<?xml version="1.0" ?>
<tei>
	<teiHeader>
		<fileDesc xml:id="0"/>
	</teiHeader>
	<text xml:lang="en">

		<p>Tumour necrosis factor receptor (TNFR)-associated<lb/> factors (TRAFs) are intracellular
			signalling molecules<lb/> with crucial functions in the signal transduction<lb/>
			pathways initiated by members of the TNFR family,<lb/> the Toll-like receptor (TLR)
			family, the interleukin-1<lb/> receptor (IL-1R) family and the RIG-I-like receptor<lb/>
			(RLR) family. The TNFR family consists of more<lb/> than 25 different receptors that are
			expressed on<lb/> different cell types and regulate diverse functions,<lb/> including
			immune responses, organ development<lb/> and tissue homeostasis <ref type="biblio"
				>1</ref> . Members of the TLR family,<lb/> which consists of 10 and 12 functional
			receptors<lb/> in humans and mice, respectively, are expressed<lb/> primarily on innate
			immune cells and B cells and<lb/> represent an integral part of the first-line host<lb/>
			defence system against infectious agents <ref type="biblio">2</ref> . The IL-1R<lb/>
			family, which includes IL-1R and IL-18R, controls<lb/> different inflammatory functions,
			in part through the<lb/> induction of other pro-inflammatory cytokines (such<lb/> as
			IL-6 and interferon-γ (IFNγ)) and also through the<lb/> activation of endothelial cells
			and the promotion of an<lb/> inflammatory tissue environment <ref type="biblio">3</ref>
			. TLRs and IL-1R<lb/> family members are structurally related and signal<lb/> through
			common cytoplasmic proteins (see below);<lb/> they are therefore collectively referred
			to here as<lb/> TLR/IL-1R proteins. The RLR family includes two<lb/> well-characterized
			members, namely, retinoic acid-<lb/>inducible gene I (RIG-I) and melanoma
			differentiation-<lb/>associated gene 5 (MDA5), which initiate innate<lb/> immune
			responses following recognition of intracellular<lb/> double-stranded RNA (dsRNA), which
			is produced<lb/> during virus infection <ref type="biblio">4</ref> .<lb/></p>

		<p>Additional functions of TRAFs in other signalling<lb/> pathways (such as the IL-17
			receptor pathway and the<lb/> T cell receptor pathway) have been described but are
			not<lb/> discussed here as only limited information is available<lb/> about the
			molecular function of TRAF3 in these path-<lb/>ways <ref type="biblio">5,6</ref> . The
			large repertoire of receptors that use and<lb/> depend on TRAFs for signal transduction
			explains why<lb/> these proteins regulate a plethora of biological functions.<lb/></p>

		<p>The first TRAFs to be identified, TRAF1 and<lb/> TRAF2, were cloned based on their
			interaction with the<lb/> intracellular domain of TNFR2. TRAFs were proposed<lb/> to be
			adaptor proteins in TNFR signalling owing to their<lb/> activation-dependent receptor
			binding and concomitant<lb/> recruitment of the constitutively associated
			TRAF-<lb/>binding proteins cellular inhibitor of apoptosis 1 (cIAP1)<lb/> and cIAP2
			(collectively referred to as cIAP here). Today,<lb/> after the discovery of five other
			TRAF family members<lb/> and the recognition that TRAFs interact with all TNFR<lb/>
			family members, it is clear that one important aspect of<lb/> their function is the
			assembly of signalling complexes at<lb/> the intracellular domains of transmembrane
			receptors,<lb/> ultimately translating receptor ligation into the activation<lb/> of
			downstream signalling pathways. In the case of some<lb/> TNFR family members, such as
			TNFR2 and CD40, the<lb/> TRAFs bind directly to the intracellular domains of the<lb/>
			receptors. In case of other receptors, such as TNFR1 and<lb/> members of the TLR/IL-1R
			family, additional adaptor<lb/> proteins are used to recruit TRAFs.<lb/></p>

		<p>Apart from their role as adaptor proteins, TRAF<lb/> proteins also act as E3 ubiquitin
			ligases, a function that is<lb/> crucial for the activation of downstream signalling
			events.<lb/> Such events include the activation of nuclear factor-κB<lb/></p>

		<figure type="table">Table 1 | Immune receptors that signal via TRAF proteins<lb/>
			Ligand<lb/> Receptor<lb/> TRAF<lb/> Receptor-expressing cell<lb/> type<lb/> TNF,
			LTα<lb/> TNFR1,<lb/> TNFR2<lb/> TRAF1, TRAF2, TRAF5 Ubiquitous expression<lb/>
			LTα–LTβ,<lb/> LIGHT<lb/> LTβR<lb/> TRAF2, TRAF3, TRAF4,<lb/> TRAF5<lb/> Stromal
			cells<lb/> CD40L<lb/> CD40<lb/> TRAF1, TRAF2, TRAF3,<lb/> TRAF5, TRAF6<lb/> Innate
			immune cells, B cells,<lb/> T cells<lb/> BAFF<lb/> BAFFR<lb/> TRAF2, TRAF3<lb/> B
			cells<lb/> LPS<lb/> TLR4<lb/> TRAF3, TRAF6<lb/> Innate immune cells, B cells,<lb/>
			activated T cells<lb/> BAFF, B cell-activating factor; BAFFR, BAFF receptor; LPS,
			lipopolysaccharide; LT, lymphotoxin;<lb/> LTβR, LTβ receptor; TLR4, Toll-like receptor
			4; TNF, tumour necrosis factor; TNFR, TNF receptor;<lb/> TRAF, TNFR-associated
			factor.<lb/></figure>

		<p>(NF-κB) and activator protein 1 (AP1), which are<lb/> key transcription factors that
			control many immune<lb/> response genes, and the activation of certain interferon<lb/>
			regulatory factors (IRFs), which transcriptionally control<lb/> the production of
			antiviral type I IFNs. Both functions<lb/> of TRAFs (as adaptor proteins and as
			ubiquitin ligases)<lb/> are essential for the activation of receptor-mediated<lb/>
			signalling responses. In addition, recent evidence shows<lb/> that at least one TRAF,
			TRAF3, has an important negative<lb/> regulatory function in unstimulated cells that is
			disrupted<lb/> following receptor activation or as a consequence of<lb/> oncogenic
			mutations.<lb/></p>

		<p>This Review briefly summarizes the principal<lb/> signalling functions of TRAF family
			members in<lb/> well-characterized signal transduction pathways. We<lb/> then describe
			recent findings regarding the molecular<lb/> mechanism and highly versatile role of
			TRAF3 in<lb/> immunity-related signal transduction, with a focus on<lb/> the signalling
			pathways and cell types that best illustrate<lb/> the different modes of TRAF3 action.
			Finally, we review<lb/> very recent observations about the role of TRAF3 in<lb/> human
			disease.<lb/></p>

		<head>Structure and function of TRAFs<lb/></head>

		<p>As mentioned above, TRAF1 and TRAF2 were first<lb/> identified as TNFR2-interacting
			proteins, and their<lb/> carboxy-terminal region (referred to as the TRAF domain)<lb/>
			became the defining feature of the TRAF family <ref type="biblio">1,7</ref> .
			TRAF2,<lb/> TRAF3 and TRAF6 are constitutively expressed in most<lb/> cell types,
			whereas TRAF5 expression is mainly restricted<lb/> to immune cells. TRAF1, which is
			constitutively<lb/> associated with TRAF2, is apparently involved in the<lb/>
			fine-tuning of TRAF2 signalling (for example, during<lb/> CD40 activation) <ref
				type="biblio">8</ref> . TRAF4 interacts with members of<lb/> the transforming growth
			factor-β (TGFβ) receptor family<lb/> and controls organ development <ref type="biblio"
				>9–11</ref> . The contributions<lb/> of TRAF1, TRAF4 and TRAF7 to
			TRAF3-regulated<lb/> functions have not been investigated so far and will not<lb/> be
			discussed further here.<lb/></p>

		<p>Soon after their original identification as TNFR-<lb/>interacting proteins, the roles of
			TRAF proteins in<lb/> the activation of the NF-κB and mitogen-activated pro-<lb/>tein
			kinase (MAPK) pathways became evident <ref type="biblio">12–14</ref> .<lb/>
			Subsequently, TRAF6 and, more recently, TRAF3 were<lb/> found to be involved in
			signalling by receptors that do<lb/> not belong to the TNFR family, including TLRs
			and<lb/> IL-1R family members <ref type="biblio">15–17</ref> . Importantly, the outcome
			of<lb/> TNFR or TLR engagement appears to be determined by<lb/> the combination of TRAF
			proteins that are recruited<lb/> by each receptor <ref type="table">(TABLE
			1)</ref>.<lb/></p>

		<p>The TRAF domain. TRAFs exhibit a modular structure<lb/> with well-defined functional
			domains <ref type="figure">(FIG. 1)</ref>. The<lb/> TRAF domain is located at the C
			terminus of all TRAF<lb/> family members, with the exception of TRAF7. The<lb/> TRAF
			domain can be divided into an amino-terminal<lb/> coiled-coil region (TRAF-N) that
			governs TRAF<lb/> homotrimerization and a C-terminal domain (TRAF-C)<lb/> that
			contributes to TRAF oligomerization and also<lb/> promotes interactions with upstream
			regulators. These<lb/> upstream regulators can be either the intracellular<lb/> domains
			of receptors (such as TNFR2, CD40 and the<lb/> BAFF receptor (BAFFR)) or intermediate
			adaptor<lb/> proteins, such as TNFR1-associated death domain<lb/> protein (TRADD; which
			functions downstream of<lb/> TNFR1) and IL-1R-associated kinase (IRAK) family<lb/>
			members (which are involved in TLR and IL-1R<lb/> signalling) <ref type="biblio"
				>18,19</ref> . In addition, both TRAF-N and TRAF-C<lb/> domains interact with
			downstream effectors <ref type="biblio">1</ref> . These<lb/> include cIAP and
			NF-κB-inducing kinase (NIK; also<lb/> known as MAP3K14), which bind to the TRAF-N<lb/>
			domain of TRAF2 and the TRAF-C domain of TRAF3,<lb/> respectively <ref type="biblio"
				>20–24</ref> . The interplay between these proteins<lb/> that occurs both
			constitutively and during receptor<lb/> activation is essential for the regulation of
			the so-called<lb/> alternative NF-κB pathway (discussed in detail below).<lb/></p>

		<p>Ubiquitylation and TRAF signalling. Most TRAFs, with<lb/> the exception of TRAF1, contain
			a similar N-terminal<lb/> domain that consists of several zinc finger domains<lb/> and
			an N-terminal really interesting new gene (RING)<lb/> finger motif. The RING finger
			motif is also found in<lb/> many E3 ubiquitin ligases <ref type="biblio">1</ref> and
			RING finger-mediated<lb/> protein ubiquitylation has emerged as a key mechanism<lb/> in
			TRAF-dependent signal transduction <ref type="biblio">25</ref> .<lb/></p>

		<p>Two major types of polyubiquitin chain have been<lb/> described, in which each subsequent
			ubiquitin molecule<lb/> is attached to the previous one either through lysine 48<lb/>
			(K48) or through K63. K48-linked polyubiquitin<lb/> chains target proteins for
			proteasome-dependent<lb/> degradation, whereas K63-linked polyubiquitin<lb/> chains
			mainly control protein–protein interactions<lb/> and thereby modify protein function
				<ref type="biblio">25–29</ref> . Both forms<lb/> of polyubiquitin modification are
			involved in TRAF-<lb/>dependent signalling. However, TRAFs themselves<lb/> seem to
			stimulate only the generation of K63-linked<lb/> polyubiquitin chains. This is mediated
			in cooperation<lb/> with the E2 ubiquitin-conjugating enzymes UBE2D3<lb/> (also known as
			UBCH5C) and UBC13–UEV1A (also<lb/> known as UBE2N–UBE2V1) <ref type="biblio">30,31</ref>
			.<lb/></p>

		<p>Non-degradative protein ubiquitylation, such as<lb/> K63-linked polyubiquitylation, was
			first proposed<lb/> as a crucial mechanism in TRAF-dependent signalling<lb/> following
			the identification of an E2 complex containing<lb/> UBC13 and UEV1A that, in concert
			with TRAF6<lb/></p>

		<figure>Figure 1 | Domain organization of mammalian TRAF proteins. The six human<lb/>
			TNFR-associated factor (TRAF) proteins that contain a C-terminal TRAF domain are<lb/>
			shown. All TRAFs (with the exception of TRAF1) contain an N-terminal RING finger<lb/>
			domain (a signature motif of E3 RING finger ubiquitin ligases; labelled R in the
			figure)<lb/> and several zinc finger motifs (labelled Z). The TRAF domain contains a
			coiled-coil<lb/> region (labelled CC) and a C-terminal TRAF-C domain (also known as a
			meprin and<lb/> TRAF homology (MATH) domain). AA, amino acids.<lb/></figure>

		<p>and ubiquitin, stimulates the catalytic activity of the<lb/> IκB kinase (IKK) complex in
			cytoplasmic extracts <ref type="biblio">28</ref> .<lb/> Although the specific targets of
			TRAF6-mediated<lb/> K63-linked ubiquitylation are still largely unknown,<lb/> a recent
			publication has confirmed K63-linked<lb/> ubiquitylation as a key mechanism for
			TRAF6<lb/> signalling <ref type="biblio">30</ref> . Based on these data, as well as on
			results<lb/> obtained from UBC13-deficient cells, it appears that<lb/> K63-linked
			polyubiquitylation mediated by TRAF6–<lb/> UBC13–UEV1A is essential for the
			IL-1-dependent<lb/> (and probably TLR4-dependent) activation of NF-κB<lb/> and MAPKs
				<ref type="biblio">30,32</ref> .<lb/></p>

		<p>Other studies have shown that TRAF2 is<lb/> decorated by K63-linked polyubiquitin chains
			in<lb/> response to CD40 engagement and can mediate the<lb/> K63-linked
			polyubiquitylation of cIAP proteins in a<lb/> UBC13-dependent manner <ref type="biblio"
				>33,34</ref> . Still, it is important to<lb/> note that comparative structural
			analyses of the RING<lb/> finger regions of TRAF2 and TRAF6 suggest that TRAF6<lb/>
			interacts directly with UBC13 but that TRAF2 does<lb/> not, indicating that additional
			cofactors are involved<lb/> in TRAF2-dependent ubiquitylation <ref type="biblio"
				>35</ref> . An interesting<lb/> recent study has shown that TRAF2 becomes a
			highly<lb/> active K63-specific ubiquitin ligase in response to<lb/> the binding of
			sphingosine-1-phosphate (S1P) to its<lb/> RING finger domain. TRAF2 then ubiquitylates
			the<lb/> protein kinase receptor-interacting protein 1 (RIP1;<lb/> also known as RIPK1),
			a protein with a crucial role in<lb/> TNFR-mediated signal transduction <ref
				type="biblio">36–38</ref> . Furthermore,<lb/> inhibition or silencing of sphingosine
			kinase 1 (SPHK1),<lb/> which catalyses S1P synthesis, blocks TNFR1-and<lb/>
			CD40-mediated IKK and MAPK activation, indicating<lb/> an important role for
			intracellular S1P as a cofactor for<lb/> efficient TRAF2 signalling.<lb/></p>

		<p>It should be mentioned that another form of non-<lb/>degradative polyubiquitylation, the
			assembly of<lb/> linear polyubiquitin chains, was recently reported to<lb/> contribute
			to TRAF-mediated signal transduction <ref type="biblio">39</ref> .<lb/> Such linear
			polyubiquitin chains are assembled by the<lb/> linear ubiquitin chain assembly complex
			(LUBAC),<lb/> which consists of haem-oxidized IRP2 ubiquitin ligase 1<lb/> (HOIL1; also
			known as RBCK1) and HOIL1-interacting<lb/> protein (HOIP; also known as RNF31) <ref
				type="biblio">40</ref> . Linear<lb/> polyubiquitin chains appear to be crucial for
			TNF-<lb/>mediated IKK activation, in part through stabilization<lb/> of the
			TRAF2-containing signalling complex and in<lb/> part through direct IKKγ ubiquitylation
				<ref type="biblio">39,41</ref> . However,<lb/> neither TRAF3 signalling nor the
			interplay between<lb/> TRAF2 and TRAF3 that controls the alternative NF-κB<lb/> pathway
			(see below) has been reported to involve this<lb/> type of ubiquitylation. Therefore, it
			will not be discussed<lb/> further here.<lb/></p>

		<p>Together, two different modes of TRAF function<lb/> can be discerned. First, TRAFs
			function as adaptor<lb/> molecules that recruit other signalling proteins (such<lb/> as
			cIAP and NIK) into protein complexes, thereby<lb/> controlling signal transduction
			pathways, both<lb/> constitutively and following receptor ligation <ref type="biblio"
				>34</ref> . Second,<lb/> TRAFs function as K63-specific ubiquitin ligases,
			which<lb/> alter the function of target proteins through their non-<lb/>degradative,
			site-specific ubiquitylation activity. In many<lb/> cases, these two functions occur
			almost simultaneously<lb/> and coordinately, as TRAF autoubiquitylation provides<lb/>
			docking sites for molecular adaptors and downstream<lb/> kinases with ubiquitin-binding
			motifs, leading to kinase<lb/> activation and subsequent effector functions <ref
				type="biblio">33,42</ref> .<lb/></p>

		<head>TRAF-dependent signalling pathways<lb/></head>

		<p>The best-characterized TRAF-dependent effector<lb/> pathways are those that lead to the
			activation of the IKK<lb/> complex and NF-κB, TANK-binding kinase 1 (TBK1)<lb/> and
			IRF3, and the MAPK cascades.<lb/></p>

		<p>Control of MAPK signalling. The MAPK signalling<lb/> pathways are hierarchically
			organized phosphorylation<lb/> cascades consisting of a MAPK, a MAPK kinase<lb/> (MAPKK)
			and a MAPKK kinase (MAP3K). The<lb/> downstream MAPKs — namely, extracellular
			signal-<lb/>regulated kinase (ERK), JUN N-terminal kinase (JNK)<lb/> and p38 — mainly
			control the activities of transcription<lb/> factors, such as AP1 and the activating
			transcription<lb/> factors (ATFs), but also phosphorylate and modify the<lb/> function
			of other substrates, such as the E3 ubiquitin<lb/> ligase ITCH <ref type="biblio"
				>43,44</ref> .<lb/></p>

		<p>TRAF proteins have been shown to activate several<lb/> MAP3Ks, including MEKK1 (also
			known as MAP3K1),<lb/> MEKK3 (also known as MAP3K3), TGFβ-activated<lb/> kinase 1 (TAK1;
			also known as MAP3K7) and tumour<lb/> progression locus 2 (TPL2; also known as MAP3K8)
				<ref type="biblio">45</ref> .<lb/> TPL2 has a role in CD40-and TLR4-induced
			ERK1<lb/> and ERK2 activation, but how TPL2 is activated is not<lb/> clear <ref
				type="biblio">45</ref> . MEKK3-deficient fibroblasts exhibit defects in<lb/>
			IL-1-and TRAF6-dependent NF-κB activation, which<lb/> depends in part on TAK1 <ref
				type="biblio">(REF. 32)</ref>. However, owing to<lb/> early embryonic lethality, no
			data are available on the<lb/> role of MEKK3 in primary immune cells <ref type="biblio"
				>46</ref> .<lb/></p>

		<p>By contrast, a large set of experimental data is available<lb/> for TRAF-dependent
			TAK1-and MEKK1-mediated<lb/> signalling <ref type="biblio">45</ref> . TAK1 is activated
			via TRAF-dependent and<lb/> -independent pathways and controls NF-κB and MAPK<lb/>
			activation in a cell-type-and stimulus-specific manner.<lb/></p>

		<figure>κ<lb/> κ<lb/> Figure 2 | Adaptors and signalling complexes used by members of the
			TLR and<lb/> IL-1R families. Engagement of Toll-like receptor (TLR) and IL-1 receptor
			(IL-1R) family<lb/> members triggers two main signalling pathways that are dependent on
			either myeloid<lb/> differentiation primary response protein 88 (MYD88) or TIR
			domain-containing adaptor<lb/> protein inducing IFNβ (TRIF) 52 . MYD88, TIR
			domain-containing adaptor protein (TIRAP),<lb/> TRIF-related adaptor molecule (TRAM) and
			TRIF are TIR domain-containing adaptor<lb/> proteins without known enzymatic activity
			that are recruited to the activated and<lb/> oligomerized receptors. TIR
			domain-containing adaptors interact with other signalling<lb/> molecules, including
			members of the IL-1R-associated kinase (IRAK) and<lb/> TNFR-associated factor (TRAF)
			families. TRAF6 is essential for the activation of most<lb/> known MYD88-dependent
			effector pathways, including the nuclear factor-κB (NF-κB),<lb/> activator protein 1
			(AP1), interferon regulatory factor 3 (IRF3) and IRF7 pathways<lb/> (REFS 47,48).
			However, TRAF6 is dispensable for TRIF-dependent NF-κB, IRF3 and<lb/> IRF7 activation,
			although it might be involved in AP1 signalling downstream of TRIF 47 .<lb/>
			TRIF-dependent NF-κB activation is still not fully understood, but depends on the<lb/>
			adaptor protein TNFR1-associated death domain protein (TRADD) and the serine/<lb/>
			threonine kinase receptor-interacting protein 1 (RIP1; also known as RIPK1), at least
			in<lb/> some cell types 91,92 . TRAF3 is recruited to both the MYD88-and
			TRIF-assembled<lb/> signalling complexes and positively controls IRF3 and IRF7
			activation 47,57 , while<lb/> negatively regulating mitogen-activated protein kinase
			(MAPK) activation and the<lb/> induction of inflammatory cytokines (not shown) 42,47
			.<lb/></figure>

		<p>TAK1 activation depends primarily on TRAF6 and occurs<lb/> downstream of both CD40 (which
			signals via TRAF2,<lb/> TRAF3 and TRAF6) and TLR4 (which uses TRAF3<lb/> and TRAF6) <ref
				type="biblio">47,48</ref> . By contrast, MEKK1 activation depends<lb/> on TRAF2
			(REF. 33) and, as a result, TLR4 does not signal<lb/> via MEKK1. TRAF2, TRAF6, MEKK1 and
			TAK1 are all<lb/> involved in JNK1, JNK2 and p38 activation, but their<lb/> relative
			contributions depend on the specific receptor<lb/> and the cell type being investigated.
			CD40-mediated<lb/> JNK activation in B cells depends primarily on TRAF2<lb/> and MEKK1,
			with only a small contribution from<lb/> TRAF6 and TAK1 <ref type="biblio">(REF.
				49)</ref>. However, TNFR1-dependent<lb/> JNK activation depends on TAK1, with a
			minor<lb/> contribution by MEKK1. Notably, TAK1, and not<lb/> MEKK1, is involved in
			TRAF2-and TRAF6-dependent<lb/> NF-κB activation pathways in some cell types <ref
				type="biblio">50</ref> . However,<lb/> in other cell types, TAK1 is clearly needed
			for JNK<lb/> activation but does not seem to have a clear role in NF-κB<lb/> activation
				<ref type="biblio">50</ref> . Different cell-type-specific effector functions<lb/>
			are regulated through MAPK pathways. In macrophages<lb/> stimulated with TLR ligands,
			the expression of several<lb/> pro-inflammatory cytokines (including TNF, IL-12<lb/> and
			IL-6) depends on MAPK activation <ref type="biblio">42,51</ref> , whereas,<lb/> in B
			cells, CD40-induced proliferation and production<lb/> of T cell-dependent antibodies
			involves MEKK1 and<lb/> MAPK activation <ref type="biblio">49</ref> . TRAF3 was
			unexpectedly found<lb/> to be a negative regulator of TNFR-and TLR-mediated<lb/> MAPK
			activation, and has to be degraded to allow MAPK<lb/> activation to occur <ref
				type="biblio">33,42</ref> (see below).<lb/></p>

		<p>TRAFs and type I IFN production. Pathogen recognition<lb/> via different receptor
			systems, including TLRs and<lb/> cytoplasmic RLRs, results in the induction of type
			I<lb/> IFNs (IFNα and IFNβ). Both receptor families have been<lb/> demonstrated to use
			TRAF3 for type I IFN induction;<lb/> however, we focus here mainly on the role of
			TRAF3<lb/> in TLR-mediated signalling, which was discovered<lb/> first, and refer to
			observations from RLR-mediated<lb/> signal transduction to illustrate the common
			molecular<lb/> mechanism of TRAF3 function in these pathways.<lb/></p>

		<p>TLRs initiate signal transduction via the recruit-<lb/>ment of TLR/IL-1R (TIR)
			domain-containing adaptor<lb/> proteins, which bind to the intracellular part of
			the<lb/> receptor via homotypic TIR–TIR interactions. Four<lb/> TIR domain-containing
			adaptor proteins are known,<lb/> and their hierarchical mode of action and usage by
			dif-<lb/>ferent TLRs is illustrated in <ref type="figure">FIG. 2</ref>. Whereas
			myeloid<lb/> differentiation primary response protein 88 (MYD88)<lb/> is used by all TLR
			and IL-1R family members with the<lb/> exception of TLR3, TIR domain-containing
			adaptor<lb/> protein inducing IFNβ (TRIF) is only used by TLR3<lb/> and TLR4. MYD88 also
			contains a death domain<lb/> through which it interacts with IRAK family members,<lb/>
			which participate in TRAF6 activation <ref type="biblio">19</ref>
			<ref type="figure">(FIG. 2)</ref>.<lb/></p>

		<p>MYD88 and TRIF interact with TRAF3 to activate<lb/> signalling pathways leading to type I
			IFN produc-<lb/>tion, although the mode of TRAF3 binding to either<lb/> MYD88 or TRIF is
			poorly defined <ref type="biblio">47</ref> . The expression of<lb/> type I IFNs and
			related molecules is controlled by IRF<lb/> proteins, in particular IRF3 and IRF7,
			together with<lb/> other transcription factors <ref type="biblio">52</ref> . Depending
			on the particu-<lb/>lar cell type and TLR agonist being examined, a specific<lb/>
			pattern of IRF activation and type I IFN gene induction<lb/> is observed. Activation of
			plasmacytoid dendritic cells<lb/> (pDCs) via TLR9 leads to a robust IRF7-driven type
			I<lb/> IFN response, which is dominated by IFNα induction,<lb/> whereas activation of
			macrophages via TLR3 or TLR4<lb/> leads to a primarily IRF3-driven response that is
			domi-<lb/>nated by the production of IFNβ <ref type="biblio">52</ref> . The differences
			in<lb/> IRF usage, and hence the IFN gene expression pattern,<lb/> are determined to
			some extent by the TIR domain-<lb/>containing adaptor proteins, as MYD88 activates<lb/>
			IRF7, whereas TRIF activates IRF3 <ref type="biblio">(REFS 53,54)</ref>.<lb/> However,
			it should be noted that IRF7 itself is encoded<lb/> by an IFN-inducible gene, the
			expression of which<lb/> is upregulated in an autocrine manner via type I<lb/> IFN
			receptor 1 (IFNR1) during the initial phases of<lb/> TLR engagement <ref type="biblio"
				>55,56</ref> .<lb/></p>

		<p>TRAFs have a crucial role in the TLR-mediated (and<lb/> RLR-mediated) IFN response. TRAF6
			is involved in<lb/> the MYD88-dependent, but not the TRIF-dependent,<lb/> type I IFN
			response, whereas TRAF3 is involved in<lb/> both MYD88-and TRIF-dependent IFN induction
				<ref type="biblio">47,57</ref> .<lb/></p>

		<figure>κα<lb/> κα<lb/> α<lb/> β α<lb/> γ<lb/> NF-κB<lb/> P<lb/> P<lb/> P<lb/> P<lb/> P<lb/>
			P<lb/> Figure 3 | The two NF-κB activation pathways.<lb/> The two nuclear factor-κB
			(NF-κB) activation pathways are<lb/> distinguished by the inhibitory proteins that they
			target<lb/> — namely, inhibitor of NF-κB (IκB) and p100 (also known<lb/> as NF-κB2), the
			precursor for the NF-κB subunit p52. The<lb/> classical NF-κB pathway, which is
			triggered by many<lb/> different receptors (including Toll-like receptors (TLRs)<lb/>
			and TNF receptors (TNFRs)), depends on the catalytic<lb/> activity of the IκB kinase
			(IKK) catalytic subunit IKKβ,<lb/> which phosphorylates IκBα or IκBβ and targets them
			for<lb/> degradative K48-linked ubiquitylation 93 . Following IκB<lb/> degradation,
			NF-κB transcription factors (such as p50–p65<lb/> or p50–cREL dimers) are free to enter
			the nucleus. IKKβ is<lb/> part of the trimeric IKK complex, which also contains the<lb/>
			IKKα catalytic subunit and the IKKγ regulatory subunit 64 .<lb/> Although IKKα can also
			phosphorylate IκB proteins, in most<lb/> cases classical NF-κB activation depends on
			IKKβ. The<lb/> alternative NF-κB pathway is mainly activated by a subset of<lb/> TNFRs
			(such as the BAFF receptor (BAFFR) and CD40) and<lb/> depends on the catalytic activity
			of IKKα dimers, which are<lb/> bound to IKKβ or IKKγ (not shown). IKKα is activated
			by<lb/> NF-κB-inducing kinase (NIK; also known as MAP3K14), the<lb/> turnover of which
			is regulated by TNFR-associated factor 3<lb/> (TRAF3). Activated IKKα phosphorylates a
			site in the C<lb/> terminus of p100, leading to its recognition and subsequent<lb/>
			ubiquitylation. This results in the degradation of the<lb/> C-terminal portion of p100,
			freeing its N-terminal portion<lb/> (p52), which enters the nucleus together with
			RELB.<lb/> TAK1, TGFβ-activated kinase 1 (also known as MAP3K7).<lb/></figure>

		<p>As discussed in more detail below, emerging evidence<lb/> demonstrates that
			non-degradative, K63-linked<lb/> polyubiquitylation by TRAF3 is a key step in the
			activation<lb/> of the type I IFN response.<lb/></p>

		<p>Classical and alternative NF‑κB signalling. TRAF2 ,<lb/> TRAF3, TRAF5 and TRAF6 are all
			involved in NF-κB<lb/> activation. Two NF-κB signalling pathways have been<lb/>
			described: the canonical or &apos;classical&apos; NF-κB signalling<lb/> pathway and the
			non-canonical or &apos;alternative&apos; NF-κB<lb/> signalling pathway. The canonical
			NF-κB pathway is<lb/> triggered by most TRAF-dependent receptors, whereas<lb/> the
			alternative NF-κB pathway is activated most potently<lb/> by a subset of TNFRs,
			including CD40, BAFFR and the<lb/> lymphotoxin-β receptor (LTβR) <ref type="biblio"
				>58</ref>
			<ref type="figure">(FIG. 3)</ref>. The classical<lb/> NF-κB pathway (together with other
			signalling pathways)<lb/> regulates the expression of a large number of target
			genes<lb/> that are of importance for innate and adaptive immune<lb/> responses,
			inflammation and cell survival. In addition,<lb/> this pathway controls the expression
			of proteins that are<lb/> involved in negative feedback regulation, such as NF-κB<lb/>
			inhibitor-α (IκBα) and the polyubiquitin-modifying<lb/> enzyme A20 (also known as
			TNFAIP3) <ref type="biblio">58</ref> . The alternative<lb/> NF-κB signalling pathway is
			mainly involved in the<lb/> regulation of lymphoid organ development, adaptive<lb/>
			immune responses and B cell survival <ref type="biblio">59</ref> . TRAF2 and<lb/> TRAF5
			participate in the regulation of both pathways <ref type="biblio">60</ref> .<lb/> TRAF3
			is mainly involved in the control of alternative<lb/> NF-κB signalling and has not been
			reported to contribute<lb/> to the classical pathway (see below), whereas TRAF6<lb/>
			activates only classical NF-κB signalling.<lb/></p>

		<head>TRAF3 and TNFR-induced MAPK activation<lb/></head>

		<p>TRAF3 is unique among the TRAF proteins, not only<lb/> because it is involved in several
			signalling pathways, but<lb/> also because its mode of action differs from one
			pathway<lb/> to another and involves both positive and negative<lb/> regulatory
			functions.<lb/></p>

		<p>TRAF3 was first described as a CD40-interacting<lb/> molecule <ref type="biblio"
				>61,62</ref> but, unlike TRAF2, TRAF5 and TRAF6,<lb/> its overexpression was found
			to inhibit CD40-mediated<lb/> NF-κB activation <ref type="biblio">62,63</ref> . The
			exact role of TRAF3 in<lb/> CD40 signalling remained a mystery until recently.<lb/>
			Based on detailed biochemical analyses of CD40-and<lb/> TLR4-mediated signalling
			pathways, it was found that<lb/> one key function of TRAF3 is to control the
			spatial<lb/> organization and composition of receptor-associated<lb/> signalling
			complexes <ref type="biblio">33,34</ref> .<lb/></p>

		<p>Following CD40 activation, many signalling proteins<lb/> (including TRAF2, TRAF6, TRAF3,
			cIAP, UBC13,<lb/> MEKK1, TAK1 and IKKγ) are recruited to the cytoplasmic<lb/> domain of
			the receptor <ref type="biblio">33</ref> . These proteins might form<lb/> two separate
			multiprotein complexes, one composed of<lb/> TRAF2, MEKK1, TRAF3, cIAP, UBC13 and IKKγ
			and<lb/> the other of TRAF6, TAK1, TRAF3, cIAP, UBC13 and<lb/> IKKγ <ref type="biblio"
				>33,34</ref> . Notably, although complex formation results in<lb/> the
			phosphorylation and eventual activation of MEKK1<lb/> and TAK1, the crucial
			phosphorylation events responsible<lb/> for the activation of these MAP3Ks and their
			downstream<lb/> targets take place only after the release of MEKK1<lb/> and TAK1 and
			their associated activation complexes<lb/> from the receptor (which remains in the
			membrane) into<lb/> the cytoplasm <ref type="biblio">33</ref> . This step is controlled
			by at least two<lb/> proteins with opposing actions. The first is TRAF3, which<lb/>
			inhibits the release of the MEKK1-and TAK1-associated<lb/> complexes into the cytoplasm,
			and the second is cIAP,<lb/> which acts as a K48-specific ubiquitin ligase that
			targets<lb/> TRAF3 for proteasomal degradation and thereby<lb/> counteracts its
			inhibitory function <ref type="biblio">33,34</ref> .<lb/></p>

		<figure>γ<lb/> κ<lb/> P<lb/> P<lb/> γ<lb/> Figure 4 | TRAF3 as a gatekeeper for TNFR-induced
			stress kinase activation. Engagement of CD40 results in the<lb/> assembly of two large
			signalling complexes. TNFR-associated factor 3 (TRAF3) , cellular inhibitor of apoptosis
			1 (cIAP1),<lb/> cIAP2, ubiquitin-conjugating enzyme 13 (UBC13; also known as UBE2N) and
			IκB kinase-γ (IKKγ) are components of both<lb/> complexes. In addition, one complex
			contains TRAF2 and MEK kinase 1 (MEKK1; also known as MAP3K1), whereas the<lb/> other
			includes TRAF6 and TGFβ-activated kinase 1 (TAK1; also known as MAP3K7). Activation of
			TRAF2 or TRAF6<lb/> through autoubiquitylation results in the activation of the classic
			NF-κB signaling pathway (a) and also in the K63-linked<lb/> ubiquitylation of the cIAP
			proteins, thereby enhancing their K48-specific ubiquitin ligase activity towards
			TRAF3,<lb/> which is then degraded by the proteasome (b). This liberates the remaining
			components of each complex, resulting in<lb/> the activation of MEKK1 and TAK1, which
			then phosphorylate and activate their effector kinases MAPKK3 and MAPKK6<lb/> (which
			activate p38) or MAPKK4 and MAPKK7 (which activate JUN N-terminal kinase
			(JNK)).<lb/></figure>

		<p>Collectively, these separate events result in a unique<lb/> two-step signalling process
				<ref type="figure">(FIG. 4)</ref>. Following the<lb/> formation of the
			receptor-associated protein complexes<lb/> described above, the K63-specific ubiquitin
			ligase activity<lb/> of TRAF2 and TRAF6 is rapidly stimulated, leading to<lb/> their
			autoubiquitylation. The resulting polyubiquitin<lb/> chains may further stabilize the
			respective signalling<lb/> complexes that contain the ubiquitin-binding adaptor<lb/>
			IKKγ. TRAF2 stimulation also leads to the activation of<lb/> cIAP, resulting in the
			K48-linked polyubiquitylation and<lb/> subsequent degradation of TRAF3. TRAF3
			degradation<lb/> is crucial for the release of the MEKK1 signalling<lb/> complex (which
			also contains IKKγ and UBC13) into the<lb/> cytoplasm, and this results in the
			activation of MEKK1<lb/> and its downstream targets MAPKK4 and MAPKK7. In<lb/> addition,
			cIAP-dependent TRAF3 ubiquitylation and<lb/> degradation leads to the release of the
			TAK1 signalling<lb/> complex, which also contains IKKγ and UBC13. But,<lb/> as TAK1
			associates with TRAF6 instead of TRAF2, it<lb/> is not clear how cIAP is incorporated
			into the receptor-<lb/>bound TAK1 complex. Nonetheless, similarly to MEKK1<lb/>
			activation, the activation of TAK1 requires its release into<lb/> the cytoplasm.
			Consistent with this model, inhibition or<lb/> ablation of cIAP or overexpression of
			TRAF3 prevents<lb/> MAPK activation, whereas knockdown of TRAF3 results<lb/> in more
			rapid MAPK activation <ref type="biblio">33,42</ref> .<lb/></p>

		<p>Many molecular aspects of this process remain unclear.<lb/> For example, the molecular
			mechanism by which TRAF3<lb/> inhibits the release and activation of TAK1 and MEKK1<lb/>
			is unknown. Also, a physiological reason that necessitates<lb/> the inhibitory function
			of TRAF3 has remained somewhat<lb/> elusive. Whether the regulation of certain
			effector<lb/> functions, such as cytokine release or cell survival,<lb/> depends on this
			fine-tuning of the kinetics and spatial<lb/> distribution of MAPK activation needs to be
			established.<lb/></p>

		<head>Dual functions of TRAF3 in TLR signalling<lb/></head>

		<p>TRAF3 and the TLR‑mediated type I IFN response.<lb/> TRAF3 was identified as an important
			component of the<lb/> TLR signalling pathways through biochemical analyses<lb/> of
			MYD88-associated signalling complexes <ref type="biblio">47</ref> . Subsequent<lb/>
			functional studies revealed that TRAF3 is crucial for TLR-<lb/>induced type I IFN and
			IL-10 production by macrophages<lb/> and DCs <ref type="biblio">47,57</ref> .
			TRAF3-deficient cells produce lower levels<lb/> of type I IFNs and IL-10 than wild-type
			cells following<lb/> TLR3, TLR4 or TLR9 engagement, but higher levels<lb/> of
			pro-inflammatory cytokines (including TNF, IL-6<lb/> and IL-12) <ref type="biblio"
				>47,57</ref> . Both MYD88 and TRIF were found<lb/> to recruit TRAF3 following the
			engagement of TLR9<lb/> or TLR4, respectively <ref type="biblio">64</ref> . However,
			with respect to the<lb/> participation of TRAF3 in type I IFN production, the<lb/> TRIF
			pathway is better understood, mainly owing to the<lb/> simpler experimental system of
			lipopolysaccharide (LPS)-<lb/>stimulated macrophages, which produce IFNs in a
			TRIF-<lb/>dependent manner. LPS-mediated TLR4 activation leads<lb/> to the
			TRIF-dependent K63-linked polyubiquitylation<lb/> of TRAF3. This modification is
			probably a result of<lb/> autoubiquitylation and is required for the activation<lb/> of
			TBK1 and IKKε, which in turn phosphorylate<lb/> and activate IRF3 <ref type="biblio"
				>(REFS 42,64)</ref>
			<ref type="figure">(FIG. 5)</ref>. Supporting the<lb/> involvement of TRAF3 ubiquitin
			ligase activity in this<lb/> process is the finding that mutation of its RING
			finger<lb/> blocks the K63-linked ubiquitylation of TRAF3 and IFN<lb/> induction <ref
				type="biblio">42</ref> . Furthermore, deubiquitylating enzyme A<lb/> (DUBA; also
			known as OTUD5), which specifically<lb/> targets and deubiquitylates TRAF3, was found to
			inhibit<lb/> TLR-induced type I IFN production <ref type="biblio">65</ref> .<lb/></p>

		<p>In addition to its function in TLR pathways, TRAF3<lb/> has been shown to control
			RLR-induced IFN produc-<lb/>tion <ref type="biblio">66–68</ref> . RLRs (including RIG-I
			and MDA5) recognize<lb/> different forms of virus-derived RNA and recruit the κ<lb/></p>

		<figure>P<lb/> κ<lb/> ε<lb/> P<lb/> P<lb/> P<lb/> P<lb/> ε<lb/> P<lb/> P<lb/> Figure 5 |
			Differential TRAF3 ubiquitylation dictates the outcome of TLR4 signalling. a | Toll-like
			receptor 4 (TLR4)<lb/> engagement by lipopolysaccharide (LPS) results in the recruitment
			of myeloid differentiation primary response protein 88<lb/> (MYD88) and the rapid
			assembly of a large multiprotein complex at the cytoplasmic face of the plasma
			membrane.<lb/> Complex formation results in TNFR-associated factor 6 (TRAF6) activation
			and K63-linked polyubiquitylation of cellular<lb/> inhibitor of apoptosis (cIAP)
			proteins, which enhances their K48-specific ubiquitin ligase activity towards
			TRAF3.<lb/> Proteasomal degradation of TRAF3 results in cytoplasmic translocation of the
			MYD88-associated multiprotein complex,<lb/> thereby allowing activation of
			TGFβ-activated kinase 1 (TAK1; also known as MAP3K7), its downstream MAPK kinases<lb/>
			(MAPKKs) and mitogen-activated protein kinases (MAPKs) and the eventual induction of
			pro-inflammatory cytokines and<lb/> chemokines. b | Activated TLR4 also translocates to
			an endosomal compartment, and this is accompanied by the<lb/> recruitment of TIR
			domain-containing adaptor protein inducing IFNβ (TRIF), TRAF3 and other proteins. Unlike
			the MYD88<lb/> complex, the TRIF-assembled complex does not contain cIAP and its
			formation results in the activation and K63-linked<lb/> polyubiquitylation of TRAF3.
			This causes the activation of TANK-binding kinase 1 (TBK1) and IκB kinase-ε (IKKε),
			which<lb/> phosphorylate interferon regulatory factor 3 (IRF3) to induce the type I
			interferon (IFN) response. TRIF-dependent<lb/> signalling also leads to the activation
			of nuclear factor-κB (NF-κB), JUN N-terminal kinase (JNK) and p38, albeit with<lb/>
			slower kinetics (&apos;late phase&apos;) than MYD88 signalling, possibly reflecting the
			involvement of additional processes, such as<lb/> endosomal maturation. c | Recognition
			of viral double-stranded RNA (dsRNA) via retinoic acid-inducible gene I (RIG-I)<lb/>
			leads to the binding and activation of mitochondrial antiviral signalling protein (MAVS;
			also known as IPS1, VISA and<lb/> CARDIF), which in turn recruits TRAF3 and induces the
			TRAF3-dependent assembly of K63-linked polyubiquitin chains.<lb/> It is notable that
			TRIF-and MAVS-mediated TRAF3 signalling are comparable, as they both involve
			non-degradative<lb/> (K63-linked) ubiquitylation and the subsequent activation of TBK1–
			IKKε and IRF3. IRAK, IL-1R-associated kinase;<lb/> TIRAP, TIR domain-containing adaptor
			protein; TRADD, TNFR1-associated death domain protein; TRAM, TRIF-related<lb/> adaptor
			molecule; UBC13, ubiquitin-conjugating enzyme 13 (also known as UBE2N).<lb/></figure>

		<p>CARD domain-containing protein mitochondrial anti-<lb/>viral signalling protein (MAVS;
			also known as IPS1,<lb/> VISA and CARDIF) to activate IRF3, IRF7 and NF-κB<lb/> pathways
				<ref type="biblio">4</ref> . TRAF3 is recruited and binds directly to<lb/> MAVS
			following RLR activation, and this binding results<lb/> in its K63-linked
			polyubiquitylation and the recruit-<lb/>ment of the IRF3-activating kinase TBK1 <ref
				type="biblio">(REFS 66,67)</ref><lb/>
			<ref type="figure">(FIG. 5)</ref>. Comparably with type I IFN induction in the<lb/> TLR
			pathway, TRAF3 autoubiquitylation is crucial for<lb/> TBK1 activation and is also
			negatively regulated by<lb/> DUBA, which removes activating K63-linked ubiquitin<lb/>
			chains from TRAF3 <ref type="biblio">(REF. 65)</ref>. During virus infection with<lb/>
			vesicular stomatitis virus or Sendai virus, the removal of<lb/> K63-linked ubiquitin
			chains from TRAF3 is followed<lb/> by K48-linked ubiquitylation by the ubiquitin
			ligase<lb/> TRIAD3A (also known as RNF216 isoform 1) and results<lb/> in proteasomal
			degradation of TRAF3 and, ultimately,<lb/> in the termination of the type I IFN response
				<ref type="biblio">69</ref> . Whether<lb/> TRIAD3A also regulates TLR-mediated type
			I IFN induc-<lb/>tion needs to be investigated. Together, the described<lb/>
			observations emphasize the role of TRAF3-dependent,<lb/> non-degradative ubiquitylation
			for the induction of the<lb/> type IFN response during TLR and RLR activation.<lb/></p>

		<p>TRAF3 negatively regulates MYD88‑mediated<lb/> MAPK activation. TLR4 engagement also
			induces<lb/> the recruitment and activation of cIAP within an<lb/> MYD88-assembled,
			receptor-proximal multiprotein<lb/> complex that also contains TRAF3, TRAF6,<lb/> UBC13,
			IKKγ and TAK1 <ref type="biblio">(REF. 42)</ref>. Similarly to CD40<lb/> engagement,
			TLR4 activation is accompanied by<lb/> TRAF3 degradation, which depends on the
			cIAP-<lb/>mediated K48-linked polyubiquitylation of TRAF3.<lb/> The degradation of TRAF3
			is required for TAK1 release<lb/> and the activation of downstream MAPK signalling <ref
				type="biblio">42</ref> .<lb/> Accordingly, the inhibition of cIAP function by<lb/>
			either cIAP inhibitors or RNA interference-mediated<lb/> silencing interferes with TRAF3
			degradation, and<lb/> thus blocks the release of TAK1 into the cytoplasm.<lb/> As a
			consequence, TAK1-mediated JNK activation<lb/> and the production of inflammatory
			cytokines (such<lb/> as TNF, IL-6 and IL-12) are reduced. However, cIAP-<lb/>mediated
			TRAF3 degradation is not required for<lb/> NF-κB activation, consistent with the
			observation that<lb/> TAK1 is primarily required for TLR-mediated JNK and<lb/> p38
			activation, but is less important for TLR-mediated<lb/> IKK activation <ref
				type="biblio">50,70</ref> .<lb/></p>

		<p>Taken together, the above findings indicate that<lb/> TRAF3 controls two independent
			signalling pathways<lb/> downstream of TLR4. Whereas TRAF3 positively<lb/> controls the
			type I IFN response (which requires TRAF3<lb/> K63-linked autoubiquitylation and
			subsequent IRF3<lb/> activation), it negatively regulates MYD88-dependent<lb/> JNK and
			p38 activation (which depends on cIAP-<lb/>mediated K48-linked polyubiquitylation and
			degradation<lb/> of TRAF3, and subsequent TAK1 activation) <ref type="figure">(FIG.
				5)</ref>.<lb/> Interestingly, the different modes of TRAF3 action<lb/> correlate
			with differences in the compartmentalization<lb/> of activated TLRs. Whereas some TLRs
			(such as TLR9)<lb/> are constitutively located at endosomal membranes <ref type="biblio"
				>71</ref> ,<lb/> TLR4 is located at the cell membrane and relocates to<lb/> the
			endosomal compartment following its engagement<lb/> by a ligand, a process that is
			required for type I IFN<lb/> induction <ref type="biblio">72</ref> . Notably,
			TRAF3-dependent activation of<lb/> the IFN response is initiated from signalling
			complexes<lb/> that are assembled by the adaptor protein TRIF at an<lb/> endosomal
			location, but not from signalling complexes<lb/> assembled by MYD88 at the plasma
			membrane. So, in the<lb/> endosomal TLR9–MYD88–TRAF3 and TLR4–TRAM–<lb/> TRIF–TRAF3
			complexes, TRAF3 acts as positive<lb/> regulator and promotes IRF3 activation and type I
			IFN<lb/> production <ref type="biblio">47,57,71</ref> . By contrast, in
			TLR4–MYD88–TRAF3<lb/> or CD40–TRAF3 complexes, which are assembled at the<lb/> cell
			membrane, TRAF3 acts as a negative regulator and<lb/> counteracts TAK1-and
			MEKK1-dependent JNK and<lb/> p38 activation. Although the molecular mechanisms for<lb/>
			this phenomenon are unclear, it appears that the cellular<lb/> compartment contributes
			to the decision between the<lb/> different modes of TRAF3 action <ref type="biblio"
				>51,73</ref> .<lb/></p>

		<head>TRAF3 and the alternative NF-κB pathway<lb/></head>

		<p>Regulation of NIK by TRAF3. Initial overexpression<lb/> experiments suggested a negative
			regulatory role for<lb/> TRAF3 in CD40-stimulated classical NF-κB signalling <ref
				type="biblio">62</ref> .<lb/> However, based on more recent evidence it appears<lb/>
			that TRAF3 is primarily involved in the regulation of<lb/> the alternative NF-κB pathway
			and has no direct role<lb/> in classical NF-κB signalling. A key insight into TRAF3<lb/>
			function was provided by the observation that TRAF3<lb/> is constitutively bound to NIK
			(an essential activator of<lb/> the alternative NF-κB pathway) in unstimulated cells
				<ref type="biblio">74,75</ref> .<lb/> Genetic ablation of Nik prevents IKKα
			activation and<lb/> disrupts CD40-or BAFFR-induced processing of the<lb/> NF-κB
			precursor protein p100 (also known as NF-κB2),<lb/> a key event in alternative NF-κB
			signalling <ref type="biblio">76,77</ref>
			<ref type="figure">(FIG. 3)</ref>.<lb/> Notably, following CD40 or BAFFR engagement,
			TRAF3<lb/> was found to be degraded by the proteasome and NIK<lb/> protein accumulation
			was observed <ref type="biblio">74</ref> .<lb/></p>

		<p>Together, these data suggest that TRAF3 regulates<lb/> the turnover and intracellular
			levels of NIK and<lb/> thereby controls IKKα activation and alternative<lb/> NF-κB
			signalling. Following receptor-induced TRAF3<lb/> degradation, the accumulation of NIK
			results in its<lb/> autophosphorylation, eventually leading to IKKα<lb/> activation and
			p100 processing. This interpretation is<lb/> consistent with the observation that
			TRAF3-deficient<lb/> B cells contain high amounts of NIK and exhibit<lb/> constitutive
			p100 processing, which can be blocked by<lb/> concomitant Nik ablation <ref
				type="biblio">34,78</ref> . Nik deletion also prevents<lb/> the perinatal lethality
			in TRAF3-deficient mice, as does<lb/> ablation of the p100-encoding gene Nfkb2 <ref
				type="biblio">(REF. 34)</ref>.<lb/> As both CD40 and BAFFR can directly
			associate<lb/> with TRAF3, these observations are consistent with<lb/> a simple linear
			model of signalling events. In this<lb/> model, engagement of one of these TNFR
			family<lb/> members leads to the recruitment, oligomerization<lb/> and degradation of
			TRAF3, thereby preventing the<lb/> constitutive proteasomal degradation of NIK.
			This<lb/> results in the autocatalytic activation of NIK, leading<lb/> to IKKα
			phosphorylation and subsequent p100<lb/> processing. However, the actual signalling
			mechanism<lb/> responsible for NIK and IKKα activation is, in fact,<lb/> more
			complex.<lb/></p>

		<p>Roles for TRAF2 and cIAP in NIK regulation. The<lb/> complexity of the signalling
			mechanism that leads to NIK<lb/> and IKKα activation was first indicated by the
			observation<lb/> that TRAF2-deficient mice exhibit perinatal lethality<lb/> and develop
			a fatal wasting disease, conditions similar<lb/> to those seen in TRAF3-deficient mice
				<ref type="biblio">78</ref> . Furthermore,<lb/> both TRAF2-and TRAF3-deficient B
			cells exhibit<lb/> NIK accumulation, constitutive p100 processing and<lb/> increased
			survival <ref type="biblio">34,78–80</ref> . Moreover, genetic ablation (or<lb/> even
			haploinsufficiency) of Nik in TRAF2-deficient mice<lb/> prevents all of these phenotypes
				<ref type="biblio">34</ref> . These observations<lb/> suggested that TRAF2 and TRAF3
			have non-redundant<lb/> roles in preventing spontaneous NIK accumulation and<lb/>
			activation. Indeed, both TRAF2 and TRAF3 were found<lb/> to be required for rapid NIK
			turnover in unstimulated<lb/> cells <ref type="biblio">34,81</ref> , and cIAP (which
			associates with TRAF2) was<lb/> also shown to be necessary <ref type="biblio">20</ref>
			.<lb/></p>

		<figure>α<lb/> α<lb/> P<lb/> P<lb/> P<lb/></figure>

		<p>The first clue that cIAP proteins are involved in<lb/> alternative NF-κB signalling and
			NIK turnover came<lb/> from studies of multiple myelomas, which are malignant<lb/>
			progeny of plasma cells <ref type="biblio">82,83</ref> . A large proportion of
			multiple<lb/> myeloma tumour samples and established cell lines<lb/> display
			significantly increased NF-κB transcriptional<lb/> activity <ref type="biblio"
				>82,83</ref> . Genetic analyses identified gain-of-<lb/>function and
			loss-of-function mutations in genes<lb/> encoding components of the alternative NF-κB
			pathway,<lb/> including NIK, CD40, NFKB1, NFKB2, LTBR, TACI,<lb/> TRAF2, TRAF3, CYLD,
			IAP1 and IAP2 <ref type="biblio">(REFS 82,83)</ref>.<lb/> Importantly, deletion of the
			closely linked IAP1 and IAP2<lb/> loci in multiple myeloma cell lines is associated
			with<lb/> accumulation of both TRAF3 and NIK and constitutive<lb/> p100 processing <ref
				type="biblio">82,83</ref> .<lb/></p>

		<p>These findings were substantiated by the use of<lb/> cIAP inhibitors termed SMAC mimetics
				<ref type="biblio">84</ref> . Treatment of<lb/> various cell lines with these
			inhibitors, or treatment with<lb/> the LTβR ligand TNF-related weak inducer of<lb/>
			apoptosis (TWEAK), results in degradation of cIAP,<lb/> NIK accumulation and activation
			of alternative NF-κB<lb/> signalling <ref type="biblio">84,85</ref> . Moreover, cIAP1
			overexpression leads<lb/> to NIK degradation, and this activity depends on the<lb/>
			cIAP1 RING finger and TRAF2-interacting domains <ref type="biblio">84</ref> ,<lb/>
			suggesting that cIAP1 induces the direct ubiquitylation<lb/> and degradation of NIK.
			Importantly, cIAP1-dependent<lb/> NIK degradation depends on TRAF2 expression,<lb/>
			indicating that TRAF2-dependent cIAP1 activation is a<lb/> prerequisite for NIK
			ubiquitylation <ref type="biblio">84</ref> .<lb/></p>

		<p>Together with earlier reports, these data suggested<lb/> that under basal conditions
			TRAF3-bound NIK is<lb/> constitutively targeted for degradation through
			cIAP-<lb/>mediated ubiquitylation, which depends on TRAF2. Still,<lb/> these data did
			not entirely explain receptor-mediated<lb/> induction of alternative NF-κB
			signalling.<lb/></p>

		<p>TRAF3 as a molecular bridge. An improved<lb/> understanding of alternative NF-κB
			signalling came from<lb/> biochemical analysis that revealed a TRAF3-dependent<lb/>
			interaction between NIK and the TRAF2–cIAP<lb/> complex in unstimulated cells <ref
				type="biblio">34,86</ref> . TRAF3 was shown<lb/> to act as a molecular bridge that
			interacts with both<lb/> NIK and TRAF2–cIAP through different binding sites.<lb/>
			Importantly, this interaction increases the proximity<lb/> of NIK to its E3 ubiquitin
			ligase cIAP. The finding that<lb/> receptor-induced TRAF2 and TRAF3 degradation is<lb/>
			inhibited in cIAP-deficient multiple myeloma cell lines<lb/> (or in SMAC mimetic-treated
			cells) strongly suggested<lb/> that receptor-mediated cIAP activation is required
			for<lb/> TRAF2 and TRAF3 degradation. TRAF3 degradation<lb/> also depends on TRAF2, but
			TRAF2 degradation is<lb/> independent of TRAF3 <ref type="biblio">(REF. 34)</ref>.
			Additional in vitro<lb/> experiments using partially purified TRAF2, TRAF3<lb/> and
			cIAP2 demonstrated that cIAP2 can directly<lb/> ubiquitylate TRAF3, decorating it with
			degradation-<lb/>inducing K48-linked ubiquitin chains <ref type="biblio">34</ref> .
			Importantly, the<lb/> TRAF3-directed E3 ligase activity of cIAP2 was strongly<lb/>
			enhanced when cIAP2 was purified from cells following<lb/> CD40 activation, suggesting
			that cIAP2 activity is<lb/> regulated by receptor stimulation. This increase in
			cIAP2<lb/> ubiquitin ligase activity depended on the expression of<lb/> TRAF2 with an
			intact RING finger domain, further<lb/> indicating that TRAF2 may upregulate the
			ubiquitin<lb/> ligase activity of cIAP by ubiquitylating it with non-<lb/>degradative
			K63-linked ubiquitin chains. Indeed, CD40<lb/> ligation induces the K63-linked
			polyubiquitylation of<lb/> cIAP2 in a TRAF2-dependent manner, and this can be<lb/>
			recapitulated by overexpression of TRAF2 and cIAP2 in<lb/> HEK293 cells <ref
				type="biblio">34</ref> .<lb/></p>

		<p>A model explaining alternative NF‑κB signalling.<lb/> Together, the findings mentioned
			above suggest<lb/> the following model of TRAF-regulated alternative<lb/> NF-κB
			signalling <ref type="figure">(FIG. 6)</ref>. Under basal conditions,<lb/> TRAF3-bound
			NIK interacts with TRAF2-bound cIAP,<lb/> an interaction that depends on the TRAF
			domain-<lb/>mediated association of TRAF3 with TRAF2. The<lb/> NIK–cIAP interaction
			results in persistent K48-linked<lb/> ubiquitylation of NIK and its subsequent
			proteasomal<lb/> degradation in unstimulated cells. Stimulation of a<lb/></p>

		<figure>Figure 6 | Role of TRAF3 in the regulation of NIK turnover and activity.<lb/> a | In
			unstimulated cells, NF-κB-inducing kinase (NIK; also known as MAP3K14)<lb/> is
			associated with TNFR-associated factor 3 (TRAF3). TRAF3 interacts with a<lb/> TRAF2–cIAP
			(cellular inhibitor of apoptosis) complex to direct cIAP-mediated<lb/> K48-linked
			polyubiquitylation of NIK. This results in the ongoing degradation of<lb/> newly
			synthesized NIK, thus keeping its intracellular concentrations below the<lb/> threshold
			required for self-activation. b | Engagement of certain TNF receptor family<lb/> members
			(such as the BAFF receptor (BAFFR), CD40 or the lymphotoxin-β receptor<lb/> (LTβR))
			results in the recruitment of the NIK–TRAF3–TRAF2–cIAP complex to the<lb/> receptor and
			the activation of TRAF2 K63-specific ubiquitin ligase activity. TRAF2<lb/> then
			ubiquitylates cIAP to enhance and redirect the K48-specific ubiquitin ligase<lb/>
			activity of cIAP towards TRAF3. Degradation of TRAF3 prevents the association<lb/> of
			newly synthesized NIK with the TRAF2–cIAP complex. Accumulation of newly<lb/>
			synthesized NIK results in its activation, the phosphorylation and processing of<lb/>
			p100 (also known as NF-κB2) and the eventual release of p52–RELB dimers,<lb/> which
			enter the nucleus and initiate gene transcription. IKKα, IκB kinase-α.<lb/></figure>

		<p>receptor (for example, CD40 or BAFFR) activates<lb/> the K63-specific ubiquitin ligase
			activity of TRAF2,<lb/> leading to the K63-linked ubiquitylation of cIAP, and<lb/> this
			enhances the K48-specific ubiquitin ligase activity<lb/> of cIAP towards TRAF3. This
			results in the K48-linked<lb/> polyubiquitylation and degradation of TRAF3 (and to
			a<lb/> lesser extent TRAF2). Once the concentration of TRAF3<lb/> falls below a certain
			threshold, newly synthesized NIK<lb/> cannot associate with the TRAF2–cIAP complex
			and<lb/> is no longer targeted for degradation. Accumulation of<lb/> NIK results in its
			autophosphorylation and activation,<lb/> leading to the phosphorylation and activation
			of<lb/> IKKα. IKKα then phosphorylates p100 and targets<lb/> it for ubiquitylation by
			the same E3 ubiquitin ligase<lb/> complex (SCF βTrCP ) that ubiquitylates IκB <ref
				type="biblio">87</ref> . This results<lb/> in p100 processing and the activation of
			the alternative<lb/> NF-κB dimer p52–RELB. Interestingly, although<lb/> TRAF2 and TRAF3
			are crucial for the activation of this<lb/> pathway, they also prevent its inadvertent
			activation in<lb/> unstimulated cells.<lb/></p>

		<p>In summary, TRAF3 serves as a platform that<lb/> mediates the coordinate binding of the
			E3 ubiquitin<lb/> ligase cIAP and its substrate NIK, thereby controlling<lb/> the
			intracellular levels of NIK and, consequently, the<lb/> activity of the alternative
			NF-κB pathway. Receptor-<lb/>triggered, TRAF2-dependent ubiquitylation of cIAP<lb/>
			redirects its substrate specificity towards TRAF3,<lb/> causing TRAF3 degradation and
			disruption of the cIAP–<lb/> NIK interaction, and this ultimately results in
			alternative<lb/> NF-κB activation.<lb/></p>

		<head>TRAF3 in disease<lb/></head>

		<p>Loss-of-function mutations that prevent the interaction<lb/> of TRAF3 with NIK or
			complete TRAF3 gene deletions<lb/> were identified in malignant cells from patients
			with<lb/> multiple myeloma <ref type="biblio">82,83</ref> . Moreover, a point mutation
			in<lb/> TRAF3 causing an amino acid substitution (R118W)<lb/> that decreases TRAF3
			stability, thereby resulting<lb/> in loss of function, was also identified in
			multiple<lb/> myelomas <ref type="biblio">82,83</ref> . This mutation is particularly
			interesting,<lb/> as it was independently identified as a germline<lb/> mutation in a
			patient with paediatric herpes simplex<lb/> encephalitis (HSE) <ref type="biblio"
				>88</ref> (see below). In multiple myeloma,<lb/> such TRAF3 mutations result in the
			accumulation of<lb/> NIK and constitutive activation of NF-κB signalling,<lb/> thereby
			promoting cancer cell survival <ref type="biblio">82,83</ref> . Interestingly,<lb/> some
			multiple myelomas exhibit small deletions in the<lb/> NIK gene that result in the
			expression of a truncated<lb/> protein that is more stable and active than
			wild-type<lb/> NIK because it cannot interact with TRAF3 <ref type="biblio">(REF.
				83)</ref>.<lb/> The end result of these mutations is the same as for
			loss-<lb/>of-function TRAF3 mutations, that is, accumulation of<lb/> activated
			NIK.<lb/></p>

		<p>Initially, the upregulation of NIK was thought to<lb/> contribute to the survival of
			multiple myeloma cells<lb/> through activation of IKKα and alternative NF-κB<lb/>
			signalling, but it was subsequently shown that NIK<lb/> exerts its pro-tumorigenic
			activity through the activation<lb/> of IKKβ and the classical NF-κB signalling pathway
				<ref type="biblio">82</ref> .<lb/> Treatment of multiple myeloma cells with a
			specific<lb/> IKKβ inhibitor, which has little inhibitory effect on IKKα<lb/> and
			alternative NF-κB signalling, resulted in the death of<lb/> at least 50% of multiple
			myeloma cell lines <ref type="biblio">82</ref> . By contrast,<lb/> short hairpin
			RNA-mediated silencing of IKKA in these<lb/> cells had little effect, if any, on their
			viability. Thus,<lb/> TRAF3 loss-of-function mutations seem to promote<lb/> classical
			NF-κB signalling, despite the fact that, under<lb/> physiological conditions, TRAF3 has
			a minimal role in<lb/> the activation of classical NF-κB signalling.<lb/></p>

		<p>As mentioned above, the R118W mutation of TRAF3<lb/> was also identified as a
			heterozygous germline mutation<lb/> in a patient with paediatric HSE <ref type="biblio"
				>88</ref> . This mutation not<lb/> only led to decreased TRAF3 stability, but it
			also reduced<lb/> the expression of wild-type TRAF3 (from the normal<lb/> allele),
			suggesting a dominant-negative effect exerted<lb/> through TRAF3 homotrimerization. As a
			result, the<lb/> total levels of TRAF3 were strongly reduced in various<lb/> cell types
			isolated from this patient. Still, unlike<lb/> TRAF3-deficient mice, which display a
			severe runting<lb/> disease that leads to premature death, this patient did<lb/> not
			display overt disease symptoms (apart from HSE),<lb/> indicating a residual function of
			the unaffected wild-<lb/>type allele. As the specific cell type that promotes<lb/>
			disease in TRAF3-deficient mice is unknown, it seems<lb/> possible that this cell type
			in the human patient still<lb/> contained functional levels of TRAF3. Consistent<lb/>
			with results from TRAF3-deficient mice, immune<lb/> cells from this patient exhibited
			defects in cytokine<lb/> production (particularly in type I IFN expression)<lb/>
			following stimulation with different TLR, RIG-I and<lb/> MDA5 agonists, and in response
			to viral infection <ref type="biblio">88</ref> .<lb/> Together with observations from
			patients with defects<lb/> in TLR3, these findings suggest that TRAF3-dependent<lb/>
			type I IFN production downstream of TLR3 and<lb/> TRIF is particularly important for
			immune defence<lb/> against HSE <ref type="biblio">89</ref> .<lb/></p>

		<p>Cells from the patient with the TRAF3 R118W<lb/> mutation also displayed constitutive
			p100 processing,<lb/> as expected <ref type="biblio">81,90</ref> , but the contribution
			of alternative<lb/> NF-κB signalling to the phenotype of this patient<lb/> remains
			obscure. It is possible that this patient will<lb/> be at a higher risk of developing
			multiple myeloma<lb/> in old age. Unlike TRAF3-deficient cells from mice,<lb/> simian
			virus 40 (SV40)-transformed human fibroblasts<lb/> expressing R118W-mutant TRAF3 showed
			defects<lb/> in TLR-induced nuclear translocation of the NF-κB<lb/> p65 subunit <ref
				type="biblio">88</ref> . However, it is not known whether<lb/> this defect in NF-κB
			activation is directly due to<lb/> the specific TRAF3 mutation (possibly as a result
			of<lb/> uncharacterized dominant-negative effects on other<lb/> TRAFs) or reflects
			differences in TRAF3 function<lb/> between mouse and human cells. Nevertheless, the<lb/>
			results confirm that TRAF3 has an essential role in<lb/> type I IFN induction and
			antiviral immunity in both<lb/> mice and humans <ref type="biblio">47,57,88</ref>
			.<lb/></p>

		<head>Remaining questions<lb/></head>

		<p>Undoubtedly, TRAF3 has gone from a poorly<lb/> understood member of the TRAF family to an
			important<lb/> multifunctional regulatory protein within a few years.<lb/> Clearly,
			TRAF3 is an essential activator of type I IFN<lb/> production and an important regulator
			of balanced<lb/> cytokine production by immune cells. Nonetheless,<lb/> there are still
			several open questions regarding the<lb/> function and regulation of TRAF3. For example,
			what<lb/> are the TRAF3-interacting proteins that control its<lb/> recruitment into
			different signalling complexes? What<lb/> are the substrates for TRAF3-dependent
			K63-linked<lb/> ubiquitylation? Are there as-yet-unidentified molecules<lb/> that
			control TRAF3 function?<lb/></p>

		<p>Given the essential functions of TRAF3 in different<lb/> signalling pathways that control
			inflammation, antiviral<lb/> immunity and cell survival, and its already
			established<lb/> roles in human health, it is likely that further mechanistic<lb/>
			insights into TRAF3 signalling will be valuable for<lb/> understanding the pathogenesis
			of TRAF3-associated<lb/> diseases and for the design of adequate therapeutic<lb/>
			approaches.</p>


	</text>
</tei>