Severe Hepatitis Promotes Hepatocellular Carcinoma Recurrence via NF-κB Pathway-Mediated Epithelial–Mesenchymal Transition after Resection

Hepatocellular carcinoma is the fifth most common type of cancer worldwide, which accounts for nearly 5.6% of all cancers, and a leading cause of cancer-related deaths (1, 2). More than 90% patients with hepatocellular carcinoma have preexisting chronic liver disease that is caused most commonly by chronic hepatitis B virus (HBV) infection, chronic hepatitis C virus (HCV) infection, and/or alcohol consumption (3, 4). Indeed, cumulative evidence indicates that chronic liver inflammation by long-term exposure to infectious agents (hepatotropic viruses) or toxins (ethanol) results in liver cirrhosis and hepatocarcinogenesis (5). While the molecular mechanisms linking chronic inflammation to hepatocellular carcinoma are not well defined, there is growing evidence to suggest that the crosstalk between tumor cells and the surrounding stroma in the tumor microenvironment serves as a key modulator in hepatocarcinogenesis and hepatocellular carcinoma progression. In the tumor stroma, hepatic stellate cells, fibroblasts, inflammatory cells, and vascular endothelia cells have been shown to secrete extracellular matrix (ECM) proteins, proteolytic enzymes, growth factors, and inflammatory cytokines that alter cancer signaling pathways to promote tumor cell initiation, invasion, and metastasis (6).

The microenvironment of inflamed liver turns on the NF-κB pathway to promote proliferation of hepatocytes, rendering them resistant to growth arrest (7). The inhibitor κB kinases (IKK) complex, which consists of three subunits, two catalytic kinases (IKKα and IKKβ), and a regulatory scaffold partner (IKKγ; ref. 8), plays a key role in the NF-κB signaling pathway that is known to induce inflammation-associated cancers (9). IKKβ-dependent NF-κB activation has been shown to promote hepatocyte survival in both developing and adult liver (10). In a study using a Mdr2-knockout mouse model, which spontaneously develops cholestatic hepatitis followed by hepatocellular carcinoma, Pikarsky and colleagues demonstrated that the inflammatory process triggers NF-κB activation in hepatocytes through upregulation of TNFα in adjacent endothelial and inflammatory cells and that inhibition of NF-κB by anti-TNFα treatment or induction of IκB super-repressor in the later stages of tumor development results in apoptosis of transformed hepatocytes, which prevents progression to hepatocellular carcinoma (11). In addition, our previous study indicated that noncanonical NF-κB activation is also important for tumor initiation. Specifically, IKKα activated by TNFα interacts with and phosphorylates FOXA2 at S107/S111, thereby suppressing FOXA2 transactivation activity that leads to decreased NUMB expression and further activating the downstream NOTCH pathway to promote hepatocellular carcinoma proliferation and tumorigenesis (12).

The long-term prognosis after surgical resection of hepatocellular carcinoma remains unsatisfactory due to high incidence of recurrence associated with hepatocellular carcinoma (13) that ranges from 50% to 70% five years after first curative hepatectomy (14). Several risk factors have been reported to associate with hepatocellular carcinoma recurrence, including tumor size, multifocal lesions, and vascular invasion, which could predict patient survival after surgical resection. In addition, investigation into the role of HBV infection in hepatocellular carcinoma recurrence following tumor resection by multivariate analysis showed that elevated hepatic inflammatory activity and HBV DNA levels as well as multinodular tumors are significantly associated with late hepatocellular carcinoma recurrence after operation (15). The severity of hepatitis may also influence the survival outcome of patients after surgery, such as sustained chronic hepatitis is associated with worse clinical outcome in hepatocellular carcinoma patients. However, the mechanisms of tumor progression in chronic hepatitis have not yet been explored. In this study, we investigate how chronic hepatitis or liver inflammation may be involved in hepatocellular carcinoma progression, in particular tumor recurrence and metastasis, after curative hepatectomy in the context of chronic inflammation in the liver microenvironment.

Cell migration and invasion assay, Western blot analysis, real-time PCR, chromatin immunoprecipitation (ChIP) assay, and luciferase reporter assay have previously been described (16). The antibodies used for immunoblotting, immunofluorescence, and immunohistochemical staining are listed in Supplementary Table S1.

Human hepatocellular carcinoma cell lines Hep3B, Huh7, Tong/HCC, PLC/PRF/5, HepG2, HA22T/VGH, HA59T/VGH, Malhavu, and SK-HEP-1 were obtained from Center for Molecular Medicine, China Medical University (Taichung, Taiwan). The Hep3B, Huh7, and HepG2 cell lines were validated by STR DNA fingerprinting using the AmpFlSTR Identifiler kit according to manufacturer's instructions (Applied Biosystems). The STR profiles were compared with known ATCC fingerprints (ATCC.org), to the Cell Line Integrated Molecular Authentication database (CLIMA) version 0.1.200808 (Nucleic Acids Research 37:D925-D932 PMCID: PMC2686526) and to the MD Anderson Cancer Center fingerprint database. The STR profiles matched known DNA fingerprints or were unique. All cell lines were maintained in DMEM/F12 medium supplemented with 10% FBS and antibiotics. IKKα^−/−, IKKβ^−/−, and wild-type mouse embryonic fibroblasts (MEF) were maintained as previously described (16). For analysis of ligand-dependent Snail expression, cells were serum-starved overnight and harvested directly at the indicated time points. Human TNFα (10 ng/mL) used for ligand stimulation was purchased from Roche. Details of various inhibitors used in this study are shown in Supplementary Table S2.

Supplementary Table S3 lists the details of the primers used in this study. The sequences of the short hairpin RNA (shRNA) used to knock down the expression of IKKβ and p65 have been previously described (16). The negative control vector expressing scrambled shRNA was obtained from Addgene (#1864). The plasmid constructs used for overexpressing IKKs were described in our published article (17). Snail-pGL2 Snail promoter luciferase reporter plasmid was purchased from Addgene (#31694; ref. 18). For ectopic overexpression of Snail, Snail-expressing plasmid (PCDH-Snail) was constructed by cloning the full-length Snail ORF into the pCDH-CMV-MCS-EF1-Puromycin vector (CD510B-1; System Biosciences) between EcoRI and BamH1 site. PCR primers for cloning are as follows:

• 5′−ATACTGAATTCATGGACTACAAAGACGATGACGACAAGATGCCGCGCTCTTTC CTCGTCAGG−3′ (forward)

• 5′− ATGATGGATCCTCAGCGGGGACATCCTGAGCA−3′ (reverse)

Between September 2004 and August 2008, 232 hepatocellular carcinoma patients undergoing hepatectomy in Chang Gung Memorial Hospital Linkou medical center were reviewed. Patients who were considered eligible for hepatectomy at our hospital had acceptable liver reservoir function [i.e., acceptable indocyanine green (ICG) retention rate at 15 minutes and CHILD A status), lacked multifocal tumors at bilateral lobe of liver, and showed no presence of extrahepatic lesion preoperatively. Patients in our cohort did not receive antiviral therapy. All patients received adequate resection margin (>1 cm). Curative resection considered by pathology-proved section margin was free of malignance. The study was approved by the Institutional review board at Chang Gung Memorial Hospital (Protocol No. 100-4467B). After surgery, follow-ups were conducted every 2 to 3 months at the outpatient clinic. Tumor recurrence was suspected if progressive elevation of serum AFP levels was present and/or if a new hepatic lesion was detected by ultrasonography. Dynamic CT scan or MRI was routinely arranged if recurrence was suspected. Intrahepatic recurrence was diagnosed by images showing contrast enhancement during arterial phase and wash-out in venous phase. Extrahepatic recurrence was diagnosed depending on its location according to the images taken by CT, MRI, Tc-99m methylene diphosphonate (Tc-99m MDP) bone scan, or 2-[¹⁸F]-fluoro-2-deoxy-D-gluocose ([¹⁸F]FDG) PET scan. Image-guided biopsy was performed only when considered necessary. Diagnosis of recurrent hepatocellular carcinoma was confirmed in accordance with the European Association for the Study of the Liver and the American Association for the Study of Liver Diseases guidelines. The last follow-up date in this study was December 31, 2011. The duration from date of operation to the date of first confirmation of recurrence (time-to-recurrence) was recorded as the primary endpoint (shown as tumor recurrence-free survival).

Surgical specimens from tumorous and nontumorous sections of liver were randomly sampled for histopathologic analyses using hematoxylin and eosin (H&E) staining. The ISHAK score was used to grade the severity of hepatitis (19) by a pathologist blinded to the patients' clinical and biochemical information. The ISHAK score contains five parameters: (i) periportal or periseptial interface hepatitis, score 0–4; (ii) confluent necrosis, score 0–6; (iii) focal (spotty) lytic necrosis, apoptosis and focal inflammation, score 0–4; (iv) portal inflammation, score 0–4; (v) architectural changes, fibrosis and cirrhosis, score 0–6. (19) In this study, severe hepatitis is defined as the sum of total ISHAK scores > 6 (15).

Immunohistochemical staining of p65 (Millipore), Snail (Abcam), and E-cadherin (Leica Biosystems) were performed on formalin-fixed paraffin-embedded tissue. A single representative block from each tissue was sectioned at 3 μm onto positively charged slides. Slides were then stained using the Bond-Max autostainer (Leica Microsystems) according to the manufacturer's protocol. Slides were dewaxed in Bond Dewax Solution (Leica Microsystems) and hydrated in Bond Wash Solution (Leica Microsystems). Antigen retrieval was performed at acidic pH using Epitope Retrieval 1 solution (Leica Microsystems) for 20 minutes at 100°C. Slides were then incubated with the primary antibody at suitable concentration for 15 minutes at room temperature. Antibody detection was performed using the biotin-free Bond Polymer Refined Detection System (Leica Microsystems). Finally, slides were counterstained with hematoxylin. The percentage of chromogen-containing cells were estimated in the ranges of 0%, <5%, 5%–25%, 26%–50%, >51% and semiquantitatively scored as 0 to 4. The expression intensity was determined from at least four fields per slide.

Median expression ratio from cytokine array and the expression level from ELISA validation were analyzed by Mann–Whitney U test for statistical significance. All data for histologic parameters and in vitro assays are expressed as mean ±SD. Data with continuous variables were analyzed by Student t test. Pearson χ² test or Fisher exact test was used to determine statistical significance on category variables. The correlation between two parameters was analyzed by Spearman correlation test. Kaplan–Meier survival curve with log-rank test and Cox regression model were used for survival analysis. All statistical analyses were performed using the SPSS software program (version 17; SPSS).

To determine whether the severity of liver inflammation affects tumor recurrence or metastasis after curative hepatectomy, we reviewed 232 hepatocellular carcinoma cases in which patients underwent curative resection from 2004 to 2008 at Chung Gung Memorial Hospital (Linkou; Taoyuan, Taiwan). The clinical features and tumor characteristics of these patients are summarized in Supplementary Table S4. We used ISHAK score (19) to grade the pathologic severity of chronic hepatitis from the patients' liver tissue specimens. Among these 232 hepatocellular carcinoma patients, 165 and 67 had tumors stemming from severe hepatitis (ISHAK score > 6) and mild/none hepatitis background (ISHAK score ≤ 6), respectively. Viral hepatitis was significantly associated with more severe liver inflammation and higher ISHAK score (Supplementary Table S5). At the median 52.8-month follow-up, 123 patients suffered from disease recurrence. The recurrence rate was significantly higher (P < 0.0001) in patients with severe hepatitis (100/165 patients, 60.6%) than those with none/mild hepatitis (23/67 patients, 34.3%). Moreover, among the 123 patients with disease recurrence, the rate of extrahepatic metastasis was also significantly higher in those who had severe hepatitis (61/100 patients, 61.0%) than those who had none/mild hepatitis (3/23 patients, 13.0%; P < 0.0001; Supplementary Table S6). There were 66 (53.6%) patients who suffered from early tumor recurrence within 2 years after hepatectomy, and 57 (46.3%) who had late disease recurrence over 2 years after resection. Interestingly, patients with severe hepatic inflammation, not necrosis or cirrhosis, in the liver microenvironment had higher incidence of both early and late tumor recurrence (Supplementary Table S7). The mean recurrence-free survival in patients who had none/mild hepatitis (total ISHAK score ≤ 6) was longer than those who had severe hepatitis (total ISHAK score > 6; 62.9 ± 4.6 years (95% CI, 53.9–71.9) vs. 44.9 ± 2.8 years (95% CI, 39.4–50.3); P = 0.002; Fig. 1A). Univariate and multivariate analyses identified severe hepatitis (total ISHAK score > 6) as an important predictor independently associated with poor hepatocellular carcinoma tumor recurrence-free survival after hepatectomy (Table 1).

Several inflammatory cytokines in tumor microenvironment are associated with cancer progression, invasion, and metastasis (6). To determine which cytokines are elevated in the liver tissues of hepatocellular carcinoma patients with severe hepatitis, we screened 8 pairs of human tissue specimens (normal and tumor) via a cytokine antibody arrays (RayBio Human Inflammation Antibody Array, Cat# AAH-INF-G3-4) and measured the expression level of 40 different inflammatory cytokines (Supplementary Table S8). Interestingly, only TNFα was consistently elevated in both normal and tumor tissues with severe hepatitis compared with mild/none hepatitis (Fig. 1B). In addition, ELISA examination further validated TNFα as the major cytokine that was significantly elevated in the microenvironment of both tumorous and nontumorous tissues with severe hepatitis (Fig. 1C).

To investigate the biologic effects of TNFα on hepatocellular carcinoma, two epithelial-type hepatocellular carcinoma cell lines, Hep3B and Tong, were stimulated by TNFα for 72 hours, which induced morphologic changes, including the loss of cell–cell adhesion, spindle shape transformation, cellular process elongation, and loss of cell polarity (Fig. 1D, top). These changes suggest that EMT may have occurred. Indeed, Western blot analysis also demonstrated elevated levels of mesenchymal markers, N-cadherin and vimentin, and decreased levels of epithelial markers, E-cadherin and plakoglobin, post-TNFα treatment for 72 hours in these cells (Supplementary Fig. S1A). The increase in vimentin and decrease in E-cadherin after TNFα stimulation were also visualized by confocal microscopy (Fig. 1D, bottom). In addition, TNFα-treated cells demonstrated increased cell migration and invasion (Fig. 1E), which was abrogated by the addition of a TNFα-neutralizing antibody. Together, these results indicate that TNFα is able to induce EMT in hepatocellular carcinoma cells in culture and suggest that the increased levels of TNFα in the microenvironment of severe hepatitis may stimulate EMT in hepatocellular carcinoma cells.

Several EMT regulators, including Twist (20, 21), Snail (21, 22), Slug (23, 24), Zeb1/2 (25, 26), β-catenin (27), FoxC1/2 (28), and Sox4 (29) are known to induce EMT in hepatocellular carcinoma cells, and their expression levels also correlate with patient survival. To further delineate the importance of these EMT regulators in hepatocellular carcinoma, we profiled their gene expression using the CCLE database and found that Twist, Snail, Slug, Zeb1/2, and β-catenin were upregulated in mesenchymal-type but downregulated in epithelial-type hepatocellular carcinoma cells (Supplementary Fig. S1B). We examined the protein expression level of these EMT regulators in a panel of TNFα-treated hepatocellular carcinoma cells to identify the potential regulator(s) responsible for TNFα-mediated EMT, As shown in Fig. 2A, only expression of Snail was consistently induced upon TNFα treatment. We then ectopically overexpressed Snail in two epithelial-type hepatocellular carcinoma cells (Hep3B and Tong) to validate its role in EMT in hepatocellular carcinoma. Confocal microscopy analysis revealed morphologic changes associated with EMT when Snail was overexpressed in Tong cells (Supplementary Fig. S2A). In addition, overexpression of Snail increased expression of N-cadherin and vimentin and decreased the expression E-cadherin and plakoglobin in Hepa3B and Tong cells as indicated by Western blot analysis (Supplementary Fig. S2B). The migration and invasion ability of these cells also increased when Snail was overexpressed (Supplementary Fig. S2C). To determine whether Snail is required for TNFα-mediated EMT, we knocked down Snail using shRNA in Tong and Hep3B cells. TNFα stimulation did not induce morphologic changes (Fig. 2B) or alter the expression levels of the EMT markers (Fig. 2C) in the absence of Snail. Furthermore, TNFα-induced cell migration (Fig. 2D) and invasion (Fig. 2E) ability was significantly reduced in these Snail knockdown cells. Collectively, these results indicate that Snail is sufficient to induce EMT in epithelial-type hepatocellular carcinoma cells and plays an essential role in TNFα-induced EMT.

Because TNFα activates various signaling pathways, we next explored which signaling cascade is responsible for TNFα-mediated upregulation of Snail expression in hepatocellular carcinoma cells. To this end, Hep3B and PLC cells were serum starved overnight and pretreated with various inhibitors prior to TNFα stimulation. As shown in Supplementary Fig. S3A, upregulation of Snail by TNFα was abolished only in the presence of the NF-κB inhibitor, Bay11-7082, but not by inhibitors against MAPK, p38, JNK kinase, PI3K/mTOR, or Akt. The NF-κB pathway can be activated by IKKα (noncanonical) and IKKβ (canonical) kinase. To determine whether IKKα or IKKβ is involved in TNFα-mediated Snail upregulation, we transiently overexpressed IKKα or IKKβ in 293T cells and examined Snail expression. We found that ectopically expressed IKKβ but not IKKα increased the basal level of Snail (Supplementary Fig. S3B). Conversely, a kinase-dead mutant of IKKβ (nIKKβ) failed to do so. Furthermore, we knocked down IKKβ by shRNA in three hepatocellular carcinoma cell lines (Hep3B, Tong, and PLC) and examined their Snail expression upon TNFα treatment. In the absence of IKKβ, TNFα failed to induce Snail expression (Fig. 3A). In a parallel experiment using IKKα or IKKβ knockout (KO) MEFs, we showed that TNFα induced Snail expression in wild-type and IKKα-deficient (IKKα^−/−) MEFs but not IKKβ-deficient (IKKβ^−/−) MEFs (Supplementary Fig. S4A). Reexpression of wild-type IKKβ but not kinase dead (KD) IKKβ mutant in IKKβ-deficient MEFs restored TNFα-induced Snail expression (Supplementary Fig. S4B). These results support IKKβ as a major downstream kinase mediating TNFα-induced Snail expression.

Next, we examined the time-dependent expression of Snail and found it increased significantly along with the canonical NF-κB cascade in which the signal initiated from IKKβ phosphorylation, followed by biphasic IκBα phosphorylation and degradation after 1 hour of TNFα stimulation (Fig. 3B). Similar results were observed in 8 of 9 liver cancer cell lines (Supplementary Fig. S5A). Snail expression reached a maximum at 2 to 3 hours after TNFα treatment (Fig. 3B) which was attenuated by pretreatment with Bay 11-7082 (Supplementary Fig. S5B). Collectively, these data suggest that TNFα upregulates Snail expression through the canonical NF-κB pathway via IKKβ.

RelA (p65) is a key factor that mediates transcriptional program in NF-κB signaling. Thus, we asked whether p65 plays a role in TNFα-induced EMT by knocking down p65 in Hep3B, Tong, and PLC cells. First, we showed that TNFα-induced Snail expression was attenuated in p65 knockdown cells (Fig. 3C). Overexpression of p65 in Tong and Hep3B cells was sufficient to drive EMT as indicated by decreased expression of E-cadherin and plakoglobin and increased expression of in N-cadherin and vimentin (Supplementary Fig. S6A). Nuclear translocation of p65 is known to correlate with its activity (8). Thus, to further validate the association between upregulated Snail and p65 activity, we examined the p65 nuclear localization in Tong and PLC cells treated with TNFα at different time points by subjecting nuclear and cytoplasmic fractions to Western blot analysis (Supplementary Fig. S6B). TNFα-induced nuclear translocation of p65 at 30 minutes led to a corresponding increase in the nuclear expression of Snail at 1 hour after treatment (Supplementary Fig. S6C). These results suggest that TNFα-mediated upregulation of Snail and subsequent EMT requires p65 in hepatocellular carcinoma cells.

Because TNFα-induced Snail expression requires p65, we asked whether it targets Snail transcriptionally. To this end, we analyzed the mRNA expression level of Snail with TNFα in Hep3B, Tong, and PLC hepatocellular carcinoma cells. Snail mRNA expression was elevated by TNFα treatment in a time-dependent manner (Fig. 4A). TNFα-induced Snail expression was attenuated in cells pretreated with actinomycin D, a transcriptional inhibitor (Fig. 4B). Next, we performed a ChIP assay of p65 in cells treated with TNFα and showed that p65 occupied the promoter region (−558 to −350) of Snail (Fig. 4C), indicating that Snail is specifically targeted by p65. Moreover, results from luciferase reporter assay also showed that p65 activated the Snail promoter. To further identify the site(s) within the Snail promoter (−558 to −350) targeted by p65, we scanned this region using the TFSEARCH program to predict potential p65-binding sequences. We identified only one putative site (−435 to −444) that shared high similarity with the canonical p65-binding site. Deletion of this region within the Snail promoter reduced its responsiveness to p65 (Fig. 4D). These results indicate that Snail is a p65 target gene downstream of TNFα signaling.

To further examine the possible clinical relevance of the aforementioned inflammation-related p65/Snail axis, we analyzed surgical specimens from 232 hepatocellular carcinoma patients who received curative hepatectomy by H&E and immunohistochemical staining. As shown in the Fig. 5A, tumors from patients with severe hepatitis background had more infiltration of inflammatory cells in the peritumoral area than those from clean portohepatic areas without hepatitis. In tumors with a severe hepatitis background, we detected high expression of p65 and Snail and low expression of membranous E-cadherin. In contrast, tumors from patient without hepatitis had no expression of p65 or Snail but had high levels of membranous E-cadherin (Fig. 5A). The severity of hepatitis was significantly correlated with the intensity of p65, Snail, and E-cadherin expression (Supplementary Table S9). Meanwhile, Snail expression correlated positively with p65 (N = 232; P < 0.0001), and E-cadherin expression correlated negatively with p65 (P < 0.0001) and Snail (P < 0.0001; Fig. 5B). Furthermore, patients with higher expression of p65 and Snail had shorter tumor recurrence-free survival, while patients with higher expression of E-cadherin had longer tumor recurrence-free survival (Fig. 5C). These data suggest the inflammation-induced hepatocellular carcinoma tumor progression is highly associated with upregulation of p65 and Snail and downregulation of E-cadherin.

Chronic or severe hepatitis is linked to increased risk of hepatocellular carcinoma development and tumor progression, such as invasion and metastasis, as well as worsened clinical outcome (30). However, the underlying mechanism governing the metastatic nature by inflammation in hepatocellular carcinoma has not been clearly defined. In the current study, we identified an important signaling axis in hepatocellular carcinoma that links inflammatory cytokine TNFα and EMT. Infiltration of inflammatory cells at the tumor site and surrounding liver (Fig. 5A) secret TNFα to promote EMT in hepatocellular carcinoma tumor cells and facilitate their migration and invasion ability. Specifically, we proposed a model in which elevated levels of TNFα in the tumor microenvironment from chronic hepatitis upregulates the canonical NF-κB signaling via activation of IKKβ but not IKKα; the liberated cytoplasmic p65 then translocates into the nucleus, binds to the Snail promoter, and rapidly turns on Snail expression which promotes tumor metastasis through EMT. Several EMT regulators have been reported to initiate EMT in hepatocellular carcinoma and correlate to patient survival in clinic; however, in our system, we found that Snail is the major regulator of EMT downstream of TNFα signaling. Of note, our earlier study showed that TNFα induces expression of Twist but not Snail in breast cancers to promote EMT (16), suggesting an intricate nature of cancer type–specific EMT program that governs inflammation-induced cancer metastasis.

NF-κB provides a mechanistic link between inflammation and cancer and is a major factor that controls the ability of both preneoplastic and malignant cells to resist apoptosis, regulates tumor angiogenesis, and promotes invasiveness (9). Upregulation of EMT regulator Snail by the NF-κB pathway in cancer cells may be through a transcriptional-dependent or -independent manner. For example, Akt and MAPK kinase can activate NF-κB–mediated Snail mRNA upregulation in squamous cell carcinoma (31) and peritoneal mesothelial cells (32), respectively. Moreover, NF-κB can stabilize Snail protein via upregulation of COP9 which subsequently blocks ubiquitination of Snail protein (33). Results from this study and from others (31, 32) have demonstrated that Snail expression is transcriptionally regulated. Interestingly, in SW480 colon cancer cells, the minimal p65-responsive promoter region of Snail was identified at −194 to −78 (34); however, in hepatocellular carcinoma cells, p65 did not occupy this region (Fig. 4C). Instead, deletion of the predicted p65-binding site in the Snail promoter at −435 to −444 substantially inactivated its responsiveness to p65 (Fig. 4C and D). These observations are consistent with previously reported complexity and cell-specific regulation of the Snail promoter by NF-κB (31).

In the current study, we found that hepatocellular carcinoma patients with more severe hepatitis had a higher tendency toward more intrahepatic recurrence and extrahepatic metastasis after curative hepatectomy. The significance of the severity of hepatitis on the clinical outcome for hepatocellular carcinoma patients after surgical resection of primary tumor may be explained by the activation of the TNFα/NF-κB/Snail pathway. First, it may be that there are more cancer cells with EMT potential in the microenvironment of hepatocellular carcinoma tumors from patients with severe hepatitis. These mesenchymal-type hepatocellular carcinoma cells within primary tumors may have already undergone micrometastases prior to operation. In addition, conventional liver resection may induce the release of cancer cells from the liver into the peripheral blood circulation, especially when liver is mobilized during hepatectomy (35). These procedure-related disseminations of cancer cells have been shown as a predicator of postsurgical recurrence of hepatocellular carcinoma (36), suggesting that the microenvironment surrounding hepatocellular carcinoma tumors from patients with severe hepatitis may shed cancer cells more easily into the systemic circulation during hepatectomy than those with mild hepatitis. Moreover, microscopic tumor cells in multifocal lesions may gain EMT potential and promote intrahepatic recurrence and distant metastasis after operation if hepatitis is sustained after resection. Thus, recurrent or metastatic hepatocellular carcinoma induced by hepatic inflammation may lead to adverse clinical outcome after surgical resection, and targeting the TNFα/NF-κB/Snail pathway or controlling hepatitis may improve patient survival after hepatectomy. Our results are in agreement with a previous a large-scale study, which profiled gene expression and survival of hepatocellular carcinoma patients, that the inflammatory signatures, TNFα and NF-κB signaling, in the surrounding liver tissue are correlated with poor survival (37).

It has become clear that dysregulation of NF-κB and the signaling pathways that control its activity are involved in cancer development and progression (38) and that targeting NF-κB may provide therapeutic benefit. One successful example is the treatment of multiple myeloma by bortezomib (Velcade; Millenium Pharmaceuticals), a reversible 26S proteasome inhibitor that effectively inhibits NF-κB activity (39). This drug also effectively induces apoptosis and inhibits growth of hepatocellular carcinoma cell in preclinical studies (40, 41). However, subsequent clinical studies of bortezomib as monotherapy for patients with unresectable hepatocellular carcinoma failed to bring survival benefit (42). In contrast, direct inhibition of NF-κB has been shown to reduce liver inflammation and attenuate liver fibrosis/cirrhosis (43). Several inhibitors of NF-κB, such as caffeic acid, captopril, curcumin, pyrrolidine dithiocarbamate, resveratrol, silymarin, and thalidomide, have demonstrated antinecrotic, anticholestatis, antifibrotic, and anticancer activities in the liver (43), but large prospective and randomized control clinical trials are still required to demonstrate their efficacy in treating hepatocellular carcinoma. Given that hepatocellular carcinoma with severe hepatitis background has more frequent vascular invasion and poorer patient survival (44), the results from our study suggest that hepatocellular carcinoma may be managed by controlling hepatitis in addition to NF-κB inhibition. As viral infection is a major cause of hepatitis, antiviral therapies may also attenuate virus-induced inflammation in the peritumor microenvironment and prevent hepatocellular carcinoma recurrence. Indeed, postoperative adjuvant antiviral therapy improved survival of hepatocellular carcinoma patients with viral infection and hepatitis (45–50).

Our observations also raised the intriguing questions of whether other cytokines are responsible for hepatocellular carcinoma recurrence and whether late recurrent hepatocellular carcinoma is a consequence of cancer metastasis. Recognizing the limited size of clinical samples due to scope of the current study, we certainly do not exclude the potential importance of cytokines other than TNFα although the results from our cytokine array analysis suggest that TNFα is correlated with the degree of hepatitis. Meanwhile, even though we observed a correlation between inflammation, EMT, and early and late tumor recurrence (2 years as the cutoff point; Supplementary Table S7), it should be noted that only early reappearance of tumors that happen within the first year are generally accepted as recurrences from the original tumors. However, late recurrences (more than 2 years) can occur as a result of de novo cancer formation in the microenvironment of sustained hepatitis. The possibility that parts of late recurrence may have originated from an early metastatic event of primary tumors after a long dormancy should not be fully excluded until a systematic genomics analysis is performed to compare the evolution of primary and late recurrent tumors.

In summary, identification of Snail as a downstream target of IKKα/p65 links the inflammatory cytokine TNFα-mediated NF-κB activation and EMT in hepatocellular carcinoma. Moreover, the findings in this study suggest that chronic liver inflammation leads to tumor metastasis and decreases patient survival after curative resection of primary hepatocellular carcinoma. Inhibition of NF-κB activation or diminishment of hepatitis after surgery, therefore, has important clinical implications for the treatment or prevention of hepatocellular carcinoma recurrence and metastasis.

No potential conflicts of interest were disclosed.

Conception and design: T.-J. Wu, C.-T. Yeh

Development of methodology: T.-J. Wu, S.-S. Chang, C.-W. Li

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-J. Wu, S.-S. Chang, T.C. Chen, W.-C. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.-J. Wu, S.-S. Chang, Y.-H. Hsu, W.-C. Lee, M.-C. Hung

Writing, review, and/or revision of the manuscript: T.-J. Wu, S.-S. Chang, Y.-H. Hsu, C.-T. Yeh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.-J. Wu, S.-S. Chang, T.C. Chen, C.-T. Yeh, M.-C. Hung

Study supervision: M.-C. Hung

We thank Dr. Jennifer L. Hsu for her assistance with manuscript preparation. We also thank the Liver Research Center at Linkou Chang Gung Memorial Hospital for providing access to the equipment to perform immunohistochemical staining.

This study was funded in part by the following: National Science Council (grant NMRP 101-2314-B-182A-032-MY2; to T.-J.Wu); NIH (CA109311; to M.-C. Hung); Ministry of Science and Technology, International Research-intensive Centers of Excellence in Taiwan (I-RiCE; MOST 104-2911-I-002-302); Ministry of Health and Welfare, China Medical University Hospital Cancer Research Center of Excellence (MOHW104-TDU-B-212-124-002); and Center for Biological Pathways.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.