target TRAFs, and local or cell-specific drug delivery needs to be exercised. Alternatively, therapeutic strategies may be designed to specifically manipulate TRAF-interacting partners or downstream signaling pathways. For example, pharmacological inhibitors for cIAP1/2 are currently at various stages of clinical trials for cancers [107], and may be applied to other TRAF-mediated diseases too. Further in-depth understanding of TRAF signaling pathways will serve as experimental framework to be translated into such therapeutic development.

Competing interests
The author declares that she has no competing financial interests.