TAK1 Regulates Reactive Oxygen Species and Cell Death in Keratinocytes, Which Is Essential for Skin Integrity*

Mice with a keratinocyte-specific deletion of Tak1 exhibit severe skin inflammation due to hypersensitivity to tumor necrosis factor (TNF) killing. Here we have examined the mechanisms underlying this hypersensitivity. We found that TAK1 deficiency up-regulates reactive oxygen species (ROS) resulting in cell death upon TNF or oxidative stress challenge. Because blockade of NF-κB did not increase ROS or did not sensitize cells to oxidative stress in keratinocytes TAK1 regulates ROS mainly through the mechanisms other than those mediated by NF-κB. We found that c-Jun was decreased in TAK1-deficient keratinocytes and that ectopic expression of c-Jun could partially inhibit TNF-induced increase of ROS and cell death. Finally, we show that, in an in vivo setting, the antioxidant treatment could reduce an inflammatory condition in keratinocyte-specific Tak1 deletion mice. Thus, TAK1 regulates ROS partially through c-Jun, which is important for preventing ROS-induced skin inflammation.

Tumor necrosis factor (TNF) 2 plays a central role in inflammation, and also regulates cell death and survival (1)(2)(3). TNF initiates intracellular signaling by binding to its receptor, initiating the formation of the TNF receptor complex, which consists of several proteins including RIP1 kinase, Fas-associated death domain (FADD), and pro-caspase-8 (also called FLICE). The TNF receptor complex in turn activates two opposing intracellular signaling pathways; one leads to up-regulation of the expression of anti-apoptotic genes such as cellular FLICEinhibitory protein (c-FLIP) (4) and caspase inhibitor IAPs (inhibitor of apoptosis proteins) (5) (anti-cell death pathway); another activates the caspase cascade to execute apoptotic cell death (pro-cell death pathway). TNF-induced activation of the transcription factor NF-B is one of the major pathways of antiapoptotic gene up-regulation and cell death inhibition. NF-B is activated upon degradation of IB (inhibitor of NF-B), which is induced by its phosphorylation by IB kinases (6).
TAK1 kinase is a member of the mitogen-activated protein kinase kinase kinase family and is activated by innate immune stimuli including bacterial components and proinflammatory cytokines such as interleukin-1 and TNF (13,14). TAK1 is an ubiquitin-dependent kinase and plays an essential role in innate immune signaling by activating both IB kinases-NF-B and mitogen-activated protein kinase (MAPK) pathways leading to activation of transcription factor AP-1 (15,16). We have recently generated mice harboring a skin keratinocyte (epidermal)-specific deletion of Tak1 and found that TAK1 is essential for keratinocyte survival in vivo (17). We have found that TAK1 deficiency causes hypersensitivity to TNF-mediated killing in keratinocytes. TNF expressed in the skin kills TAK1-deficient keratinocytes and induces an inflammatory condition in the Tak1 mutant skin. However, the mechanism by which TAK1 deficiency increases the susceptibility to TNF killing has not yet determined. One plausible mechanism is that TAK1 deficiency impairs TNF-induced NF-B resulting in activation of the prodeath caspase pathway. However, the degree of cell death is somewhat greater in TAK1-deficient mouse embryonic fibroblasts (MEFs) than NF-B-deficient MEFs as described previously (18). Furthermore, we have recently found that Tak1 deletion in the intestinal epithelium causes epithelial cell death at a much higher degree compared with the intestinal epithelium-specific deletion of NEMO (19,20). Therefore, we speculate that TAK1 may participate in cell survival pathways other than those mediated by NF-B.
In this study, we investigated the mechanisms by which TAK1 regulates sensitivity to TNF-induced cell death. Particularly, we are interested in whether TAK1 deficiency induces cell death through the mechanisms mediated by other than lack of NF-B activation in TAK1-deficient keratinocytes. To this end, we generated NF-B-deficient keratinocytes by expressing non-degradable IB (IB⌬N) that completely blocks activation of NF-B, and compared them with TAK1-deficient keratinocytes. We show here that TNF-induced ROS is mainly regulated by a non-NF-B mechanism in TAK1-deficient keratinocytes, and that this TAK1-dependent ROS regulation is important for preventing the inflammatory conditions in mice having an epidermal specific Tak1 deletion.

EXPERIMENTAL PROCEDURES
Mice-Tak1-floxed (Tak1 flox/flox ) mice were from Dr. Akira, Osaka University (14). K14-CreERT mice were obtained from the Jackson Laboratories (21). Mice harboring an epidermalspecific inducible Tak1 deletion (K14-CreERT Tak1 flox/flox ) were generated and at 5 weeks of age were topically treated on the dorsal skin once a day for 5 consecutive days with 4-hydroxytamoxifen (1 mg/mouse). Some mice were fed with food containing 0.7% butylated hydroxyanisole (BHA) starting at 3 days prior to the tamoxifen treatment. Deletion of the Tak1 gene was confirmed by PCR using the following primer set that can detect the truncated from of Tak1 (Tak1⌬): CACCAGT-GCTGGATTCTTTTTGAGGC and GGAACCCGTGGATA-AGTGCACTTGAAT. TNF receptor was done as a loading control using the following primer set: TGTGAAAAGGGCA-CCTTTACGGC and GGCTGCAGTCCACGCACTGG. All animal experiments were done with the approval of the North Carolina State University Institutional Animal Care and Use Committee.
Annexin V Binding Assay-To determine apoptotic cells, Annexin V-Alexa Fluor 488 binding and propidium iodide staining were performed according to the manufacturer's protocol (Invitrogen). Images were taken using an inverted fluorescent microscope (TE2000-S; Nikon). 3-5 randomly selected areas were photographed with the same exposure time. The images were processed using the fixed threshold in all samples in each experiment, and more than 1000 cells were counted for each sample.
Crystal Violet Assay-The viable adherent cells were fixed with 10% formalin and stained with 0.1% crystal violet. The stain was solubilized by adding 25% ethanol containing 50 mM sodium citrate, and the absorbance of each plate was determined at 595 nm.
Real-time PCR Analysis-Total RNA was prepared from keratinocytes using the RNeasy protect mini kit (Qiagen). cDNA was synthesized using the reverse transcription reagents (Applied Biosystems). Real-time PCR analysis was performed using the ABI PRISM 7000 sequence detection system and the Assays-on-Demand gene expression kit (Applied Biosystems). All samples were normalized to the signal generated from glyceraldehyde-3-phosphate dehydrogenase.
ROS Measurement-Keratinocytes were stimulated with TNF and incubated with 10 M CM-H 2 DCFDA (Invitrogen) for 30 min at 37°C, harvested, and analyzed by flow cytometry.
Immunoblotting-Cells were washed once with ice-cold phosphate-buffered saline and whole cell extracts were prepared using a lysis buffer (20 mM HEPES (pH 7.4), 150 mM NaCl, 12.5 mM ␤-glycerophosphate, 1.5 mM MgCl 2 , 2 mM EGTA, 10 mM sodium fluoride, 2 mM dithiothreitol, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 20 M aprotinin, 0.5% Triton X-100). Cell extracts were resolved on SDS-PAGE and transferred to Hybond-P membranes (GE Healthcare). The membranes were immunoblotted with various antibodies, and the bound antibodies were visualized with horseradish peroxidaseconjugated antibodies against rabbit or mouse IgG using the ECL Western blotting system (GE Healthcare).
Analysis of Cytochrome c Release-Cytosolic fraction was obtained using cytochrome c release apoptosis assay kit (Calbiochem). Briefly, keratinocytes were collected by trypsinization and sonicated in cytosol extraction buffer. Homogenates were fractionated into cytosolic and mitochondrial fractions. Immunoblot was performed using monoclonal antibody against cytochrome c (BD Bioscience).
Retroviral Infection-Retroviral vectors for c-Jun (pMXpuro-c-Jun) and Ib⌬n (pQCXIP-IB⌬N) was generated by inserting c-Jun cDNA into the retroviral vector pMX-puro (22) or IB⌬N into the retroviral vector pQCXIP (Clontech). IB⌬N cDNA was a gift from Dr. Ballard, Vanderbilt University (23). EcoPack293 cells (BD Biosciences) were transiently transfected with pMX-puro-c-Jun or pQCXIP-IB⌬N. After 48 h culture, growth medium containing retrovirus was collected and filtered with 0.45-m cellulose acetate membrane to remove packaging cells. Keratinocytes were incubated with the collected virus-containing medium with 8 g/ml Polybrene for 24 h. Uninfected cells were removed by puromycin selection.
Histology and TUNEL Staining-Paraffin sections were stained with hematoxylin and eosin for histological analysis. dUTP nick-end labeling (TUNEL) assay was performed on frozen sections using an apoptotic cell death detection kit (Promega) according to the manufacturer's instructions. Immunohistochemical analysis was performed on paraffin sections using polyclonal antibodies against K5, K6, Loricrin (Convance), and cleaved caspase-3 (Cell Signaling). Sections were counterstained with hematoxylin.

RESULTS
TNF Induces Pro-cell Death Events in Tak1⌬/⌬ Keratinocytes-We first determined which cell signaling events were induced in Tak1⌬/⌬ keratinocytes following TNF stimulation. Tak1 wild-type and ⌬/⌬ keratinocytes were isolated from Tak1 flox/flox or K5-Cre Tak1 flox/flox mice, respectively. In this floxed Tak1 system, Cre recombinase catalyzes the deletion of the TAK1 ATP binding site, amino acids 41-77, resulting in the generation of a truncated form of TAK1 (TAK1⌬) (Fig. 1A, bottom panel). As reported previously (17), TNF activated NF-B in Tak1 wild-type keratinocytes, but this activation was largely abolished in Tak1⌬/⌬ keratinocytes (Fig. 1A, top panel). We next examined pro-cell death events, including activation of JNK, degradation of c-FLIP, and activation of caspase in Tak1 wild-type and ⌬/⌬ keratinocytes. TNF activated JNK in a transient manner, peaking at 10 min post-stimulation in Tak1 wildtype keratinocytes, whereas no transient JNK activation was detected in Tak1⌬/⌬ keratinocytes (Fig. 1A, third panel). Thus, TAK1 is essential for TNF-induced transient JNK activation, consistent with earlier studies using TAK1-deficient MEFs (14,18). However, JNK was greatly activated at 3-6 h after TNF stimulation in Tak1⌬/⌬ but not in wild-type keratinocytes. It has been reported that ROS accumulation activates JNK in a prolonged manner following TNF stimulation (9, 10). Prolonged activation of JNK has been linked to c-FLIP degradation and subsequent caspase activation (10). We examined c-FILP degradation and caspase-3 activation (Fig. 1A, fifth and sixth panels). As anticipated, c-FLIP was degraded and caspase-3 was activated in Tak1⌬/⌬ but not in wild-type keratinocytes. We measured ROS production using CM-H 2 DCFDA, a substrate that exhibits increased fluorescence when oxidized by intracellular ROS (Fig. 1B). The levels of ROS were greatly increased in Tak1⌬/⌬ but not in wild-type keratinocytes following TNF stimulation. We also examined whether cell death is through the apoptotic or necrotic pathways. Apoptosis and necrosis were assessed by Annexin V binding and propidium iodide staining (Fig. 1C). The numbers of both apoptotic and necrotic cells were greatly increased in Tak1⌬/⌬ keratinocytes at 2-6 h post-TNF stimulation. Cytochrome c was increased in the cytoplasm in TNF-stimulated Tak1⌬/⌬ but not in wild-type keratinocytes (Fig. 1D), suggesting that TNF activates the mitochondrial apoptosis pathway in Tak1⌬/⌬ keratinocytes. Collectively, these results demonstrate that TNF induces pro-cell death events including activation of JNK and caspase-3 in TAK1⌬/⌬ keratinocytes, which occur concomitantly with accumulation of ROS, resulting in apoptotic and necrotic cell death.

Inhibition of ROS Completely Blocks TNF-induced Cell Death
in Tak1⌬/⌬ Keratinocytes-It is known that ROS activates JNK resulting in activation of caspases through c-FLIP degradation. wild-type (ϩ/ϩ) and ⌬/⌬ keratinocytes were stimulated with 20 ng/ml TNF and Annexin V-Alexa Fluor 488 binding and propidium iodide staining was performed. Annexin V-positive and propidium iodide-negative cells were counted as apoptotic cells, whereas double Annexin V-and propidium iodide-positive cells were counted as necrotic cells. Data are representative of two independent experiments with similar results. D, Tak1 wild-type (ϩ/ϩ) and ⌬/⌬ keratinocytes were stimulated with 20 ng/ml TNF for 2.5 or 5 h and fractionated into cytosolic and mitochondrial fractions. Cytochrome c release was analyzed by immunoblot analysis using cytosolic fraction. ␤-Actin was used as a loading control.
Furthermore, the activation of JNK and caspase-3 in Tak1⌬/⌬ keratinocytes occurred concomitantly with ROS accumulation. Therefore, we assume that ROS is the cause of these pro-cell death events. We examined the effect of inhibiting ROS with antioxidants on TNF-induced pro-cell death events. Cells were treated with the antioxidant BHA and the activation of JNK and caspase-3 and c-FLIP degradation examined ( Fig. 2A). BHA could abolish all these pro-cell death events. We also measured apoptosis and cell viability. BHA could completely inhibit an increase of apoptotic cells following TNF stimulation in Tak1⌬/⌬ keratinocytes (Fig. 2B). Furthermore, BHA and another antioxidant N-acetylcysteine could totally block TNFinduced cell death (Fig. 2C). These results suggest that ROS is the major mediator of TNF-induced cell death in Tak1⌬/⌬ keratinocytes.
NF-B Deficiency Does Not Effectively Induce ROS or Apoptosis in TNF-stimulated Keratinocytes-Our results thus far demonstrated that TAK1 deficiency causes ROS accumulation upon TNF stimulation, which results in apoptotic and necrotic cell death. One likely mechanism by which TAK1 deficiency causes ROS accumulation is a lack of activation of NF-B. It is known that NF-B regulates several enzymes and molecules that are involved in ROS metabolism such as manganese superoxide dismutase (MnSOD) and ferritin heavy chain (10,24). To examine the role of NF-B in TNF-induced cell death in keratinocytes, we generated wild-type keratinocytes stably expressing the repressor of NF-B, IB⌬N (IB⌬N keratinotyes). We used Tak1 wild-type keratinocytes isolated from the littermate mice of those we isolated from our Tak1⌬/⌬ keratinocytes. Therefore, the backgrounds of Tak1⌬/⌬ and IB⌬N keratinocytes are similar. We confirmed that TNFinduced NF-B activation was completely abolished in IB⌬N keratinocytes (Fig. 3A). We examined TNF-induced ROS accumulation, apoptosis, and cell death. Tak1⌬/⌬ keratinocytes exhibited a large increase of ROS at 5 h and ROS could not be measured due to cell death at 16 h after TNF stimulation (Fig. 3B). In contrast, ROS was not increased even at 16 h in IB⌬N keratinocytes (Fig. 3B). Annexin V-binding positive cells were also not increased in IB⌬N keratinocytes (Fig. 3C). These results indicate that TAK1 regulates ROS and cell death not through NF-B. In MEFs, MnSOD is one of the major targets of NF-B and regulates ROS (10). To further examine the involvement of NF-B in ROS regulation in keratinocytes, we observed the expression levels of MnSOD following TNF stimulation (Fig. 3D). A chemokine MIP2, which is a NF-B target, was greatly increased by TNF in Tak1 wild-type keratinocytes, indicating that NF-B was activated under this experimental condition. However, unlike MEFs, MnSOD was not increased by TNF in keratinocytes and TAK1 deficiency did not decrease the level of MnSOD. Collectively, we conclude that NF-B is not primarily important for ROS regulation at least in keratinocytes. Finally, we examined TNF-induced cell death at 24 h after stimulation. Cell death was moderately increased in IB⌬N keratinocytes, however, cell death was significantly less compared with that in Tak1⌬/⌬ keratinocytes (Fig. 3E). These results suggest that TAK1 deficiency causes an increase of ROS and cell death mainly not through the lack of NF-B activation. We next examined the relationship between TAK1 and ROS. If TAK1 regulates ROS, TAK1 deficiency should cause an increase of sensitivity not only to TNF but also to ROS-induced cell death. We treated kerati-

TAK1 Regulates ROS and Cell Death
nocytes with tBHP, a prototypical organic oxidant. Tak1⌬/⌬ keratinocytes died by 0.3 mM tBHP treatment (Fig. 3E, left  panel), whereas Tak1 wild-type and IB⌬N keratinocytes were resistant to 0.5 mM tBHA and died with 1.0 mM tBHP treatment (Fig. 3F, right panel). These results demonstrate that TAK1 but not NF-B affects sensitivity to ROS-induced cell death in keratinocytes, suggesting that TAK1 regulates ROS mainly through non-NF-B mechanisms.

c-Jun Is Partially Involved in TAK1 Regulation of ROS-We
next attempted to determine the mechanism by which TAK1 regulates ROS. TAK1 can activate two groups of transcription factors, namely NF-B and AP1. As described above, we found that ROS is regulated through non-NF-B mechanisms. Therefore, we examined the levels of AP1 family transcription factors including the Jun family and c-Fos. We also examined CREB. Among these, we found that the levels of c-Jun were reduced in TAK1-deficient keratinocytes under the basal condition (Fig. 4, A and B). The level of c-Jun was not significantly altered by TNF stimulation both in wildtype and Tak1⌬/⌬ keratinocytes. Thus, only the basal level of c-Jun is regulated by TAK1. TAK1 is activated by the number of stimuli including cytokines and stresses (13,25,26), which presumably are constitutively present at low levels even in unstimulated conditions. We speculate that TAK1 may be constitutively activated at a low level under the normal culture conditions, and this constitutive activity of TAK1 is important to maintain the basal level of c-Jun. We hypothesized that TAK1-dependent basal expression of c-Jun is involved in ROS regulation and cell survival. To examine this possibility, we generated Tak1⌬/⌬ keratinocytes that ectopically express c-Jun (Fig. 4C), and examined accumulation of ROS, apoptosis, and cell death following TNF stimulation.
We found that the levels of ROS were reduced by c-Jun expression in TNF-treated Tak1⌬/⌬ keratinocytes (Fig. 4D). Concomitantly, TNF-induced apoptosis was significantly reduced in c-Jun expressing keratinocytes (Fig. 4E). However, cell viability at 24 h after TNF stimulation was not greatly rescued by c-Jun expression (Fig. 4F). We note that higher levels of c-Jun expression did not increase the level of rescuing the TNF-induced cell death (data not shown). These results suggest that the TAK1-c-Jun pathway participates in ROS regulation but is not sufficient to rescue cell death at later time points.
Antioxidants Rescue the Skin Disorder in Mice Caused by Epidermal-specific Deletion of TAK1-Our results in cultured keratinocytes indicate that TAK1-dependent ROS regulation is important for cell survival. Finally, we asked whether ROS are the cause of keratinocyte death and subsequent skin inflammation in mice harboring an epidermal-specific deletion of TAK1. Mice with an epidermal-specific deletion of Tak1 were born at the expected Mendelian ratio but developed a severe skin inflammatory condition and were lethal by postnatal days 6 -7. Because small pups are difficult to effectively treat with antioxidants, we generated mice with epidermal-specific inducible Tak1 deletion. We crossed homozygous Tak1-floxed (Tak1 flox/flox ) mice to transgenic mice expressing the tamoxifen-dependent Cre-ERT recombinase under the control of the cytokeratin K14 promoter (K14-CreERT) (21). The resulting mice (K14-CreERT Tak1 flox/flox ) were born at the expected Mendelian ratio and grew normally, at least until the age of 6 -7 months. To induce deletion of the Tak1 gene, we topically treated the K14-CreERT Tak1 flox/flox mice with 4-hydroxytamoxifen (4-OHT), an active metabolite of tamoxifen. We confirmed that this treatment induced deletion of TAK1 by PCR, using a primer set that could only detect the deleted allele of Tak1 (Fig. 5B). The mutant mice harboring an epidermal-specific Tak1 deletion developed an inflammatory skin condition 2-5 days after treatment with 4-OHT. The inflammatory conditions were severe and similar to the conditions in the constitutive epidermal specific Tak1 deletion mice at postnatal days 5-7 (17). The mutant epidermis was hyperplastic, hypertophic, and hyperkeratotic, and microabscesses were commonly found (supplementary Fig. S1A). An inflamed skin marker K6 was highly expressed in the entire area where 4-OHT was treated (supplementary Fig. S1B). Expression of the epidermal differentiation marker loricrin was diminished and apoptotic keratinocytes are greatly increased all over the 4-OHT-treated skin (supplementary Fig. S1B). To examine whether ROS are the cause of this inflammation, we fed the mice food containing BHA. The mice consuming food containing BHA showed less flaky skin compared with the TAK1-deficient littermates who did not consume BHA (Fig. 5A). Histological analysis revealed that BHA decreased epidermal hyperplasia, infiltration of immune cells, and hyperkeratotic stratum corneum that are frequently observed in TAK1-deficient epidermis (Fig. 5C). Importantly, keratinocyte death was significantly reduced in BHAtreated mice (Fig. 5C). These results demonstrated that ROS are the cause of the skin inflammation in mice having TAK1deficient epidermis.

DISCUSSION
TNF induces ROS, activation of JNK, degradation of c-FLIP, and activation of caspases in Tak1⌬/⌬ keratinocytes, and all of these events cooperatively promote cell death. Because inhibition of ROS could abolish pro-cell death events and ultimately blocked cell death in Tak1⌬/⌬ keratinocytes, we surmise that accumulation of ROS in the TAK1-deficient keratinocytes is the central mediator of keratinocyte death.
How Does TAK1 Regulate ROS?-ROS are regulated by a number of enzymes including NF-B target gene products such as MnSOD and ferritin heavy chain (10,24,27). TAK1 is an essential intermediate in the TNF signaling pathway leading to activation of NF-B transcription factors (Fig. 1A). Therefore, it was anticipated that TAK1 would influence the expression of MnSOD and/or ferritin heavy chain. However, we found that deletion of TAK1 did not affect their expression ( Fig. 3D and data not shown). Moreover, we found that NF-B-deficiency did not induce ROS following TNF stimulation (Fig. 3B). Thus, NF-B does not regulate the level of ROS in keratinocytes, and the mechanisms mediated by other than NF-B should be important for ROS regulation. We found that TAK1-dependent c-Jun expression may partially contribute to ROS regulation in keratinocytes. However, we could not totally block either TNF-induced ROS accumulation or cell death by ectopic expression of c-Jun. mice from the same litter were fed with BHA-containing or regular food, and treated with 4-OHT for 5 consecutive days. The mice at 2 days after completion of 4-OHT treatment are shown. B, genomic DNA was isolated from skin (S), liver (L), and kidney (K). Tak1⌬ was analyzed by PCR. TNF receptor (TNFR) was used as a loading control. C, the Dorsal skin sections were stained with hematoxylin/eosin and apoptotic cells were detected with TUNEL staining (green). Nidogen was used as basement membrane marker (red) brackets indicate epidermis. Scale bars, 20 m. The results are representatives of three independent experiments with similar results. Therefore, TAK1 activates several mediators including c-Jun, and unidentified factors that may cooperatively regulate the levels of ROS in keratinocytes. We have recently found that intestinal epithelial specific Tak1 mutant mice die immediately after birth and they show massive epithelial cell death (19). In contrast, intestinal epithelial specific NF-B-deficient (NEMO deletion) mice slowly develop intestinal inflammation by 6 weeks of age (20). Our current results raise the possibility that Tak1 deletion increases ROS, thereby inducing cell death and tissue damage in intestinal epithelial specific Tak1 deletion mice more severely compared with the epithelial specific NEMO deletion mice.
TAK1, ROS, and Psoriasis-ROS has been associated with psoriasis; however, it has not been determined whether ROS is the cause or the consequence of inflammation (28). We have shown here that Tak1 deletion causes dysregulation of ROS in keratinocytes, which is causally associated with skin inflammation (Ref. 17 and Fig. 5). TAK1 can be activated by cytokines and innate immune stimuli including interleukin-1, TNF, and microbial components and stress conditions (13,25,26). The levels of those TAK1 stimuli in the epithelial tissues may be fluctuated in vivo. For example, reduction of commensal bacteria decreases TAK1 activity, thereby increasing the accumulation of ROS. Such dysregulation in TAK1 activity may contribute to ROS-mediated epithelial inflammation. Our results warrant further studies on the relationship of the TAK1 signaling pathway, ROS metabolism, and psoriasis.