High-mobility group box 1 activates caspase-1 and promotes hepatocellular carcinoma invasiveness and metastases

Authors


  • Potential conflict of interest: Nothing to report.

  • This work was supported by a Howard Hughes Medical Institute Physician-Scientist Award (to A.T.) and an American College of Surgeons Research Fellowship (to A.T.).

Abstract

Hypoxia is often found in solid tumors and is associated with tumor progression and poor clinical outcomes. The exact mechanisms related to hypoxia-induced invasion and metastasis remain unclear. We elucidated the mechanism by which the nuclear-damage–associated molecular pattern molecule, high-mobility group box 1 (HMGB1), released under hypoxic stress, can induce an inflammatory response to promote invasion and metastasis in hepatocellular carcinoma (HCC) cells. Caspase-1 activation was found to occur in hypoxic HCC cells in a process that was dependent on the extracellular release of HMGB1 and subsequent activation of both Toll-like receptor 4 (TLR4)- and receptor for advanced glycation endproducts (RAGE)-signaling pathways. Downstream from hypoxia-induced caspase-1 activation, cleavage and release of proinflammatory cytokines interleukin (IL)-1β and -18 occurred. We further demonstrate that overexpression of HMGB1 or treatment with recombinant HMGB1 enhanced the invasiveness of HCC cells, whereas stable knockdown of HMGB1 remarkably reduced HCC invasion. Moreover, in a murine model of HCC pulmonary metastasis, stable knockdown of HMGB1 suppressed HCC invasion and metastasis. Conclusion: These results suggest that in hypoxic HCC cells, HMGB1 activates TLR4- and RAGE-signaling pathways to induce caspase-1 activation with the subsequent production of multiple inflammatory mediators, which, in turn, promote cancer invasion and metastasis. (HEPATOLOGY 2012;55:1866–1875)

A worldwide increase in mortality associated with hepatocellular carcinoma (HCC) has recently been reported.1 Clinical treatment of HCC remains challenging because of a lack of effective chemotherapy and clearly defined endpoints for clinical protocols.2, 3 The advanced nature of disease at presentation, often in the background of steatosis4 and chronic hepatitis C,5 is a major change in the etiology from that observed in the past.6 The main cause of death in HCC patients, irrespective of etiology, is cancer metastasis within or outside the liver. The underlying mechanisms responsible for invasiveness and metastatic spread of HCC are still not fully understood.

Hypoxia is found in a wide range of human malignancies, including liver, breast, prostate, and pancreatic cancers as well as brain tumors and melanoma.7 The extent of hypoxia is associated with tumor progression and poor clinical outcomes. Hypoxia has a dual role: Insufficient oxygen limits tumor cell division while, at the same time, selecting for more malignant cells and induces cell adaptations that allow for more invasive behavior.8 Cancer cells may promote tumor metastasis through the release of paracrine or endocrine signals that enhance invasiveness and promote the tumor premetastatic niche.9-11 Tumor-derived mediators not only inhibit apoptosis of tumor cells, but also promote cell invasiveness and metastasis.

High-mobility group box 1 (HMGB1) is an evolutionarily conserved, chromatin-binding protein that has been implicated in several disease states, including sepsis, arthritis, ischemia-reperfusion injury, and cancer.12, 13 Cancer cells that have undergone necrotic cell death can release HMGB1 into the local microenvironment. HMGB1 is also actively secreted by inflammatory cells, acting as an endogenous danger signal and binding with high affinity to several receptors, including the receptor for advanced glycation endproducts (RAGE) as well as Toll-like receptors (TLRs)-2, -4, and -9.12 Extracellular HMGB1 can lead to chronic inflammatory-reparative responses that, in the setting of cancer, may lead to tumor cell survival, expansion, and metastases. Overexpression of HMGB1 is associated with all of the central hallmarks of cancer.13 Interestingly, numerous studies suggest that HMGB1 plays a role in metastasis development and thus link it to poor prognosis in a variety of cancers, including prostate, breast, liver, and colon.13

Caspase-1, a prototypical member of the inflammatory caspases, is responsible for the maturation of prointerleukin (IL)-1β and pro-IL-18.14 Once activated by cellular infection or stress, including hypoxia, caspase-1 cleaves IL-1β and -18 in inflammasomes. Endogenous damage-associated molecular pattern (DAMP) molecules, such as HMGB1, are either normal cell constituents or derived from the extracellular matrix and can be released into the extracellular milieu during states of cellular stress or damage. DAMPs subsequently activate inflammasomes and promote proinflammatory responses. Cytokines downstream of caspase-1 can attract and activate immune cells to induce inflammation and affect tumor progression, including cell survival and metastasis.15, 16

Therefore, we evaluated whether HMGB1 could induce caspase-1 activation and subsequently contribute to tumor invasiveness in hypoxic HCC cells. We show that HMGB1 is overly expressed in human HCC tumors, compared to noncancerous liver tissue, and is higher in the serum of patients with HCC. HMGB1 is also released from hypoxic HCC cells and then subsequently activates caspase-1. In addition, both HMGB1 receptors, TLR4 and RAGE, play important roles in hypoxia-induced caspase-1 activation. Inhibition of HMGB1 or caspase-1 can decrease cell invasiveness in the setting of hypoxia, and knockdown of HMGB1 in HCC cells suppresses lung metastasis in a murine model of liver cancer.

Abbreviations

AIM2, absent in melanoma 2; DAMP, damage-associated molecular pattern; ELISA, enzyme-linked immunosorbent assay; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HMGB1, high-mobility group box 1; IL, interleukin; NLRP, NOD-like receptor family, pyrin domain-containing protein; RAGE, receptor for advanced glycation endproducts; rhHMGB1, recombinant human high-mobility group box 1; SEM, standard error of the mean; shRNA, short hairpin RNA; siRNA, short interfering RNA; sRAGE, the soluble isoform inhibitor of the RAGE receptor; TLR, Toll-like receptor; Z-YVAD-FMK, carbobenzoxy-valyl-alanyl-aspartyl-(O-methyl)-fluoromethylketone.

Materials and Methods

Patient Samples, Cell Lines, and Reagents.

Patient serum, human HCC, and their corresponding nontumorous liver samples were collected at the time of surgical resection at the University of Pittsburgh (institutional review board–approved serum and tumor registry). Cell lines and reagents information are described in the Supporting Materials and Methods.

Animals.

Male wild-type (C57BL/6) mice (8 weeks old) were purchased from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA). Animal protocols were approved by the Animal Care and Use Committee of the University of Pittsburgh, and the experiments were performed in adherence to the National Institutes of Health Guidelines for the Use of Laboratory Animals.

Establishment of Stable HMGB1-Expressing Cells and HMGB1 Knockdown Cells.

Stable transfection is described in the Supporting Materials and Methods.

Immunoblotting Analysis.

Whole cell protein was extracted with cell lysis buffer (Sigma-Aldrich, St. Louis, MO). Cytoplasmic protein extraction and western blotting analysis were performed by following a standard protocol, as described previously.17

RNA Interference by Short Interfering RNA.

The RNA interference experiment protocol is described in the Supporting Materials and Methods.

Enzyme-Linked Immunosorbent Assay.

HMGB1 level in serums from humans and mice was detected by enzyme-linked immunosorbent assay (ELISA) (IBL, Toronto, Ontario, Canada), according to the manufacturer's instructions.

Confocal Microscopy.

Cultures were fixed, stained, and examined under a confocal microscope (Olympus, Tokyo, Japan), as described in the Supporting Materials and Methods.

Caspase-1 Colorimetric Assay.

The Caspase-1 Colorimetric Assay kit (R&D Systems, Minneapolis, MN) was used according to the manufacturer's protocol.

Cell Migration and Invasion Assays.

We determined migration and invasion as previously described.18

Experimental Metastasis Model.

To examine the metastatic potential of stable HMGB1 knockdown clones, 2 × 106 Hepa1-6, in 0.3 mL of phosphate-buffered saline, were injected into the tail vein of C57BL/6 mice. For in vivo tracking, the Hepa1-6 cells were stably transfected with firefly luciferase. One hundred milligrams per kilogram of D-luciferin (Caliper Life Sciences, Hopkinton, MA) were injected into the peritoneal cavities of mice, and bioluminescence was detected with the IVIS 100 Imaging System (Caliper Life Sciences).

Statistical Analysis.

Results are expressed as the mean ± standard error of the mean (SEM). Statistical analysis was performed using the Student's t test or one-way analysis of variance test. All statistical analyses were performed using Sigma Stat v.3.5 (Systat Software, Inc., Chicago, IL). Graphs were generated using Sigma Plot v.10 (Systat Software). P < 0.05 was denoted as statistically significant.

Results

HMGB1 Is Overexpressed in HCC.

Overexpression of HMGB1 is associated with tumor progression.13 To study the role of HMGB1 in HCC, we first examined the amount of HMGB1 in 20 HCC tissue samples and their corresponding nontumor liver by immunoblotting analysis. The detailed clinicopathological information of 20 cases is shown in Supporting Table 1. We found that the expression of HMGB1 was higher in all HCC tissues (Fig. 1A). Compared to normal primary hepatocytes, the expression of HMGB1 was also much stronger in five HCC cell lines (Fig. 1B). We then examined the level of nuclear and cytoplasmic HMGB1 by the fractionation of nuclear and cytoplasmic proteins in HCC tissues and nontumor liver tissues. The amount of nuclear HMGB1 in HCC tissues and nontumor tissues was not significantly different (data not shown). However, cytoplasmic HMGB1 was absent or present at low levels in nontumor tissues, whereas cytoplasmic HMGB1 was found at high levels in HCC tissues (Fig. 1C). High cytoplasmic levels of HMGB1 usually occur in the context of active HMGB1 release. Indeed, compared to serum obtained from healthy volunteers, HMGB1 levels were significantly increased in patients with HCC (Fig. 1D). These findings suggest that HMGB1 may, in fact, be actively released into the circulation of patients with HCC.

Figure 1.

Overexpression and nuclear and cytoplasmic localization of HMGB1 in HCC cells. (A) HMGB1 protein levels were measured with western blotting in 20 paired HCC samples and their nontumor counterparts. Protein expression results were normalized to internal control β-actin. ***P<0.001. N, nontumorous liver; T, tumor. (B) Expressions of HMGB1 in two primary hepatocytes (mouse and human hepatocytes) and five HCC cell lines were detected by western blotting analysis. (C) Expression of HMGB1 in cytoplasm was measured within HCC samples and their nontumor counterparts. (D) Expression of HMGB1 in normal and HCC serums were detected by ELISA. **P <0.01. (E) Nontumor and HCC tissues were stained for HMGB1. Red, HMGB1; blue, nuclei. (F) Hepa1-6 and Huh7 cells were analyzed in normoxia or hypoxia by immunostaining. Green, HMGB1; blue, nuclei; red, F-actin. Imaging shown is representative of three experiments with similar results.

Hypoxia Induces HMGB1 Release From HCC Cells.

Hypoxia is a hallmark of solid tumors, including HCC.8 Accordingly, we found hypoxia-inducible factor-1 alpha levels to be substantially increased in liver homogenates of HCC specimens, compared to nontumor tissue (Supporting Fig. 1). Immunofluorescence showed that HMGB1 mainly localized in the nucleus of hepatocytes in the nontumorous liver. However, in addition to nuclear HMGB1, cytoplasmic HMGB1 was also present at high levels in HCC cells (Fig. 1E). To determine whether hypoxia plays a role in inducing the expression or translocation of HMGB1 in HCC cells, Hepa1-6 and Huh7 cells were cultured under normoxic (21% O2) or hypoxic (1% O2) conditions. Under normoxia in both cell lines, HMGB1 remained located predominantly in the nucleus. After exposure to hypoxia, the expression of HMGB1 increased in the cytoplasm (Fig. 1F). These findings indicate that hypoxia leads to the nuclear to cytoplasmic translocation of HMGB1 in HCC cells.

To determine whether hypoxia induces the expression or translocation of HMGB1 in HCC, western blotting analysis was performed. The expression of HMGB1 in whole cell lysates was not significantly changed during hypoxia at either the protein or mRNA level (Fig. 2A; Supporting Fig. 2). In contrast, hypoxia led to a time-dependent increase of HMGB1 translocation to the cytoplasm and HMGB1 release into the media in both cell lines (Fig. 2B,C). In addition, cell viability was not substantially different, when comparing HCC cells exposed to hypoxia and normoxia, as assessed in a crystal violet viability assay (Fig. 2D). These findings demonstrate that hypoxia leads to the translocation and release of HMGB1 in HCC cells.

Figure 2.

Hypoxia induces translocation and release of HMGB1. (A) Expression of HMGB1 was determined by western blotting in Hepa1-6 and Huh7 cells subjected to 24 hours of hypoxia (1% O2). Cell supernatants and cytoplasmic extracts from Hepa1-6 cells (B) or Huh7 cells (C) that were subjected to a time course of hypoxia were collected. HMGB1 was determined by western blotting. Blotting shown is representative of three experiments with similar results. (D) Viability was determined by crystal violet assay on Hepa1-6 and Huh7 cells subjected to 24 hours of hypoxia.

Hypoxia Induces Caspase-1 Activation in HCC Cells.

Caspases play an important role in programmed cell death and inflammation.16, 19 Though apoptosis inducers caspase-3 and -9 were activated in our hypoxic HCC cells (data not shown), we focused on caspase-1, which can induce inflammation and affect tumor progression. To characterize caspase-1 activation during hypoxia, Hepa1-6 and Huh7 cells were cultured under hypoxic conditions for various times. The 45-kDa procaspase-1 was cleaved into the active 20-kDa caspase-1 during hypoxia and was increased in a time-dependent manner during hypoxia (Fig. 3A,B). Using a colorimetric assay to assess caspase-1 activity during hypoxia, we found that caspase-1 activity was also significantly increased in both Hepa1-6 and Huh7 cells subjected to hypoxia, compared to normoxic controls (Fig. 3C). To confirm the localization of cleaved caspase-1, we performed immunofluorescent staining for cleaved caspase-1. After hypoxia, cleaved caspase-1 significantly increased in the cytoplasm diffusely in both cell lines (Supporting Fig. 3). These findings demonstrate that hypoxia induces caspase-1 activation in HCC cells.

Figure 3.

Hypoxia induces caspase-1 activation in HCC cells. Caspase-1 expression was determined by western blotting in Hepa1-6 (A) or Huh7 cells (B) subjected to a time course of hypoxia. Blotting shown is representative of three experiments with similar results. (C) Whole cell lysates were extracted from Hepa1-6 and Huh7 cells. Caspase-1 activity was determined by colorimetric assay. *P <0.05. Assay shown is representative of three experiments with similar results.

HMGB1 Is Necessary for Hypoxia-Induced Activation of Caspase-1 and Also Activates Caspase-1 Under Normoxia.

The previous results suggest that an association exists between HMGB1 release and caspase-1 activation in hypoxic HCC cells. Because HMGB1 can promote inflammation and inhibit apoptosis, we next sought to study whether HMGB1 participates in the hypoxia-induced activation of the inflammation-related caspase-1. Hepa1-6 cells were treated with ethyl pyruvate or an anti-HMGB1 neutralizing antibody to inhibit HMGB1 release or block HMGB1, respectively. Either inhibiting HMGB1 release or blocking HMGB1 significantly decreased the production of cleaved caspase-1 in hypoxia (Fig. 4A). Treatment with ethyl pyruvate or anti-HMGB1 neutralizing antibody also resulted in a dramatic decrease in caspase-1 activity, compared with hypoxic controls (Fig. 4B). These results suggest that HMGB1 released from hypoxic HCC cells is necessary for caspase-1 activation.

Figure 4.

HMGB1 is necessary for hypoxia-induced activation of caspase-1 and induces activation of caspase-1 in normoxia. (A) Western blotting analysis for caspase-1 from Hepa1-6 cells treated with 10 mM of ethyl pyruvate (EP) or 1 μg/mL of anti-HMGB1 neutralizing antibody (a-HMGB1) in hypoxia. Blotting shown is representative of three experiments with similar results. (B) Caspase-1 activity was determined by colorimetric assay after various treatments. *P < 0.05 versus Hepa1-6 cells that were subjected to hypoxia. (C) Western blotting analysis of caspase-1 in Hepa 1-6 cells treated with various doses of rhHMGB1 for 24 hours. (D) Western blotting analysis of caspase-1 in Hepa1-6 cells treated with 1 μg/mL of rhHMGB1 for various times. (E) Ectopic overexpression of HMGB1-GFP fusion protein in Hepa1-6 stable clones. GFP and HMGB1-GFP fusion expression were confirmed via western blotting analysis with GFP antibody. (F) Cleaved caspase-1 was significantly increased in HMGB1 stable transfectants, as compared with vector controls, via western blotting analysis. GFP, green fluorescent protein.

To further confirm that HMGB1 activates caspase-1, we treated Hepa1-6 cells with recombinant human HMGB1 (rhHMGB1) and studied these cells under normoxic cell culture conditions. rhHMGB1 treatment in normoxia induced a dose- and time-dependent significant increase in cleaved caspase-1 in Hepa1-6 cells (Fig. 4C,D). Constitutively active HMGB1 was also stably transfected into the Hepa1-6 cell line and the expression was confirmed via western blotting (Fig. 4E). HMGB1 stably expressing cells displayed a significant increase of cleaved caspase-1, compared with the vector control (Fig. 4F). These results indicate that HMGB1 is required for hypoxia-induced caspase-1 activation and that HMGB1 overexpression independently induces caspase-1 activation in Hepa1-6 cells, even without exposure to hypoxia.

Signaling Pathways Downstream of HMGB1 Are Involved in Hypoxia-Induced Activation of Caspase-1.

Several important receptors have been implicated in HMGB1 signaling, including RAGE, TLR2, and TLR4.12 To investigate whether these receptors are involved in hypoxia-induced caspase-1 activation, western blotting analysis was performed on whole cell protein from Hepa1-6 cells subjected to hypoxia. TLR4 and RAGE, but not TLR2, were detected in Hepa1-6 cells. The expression of TLR4 increased in a time-dependent manner in Hepa1-6 cells subjected to hypoxia (Fig. 5A). To determine whether hypoxia-induced caspase-1 activation is TLR4 dependent, Hepa1-6 cells were treated with TLR4 short interfering RNA (siRNA). After TLR4 siRNA treatment, the expression of TLR4 was significantly decreased (Supporting Fig. 4A), and the hypoxia-induced expression of cleaved caspase-1 was also significantly diminished (Fig. 5B). Anti-TLR4 neutralizing antibody was also used to confirm this result.

Figure 5.

Signaling pathways downstream of HMGB1 are involved in hypoxia-induced activation of caspase-1. (A) Expression of TLR4 from Hepa1-6 cells subjected to hypoxia for various time points. (B) Western blotting analysis for caspase-1 from hypoxic Hepa1-6 cells transfected with scrambled siRNA or TLR4 siRNA or treated with 5 μg/mL of anti-TLR4 neutralizing antibody (a-TLR4). (C) Expression of RAGE from Hepa1-6 cells subjected to hypoxia for various time points. (D) Western blotting analysis for caspase-1 from hypoxic Hepa1-6 cells transfected with scrambled siRNA or RAGE siRNA or treated with 5 μg/mL of sRAGE. (E) Western blotting analysis for caspase-1 from Hepa1-6 cells treated with a-TLR4 or/and sRAGE. Blotting shown is representative of three experiments with similar results.

RAGE regulates metabolism, inflammation, and epithelial survival in the setting of stress.12 The expression of RAGE increased in a time-dependent manner in Hepa1-6 cells subjected to hypoxia (Fig. 5C). To further study whether hypoxia-induced caspase-1 activation is RAGE dependent, Hepa1-6 cells were treated with RAGE siRNA. Compared with scrambled siRNA, treatment with specific siRNA against RAGE resulted in a significant decrease of RAGE protein (Supporting Fig. 4B). The expression of cleaved caspase-1 was also significantly decreased after RAGE siRNA transfection. The soluble isoform inhibitor of the RAGE receptor (sRAGE) was also used to confirm the result. As was shown, cleaved caspase-1 was significantly inhibited after blocking RAGE with sRAGE treatment (Fig. 5D). To further confirm that both TLR4 and RAGE receptors are involved in hypoxia-induced caspase-1 activation, Hepa1-6 cells were simultaneously treated with anti-TLR4 neutralizing antibody and sRAGE. Cleaved caspase-1 was almost completely inhibited by blocking both TLR4 and RAGE (Fig. 5E). These results suggest that hypoxia-induced caspase-1 activation occurs through both TLR4- and RAGE-signaling pathways.

NOD-Like Receptor Family, Pyrin Domain-Containing Protein 3–Dependent Caspase-1 Activation Promotes Maturation of IL-1β and -18 Under Hypoxia.

To date, four cytoplasmic receptors have been described that form an inflammasome complex: NOD-like receptor family, pyrin domain-containing protein (NLRP)1, NLRP3, IPAF, and absent in melanoma 2 (AIM2).14 To investigate which one is involved in hypoxia-induced caspase-1 activation, Hepa1-6 cells were transfected with NLRP1 siRNA, NLRP3 siRNA, IPAF siRNA, and AIM2 siRNA. All receptors were efficiently silenced by their respective siRNA (data not shown), but only NLRP3 siRNA was found to inhibit hypoxia-induced caspase-1 activation (Fig. 6A; Supporting Fig. 4C,D). These results suggest that hypoxia-induced caspase-1 activation in HCC cells is NLRP3 dependent.

Figure 6.

NLRP3-dependent caspase-1 activation promotes cleavage of IL-1β and-18 in hypoxia. (A) Western blotting analysis for caspase-1 from Hepa1-6 cells treated with NLRP3 siRNA in hypoxia. (B) Hepa1-6 cells were treated with a caspase-1 inhibitor (Z-YVAD-FMK; 10 μM), then analyzed by western blotting for caspase-1. (C) Western blotting analysis for IL-1β after treatment with Z-YVAD-FMK. (D) Expression of IL-18 from Hepa1-6 cells was determined by western blotting.

Next, we investigated signaling pathways downstream of caspase-1. The cytokines, IL-1β and -18, begin as cytosolic precursors that require cleavage by the cysteine protease, caspase-1, to generate biologically active molecules. Cleaved caspase-1, IL-1β, and IL-18 were all increased after hypoxia (Fig. 6B-D). In contrast, all three were decreased after treatment with the caspase-1 inhibitor, Z-YVAD-FMK (carbobenzoxy-valyl-alanyl-aspartyl-[O-methyl]-fluoromethylketone) (Fig. 6B-D). As downstream effectors of caspase-1, both cytokines released from cancer cells can induce the production and secretion of other cytokines and chemokines, recruit stromal cells, and induce angiogenesis, leading to tumor progression.15, 16

HMGB1-Induced Caspase-1 Activation Is Involved in Hypoxia-Induced Cell Migration and Invasion.

We next sought to determine the effects of hypoxia on HCC cell migration and invasion. Under hypoxia, Hepa1-6 cell migration was increased 2.04- ± 0.14-fold (P < 0.01; n = 6) in 1% O2 (Fig. 7A. Similarly, invasion of Hepa1-6 cells through reconstituted three-dimensional Matrigel matrices was increased 1.72- ± 0.12-fold (P < 0.01; n = 6) in 1% O2.

Figure 7.

HMGB1-induced caspase-1 activation is involved in hypoxia-induced cell invasion. (A) Transwell chamber studies were performed for 24 hours in 21% O2 or 1.0% O2. Migration and invasion are expressed relative to the levels observed with normoxic controls. (mean ± SEM; n = 6). **P < 0.01. (B) Hepa1-6 cells were treated with rhHMGB1 (1 μg/mL) or Z-YVAD-FMK (10 μM) in normoxia or a-HMGB1 (1 μg/mL) or Z-YVAD-FMK in hypoxia. Cells were allowed to invade Matrigel for 24 hours. (C) The number of invasive cells was quantified. Invasion is expressed relative to the levels observed with normoxic controls. *P < 0.05; **P < 0.01. (D) Huh7 cells were treated with a-HMGB1 or Z-YVAD-FMK in hypoxia. Cells were allowed to invade Matrigel for 24 hours. (E) The number of invasive cells was quantified. Invasion is expressed relative to the levels observed with normoxic controls. *P < 0.05.

To elucidate whether HMGB1-induced caspase-1 activation was responsible for the increase in cell invasion observed in hypoxia, Hepa1-6 cells were treated with anti-HMGB1 neutralizing antibodies. Blockade of HMGB1 inhibited the increase in cell invasion observed during hypoxia (Fig. 7B,C). We also examined caspase-1 inhibition in hypoxia-induced invasion. Hepa1-6 cells treated with the caspase-1 inhibitor, Z-YVAD-FMK, also exhibited significantly decreased hypoxia-induced invasion (Fig. 7B,C). In contrast, HCC cell invasiveness was increased after rhHMGB1 treatment, but this effect was abolished by caspase-1 inhibition (Fig. 7B,C). These findings were further confirmed using Huh7 HCC cells (Fig. 7D,E). Collectively, these results demonstrate that HMGB1-mediated caspase-1 activation is required for hypoxia-induced invasion in HCC cells.

HMGB1 Promotes HCC Cell Invasion and Metastasis In Vitro and In Vivo.

To further assess the effect of HMGB1 on HCC cell invasion, constitutively active HMGB1 was stably transfected into the Hepa1-6 cell line. HMGB1 stably expressing cells (pEGFPN1-HMGB1) displayed a significant increase in cell-invasion ability, compared with vector controls (Fig. 8A). In contrast to HMGB1 overexpression, stable knockdown of HMGB1 in Hepa1-6 cells, using short hairpin RNA (shRNA) (Supporting Fig. 5), considerably decreased the invasiveness of Hepa1-6 cells, as evidenced by the Transwell assay (Fig. 8B).

Figure 8.

HMGB1 promotes HCC cells invasion and metastasis in vitro and in vivo. Stable HMGB1 overexpressing (A) or knockdown cells (B) were seeded onto the upper chamber of the Transwell and allowed to invade Matrigel for 24 hours. The number of invasive cells was quantified. Invasion is expressed relative to the levels observed with vector controls. **P < 0.01. (C) A luciferase signal was generated from luciferase-labeled Hepa1-6 cells. The signal was observed in the lungs of the C57BL/6 mice. A stronger signal was observed in the vector control group. (D) The luciferase signal was determined once a week in both the vector control and the shRNA groups. Each group contained 5 mice. (E) HMGB1 in serum was measured by ELISA in different groups. (F) Metastatic nodules on the surface of lung are shown (indicated by arrows). Numbers of tumor nodules were quantified.

To determine whether HMGB1 participates in HCC metastasis in vivo, a murine lung metastasis model was utilized. Mice were injected via the tail vein with luciferase-expressing tumors derived from HMGB1 shRNA and vector control clones and were monitored weekly for bioluminescent signals. Four weeks after injection, mice were sacrificed and their lungs were examined. Bioilluminescent signals in the lungs from the control group were much stronger than from the HMGB1 shRNA group (Fig. 8C,D). Furthermore, serum HMGB1 levels from the control group were 43.48 ±10.91 ng/mL, which were much higher than that from the HMGB1 shRNA-treated group (17.12 ± 4.56 ng/mL) (Fig. 8E). Tumor nodules were also more numerous in the control than in the shRNA group (Fig. 8F) and were confirmed with histology (data not shown), demonstrating that HMGB1 can promote metastasis.

Discussion

HCC remains a leading cause of cancer-related death worldwide. This is despite the fact that a number of advances in both surgical (e.g., transplantation or resection) and ablative (e.g., transcatheter arterial chemoembolization or radiofrequency ablation) techniques have developed in the past several decades20 and is reflective of the advanced nature of disease with which many patients present as well as the lack of effective chemotherapeutic agents aimed at the treatment of widely metastatic disease. Though loss of tumor-suppressor gene function, oncogene activation, direct viral effects, and angiogenesis all appear to be involved in the development of HCC,21 the lack of effective chemotherapy speaks to a gap in knowledge as to the precise molecular events and pathways involved in tumor development and progression. Therefore, further elucidating such a mechanism is an important goal in developing novel strategies to both prevent and treat HCC.

Hypoxia is a hallmark of diverse human solid tumors and is associated with tumor progression.8 The extent of hypoxia in a tumor may represent an independent indicator of poor prognosis22; however, the mechanism by which hypoxia affects cancer progression is still unclear. Therefore, understanding the effects of hypoxia on cancer cell biology is an important goal; however, it is also important to identify individual gene products that may be targeted to counteract the adverse effects of hypoxia in cancer. Our study sought to determine how HMGB1, a proinflammatory cytokine, affects tumor invasion and metastasis induced by hypoxia. We found that hypoxia induces casapse-1 activation in HCC cells. For the first time, we report that HMGB1 is essential for hypoxia-induced caspase-1 activation in HCC cells. HMGB1 translocates from the nucleus to the cytoplasm in hypoxic HCC cells and inhibiting its release prevents hypoxia-induced caspase-1 activation. Furthermore, treatment with rhHMGB1 or overexpression of HMGB1 in HCC cells induces caspase-1 activation, even in normoxic cell culture.

HMGB1 is recognized as the prototypical DAMP.23 Initially identified as a chromatin-binding protein, HMGB1 can be released by both passive24 and active25 pathways. The passive pathway requires loss of cell-membrane integrity, as observed in necrosis. The active pathway appears to involve HMGB1 hyperacetylation and packaging into secretory vesicles. Our previous studies showed that HMGB1 release from cultured hepatocytes is an active process regulated by reactive oxygen species in the setting of hypoxia.17 In cancer cells, HMGB1 is actively released after anticancer agent treatment and promotes cell survival.26 We postulated that HMGB1 release in HCC cells would be an active process under hypoxic conditions. Indeed, hypoxia did not promote the expression of HMGB1, but induced the nuclear to cytoplasmic translocation of HMGB1 and release of HMGB1 into the extracellular space in two HCC cell lines, indicating that HMGB1 release in this setting may affect the biologic behavior of tumors.

Tumor development and progression is associated with caspase activation, which regulates apoptosis and inflammation.27 One hallmark of tumor cells is the intrinsic or acquired resistance to apoptosis. Surprisingly, recent studies demonstrate that apoptosis promotes early tumorigenesis.19 Apoptosis-related caspases (caspase-3 and -9) were activated in hypoxic HCC cells (data not shown). In our study, cell invasiveness was increased after inhibiting caspase-3 in hypoxic HCC cells, which suggests that apoptosis seems to inhibit hypoxia-induced invasion (data not shown).

Studies have implicated caspase-1, an inflammation-related caspase, in a variety of responses, including the host response to microbial pathogens, inflammatory diseases, and metabolic and autoimmune disorders.28 Recent studies have also shown that caspase-1 is involved in tumorigenesis and tumor progression.29 Caspase-1 activation, regulated by the inflammasome, promotes the maturation of proinflammatory cytokines, such as IL-1β and -18, and induces inflammatory responses. Caspase-1 is activated not only in immune cells, but also in epithelial and mesenchymal cells in conditions such as tissue repair.30 Recently, Okamoto et al.16 found that fresh melanoma biopsies constitutively express activated caspase-1 and secrete IL-1β. After caspase-1 activation, IL-1β-secreting human melanoma cells recruit stromal cells and induce angiogenesis, leading to tumor progression. However, the mechanisms by which caspase-1 affects tumor cancer progression remain incompletely understood. We speculate that hypoxia promotes the caspase-1 activation and maturation of IL-1β and -18, thus contributing to invasion and metastasis.

Inflammation can promote tumorigenesis. Robust epidemiological data support the role of inflammation induced by chronic hepatitis B or C viral (HBV/HCV) infections and alcohol abuse as key players in HCC development. Lymphotoxin, the proinflammatory and homeostatic cytokine, is induced by HBV or HCV infection and can promote HCC development.31 Similarly, HMGB1 can be secreted in response to HBC or HCV infection and can contribute to the pathogenesis of these infections.32, 33 Additionally, HMGB1 may be involved in the initial phases of tumorigenesis associated with these viral infections. Indeed, activation of the HMGB1 receptor, RAGE, significantly affects tumorigenesis and hepatic tumor growth.34 Studies have shown that, when HMGB1 is overexpressed, the oncoproteins, Cyclin D and E, which regulate cell proliferation, are overexpressed, whereas tumor-suppressor protein p53 is repressed.35

Overexpression of HMGB1 is also associated with tumor progression, including invasion and metastasis13; however, the mechanism is still not fully understood. We hypothesized that during hypoxia, the release of HMGB1 may promote caspase-1 activation and may thus contribute to invasion and metastasis of liver cancer. Our results demonstrate that inhibiting HMGB1 release or blocking its effects inhibits the activation of caspase-1 in hypoxia. In normoxia, rhHMGB1 can induce caspase-1 activation. This confirms that released HMGB1 from HCC cells can induce caspase-1 activation in both normoxia and hypoxia.

Using Transwell experiments, we confirmed that hypoxia promotes liver cancer cell migration and invasion in vitro. Therefore, we wished to study what role HMGB1 plays in hypoxia-induced invasion. We found that HMGB1-induced caspase-1 activation promoted invasion in hypoxia. Conversely, knockdown of endogenous HMGB1, specifically using shRNA, significantly reduced the invasiveness of HCC cells, indicating that HMGB1 is closely involved in HCC invasion. Furthermore, an in vivo murine model of HCC lung metastases also confirmed that HMGB1 is associated with tumor invasion and metastasis. Taken together, our data demonstrate that HMGB1 plays a pivotal role in HCC invasion and metastasis by way of enhancing invasiveness and activating caspase-1, with the subsequent production of multiple mediators. These findings support the notion that HMGB1 may serve as a suitable target for the development of novel anticancer agents.

Acknowledgements

Expert technical assistance from X. Liao, L. Shao, and J. Chen is appreciated. The authors thank J. Evankovich, L. Zhang, G. Nace, and J. Klune for their valuable advice and discussion.

Ancillary