TAK1 signaling maintains intestinal integrity by preventing accumulation of reactive oxygen species in the intestinal epithelium

Abstract

The intestinal epithelium is constantly exposed to inducers of reactive oxygen species (ROS) such as commensal microorganisms. Levels of ROS are normally maintained at non-toxic levels, but dysregulation of ROS is involved in intestinal inflammatory diseases. Here we report that TGFβ-activated kinase 1 (TAK1) is a key regulator of ROS in the intestinal epithelium. tak1 gene deletion in the mouse intestinal epithelium caused tissue damage involving enterocyte apoptosis, disruption of tight junctions and inflammation. Disruption of TNF signaling, which is a major intestinal damage inducer, rescued the inflammatory conditions but not apoptosis or disruption of tight junctions in the TAK1-deficient intestinal epithelium, suggesting that TNF is not a primary inducer of the damage noted in TAK1-deficient intestinal epithelium. We found that TAK1 deficiency resulted in reduced expression of several antioxidant responsive genes and reduced the protein level of a key antioxidant transcription factor NF-E2-related factor 2 (Nrf2), which resulted in accumulation of ROS. Exogenous antioxidant treatment could reduce apoptosis and disruption of tight junctions in the TAK1-deficient intestinal epithelium. Thus, TAK1 signaling regulates ROS through Nrf2, which is important for intestinal epithelial integrity.

Introduction

Integrity of the epithelial barrier is essential for preventing invasion of microorganisms and development of chronic inflammatory conditions in the intestine. A single layer of enterocytes separates the lamina propria from the gut lumen, thereby functioning as a physical barrier. Enterocytes are derived from intestinal epithelial stem cells which are localized to the crypts. Proliferating enterocytes differentiate and migrate toward the villus tips in the small intestine (1). Enterocytes undergo apoptosis only at the apical component of the villi. Although the intestine is constantly exposed to high levels of cell death inducers such as TNF and bacteria-derived stressors, enterocytes are resistant to those inducers and survive until they reach the tip of villus. Dysregulated apoptosis during the periods of proliferation and migration disrupts the intestinal barrier. In addition to enterocyte survival, tightly connected enterocyte-enterocyte junctions (tight junctions) are essential to form the physical barrier (2). Recently, it has become evident that commensal bacteria play a protective role in intestinal epithelial barrier function (3, 4). For example, depletion of commensal bacteria greatly increases sensitivity to stress-induced intestinal damage (4). Ablation of Toll-like receptor (TLR) signaling from commensal bacteria disrupts integrity of tight junctions in enterocytes (5). Therefore, commensal bacteria-derived cell signaling is likely to play a crucial role in cell survival and regulation of tight junctions in enterocytes. However the intracellular signaling pathways that maintain enterocyte survival and tight junctions remain elusive.

TGF®-activated kinase 1 (TAK1) plays an important role in several innate immune signaling pathways. TAK1 is activated by TLR ligands, the intracellular bacteria sensor NOD2 and cytokines such as TNF and IL-1 (6–8). The TAK1 signaling pathway leads to activation of two groups of transcription factors, AP-1 and NF-κB, which are intimately involved in immune responses in immune cells (9). We have recently reported that TAK1 deficiency results in dysregulation of cell survival in two types of epithelial cells: keratinocytes and enterocytes (10, 11). These findings raised the possibility that commensal bacteria-induced TAK1 signaling regulates enterocyte survival.

TNF is constitutively expressed in the intestine and is essential for preventing invasion of microorganisms (12). However, dysregulation in TNF signaling plays a major role in intestinal damage by inducing inflammation and apoptosis in inflammatory diseases such Crohn's disease, and inhibition of TNF signaling is one of the most effective approaches to prevent intestinal damage (13). TNF transcriptionally induces inflammatory genes through TAK1-NF-κB and TAK1-AP-1 pathways in immune cells (7, 14–16). TNF can activate caspase-dependent apoptosis through the Fas-associated death domain (FADD) and pro-caspase 8 (also called FLICE) (17). TNF also activates NADPH oxidase which generates reactive oxygen species (ROS) (18, 19). We recently reported that ablation of TAK1 causes hypersensitivity to TNF-induced apoptosis in keratinocytes (20). In the epidermal-specific TAK1 deletion mice, the tissue damage in the skin is largely rescued by deletion of the TNF receptor 1 (TNFR1) gene (11). However, in the intestinal epithelium, although deletion of TNFR1 greatly reduces intestinal damage in neonatal intestinal epithelial-specific TAK1 deletion mice, the mice spontaneously develop ileitis and colitis at approximately postnatal day 15 (10). Thus, ablation in enterocyte-derived TAK1 signaling results in both TNF-dependent and independent intestinal damage. In the present study, we investigated the mechanism by which TAK1 prevents TNF-dependent and - independent intestinal epithelial damage.

Materials and Methods

Mice

Mice carrying a floxed Map3k7 allele (TAK1^FL/FL) (7) were backcrossed to C57BL/6 for at least 5 generations. TNFR1-deficient C57BL/6 mice Tnfrsf1atm1Mak (TNFR1^−/−) (21) and villin-Cre transgenic mice (22) with a C57BL/6 background were from Jackson Laboratory. villin-CreER^T2 transgenic mice with a C57BL/6 background have been described previously (23). The backcrossed TAK1^FL/FL mice were used to generate villin-CreTAK1^FL/FL (TAK1^IE-KO), villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO) and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) mice. In all experiments, littermates were used as controls. To induce TAK1 gene deletion, 4-week-old mice were given intraperitoneal injections of tamoxifen (1 mg/20 g body weight) for 2–5 consecutive days. Some mice were fed with food containing 0.7% butylated hydroxyanisole (BHA) from 1 week prior to the tamoxifen treatment. The following primers were used for genotyping: floxed TAK1, CACCAGTGCTGGATTCTTTTTGAGGC and GGAACCCGTGGATAAGTGCACTTGAAT; villin-CreER^T2, CAAGCCTGGCTCGACGGCC and CGCGAACAT CTTCAGGTTCT; TNF receptor 1; TGTGAAAAGGGCACCTTTACGGC and GGCTGCAGTCCACGCACTGG. Mice were bred and maintained under specific pathogen-free conditions. All animal experiments were done with the approval of the North Carolina State University Institutional Animal Care and Use Committee.

shRNA and cell culture

Caco-2 cells were cultured in DMEM with 10% bovine growth serum (Hyclone), and penicillin-streptomycin at 37oC in 5% CO2. TAK1 siRNA target sequence corresponded to nucleotides 88–106 of the TAK1 cording region was used to generate a retrovirus vector expressing shRNA against TAK1, pSUPERRetro-puro-shTAK1 (24). Caco-2 cells were transfected with pSUPERRetro-puro vector or pSUPERRetro-puro-shTAK1, and selected with puromycin for 2 weeks. Pools of puromycin resistant Caco-2 cells were used. TAK1+/+ and TAK1Δ/Δ keratinocytes were isolated from TAK1^FL/FL and K5-Cre TAK1^FL/FL mice described previously (11). Spontaneously immortalized keratinocytes derived from the skin of postnatal day 0–2 mice were cultured in Ca2+-free minimal essential medium (Lonza) supplemented with 4% Chelex-treated bovine growth serum (Hyclone), 10 ng/ml human epidermal growth factor (Invitrogen), 0.05 mM calcium chloride, and penicillin-streptomycin at 33 °C in 8% CO2. Reagents used were butylated hydroxyanisole (BHA, Sigma), tert-butylhydroperoxide (tBHP, Sigma), MG-132 (Merck) and cycloheximide (Merck).

Histology and immunohistochemistry

Sections were stained with hematoxylin and eosin for histological analysis. Sections were scored in a blinded fashion on a scale from 0 to 4, based on the degree of lamina propria mononuclear cell infiltration, crypt hyperplasia, goblet cell depletion, and architectural distortion, as previously described (25). To detect apoptotic cells, the TUNEL assay was performed on paraffin sections using the DeadEnd™ Colorimetric TUNEL System (Promega) according to the manufacturer's instructions. Immunofluorescent staining was performed on paraffin-embedded sections or cryosections using polyclonal antibodies against claudin-3 (1:500, Zymed), occludin (1:50, Zymed), ZO-1 (1:50, Zymed) and cleaved-caspase 3 (1:100, Cell Signaling). Bound antibodies were visualized by Cy3 or Cy2-conjugated secondary antibodies against rabbit (1:500, GE healthcare). Nuclei were counterstained with DAPI. Images were visualized using a microscope (BX41; Olympus) controlled by the IPLab imaging software (Scanalytics).

Isolation of enterocytes

The small intestine was harvested and flushed with PBS to remove fecal contents. One end of the intestine was tied off, filled with Hanks' Balanced Salt Solution (HBSS, Sigma) containing 10mM EDTA and incubated in a PBS bath at 37°C for 5 min. After removing the contents, the intestine was filled with 10mM EDTA in HBSS and incubated in PBS again for 10 min. The contents were collected into tubes and centrifuged at 1,200 rpm for 5 min. The resulting pellets containing predominantly epithelial cells were washed twice in cold PBS.

Immunoblot analyses

Nuclear and cytoplasmic extracts from enterocytes and keratinocytes were prepared using a Nuclear Extract Kit (Active Motif). Proteins from cell lysates were electrophoresed on SDS-PAGE and transferred to Hybond-P (GE Healthcare). The membranes were immunoblotted with polyclonal antibodies against NF-E2-related factor 2 (Nrf2) (Santa Cruz), TAK1 (8), TAK1 binding protein 2 (TAB2) (26), and monoclonal antibodies against Lamin B₁ (Zymed). Bound antibodies were visualized with horseradish peroxidase-conjugated antibodies against rabbit or mouse IgG using the ECL western blotting system (GE).

Quantitative real-time PCR

Total RNA from the small intestine was isolated using RNeasy Mini (Qiagen). cDNA was synthesized using TaqMan reverse transcription reagents (Applied Biosystems). mRNA levels of NQO1 and β-actin were analyzed by real-time PCR with SYBR Green (Applied Biosystems). NQO1 primer, CATTCTGAAAGGCTGGTTTGA and CTAGCTTTGATCTGGTTGTCAG; GST-M1 primers, CTCCCGACTTTGACAGAAGC and CAGGAAGTCCCTCAGGTTTG; GST-A4 primers, GCCAAGTACCCTTGGTTGAA and AATCCTGACCACCTCAACA and β-actin primers, CCCAGAGCAAGAGAGGTATC and AGAGCATAGCCCTCGTAGAT were used. Expression levels of Bcl2, BclxL, glutamylcysteine ligase catalytic subunit (GCLC), IL-1®, IL-6, MIP2, Nrf2 and GAPDH were also analyzed by TaqMan gene expression assay (Applied Biosystems). Results were analyzed using the comparative Ct Method. Values were normalized to the level of β-actin mRNA in SYBR Green and to the level of GAPDH mRNA in TaqMan gene expression assays.

Transepithelial electrical resistance

The ileal tissues were harvested immediately after euthanasia, cut longitudinally, and placed on 0.12cm²-aperture Ussing chambers (27). Tissues were bathed on the serosal and mucosal sides with Ringer's solution. The serosal bathing solution contained 10 mM glucose, which was osmotically balanced on the mucosal side with 10 mM mannitol. Bathing solutions were oxygenated (95% O2–5% CO2) and circulated in water-jacketed reservoirs maintained at 37°C. The spontaneous potential difference (PD) was measured using Ringer-agar bridges connected to calomel electrodes, and the PD was short circuited through Ag-AgCl electrodes using a voltage clamp that corrected for fluid resistance. Transepithelial electrical resistance (TER; ·cm²) was calculated from the spontaneous PD and short-circuit current. The Ussing chamber experiments were run for up to 3-hours after initial equilibration period of 15-min. In each Ussing chamber experiment, duplicate tissues were studied from each animal.

Analysis of ROS production

Harvested small intestines were embedded and frozen in OCT compound and frozen sections were prepared. Sections were stained with 5μM CM-H₂DCFDA (Invitrogen) for 40 min at 37°. Images were taken using a fluorescent microscope (BX41, Olympus) controlled by IPlab (Scanalytics). 3–5 randomly selected areas were photographed with the same exposure time. The images were processed using the same fixed threshold in all samples by Photoshop software, and representative images are shown.

Statistical analyses

Statistical comparisons were made using independent, Student's t-tests on data with normal variance.

Results

TAK1 prevents both TNF-dependent and independent intestinal damage

Mice with intestinal epithelial-specific deletion of TAK1 (TAK1^IE-KO) has a lethal defect within 24–48 h after birth due to severe damage in the intestine (10). We have previously demonstrated that TNFR1 deletion rescues this early lethality (10), indicating that the severity of damage is mainly caused by TNF. However, mice having intestinal epithelial specific TAK1 deletion, even on a TNFR1^−/− background, develop ileitis and colitis at postnatal day 15–17 (10). This suggests that TAK1 not only prevents TNF-dependent damage but that TAK1 is also important for blockade of TNF-independent epithelial dysregulation. In this study, we aimed to determine the mechanism by which TAK1 prevents TNF-dependent and independent epithelial dysregulation. Enterocyte survival and tightly connected cell-cell junctions are essential for maintenance of intestinal epithelial integrity. Therefore, we first examined apoptosis and tight junctions in control and TAK1-deficient intestinal epithelium on a wild type or TNFR1^−/− background. In order to compare the level of damage at the same age in the mouse model, we have generated intestinal epithelial-specific inducible TAK1 deletion mice on a wild type background (TAK1^IE-IKO) and TNFR1^−/− background (TAK1^IE-IKO TNFR1^−/−). In this system, we induce tak1 gene deletion by tamoxifen injection. We used 4-week-old mice for all experiments and initially analyzed the small intestine at day 3 of tamoxifen injection. TAK1 deletion caused epithelial damage, including extensive evidence of apoptosis in the crypts (Fig. 1a and b). The TAK1-deficient epithelium also exhibited extensive separation of villus epithelium from the lamina propria throughout the upper two thirds of the affected villi. This damage was observed regardless of TNFR1 status in the ileum (Fig. 1a and b). The numbers of apoptotic enterocytes were similar in the TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− ileum epithelium. These results indicate that apoptosis was induced mainly through a TNF-independent mechanism in the TAK1-deficient ileum. We noted that TAK1^IE-IKO mice exhibited damage in both the small intestine and colon, whereas damage in TAK1^IE-IKO TNFR1^−/− mice was observed primarily in the ileum (Supplementary Fig. S1). As reported previously, TAK1^IE-IKO mice develop severe damage, which becomes lethal at 4–5 days after the initiation of tamoxifen injection (10). In contrast, the level of damage in TAK1^IE-IKO TNFR1^−/− intestinal epithelium was not changed following 3 days of tamoxifen injection. We injected TAK1^IE-IKO TNFR1^−/− mice with tamoxifen for 5 consecutive days and maintained them without further tamoxifen injection. We confirmed that the tak1 gene was deleted at 3 days and at 8 weeks after the termination of tamoxifen injection. TAK1^IE-IKO TNFR1^−/− mice were viable at least 6 months without showing any clinical signs but exhibited tissue damage in the ileum at similar levels to those at 3 days after initiation of tamoxifen injection. These results indicate that TAK1 prevents TNF-independent apoptosis in the ileum and that TNF is not a primary mediator of intestinal damage but amplifies the damage.

Figure 1. TAK1 deletion causes TNF-dependent and independent intestinal damage.

(a) Hematoxylin and eosin (H&E) staining of ileal sections from littermate control (TAK1^F/F, CT) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO) or littermate control TNFR1^−/− (TNFR1^−/−) and villin-CreER^T2TAK1^FL/FL TNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) mice. 4-week-old mice were treated with tamoxifen for 2 consecutive days and samples were prepared one day after the 2nd injection (day 3). Scale bars, 50 μm. Arrows indicate examples of apoptotic enterocytes.

(b) TUNEL staining of (a) sections. Scale bars, 20 μm. TUNEL positive cells were counted in at least 180 crypts of each ileum. The numbers shown are means of TUNEL positive cells per crypt and SEM; n=3.

(c) Immunofluorescence staining of claudin-3 using the sections prepared as described in (a). Scale bars, 20 μm.

(d) Real-time PCR analysis of TAK1^FL/FL (CT, open bars) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO, filled bars) or TNFR1^−/− (TNFR1^−/−, open bars) and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−, filled bars) small intestine. 4-week-old mice were treated with tamoxifen for 2 consecutive days and mRNA samples were prepared one day after the 2nd injection (day 3). mRNA levels relative to GAPDH mRNA are shown. The data represent means ± SEM; n=7. *, p<0.05; ns, not significant.

We examined whether the tight junctions were also affected by ablation of TAK1. We analyzed the localization of three major tight junction-associated proteins in enterocytes: claudin-3, occludin and tight junction plaque protein ZO-1. All three proteins were diffusely localized in the ileum of both TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− mice (Supplementary Fig. S2 and data not shown). While localization of occludin and ZO-1 was marginally altered, the mislocalization of claudin-3 was striking in TAK1-deficient intestinal epithelium (Fig. 1c) in that there was very little evidence of claudin-3 precisely at the region of the tight junction apical lateral membrane, but rather appeared to be within the cytoplasm. The level of claudin-3 mislocalization was not notably different between TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− mice.

To assess the levels of inflammation, we measured the mRNA levels of inflammatory cytokines that were expressed in the small intestine in TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− at day 3 of tamoxifen injection (Fig. 1d). TAK1^IE-IKO mice had markedly increased levels of inflammatory cytokines, whereas TAK1^IE-IKO TNFR1^−/− mice did not significantly increase those cytokines. Collectively, TAK1 appears to be essential for enterocyte survival and integrity of tight junctions, primarily in the ileum. This effect is independent of TNF signaling. When TNF signaling is intact, TAK1 deletion causes more severe tissue damage not only in the ileum but also in other regions of the intestine. These results suggest that TAK1 is primarily important for maintenance of enterocyte survival and tight junction integrity, and that TNF signaling amplifies TAK1-deficiency-induced damage, by promoting inflammation.

TAK1 deficiency reduces transepithelial electrical resistance

To verify whether the increased apoptosis and disruption of tight junction impairs intestinal barrier function, we measured transepithelial electrical resistance (TER) in TAK1-deficient intestinal epithelium. We chose TNFR1−/− background (TAK1^IE-IKO TNFR1^−/−) mice to rule out the possibility that inflammatory conditions could indirectly affect the barrier function. TER was significantly low in TAK1^IE-IKO TNFR1^−/− mice compared to control TNFR1^−/− mice (Fig. 2). Thus, TAK1 is important for maintenance of the intestinal barrier function through modulating enterocyte survival and tight junctions.

Figure 2. TAK1 deficiency impairs intestinal barrier.

TNFR1^−/− (TNFR1^−/−, open circle) and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−, filled circle) mice were treated with tamoxifen for 5 consecutive days. At 2 weeks post-tamoxifen treatment, the ileum was isolated and analyzed transepithelial electrical resistance by using Ussing chamber.

TAK1 deficiency downregulates the levels of antioxidant responsive genes

To determine the mechanism by which TAK1 mediates enterocyte survival, we measured the mRNA levels of genes associated with cell survival in control and TAK1-deficient intestine. We initially analyzed samples from the intestine having control genotype and intestinal epithelial-specific constitutive deletion of TAK1 (TAK1^IE-KO). The expression levels of anti-apoptotic genes, including Bcl2 and BclxL, were not altered (Supplementary Fig. S3). We found that several antioxidant responsive genes, namely NAD(P)H:quinone oxidoreductase 1 (NQO1), glutathione S-transferases (GST) M1 and A4, were downregulated in the constitutive and inducible TAK1-deficient small intestine and the colon (Fig. 3a). These antioxidant responsive genes are known to be regulated by a key antioxidant transcription factor, Nrf2. Therefore, we examined the levels of Nrf2 in inducible TAK1-deficient intestinal epithelium both on wild type and TNFR1^−/− background (TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/−) (Fig. 3b and c). While the mRNA levels of Nrf2 were not significantly altered by ablation of TAK1, the protein level of nuclear Nrf2 was greatly reduced in the TAK1-deficient intestine. Nrf2 was not detectable in the cytoplasmic fraction in intestinal epithelium (data not shown). Downregulation of Nrf2 was independent of TNFR1 status. Nrf2 regulation of antioxidant responsive genes plays an integral role in ROS metabolism (28, 29), which is critically involved in cell viability and tight junction integrity. We postulated that TAK1 might regulate the protein level of Nrf2, thereby modulating cell survival and tight junctions.

Figure 3. Antioxidant responsive genes are downregulated in TAK1-deficient intestinal epithelium.

(a) Real-time PCR analysis to quantify GST-M1, GST-A4 and NQO1 mRNA levels in the TAK1^FL/FL (CT) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO) small intestine and colon. The mice were treated with tamoxifen for 3 consecutive days and mRNA samples were prepared one day after the 3rd tamoxifen injection (day 4). Relative mRNA levels were determined using β-actin mRNA. The data represent means ± SEM; n=4. *, p<0.05; ** P<0.01; ns, not significant.

(b) (left) Nrf2 mRNA level were determined by real-time PCR analysis in the TAK1^FL/FL (CT, open bars) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO, filled bars) small intestine. The mice were treated with tamoxifen for 2 consecutive days and mRNA samples were prepared one day after the 2nd tamoxifen injection (day 3). Relative mRNA levels were calculated using GAPDH mRNA. The data represent means ± SEM; TAK1^FL/FL, n=7. ns, not significant. Immunoblot analysis of Nrf2 in the nuclear extracts (Nuc) from TAK1^FL/FL (CT) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO) enterocytes. The enterocytes were prepared at day 3 of tamoxifen injection. Lamin B1 was used as a loading control.

(c) Nrf2 mRNA and protein levels of TNFR1^−/− and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) small intestine were analyzed as described in (b). The mRNA data are means ± SEM; n=7. ns, not significant.

Ablation of TAK1 downregulates Nrf2 and sensitizes cells to oxidative stress

To further investigate the mechanism by which TAK1 regulates Nrf2 and cell survival, we utilized two lines of cultured epithelial cells having TAK1 ablation. One is Caco-2 cells stably expressing an shRNA targeted against TAK1; and the other is TAK1-deficient skin epithelial cells (keratinocytes) that were isolated from the epidermal-specific TAK1 deletion mice (11). In both cells, Nrf2 was not detectable in the cytoplasmic fractions (data not shown), and the protein level of nuclear Nrf2 was lower in TAK1-deficient cells compared to control cells (Fig. 4a and b). The protein level of Nrf2 is known to be primarily regulated by protein degradation through the proteasome pathway (30). Blockade of proteasome pathway by MG-132 treatment could greatly increase the levels of Nrf2 in both control and TAK1-deficient cells (Fig. 4a and b). These suggest that Nrf2 is always highly degraded through the proteasome pathway, and that TAK1 might be in part involved in Nrf2 stability. We asked whether activation of TAK1 could alter Nrf2 stability. Co-expression of TAK1 together with TAK1 binding protein 1 (TAB1) highly activates TAK1 (31). We determined the Nrf2 stability with and without co-expression of TAK1 and TAB1 in 293 cells (Fig. 4c). The protein level of Nrf2 was almost completely diminished within 5 h after blockade of protein synthesis when Nrf2 alone was expressed, whereas Nrf2 was much more slowly decreased in cells with co-expression of TAK1 and TAB1. These results suggest that TAK1 may participate in Nrf2 stability. Therefore, ablation of TAK1 might cause increased degradation of Nrf2.

Figure 4. TAK1 deficiency downregulates Nrf2 and cause hypersensitivity to oxidative stress in cultured epithelial cells.

(a) Caco-2 cells stably expressing a control vector or shRNA targeted against TAK1 were untreated or treated with10 μM MG-132 for 5 h. The nuclear and cytosolic fractions were prepared from those cells. The levels of Nrf2 were analyzed by immunoblots in the nuclear fraction (Nuc) and the levels of TAK1 was analyzed in the cytosolic fraction (Cyto). Lamin B was used as a loading control.

(b) TAK1 wild type (+/+) and deficient (Δ/Δ) keratinocytes left untreated or were treated with 10 μM MG-132 for 5 h. The nuclear and cytosolic fractions were analyzed as described in (a). TAK1 Δ/Δ keratinocytes expressed a truncated from of TAK1 (TAK1 Δ) and the expression level of TAK1 Δ was low compared to TAK1 wild type due to instability of the truncated form.

(c) 293 cells were transfected with expression vectors for 3 g Nrf2 or 0.33 μg Nrf2, 0.33 μg TAK1 and 0.33 μg TAB1. At 48 h post-transfection, cells were treated with 100 μg/ml cycloheximide (CHX) and harvested at 0, 2.5 and 5 h. Whole cell extracts were analyzed by immunoblots. The amount of endogenous TAB2 are also shown as a loading control.

(d) (left) Caco-2 cells stably expressing a control vector or shRNA targeted against TAK1 were treated with 1 mM tBHP for 15 h. Cells were analyzed by immunofluorescence staining with anti-ZO-1. (right) TAK1 wild type (+/+) and deficient (Δ/Δ) keratinocytes were treated with 0.3 mM tBHP for 24 h. Cells were analyzed by immunofluorescence staining with anti-ZO-1. Scale bars, 40 μm.

(e) (upper panels) Caco-2 cells stably expressing a control vector or shTAK1 and TAK1 wild type (+/+) and deficient (Δ/Δ) keratinocytes were treated as described in (c). Cells were analyzed by immunofluorescence staining with cleaved-caspase 3. Cleaved-caspase 3 positive cells were counted in at least 300 cells of each sample. The data represent means ± SEM. n=5. **, p<0.01; ***, P<0.001. (lower panels) Representative images of TAK1 wild type (+/+) and deficient (Δ/Δ) keratinocytes with cleaved-caspase 3 staining (red). DAPI staining (blue) was used to visualize nuclei. Scale bars, 40 μm.

Nrf2 is important for preventing oxidative stress (29, 32). We next examined whether TAK1 deficiency could cause increased apoptosis and disruption of tight junctions in response to oxidative stress. We treated Caco-2 and keratinocytes with tert-butylhydroperoxide (tBHP), a prototypical organic oxidant, and the tight junctions and apoptotic cells were observed by immunostaining with anti-cleaved caspase 3 and anti-ZO-1, respectively (Fig. 4d and e). We noted here that claudin-3 was less clearly detected in Caco-2 cells compared to ZO-1, and we used ZO-1 to visualize the tight junctions in those cultured cells. The tight junctions were relatively more damaged and apoptotic cells were significantly increased in TAK1-deficient cells compared to control cells. These results indicate that TAK1-deficiency caused hyper-sensitive to oxidative stress in cultured epithelial cells. Collectively, we postulated that ablation of TAK1 causes dysregulation of Nrf2 stability and ROS, which results in impaired tight junctions and increased apoptosis in the intestinal epithelium.

ROS is the cause of both TNF-dependent and independent intestinal damage in TAK1-deficient intestinal epithelium

We examined the hypothesis that TAK1 regulates ROS levels in the intestinal epithelium. ROS were measured in the TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− intestinal epithelium at day 3 of tamoxifen injection. The unfixed fresh cryosections of the ileum were used to detect ROS by CM-H₂DCFDA staining (Fig. 5a). The levels of ROS were greatly increased in TAK1-deficient intestinal epithelium. The increased ROS might be generated from infiltrated myeloid cells, because TAK1^IE-IKO intestinal epithelium was highly inflamed as shown in Fig. 1. However, the levels of ROS were not different between highly inflamed TAK1^IE-IKO mice and TAK1^IE-IKO TNFR1^−/− mice which did not exhibit significant inflammatory conditions. We detected some Gr-1 positive cells in both control TNFR1^−/− and TAK1-deficient TAK1^IE-IKO TNFR1^−/− intestinal epithelium; however, the number of Gr-1 positive cells were greatly smaller than that of ROS positive cells in TAK1^IE-IKO TNFR1^−/− intestinal epithelium (Supplementary Fig. S4). We found that the ROS positive cells were highly overlapped with the cells having cleaved-caspase 3 (Fig. 5b), suggesting that ROS are produced in apoptotic intestinal epithelial cells. These results indicate that TAK1 signaling is essential for preventing ROS accumulation in the intestinal epithelial cells. We next attempted to reduce ROS and tested whether reduction of ROS could rescue the apoptosis and tight junction disruption in the TAK1-deficient epithelium. We fed the mice with a chow diet containing the antioxidant butylated hydroxyanisole BHA for 1 week prior to tak1 gene deletion. We found that the levels of ROS were greatly reduced by BHA feeding in both TAK1^IE-IKO and TAK1^IE-IKO TNFR1^−/− mice (Fig. 5a). Histological evaluation and TUNEL assays revealed that intestinal damage and apoptotic enterocytes were significantly reduced in BHA feeding (Fig. 6). BHA feeding was equally effective in TAK1-deficient intestinal epithelium on either a TNFR1^+/+ or TNFR1^−/− background. Furthermore, BHA feeding greatly improved tight junction integrity (Fig. 7a). The mRNA levels of inflammatory cytokines were not upregulated in BHA treated TAK1^IE-IKO mice (Fig. 7b). These results indicate that ablation of TAK1 causes enterocyte apoptosis and impairs barrier function most likely due to increased ROS. .

Figure 5. TAK1 regulates ROS in the intestinal epithelium.

(a) (Upper panels) TAK1^FL/FL (CT) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO) mice were fed with normal or BHA chow diet for 1 week and subsequently treated with tamoxifen for 2 consecutive days. Unfixed cryosections were prepared from the ileum one day after the 2nd injection (day 3) and were incubated with 5 μM CM-H₂DCFDA for 40 min at 37°C to detect ROS (green). (Lower 2 panels) CM-H₂DCFDA staining of TNFR1^−/− and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) samples prepared by the same procedure as above. Scale bars, 20 μm.

(b) Unfixed cryosections were prepared from the ileum of villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) mice and stained by CM-H₂DCFDA (left panel), and the sections were subsequently fixed and cleaved-caspase 3 was detected by immunofluorescent staining (middle panel). Scale bars, 40 μm.

All images were photographed with the same exposure time. The data shown are representatives of 3–4 separate experiments with similar results.

Figure 6. Antioxidant prevents both TNF-dependent and independent intestinal damage in the TAK1-deficient intestinal epithelium.

(a) Hematoxylin and eosin (H&E) staining of the ileal sections from control (TAK1^F/F, CT), villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO), TNFR1^−/− and villin-CreER^T2TAK1^FL/FLTNFR1^−/−(TAK1^IE-IKO TNFR1^−/−) mice fed with normal or BHA chow diet. The mice were fed with a normal or BHA chow diet for 1 week and subsequently treated with tamoxifen for 2 consecutive days. Sections were prepared form the ileum one day after the 2nd injection (day 3). Scale bars, 50 μm. Arrows indicate examples of apoptotic enterocytes.

(b) Histological scores of (a). Data show the means ± SEM; CT, TAK1^IE-IKO and TAK1^IE-IKO +BHA, n=3; TNFR1^−/−, TAK1^IE-IKO TNFR1^−/− and TAK1^IE-IKO TNFR1^−/− +BHA, n=4. ***, P<0.001

(c) TUNEL staining of the sections shown in panel (a).

(d) TUNEL positive cells were counted in at least 180 crypts of each small intestinal segment. The numbers shown are means of TUNEL positive cells per crypt ± SEM; n=4. *, P<0.05; **, P<0.01

Figure 7. BHA feeding blocks disruption of tight junctions and inflammation.

(a) Immunofluorescence analysis of claudin-3 of the ileum sections from control (TAK1^F/F, CT), villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO), TNFR1^−/− and villin-CreER^T2TAK1^FL/FLTNFR1^−/− (TAK1^IE-IKO TNFR1^−/−) mice fed with normal or BHA chow diet. Scale bars, 20 μm.

(b) Real-time PCR analysis was performed in the small intestine from control (TAK1^F/F, CT) and villin-CreER^T2TAK1^FL/FL (TAK1^IE-IKO), mice with normal or BHA chow diet. The mice were treated with the procedure described in (a). mRNA levels relative to GAPDH mRNA are shown. The data are means ± SEM; CT, n=7; TAK1^IE-IKO, n=7, CT +BHA, n=4; TAK1^IE-IKO +BHA, n=7. ns, not significant.

Discussion

In this study, we demonstrated that enterocyte-derived TAK1 signaling plays a critical role in ROS metabolism, possibly through transcription factor Nrf2 and its target genes. Ablation of this TAK1 pathway caused accumulation of ROS resulting in enterocyte apoptosis and disruption of tight junctions. We have previously reported that deletion of TNFR1 can rescue TAK1 deficiency-induced apoptosis and inflammatory conditions in the intestinal epithelium in neonatal mice (10) and in the epidermis of the skin (11). In cultured cells, we have demonstrated that TNF greatly increases ROS in TAK1-deficient keratinocytes, which causes TNF-induced apoptosis (11, 20). Thus, we concluded that TAK1 signaling principally reduces TNF-induced ROS and prevents TNF-induced apoptosis in the intestine of neonatal mice and epidermis of the skin. However, in the adult intestinal epithelium, we found that the TAK1 deficiency-induced ROS was not altered by TNFR1 deletion (Fig. 5a). This indicates that TNF is not the major inducer of ROS in the adult intestinal epithelium. Mice are sterile in utero and are inoculated with bacteria at birth, and populations of intestinal commensal bacteria species are known to be dramatically altered during postnatal development (33). We speculate that commensal bacteria in the adult intestines may be the major trigger of ROS. TAK1 signaling reduces those non-TNF-induced ROS, which is essential for enterocyte survival and integrity of tight junctions.

How does TAK1 regulate ROS? In this study, we show that the level of Nrf2 was down regulated in TAK1-deficient intestinal epithelium. Although Nrf2 knockout increases susceptibility to intestinal injury, it alone does not increase enterocyte apoptosis or cause inflammatory conditions (34). Therefore, we would not expect that ablation of Nrf2 alone would be sufficient to induce all of the noted disruptions caused by intestinal epithelial-specific deletion of TAK1. In our previous study, we found that an AP-1 family transcription factor c-Jun is down regulated in TAK1-deficient keratinocytes (20). Similar to Nrf2, AP-1 family transcription factors are critical to the transcriptional regulation of antioxidant responsive genes (35–38). Over expression of c-Jun partially blocks accumulation of TNF-induced ROS (20). In addition, TAK1 is an integral upstream kinase of IκB kinases leading to activation of transcription factor NF-κB (11, 39), which is also an major transcription factor for several cellular antioxidant genes (40, 41). Taken together, we believe that TAK1 signaling regulates multiple antioxidant transcription factors, including Nrf2, c-Jun and possibly other unidentified factors which modulate the level of ROS.

We showed that TAK1 deficiency downregulates the protein levels of Nrf2 in the nucleus. Nrf2 is normally localized in the cytoplasm by its binding partner Keap1 and constantly degraded through the proteasome pathway (28). In the intestinal epithelium and cultured Caco-2 and keratinocytes, Nrf2 was not detectable in the cytoplasmic fraction. Antioxidants and oxidative stress oxidize Keap1 which results in the release of Nrf2. Dissociation from Keap1 stabilizes and translocates Nrf2 into the nucleus (28). We show that TAK1 regulates Nrf2 stability; therefore ablation of TAK1 downregulates the level of Nrf2. TAK1 signaling is likely to modulate Nrf2 or Keap1 and blocks Keap1-dependent Nrf2 degradation. Further studies will be needed to define the mechanism by which TAK1 modulates the Nrf2-Keap1 complex.

Intestinal epithelial-specific deletion of TAK1 causes increased apoptosis and disruption of cell-cell tight junctions primarily in the ileum. Those pathological conditions are very much similar to the pathology noted in inflammatory bowel disease (IBD). Anti-TNF therapy has recently been extensively used for the effective treatment of IBD (13). In the TAK1-deficient intestinal epithelium, while the deletion of TAK1 causes sustained severe intestinal damage, additional gene deletion of TNFR1 greatly reduces the inflammatory conditions, enabling the mice to survive. Thus, the effects of TNF down regulation have some similarities between IBD and the mouse model with intestinal epithelial-specific deletion of TAK1. In the intestinal epithelial-specific TAK1 deletion mouse model, TNFR1 deletion did not block increased ROS, apoptosis or disruption of tight junctions. In IBD patients, although anti-TNF therapy is effective in the treatment of select IBD patients, it does not block all of the associated pathologic conditions. Our results raise the possibility that upregulation of the TAK1-Nrf2 pathway could reduce the level of ROS and enhance enterocyte survival and integrity of tight junctions. The TAK1 pathway may be a novel target involved in regulating intestinal barrier function.