p38 maintains E-cadherin expression by modulating TAK1–NF-κB during epithelial-to-mesenchymal transition

Epithelial-to-mesenchymal transition (EMT) of peritoneal mesothelial cells is a pathological process that occurs during peritoneal dialysis. EMT leads to peritoneal fibrosis, ultrafiltration failure and eventually to the discontinuation of therapy. Signaling pathways involved in mesothelial EMT are thus of great interest, but are mostly unknown. We used primary mesothelial cells from human omentum to analyze the role of the p38 MAPK signaling pathway in the induction of EMT. The use of specific inhibitors, a dominant-negative p38 mutant and lentiviral silencing of p38α demonstrated that p38 promotes E-cadherin expression both in untreated cells and in cells co-stimulated with the EMT-inducing stimuli transforming growth factor (TGF)-β1 and interleukin (IL)-1β. p38 inhibition also led to disorganization and downregulation of cytokeratin filaments and zonula occludens (ZO)-1, whereas expression of vimentin was increased. Analysis of transcription factors that repress E-cadherin expression showed that p38 blockade inhibited expression of Snail1 while increasing expression of Twist. Nuclear translocation and transcriptional activity of p65 NF-κB, an important inducer of EMT, was increased by p38 inhibition. Moreover, p38 inhibition increased the phosphorylation of TGF-β-activated kinase 1 (TAK1), NF-κB and IκBα. The effect of p38 inhibition on E-cadherin expression was rescued by modulating the TAK1–NF-κB pathway. Our results demonstrate that p38 maintains E-cadherin expression by suppressing TAK1–NF-κB signaling, thus impeding the induction of EMT in human primary mesothelial cells. This represents a novel role of p38 as a brake or ‘gatekeeper’ of EMT induction by maintaining E-cadherin levels.


Introduction
EMT is a complex, stepwise phenomenon that occurs during embryonic development and tumor progression (Thiery et al., 2009), and is also associated with chronic inflammatory and fibrogenic diseases affecting lung, liver and the peritoneum of patients undergoing peritoneal dialysis (Kalluri and Weinberg, 2009;Aroeira et al., 2007). During dialysis, the peritoneum is exposed to continuous inflammatory stimuli such as hyperosmotic, hyperglycemic and acidic dialysis solutions, as well as episodes of peritonitis and hemoperitoneum, which might cause acute and chronic inflammation and progressively lead to fibrosis, angiogenesis and hyalinizing vasculopathy. Our previous work demonstrated that effluent-derived mesothelial cells (MCs) from peritoneal dialysis patients show phenotypic changes, reminiscent of EMT, which are associated with the time of peritoneal dialysis treatment and with episodes of peritonitis or hemoperitoneum (Yanez-Mo et al., 2003). Moreover, the appearance of EMT signs correlates with structural and functional deterioration of the peritoneal membrane (Aroeira et al., 2007).
EMT is characterized by the disruption of intercellular junctions, replacement of apical-basolateral polarity with front-to-back polarity, and acquisition of migratory and invasive phenotypes.
Cells that have undergone EMT also acquire the capacity to produce extracellular matrix (ECM) components and a wide spectrum of inflammatory, fibrogenic and angiogenic factors. EMT is triggered by an interplay of extracellular signals, including components of the ECM, as well as soluble growth factors and cytokines, including members of the transforming growth factor (TGF)- and fibroblast growth factor families, epidermal growth factor and hepatocyte growth factor (Thiery et al., 2009).
The molecular mechanisms controlling the establishment and progression of EMT appear to be multifold and cell-type specific. One of the key events in EMT is the disruption of cadherin junctions between epithelial cells, and a pivotal role is played by families of transcription factors that repress E-cadherin expression, such as Snail, ZEB and basic helix-loop-helix (bHLH) families (Peinado et al., 2007). The expression of these transcription factors is controlled by a complex network of signaling molecules, including SMADs, integrin-linked kinase, phosphatidylinositol 3kinase, mitogen-activated protein kinases (MAPKs), glycogen synthase kinase 3 , and nuclear factor  B (NF-B) (Thiery et al., 2009;Zavadil and Bottinger, 2005).
MAPKs are serine/threonine kinases that play important roles in a vast array of pathophysiological processes. The family is divided into three main subfamilies: extracellular regulated kinase (ERK), Jun N-terminal kinase (JNK) and p38. All are characterized by the presence of a typical activation module and a conserved activation domain (Chang and Karin, 2001). ERK1 and ERK2 are activated by mitogenic stimuli, whereas JNK and p38, also called stressactivated protein kinases (SAPKs), are activated by environmental and genotoxic stresses (Wagner and Nebreda, 2009). Besides being a central mediator of the inflammatory and stress response, p38 plays an important role in non-inflammatory processes such as cell-cycle regulation and cell differentiation (Cuenda and Rousseau, 2007). Once activated, p38 phosphorylates a wide array of substrates in the cytoplasm and in the nucleus, thus regulating gene expression, cell cycle and cellular polarization. The induction or activation of phosphatases is an important mechanism of crosstalk between MAPKs, and allows p38 to regulate the activity of ERK and JNK (Junttila et al., 2008;Perdiguero et al., 2007). Studies into the roles of MAPK families in the genesis of EMT have produced conflicting results, due to the heterogeneity of the cellular models and the different experimental approaches used. There is compelling evidence that ERKs drive EMT in many experimental systems (Zavadil and Bottinger, 2005), and our previous work has demonstrated the role of ERKs in E-cadherin downregulation and EMT induction (Strippoli et al., 2008). However, the role of SAPKs is less studied, although some reports indicate that JNK is an EMT inducer, also in MCs (Alcorn et al., 2008;Liu, Q. et al., 2008). p38 appears to promote EMT during development and in tumors (Zavadil and Bottinger, 2005;Zohn et al., 2006;Liu, Y. et al., 2008); however, the role of p38 in EMT during chronic inflammatory disease has not been analyzed. Here, we demonstrate that p38 promotes E-cadherin expression by suppressing TGF-activated kinase 1 (TAK1)-NF-B signaling, thus impeding the induction of EMT in human primary mesothelial cells.

Inhibition of p38 MAPK represses E-cadherin and cytokeratin expression in primary MCs
The main inducers of EMT in MCs in vivo are thought to be combinations of inflammatory and profibrotic cytokines, such as TGF-1 and IL-1. We have shown that this combination can induce a genuine EMT in primary omentum-derived MCs (Yañez Mo et al., 2003) and described the role of ERK MAPK in EMT induction in MCs (Strippoli et al., 2008). To analyze the role of SAPKs in E-cadherin expression, we pretreated MCs with specific inhibitors of p38 and JNK before stimulation with a combination of TGF-1 (0.5 ng/ml) and IL-1 (2 ng/ml) for 24 hours (T/I stimulation). These stimuli are able to induce a sustained activation of ERK (Strippoli et al., 2008) and p38 (supplementary material Fig. S1A). Moreover, p38 is stably activated in quiescent MCs and, differently from ERK, its activation levels are increased upon cellular confluency in MCs and in other experimental systems (supplementary material Fig. S1B) (Faust et al., 2005). As previously described, inhibition of ERK signaling with U0126 increased basal E-cadherin expression and limited its downregulation upon cytokine treatment (Fig. 1A) (Strippoli et al., 2008). On the other hand, treatment with p38 inhibitor SB203580 markedly reduced E-cadherin expression in both untreated and cytokine-treated cells (Fig. 1A), whereas inhibition of JNK with SP600125 had some effect in limiting E-cadherin downregulation (Fig. 1A). The effects of p38 inhibition were confirmed by using BIRB 796 (Pargellis et al., 2002), a more specific p38 inhibitor (Fig. 1A). Similar results were obtained in MCs infected with a retroviral vector encoding p38AGF, a non-phosphorylatable and inactive mutant form of p38 (Fig. 1B). Specific blockade of p38 activity with the p38AGF construct was confirmed in cells treated with sodium arsenite, a p38 activator (Westermarck et al., 2001) (supplementary material Fig. S2). The reduction in E-cadherin expression upon p38 inhibition was also demonstrated by infecting primary omentum-derived MCs with two specific lentiviral vectors encoding small hairpin RNA (shRNA) specific for p38 (Fig. 1C), and lentiviral silencing of p38 expression led to the disappearance of E-cadherin from cell membrane junctions ( Fig. 1D; supplementary material Fig. S3A). Inhibition of p38 with pharmacologic inhibitors (Fig. 1E) or by lentiviral-mediated shRNA silencing (Fig. 1F) also reduced the expression of E-cadherinencoding mRNA. On the other hand, treatment of MCs with sodium arsenite led to blockade of T/I-induced E-cadherin downregulation ( Fig. 2A). Confocal microscopy analysis showed increased membrane localization of E-cadherin upon treatment with sodium arsenite in cells stimulated with T/I ( Fig. 2B; supplementary material Fig. S3B). These results strongly suggest that p38 promotes E-cadherin expression and plasma membrane localization in human primary MCs.
Pharmacological inhibition of p38 also reduced the expression of cytokeratin, another epithelial marker whose downregulation is often associated with EMT, and resulted in a disorganized cytokeratin skeleton (Fig. 3A,B). The expression of ZO-1, a component of tight junctions, was also reduced upon p38 inhibition (Fig. 3B). Confocal analysis showed an altered plasma membranecytosolic distribution of ZO-1 in cells treated with p38 inhibitors ( Fig. 3C; supplementary material Fig. S3C). Treatment with p38 inhibitors in combination with T/I stimulation led to the appearance of cells with a fibroblastoid shape, and which no longer expressed ZO-1 at the plasma membrane cell edges (Fig, 3C, arrows). On the other hand, the expression of vimentin, a mesenchymal marker, increased upon p38 inhibition ( Fig. 3B; supplementary material Fig. S4). These results suggest that p38 inhibition might oppose the expression of different epithelial proteins during EMT induction, while increasing levels of mesenchymal markers.

Blockade of p38 decreases expression of Snail1 and induces expression of Twist
We next analyzed the role of p38 in the expression of transcription factors known to directly regulate E-cadherin expression in our experimental conditions. We focused first on Snail1, which is a main regulator of E-cadherin expression and is upregulated upon stimulation with TGF-1 plus IL-1 in our experimental system (Yanez Mo et al., 2003). Stimulation of primary MCs with T/I induced a rapid increase in Snail1-encoding mRNA expression that was inhibited by pretreatment with SB203580 (Fig. 4A), and also by pretreatment with the ERK pathway inhibitor U0126, as previously reported (Strippoli et al., 2008). This effect was also reflected in reduced Snail1 protein expression detected by western blot (Fig. 4B) and by immunofluorescence (Fig. 4C). Interestingly, simultaneous pretreatment with SB203580 and U0126 caused a stronger inhibition of the induction of mRNA encoding Snail1. We analyzed whether this effect was due to an involvement of mitogenand stress-activated kinases 1 and 2 (MSK1/2), which are activated by both ERK and p38 (Vermeulen et al., 2009). However, pretreatment with H89, an inhibitor of MSK1/2 and protein kinase A, ruled out this hypothesis (Fig. 4D). Because increased expression of Snail1 cannot account for decreased E-cadherin upon p38 inhibition, we analyzed the effect of p38 inhibitors on the expression of the zinc finger transcription factors Twist, ZEB1, ZEB2, E47 and Slug. Pretreatment with SB203580 or BIRB 796 had no significant effect on ZEB1 and ZEB2, E47 and Slug, but increased the expression of the bHLH protein Twist (Fig. 4E). The role of p38 on Twist expression was also confirmed by p38 silencing (Fig.  4F). p38 thus has opposing actions on the expression of Snail1 and Twist in MCs.

p38 inhibition increases NF-B nuclear translocation, DNA binding and transcriptional activity in MCs
NF-B plays a major role in EMT induction in many experimental settings, including MCs (Huber et al., 2004;Strippoli et al., 2008;Solanas et al., 2008). Moreover, Dorsal, a Drosophila homolog of p65 NF-B, has been shown to control the transcription of Snail and Twist during development and innate immunity in Drosophila (Furlong et al., 2001), indicating a role in the expression of Snail and bHLH factors; and p38 activity can, depending on the experimental conditions, enhance or limit NF-B function (Orr et al., 2005;Gazel et al., 2008). Pretreatment of MCs with SB203580 or BIRB 796 increased the intensity of NF-B nuclear staining induced by cytokine treatment (Fig. 5A). To analyze NF-B interaction with specific DNA sequences upon p38 inhibition, we performed a p65 NF-B binding essay (supplementary material Fig. S5) using MeT5A cells, an untransformed MC line widely used in peritoneal MC research (Rampino et al., 2001). BIRB 796 pretreatment moderately increased NF-B binding to a specific probe both in unstimulated cells and upon T/I stimulation (Fig.  5B). To test the effect of p38 on NF-B transcriptional activity, we transfected MeT5A cells with a luciferase reporter construct containing multiple NF-B binding sites (KBF-luc). Pretreatment of these cells with BIRB 796 increased basal and cytokine-induced luciferase activity (Fig. 5C). This experiment was reproduced in MeT5A cells infected with the p38AGF dominant-negative construct (Fig. 5D). These results strongly support a role for p38 in limiting NF-B activity in MCs during EMT.

p38 inhibition enhances phosphorylation of TAK1 on Thr187 and of p65 NF-B on Ser536, and reduces PP2A phosphatase activity in cytokine-stimulated MCs
We next examined the possible role of TAK1 in p38-regulated NF-B activation. TAK1 plays a major role in the activation of NF-B and MAPK pathways in response to TGF-1 and IL-1 (Shim et al., 2005). Pharmacologic inhibition of p38 before T/I stimulation increased the phosphorylation of TAK1 at Thr187, which is necessary for TAK1 kinase activity (Singhirunnusorn et al., 2005), and also enhanced the phosphorylation of p65 NF-B at Ser536 (Fig. 6A). These changes coincided with a reduction in the levels of the inhibitor of B (IB), probably due to increased IB phosphorylation and subsequent degradation. Moreover, a timecourse study showed a prolonged TAK1 phosphorylation in cells pretreated with BIRB 796 (Fig. 6B). Because okadaic acid, an inhibitor of serine/threonine phosphatases [especially protein phosphatase 2A, PP2A (Westermarck et al., 2001)] strongly increased TAK1 phosphorylation (Fig. 6B), we analyzed whether p38 inhibition might affect PP2A activity in this experimental system. As shown in Fig. 6C, p38 inhibition by BIRB 796 reduced PP2A activity in cells treated with T/I. Collectively, these results suggest that inhibition of p38 leads to the activation of the TAK1-NF-B pathway, and that this effect might be mediated by inhibition of protein phosphatases, such as PP2A.

TAK1 and NF-B mediate the effects of p38 inhibition on E-cadherin expression and on the acquisition of a spindle-like phenotype
To directly demonstrate the role of enhanced NF-B signaling in the downregulation of E-cadherin induced by p38 inhibition, we 4324 Journal of Cell Science 123 (24) infected MCs with a retroviral vector encoding an IB superrepressor, a non-degradable mutant form (Ser32Ala and Ser36Ala) of the repressor IB. The IB super-repressor rescued the enhanced downregulation of E-cadherin expression observed upon pharmacologic inhibition of p38 (Fig. 7A). Moreover, E-cadherin expression was also rescued by retroviral infection with TAK1 D175A, a catalytically inactive mutant of TAK1 that behaves as a dominant-negative mutant, and a similar result was obtained by pretreating MCs with a derivative of 5Z-7-oxoeanol, a highly specific and potent inhibitor of TAK1 catalytic activity (Yao et al., 2007), both for protein and mRNA expression (Fig. 7B,C). Interestingly, 5Z-7-oxoeanol inhibited the basal and cytokineinduced phosphorylation of p38 (Fig. 7B), supporting our hypothesis that p38 MAPK and the TAK1-NF-B activation pathway are linked by a negative feedback loop. Treatment with 5Z-7-oxoeanol also totally blocked the acquisition of a spindle-like phenotype, characteristic of EMT, upon cytokine stimulation in cells treated with p38 inhibitors (Fig. 8). These experiments demonstrate a causal link between p38 and NF-B, through the modulation of TAK1 activity, in the control of E-cadherin downregulation and EMT induction in MCs.

Discussion
This study aimed to characterize the role of p38 MAPK in the EMT undergone by primary mesothelial cells, in particular in the control of E-cadherin expression. Our results demonstrate that p38 maintains E-cadherin levels in MCs, acting as a break on cytokineinduced E-cadherin downregulation by modulating the TAK1-NF-B activation pathway. To our knowledge, this is the first study to show a role of p38 in supporting E-cadherin levels, and indicates a previously unknown role for p38 MAPK in impeding the progression of EMT. As demonstrated with different experimental approaches, inhibition of p38 activity led to downregulation of Ecadherin expression in MCs. RT-PCR experiments showed a parallel reduction of E-cadherin-encoding mRNA, suggesting that p38 acts in this experimental system at the level of mRNA expression. p38 inhibition was able to further reduce E-cadherin levels upon T/I treatment, which induces EMT in MCs (Yañez-Mo et al., 2003). We obtained opposite results by treating MCs with sodium arsenite, a known p38 activator. Hence, p38 activation counteracted E-cadherin downregulation and disappearance from the plasma membrane, as shown by western blot and confocal microscopy analysis.

4325
p38 maintains E-cadherin expression Treatment with p38 inhibitors also induced a disassembly of the cytokeratin filament network and a reduction of cytokeratin levels. This phenomenon was slower (56 hours) than E-cadherin downregulation (24-48 hours). p38 has been reported to phosphorylate cytokeratin at Ser73, this event allowing the formation of keratin granules, or Mallory bodies in the hepatocytes (Wöll et al., 2007). In MCs, cytokeratin disassembly and downregulation might favour the acquisition of a more motile phenotype, sustained by the increase of other intermediate filaments such as vimentin (Yañez-Mo et al., 2003). Also, prolonged treatment with p38 inhibitors reduced the expression and cellular localization of ZO-1. In particular, the appearance after treatment with T/I plus p38 inhibitors of cells with a fibroblastoid shape, which no longer localize ZO-1 at cell-cell junctions, suggests that an EMT process is taking place in these experimental conditions.
On the other hand, p38 inhibition increased the expression of vimentin, a mesenchymal marker. Vimentin has been reported to be phosphorylated upon p38 activation, and recent studies demonstrate that the p38-hsp27 pathway affects vimentin expression and filament assembly (Liu et al., 2010).
We focused our study on the control mechanism of E-cadherin transcription by p38. We analyzed the expression of transcriptional factors relevant for E-cadherin downregulation, such as zinc finger proteins of Snail, ZEB and bHLH families (Peinado et al., 2007). We found that p38 inhibition limited the expression of Snail1, a factor that is rapidly induced by T/I treatment and plays a major role in the induction of EMT in MCs (Strippoli et al., 2008). With this respect, p38 inhibition is similar, albeit less strong, to the inhibition operated by ERK, which is a major Snail inducer. This result does not account for the role of p38 in controlling E-cadherin levels. Besides Snail, E-cadherin repression could be regulated by a wide array of factors. We analyzed the expression of other Ecadherin repressors and found that p38 inhibition leads to an increase of a bHLH factor, Twist, whereas other transcription factors such as ZEB1, ZEB2, E47 and Slug were not significantly affected by p38 inhibition. The expression of Twist upon p38 inhibition might account for the downregulation of E-cadherin during the induction of EMT in MCs. Interestingly, Twist has been reported to regulate Snail1 expression (Smit et al., 2009). In Drosophila, Twist is induced by Dorsal, a NF-B analog (Jiang et al., 1991). NF-B is an important mediator of inflammation, has been shown to play a role in a tumor model of EMT, is emerging as a central player in many models of EMT, and controls EMT in MCs (Huber et al., 2004;Strippoli et al., 2008). Similarly to E- immunofluorescence. MCs were treated as for B. Cells were fixed, permeabilized, stained with a mouse monoclonal antibody against Snail1 (red) and a rabbit monoclonal against E-cadherin (green) and subjected to confocal microscopy analysis. Panel shows immunofluorescence staining of Snail1 and E-cadherin expression. Cell nuclei are shown in blue (Hoechst 33342). Only Snail1 from T/I-stimulated cells is shown. The results shown are from a representative independent experiment of three performed. (D)Effect of MSK1/2 inhibition on Snail1 expression. MeT5A cells were pretreated with DMSO, 10M H89 or SB203580 plus U0126 and stimulated for 12 hours with T/I as indicated. Quantitative RT-PCR was performed on total RNA, with Snail1-encoding mRNA amounts normalized to expression of histone H3-encoding mRNA. (E)Effect of pharmacologic p38 inhibition on the mRNA expression of the zinc finger transcription factors Twist, Slug, ZEB2, E47 and ZEB1 in MCs. Cells were pretreated as for B and mRNA expression determined by quantitative RT-PCR on total RNA, with normalization to histone H3-encoding mRNA levels. (F)Effect of p38 shRNA silencing on Twist expression. Cells were infected with lentiviral vector encoding a specific p38-targeting shRNA. Quantitative RT-PCR was performed on total RNA, with normalization to histone H3-encoding mRNA levels. Control cells were infected with nonspecific shRNA. cadherin, NF-B has also been demonstrated to inhibit cytokeratin expression in mammary epithelial cells (Chua et al., 2007). Thus, we hypothesize that p38 might control E-cadherin and cytokeratin levels through its effect on NF-B activity.
We investigated the relationship between p38 and NF-B in our experimental system. We found that p38 inhibition led to increased NF-B nuclear translocation, binding to specific DNA sequences and transcriptional activity. These effects were seen both at basal levels and in cells stimulated with T/I. Depending on the stimuli and the experimental models, p38 can both activate or inhibit NF-B transcriptional activity. Various stimuli known to induce p38 activation, such as salicylate, sorbitol, arsenite and UV radiation, can induce NF-B inhibition (Orr et al., 2005;Ivanov, 2000).
Interestingly, Twist plays a role in a major negative feedback for NF-B activity. Twist2 -/mice show uncontrolled inflammation due to aberrant NF-B activity (Sosic et al., 2003). We therefore studied how p38 inhibition might lead to increased NF-B activity. We focused on TAK1, a MAPK kinase kinase rapidly induced by both IL-1 and TGF-1 and playing a role in the activation of p38, JNK and NF-B signaling (Shim et al., 2005). Pretreatment with 4327 p38 maintains E-cadherin expression  p38 inhibitors increased TAK1 phosphorylation at Thr187 induced by T/I. This event correlated with increased phosphorylation of NF-B at Ser536, which is linked to NF-B activation, and with increased IB phosporylation and degradation. All these events might account for increased NF-B activity in cells pretreated with p38 inhibitors. TAK1 controls the NF-B pathway upon TNF stimulation by activating the high-molecular-weight IB kinase (IKK) complex, which phosphorylates IB at Ser32 and Ser36 and NF-B at Ser536 (Sakurai et al., 2003). Moreover, TAK1 has been recently demonstrated to mediate TGF-1-induced NF-B effects in a tumor model of EMT (Neil and Schiemann, 2008). Our results thus suggest that the effects of p38 inhibition on NF-B are related to increased TAK1 activity. Because TAK1 phosphorylation is increased upon treatment with okadaic acid, a serine/threonine phosphatase inhibitor, we hypothesized that p38 might regulate TAK1 activity via this route. In agreement with this hypothesis, we found that treatment with BIRB 796 markedly inhibits PP2A activity in MCs stimulated with T/I. It has already been demonstrated in other experimental settings that p38 coimmunoprecipitates with PP2A and regulates its activity in response to diverse stimuli, including TNF and sodium arsenite (Grethe et IB and TAK1 levels were analyzed to confirm overexpression, and -tubulin was detected as a loading control. (B)MCs were pretreated with the TAK1 inhibitor 9-epimer-11,12-dihydro-(5Z)-7-oxozeanol for 1 hour, followed by pretreatment for 24 hours with p38 inhibitors as in A. Controls were pretreated with DMSO. Cells were then either not treated (NT) or stimulated for 24 hours with T/I. The representative western blot shows the levels of E-cadherin expression and phospho-p38. Expression of -tubulin was detected as a loading control. Densitometric analysis of this blot is shown underneath. The results shown are from a representative independent experiment of three performed. (C)MCs were treated as in B, and E-cadherin-encoding mRNA expression was measured by quantitative RT-PCR performed on total RNA. Histone H3-encoding mRNA expression was used for normalization. Bars represent means + s.e.m. of three independent experiments. Images show the results of a representative independent experiment of three performed. Bottom left: mask of cell areas used for MetaMorph analysis of cell morphology. Borders were tracked and masks were created from representative fields from pictures taken with a phase-contrast microscope (shown in supplementary material Fig. S6). Bottom right: elliptical factor quantification of the experiment described above by using the MetaMorph software. Bars represent means + s.e.m. of three independent experiments. *P<0.05 for DMSO-treated cells compared with cells treated with SB203580 or BIRB 796. Avdi et al., 2002;Westermarck et al., 2001). Moreover, PP2A is stably associated with TAK1 and negatively regulates its activation by TGF-1 (Kim, S. I. et al., 2008). Alternatively, p38 might affect TAK1 activity through the phosphorylation of the TAK1 subunit TAB1, as proposed by Cheung and co-workers (Cheung et al., 2003).
Our analysis of TAK1-NF-B signaling showed that inhibition of the TAK1-NF-B pathway rescues the downregulation of Ecadherin induced by p38 inhibition. These results are summarized in the model shown in Fig. 9. TAK1 and NF-B play a major role, according to our results, in mediating the effect of p38 on Ecadherin expression; however, we cannot exclude TAK1-NF-Bindependent effects of p38 on E-cadherin in our experimental system. In this regard, TAK1 would activate p38, which in turn would trigger a negative feedback loop on the TAK1-NF-B pathway to control excessive EMT.
Our results show that p38 promotes E-cadherin expression in MCs both under basal conditions and upon cytokine treatment. However, the difference between basal and cytokine-stimulated conditions in this experimental system might be simply a quantitative matter because MCs constitutively produce low levels of TGF-1 (Loureiro et al., 2010) that might favor the transition towards a mesenchymal phenotype. Because the decision between epithelial and mesenchymal fate is the outcome of a balance of numerous extracellular signals (Kalluri and Weinberg, 2009), we hypothesize that basal p38 activity in untreated MCs might contribute to the maintenance of the epithelial state, helping to prevent excessive transdifferentation when the epithelium is exposed to inflammatory stimuli.
In this regard, unlike ERK, p38 activity is increased in confluent cultures with respect to dispersed cells (Houde et al., 2001;Faust et al., 2005); this is a particularly interesting finding because cell scattering is characteristic of EMT. We confirmed that p38 is activated in MCs under resting conditions, and that its activation increases with confluency. p38-mediated inhibition of EMT induction might reflect findings in other experimental systems, which show that p38 controls cell differentiation and might contribute to proliferation arrest and tumor suppression (Houde et al., 2001;Perdiguero et al., 2007;Wagner and Nebreda, 2009). Moreover, a recent study identifies p38 at a crossroad of proinflammatory and anti-inflammatory responses (Kim, C. et al., 2008). On the other hand, p38 activity induced by pro-inflammatory stimuli, concomitant with other pathways (such as JNK and ERK), might be responsible for the effects linked to chronic inflammation and EMT, such as increased cell motility and invasion or extracellular matrix protein production, as demonstrated in other experimental systems (Bakin et al., 2002;Bhowmick et al., 2001). In contrast to our results, recent studies (Zohn et al., 2006;Liu, Y. et al., 2008) reported a downregulation of E-cadherin caused by p38 activation during mouse gastrulation and in lung morphogenesis. Interestingly, in these and other studies, mainly performed in tumoral or developmental models, E-cadherin is not regulated by p38 at a transcriptional level. These discrepancies with our results can be explained by the heterogeneity of different experimental approaches and EMT models.
Overall, a deeper understanding of the role of p38 in a chronic inflammatory model (such as the EMT of MCs) might be useful to better understand the role of this kinase in inflammation and other pathophysiological systems, in order to design more efficacious and selective anti-inflammatory strategies.

Isolation and culture of mesothelial cells
Human mesothelial cells (MCs) were obtained by digestion of omentum samples from patients undergoing unrelated abdominal surgery (Stylianou et al., 1990). Samples were digested in 0.125% trypsin containing 0.01% EDTA. Cells were cultured in Earle's M199 medium supplemented with 20% fetal calf serum, 50 U/ml penicillin, 50 g/ml streptomycin, and 2% Biogro-2 (containing insulin, transferrin, ethanolamine and putrescine) (Biological Industries, Beit Haemek, Israel). To induce EMT, MCs were treated with a combination of human recombinant TGF-1 (0.5 ng/ml) and IL-1 (2 ng/ml) (R&D Systems, Minneapolis, MN) as described previously (Yanez-Mo et al., 2003). Although TGF-1 and IL-1 both induce EMT phenotypic changes when administered separately, combined stimulation is necessary to induce genuine EMT. The cytokine doses used were within the range of those detected in peritoneal dialysis fluids from peritonitis patients (Lai et al., 2000) and are similar to those used in previous studies (Yanez-Mo et al., 2003;Yang et al., 1999). The human mesothelial cell line MeT-5A (ATCC, Rockville, MD) was cultured in Earle's M199 medium as above and stimulated with the same doses of TGF-1 and IL-1.

4329
p38 maintains E-cadherin expression Equal amounts of protein were resolved by SDS-PAGE. Proteins were transferred to nitrocellulose membranes (Amersham Life Sciences, Little Chalfont, UK) and probed with antibodies using standard procedures. Nitrocellulose-bound antibodies were detected by chemiluminescence with ECL (Amersham Life Sciences).

Confocal microscopy and immunofluorescence
Cells were fixed for 20 minutes in 3% formaldehyde in PBS, permeabilized in PBS containing 0.2% Triton X-100 for 5 minutes, and blocked with 2% BSA for 20 minutes. For E-cadherin staining, cells were fixed and permeabilized in cold methanol for 5 minutes. Secondary antibodies (conjugated to Alexa Fluor 647, Alexa Fluor 488 and Alexa Fluor 541) and Hoechst 33342 were from Pierce Chemical (Rockford, IL). Confocal images were acquired with a Leica SP5 Confocal Microscope. The elliptical factor (length/width) of cells as a measure of elongation was determined by using the MetaMorph software (Universal Imaging). For the membrane and cytoplasm measurements, fluorescence intensity profiles were analyzed with Leica LAS-AF software and represented graphically with Prism Graph-Pad 5.0.

Infection of MeT5A cells and MCs with retroviral and lentiviral vectors
MCs were infected with pRV-IRES-CopGreen retroviral vectors (Genetrix, Madrid, Spain) encoding p38AGF, a dominant-negative mutant of p38; WT p38 . MCs were infected with a super-repressor IB mutant harboring mutations Ser32Ala and Ser36Ala. The IB super-repressor blocks NF-B nuclear translocation induced by TGF-1 and IL-1, whereas empty vector has no effect (Strippoli et al., 2008). MCs were also infected with a catalytically inactive TAK1 mutant (Asp175Ala, from Phil Cohen's laboratory, University of Dundee, Scotland); or with empty pRV-IRES-CopGreen vector as control. Twenty-four hours before infection, MCs were seeded into six-well plates (2ϫ10 5 cells per well), and retrovirus-producing 293T cells were seeded at 3ϫ10 6 cells per 10-cm plate. For infection, 293T cell supernatants were filtered through a 0.45-m filter (Whatman, Dassel, Germany), and 5 g/ml polybrene (Sigma-Aldrich, St Louis, Missouri) was added to the filtrate. Thereafter, medium was removed from the MCs and replaced with 293T cell supernatants containing the retrovirus. This process was repeated twice at 24-hour intervals. Twenty-four hours after the last exposure to retrovirus, infection efficiency was monitored by fluorescence microscopy (Carl Zeiss, Standort Göttingen, Germany) or FACS analysis (BD FACS Canto, Becton-Dickinson Laboratories, Mountain View, CA). Alternatively, MCs were infected with virus encoding shRNA specific for p38, which was a kind gift from Angel Nebreda, CNIO, Madrid, Spain. MC infection was performed according to the manufacturer's protocols.

RNA interference
MCs were grown on 48-well plates (4ϫ10 4 cells per well) for 24 hours and transfected twice with a 72-hour interval using Dharmafect 1 (0.8 l per well) and the ON-TARGETplus SMARTpool directed against human Twist1 (60 pmol per well) or a nonspecific control siRNA (Dharmacon, Lafayette, CO). At 24 hours after the last transfection, cells were treated as indicated and RNA was extracted 48 hours later. Knockdown efficiency and E-cadherin levels were determined by quantitative PCR (see above).

NF-B p65 DNA binding assay
MeT5A cells were grown to confluence in 10-cm dishes and treated as indicated. Cells were scrapped in 1 ml of cold complete medium, pelleted and washed with cold PBS. Nuclear extracts were prepared using the NE-PER nuclear and cytoplasmic extraction reagents (Thermo Scientific, Rockford, IL) and protein concentration determined with the Bio-Rad protein assay (Bio-Rad Laboratories, Munich, Germany). Nuclear extracts (7 g) were used to determine p65 DNA binding activity with the NF-B p65 transcription factor kit (Thermo Scientific) according to the manufacturer's instructions. The chemiluminiscent signal was measured with the GloMax-Multi Microplate Multimode Reader (Promega).

Cell transfection and luciferase assays
NF-B transcriptional activity was measured by transient transfection of MeT5A cells with the KBF-luc reporter plasmid and subsequent luciferase activity assay (Castellanos et al., 1997). Briefly, 2ϫ10 5 cells were transfected with 2 g of the KBF-luc reporter plasmid together with 500 ng of the reporter plasmid pRL-null, which bears a promoterless Renilla Luciferase gene (Promega, Madison, WI). Transfections were performed by incubating cells for 4 hours with a mixture of DNA and Lipofectamine 2000 at a ratio of 1:2.5 (Invitrogen, Carlsbad, CA) in serum-free medium. After transfection, cells were pretreated overnight with vehicle (DMSO) or U0126 (20 M). Cells were then stimulated with T/I for the times indicated.
Luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions and determined in a Sirius Single Tube Luminometer (Berthold Detection Systems, Pforzheim, Germany). All experiments were performed in triplicate.

Statistical analysis
Statistical significance was determined with a Student's t-test (unpaired t-test) that was performed using Graph pad 4.0 software, where P values of <0.05 were considered significant.