The TGFβ-induced phosphorylation and activation of p38 mitogen-activated protein kinase is mediated by MAP3K4 and MAP3K10 but not TAK1

The signalling pathways downstream of the transforming growth factor beta (TGFβ) family of cytokines play critical roles in all aspects of cellular homeostasis. The phosphorylation and activation of p38 mitogen-activated protein kinase (MAPK) has been implicated in TGFβ-induced epithelial-to-mesenchymal transition and apoptosis. The precise molecular mechanisms by which TGFβ cytokines induce the phosphorylation and activation of p38 MAPK are unclear. In this study, I demonstrate that TGFβ-activated kinase 1 (TAK1/MAP3K7) does not play a role in the TGFβ-induced phosphorylation and activation of p38 MAPK in MEFs and HaCaT keratinocytes. Instead, RNAi-mediated depletion of MAP3K4 and MAP3K10 results in the inhibition of the TGFβ-induced p38 MAPK phosphorylation. Furthermore, the depletion of MAP3K10 from cells homozygously knocked-in with a catalytically inactive mutant of MAP3K4 completely abolishes the TGFβ-induced phosphorylation of p38 MAPK, implying that among MAP3Ks, MAP3K4 and MAP3K10 are sufficient for mediating the TGFβ-induced activation of p38 MAPK.


Summary
The signalling pathways downstream of the transforming growth factor beta (TGFb) family of cytokines play critical roles in all aspects of cellular homeostasis. The phosphorylation and activation of p38 mitogen-activated protein kinase (MAPK) has been implicated in TGFb-induced epithelial-to-mesenchymal transition and apoptosis. The precise molecular mechanisms by which TGFb cytokines induce the phosphorylation and activation of p38 MAPK are unclear. In this study, I demonstrate that TGFb-activated kinase 1 (TAK1/MAP3K7) does not play a role in the TGFb-induced phosphorylation and activation of p38 MAPK in MEFs and HaCaT keratinocytes. Instead, RNAi-mediated depletion of MAP3K4 and MAP3K10 results in the inhibition of the TGFbinduced p38 MAPK phosphorylation. Furthermore, the depletion of MAP3K10 from cells homozygously knocked-in with a catalytically inactive mutant of MAP3K4 completely abolishes the TGFb-induced phosphorylation of p38 MAPK, implying that among MAP3Ks, MAP3K4 and MAP3K10 are sufficient for mediating the TGFb-induced activation of p38 MAPK.

Introduction
Members of the transforming growth factor beta (TGFb) family of cytokines regulate a plethora of cellular processes, including growth control, differentiation, extracellular matrix production, migration, survival and apoptosis [1]. These pleotropic effects are due, in part, to the ability of the TGFb cytokines to exert direct control over multiple signalling networks in addition to the control of the canonical SMAD-dependent signalling pathway [2,3]. Aberrations of both canonical and non-canonical signalling pathways downstream of the TGFb cytokines often result in the manifestation of various human diseases, including fibrosis, cancer progression and metastasis [4,5].
TGFb ligands initiate cellular responses by binding to their cognate type II and type I transmembrane receptor serine threonine protein kinases. Upon ligand binding, the type II receptor phosphorylates and activates the type I receptor kinase [1]. In the canonical pathway, the type I receptors then phosphorylate the SMAD transcription factors-1,5 and 8 in the bone morphogenetic protein (BMP) pathway, and 2 and 3 in the TGFb pathway-which leads to their association with SMAD4 and entry into the nucleus. In the nucleus, together with the other transcription cofactors, SMADs control the expression of hundreds of target genes [1]. the SMAD-independent pathways, often referred to as noncanonical pathways, that are directly modulated by the TGFb ligands include various layers of the mitogen-activated protein kinase (MAPK) pathways, the PI3K/AKT/mTOR pathways and the RhoA-dependent signalling pathways [2]. A combination of the canonical and non-canonical signalling outputs, as well as context-dependent crosstalk inputs from other signalling networks, probably define the nature of cellular responses to TGFb ligands [2,[6][7][8][9].
Among the MAPK pathways activated by the TGFb ligands, the activation of p38 MAPK has been the best characterized and widely investigated. Both SMAD-dependent and -independent mechanisms have been proposed for the TGFb-induced phosphorylation and activation of p38 MAPK [10][11][12][13][14][15][16][17]. Inhibitors of p38 MAPK inhibit the TGFbinduced epithelial-to-mesenchymal transition (EMT), as well as cell death, implying that the activation of p38 MAPK is crucial in regulating these cellular responses to TGFb [14,16]. There are four mammalian p38 MAPK isoforms, namely p38a, p38b, p38d and p38g [18]. While unique roles for different isoforms have been reported [18,19], primarily most stimuli result in the activation of p38a MAPK by the MAP kinase kinases (MKKs) MKK3, MKK6 or MKK4 [20,21]. The MKKs phosphorylate the activation loop residues, Thr180 and Tyr182, of p38a MAPK [20]. Further upstream, MKKs are activated by various MAP3Ks in response to different stimuli [18,20]. The search for the MAP3Ks responsible for mediating the TGFb-induced phosphorylation of p38 MAPK has generated great interest in the field and has led to numerous publications [11,15,17,[22][23][24]. As the name indicates, TGFb-activated kinase 1 (TAK1, also known as MAP3K7) was the first MAP3K reported to mediate the activation of p38 MAPK in response to TGFb [17]. Subsequent reports have claimed that TGFb receptor complexes bind to and activate TNF-receptor-associated factor 6 (TRAF6), resulting in its autoubiquitylation through K63linked ubiquitin chains, and this allows TRAF6 to recruit and activate TAK1 [11,24]. Catalytically active TRAF6 is indispensable for mediating the interleukin-1 receptor (IL-1R) and toll-like receptor (TLR)-mediated activation of TAK1, and subsequent downstream signalling events such as the activation of p38 MAPK and the production of nuclear factor kB (NF-kB) and IFN regulatory factors [25,26]. However, autoubiquitylation of TRAF6 is unlikely to play a role in recruiting TAK1, as lysine-free TRAF6 has been demonstrated to restore IL-1-stimulated TAK1 activation to TRAF6 2/2 cells [25]. Another MAP3K, MEKK4 (MAP3K4), has also been proposed to mediate the TGFb-induced phosphorylation of p38 MAPK through SMAD-dependent expression of GADD45b, which associates with and activates MAP3K4 [15]. MEKK1 (MAP3K1) has been proposed to mediate the TGFb-induced activation of c-Jun N-terminal kinase (JNK) isoforms [27].
In this paper, I examine the roles of various MAP3Ks in mediating the TGFb-induced phosphorylation of p38 MAPK. I demonstrate that the loss of TAK1 (MAP3K7) in mouse embryonic fibroblasts (MEFs) or human keratinocytes (HaCaT) does not affect the levels of TGFb-induced phosphorylation of p38 MAPK. Furthermore, restoring wild-type (WT) or catalytically inactive TAK1 in TAK1-deficient MEFs does not alter the ability of TGFb to induce the phosphorylation of p38 MAPK. By using a comprehensive RNAi screen to knockdown all human MAP3Ks, I demonstrate that the depletion of MEKK4 (MAP3K4) and MLK2 (MAP3K10) results in a moderate reduction in the TGFb-induced phosphorylation of p38 MAPK. The depletion of MLK2 (MAP3K10) in cells with homozygous knockin of catalytically inactive MEKK4 (MAP3K4) results in a complete loss of the TGFb-induced phosphorylation of p38 MAPK, implying that MEKK4 and MLK2 mediate the TGFb-induced phosphorylation and activation of p38 MAPK in MEFs and HaCaT keratinocytes.

TAK1 (MAP3K7) does not mediate the TGFbinduced phosphorylation of p38 MAPK
In order to investigate the contribution of TAK1 in mediating the TGFb-induced phosphorylation of p38 MAPK, I obtained WT and TAK1-deficient MEFs [28]. Additionally, using these cells, I generated TAK1-deficient MEFs stably expressing a control vector, or N-terminal HA-tagged human WT TAK1 or catalytically inactive (kinase dead, KD) TAK1 (figure 1a). Treatment of these cells with TGFb for 45 min resulted in phosphorylation of SMAD2 to the same extent (figure 1b).
Rather surprisingly, the levels of TGFb-induced phosphorylation of p38 MAPK observed in TAK1 2/2 MEFs were similar to those seen in WT MEFs (figure 1b). Restoration of WT TAK1 or KD TAK1 in TAK1-deficient MEFs did not result in significant changes to the levels of TGFb-induced phosphorylation of p38 MAPK, albeit the expression of HA KD TAK1 was less than HA WT TAK1 (figure 1b). In order to determine whether TAK1 plays a role in modulating a time-dependent activation of p38 MAPK in response to TGFb, a time course of TGFb treatment was performed in these cells (figure 1c). No significant differences in TGFb-induced phosphorylation of SMAD2 or p38 MAPK were observed in WT or TAK1deficient MEFs, nor in TAK1-deficient MEFs restored with WT or KD TAK1 at any of the time points assayed (figure 1c). As expected, interleukin-1a (IL-1a) induced a robust p38 MAPK phosphorylation and loss of IkBa only in WT MEFs but not in TAK1-deficient MEFs (figure 1c). The IL-1a-induced p38 MAPK phosphorylation, and loss of IkBa was partially rescued in TAK1-deficient MEFs expressing HA WT TAK1 but not HA KD TAK1 (figure 1c). In order to complement the above findings and definitively establish that TAK1 does not mediate the TGFb-induced phosphorylation of p38 MAPK, I obtained a different set of MEFs from WT and TAK1-deficient mice that were generated independently using different targeting strategy [29]. By using these MEFs, I was able to demonstrate that there was no difference in the levels of TGFb-induced p38 MAPK phosphorylation between WT and TAK1-knockout MEFs (figure 1d). TGFb induced similar levels of SMAD2 phosphorylation in WT or TAK1-knockout MEFs (figure 1d). As expected, IL-1a-induced p38 MAPK was significantly inhibited in TAK1-knockout MEFs compared with the WT (figure 1d).

TGFb does not activate TAK1
Next, in order to assess whether TGFb induces TAK1 kinase activity, I used an in vitro kinase assay developed for the measurement of TAK1 activity from cell extracts [30]. As expected, TGFb or IL-1a did not stimulate any TAK1 activity in TAK1-deficient cells or TAK1-deficient cells stably rsob.royalsocietypublishing.org Open Biol 3: 130067 expressing KD TAK1 (figure 2). In TAK1-deficient cells stably expressing WT TAK1, a basal TAK1 kinase activity was detected under ambient conditions (figure 2). Treatment of these cells with IL-1a stimulated a significant increase in TAK1 kinase activity (figure 2). However, treatment of these cells with TGFb did not induce TAK1 activity over basal untreated conditions (figure 2). In all cases, TGFb induced similar levels of p38 MAPK and SMAD2 phosphorylation. Treatment of cells with IL-1a resulted in the phosphorylation of p38 MAPK only in TAK1-deficient cells stably expressing WT TAK1 (figure 2), but not in TAK1-deficient cells or TAK1-deficient cells expressing KD TAK1 (figure 2).

TAK1 does not affect BMP-induced phosphorylation of SMAD1 in mouse embryonic fibroblasts
It has been reported that TAK1 impacts the BMP pathway in chondrocytes in part by directly phosphorylating the BMPactivated SMADs at their activating SXS motif [31]. Treatment of both WT MEFs and TAK1-deficient MEFs with BMP-2 led to phosphorylation of SMAD1 at Ser463 and Ser465 to the same extent (figure 3). Furthermore, restoration of WT TAK1 or KD TAK1 in TAK1-deficient MEFs did not alter the levels of BMP-induced phosphorylation of SMAD1, indicating that TAK1 does not mediate the BMP-induced phosphorylation of SMAD1 in MEFs (figure 3). It is therefore likely that any effect that TAK1 has on BMP signalling does not involve direct phosphorylation of SMAD proteins. rsob.royalsocietypublishing.org Open Biol 3: 130067 p38 MAPK phosphorylation. As control, I treated cells with human IL-1b, which is known to promote p38 MAPK phosphorylation through activation of TAK1 [32]. For each MAP3K target and a non-MAP3K control target, a pool of four siRNAs were transfected into HaCaT cells. As anticipated, the IL-1binduced phosphorylation of p38 MAPK was substantially depleted only upon TAK1 (MAP3K7) knockdown, but was unaffected by knockdown of other MAP3Ks ( figure 4a,b). The siRNA pool targeting TAK1 resulted in a robust depletion in expression of endogenous TAK1 protein (figure 4b). In all cases, the treatment of cells with TGFb resulted in the phosphorylation of SMAD2 (figure 4a). The TGFb-induced p38 MAPK phosphorylation was not affected by depletion of the majority of MAP3Ks, including TAK1 (figure 4a,b). However, the siRNA depletion of MAP3K4 (MEKK4) and MAP3K10 (MLK2) resulted in a significant reduction in the levels of TGFb-induced phosphorylation of p38 MAPK (figure 4a), an observation that is further evident when MAP3K4, TAK1 and MAP3K10 siRNA screens are compared together (figure 4b).
No immunoblotting antibodies were available to detect endogenous levels of MAP3K4 and MAP3K10 proteins (figure 4b).

RNAi knockdown of MAP3K10 in cells expressing
catalytically inactive MAP3K4 (MAP3K4-KD) completely abolishes TGFb-induced phosphorylation of p38 MAPK MAP3K4 (MEKK4) has previously been implicated in mediating SMAD-dependent activation of p38 MAPK [15]. This has, however, never been confirmed in cells derived from mice in which a catalytically inactive MAP3K4 (MAP3K4-KD) has replaced the WT protein [33]. I obtained WT and MAP3K4-KD MEFs, and investigated the TGFb-induced phosphorylation of p38 MAPK in these MEFs.

TGFb-induced p38 MAPK activation impacts CREB phosphorylation and transcription
I investigated whether TGFb-induced activation of p38 MAPK impacted downstream signalling in MEFs. cAMP response element-binding protein (CREB) is phosphorylated at Ser133 upon activation of p38 MAPK and drives the expression of multiple target genes, including AREG and IL-6 [35][36][37][38]. In MEFs, TGFb induced a robust CREB phosphorylation (figure 6a). This phosphorylation was inhibited by VX745, a selective and potent inhibitor of p38 MAPK (figure 6a) [39]. VX745 had no effect on TGFb-induced phosphorylation of SMAD2 (figure 6a). By RT-PCR, I demonstrate that TGFb induces the expression of AREG and IL-6 transcripts in MEFs (figure 6b). VX745 completely abolished the TGFbinduced expression of AREG and IL-6 transcripts, suggesting that TGFb-induced activation of p38 MAPK is necessary and sufficient for expression of these transcripts (figure 6b).

Discussion
In  have demonstrated that the TGFb-induced phosphorylation of SMAD2 and p38 MAPK remain unaffected in these cells. Furthermore, while the treatment of cells with IL-1a resulted in a robust activation of TAK1 in cells expressing WT TAK1, TGFb stimulation did not enhance TAK1 activity. These findings contradict many reports implying a role for TAK1 in mediating the TGFb-induced phosphorylation of p38 MAPK [10,11,17,25]. The report describing the discovery of TAK1 assigned its role in the TGFb pathway based primarily on the ability of overexpressed TAK1 mutants to drive TGFb-induced transcriptional reporter activity [17]. However, at the time many of the MAP3Ks were yet to be discovered. Even the discovery of SMAD proteins as key mediators of the TGFb pathway was made a year after the first reported role for TAK1 in the TGFb pathway [40,41]. A robust validation of the original findings with improved tools, including TAK1-knockout cells, was therfore long overdue. More recent reports addressing the possible role of TAK1 in TGFb-induced phosphorylation of p38 MAPK have focused mainly on TRAF6 as an upstream activator of TAK1 [11,24]. While both reports describe how TRAF6 is essential for TGFb-induced phosphorylation of p38 MAPK in MEFs and autoubiquitylation of TRAF6 in response to TGFb recruits TAK1, neither report assesses the TGFb-induced activation of TAK1 [11,24]. Furthermore, TRAF6 autoubiquitylation has been shown to be dispensable for IL1 and TLR-mediated activation of TAK1 [25]. It is therefore likely that TRAF6 could mediate the TGFb-induced phosphorylation of p38 MAPK through another MAP3K.
By using an unbiased siRNA screen targeting all the human MAP3Ks, I have demonstrated that depletion of MAP3K4 (MEKK4) and MAP3K10 (MLK2) results in a significant inhibition of the TGFb-induced phosphorylation of p38 MAPK, whereas the depletion of other MAP3Ks, including TAK1, did not significantly affect the p38 MAPK phosphorylation in response to TGFb. As a proof that the siRNA screen used was an acceptable approach, I show that IL-1b-induced phosphorylation of p38 MAPK, which requires TAK1, is inhibited upon depletion of TAK1 by siRNA pool. However, some off-target effects of the siRNAs in the screen cannot be ruled out. Furthermore, because I was unable to verify the expression of intended targets by immunoblotting or RT-PCR, it is possible that certain targets may not have been depleted by certain siRNAs. Despite the limitations of the siRNA screen, it was very interesting that depletion of MAP3K4 resulted in the inhibition of TGFbinduced p38 MAPK phosphorylation. MAP3K4 has previously been reported to mediate SMAD-dependent activation of p38 MAPK, in part by associating with an activator GADD45b, which is transcribed in response to TGFb [15]. By using MEFs isolated from knockin mice expressing rsob.royalsocietypublishing.org Open Biol 3: 130067 catalytically inactive MAP3K4, I have demonstrated for the first time that these MEFs display attenuated levels of p38 MAPK phosphorylation in response to TGFb compared with the WT MEFs. This suggested that MAP3K4 was not sufficient to mediate the TGFb-induced phosphorylation of p38 MAPK and implied a role for additional MAP3Ks in this process. When I depleted MAP3K10 with two independent siRNAs from MAP3K4-KD MEFs, the TGFb-induced p38 MAPK phosphorylation was completely inhibited. However, depletion of MAP3K10 from WT MEFs, which still have intact MAP3K4, resulted in partial inhibition of the TGFbinduced p38 MAPK phosphorylation. Taken together, it is clear that MAP3K4 and MAP3K10 are the two MAP3Ks that sufficiently mediate the TGFb-induced phosphorylation of p38 MAPK in MEFs and HaCaT cells.
One of the most interesting observations from this study is that MAP3K10 (MLK2) lies upstream of p38 MAPK phosphorylation in response to TGFb. MLK2 is a member of the mixed-lineage subfamily of kinases that includes three other members, namely MLK1 (MAP3K9), MLK3 (MAP3K11) and MLK4 (KIAA1804). It is interesting, however, that depletion of MLK1, MLK3 or MLK4 by siRNAs did not result in the inhibition of the TGFb-induced p38 MAPK phosphorylation, indicating that MLK2 is sufficient for TGFb-induced phosphorylation of p38 MAPK. In fact, depletion of MLK2 by two independent siRNAs, despite increasing the levels of MLK3 mRNA, inhibited TGFb-induced p38 MAPK phosphorylation. It has been reported that overexpressed MAP3K11 (MLK3) was activated by TGFb and was able to mediate the TGFb-induced phosphorylation of p38 MAPK [12]. However, in my hands, the depletion of MAP3K11 in HaCaT cells did not result in any inhibition of the TGFbinduced p38 MAPK phosphorylation. Furthermore, the depletion of MAP3K10 (MLK2) enhanced the levels of MAP3K11 transcripts in MEFs, but still resulted in the inhibition of TGFb-induced p38 MAPK phosphorylation. Collectively, these results imply that MAP3K10 (MLK2), but not MAP3K11, mediates the TGFb-induced phosphorylation of p38 MAPK. Investigation of TGFb-induced p38 MAPK phosphorylation in MAP3K10-null MEFs would provide a definitive answer to the extent of its involvement in the TGFb pathway. It will also be interesting to establish whether TGFb activates the MLK family of protein kinases.
Because p38 MAPK lies at the heart of multiple pro-inflammatory signalling inputs, it has been an attractive target for inhibition to treat inflammatory diseases such as rheumatoid arthritis, psoriasis and chronic obstructive pulmonary disease [18,42]. As a result, numerous small molecule inhibitors of p38 MAPK, including VX745, have been developed and have entered clinical trials. Given the critical roles of p38 MAPK in mediating TGFb-induced CREB phosphorylation and transcription (figure 6), as well as EMT [16] and cell death [13,43], the use of these inhibitors in a clinical context may have consequences on the TGFb responses as well. Depending on different biological contexts, p38 MAPK inhibitors may prove to be useful as inhibitors of TGFb-induced metastasis (by inhibiting rsob.royalsocietypublishing.org Open Biol 3: 130067 EMT) or may prove less useful by promoting tumour proliferation through blocking TGFb-induced apoptosis.

Materials
Antibodies recognizing total p38 MAPK, phospho-p38 MAPK (Thr180 and Tyr182), total SAPK/JNK, phospho-SMAD1 (Ser463/465), phospho-SMAD2 (Ser465/467), total SMAD2/ 3, phospho-CREB (Ser133), total CREB and MAP3K3 (MEKK3) were purchased from Cell Signaling. Antibodies recognizing phospho-JNK1/2 (Thr183/Tyr185) and Lipofectamine 2000 reagent were purchased from Invitrogen. Antibodies recognizing TAK1, MAP3K8, MAP3K4, MAP3K10 (N-and C-terminus) and MAP3K11 were purchased from Santa Cruz Biotechnology. 32 P g-ATP was from Perkin-Elmer. BMP-2, TGFb1 and mouse IL-1a were from R&D Biosystems. Polybrene, Puromycin, DMSO and Tween-20 were from Sigma. Antibody recognizing SMAD1 was generated in sheep against a glutathione S-transferase-tagged SMAD1(144-268) fragment and affinity purified [44]. Immunoprecipitating anti-TAB1 antibody was generated in sheep against a His-TAB1 protein and affinity purified. Pre-immune IgG was purified and used as control. Human IL-1b was expressed in bacteria. These antibodies and proteins were generated at the Division of Signal Transduction Therapy at the University of Dundee (UK). TAK1-deficient (TAK1 2/2 ) and corresponding WT MEFs were a generous gift from S. Akira (Osaka University, Japan) [32]. Independently generated TAK1-null and corresponding WT MEFs were obtained from Prof. Sankar Ghosh (Columbia University, New York, NY) [29]. The MAP3K4-KD MEFs, derived from mouse embryos in which the WT protein is replaced by the catalytically inactive MA3K4 (K1361R) mutant, and the corresponding WT MEFs, were a generous gift from G. Johnson (University of North Carolina, NC) [33]. VX745 was generated by Natalia Shpiro (University of Dundee). For real-time PCR, isolated RNA (1 mg) was used to prepare cDNA using I-script kit (BioRad). Reactions from three biological replicates were performed in triplicates of 20 ml each, including 0.5 per cent cDNA, 1 mM forward and reverse primers, and 50 per cent SYBR Green (Quanta). RT-PCR was performed using a standard protocol in a CFX 384 real-time system (BioRad). Data were normalized to a housekeeping gene (GAPDH or 18S). The data were analysed as reported previously [7,45].

Cell culture, stimulation and lysis
HaCaT cells, two independent sets of WT MEFs and TAK1deficient MEFs, MAP3K4 þ/þ , and MAP3K4-KD MEFs were cultured in dishes of 10 cm diameter in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10 per cent foetal bovine serum (FBS), 1 per cent penicillin/streptomycin mix and 2 mM L-glutamine (D10F). All cells were grown under a humidified atmosphere with 5 per cent CO 2 at a constant temperature of 378C. TAK1-deficient MEFs stably expressing a control vector or WT TAK1 (TAK1-WT MEFs) or KD TAK1 (TAK1-KD MEFs) were cultured as above except that the medium was supplemented with 2 mg ml 21 puromycin. Individual or pools of siRNAs (300 pmoles final per 10 cm 50% confluent dish) were transfected using Lipofectamine 2000 reagent as described previously [34,46]. Cells were cultured in DMEM containing 0.1 per cent FBS for 16 h prior to treatment with appropriate ligands. Unless stated otherwise, cells were treated with TGFb (50 pM rsob.royalsocietypublishing.org Open Biol 3: 130067 final) or BMP-2 (25 ng ml 21 final) for 45 min, and human IL-1b (1 mg ml 21 final) or mouse IL-1a (5 ng ml 21 final) for 10 min. Cells were then washed once with ice-cold PBS and lysed in 0.5 ml ice-cold complete lysis buffer (50 mM Tris -HCl pH 7.5, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, 1 mM sodium orthovanadate, 50 mM sodium flouride, 5 mM sodium pyrophosphate, 0.27 M sucrose, 5 mM bglycerophosphate, 0.1% (v/v) 2-mercaptoethanol, 0.5 mM microcystin-LR, one tablet per 25 ml of complete protease inhibitor cocktail). The extracts were cleared by centrifuging at 16 000g at 48C for 10 min and snap-frozen in liquid nitrogen for storage at 2808C or processed immediately. For RNA isolation, cells were processed using Nucleospin II RNA Isolation kit (Macherey-Nagel) according to the manufacturer's instructions.

SDS -PAGE and immunoblotting
Cleared cell extracts (20 mg) were heated at 958C for 5 min in 1Â SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 10% (v/ v) glycerol, 2% (w/v) SDS, 0.02% (w/v) bromophenol blue and 1% (v/v) b-mercaptoethanol), resolved on a 10 per cent polyacrylamide gel by electrophoresis and transferred to nitrocellulose membranes. Membranes were blocked in TBS-T buffer (50 mM Tris-HCl pH 7.5, 0.15 M NaCl and 0.1% (w/ v) Tween-20) containing 10 per cent (w/v) non-fat milk for 1 h at room temperature. The membranes were then incubated with the indicated antibodies (diluted in TBS-T containing 10% (w/v) milk) for 16 h at 48C, washed 2 Â 10 min in TBS-T buffer, probed with the HRP-conjugated secondary antibodies (diluted 1 : 5000 in TBS-T/5% milk) for 1 h at room temperature, and washed 3 Â 10 min in TBS-T buffer. Enhanced chemiluminescence reagent was used to detect the signals.

5.6.
In vitro TAK1 kinase assay TAK1 assays were performed as described previously [30]. Briefly, TAK1 associated with TAB1 was immunoprecipitated from cell extracts (1 mg total protein) using 2 mg of anti-TAB1 antibody coupled to 5 ml of protein G-sepharose beads for 2 h at 48C. The immunoprecipitates were washed twice with 1 ml of lysis buffer containing 0.5 M NaCl followed by two further washes with 1 ml of 50 mM Tris -HCl pH 7.5, 0.1 mM EGTA and 0.1% (v/v) 2-mercaptoethanol. The TAK1 activity in the immunoprecipitates was assessed by its ability to activate MKK6 as judged by its activation of p38a MAPK. The activity of p38a MAPK was then assayed by measuring its ability to phosphorylate MBP as described previously [30].