Negative regulation of NF-κB and its involvement in rheumatoid arthritis

The transcription factor NF-κB plays crucial roles in the regulation of inflammation and mmune responses, and inappropriate NF-κB activity has been linked with many autoimmune and inflammatory diseases, including rheumatoid arthritis. Cells employ a multilayered control system to keep NF-κB signalling in check, including a repertoire of negative feedback regulators ensuring termination of NF-κB responses. Here we will review various negative regulatory mechanisms that have evolved to control NF-κB signalling and which have been implicated in the pathogenesis of rheumatoid arthritis.


Introduction
Rheumatoid arthritis (RA) is the most common infl ammatory arthritis, aff ecting up to 1% of the adult population. RA is characterised by a symmetrical polyarthritis in which chronic infl ammation of joints is associated with a progressive destruction of cartilage and bone, leading to functional decline and disability. Infi ltration of cells of the innate and adaptive immune system into the joint space drives the local production of proinfl ammatory T-helper type 1 and T-helper type 17 cytokines, chemokines, and matrix metalloproteinases by infi ltrating monocytes and synovial cells. Proliferation of synovial fi broblasts leads to the formation of pannus tissue, which invades and degrades articular cartilage and subchondral bone.
Th e aetiology of RA is still not understood, but it is well accepted that activation of NF-κB-dependent gene expres sion plays a key role in the development of RA and many other autoimmune diseases. NF-κB represents a family of structurally related and evolutionarily conserved proteins (p100 or NF-κB2, p105 or NF-κB1, p65 or RelA, RelB, c-Rel) that function as homodimers or heterodimers [1], and that regulate the expression of a large number of genes -such as TNF, IL-1, IL-6, cyclooxygenase-2, chemokines, inducible nitric oxide synthase, and matrix metalloproteinases -that are involved in RA. In addition, TNF and IL-1 are themselves very potent activators of NF-κB (reviewed in [2,3]).
NF-κB activation can be detected in cultured synovial fi broblasts and synovial tissue from RA patients, and animal models of infl ammatory arthritis also demonstrate the active role of NF-κB in the development and progression of RA (reviewed in [4]). Th e time course of NF-κB activation appears to precede the onset of disease, and blockade of NF-κB by diff erent means decreases disease severity [5,6].
Next to its role in proinfl ammatory gene expression, NF-κB is also essential for osteoclastogenesis, mainly by mediating the eff ects of receptor activator of NF-κB ligand (RANKL). Defects in the regulation of osteo clastogenesis are the major cause of bone erosion in osteolytic diseases such as RA [7].
Finally, recent discoveries revealing a genetic association with several genes relevant to NF-κB signalling, including CD40, TRAF1, TNFAIP3, and c-REL, further highlight the importance of NF-κB activation in RA pathogenesis [8].

Pathological triggers of NF-κB signalling in RA
Since NF-κB is central to the process of infl ammation in RA, much research deals with the identifi cation of the molecular triggers that activate NF-κB in RA. It is well accepted that proinfl ammatory cytokines such as TNF and IL-1 play an important role, and administration of TNF antagonists is an eff ective treatment for severe RA (reviewed in [9]). TNF and IL-1 are both very potent activators of NF-κB and it can be expected that NF-κB activation by these cytokines mediates most of their proinfl ammatory activities in RA (reviewed in [3]). NF-κB activation by receptor activator of NF-κB (RANK), a TNF receptor family member, is important for osteoclastogenesis, and defects in proper RANK-NF-κB signalling are likely to be involved in RA pathology and other diseases associated with bone loss [7]. CD40 is another TNF receptor family member that is functionally expressed on a variety of cell types, including smooth muscle fi bro blasts from normal and RA patients and RA synovial cells, B cells, macrophages, and dendritic cells, and can be upregulated by proinfl ammatory cytokines including TNF [10]. Binding of the CD40 ligand (CD154), which is transiently expressed on the surface of activated CD4 + T cells, triggers NF-κB activation resulting in fi broblast proliferation and secretion of proinfl ammatory cytokines and chemokines, which contributes to joint destruction. However, studies with antagonistic anti-CD40 or anti-CD154 antibodies led to the conclusion that CD40 signals may be important at the initial stages of arthritis induction, but are not required once disease is established and pathogenic antibodies are already present [11,12]. Enhanced expression of the TNF receptor family member B-cell activating factor (BAFF), allowing the survival of autoantibody-producing B lymphocytes, is also characteristic for RA, and antagonists of BAFF have been developed to counter RA [13]. Finally, lymphotoxin β receptor signalling has been implicated in tertiary lymphoid organ formation at sites of chronic infl ammation including RA [13].
Toll-like receptors (TLRs) have been implicated in a variety of autoimmune diseases and are potential candidates for inducing NF-κB-dependent infl ammation in RA. In addition to microbial ligands, an increasing number of endogenous ligands -a group of proteins derived from host tissues and cells -have been reported as candidate activators of TLRs inducing so-called sterile infl ammation (reviewed in [14,15]). TLRs are expressed in RA synovial tissues and various endogenous ligands are present within the infl amed joints of RA patients. Moreover, animal models using TLR knockout mice or strategies to block TLR signalling clearly identify TLRdependent infl ammation as being important in the pathogenesis of the disease.
High mobility group box chromosomal protein 1 (HMGB1), a highly conserved chromatin component that can be actively secreted by macrophages or passively released by necrotic cells, is one of the most putative endogenous TLR4 ligands involved in RA pathology. HMGB1 is increased in RA synovial tissue and HMGB1 neutralising antibodies or the antagonistic BoxA domain of HMGB1 protect against collagen-induced arthritis in mice [16]. Myeloid-related protein 8 and myeloid-related protein 14, damage-associated molecular pattern molecules belonging to the S100 family of calcium-binding proteins, are also abundantly present in RA synovial fl uid, and have been suggested to be involved in TLR4induced chronic infl ammation in RA [17,18]. Other endogenous TLR ligands that may be involved in RA pathology are extracellular matrix components such as fi brinogen, fi bronectin, biglycan, tenascin C, and hyaluronic acid fragments (reviewed in [14,15]). Together, these studies suggest that several TLR ligands in the infl amed joint tissue may contribute to NF-κB activation and infl ammatory gene expression in RA.
Th e canonical NF-κB pathway plays a major role in innate and adaptive immunity, and is triggered by many stimuli including proinfl ammatory cytokines (for example, TNF, IL-1), antigens, RANKL, and TLR ligands. NF-κB signalling initiated by diff erent receptors requires the formation of proximal protein-protein interactions that are often receptor specifi c, but ultimately converge in the activation of the IκB kinase (IKK) complex, which mediates phosphorylation of the inhibitory IκB protein leading to its K48-polyubiquitination and degradation by the proteasome [20]. Th e IKK complex is comprised of the two catalytic subunits IKK1 and IKK2 (also known as IKKα and IKKβ) and the regulatory subunit NF-κB essential modulator (NEMO -also known as IKKγ) ( Figure 1). Gene targeting experiments showed that IKK2 and NEMO, but not IKK1, are required for canonical NF-κB activation [21].
One of the best studied NF-κB signalling pathways is the TNF pathway. TNF stimulation results in the recruitment of TNF receptor-1-associated death domain (TRADD) protein and of receptor interacting protein 1 (RIP1), which function as adaptor proteins for the E3 ubiquitin ligases TNF receptor-associated factor (TRAF) 2 and TRAF5, which in turn bind the E3 ubiquitin ligases cellular inhibitor of apoptosis (cIAP) 1 and cIAP2 ( Figure 1). On TNF stimulation, TNF-receptor bound RIP1 is rapidly modifi ed by K63-linked polyubiquitin chains. TRAF2/5 and cIAP1/2 are good candidates for RIP1 ubiquitination, but the specifi c role of each is still unclear. Th e poly ubiquitin chains on RIP1 are believed to create a scaff old to recruit the IKK and TAK1 complex via the ubiquitin-binding proteins NEMO and TAB1/2, respec tively. Th e recent identifi cation of a distinct E2/E3 enzyme complex that modifi es NEMO with linear K63-linked polyubiquitination of TRAFs, RIP1 and IRAK1, is recognised by NEMO and TAB proteins, resulting in the recruitment and activation of respectively IKK2 and TAK1. TAK1 then phosphorylates and activates IKK2, which in turn phosphorylates IκBα, triggering its K48-linked ubiquitination and proteasomal degradation. This allows NF-κB (here shown as a heterodimer of p65 and p50) to translocate to the nucleus and promote target gene expression. TRAF1, which has no ubiquitin ligase activity, can negatively regulate NF-κB activation, most probably by competing with other TRAFs. A20 and CYLD are deubiquitinating enzymes that control NF-κB activation by targeting specifi c signalling proteins including RIP1 and TRAF6, to which they are recruited using specifi c ubiquitin-binding adaptor proteins such as ABIN-1 and p62. miR-146 is thought to negatively regulate TLR signalling by inhibiting expression of IRAK1 and TRAF6. Finally, TLR signalling can also be inhibited by the transmembrane protein SIGIRR, which has been proposed to compete with TLR4 for binding to IRAK1 and TRAF6. The expression of many of these negative regulatory molecules is NF-κB dependent, implicating them in the negative feedback regulation of NF-κB activation. ABIN, A20-binding inhibitor of NF-κB; cIAP, cellular inhibitor of apoptosis; CYLD, cylindromatosis; IKK, IκB kinase; IκB, inhibitor of NF-κB; IRAK, IL-1R-associated kinase; MyD88, myeloid diff erentiation primary response gene 88; NEMO, NF-κB essential modulator; NF, nuclear factor; RANK, receptor activator of NF-κB; RIP1, receptor interacting protein 1; SIGIRR, single-immunoglobulinIL-1 receptor-related; TIR, Toll-like receptor/IL-1R; TRAF, TNF receptor-associated factor; TLR, Toll-like receptor; TNF, tumour necrosis factor. poly ubiquitin chains and is essential for TNF-activated NF-κB signalling adds further complexity [22]. Th e exact role of protein-anchored polyubiquitin chains remains unclear, as it was recently suggested that unanchored polyubiquitin chains can directly activate the TAK1 complex [23]. Similar signalling principles apply to other receptors. For example, TLR4 stimulation by lipopolysaccharide induces the recruitment of Toll/IL-1 receptor adaptor protein (also referred to as Mal) and TRIF-related adap tor molecule (TRAM), which most probably serve as bridg ing factors to recruit myeloid diff erentiation primary response gene 88 (MyD88) and TIR domain-contain ing adaptor-inducing IFNβ (TRIF), respectively. MyD88 in turn recruits members of the IL-1R-associated kinase (IRAK) family and TRAF6, leading to oligomerisa tion and selfubiquitination of TRAF6 [24]. TRIF also recruits TRAF6 [25] and RIP1 [26] via a direct inter action. Both pathways then activate TAK1 and IKK in a ubiquitination-dependent manner similar to the TNF pathway ( Figure 1).
Th e noncanonical NF-κB pathway can be activated by the lymphotoxin β receptor, BAFF receptor, CD40, and RANK ( Figure 2). In this pathway, p100 is processed by the protea some to p52, which together with the RelB NF-κB subunit regulates a distinct set of target genes that control B-cell development, secondary lymphoid organ development, and osteoclastogenesis [27] Th e noncanonical NF-κB path way is strictly dependent on IKK1, which is acti vated upon phosphorylation by NF-κB inducing kinase (NIK). NIK is predominantly regulated at the post-translational level and is present at extremely low levels in most cell types. In unstimulated cells, NIK occurs in a cytoplasmic complex with TRAF2, TRAF3, and cIAP1/2, which K48polyubiquitinates NIK, leading to its continuous degra dation by the proteasome. Receptor ligation has been shown not only to remove TRAF3 from this complex by recruiting it to the receptor, but also to attract TRAF2 and cIAP1/2, which are essential for subsequent TRAF3 degradation. All this contributes to releasing NIK from its constitutive degradation, resulting in NIK accumu lation and IKK1 phosphorylation [28,29] (Figure 2).
It should be mentioned that CD40, lymphotoxin β receptor and RANK mediate the activation of both canonical and noncanonical NF-κB signalling pathways. Upon binding of their ligand, CD40 and RANK interact with several TRAF members, including TRAF1, TRAF2, TRAF3, TRAF5, and TRAF6, and this leads to the proteolysis of both TRAF2 and TRAF3, which represents an important step in the activation of the noncanonical pathway as described above. Specifi c TRAF molecules are associated with overlapping and distinct CD40mediated functions. For example, in B cells TRAF6 is required for CD40-mediated JNK activation and IL-6 production, while TRAF2 is required for activation of NF-κB, and TRAF3 serves as a negative regulator of CD40 signalling [30,31].

Negative regulation of NF-κB signalling in RA
Since NF-κB activation is so crucial to many cellular processes, a tight regulation of the NF-κB signalling pathway and the genes it induces is an absolute requirement to fi ne-tune the infl ammatory response. Moreover, terminating an NF-κB response is essential to prevent persistent NF-κB activation that may lead to chronic infl ammation and/or tumorigenesis. To achieve this, cells employ diff erent mechanisms, including the expression of inhibitory proteins that downregulate NF-κB signalling [32]. Below we give an overview of a number of proteins that are involved in the dampening or termination of the NF-κB response, some of them under the control of NF-κB itself and thus acting in a negative feedback loop. In addition, we discuss the potential role of these NF-κB inhibitory factors in the immuno pathology of RA. Several other proteins involved in the negative regulation of NF-κB-dependent infl ammatory responses, such as MyD88s, IRAK-M, and TOLLIP, have been described (reviewed in [33]). Th ese proteins are not discussed here, since a link with RA pathology has not yet been reported.
Although IKK1 is a critical component of the noncanonical NF-κB pathway, it should be mentioned that this kinase also plays a prominent role in the negative regulation of both canonical and noncanonical NF-κB pathways. Macrophages from IKK1-defi cient mice or knockin mice expressing inactive IKK1 show increased production of proinfl ammatory cytokines as a result of enhanced IKK2 activation and IκBα degradation [34]. IKK1 has also been shown to inhibit nuclear NF-κB and to downregulate proinfl ammatory signalling by phosphory lating STAT1 [35]. Interestingly, a recent study has demonstrated that IKK1 phosphorylates NIK in negative feedback regulation of the noncanonical NF-κB pathway [36], supporting the idea that IKK1 plays important roles in terminating both canonical and noncanonical NF-κB pathways with possible implications for chronic infl ammatory diseases like RA.
Next to its role in suppressing NF-κB activation, A20 is also a strong inhibitor of apoptosis, at least in some cell types. Th e mechanisms by which A20 regulates apoptotic signalling, however, are still elusive. A20-defi cient mice spontaneously develop multiorgan infl ammation and cachexia and die within 2 weeks of birth, illustrating the potent anti-infl ammatory function of this molecule [46]. CD40 and RANK can activate the noncanonical NF-κB pathway that is dependent on NF-κB inducing kinase (NIK) expression levels. In unstimulated cells NIK forms a cytosolic complex with the ubiquitin ligases TRAF2, TRAF3 and cIAP1/2, which facilitates the K48-linked polyubiquitination and proteasomal degradation of NIK, keeping NIK levels low. Upon ligand binding, TRAF3 is recruited to the receptor, where TRAF2 directs nondegradative K63-linked polyubiquitination of cIAP1/2, resulting in their activation. Subsequently cIAP1/2 directs its K48-linked polyubiquitination to TRAF3, rather than NIK. As a result, TRAF3 is degraded and NIK is stabilised, resulting in increased NIK levels in the cell. NIK then phosphorylates and activates IKK1, which mediates NF-κB p100 phosphorylation. This is followed by K48-linked polyubiquitination and partial proteasomal degradation of p100 to p52, which forms a heterodimer with RelB to activate transcription. Next to TRAF3, TRAF1 has also been identifi ed as a negative regulator of this pathway, most probably by competing with other TNF receptor-associated factors. cIAP, cellular inhibitor of apoptosis; IKK, IκB kinase; NF, nuclear factor; RANK, receptor activator of NF-κB; TRAF, TNF receptor-associated factor; TNF, tumour necrosis factor. A20-defi cient cells are also more susceptible to TNFmediated apoptosis, confi rming its role as an anti apop totic protein. We recently showed that mice specifi cally lacking A20 in intestinal epithelial cells exhibit increased susceptibility to experimental colitis due to the hyper sensitivity of their intestinal epithelial cells to TNF-induced apoptosis, confi rming A20 as a major anti apoptotic protein in the intestinal epithelium [47]. Two independent studies showed that mice lacking A20 in B cells develop autoimmunity due to hyperactive NF-κB responses in B cells leading to unrestricted B-cell survival [48,49]. A20 expression has been observed in several cell types that play important roles in the pathophysiology of RA, such as fi broblasts, synoviocytes and lymphocytes. Interest ingly, A20 expression is itself regulated by NF-κB [50], implicating A20 in the negative feedback regulation of NF-κB signalling. Recently, intra-articular injection of an A20-expressing adenovirus was shown to reduce the severity of synovial infl ammation and joint destruction in a mouse model of collagen-induced arthritis, even in untreated joints, in both a prophylactic and therapeutic setting. A20 expression in synovial tissue was associated with inhibition of NF-κB activity and decreased levels of TNF, IL-1β, IL-6, soluble RANKL, monocyte chemoattractant protein 1, and IL-17, suggesting that A20 induces a protective eff ect in collagen-induced arthritis mice through suppression of NF-κB activation and NF-κB-dependent gene expression. [51]. Because TNF and IL-1β are known to mediate synovitis, pannus formation, and erosion of cartilage and bone in RA, the decreased serum levels of TNF and IL-1β in A20transduced mice might explain the benefi cial eff ects in the clinical, pathological, and radiological fi ndings. Th is study also demonstrated that A20 overexpression leads to a decrease in the number of activated osteoclasts in joint tissue. A20 might therefore minimise joint destruction through decreasing the osteoclast number and activity. It will be interesting to analyse in future the susceptibility of conditional knockout mice that lack A20 in specifi c cell types such as synovial fi broblasts, macrophages, dendritic cells, B cells or T cells.
Importantly, several SNPs in the human A20 locus have been associated with increased susceptibility to development of autoimmune pathologies (reviewed in [52]). Several genome-wide association studies also revealed a clear association between mutations in the A20 locus in the 6q23 chromosome and susceptibility to RA [52]. Although the identifi ed variants are not located in a gene, they are thought to infl uence A20 as its nearest gene (~150 kb downstream of A20), probably by the presence of potential regulatory DNA elements in this region. As A20 is required for termination of TNF-induced signals, and TNF is the primary infl ammatory cytokine in RA, these fi ndings reveal A20 as a candidate susceptibility locus for RA. How these variants aff ect normal A20 activity and how they cause RA, however, remain unclear. Recently, Elsby and colleagues functionally evaluated in vitro the regulatory ability of RA-associated SNP variants on A20 promoter activity, and could show repressed A20 trans cription for some of the SNPs investigated [53]. It will be of interest to identify the actual causal variants and to elucidate the functional consequences of these variants. In this context, knockin mice for the corresponding A20 SNPs, combined with mouse models for RA, will be very valuable tools.
Using oligonucleotide microarray analysis, ABIN-1 was identifi ed among TNF-induced genes in human synoviocytes, and high levels of ABIN-1 mRNA were detected in RA tissue biopsies, indicating a potential role for ABIN-1, together with A20, in the negative feedback regulation of NF-κB signalling and the pathogenesis of RA [59]. In a recent study, inhibition of TLR responses by immunoreceptor tyrosine-based activation motif (ITAM)-coupled recep tors was shown to depend on the expression of the NF-κB inhibitory proteins ABIN-3 and A20. Moreover, this protective eff ect of the ITAM was strongly suppressed in infl ammatory arthritis synovial macrophages [60]. Interest ingly, ABIN-1 poly morphisms have been associated with psoriasis and systemic lupus erythema tosus in humans [61,62]. A high degree of overlap between systemic lupus erythematosus and RA susceptibility loci might be expected as the two diseases show some clinical overlap in joint involvement, autoantibody production, systemic features and response to treatments such as Bcell depletion (rituximab). It will therefore be interesting to investigate whether SNPs that have been reported to be associated with systemic lupus erythematosus are also associated with RA.

Cylindromatosis protein
Cylindromatosis (CYLD) protein was originally identifi ed as a tumour suppressor involved in familial cylindromatosis [63]. CYLD is also involved, however, in diverse physiological processes ranging from immunity and infl ammation to cell cycle progression, spermatogenesis, and osteoclastogenesis (reviewed in [64]). CYLD is a deubiquitinating enzyme that negatively regulates NF-κB signalling initiated by TNFR, RANK and T-cell receptor stimulation (reviewed in [64]) (Figure 1), by deubiquitinating several NF-κB signalling proteins including NEMO, TRAF2, TRAF6, TRAF7, RIP1 and TAK1. Many of these are also targeted by A20 and it is still not clear why the cell needs A20 as well as CYLD to control NF-κB activation. As A20 is only expressed in many cell types upon stimulation, it has been suggested that this protein mainly regulates later phases of NF-κB signalling, whereas CYLD would regulate constitutive and early signalling. In addition, their relative activity might also be cell-type dependent [64].
Interestingly, CYLD has been shown to negatively regulate RANK signalling and osteoclastogenesis in mice [65]. Mice with a genetic defi ciency of CYLD have aberrant osteoclast diff erentiation and develop severe osteoporosis. Osteoclast precursors of these mice are hyper-responsive to RANKL-induced diff erentiation and produce more and larger osteoclasts. CYLD expression is markedly upregulated under conditions of RANKLinduced osteoclastogenesis and is recruited to ubiquitinated TRAF6 via the ubiquitin-binding adaptor protein p62 (also known as sequestosome 1) [65], followed by the CYLD-mediated deubiquitination of TRAF6. In this context, it is worth mentioning that transgenic mice expressing a mutated form of p62 also display abnormal osteoclastogenesis and develop progressive bone loss [66]. Th ese fi ndings suggest that CYLD-mediated inhibition of RANK-induced NF-κB signalling plays a key role in the negative regulation of osteoclastogenesis and indicate CYLD as a potential genetic factor involved in the pathology of bone disorders such as RA.

Single-immunoglobulin IL-1 receptor-related protein
Single-immunoglobulin IL-1 receptor-related (SIGIRR) protein, also known as TIR-8, is a member of the TLR/ IL-1R family that has been extensively characterised as an inhibitor of IL-1R and TLR signalling, probably through direct interaction with these receptors, MyD88, IRAK1 or TRAF6 [67] (Figure 1). Given the important role of IL-1R and TLR signalling in the chronic infl ammation observed in RA [68], a regulatory role for SIGIRR in RA is not unlikely. SIGIRR has a very restricted expression pattern, being expressed in epithelial cells, monocytes and immature dendritic cells, but not in mature macrophages [69,70]. Recently, SIGIRR overexpression was shown to inhibit the spontaneous release of infl ammatory cytokines by human RA synovial cells. Th is inhibitory function of SIGIRR was further confi rmed in vivo, since SIGIRR-defi cient mice developed a more severe disease in zymosan-induced arthritis, as well as collagen-induced arthritis mouse models [70]. It will be interesting to compare the expression of SIGIRR in RA patients with its expression in control patients, or to investigate whether the function of SIGIRR is compromised in RA patients. Because of its restricted expression pattern, SIGIRR may also be an interesting therapeutic target in RA.

TNF receptor-associated factor 1
TRAF1 is a unique member of the TRAF protein family because it lacks a RING fi nger domain and therefore lacks ubiquitin ligase activity. Accumulating data support a role for TRAF1 as both a negative and a positive modulator of NF-κB signalling by certain TNF family receptors, possibly in a cell-type-dependent manner [71,72]. Expression of TRAF1 is inducible by TNF and overexpression of TRAF1 inhibits TNF-induced NF-κB activation. TRAF1-defi cient T cells are hyper-responsive to TNF, with enhanced proliferation and activation of the NF-κB signalling pathway. TRAF1 also functions as a negative regulator of CD40-induced NF-κB activation. TRAF1-defi cient dendritic cells, however, show attenuated responses to secondary stimulation by TRAF2-dependent factors, suggesting a positive regulatory role in these cells. Th e mechanism by which TRAF1 modulates NF-κB activation is still unclear. Most probably, TRAF1 competes with TRAF family members for binding to the receptor or other signalling proteins. Alternatively, TRAF1 might recruit A20 with which it can physically interact [73]. A genome-wide association study examining more than 300,000 SNPs among approximately 1,500 autoantibody-positive RA cases and 1,800 controls identifi ed a genetic variation at the TRAF1-complement component 5 locus as an impor tant RA risk locus [74]. Subsequent work indicates that TRAF1 is more likely to be the causative locus [75]. Recent work in a Korean population also demonstrates genetic association of the TRAF1 region with RA [76]. protein expression and the promotion of target mRNA degradation [77]. miR-146a/b and miR-155 are of particular interest for infl ammatory signalling to NF-κB, since these miRNAs can be induced by infl ammatory stimuli such as IL-1β, TNF and TLRs [78,79]. In addition, miR-146a is an NF-κB-dependent gene, and the NF-κB signalling molecules IRAK1 and TRAF6 were identifi ed as target genes of miR-146a [78] (Figure 1). Similarly, miR-155 was shown to target transcripts for the NF-κB signalling molecules IKKε and RIP1 [79,80]. Notably, both miR-146 and miR-155 are expressed at higher levels in RA synovial fi broblasts and synovial tissue [81,82], as well as in peripheral blood mononuclear cells of RA patients [83]. miR-146a is also overexpressed in CD4 + and IL-17producing T cells from RA patients [84,85]. Interestingly, a polymorphism in the 3'-UTR of the miR-146a target gene was recently shown to be associated with RA susceptibility [86]. miR-155 overexpression in synovial fi broblasts was able to prevent the TLR and cytokineinducible expression of specifi c matrix metalloproteinases that mediate tissue destruction in RA [81]. Moreover, miR-155 was shown to promote TNF production, a key process in the pathogenesis of RA [87]. miR-146 and miR-155 may therefore be important negative regulators of infl ammation in RA and their potential for the development of new treatments is substantial. In addition, their increased expression in RA patients is potentially useful as a marker for disease diagnosis, progression, or treatment effi cacy [88], but this will require confi rmation using a large and well-defi ned cohort.
Besides miR-146 and miR-155, a number of other miRNAs with a potential role in the control of NF-κBdependent infl ammatory responses in RA pathology were recently identifi ed. In this context, miR-124a -a key regulator of the chemokine monocyte chemoattractant protein 1 -was shown to be decreased in synoviocytes from RA patients [89]. Similarly, elevated levels of miR-203 -leading to increased secretion of matrix metallo proteinase-1 and IL-6 -were detected in RA synovial fi broblasts [90]. Finally, miR-16, miR-132, and miR-223 were also shown to have an altered expression in RA patients, indicating their potential as diagnostic biomarkers for pathogenesis [83,88,91].

Conclusions
Th e NF-κB family of transcription factors plays crucial roles in the infl ammatory processes in RA leading to cartilage and bone destruction. Keeping NF-κB activation under control can thus be very important for the design of specifi c therapeutics. Th e existence of multiple negative regulators ensuring a tight regulation of the NF-κB pathway, however, raises the question of the specifi c role of each of these regulators and the relationship between them. In addition, given the number of miRNAs in humans and the multiple mRNAs they target, intense complexity can be expected. How all these regulatory signals are themselves regulated will be an important question in order to clarify how NF-κB signalling is organised, and, more importantly, how this knowledge may lead to new treatments for infl ammatory diseases such as RA.

Competing interests
The authors declare that they have no competing interests.