Glycogen synthase kinase 3β ubiquitination by TRAF6 regulates TLR3-mediated pro-inflammatory cytokine production

TRAF6 is critical for the production of inflammatory cytokines in various TLR-mediated signalling pathways. However, it is poorly understood how TRAF6 regulates TLR3 responses. Here we demonstrate that GSK3β interacts with TRAF6 and positively regulates the TLR3-mediated signalling. Suppression of GSK3β expression or its kinase activity drastically reduces the production of inflammatory cytokines and the induction of c-Fos by decreasing ERK and p38 phosphorylation. GSK3β physically associates with TRAF6 in a TLR3 ligand poly I:C-dependent manner. TRAF6 is determined to be a direct E3 ligase for GSK3β, and TRAF6-mediated GSK3β ubiquitination is essential for poly I:C-dependent cytokine production by promoting the TLR3 adaptor protein TRIF-assembled signalling complex.

T oll-like receptors (TLRs) comprise a class of conserved type I transmembrane pattern recognition receptors 1 that recognize pathogen-associated molecular patterns and play a critical role in the host cell defence against microbial pathogens 2 . The recognition of pathogen-associated molecular patterns by TLRs activates multiple pathways that mediate immune responses to produce immune mediators, including pro-inflammatory cytokines, chemokines and type I interferons (IFNs) [2][3][4] . In particular, TLR3 signalling through the recognition of double-stranded RNA is crucial for antiviral responses [5][6][7] . Upon ligand binding to TLR3, the sole cytoplasmic adaptor molecule toll-interleukin 1 receptor homology-domaincontaining adapter-inducing interferon-b (TRIF) is recruited to the TLR signalling complex 8,9 . The TLR3-TRIF signalling complex further triggers the recruitment of downstream signalling molecules, including tumour necrosis factor (TNF) TNF-α mRNA (fold) ** Figure 1 | GSK3b but not GSK3a is involved in TLR3-mediated pro-inflammatory cytokine production. (a) RAW264.7 cells stably expressing control shRNA (Con) or GSK3b-specific shRNA (sh-GSK3b) were stimulated with 10 mg ml À 1 poly I:C for 1 h. Levels of IL-6, TNF-a, IP-10, IL-12 and IL-10 mRNA were determined by real-time PCR analysis, and the values were normalized to b-actin mRNA expression. GSK3b knockdown was confirmed by reverse transcription-PCR and western blotting. (b) BMDMs were treated with 10 mg ml À 1 poly I:C for the indicated time points. Whole-cell lysates were immunoblotted with antibodies to the molecules along the right margin. (c) HEK293-TLR3 cells were transiently transfected with V5-GSK3a or HA-GSK3b plasmids. Levels of IL-6 and TNF-a mRNA were determined as described in a. Expression of the transduced proteins was detected by western blotting with anti-V5 (for GSK3a) and anti-HA (for GSK3b). (d) BMDMs were preincubated for 1 h with or without 10 mM SB216763 and stimulated with or without 10 mg ml À 1 poly I:C for 20 h. Levels of IL-6, TNF-a and IL-10 in culture supernatants were determined by enzyme-linked immunosorbent assay. (e) Gsk3b À / À MEFs were transiently transfected with V5-GSK3a, HA-GSK3b (WT) or GSK3b (K85A) plasmids, and cells were stimulated with 10 mg ml À 1 poly I:C for 1 h. Levels of IL-6, TNF-a and IL-10 mRNA were determined as described in a. Expression of the transduced proteins was detected by western blotting as described in c. Data are presented as the mean ± s.d. from at least three independent experiments. Statistical analyses were calculated using the Student's t-test (*Po0.05; **Po0.01; NS, not significant). receptor-associated factor 3 (TRAF3) 10 , TRAF6 (ref. 11) and receptor-interacting protein 1 (RIP1) 12 , which lead to the activation of IFN regulatory factor 3 (IRF3) 13 , activator protein 1 (AP1) 14 and nuclear factor-kappa B (NF-kB) 15 . While the TLR3-mediated signalling pathways in type I IFN production have been well explored, little is known about their regulatory mechanisms in pro-inflammatory cytokine production.
Glycogen synthase kinase 3 (GSK3) is a highly conserved serine/threonine kinase that was originally identified as a regulator of glycogen metabolism 16 . Two highly related isoforms of GSK3 exist, GSK3a and GSK3b, and they are ubiquitously expressed in mammalian tissues 17 . Although both isoforms share similar structural features, they are not functionally identical 18 . GSK3b plays crucial roles in various signal pathways that regulate multiple cellular functions, including metabolism, cell proliferation, differentiation and development [19][20][21] . GSK3b is also involved in diverse TLR signalling 22,23 . For example, GSK3b has been identified as a key mediator of pro-inflammatory cytokine production, including interleukin (IL)-6, TNF-a, IL-12p40, IL-1b and IFN-g, and anti-inflammatory cytokine IL-10 production by regulating CREB activity in Myd88-dependent TLR pathways 24,25 . In addition, GSK3b differentially regulates the production of lipopolysaccharide (LPS)-induced IL-Ib and the endogenous IL-1 receptor antagonist through ERK activity 26 . Another study demonstrated that GSK3b regulated IFN-g-induced signal transducer and activator of transcription 3 (STAT3) activity and was required for the synergistic action of LPS and IFN-g on IL-6 cytokine production 27 . Although these studies clearly document the importance of GSK3b in TLR-mediated cytokine production, little is known about the role of GSK3b in TLR3 signalling.
In this report, we show that GSK3b is essential for TLR3-mediated ERK and p38 activation, c-Fos induction and pro-inflammatory cytokine production. We also find that GSK3b undergoes a lysine (K)-63 chain ubiquitination, which is important for assembling the TRIF signalling complex for TLR3 signalling. Our findings provide insights into the molecular mechanisms underlying the regulatory function of GSK3b in TLR3-mediated pro-inflammatory cytokine production.
Although both GSK3a and GSK3b were phosphorylated in response to poly I:C (Fig. 1b), overexpression of GSK3b but not its homologue GSK3a significantly elevated IL-6 and TNF-a mRNA expression in a dose-dependent manner (Fig. 1c). We next used a GSK3 inhibitor SB216763 to determine whether the kinase activity of GSK3 was responsible for inflammatory cytokine production. GSK3 inhibition with SB216763 resulted in a substantial reduction in IL-6, TNF-a and IL-10 levels compared with the levels in untreated cells after poly I:C stimulation (Fig. 1d). Importantly, poly I:C-induced mRNA and protein expression levels of IL-6, TNF-a and IL-10 were impaired in Gsk3b À / À mouse embryonic fibroblasts (MEFs) ( Fig. 1e; Supplementary Fig. 2). Reconstitution with GSK3b but not GSK3a or the kinase inactive GSK3b (K85A) mutant into Gsk3b À / À MEFs by transient overexpression restored the poly I:C-induced mRNA expression of IL-6, TNF-a and IL-10 ( Fig. 1e). Consistently, silencing of GSK3b but not GSK3a in bone marrow-derived macrophages (BMDMs) inhibited induction of inflammatory cytokines, including IL-6, TNF-a and IL-10 ( Supplementary Fig. 3a-c). Thus, these results suggest that GSK3b positively regulates TLR3-mediated inflammatory cytokine production and that the kinase activity of GSK3b is required for its role in poly I:C-induced cytokine production.
GSK3b is involved in retinoic acid-inducible gene 1-like receptor (RLR)-mediated antiviral response 31 . Since TLR-and RLR-mediated signalling pathways share a number of components, we examined whether GSK3b is also involved in poly I:C-stimulated IRF3 activation and IFN-b induction. We found that poly I:C stimulation induced increased phosphorylation of TBK1, a kinase responsible for phosphorylating IRF3 after poly I:C stimulation, in wild-type MEFs but not Gsk3b À / À MEFs ( Supplementary Fig. 4a). Consistently, deficiency of GSK3b markedly inhibited poly I:C-induced IRF3 phosphorylation and nuclear translocation ( Supplementary Fig. 4a,b) as well as dimerization of IRF3 ( Supplementary Fig. 4c). Consistently, neither Gsk3b À / À MEFs nor GSK3b knockdown RAW264.7 cells showed significant induction of IFN-b upon poly I:C stimulation ( Supplementary  Fig. 4d). We further confirmed that silencing of GSK3b but not GSK3a in BMDMs inhibited IFN-b induction significantly ( Supplementary Fig. 4e). It is interesting to note that GSK3 inhibition with SB216763 did not alter poly I:C-induced IFN-b mRNA expression ( Supplementary Fig. 4f), suggesting that the effect of GSK3b on TLR3-mediated IFN-b induction is independent of its kinase activity. Together, these data indicate that GSK3b is required for TLR3-mediated IRF3 activation and the type I IFN-b induction.
GSK3b regulates TLR3-mediated ERK and p38 activation. The mitogen-activated protein kinases (MAPKs) and the NF-kB signalling pathways are important for inflammatory cytokine production in TLR signalling [32][33][34] . To examine whether GSK3b regulates MAPKs and NF-kB activation in TLR3 signalling, we analysed the phosphorylation levels of ERK, p38, JNK and NF-kB p65 in GSK3b knockdown RAW264.7 cells. As shown in Fig. 2a, GSK3b knockdown markedly decreased poly I:C-induced ERK and p38 phosphorylation levels, whereas there were no significant differences in the phosphorylation levels of JNK and NF-kB p65. Similar results were observed in BMDMs silenced by short interfering RNA (siRNAs) specific for GSK3b. Although knockdown of both forms of GSK3 appears to delay poly I:Cinduced IkB-a degradation, silencing of GSK3b but not GSK3a significantly reduced poly I:C-induced ERK and p38 activation ( Supplementary Fig. 5). These results were further confirmed in Gsk3b À / À MEFs. Phosphorylation levels of ERK and p38 but not JNK, IKKa/b, IkB-a and NF-kB p65 were significantly decreased in Gsk3b À / À MEFs compared with the levels in Gsk3b þ / þ MEFs ( Fig. 2b; Supplementary Fig. 6a). In reporter assays, knockdown or overexpression of GSK3b did not affect poly I:Cinduced NF-kB activation, further suggesting that GSK3b is not involved in TLR3-mediated NF-kB signalling ( Supplementary  Fig. 6b,c). Because a TRIF-dependent pathway is also involved in TLR4 signalling 35,36 , we examined the effects of GSK3b on LPSinduced MAPKs and NF-kB p65 phosphorylation levels. Interestingly, the GSK3b knockdown decreased LPS-induced phosphorylation levels of p38 and JNK but not ERK and NF-kB p65 (Fig. 2c). In contrast, there were no significant differences between GSK3b knockdown RAW264.7 and control cells after a TLR2 ligand, Pam3CSK4, stimulation (Fig. 2d). It should be noted that Pam3CSK4-induced MAPKs and NF-kB p65 phosphorylation is mediated by a Myd88-dependent, but not a TRIF-dependent, pathway. Thus, these data indicate that GSK3b regulates ERK and p38 MAPK activation in TRIF-dependent TLR3 signalling.
GSK3b regulates expression of c-Fos through ERK and p38. AP1, comprising Jun, c-Fos and ATF2, is activated in response to TLR stimulation and regulates the production of pro-inflammatory cytokines 14,37,38 . To identify transcription factors regulated by the TLR3-GSK3b axis, we separated the cytosolic and nuclear fractions of GSK3b knockdown RAW264.7 cells after poly I:C stimulation. Interestingly, the c-Fos protein levels in the nuclear fraction appeared drastically reduced in GSK3b knockdown RAW264.7 cells compared with the levels in control cells, whereas nuclear ATF2, c-Jun and NF-kB p65 protein levels were comparable in both cell types ( Fig. 3a; Supplementary Fig. 7), suggesting that GSK3b regulates TLR3-mediated c-Fos expression but not ATF2, c-Jun and NF-kB p65. Consistently, treatment with the GSK3 inhibitor SB216763 or silencing of GSK3b but not GSK3a significantly reduced c-Fos mRNA or protein after TLR3 stimulation (Fig. 3b,c), whereas overexpression of GSK3b but not GSK3a produced a substantial increase in c-Fos mRNAs ( Supplementary Fig. 8). Reconstitution of GSK3b but not GSK3a or the kinase inactive GSK3b (K85A) mutant into Gsk3b À / À MEFs by transient transfection restored poly I:C-induced c-Fos mRNA expression (Fig. 3d). Moreover, the knockdown of c-Fos markedly reduced poly I:C-induced IL-6 and TNF-a expression ( Supplementary Fig. 9), confirming the importance of GSK3b-mediated c-Fos expression in TLR3 signalling.
Considering the defects in ERK and p38 MAPK signalling in GSK3b knockdown RAW264.7 cells and Gsk3b À / À MEFs (Fig. 2a,b), it is likely that poly I:C-mediated ERK and p38 MAPK activation regulates c-Fos expression. To test this hypothesis, we assessed c-Fos expression in the presence of the ERK inhibitor PD98059, the p38 inhibitor SB203580, the JNK inhibitor SP600125 or the NF-kB inhibitor BAY 11-7085. Expectedly, ERK or p38 inhibition markedly reduced poly I:C-induced c-Fos protein levels compared with the levels of the dimethylsulphoxide control (Fig. 3e,f). In contrast, nuclear c-Fos Poly I:C (min) expression was not affected by the inhibition of JNK or NF-kB ( Fig. 3g,h). Together, these data indicate that GSK3b regulates poly I:C-induced c-Fos expression via ERK and p38 activation.
GSK3b is required for TLR3 signalling complex formation. Upon poly I:C stimulation, activated TLR3 initiates the interaction of TRIF with TRAF6, transforming growth factor b-activated kinase 1 (TAK1) and RIP1 to activate MAPKs and NF-kB signalling cascades 11,12,39 . Specifically, the TRIF-TRAF6-TAK1 axis is important for the activation of MAPKs, which in turn leads to AP1 activation 34,39,40 . To examine how GSK3b regulates MAPKs in TLR3 signalling, we first determined the roles of GSK3b in forming the TLR3 signalling complex following poly I:C stimulation in Gsk3b À / À MEFs. In control cells, poly I:C stimulation induced recruitment of TRIF, RIP1, TRAF6, TAK1, TAK1-binding protein 1 (TAB1), TAB2 and GSK3b to TLR3, which persisted for at least 30 min (Fig. 4). Interestingly, GSK3b deficiency prevented the recruitment of TRAF6, TAK1, TAB1 and TAB2 to TLR3, but had no effect on the recruitment of TRIF   Fig. 10a). Notably, pretreatment with the GSK3 inhibitor SB216763 in BMDMs also blocked the formation of the poly I:C-induced TLR3 signalling complex ( Supplementary Fig. 10b). Thus, these results suggest that GSK3b is important for recruiting the TRAF6-TAK1-TAB1-TAB2 complex to TLR3.
GSK3b interacts with TRAF6 and TAK1. To further explore the regulatory mechanisms of GSK3b in TLR3 signalling, we investigated the interaction of GSK3b with TRAF6 and TAK1. TAK1, a MAP3K family member 41 , is a critical transducer molecule downstream of TRAF6, and the TRAF6-TAK1 axis activates MAPKs in TLR signalling 2,42,43 . We first examined by coimmunoprecipitation experiments whether GSK3b associates with TRAF6 or TAK1 upon poly I:C stimulation in BMDMS.
The results indicated that GSK3b indeed associates with TRAF6 and/or TAK1 under physiological conditions (Fig. 5a). We next examined whether TRAF6, GSK3b and TAK1 bind to one another, forming a ternary complex. To examine this binding, HEK293T cells were transfected with a Flag-TRAF6 construct with or without a Myc-TAK1 or HA-GSK3b construct. We observed that Flag-TRAF6 co-immunoprecipitated with Myc-TAK1 and HA-GSK3b, indicating that TRAF6 forms a ternary complex with GSK3b and TAK1 (Fig. 5b). Similarly, we further confirmed the association of GSK3b with TRAF6 and TAK1 as well as TAK1 with TRAF6 and GSK3b by cotransfection studies in mammalian cells ( Supplementary Fig. 11a,b). Notably, deletion of GSK3b prevented association of TRAF6 with TAK1 upon poly I:C stimulation (Fig. 5c). We also investigated the interaction of GSK3b with RIP1, an essential mediator of TLR3-induced NF-kB activation 12 . Unlike TRAF6 and TAK1, however, RIP1 did not interact with GSK3b ( Supplementary Fig. 12a,b), further confirming that GSK3b is not involved in NF-kB signalling mediated by the TLR3-TRIF-RIP1 axis. TAK1 forms a complex with its adaptor proteins TAB1 and TAB2. TAB2 bridges TRAF6 to TAK1, allowing TAK1 activation [44][45][46] . We therefore investigated whether GSK3b could associate with TAB1 or TAB2 and affect the formation of the signalling complex containing TRAF6, TAK1, TAB1 and TAB2. In contrast to the GSK3b-TAK1 interaction, there was little association between GSK3b and TAB1 or TAB2 (Fig. 5d, lanes  3 and 4). However, GSK3b could associate with TAB1 or TAB2 in the presence of TAK1, indicating that the association of GSK3b with TAB1 or TAB2 requires TAK1. Notably, forced expression of GSK3b promotes the association of TRAF6 with the TAK1-TAB1-TAB2 complex (Fig. 5e, lane 3). Furthermore, overexpressed GSK3b binds to TRIF, TRAF6 and TAK1 as a complex (Fig. 5f). Altogether, these results demonstrate that GSK3b could form a TRIF-assembled signalling complex containing TRAF6-TAB1-TAB2-TAK1.
TRAF6 is required for ubiquitination of GSK3b. Because TRAF6 possesses E3 ubiquitin ligase activity 47-49 , we next questioned whether GSK3b is ubiquitinated by TRAF6 in TLR3 signalling. We first identified the binding regions between TRAF6 and GSK3b. GSK3b interacted with the TRAF6 (289-530) derivative containing just the coiled-coil TRAF-N domain and the conserved TRAF-C domain, whereas the N-terminal ring and zinc-finger domains of TRAF6 spanning amino acids (aa) 1-289 failed to interact with GSK3b ( Supplementary Fig. 13a). On the other hand, serial deletion constructs of GSK3b revealed that the N-terminal region spanning aa 1-120 is necessary for TRAF6 interaction ( Supplementary Fig. 13b). We next tested whether GSK3b is ubiquitinated upon poly I:C stimulation in BMDMs and found that poly I:C triggered polyubiquitination of GSK3b (Fig. 6a). Furthermore, overexpression of TRAF6 induced GSK3b ubiquitination (Fig. 6b). In contrast, the catalytically inactive TRAF6 (C70A) mutant, which has lost its E3 ligase activity, lost the ability to promote GSK3b ubiquitination (Fig. 6c). Ubiquitination of GSK3b occurred mainly through K63 linkage ( Supplementary Fig. 14). Unlike TRAF6, NEDD4-1 and TRAF3, two E3 ubiquitin ligases that catalyse K63 ubiquitination and function in CD40 and TLR2 signalling, respectively 50,51 , did not promoted GSK3b ubiquitination (Supplementary Fig. 15). TRAF6 deficiency markedly decreased ubiquitination of endogenous GSK3b compared with the ubiquitination of wild-type cells upon poly I:C stimulation (Fig. 6d). Moreover, TRAF6 induced GSK3b ubiquitination in vitro (Fig. 6e). These results suggest that TRAF6 is an E3 ubiquitin ligase for GSK3b.
To search for the sites of GSK3b that are responsible for ubiquitination, we generated a series of full-length GSK3b variants containing a lysine residue mutation based on mass spectrometry (MS) analysis (Supplementary Table 1). We mutated K85, K86, K91 or K183 to arginine and then tested the susceptibility of these mutants, when expressed ectopically, to be ubiquitinated by TRAF6. TRAF6-mediated ubiquitination of GSK3b was substantially reduced when with K183R, but not with K85R, K86R or K91R (Fig. 6f). Importantly, overexpression of the GSK3b K183R mutant significantly reduced mRNA expression of IL-6 and TNF-a as well as c-Fos in a dose-dependent manner compared with that of the overexpressed wild-type GSK3b (Fig. 6g). In similar experiments, reconstitution of GSK3b but not GSK3b (K183R) mutant into Gsk3b À / À MEFs restored the poly I:C-induced IL-6, TNF-a, IL-10 and c-Fos mRNA expression ( Supplementary Fig. 16a,b) and GSK3b ubiquitination ( Supplementary Fig. 16c).
Because GSK3b (1-120) lacking the K183 residue could bind to TRAF6 ( Supplementary Fig. 13b), we tested whether GSK3b (1-120) can be ubiquitinated by TRAF6. As expected, GSK3b (1-120) failed to be ubiquitinated by TRAF6 ( Supplementary  Fig. 17). Interestingly, GSK3b (1-120) could act as a dominantnegative mutant to inhibit poly I:C-induced mRNA expression of IL-6, TNF-a and c-Fos (Supplementary Fig. 18). GSK3b ubiquitination is required for interaction with TRAF6. In our earlier experiments, we showed that GSK3b, TRAF6 and TAK1 bind to one another to form a ternary complex ( Fig. 5b; Supplementary Fig. 11a,b) and that GSK3b associates with TRIF, TRAF6 and TAK1 as a complex (Fig. 5f). Therefore, we tested whether GSK3b ubiquitination influenced its interaction with TRIF, TRAF6 and TAK1. Overexpression of the GSK3b K183R mutant, which is defective in GSK3b ubiquitination by TRAF6, showed impaired association with TRAF6 compared with the association of wild-type GSK3b (Fig. 7a). However, the K183R mutation of GSK3b did not abrogate interactions with TRIF or TAK1 (Fig. 7b,c), suggesting that GSK3b ubiquitination by TRAF6 is essential for its interaction with TRAF6. Because either deficiency or a pharmacological inhibition of GSK3b impaired the poly I:C-triggered formation of the TLR3 signalling complex (Fig. 4), we tested whether the GSK3b K183R mutant produced similar effects. Forced expression of the GSK3b K183R mutant markedly decreased the association of GSK3b with TRIF, TRAF6 and TAK1 ( Fig. 7d; Supplementary Fig. 19), as well as the formation of the TRIF-assembled signalling complex containing TRAF6, GSK3b and TAK1 (Fig. 7e). Together, our results show that TRAF6-mediated GSK3b ubiquitination is essential for TRAF6 interaction, thereby contributing to the formation of the TRIF-GSK3b-TRAF6-TAK1 complex.

Discussion
In general, TLRs that recognize bacteria induce pro-inflammatory cytokines, whereas those TLRs that detect viruses trigger the IFN response 2-4 . These two responses depend on the engagement of the major two adaptor molecules, MyD88 (refs 28,29) and TRIF 8,9 . All TLRs, with the exception of TLR3, signal through the MyD88-dependent pathway 1,2 , whereas TRIF-mediated signalling includes TLR3 (refs 6,7) and TLR4 (refs 35,36).GSK3b is a crucial regulator in the balance between pro-and anti-inflammatory cytokines in MyD88-dependent TLR signalling 24 , as well as viraltriggered RLR-mediated activation 31 . However, how GSK3b controls TLR3-mediated pro-inflammatory signalling mediated by TRIF, but not by MyD88, is still unknown. We therefore examined whether GSK3b is involved in pro-inflammatory cytokine production in TLR3 signalling and explored the molecuar basis of GSK3b's role. We found that GSK3b selectively regulated the TLR3-mediated activation of ERK and p38 but not JNK or NF-kB. Notably, the ERK and p38 pathways were required for the induction of c-Fos, which forms the AP1 complex 2,14,38 . We also found that GSK3b was incorporated into a TLR3-assembled multiprotein complex, and its signalling funtion was regulated by TRAF6-mediated ubiquitination. In the TLR3-assembled signalling complex, GSK3b undergoes polyubiquitination, which is dependent on the E3 ligase activity of TRAF6, and thereby promotes the formation of the TRIF-GSK3b-TRAF6-TAK1 complex. Notably, a ubiquitinationdefective GSK3b mutant acts as a dominant-negative form of GSK3b regarding the induction of pro-inflammtory cytokines, as well as c-Fos. The present study establishes an important role for GSK3b in poly I:C-triggered inflammatory cytokine production and provides a mechanistic explanation for how the TRAF6-GSK3b axis selectively regulates TLR3 signalling. In addition, we found that GSK3b in TLR3 signalling had a positive role in regulating the production of TBK1-mediated type I IFN-b. Similar to our results, it was previously shown that GSK3b functions in RLR-mediated IFN-b production by promoting TBK1 and IRF3 activation 31 , suggesting that the IFN-b production triggered by TLR3 and RLRs share a common pathway that converge upon TBK1, which is regulated by GSK3b through its kinase activity-independent mechanism. Among TRAF family members, TRAF3 positively regulates IRF3 and IFN-b response through TRIF interaction 10,52 . Therefore, it is likely that the TRAF6-GSK3b axis controls MAPK signalling and c-Fos expression by TLR3, while the TRAF3-GSK3b-TBK1 axis regulates IRF3 activation and IFN-b induction. Indeed, we found that GSK3b interacts with TRAF3 ( Supplementary Fig. 20a). Furthermore, GSK3b can form a complex containing TRIF, TRAF3 and TBK1 (Supplementary Fig. 20b).
Interestingly, GSK3b differentially regulates TLR-induced signalling. Our results have demonstrated that GSK3b regulates TLR3-and TLR4-mediated MAPK activation but is not required for TLR2 signalling [53][54][55] . Because TLR-mediated responses are controlled mainly by the MyD88-dependent pathway 2,28,29 , which is used by all TLRs except TLR3, and the TRIFdependent pathway 8,9,30 , which is used by TLR3 and TLR4 (refs 35,36), we propose that GSK3b selectively regulates TRIFdependent MAPK activation. Our results also showed that GSK3b differentially regulates TLR3-and TLR4-mediated MAPKs signalling. GSK3b was required for ERK and p38 pathway activation downstream of TLR3, and for the JNK and p38 pathway activation downstream of TLR4. It is probable that endosomal TRIF signalling complexes downstream of TLR3 and TLR4 are not identical, and differences in their signalling potentials correlate with their ability to selelctively engage GSK3b and thereby dictate downstream MAPK activation. In TRIF-dependent TLR3 signalling, TRIF directly recruits TRAF6 and RIP1, which work cooperatively to activate TAK1, eventually leading to activation of NF-kB and AP1 (refs 11,39,40). In the case of TLR4 stimulation, the initial step of TRIF signalling is mediated through an adaptor TRAM 56,57 . Internalized TLR4 recruits TRAF6 to the endosome via TRAM-TRIF 58,59 . Consequently, TRAF6 and RIP1 mediate the TRIF-induced activation of MAPKs and NF-kB, respectively 39,40 . In this regard, the TRIF-assembled signalling complex of TLR3 or TLR4 formed in different ways, and this may have accounted for the differential regulation of GSK3b in MAPK activation. Although the exact mechanisms of the GSK3b-mediated differential activation of MAPKs in the TLR3 and TLR4 signalling pathways require further investigations, our study suggests that GSK3b selectively regulates TRIF-dependent MAPK activation pathways.
Phosphorylation by GSK3b results in activation or inhibition of many its substrates 18,20,21 . Our results have demonstrated that the kinase activity of GSK3b is required for its ability to induce inflammatory cytokine production. These findings suggest that kinase activity of GSK3b in TLR3-TRIF signalling complex is ARTICLE involved in the cytokine production. Indeed, we found that Ser/Thr phosphorylation of TRAF6 was enhanced in control cells after poly I:C stimulation compared with that in the knockdown of GSK3b ( Supplementary Fig. 21), suggesting that TRAF6 might be a candidate among GSK3b substrates. Alternatively, it has been recently reported that Bruton's tyrosine kinase is required for the production of inflammatory cytokines in TLR3-stimulated macrophages 60 , suggesting that Bruton's tyrosine kinase acts in TLR3/TRIF signalling. Further studies will be required for characterizing GSK3b substrate(s) in the TLR3-TRIF complex. TRAF6, as an E3 ubiquitin ligase, is known to be a common signalling adaptor for cytokine production in response to various TLR ligands 61,62 . TRAF6 can ubiquinate itself on lysine 63, and TRAF6 autoubiquitination in turn recruits mediators for the activation of downstream MAPKs and NF-kB signalling pathways 2,48,63 . TRAF6 mediates both MyD88-dependent and TRIF-dependent activation of NF-kB and AP1. In MyD88dependent and TRIF-dependent TLR signalling, ubiquitinated TRAF6 serves as a signalling scaffold to recruit TAK1 via TAB2 and TAB3 (refs 11,45,46,64). The TAK1 signalling complex, including TRAF6-TAB2-TAB3-TAB1-TAK1 is subsequently released into the cytosol, where TAK1 activates MAPK cascades 43,46,64 . Although TRAF6 appears to be a common factor employed by MyD88-and TRIF-dependnent signalling, it is probable that a specific signalling partner, substrate or other signalling protein(s) in each signalling complex is needed for signalling specificity or fine tuning of signalling. In this regard, we have now demonstrated that GSK3b underwent K63 chain ubiquitination. TRAF6 was found to be a direct E3 ligase for GSK3b and was essential for GSK3b ubiquitination, TRIFassembled signalling complex formation and pro-inflammtory cytokine production upon TLR3 stimulation. This mechanism involves the TLR3-induced assembly of a multiprotein complex containing TRIF, TRAF6, TAB1, TAB2, TAK1 and GSK3b. Complex assembly resulted in TRAF6 autoubiquitination and activation, which led to K63-linked ubiquitination of GSK3b. Ubiquitinated GSK3b promoted a multiprotein-assembled signalling complex, where TAK1 and its subordinate MAPKs are activated. Notably, during IL-1 and receptor activator of nuclear factor-kB ligand signalling, TRAF6 autoubiquitination was dispensable for both interaction with and activation of the TAK1 signalling complex 65 . It has been suggested that TRAF6mediated K63-linked ubiquitination instead targets relevant protein substrates during activation 63,64 . Accordingly, we propose that TRAF6 mediates the K63-linked ubiquitination of GSK3b, which would form a signalling complex sufficient to meet activation thresholds and/or to generate signalling specificity. It should be noted that TRAF6-mediated GSK3b ubiquitination proceeds through a two-stage mechanism. This mechanism involves an initial interaction prior to ubiquitination between TRAF6 and GSK3b. The interaction of TRAF6 (through a coiledcoil TRAF-N domain and a conserved TRAF-C domain) with GSK3b (through a N-terminal region spanning aa 1-120) may lead to K63-linked ubiquitination of GSK3b by E3 ligase activity of TRAF6. We thus propose that GSKb is a novel TRAF6 substrate downstream of TLR3. The establishment of a regulatory role for GSK3b in TLR3-mediated signalling contributes to the elucidation of the complicated molecular mechanisms of inflammatory and antiviral responses.
Immunoprecipitation and western blot analysis. Cells were washed with cold PBS (Hyclone) and lysed with lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% deoxycholate acid, 1% NP-40) containing phosphatase and protease inhibitors. For immunoprecipitation, lysates were incubated with the indicated primary antibodies at 4°C for 16 h, and were further incubated with protein A-agarose (Millipore, Billerica, MA, USA) at 4°C for 1 h with gentle shaking. After washing five times with lysis buffer, immunoprecipitated proteins were boiled with 2 Â SDS loading buffer, and separated on SDS-polyacrylamide gel electrophoresis (PAGE) and electrophoretically transferred to polyvinylidene difluoride membranes (Millipore). Membranes were blocked with 5% BSA in Tris-buffered saline containing 0.1% Tween-20 and were immunoreacted with the indicated primary antibodies and secondary antibodies conjugated to HRP. Images have been cropped for presentation. Full-size images of all western blots are provided in Supplementary Fig. 22.
Enzyme-linked immunosorbent assay. To measure mouse IL-6, TNF-a and IL-10 levels, BMDMs were preincubated with the indicated inhibitors and then stimulated with or without 10 mg ml À 1 poly I:C for 20 h. Cell culture supernatants were assessed using ELISA kits from R&D Systems according to the manufacturer's instructions.
Real-time PCR analysis. Total RNA was extracted from cells using the TRIzol reagent (Invitrogen) and reverse transcribed to complementary DNA using the Superscript cDNA synthesis kit (Invitrogen) following the manufacturer's instructions. Real-time PCR analysis was performed using the KAPA SYBR green FAST qPCR kit (Kapa Biosystems, Boston, MA, USA) on an ABI 7300 real-time PCR machine (Applied Biosystems, Foster City, CA). Samples were analysed in triplicate and normalized to b-actin mRNA expression. Primer sequences are listed in Supplementary Table 2.
In vitro ubiquitination assay. The GSK3b protein was obtained from Invitrogen (cat # PV3365) and the Flag-TRAF6 protein was purified as previously described 68 . Briefly, HEK293T cells transfected with Flag-TRAF6 for 48 h were lysed in a lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 1.5 mM MgCl 2 ) containing protease inhibitors and immunoprecipitated with an anti-Flag affinity gel (Sigma) for 16 h at 4°C. Immunoprecipitates were washed and eluted with 300 mg ml À 1 Flag peptide according to the manufacturer's instructions. The in vitro ubiquitination assay was performed as previously described with minor modifications 67 . Briefly, 5 nM Flag-TRAF6 and 1 mM GSK3b protein were mixed with 100 nM His-E1, 1 mM His-E2 (Ubc13/Mms2) and 2.5 mM Bt-Ub in 50 mM ubiquitination reaction buffer from the ubiquitination kit (Enzo Life Sciences, Farmingdale, NY, USA) according to the manufacturer's instructions. Samples were subsequently immunoprecipitated with an anti-GSK3b antibody and separated on SDS-PAGE followed by streptavidin conjugated to HRP (Thermo Fisher Scientific).
Luciferase assay. RAW264.7 cells were transiently transfected with pGL3-Basic or pGL3-NF-kB-luc along with pRL-TK-renilla luciferase plasmids. HEK293-null or HEK293-TLR3 cells were transiently transfected with pGL3-NF-kB-luc and pRL-TK-renilla luciferase along with pEGFP-N3 or pEGFP-N3-GSK3b plasmids. After 24 h transfection, cells were stimulated with 10 mg ml À 1 poly I:C for 4 h and luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA) according to the manufacturer's instructions. Samples were analysed in triplicate and normalized to renilla luciferase activity.
Mapping of ubiquitination sites on GSK3b. For identification of ubiquitination sites on GSK3b, HEK293T cells were transfected with HA-tagged GSK3b expression plasmids for 36 h. Cells were lysed with lysis buffer (20 mM Tris-HCl, pH 7.4, 1% SDS), boiled for 5 min, sonicated and then diluted 10-fold with NP-40 lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% NP-40). After centrifugation at 15,000g for 15 min, the lysates were incubated with 120 ml of anti-HA Agarose (Thermo scientific) at 4°C overnight with rotation. The beads were washed four times with NP-40 lysis buffer and eluted with 2 Â SDS-PAGE sample buffer without a reducing agent. HA-GSK3b proteins eluted from the beads were subjected to SDS-PAGE followed by Coomassie Blue staining or western blot analysis. For MS analysis, gel bands from the Coomassie Blue stained-gel were excised and subjected to trypsin digestion and liquid chromatography-MS/MS. MS/MS data were analysed using SEQUEST (Thermo Finnigan, San Jose, CA) software to identify ubiquitin modification with the GG or LRGG remnant tag on lysine residues of GSK3b.
Statistical analysis. Data are presented as the mean±s.d. from at least three independent experiments. Statistical significance was determined using Student's t-test.