Transforming Growth Factor–β Signaling in Hepatocytes Promotes Hepatic Fibrosis and Carcinogenesis in Mice With Hepatocyte-Specific Deletion of TAK1

Abstract

BACKGROUND & AIMS

Transforming growth factor (TGF)-β–activated kinase 1 (TAK1) is activated in different cytokine signaling pathways. Deletion of Tak1 from hepatocytes results in spontaneous development of hepatocellular carcinoma (HCC), liver inflammation, and fibrosis. TGF-β activates TAK1 and Smad signaling, which regulate cell death, proliferation, and carcinogenesis. However, it is not clear whether TGF-β signaling in hepatocytes, via TGF-β receptor–2 (Tgfbr2), promotes HCC and liver fibrosis.

METHODS

We generated mice with hepatocyte-specific deletion of Tak1 (Tak1ΔHep), as well as Tak1/Tgfbr2DHep and Tak1/Smad4ΔHep mice. Tak1flox/flox, Tgfbr2ΔHep, and Smad4ΔHep mice were used as controls, respectively. We assessed development of liver injury, inflammation, fibrosis, and HCC. Primary hepatocytes isolated from these mice were used to assess TGF-β–mediated signaling.

RESULTS

Levels of TGF-β, TGF-βR2, and phospho-Smad2/3 were increased in HCCs from Tak1ΔHep mice, which developed liver fibrosis and inflammation by 1 month and HCC by 9 months. However, Tak1/Tgfbr2ΔHep mice did not have this phenotype, and their hepatocytes did not undergo spontaneous cell death or compensatory proliferation. Hepatocytes from Tak1ΔHep mice incubated with TGF-β did not activate p38, c-Jun N-terminal kinase, or nuclear factor-κB; conversely, TGF-β–mediated cell death and phosphorylation of Smad2/3 were increased, compared with control hepatocytes. Blocking the Smad pathway inhibited TGF-β–mediated death of Tak1−/− hepatocytes. Accordingly, disruption of Smad4 reduced the spontaneous liver injury, inflammation, fibrosis, and HCC that develops in Tak1ΔHep mice. Levels of the anti-apoptotic protein Bcl-xL, β-catenin, connective tissue growth factor, and vascular endothelial growth factor were increased in HCC from Tak1ΔHep mice, but not in HCCs from Tak1/Tgfbr2ΔHep mice. Injection of N-nitrosodiethylamine induced HCC formation in wild-type mice, but less in Tgfbr2ΔHep mice.

CONCLUSIONS

TGF-β promotes development of HCC in Tak1ΔHep mice by inducing hepatocyte apoptosis and compensatory proliferation during early phases of tumorigenesis, and inducing expression of anti-apoptotic, pro-oncogenic, and angiogenic factors during tumor progression.

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related deaths in the United States.¹ The increasing incidence of HCC is a result of the increased prevalence of hepatitis C virus infection, alcoholic cirrhosis, and nonalcoholic steatohepatitis. In humans, development of HCC is associated with chronic liver inflammation, fibrosis, and its subsequent irreversible cirrhosis.¹ In contrast, a mouse model for liver fibrosis induced by carbon tetrachloride does not induce HCC, and a mouse model for carcinogen-induced HCC is not accompanied by fibrosis. The current lack of an ideal animal model for liver fibrosis–associated HCC limits the mechanistic study of HCC.² We and others have recently established a mouse model that spontaneously develops HCC accompanied by liver fibrosis by deleting Tak1 in hepatocytes, indicating that TAK1 is a tumor suppressor in the liver.^3,4

Transforming growth factor-β (TGF-β)–activated kinase 1 (TAK1) is activated in various cytokine signaling, such as tumor necrosis factor (TNF), interleukin (IL)-1, Toll-like receptors, and transforming growth factor (TGF)–β. TAK1 then activates downstream kinases such as IκB kinase (IKK), c-Jun-N-terminal kinase (JNK), and p38 that regulate cell survival, proliferation, and tumor growth.^5,6 TNF receptor signaling activates both TAK1-IKK-nuclear factor–κB (NF-κB) pathway and caspase-mediated cell death pathway.⁷ Inactivation of TAK1 increases the susceptibility of hepatocytes to TNF-induced cell death through overactivation of caspases caused by lack of the TAK1-IKK-NF-κB activation.³ Notably, deletion of TNF receptor prevented hepatocyte death in hepatocyte-specific Tak1-deleted (Tak1^ΔHep) mice, indicating that TNF receptor signaling is responsible for spontaneous liver injury and inflammation in Tak1^ΔHep mice.³

TGF-β signaling activates Smad-dependent canonical and Smad-independent noncanonical TAK1-mediated signal pathways that activate JNK and p38.^8–10 TGF-β signaling regulates cell apoptosis, proliferation, differentiation, and extracellular matrix production. The importance of TGF-β signaling in cancer development has been demonstrated in several types of human cancers, including colon cancer, prostate cancer, and breast cancer.¹¹ In HCC, the functions of TGF-β signaling are paradoxical. In human HCC, persistent up-regulation of TGF-β was observed, and several reports indicate TGF-β signaling functions as a tumor suppressor. Alternatively, TGF-β signaling promotes the growth and migration of HCC through the induction of connective tissue growth factor (CTGF).^9,12,13 It is unclear whether TGF-β signaling promotes or suppresses HCC development in Tak1^ΔHep mice. In the present study, we investigated the role of TGF-β signaling in hepatocyte death, liver inflammation, fibrosis, and HCC spontaneously developed in Tak1^ΔHep mice. We found that TGF-β signaling promoted HCC development through induction of hepatocyte death and its subsequent compensatory proliferation in early phase. Importantly, in HCC of Tak1ΔHep mice, TGF-β signaling is associated with overexpression of anti-apoptotic Bcl-xL, procarcinogenic β-catenin, and CTGF, which prevent tumor apoptosis and promote cancer growth. In addition, TGF-β signaling also contributes to angiogenesis through vascular endothelial growth factor (VEGF) expression to promote HCC growth in Tak1^ΔHep mice.

Materials and Methods

Mouse Colonies

Albumin-Cre recombinase transgenic mice (Alb-Cre Tg mice), TGF-β receptor–2 (Tgfbr2)^flox/flox mice, and Smad4^flox/flox mice were purchased from Jackson Laboratories (Bar Harbor, ME). Mice carrying the floxed allele of Tak1 (Tak1^flox/flox mice) have been described previously.³ These mouse lines were crossed to generate the following types of mice with genes specifically deleted in hepatocytes: Alb-cre/+Tak1^flox/flox mice (Tak1^ΔHep), Alb-Cre/+Tgfbr2^flox/flox mice (Tgfbr2^ΔHep), Alb-cre/+Tak1^flox/flox Tgfbr2^flox/flox (Tak1/Tgfbr2^ΔHep), and Alb-cre/+ Tak1^flox/floxSmad4^flox/flox (Tak1/Smad4^ΔHep) mice on C57BL6 background. Both males and females were used in the study and Cre-negative animals were used as WT controls. In our previous report, we demonstrated that Tak1^ΔHep mice spontaneously develop liver injury, inflammation, fibrosis, and tumors without sex disparity.^3,4 One-month-old mice were used for assessing spontaneous liver injury, inflammation, and fibrosis, and 9-month-old mice were used for examining spontaneous liver fibrosis and HCC. In some experiments (Figure 7E), HCC was induced chemically via intraperitoneal injection of 25 mg/kg n-nitrosodiethylamine (DEN) (Sigma–Aldrich, St Louis, MO) into 14-day-old male Tgfbr2^flox/flox and Tgfbr2^ΔHep mice.¹⁴ Nine months after DEN injection, the chemically induced HCC was evaluated. All mice received humane care according to the National Institutes of Health recommendations outlined in their Guide for the Care and Use of Laboratory Animals. All animal experiments were approved by the University of California San Diego Institutional Animal Care and Use Committee.

Figure 7.

Oncogenic gene expression in HCC from Tak1^ΔHep mice, and TGF-β signaling in DEN-induced murine HCC and in human HCC. (A–D) Liver tissues in 9-month-old WT and Tgfbr2^ΔHep mice, and nontumor liver tissues and tumor tissues in 9-month-old Tak1^ΔHep, Tak1/Tgfbr2^ΔHep were harvested. (n=7, each samples) (A) Hepatic messenger RNA expression of pro-oncogenes (Ctnnb1, Myc, Yap1, and Wisp1) was determined by quantitative real-time polymerase chain reaction (qPCR). (B) Hepatic messenger RNA expression of Bax and Bcl-xl was assessed by qPCR, and protein expression of Bcl-2 and Bcl-xL are shown by immunoblotting. (C, D) Hepatic messenger RNA expressions of Ctgf (C), Vegfa, and Vegfr2 (D, upper) were assessed by qPCR. (D, lower) Immunostaining for VEGFa (upper), VEGFR2, and CD31 (lower) are shown. Original magnification 320×. (E) Diethylnitrosamine was injected (25 mg/kg) in 14-day-old WT and Tgfbr2^ΔHep mice and their livers were harvested at 9 months after DEN injections. Representative macroscopic pictures (left). The number of tumors per mouse was counted and the maximum diameter of individual tumor nodules was measured (right). (WT, n = 16; Tgfbr2^ΔHep, n = 16). NT, nontumor liver, T, tumors. (F) Immunohistochemistry for phospho-Smad2/3 in liver biopsy samples from patients with chronic hepatitis C (n = 4) and in liver tissues from patients with HCC (n = 4). Data are represented as mean ± standard error of mean. *P < .05; **P < .01.

Human Liver and HCC Samples

Paraffin-embedded human liver tissues were acquired from liver biopsy samples of patients with chronic hepatitis C and from explanted or surgically resected livers of patients with HCC. All patients provided written informed consent, and the study was approved by the University of California San Diego Institutional Review Board.

Other Materials and Methods

Other materials and methods for mouse tissue processing, histologic examination, Western blot analysis, quantitative real-time polymerase chain reaction, cell isolation, and treatment are described in the Supplementary Materials and Methods.

Statistical Analysis

Differences between 2 groups were compared using the Mann-Whitney U test or 2-tailed unpaired Student t test. Differences between multiple groups were compared using 1-way analysis of variance using SPSS software (SPSS Inc, Chicago, IL). P values <.05 were considered significant.

Results

Expression of TGF-β Signaling Components in Tumors From Tak1^ΔHep Mice

As reported previously, Tak1^ΔHep mice develop spontaneous liver inflammation, fibrosis, and HCC without treatment with carcinogens.³ In contrast, their WT counterparts did not develop liver tumors and fibrosis, even at 12 months of age.³ We initially investigated the expressions of TGF-β signaling components in liver tumors of 9-month-old Tak1^ΔHep mice. An immunohistochemical study demonstrated that expressions of TGF-βR type 2 and phosphorylated Smad2/3 were modestly increased in nontumor livers of Tak1^ΔHep mice and remarkably increased in HCC lesions of Tak1^ΔHep mice in comparisons with WT normal livers (Figure 1A). Immunoblots showed dramatically increased phosphorylation and nuclear translocation of Smad2/3, as well as blunted expression of TAK1 protein in tumors of Tak1^ΔHep mice (Figure 1B). Expressions of genes related to TGF-β signaling, including Tgfb1, Tgfbr2, and Tgfbr1, were also increased in tumors of Tak1^ΔHep mice (Figure 1C). These results suggest that activation of TGF-β signaling is associated with spontaneous HCC development in Tak1^ΔHep mice.

Figure 1.

TGF-β signaling is required for spontaneous hepatocarcinogenesis in Tak1^ΔHep mice. (A) Immunohistochemistry for TGFβR2 and phosphorylated Smad2/3 in liver tissues from 9-month-old WT, and nontumor livers and tumors from 9-month-old Tak1^ΔHep mice is shown. Original magnification ×200 for TGFβR2 staining and 400× for Smad2/3 staining. (B) Western blotting for phosphorylated Smad2 and Smad3 and TAK1 in liver tissues from 9-month-old WT, and nontumor liver tissues and tumors from 9-month-old Tak1^ΔHep mice. β-actin was used as a loading control. (C) Expression of Tgfb1, Tgfbr2, and Tgfbr1 messenger RNA in liver tissues from 9-month-old WT, and nontumor liver tissues and tumors from 9-month-old Tak1^ΔHep mice was measured by quantitative real-time polymerase chain reaction. NT, nontumor liver, T, tumors. (n=6, each samples) (D, left) Immunoblots for TAK1, TGFβR2, phospho-Smad2, phospho-Smad3, Smad4 in the livers of WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at 1 month of age are shown. β-actin was used as a loading control. (D, right) Representative macroscopic pictures of livers of 9-month-old WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice. (E) The number of tumors per mouse was counted and the maximum diameter of individual tumor nodules was measured (WT, n = 10; Tak1^ΔHep, n = 29; Tak1/Tgfbr2^ΔHep, n = 27; Tgfbr2^ΔHep, n = 10). (F) H&E staining. Data are represented as mean ± standard error of mean.*P < .05; **P < .01.

Ablation of Tgfbr2 Prevents Spontaneous HCC Development in Tak1^ΔHep Mice

To investigate the contribution of TGF-β signaling in spontaneous hepatocarcinogenesis in Tak1^ΔHep mice, we generated Tak1/Tgfbr2^ΔHep mice by crossing Tak1^ΔHep mice with Tgfbr2^ΔHep mice. Deletion of hepatic TAK1 and/or TGFβR2 protein in Tak1^ΔHep, Tgfbr2^ΔHep, or Tak1/Tgfbr2^ΔHep mice was confirmed by immunoblotting (Figure 1D). In the liver from 1-month-old Tak1^ΔHep mice, Smad2 and Smad3 were strongly phosphorylated, however, such signaling was diminished by Tgfbr2 deletion (Figure 1D). This demonstrates that the deletion of Tgfbr2 blocks TGF-β signaling in Tak1^ΔHep mice. Tak1^flox/flox mice and Tgfbr2^ΔHep mice were used as corresponding WT controls, and these mice showed no appearances of spontaneous liver tumors, even at 12 months of age. Tak1^ΔHep mice developed spontaneous HCC at 9 months of age; in comparisons, Tak1/Tgfbr2^ΔHep mice developed tumors significantly lower in multiplicity and maximal size of tumors (Figure 1D–F). These results indicate that TGF-β signaling is responsible for spontaneous HCC development in Tak1^ΔHep mice. Additionally, we confirmed no significant sex disparity in HCC formation in both Tak1^ΔHep mice and Tak1/Tgfbr2^ΔHep mice (Supplementary Figure 1).

TGF-β Signaling Is Required for Spontaneous Liver Injury and Inflammation in Tak1^ΔHep Mice

Next, we investigated whether TGF-β signaling contributes to spontaneous liver injury and inflammation in Tak1^ΔHep mice. Histologic analysis showed evident hepatocyte death and inflammatory cell infiltration to have occurred in both 1- and 9-month-old Tak1Δ^Hep mice (Figure 2A). These phenotypes were abolished in the livers of Tak1/Tgfbr2^ΔHep mice (Figure 2A). Elevation of serum alanine aminotransferase (ALT) levels in Tak1^ΔHep mice were also suppressed by Tgfbr2 deletion (Figure 2B). Kupffer cells/macrophages are the major source of inflammatory cytokines produced in response to hepatocyte damage.^15,16 According to the immunohistochemical staining of F4/80 in 1-month-old mice, the number of infiltrated hepatic macrophages in Tak1^ΔHep mice was significantly lowered in Tak1/Tgfbr2^ΔHep mice (Figure 2C and D). Accordingly, Tak1/Tgfbr2^ΔHep livers had attenuated expression of inflammatory genes, including TNFα, IL-6, CCL2, and IL-1β compared with Tak1^ΔHep livers (Figure 2E). These results indicate that spontaneous liver injury and inflammation in Tak1^ΔHep mice are associated with TGF-β–mediated hepatocyte damage, which drives Kupffer cell/macrophage activation and production of inflammatory cytokines.

Figure 2.

Ablation of Tgfbr2 in Tak1^ΔHep mice reduces spontaneous liver inflammation. (A) H&E staining in the livers of 1- and 9-month-old WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice. Original magnification 200×. (B) Serum ALT levels were measured in WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at age 1, 4, and 9 months (n = 8 at the each time point). (C, D) Immunohistochemistry for F4/80 in WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at 1 month of age. Quantification (C) and representative pictures (D). Original magnification 200×. (E) Hepatic messenger RNA expression of inflammatory genes (Tnf, Il6, Ccl2, and Il1b) in mice at the age of 1 month was determined by quantitative real-time polymerase chain reaction. Data are represented as mean ± standard error of mean. *P < .05; **P < .01.

TGF-β Signaling in Hepatocytes Drives Spontaneous Liver Fibrosis in Tak1^ΔHep Mice

Subsequently, we investigated the role of hepatocyte’s TGF-β signaling in the development of liver fibrosis in Tak1^ΔHep mice. Spontaneous liver fibrosis in Tak1^ΔHep mice was significantly inhibited by the additional deletion of Tgfbr2 (Figure 3A and B). The number of activated hepatic stellate cells that express α–smooth muscle actin was also decreased in Tak1/Tgfbr2^ΔHep mice (Figure 3C and D). Expressions of fibrogenic genes, such as Col1a1, Acta2, Timp1, and Tgfb1 were significantly reduced in Tak1/Tgfbr2^ΔHep mice compared with the Tak1^ΔHep mice (Figure 3E). These findings demonstrate that TGF-β signaling in hepatocytes is required for the development of spontaneous liver fibrosis in Tak1^ΔHep mice.

Figure 3.

Loss of Tgfbr2 in Tak1^ΔHep mice suppresses spontaneous liver fibrosis. (A, B) WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at the age of 1, 4, and 9 months (n = 7 at the each time point) were used for analysis. Fibrillar collagen deposition was determined by Sirius red staining (A) and its quantification are shown in (B). Original magnification 100×. (C, D) Expression of α–smooth muscle actin in 9-month-old mice was determined by immunohistochemistry. Quantification (C) and representative pictures (D). Original magnification 320×. (E) Hepatic messenger RNA expressions of fibrogenic markers, including Col1A1, Acta2, Timp1, and Tgfb1, were determined by quantitative real-time polymerase chain reaction in 1-month-old WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice. Data are reported as mean ± standard error of mean. *P < .05; **P < .01.

TGF-β Signaling Is Associated With Spontaneous Hepatocyte Apoptosis and Compensatory Proliferation in Tak1^ΔHep Mice

Hepatocyte death and compensatory proliferation are key drivers for the initiation of HCC after exposure to carcinogen or chronic liver inflammation.^17,18 According to the elevated serum ALT levels caused by the deletion of hepatocyte’s Tak1 (Figure 2B), spontaneous hepatocyte apoptosis was remarkably increased in Tak1^ΔHep mice, which was suppressed in Tak1/Tgfbr2^ΔHep mice (Figure 4A and B). Decreased hepatocyte apoptosis in Tak1/Tgfbr2^ΔHep mice was also demonstrated by the lack of cleaved caspase 3 induction (Figure 4E). The increased pro-apoptotic gene, Bax, in Tak1^ΔHep livers was decreased in Tak1/Tgfbr2^ΔHep livers, as measured by quantitative real-time polymerase chain reaction (Figure 4D). Because the compensatory hepatocyte proliferation in response to massive liver cell death is associated with spontaneous hepatocarcinogenesis in Tak1^ΔHep mice,³ we investigated the regenerative responses of Tak1/Tgfbr2^ΔHep mice. Cyclin D1 expression and proliferation cell nuclear antigen–positive cells were increased in Tak1^ΔHep mice, but not in Tak1/Tgfbr2^ΔHep mice (Figure 4A, C, and E). These findings demonstrated that the loss of TGF-β receptor signaling in hepatocytes inhibits spontaneous hepatocyte death and its subsequent proliferation in Tak1^ΔHep mice.

Figure 4.

Additional deletion of Tgfbr2 in hepatocytes inhibits spontaneous apoptosis and compensatory regeneration in the livers of Tak1^ΔHep mice. (A) Apoptotic hepatocytes were evaluated by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining (A, upper) and proliferating hepatocytes were evaluated by immunohistochemistry for proliferation cell nuclear antigen (PCNA) (A, lower) in the livers of WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at the 1 month of age (n = 7). Original magnification 320×. (B, C) Quantification for TUNEL staining (B) and staining for PCNA (C). (D) Hepatic messenger RNA expression of apototic genes (Bcl-2, Bax) in the livers of WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at 1 month of age was determined by quantitative real-time polymerase chain reaction. (E) Immunoblots for caspase 3, cleaved caspase 3, PCNA, and cyclin D1 in the livers of WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice at 1 month of age. Data are reported as mean± standard error of mean. *P < .05; **P < .01.

TAK1-NF-κB Axis Protects Hepatocytes From Apoptosis Induced by TGF-β

We investigated whether TGF-β signaling participates in hepatocyte death under Tak1-deficient condition using primary culture hepatocytes. WT hepatocytes did not undergo apoptosis with or without TGF-β treatment. Tak1 deficiency induced modest hepatocyte apoptosis, and TGF-β challenge dramatically increased apoptosis in Tak1^−/− hepatocytes (Figure 5A). Consistently, TGF-β treatment induced cleavage of caspase 3 in Tak1^−/− hepatocytes, but not in WT hepatocytes (Figure 5B). WT hepatocytes displayed phosphorylation of JNK and p38 induced by TGF-β treatment, but Tak1^−/− hepatocytes did not. Interestingly, the signal intensity of phosphorylation of Smad2 and Smad3 was stronger in Tak1^−/− hepatocytes than in WT hepatocytes (Figure 5B), suggesting that the TAK1-dependent pathway negatively regulates Smad2/3 activation and cell death signaling. To determine whether TGF-β signaling regulates NF-κB activation, hepatocytes isolated from NF-κB–green fluorescent protein (GFP) reporter mice were used. When the WT hepatocytes were challenged with TGF-β, GFP reporter expression increased. In contrast, this was not observed in the hepatocytes isolated from Tak1^−/− NF-κB-GFP reporter mice (Figure 5C and D). To test whether NF-κB–mediated survival signaling prevents TGF-β–mediated hepatocyte apoptosis, NF-κB activation was blocked via infection with adenoviral vector expressing IκB super-repressor, a potent inhibitor of NF-κB. Upon inhibition of NF-κB, TGF-β treatment induced hepatocyte apoptosis (Figure 5E). Deletion of Tak1 abolishes TGF-β–mediated NF-κB activation and thereby increases susceptibility to TGF-β–mediated hepatocyte apoptosis.

Figure 5.

Ablation of Tak1 in hepatocytes abolishes NF-κB activation and increases susceptibility to TGF-β–mediated apoptosis. (A) Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining (left) and its quantification (right) in the primary hepatocytes from WT and Tak1^ΔHep mice after incubation with TGF-β1 (10 ng/mL) for 24 h. Original magnification ×320. (B) Phospho-p38, phospho-JNK, phospho-Smad2, phospho-Smad3, caspase 3, cleaved caspase 3, TAK1, and β-actin were determined by immunoblot analysis after primary hepatocytes from WT and Tak1^ΔHep mice were incubated with TGF-β1 (10 ng/mL) for the indicated time periods. (C, D) When stimulated with TGF-β (10 ng/mL) for 24 h, NF-κB was activated in the primary hepatocytes from the NF-κB-GFP reporter mice with Tak1 sufficiency and deficiency, and their GFP expression was determined by microscopy (C) and Western blotting (D). Original magnification ×400. (E) Adenoviral-IκB super repressor inhibited NF-κB activation, which, as a result, caused apoptosis in primary hepatocytes from WT when stimulated with TGF-β (10 ng/mL) for 24 h. Apoptosis was determined with TUNEL staining. Data are reported as mean ± standard error of mean. *P < .05; **P < .01.

Smad Signaling Is Required for TGF-β–Mediated Apoptosis in Tak1^−/− Hepatocytes

As shown in Figures 1D and 5B, Smad2 and Smad3 were overactivated in Tak1^−/− livers and in Tak1^−/− hepatocytes treated with TGF-β. This led us to investigate whether Smad activation contributes to TGF-β–mediated hepatocyte apoptosis. To inactivate Smad signaling in Tak1^−/− hepatocytes, we used primary hepatocytes isolated from Tak1/Smad4^ΔHep mice. Although TGF-β treatment increased the number of apoptosis in both Tak1^−/− and Tak1/Smad4 double-knockout hepatocytes in comparison with the nontreatment group, apoptosis was suppressed in Tak1/Smad4 double-knockout hepatocytes compared with Tak1^−/− hepatocytes (Figure 6A), indicating that Smad activation is required for the induction of apoptosis in Tak1^−/− hepatocytes mediated by TGF-β. We further confirmed the requirement of Smad pathways by silencing Smad2 using small interfering RNA in primary hepatocytes; TGF-β–induced apoptosis in Tak1^−/− hepatocytes was suppressed in Smad2-silenced hepatocytes (Supplementary Figure 2).

Figure 6.

Inactivation of Smad molecules reduces spontaneous liver injury, inflammation, fibrosis, and HCC in Tak1^ΔHep mice. (A) Apoptosis in the primary hepatocytes from WT, Tak1^ΔHep, Smad4^ΔHep, and Tak1/Smad4^ΔHep mice after being treated with TGF-β (10 ng/mL) for 24 h were analyzed by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining. The number of TUNEL-positive cells was counted. (B–E) WT, Tak1^ΔHep, and Tak1/Smad4^ΔHep mice at the age of 1 month (n = 8) were analyzed. (B) Apoptotic hepatocytes evaluated by TUNEL staining (left) and its quantification (lower right). Original magnification 200×. Immunoblots for cleaved caspase 3, caspase 3, Smad4, TAK1, and β-actin are shown (upper right). (C) Serum ALT levels. (D) Hepatic messenger RNA expression of inflammatory genes (Tnf, Il1b, and Ccl2) and fibrogenic genes (Col1A1, Acta2, and Tgfb1) determined by quantitative real-time polymerase chain reaction. (E) Fibrillar collagen deposition was determined by Sirius red staining (left) and its quantification (right). Original magnification ×100. (F) The number of tumors per mouse was counted and the maximum diameter of individual tumor nodules was measured. (WT, n = 10; Tak1^ΔHep, n = 42; Tak1/Smad4^ΔHep, n = 15). Data are represented as mean± standard error of mean. *P < .05; **P < .01.

We then investigated the in vivo roles of Smad pathway for the liver pathology in Tak1^ΔHep mice by using Tak1/Smad4^ΔHep mice. Spontaneous hepatocytes apoptosis was increased in 1-month-old Tak1^ΔHep mice, which was suppressed in Tak1/Smad4^ΔHep mice as assessed by counting the number of apoptotic terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling (TUNEL)–positive hepatocytes (Figure 6B). Consistent with the results from TUNEL staining, serum ALT levels in Tak1/Smad4^ΔHep mice were significantly lower than those in Tak1^ΔHep mice (Figure 6C). Expression of inflammatory and fibrogenic genes, as well as fibrillar collagen deposition, were significantly suppressed by Smad4 deletion in 1-month-old Tak1^ΔHep mice (Figure 6D, E). Notably, HCC formation in Tak1/Smad4^ΔHep mice was significantly reduced compared with Tak1^ΔHep mice at 9 months of age (Figure 6F). These results indicate that Smad activation in hepatocytes is crucial for spontaneous liver injury, inflammation, fibrosis, and tumorigenesis in Tak1^ΔHep mice.

Anti-Apoptotic and Pro-Oncogenic Gene Expression and Angiogenesis in HCC From Tak1^ΔHep Mice are TGF-β Signaling-Dependent

To explore additional TGF-β–dependent mechanisms that drive hepatocarcinogenesis, we further analyzed HCC tissues in 9-month-old Tak1^ΔHep mice. We investigated pro-oncogene and anti-apoptotic gene expression in HCC from Tak1^ΔHep mice. Although expressions of pro-oncogenes, such as Ctnnb1, Yap1, and Wisp1, were significantly increased in tumors from Tak1^ΔHep mice, these expressions were decreased in tumors from Tak1/Tgfbr2^ΔHep mice (Figure 7A). Expression of anti-apoptotic Bcl-2 was not affected in the tumors from Tak1^ΔHep or Tak1/Tgfbr2^ΔHep mice. Notably, anti-apoptotic Bcl-xL increased significantly, and pro-apoptotic Bax decreased in the tumors from Tak1^ΔHep mice compared with WT normal livers and nontumor liver tissues from Tak1^ΔHep mice; however, these changes were not seen in Tak1/Tgfbr2^ΔHep mice (Figure 7B). Because CTGF is known to promote growth and migration of HCC, Ctgf messenger RNA expression was assessed.^9,12,13,19 Although tumors from Tak1^ΔHep mice exhibited increased Ctgf messenger RNA expression compared with nontumor tissues, tumors from Tak1/Tgfbr2^ΔHep mice did not (Figure 7C). Due to the pivotal role of angiogenesis in the pathogenesis of HCC, we examined the expression of VEGFa, a potent proangiogenic factor, and its receptor, VEGFR2, in the tumors of Tak1^ΔHep mice and Tak1/Tgfbr2^ΔHep mice. VEGFa was overexpressed in tumors from Tak1^ΔHep mice, but not from Tak1/Tgfbr2^ΔHep mice (Figure 7D). VEGFR2-expressing endothelial cells were also increased in tumors from Tak1^ΔHep mice compared with tumors from Tak1/Tgfbr2^ΔHep mice (Figure 7D). These results suggest that expression of Bcl-xL, CTGF, and VEGF-mediated angiogenesis are TGF-β–dependent in the development of HCC in Tak1^ΔHep mice.

TGF-β Signaling in a Mouse Model for Chemically Induced HCC and in Human HCC

To examine whether TGF-β signaling is a general promoter for HCC growth in vivo, we assessed DEN-induced HCC mouse model in WT mice and Tgfbr2^ΔHep mice. Neonatal DEN treatment induced multiple, large-sized HCC in WT mice at 9 months after DEN injection. In contrast, inactivation of TGF-β signaling by deletion of Tgfbr2 reduced HCC formation induced by DEN (Figure 7E). Consistently, serum ALT levels were higher in WT mice than in Tgfbr2^ΔHep mice after DEN treatment (Supplementary Figure 3). Analogous in pattern to HCC from Tak1^ΔHep and Tgfbr2^ΔHep mice, HCC from DEN-treated WT mice had increased expressions of prooncogenes (Ctnnb1, Yap1, and Wisp1), Bcl-xl, and Ctgf, and these expressions were suppressed in HCC from Tgfbr2^ΔHep mice treated with DEN (Supplementary Figure 4A–D). VEGFa expression and angiogenesis were increased in DEN-induced HCC from WT mice, but not in HCC from Tgfbr2^ΔHep mice (Supplementary Figure 4E). These results indicate that TGF-β is a promoter for DEN-induced HCC. Finally, we assessed activation of TGF-β signaling by comparing liver biopsy specimens of patients with chronic hepatitis C with hepatic fibrosis (hepatitic C virus–non-HCC controls) with patients with chronic hepatitis C infection and HCC (hepatitic C virus–HCC cases) with immunohistochemistry for phosphorylation of Smad2/3. In liver tissues from patients with chronic hepatitis C–induced fibrosis, phosphorylation of Smad2/3 was seen only in nonparenchymal cells and not in liver parenchymal cells. As we predicted, in patients with chronic hepatitis C and HCC, we found increased phosphorylation and nuclear translocation of Smad2/3 (Figure 7F), providing clinical relevance of activation of TGF-β signaling in human HCC.

Discussion

TAK1 is a mitogen-activated protein kinase kinase kinase that is activated in the signaling cascade of IL-1β, Toll-like receptors, TNF, and TGF-β. TAK1 requires binding to TAB2 and TAB3 for activation of the downstream kinases JNK, p38, and IKK,^5,6 These kinases further activate transcription factors AP-1 and NF-κB. The IKK–NF-κB pathway is anti-apoptotic and anti-oncogenic, whereas the JNK signaling is pro-apoptotic and pro-carcinogenic.⁷ Because TAK1 regulates both IKK–NF-κB and JNK pathways, the functions of TAK1 in hepatocyte death and carcinogenesis were uncertain. We and others have developed a genetically engineered mouse model that specifically knocked out Tak1 gene in the liver.^3,4 These mice spontaneously develop HCC accompanied by liver inflammation and fibrosis, which best mimics the progression of human HCC; therefore, Tak1^ΔHep mice serve as an excellent mouse model to analyze the mechanisms of hepatocarcinogenesis, especially for the link between inflammation, fibrosis, and carcinogenesis.

TGF-β signaling regulates cell proliferation, apoptosis, migration, angiogenesis, immunity, fibrosis, and cancer development.^8,10,20 TGF-β binds to the receptor subunit TGF-β receptor 2, and then activates TGF-β receptor 1. Intracellular domain of TGF-β receptor 1 phosphorylates Smad2 and Smad3, and forms a complex with Smad4 that translocates into the nucleus to regulate gene transcription. TGF-β signaling also activates the Smad-independent, TAK1-dependent pathway that activates JNK and p38.^8,20 In Tak1^ΔHep mice, increased production of TGF-β and overactivation of Smad2 and Smad3 were observed, indicating that the TGF-β receptor signaling is activated in Tak1^ΔHep livers. Due to the lack of TAK1, TGF-β–induced JNK and p38 phosphorylation were impaired. We demonstrate that TGF-β stimulation induced NF-κB activation in WT hepatocytes, and inhibition of NF-κB in TGF-β–stimulated hepatocytes caused hepatocyte apoptosis (Figure 5C and E). In contrast, induction in hepatic apoptosis was not seen with inhibition of JNK or p38 in TGF-β–stimulated hepatocytes (Supplementary Figure 5). Because Tak1^−/− hepatocytes are more susceptible to TGF-β–mediated cell death, the TAK1-NF-κB pathway is an essential survival signal in the TGF-β signaling. Tak1^−/− deficiency augmented TGF-β–mediated Smad2/3 phosphorylation. This suggests that the TAK1-NF-κB pathway can negatively regulate Smad2/3 activation, and overactivation of Smad2/3 is associated with TGF-β–mediated cell death in Tak1^−/− hepatocytes. In fact, inactivation of Smad pathway attenuated TGF-β–mediated cell death in Tak1^−/− hepatocytes (Figure 6A, Supplementary Figure 2). The importance of Smad signaling in TGF-β–mediated hepatocyte apoptosis was also confirmed by decreased hepatocyte apoptosis in Tak1/Smad4^ΔHep mice (Figure 6B).

To determine the role of TGF-β-Smad signaling in hepatocytes for the regulation of spontaneous liver inflammation, fibrosis, and carcinogenesis in Tak1^ΔHep mice, we have generated hepatocyte-specific Tak1 and Tgfbr2 double-knockout (Tak1/Tgfbr2^ΔHep) mice, and Tak1/Smad4^ΔHep mice. The additional deletion of hepatocyte’s TGF-β-Smad signaling in Tak1^ΔHep mice resulted in decreased spontaneous carcinogenesis, fibrosis, inflammation, and hepatocyte apoptosis. This implicates that TGF-β-Smad signaling in hepatocytes promotes liver fibrosis and formation of liver tumors that spontaneously develop in the setting of TAK1 inactivation (Figures 1–4). Notably, the suppression of liver pathology in Tak1/Smad4^ΔHep mice was less than that in Tak1/Tgfbr2^ΔHep mice. This suggests that, in addition to Smad, TGF-β–mediated unknown pathways, such as phosphatidylinositol 3 kinase or RhoA/ROCK pathway, are involved in TGF-β–mediated liver phenotype in Tak1^ΔHep mice.

Because spontaneous hepatocyte death and carcinogenesis in Tak1^ΔHep mice were reduced in Tak1/Tgfbr2^ΔHep and Tak1/Smad4^ΔHep mice, we concluded that TGF-β-Smad–mediated hepatocyte injury and compensatory proliferation of the surviving hepatocytes are the probable cause of spontaneous HCC development in Tak1^ΔHep mice. Particularly, HCC from 9-month-old Tak1^ΔHep mice overexpressed anti-apoptotic Bcl-xL, but not HCC from Tak1/Tgfbr2^ΔHep mice. This suggests that hepatocytes lacking Tak1 become resistance to TGF-β–mediated apoptosis when they transform to HCC. We also characterized the oncogenic properties of HCC developed in Tak1^ΔHep mice; the HCC of Tak1^ΔHep mice highly expressed pro-oncogene β-catenin (Figure 7). Yes-associated protein has been identified as an oncoprotein that is regulated by the Hippo signaling pathway.²¹ Yes-associated protein is overexpressed in lung, ovarian, colon, prostate, and liver cancer, and is associated with the prognosis for HCC patients.^22–24 Our data demonstrated that Yes-associated protein 1 is highly expressed in HCC from Tak1^ΔHep mice (Figure 7). Wnt signaling downstream target, WISP1, is a member of the CTGF family, and is associated with HCC development.²⁵ Although WISP1 expression was very high in tumors from Tak1^ΔHep mice, expression of this oncogene was significantly suppressed in tumors from Tak1/Tgfbr2^ΔHep mice, suggesting that TGF-β signaling regulates WISP1 expression in HCC. TGF-β signaling induces CTGF production in hepatocytes, which is associated with fibrosis, and HCC growth, and migration.^19,26 Hepatocytes and HCC of Tak1^ΔHep mice express CTGF through TGF-β signaling. It is postulated that, in combination with TGF-β, CTGF promotes HCC growth and liver fibrosis in Tak1^ΔHep mice. Consistent with previous reports,^27,28 TGF-β induced production of VEGF in hepatocytes that augmented angiogenesis in tumors and assisted tumor growth in Tak1^ΔHep mice.

There exist a number of studies demonstrating that TGF-β signaling is a tumor suppressor.⁹ Human gastric, colon, and pancreatic cancer express mutant forms of TGF-β signaling components that inhibit activity of TGF-β signaling.²⁹ Mice with inactivation of TGFβR2 by the ectopic expression of soluble TGFβR2, or haploinsufficiency for Tgfbr2 or elf, a signaling molecule essential for TGF-β signaling, augmented HCC development.^12,13,30 Tumor suppressive effect of TGF-β signaling can be mediated by increasing tumor-suppressing genes (eg, p15, p21) and inducing apoptosis of cancer cells.⁹ In contrast, sustained elevation of TGF-β is considered to promote malignancies and metastases in glioma cells and breast cancer.^31–33 TGF-β signaling induces cell migration of HCC cells through induction of epithelial to mesenchymal transition, and expressions of anti-apoptotic genes and CTGF to promote HCC growth.^19,34,35 Interestingly, additional deletion of Tgfbr2 in p53^ΔHep mice inhibits spontaneous development of HCC.³⁶ A recent publication demonstrated that TGF-β signaling is required for liver progenitor cell–associated hepatocarcinogenesis, which corroborates our results.³⁷ These findings serve as additional evidence that TGF-β signaling promotes HCC formation in vivo. To address whether TGF-β signaling promotes HCC only in the specific genetic conditions, we provided additional evidence that Tgfbr2 ablation in hepatocytes inhibits chemically induced HCC (Figure 7E). In addition to previous reports,^{38 – 41} our data demonstrated that TGF-β signaling is activated in human HCC as increased phosphorylation and nuclear translocation of Smad2/3 (Figure 7F), which provided evidence in the activation of TGF-β signaling in human HCC. It is postulated that function of TGF-β signaling as a tumor promoter or a tumor suppressor is dependent on cell types, tissues, organs, and stage of cancer. Our data clearly demonstrate that TGF-β signaling in hepatocytes contributes to tumor promotion.

Supplementary Material

01

Supplementary Figure 1. No sex disparity in Tak1^ΔHep and Tak1/Tgfbr2^ΔHep mice with respect to hepatocarcinogenesis. Male and female WT, Tak1^ΔHep, Tak1/Tgfbr2^ΔHep, and Tgfbr2^ΔHep mice WT and Tak1^ΔHep mice were analyzed at 9 months of age (male WT, n = 5; female WT, n = 5; male Tak1^ΔHep, n = 17; female Tak1^ΔHep, n = 25; male Tak1/Tgfbr2^ΔHep, n = 14; female Tak1/Tgfbr2^ΔHep, n = 21; male Tgfbr2^ΔHep, n = 5; and female Tgfbr2^ΔHep, n = 5). (A) The number of tumors per mouse was examined. (B) The maximum diameter of individual tumor nodules is presented. Data are presented as mean ± standard error of mean. *P < .05; **P < .01.

Supplementary Figure 2. Silencing Smad2 reduces apoptosis in Tak1^−/− hepatocytes. Smad2 and β-actin were determined by immunoblot analysis in WT and Tak1^−/− primary hepatocytes transfected with small interfering RNA (siRNA) for Smad2. Apoptosis were analyzed by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling staining after transfection with siRNA for Smad2 and then stimulation with TGF-β1 (10 ng/mL) for 24 h. Data are presented as mean ± standard error of mean. *P < .05; **P < .01.

Supplementary Figure 3. Liver damage is suppressed in Tgfbr2^ΔHep mice at 9 months after DEN injection. Diethylnitrosamine was injected (25 mg/kg) in 14-day-old WT and Tgfbr2^ΔHep mice and serum were sampled at 9 months after DEN injection. Serum ALT levels were measured in WT (n = 11), and Tgfbr2^ΔHep mice (n = 10). Data are represented as mean ± standard error of mean. *P < .05.

Supplementary Figure 4. TGF-β signaling mediates pro-oncogene and anti-apoptotic gene expression and angiogenesis in DEN-induced HCC. (A–E) Diethylnitrosamine was injected (25 mg/kg) in 14-day-old WT and Tgfbr2^ΔHep mice, and livers were harvested at 9 months after DEN injection. (WT, n = 11; Tgfbr2^ΔHep, n = 10). (A) Hepatic messenger RNA (mRNA) expression of pro-oncogenes (Ctnnb1, Yap1, and Wisp1) was determined by quantitative real-time polymerase chain reaction (qPCR). (B) Hepatic mRNA expression of Bax and Bcl-xl was assessed by qPCR. (C, D) Hepatic mRNA expressions of Ctgf (C), Vegfa, and Vegfr2 (D) were assessed by qPCR. (E) Immunostaining for VEGFa (upper), VEGFR2, and CD31 (lower) are shown. Original magnification used ×320. NT, nontumor liver; T, tumors. Data are presented as mean ± standard error of mean. *P < .05; **P < .01.

Supplementary Figure 5. No significant impact of cell death with inactivation of JNK and p38. (A) WT hepatocytes were pretreated with 20 μM SP600125 (JNK inhibitor) or (B) 10 μM SB203580 (p38 inhibitor). After stimulation with TGF-β (10 ng/mL) for 24 h, apoptotic hepatocytes were analyzed by counting cells positive for terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling staining.

Supplementary Table 1. Sequence of Primers Used for Real-Time Quantitative Polymerase Chain Reaction

Abbreviations used in this paper

ALT

alanine aminotransferase

CTGF

connective tissue growth factor

DEN

N-nitrosodiethylamine

GFP

green fluorescent protein

HCC

hepatocellular carcinoma

IKK

IκB kinase

IL

interleukin

JNK

c-Jun-N-terminal kinase

NF-κB

nuclear factor–κB

TAK1

TGF-β–activated kinase 1

Tak1^ΔHep mice

hepatocyte-specific Tak1-deleted mice

TGF

transforming growth factor

Tgfbr2

TGF-β receptor–2

TNF

tumor necrosis factor

TUNEL

terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick-end labeling

VEGF

vascular endothelial growth factor

WT

wild-type