The protein kinase TAK1 (transforming growth factor-β-activated kinase 1), which has been implicated in the activation of MAPK (mitogen-activated protein kinase) cascades and the production of inflammatory mediators by LPS (lipopolysaccharide), IL-1 (interleukin 1) and TNF (tumour necrosis factor), comprises the catalytic subunit complexed to the regulatory subunits, termed TAB (TAK1-binding subunit) 1 and either TAB2 or TAB3. We have previously identified a feedback-control mechanism by which p38α MAPK down-regulates TAK1 and showed that p38α MAPK phosphorylates TAB1 at Ser423 and Thr431. In the present study, we identified two IL-1-stimulated phosphorylation sites on TAB2 (Ser372 and Ser524) and three on TAB3 (Ser60, Thr404 and Ser506) in human IL-1R cells [HEK-293 (human embryonic kidney) cells that stably express the IL-1 receptor] and MEFs (mouse embryonic fibroblasts). Ser372 and Ser524 of TAB2 are not phosphorylated by pathways dependent on p38α/β MAPKs, ERK1/2 (extracellular-signal-regulated kinase 1/2) and JNK1/2 (c-Jun N-terminal kinase 1/2). In contrast, Ser60 and Thr404 of TAB3 appear to be phosphorylated directly by p38α MAPK, whereas Ser506 is phosphorylated by MAPKAP-K2/MAPKAP-K3 (MAPK-activated protein kinase 2 and 3), which are protein kinases activated by p38α MAPK. Studies using TAB1−/− MEFs indicate important roles for TAB1 in recruiting p38α MAPK to the TAK1 complex for the phosphorylation of TAB3 at Ser60 and Thr404 and in inhibiting the dephosphorylation of TAB3 at Ser506. TAB1 is also required to induce TAK1 catalytic activity, since neither IL-1 nor TNFα was able to stimulate detectable TAK1 activity in TAB1−/− MEFs. Surprisingly, the IL-1 and TNFα-stimulated activation of MAPK cascades and IκB (inhibitor of nuclear factor κB) kinases were similar in TAB1−/−, MEKK3−/− [MAPK/ERK (extracellular-signal-regulated kinase) kinase kinase 3] and wild-type MEFs, suggesting that another MAP3K (MAPK kinase kinase) may mediate the IL-1/TNFα-induced activation of these signalling pathways in TAB1−/− and MEKK3−/− MEFs.
The uncontrolled production of TNFα (tumour necrosis factor α) and other pro-inflammatory cytokines is a major cause of chronic inflammatory diseases, such as rheumatoid arthritis and inflammatory bowel disease. TAK1 (transforming growth factor-β-activated kinase 1) is a MAP3K [MAPK (mitogen-activated protein kinase) kinase kinase], which becomes activated when cells are stimulated with bacterial LPS (lipopolysaccharide) or the pro-inflammatory cytokines TNFα or IL-1 (interleukin-1) [1–3]. Along with other MAP3Ks, such as MEKK3 [MAPK/ERK (extracellular-signal-regulated kinase) kinase kinase 3] , TAK1 is thought to play a key role in the production of TNFα and other inflammatory mediators by activating several MAPKs, termed p38α MAPK, JNK (c-Jun N-terminal kinase) 1/JNK2 and ERK1/2, and the transcription factor NF-κB (nuclear factor κB) [3,5–7], via the signalling pathways shown in Figure 1.
Pro-inflammatory protein kinase cascades whose activation by IL-1 requires the expression of the TAK1 catalytic subunit
The native forms of TAK1 comprise the catalytic subunit complexed to a regulatory subunit TAB (TAK1-binding subunit) 1, which is an inactive pseudophosphatase structurally related to members of the MPP (Mg2+-dependent protein phosphatase) family of protein serine/threonine phosphatases , and either of two structurally related proteins, termed TAB2 and TAB3 [9–11]. The activation of TAK1 by LPS or IL-1 is thought to be triggered by the formation of Lys63-linked polyubiquitinated TRAF6 (TNF receptor-associated factor 6), which binds to the C-terminal NZF (nuclear zinc finger) motifs of TAB2 and TAB3, leading to the autophosphorylation and activation of TAK1 . We have previously identified a feedback-control loop in which p38α MAPK suppresses the activation of TAK1, and which correlates with the phosphorylation of TAB1 at Ser423 and Thr431 . This feedback loop is prevented by the p38α MAPK inhibitor SB 203580 , and is not observed in p38α MAPK−/− fibroblasts . For this reason, the inhibition of p38α MAPK causes the hyperactivation of TAK1, and hence the hyperactivation of JNK. This may contribute to side effects of p38α MAPK inhibitors, such as hepatotoxicity , which have prevented these drugs from progressing to later-stage clinical trials for the treatment of chronic inflammatory diseases.
Agonists that trigger the phosphorylation of TAB1 also induce the phosphorylation of the TAB2 and TAB3 regulatory subunits, as judged by a decrease in the electrophoretic mobility of these proteins that is reversed by incubation with a protein serine/threonine phosphatase [10,11]. This decreased mobility is partially (TAB2) or completely (TAB3) prevented by SB 203580, indicating that the p38α MAPK-induced feedback control of TAK1 may involve the phosphorylation of TAB2/TAB3 as well as TAB1 . In the present study, we demonstrated that TAB2 and TAB3 are phosphorylated at multiple sites in human IL-1R cells [HEK-293 (human embryonic kidney) cells that stably express the IL-1 receptor] and MEFs (mouse embryonic fibroblasts) in response to IL-1. We also identified three IL-1-stimulated sites on TAB3 whose phosphorylation is dependent on p38α MAPK activity and two phosphorylation sites on TAB2 that are phosphorylated by p38α MAPK-independent mechanisms. Studies with TAB1−/− MEFs suggest that TAB1 plays several roles in the regulation of the TAK1 complex, namely to recruit p38α MAPK to the TAK1 complex for the phosphorylation of TAB3, to suppress the dephosphorylation of TAB3, and to induce TAK1 catalytic activity. We find that IL-1 and TNFα do not activate TAK1 in either TAB1−/− MEFs or MEKK3−/− MEFs, despite the unimpaired activation of MAPK cascades in these cells.
MATERIALS AND METHODS
PD 184352 , synthesized by an improved method , and BIRB 0796 , synthesized from 4,4-dimethyl-3-exopentanenitrile , were provided by Dr Natalia Shpiro (College of Life Sciences, University of Dundee). SB 203580 was purchased from Promega, mouse IL-1α and IL-1β and human IL-β were from Sigma, LPS was from Invivogen and glutathione–Sepharose was from GE Healthcare.
TAB2 and TAB3 were cloned and inserted into pGEX6P-1 as described previously . Both constructs were digested with BamHI and NotI and ligated into the same sites in pEBG6P to form pEBG6P-TAB2 and pEBG6P-TAB3 respectively for transfection into IL-1R cells. Mutations in phosphorylation sites in TAB2 (S372A, S524A and S582A) and TAB3 (S60A, T404A and S506A) were created using the QuikChange® site-directed mutagenesis method (Stratagene), but using KOD Hot Start DNA polymerase (Novagen). DNA encoding TAK1 was amplified from IMAGE EST (expressed sequence tag) 3906837 with Expand HiFi (Roche) using standard methods and subcloned into the BamHI site in pEBG6P.
The phosphopeptides YIAApSPPNTD, RKLpSMGSDD and LKRSNpSISQIP, corresponding to residues 368–377, 521–529 and 576–587 of TAB2 respectively, YMEYHpSPDDNR and YMEYHpSPEDNR, corresponding to residues 55–65 of human and mouse TAB3 respectively, LYTATpTPPSSS and KYQRSSpSSGSDD, corresponding to residues 399–409 and 500–511 of TAB3 respectively, and IQTHMpTNNKGS, corresponding to residues 182–192 of TAK1 (where pS and pT correspond to phosphoserine and phosphothreonine respectively), were synthesized by Dr Graham Bloomberg (Department of Biochemistry, University of Bristol, Bristol, U.K.), coupled to both BSA and keyhole-limpet haemocyanin and injected into sheep at Diagnostics Scotland (Edinburgh, U.K.). The antibodies were affinity-purified from the antisera on CH-Sepharose to which the relevant phosphorylated peptide had been coupled covalently, and used for immunoblotting at 3 μg/ml in the presence of 30 μg/ml unphosphorylated peptide immunogen to neutralize any antibodies that recognize the unphosphorylated form of the protein. His6-tagged full-length MEKK3 expressed from a baculovirus vector in insect Sf21 cells was also injected into sheep, and the antisera were affinity-purified against purified MEKK3 immobilized on CH-Sepharose and used for immunoblotting at 1 μg/ml. The antibodies that recognize TAB3, TAK1, TAB1, TAB1 phosphorylated at Ser423 and TAB1 phosphorylated at Ser438 were raised in sheep and affinity-purified as described previously . Antibodies recognizing all forms of JNK1/2, ERK1/2, MAPKAP-K (MAPK-activated protein kinase) 2, MAPKAP-K2 phosphorylated at Thr334, the active phosphorylated forms of p38α MAPK and ERK1/2 and c-Jun phosphorylated at Ser63 were from Cell Signaling Technologies. An antibody recognizing the phosphorylated forms of JNK1 and JNK2 was purchased from Biosource, and anti-TAB2, anti-TAK1 and anti-IκBα (inhibitor of NF-κB α) antibodies were from Santa Cruz.
Generation of MAPKAP-K2−/−/MAPKAP-K3−/− mice
Mice were maintained under specific pathogen-free conditions and work was carried out in accordance with U.K. legislation. MAPKAP-K3 gene-targeted ES (embryonic stem) cells were generated using standard techniques in 129Sv ES cells. Exon 3, encoding subdomains III and IV, which form the major part of the ATP-binding site of MAPKAP-K3, was replaced with a neomycin cassette. These ES cells were used to produce MAPKAP-K3−/− mice using established protocols . Full details of the generation of these mice are available from J. S. C. A. MAPKAP-K2−/−/MAPKAP-K3−/− mice were obtained by inter-crossing with MAPKAP-K2−/− mice , and the mice were genotyped by PCR-based methods from ear biopsies.
Generation of Cre-conditional MEKK3−/− mice
Genomic DNA encoding exons 9–16 of the MEKK3 gene from 129Sv wild-type mouse DNA was cloned into the Bluescript targeting vector. The neomycin-resistance gene (PGKneo) flanked by FRT (Flp recombinase recognition target) sites was cloned in reverse orientation into the intron flanked by exons 11 and 12. LoxP sites were introduced 3′ to the neomycin-resistance gene in the intron between exons 11 and 12 and the intron flanked by exons 13 and 14. The linearized targeting vector was electroporated into 129Sv ES cells. Homologous recombination in three ES cell clones was verified by PCR and Southern blotting. Flp recombinase was used to excise the neomycin-resistance gene and was confirmed by PCR. ES cells were injected into C57/Bl6 blastocysts and founder males were crossed with 129Sv females. Heterozygotes were crossed, and germline transmission was verified by Southern blotting and PCR. MEFs were isolated from E14.5 (embryonic day 14.5) embryos, and immortalized MEFs were isolated by serial passage. Adenovirus encoding Cre recombinase was used to infect MEFs for deletion of exons 12 and 13 which encode kinase subdomains I–V within the MEKK3 coding sequence. Deletion of exons 12 and 13 was confirmed by PCR. MEKK3 mRNA having exons 12 and 13 deleted was detected following MEF treatment with Cre recombinase. A truncated MEKK3 protein was detected by immunoblotting (see the Results section) and was demonstrated to be kinase-inactive by an in vitro kinase assay of immunoprecipitated protein.
Cell culture, stimulation and lysis
IL-1R cells (provided by Tularic), immortalized MEFs deficient in JNK1 and JNK2 (JNK1−/−/JNK2−/−) , immortalized TAB1−/− MEFs  or immortalized MEKK3−/− MEFs were cultured in DMEM (Dulbecco's modified Eagle's medium) containing 10% (v/v) heat-inactivated FCS (foetal calf serum), 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine. Primary MEFs were isolated from E13.5 embryos of the MAPKAP-K2−/−/MAPKAP-K3−/− mice as reported previously for MSK−/− MEFs , and cultured for up to five passages as described above for immortalized MEFs. After 6 h or 16 h before stimulation with human IL-1β in IL-1R cells or murine IL-1α, IL-1β or sorbitol in MEFs, the medium was removed and replaced with DMEM from which FCS had been omitted. At 1 h before stimulation with agonists, an aliquot of concentrated (10–20 mM) solutions of SB 203580, PD 184352 or BIRB 0796 dissolved in DMSO was added to the culture medium to achieve the final concentrations indicated in the Results section. The equivalent volume of DMSO was added to control cells. Cells were lysed in ice-cold lysis buffer [50 mM Tris/HCl, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1 mM sodium orthovanadate, 10 mM sodium 2-glycerophosphate, 50 mM NaF, 5 mM sodium pyrophosphate, 0.27 M sucrose, 1% (v/v) Triton X-100, 0.1% 2-mercaptoethanol and one tablet of Complete™ proteinase inhibitor cocktail (Roche) per 50 ml of buffer]. Lysates were centrifuged at 13000 g for 10 min at 4 °C, and the supernatants were used immediately or snap-frozen in liquid nitrogen and stored in aliquots at −20 °C until use. Protein concentrations were determined using the method of Bradford.
Transfection of IL-1R cells
Cells were transfected at 40–50% confluence using polyethyleneimine. To express TAK1 complexes for purification and mass spectrometric analysis of TAB3 and TAB2, a single 15-cm-diameter dish of IL-1R cells was transfected with DNA encoding GST (glutathione transferase)–TAK1 (2 μg), GST–TAB1 (8 μg), and either GST–TAB3 or GST–TAB2 (20 μg). In order to test the specificities of the TAB3 or TAB2 phospho-specific antibodies, 10-cm-diameter dishes of IL-1R cells were transfected with DNA encoding GST–TAK1 (2.5 μg), TAB1 (2.5 μg) and wild-type or mutant GST–TAB2 or GST–TAB3 (10 μg).
Purification of complexes
To purify GST-tagged TAK1 complexes, a 100 μl volume of a 50% slurry of glutathione–Sepharose 4B was incubated for 2 h at 4 °C with 8 mg of cell extract protein. After brief centrifugation, the supernatant was removed, and the beads were washed with 4 ml of 50 mM Tris/HCl, pH 7.5, 250 mM NaCl, 0.03% Brij-35, 0.2 mM PMSF, 1 mM benzamidine and 0.1% 2-mercaptoethanol, followed by two further washes with 1 ml of 50 mM Tris/HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 1 mM DTT (dithiothreitol), 0.2 mM PMSF and 1 mM benzamidine. The beads were then resuspended in 100 μl of the same buffer, and the GST–TAK1 complexes were digested for 16 h at 4 °C with 4 μg of PreScission protease to release the TAK1 complexes from the beads. They were then denatured in SDS, subjected to SDS/PAGE and stained with colloidal Coomassie Blue.
Immunoprecipitation of TAK1 complexes
The endogenous TAB1–TAK1–TAB3 or TAB1–TAK1–TAB2 complexes were immunoprecipitated from 3 mg of IL-1R cell extract protein or 2 mg of MEF extract protein using an anti-TAB1 antibody . A 3 mg amount of cell extract protein was incubated for 2 h at 4 °C with 10 μg of anti-TAB1 antibody coupled with 10 μl of Protein G–Sepharose. The immunoprecipitates were washed twice with 1 ml of lysis buffer containing 0.25 M NaCl, followed by two washes with 1 ml of 50 mM Tris/HCl, pH 7.5, 50 mM NaCl and 0.1% (v/v) 2-mercaptoethanol.
Measurement of TAK1 activity
TAK1 activity in TAB1 immunoprecipitates was measured by the activation of MKK (MAPK kinase) 6 and coupled to the activation of p38α MAPK. The active p38α MAPK generated in this reaction was then quantified in a second assay by the phosphorylation of myelin basic protein. This coupled assay has been detailed and validated previously .
Expression of TAB2 and TAB3
In order to identify phosphorylation sites on TAB2 and TAB3, we first expressed these proteins in IL-1R cells, alone or in the presence of the other subunits of the TAK1 complex. These studies revealed that the expression of either TAB3 or TAB2 was greatly enhanced when they were co-expressed with TAK1 and TAB1 (results not shown). The co-expression of vectors encoding GST-fusions of TAB3, TAK1 and TAB1 or TAB2, TAK1 and TAB1 in a single 15-cm-diameter dish of cells generated sufficient amounts of TAB3 and TAB2 (Figure 2A) for the phosphorylated peptides in each protein to be identified by tryptic digestion and MS.
The expression of TAK1 complexes in IL-1R cells, MS of TAB2 and TAB3, and characterization of phospho-specific antibodies
Identification of phosphorylation sites on TAB2
Nine tryptic phosphopeptides were detected by precursor ion-scanning MS, all of which were identified (Figure 2B). Phosphopeptides 1a and 1b both comprised residues 369–382 of TAB2 phosphorylated at Ser372, these peptides only differing by the presence of methionine sulfoxide in peptide 1b compared with unoxidized methionine in peptide 1a. Peptides 1a and 1b arise from a chymotryptic-like cleavage of the Tyr–Ile bond between residues 368 and 369 of TAB2. Phosphopeptides 5a and 5b both comprised residues 363–382 of TAB2 phosphorylated at Ser372, these peptides also only differing by the presence of methionine sulfoxide in peptide 5b compared with unoxidized methionine in peptide 5a. Phosphopeptide 2 comprised residues 355–369 and contained one phosphate, but which of the six serine residues and three threonine residues was phosphorylated could not be identified. Phosphopeptides 3a and 3b both comprised residues 580–596 of TAB2, these peptides again differing by the presence of methionine sulfoxide in peptide 3b compared with unoxidized methionine in peptide 3a. Phosphopeptides 3a and 3b were not phosphorylated at Ser588, indicating that the site(s) of phosphorylation was residues 580, 582 or 584. Phosphopeptide 4b was an oxidized form of phosphopeptide 4a comprising residues 522–541 of TAB2. MS data showed that these peptides were not phosphorylated at Thr533, indicating that the site(s) of phosphorylation was Ser524 and/or Ser527.
Identification of phosphorylation sites on TAB3
Nine tryptic phosphopeptides were detected by precursor ion-scanning MS (Figure 2C), eight of which were identified. Phosphopeptides 1a and 1b both comprised residues 53–65 of TAB3 phosphorylated at Ser60, these peptides only differing by the presence of methionine sulfoxide in peptide 1b compared with unoxidized methionine in peptide 1a. Peptide 3, resulting from an incomplete tryptic cleavage of the Arg–Met bond between residues 65 and 66, comprised residues 53–68 with both methionine residues oxidized to the sulfoxide derivative, and was also phosphorylated at Ser60.
Peptides 2a and 2b both corresponded to residues 395–412 of TAB3, these peptides containing one and two phosphorylated residues respectively, which were not located at Ser407, Ser408, Ser409 or Ser411. This narrowed down the potential sites of phosphorylation to Ser398, Tyr400, Thr401, Thr403 or Thr404. Peptides 2a* and 2b* also corresponded to singly and doubly phosphorylated forms of the peptide comprising residues 395–412. However, their molecular masses were 203 Da greater than those of peptides 2a and 2b respectively, which equates to the mass predicted for an N-acetylhexosamine residue, probably an O-linked N-acetylglucosamine residue, attached covalently to a serine or threonine residue in the peptide.
Peptide 4 was found to comprise residues 504–523 phosphorylated at one residue, which was not Thr515. This indicated that the phosphorylation site(s) was located at Ser504, Ser505, Ser506, Ser507 or Ser509.
The phosphopeptide indicated by a double asterisk (Figure 2C) could not be identified by MS.
Production and specificity of phospho-specific antibodies
Phospho-specific antibodies were raised against the two phosphopeptides containing the sites of phosphorylation that were identified definitively by MS, namely Ser60 on TAB3 and Ser372 on TAB2. In addition, we also raised a phospho-specific antibody capable of recognizing TAB3 if it were phosphorylated at Thr404. This residue was selected as the likely site of phosphorylation in peptide 2a/2b (Figure 2C) because MAPKs are proline-directed protein kinases and Thr404 not only lies in a Thr-Pro sequence, but also is located in a position equivalent to Ser372 of TAB2. We also raised phospho-specific antibodies that should recognize TAB3 phosphorylated at Ser506 and TAB2 phosphorylated at either Ser524 or Ser582, which were among the potential sites of phosphorylation in phosphopeptide 4 from TAB3 (Figure 2C) and phosphopeptides 4a/4b and 3a/3b from TAB2 (Figure 2B) respectively. These serine residues were selected as likely sites of phosphorylation because they are located in Arg-Xaa-Xaa-Ser sequences, which are targeted by many protein kinases.
All six phospho-specific antibodies that were generated recognized GST–TAB2 or GST–TAB3 purified from transfected IL-1R cells, but did not recognize mutant forms of GST–TAB2 or GST–TAB3 in which the particular phosphorylation site had been changed to alanine (Figures 2D and 2E). These experiments established that TAB3 was phosphorylated at Ser60, Thr404 and Ser506 and TAB2 at Ser372, Ser524 and Ser582, at least when overexpressed in IL-1R cells, and that each antibody recognized only a single phosphorylated residue in each protein.
Effect of IL-1β on the site-specific phosphorylation of TAB2
The phosphorylation state of TAB2 in the endogenous TAK1 complex was analysed after its immunoprecipitation with the TAK1 complex using an anti-TAB1 antibody. These studies showed that TAB2 became phosphorylated at Ser372 and Ser524 in response to IL-1β. Phosphorylation was maximal after approx. 10 min and was sustained for at least 30 min (Figure 3A). The phosphorylation of TAB2 at Ser372 or Ser524 was not prevented by 5 μM SB 203580 (Figure 3B) or by 0.1 μM BIRB 0796 (results not shown), which are structurally unrelated and relatively [13,22] and highly [17,22] specific inhibitors of p38α MAPK, although the phosphorylation of TAB1 at Ser423, which is known to be phosphorylated by p38α MAPK in cells , was blocked by these compounds as expected (Figure 3B and results not shown). Moreover, the phosphorylation of TAB2 at Ser372 or Ser524 was not prevented by the exquisitely specific MKK1 inhibitor PD 184352 [22,23] in the absence or presence of SB 203580 (Figure 3B), excluding the involvement of ERK1/2, or kinases activated by ERK1/2, in the phosphorylation of these residues. PD 184352 prevented the IL-1-stimulated phosphorylation of ERK1/2, as expected (Figure 3B).
Effect of IL-1β on the site-specific phosphorylation of TAB2
If the concentration of BIRB 0796 in the culture medium is increased from 0.1 to 10 μM, it suppresses the activity not only of p38α MAPK and p38β MAPK, but also of the MAPKs termed JNK1 and JNK2 . Pre-incubation of the IL-1R cells with 10 μM BIRB 0796 suppressed the IL-1-induced phosphorylation of TAB2 at Ser372, as well as c-Jun at Ser63 (a physiological substrate of JNKs), but did not affect the phosphorylation of Ser524 (Figure 4A), indicating that Ser372 and Ser524 are phosphorylated by distinct protein kinases and that Ser372 might be phosphorylated by a JNK isoform(s). However, the IL-1-stimulated phosphorylation of Ser372 in immortalized JNK1−/−/JNK2−/− MEFs was similar to that observed in wild-type MEFs (Figure 4B). Moreover, the IL-1-induced phosphorylation of TAB2 at Ser372 was suppressed by 10 μM BIRB 0796 in the JNK1−/−/JNK2−/− MEFs (Figure 4C). Since fibroblasts do not express the JNK3 isoform, these results demonstrate that 10 μM BIRB 0796 prevents the phosphorylation of Ser372 in JNK−/− MEFs (and presumably also in wild-type MEFs) by inhibiting a protein kinase distinct from JNK.
Effect of IL-1β on the phosphorylation of TAB2 in wild-type and JNK−/− cells
The phospho-specific antibody that recognizes TAB2 phosphorylated at Ser582 in transfected cells (Figure 2D) did not recognize the endogenous TAK1 complex in either unstimulated or IL-1-stimulated IL-1R cells.
Effect of IL-1β on the site-specific phosphorylation of TAB3
The phosphorylation of TAB3 in the endogenous TAK1 complex was studied after immunoprecipitating the TAK1 complex from extracts of IL-1R cells using an anti-TAB1 antibody. These studies showed that TAB3 was minimally phosphorylated at Ser60, Thr404 and Ser506 in unstimulated cells, but these residues became phosphorylated in response to IL-1β, with maximal phosphorylation after 20 min (Figure 5A). Moreover, the IL-1-induced phosphorylation of all three sites was suppressed by either 5 μM SB 203580 or 0.1 μM BIRB 0796. At the same concentrations, these compounds inhibited the phosphorylation of TAB1 at Ser423, a known physiological substrate for p38α MAPK (Figure 5B).
Effect of interleukin 1β on the site-specific phosphorylation of TAB3
The experiments presented in Figure 5(B) demonstrated that Ser60, Thr404 and Ser506 are phosphorylated by p38α MAPK either directly or indirectly by another protein kinase that is activated by p38α MAPK. Since p38α MAPK is a proline-directed protein kinase and Ser60 and Thr404 are followed by a proline residue, it seems likely that these sites are phosphorylated directly by p38α MAPK. However Ser506 is not followed by a proline residue, but lies in a Hyd-Xaa-Arg-Xaa-Xaa-Ser (where Hyd is a hydrophobic residue) motif reminiscent of the consensus sequence for phosphorylation by MAPKAP-K2 and MAPKAP-K3 , two closely related protein kinases that are activated by p38α MAPK in vivo. We therefore studied the effect of IL-1 using MAPKAP-K2−/−/MAPKAP-K3−/− MEFs. These experiments revealed that the IL-1-stimulated phosphorylation of Ser506 did not occur in these cells (Figure 5C), indicating that MAPKAP-K2 and/or MAPKAP-K3 are likely to mediate the phosphorylation of TAB3 at this site in wild-type cells. The expression of p38α MAPK is reduced in MAPKAP-K2−/−/MAPKAP-K3−/− cells , which accounts for the reduced IL-1-stimulated phosphorylation of p38α MAPK observed in our knockout cells. Consistent with this, the IL-1-stimulated phosphorylation of TAB3 at Ser60 and Thr404, which is probably catalysed by p38α MAPK, was reduced, but not abolished, in the MAPKAP-K2−/−/MAPKAP-K3−/− cells (Figure 5C).
IL-1 stimulated phosphorylation of TAB3 in TAB1−/− MEFs
It has been reported that the IL-1-stimulated activation of MAPK cascades and NF-κB in MEFs from TAB1−/− cells is similar to wild-type cells , which was confirmed in the present study (Figure 6A). It had been inferred from these observations that TAB1 is not required for the IL-1-stimulated activation of TAK1, which seemed to offer the opportunity of studying how the p38α MAPK-dependent phosphorylation of TAB3 regulates the activity of TAK1 independently of the phosphorylation of TAB1. The results of these experiments are presented in Figure 6(B). Interestingly, these studies revealed that, although the expression of TAB3 in TAB1−/− and wild-type cells was similar, the IL-1-induced phosphorylation of TAB3 at Ser60 and Thr404 was greatly reduced in the TAB1−/− cells. IL-1 still stimulated the phosphorylation of TAB3 at Ser506 in TAB1−/− cells, but, in contrast with wild-type MEFs, phosphorylation of this site was transient. The IL-1-induced phosphorylation of MAPKAP-K2 was similar in wild-type and TAB1−/− cells. The lack of expression of TAB1 in the TAB1−/− cells explains why IL-1-induced phosphorylation of TAB1 at Ser423 was not detected (Figure 6B).
Phosphorylation of TAB3 in wild-type and TAB1−/− MEFs
TAK1 is not activated by IL-1 or TNFα in TAB1−/− MEFs
Since the IL-1-stimulated phosphorylation of TAB3 at Ser60 and Thr404 did not occur, and the phosphorylation of Ser506 was reduced in TAB1−/− MEFs, it was of interest to investigate how the feedback control of TAK1 by p38α MAPK was affected in these cells. However, this experiment could not be performed, because, surprisingly, the TAK1 immunoprecipitated from the TAB1−/− MEFs did not have detectable activity, whether or not the cells were stimulated with IL-1 (Figure 7A) or TNFα (Figure 7B). We also assessed the phosphorylation of the TAK1 catalytic subunit at Thr187. The phosphorylation of this residue correlates with activation and is thought to be an autophosphorylation event catalysed by TAK1 itself . Although no TAK1 activity could be detected in the immunoprecipitates, some residual IL-1- (Figure 7C) or TNFα- (Figure 7D) stimulated phosphorylation of Thr187 appeared to be present, although it was reduced compared with that observed in wild-type cells.
IL-1α- and TNFα-induced activation of TAK1 in wild-type and TAB1−/− MEFs
IL-1-induced signalling in MEKK3−/− MEFs
The observation that no TAK1 activity could be detected when this protein kinase was immunoprecipitated from IL-1- or TNFα-stimulated TAB1−/− MEFs raised the question of why the IL-1/TNFα-stimulated activation of MAPK cascades was normal in TAB1−/− cells. One possibility was that MEKK3 could substitute for TAK1 in TAB1−/− MEFs, because it has been reported that the IL-1- or LPS-stimulated activation of p38α MAPK, JNK and NF-κB  or the TNFα-stimulated activation of NF-κB  does not occur in MEKK3−/− cells. We therefore generated mice expressing a truncated inactive form of MEKK3 and generated immortalized MEKK3−/− MEFs as well as control wild-type MEFs (see the Materials and methods section, and Figure 8A). However, in contrast with the earlier reports [4,28], we found that IL-1 (Figure 8B) or TNFα (Figure 8C) activated p38α MAPK, JNK and ERK1/2 and induced the degradation of IκBα similarly in MEKK3−/− and wild-type MEFs. These cells express a truncated form of MEKK3 (Figure 8C), which is catalytically inactive (results not shown).
Characterization of IL-1-stimulated signalling in MEKK3−/− MEFs
Identification of the protein kinases that phosphorylate TAB1 at Ser438
We have reported previously that LPS, IL-1, TNF and several cellular stresses induce the phosphorylation of TAB1 at Ser438 and that this is catalysed by a protein kinase(s) distinct from p38α MAPK, because phosphorylation is not suppressed by SB 203580 or in p38α MAPK−/− MEFs . Ser438 is followed by a proline residue, implying that it is phosphorylated by a proline-directed protein kinase, such as another MAPK. Indeed, we have reported previously that TAB1 can be phosphorylated at Ser438in vitro by ERK2, JNK1, JNK2 and all four isoforms of p38 MAPK .
In order to identify the protein kinase that phosphorylates TAB1 at Ser438 in cells, we studied the effect of the MKK1 inhibitor PD 184352 in wild-type MEFs and MEFs that do not express JNK1 or JNK2. These experiments revealed that the IL-1α-stimulated phosphorylation of TAB1 at Ser438 was partially reduced in the JNK1−/−/JNK2−/− MEFs, partially suppressed by PD 184352 in the wild-type MEFs, and greatly reduced by PD 184352 in the JNK1−/−/JNK2−/− MEFs (Figure 9A). Similar results were obtained if the phosphorylation of Ser438 was induced by osmotic shock instead of IL-1 (Figure 9B). These experiments indicate that the phosphorylation of Ser438 in MEFs is catalysed by both ERK and JNK isoforms and that TAB1 is therefore targeted by three different MAPKs in MEFs. However, we have been unable, so far, to detect a significant difference in IL-1-stimulated TAK1 activation between JNK1−/−/JNK2−/− MEFs and wild-type MEFs (results not shown). The role of Ser438 phosphorylation has therefore still to be determined.
Phosphorylation of TAB1 at Ser438 in wild-type and JNK−/− MEFs
Our previous studies have identified a feedback control mechanism by which p38α MAPK down-regulates TAK1 and which correlates with the phosphorylation of TAB1 at Ser423 and Thr431  and a decrease in the electrophoretic mobilities of TAB2 and TAB3 that could be reversed by protein phosphatase treatment . In the present study, we sought to identify the phosphorylation sites on TAB2 and TAB3 in order to evaluate their importance in the feedback control of TAK1.
Previous studies have also shown that the LPS-, IL-1- or osmotic-shock-induced decrease in the electrophoretic mobility of TAB2 was only partially prevented by SB 203580 or in p38α MAPK−/− MEFs, indicating that TAB2 was likely to be phosphorylated in cells by p38α MAPK and at least one other protein kinase . In the present study, we identified two IL-1-stimulated phosphorylation sites on TAB2 in IL-1R cells, Ser372 and Ser524, whose phosphorylation was not prevented by specific inhibition of p38α MAPK (Figures 3 and 4) and which may underlie the p38α MAPK-insensitive decrease in the electrophoretic mobility of TAB2 noted previously .
Ser372 of TAB2 is followed by a proline residue, but several proline-directed protein kinases known to be activated by IL-1 do not appear to phosphorylate Ser372. Thus the MKK1 inhibitor PD 184352 (Figure 1) did not prevent the IL-1-induced phosphorylation of Ser372 in the presence or absence of SB 203580 (Figure 3) and IL-1 stimulated the phosphorylation of Ser372 in JNK−/− cells in a similar manner to that in wild-type cells (Figure 4). Moreover, IL-1 did not activate ERK5 in IL-1R cells (results not shown). It would therefore appear that IL-1 activates a proline-directed Ser372 kinase that has yet to be identified, perhaps p38γ MAPK and/or p38δ MAPK, whose activities are suppressed by high concentrations of BIRB 0796 .
In contrast with Ser372, Ser524 is not followed by a proline residue and its phosphorylation was unaffected by PD 184352, SB 203580 and low or high concentrations of BIRB 0796. These observations indicate that this residue is phosphorylated by an IL-1-stimulated protein kinase that does not lie downstream of ERK1/2, JNK1/2 or p38 MAPK isoforms.
The residue(s) in TAB2 that is targeted by the p38 MAPK pathway remains to be identified, but one site could be Ser582, which became phosphorylated when the TAK1 complex was overexpressed in IL-1R cells (Figure 2D). Although the phosphorylation of the endogenous TAB2 at this site was not detected in IL-1-stimulated IL-1R cells, we have observed the LPS-induced phosphorylation of Ser582 in murine RAW macrophages (results not shown). Moreover, the LPS-induced phosphorylation of Ser582 in RAW cells was prevented by SB 203580, indicating that the phosphorylation of this site may underlie the SB 203580-sensitive decrease in the electrophoretic mobility of TAB2. Ser582 is not followed by a proline residue and is presumably phosphorylated by a protein kinase that is activated by p38α MAPK. Ser582 lies in a Leu-Xaa-Arg-Xaa-Xaa-Ser-Ile sequence, which is an optimal consensus for phosphorylation by MAPKAP-K2/MAPKAP-K3 , but further studies will be needed to establish whether these protein kinases mediate the phosphorylation of Ser582 in LPS-stimulated macrophages.
We identified three residues on TAB3, namely Ser60, Thr404 and Ser506 (Figures 5 and 6) that became phosphorylated in response to IL-1, two of which (Thr404 and Ser506) were located in positions equivalent to Ser372 and Ser524 of TAB2 (Figure 10A). However, in contrast with the phosphorylation sites on TAB2, the phosphorylation of Thr404 and Ser506 of TAB3, as well as Ser60, was prevented by either of two structurally unrelated inhibitors of p38α MAPK: SB 203580 and BIRB 0796. The phosphorylation of these sites presumably accounts for the previously observed IL-1- or LPS-dependent reduction in the electrophoretic mobility of TAB3, which was prevented by SB 203580 and did not occur in p38α MAPK−/− MEFs . Ser60 and Thr404 are followed by a proline residue, and since MAPKs normally phosphorylate Ser-Pro and Thr-Pro sequences, these residues are likely to be phosphorylated directly by p38α MAPK in cells. However, Ser506 is not followed by a proline residue, suggesting that it is not phosphorylated directly by p38α MAPK, but by another protein kinase that is activated by p38α MAPK. This led us to discover that the IL-1-stimulated phosphorylation of Ser506 did not occur in MEFs that do not express MAPKAP-K2 and MAPKAP-K3 (Figure 5C). Thus these protein kinases are likely to mediate the IL-1-induced phosphorylation of Ser506 in wild-type MEFs. The protein kinases that are likely to phosphorylate the different sites on TAB2 and TAB3 in MEFs are summarized in Figure 10(B).
Location of the phosphorylation sites in TAB2 and TAB3 and the protein kinases involved in their phosphorylation
We  and others  have shown that p38α MAPK interacts strongly with a region near the C-terminus of TAB1 and that, in contrast with other MAPKs, including the very closely related p38β MAPK, it can be pulled down on glutathione–Sepharose together with GST–TAB1 . This observation suggested that one role of TAB1 is to recruit p38α MAPK to the TAK1 complex , a hypothesis now supported by the finding that the IL-1-stimulated phosphorylation of TAB3 at Ser60 and Thr404 is greatly reduced in TAB1−/− MEFs (Figure 6B). This implies that the phosphorylation of TAB3 is dependent on the recruitment of p38α MAPK to the TAB1 component of the TAK1 complex.
The IL-1-induced MAPKAP-K2/MAPKAP-K3-catalysed phosphorylation of TAB3 at Ser506 was still observed in TAB1−/− MEFs, but was far more transient than in wild-type MEFs (Figure 6B). The structure of TAB1 has revealed that it is an inactive pseudophosphatase of the MPP family of serine/threonine-specific protein phosphatases and we have suggested that a further role of TAB1 may be to interact with phosphorylated residues in the TAK1 complex, thereby protecting them against dephosphorylation . The failure of TAB1 to protect Ser506 from dephosphorylation in TAB1−/− MEFs could account for its transient phosphorylation compared with the more sustained activation seen in wild-type cells, and may also contribute to the lack of phosphorylation at Ser60 and Thr404.
TAK1 is widely believed to be a ‘master’ kinase that switches on MAPK cascades and NF-κB in response to IL-1, LPS and TNFα, because the activation of these signalling pathways is abolished or reduced in TAK1−/− cells [5–7]. Like the knockout of TAK1, the knockout of TAB1 is embryonic lethal [5,30], and the generation of active TAK1 requires its co-expression with TAB1 [9,31–33] or fusion to the TAK1-binding C-terminal domain of TAB1 [33,34]. These studies implied an essential role for TAB1 in activating the TAK1 complex. However, more recently, the IL-1- and TNFα-induced activation of MAPK cascades and NF-κB were reported to be unaffected by siRNA (small interfering RNA) knockdown of TAB1 in HeLa cells, which reduced the level of expression of this protein by 80%, whereas siRNA knockdown of TAK1 did inhibit the activation of these signalling pathways . Similarly, the IL-1- or TNFα-induced activation of JNK and IKK, or IL-1-, TNFα- or LPS-stimulated NF-κB gene transcription, was not impaired in TAB1−/− MEFs, although it was suppressed in TAK1−/− MEFs . It had been inferred from the latter study that TAB1 was dispensable for the activation of TAK1 by IL-1 and TNFα in these cells. However, consistent with the earlier reports that TAB1 is required for TAK1 catalytic activity, we found here that IL-1 and TNFα were unable to stimulate detectable TAK1 activity in the same TAB1−/− MEFs (Figure 7). The simplest explanation that can account for these observations is that a protein kinase distinct from TAK1 activates MAPK cascades and IKK in TAB1−/− MEFs. However, we cannot exclude the possibility that TAK1 is active in TAB1−/− cells, but is unstable without TAB1 so that activity disappears during its immunoprecipitation from the cell extracts before it can be assayed. Alternatively, the TAK1 in TAB1−/− cells may have trace residual activity (5% or less) that would be difficult to detect in our assay, this slight activity nevertheless being sufficient to sustain normal downstream signalling. Another possible explanation of the data is that the lack of TAB1 prevents TAK1 from activating MKK6 in vitro, but does not prevent TAK1 from activating another protein(s), such as another MAP3K, required to switch on MAPK cascades and NF-κB in response to IL-1 and TNFα. This scenario, or an essential role for TAK1 that does not require its protein kinase activity, could explain why downstream signalling does not occur in TAK1−/− MEFs, but is similar in TAB1−/− and wild-type MEFs. It is also possible that, in the absence of TAK1, TAB1 and/or TAB2/TAB3 become dominant-negative inhibitors of another protein kinase required for IL-1-induced ‘downstream’ signalling.
It has been reported that TAK1 co-operates with MEKK3 in mediating TNFα-induced activation of NF-κB , raising the possibility that MEKK3 mediates IL-1- and TNFα-stimulated ‘downstream’ signalling in TAB1−/− MEFs. Indeed, it has been reported that the IL-1 or LPS-stimulated activation of p38α MAPK, JNK and NF-κB , or the TNFα-stimulated activation of NF-κB , does not occur in MEKK3−/− MEFs. However, using MEFs that express a truncated, but catalytically inactive, form of MEKK3, we found that the IL-1- and TNFα-stimulated activation of p38α MAPK, JNK and ERK1/2 and the degradation of IκBα occurred similarly in the MEKK3−/− and wild-type MEFs (Figure 8). ASK1 (apoptosis signal-regulating kinase 1) is another MAP3K that may mediate ‘downstream’ signalling, since it is activated in response to LPS and LPS-induced activation of p38α MAPK is suppressed in ASK1−/− splenocytes and dendritic cells . However, LPS-stimulated activation of JNK and IKK was similar in ASK1−/− and wild-type splenocytes and dendritic cells , therefore ASK1 may not be the MAP3K that mediates IL-1-induced activation of JNK and IKK in MEFs. Taken together, the present study suggests that further work is needed to clarify which MAP3K mediates the IL-1-stimulated activation of MAPK cascades and IKK in TAB1−/− and wild-type MEFs.
We are grateful to Dr Mark Peggie (University of Dundee) for site-directed mutagenesis of the phosphorylation sites on TAB2 and TAB3, Simon Rousseau (University of Dundee) for performing the immunoblotting for Ser60 and Thr404 phosphorylation in Figure 5(C) and the Sequencing Service, University of Dundee (http://www.dnaseq.co.uk), for DNA sequencing. This study was funded by the UK Medical Research Council, The Royal Society, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Merck and Co., Merck KGaA and Pfizer.
apoptosis signal-regulating kinase 1
Dulbecco's modified Eagle's medium
foetal calf serum
inhibitor of nuclear factor κB α
- IL-1R cells
HEK-293 (human embryonic kidney) cells that stably express the IL-1 receptor
c-Jun N-terminal kinase
mitogen-activated protein kinase
MAPK kinase kinase
MAPK-activated protein kinase
mouse embryonic fibroblast
MAPK/ERK kinase kinase 3
Mg2+-dependent protein phosphatase
nuclear factor κB
small interfering RNA
transforming growth factor-β-activated kinase 1
tumour necrosis factor