High-Mobility Group Box 1 Promotes Hepatocellular Carcinoma Progression through miR-21–Mediated Matrix Metalloproteinase Activity

Hepatocellular carcinoma (HCC) is a common complication of chronic liver disease and the third leading cause of cancer deaths worldwide (1). Despite advances in diagnosis and treatment, the majority of patients with HCC have an extremely poor prognosis due to advanced disease at presentation and lack of effective treatment modalities (2). HCC has been closely linked to underlying chronic liver diseases such as hepatitis B virus (HBV), hepatitis C virus (HCV), alcoholic liver disease, and nonalcoholic steatohepatitis (NASH; ref. 1). Inflammation is a unifying feature of these disease processes and has also been implicated as a critical factor in carcinogenesis and progression of HCC (3). Understanding the relationship between inflammatory stimuli and HCC tumor biology is therefore critical to the development of novel therapeutic strategies for HCC.

The role of proinflammatory transcription factors, such as NF-κB, HIF1α, and STAT3, in the development and progression of cancer is well established, and has prompted active investigation into the regulatory mechanisms of these pathways (4). Previous work from our group and others suggest that high-mobility group box-1 protein (HMGB1) is an important member of this inflammatory cascade (5, 6). HMGB1 is a highly conserved nuclear protein that acts as a chromatin-binding factor by bending DNA and promoting access to transcriptional protein assemblies on specific DNA targets (7). More importantly, the role of HMGB1 as a damage-associated molecular pattern (DAMP) has led to its implication in a variety of disease states mediated by inflammatory stimuli, including sepsis, arthritis, ischemia-reperfusion injury, and cancer (8, 9). After passive release from necrotic cells or active release from inflammatory cells, HMGB1 binds pattern recognition receptors (PRR), including the receptor for advanced glycation end products (RAGE), Toll-like receptors (TLR)-2, TLR-4, and TLR-9, and, as a negative signaling molecule, CD24. These interactions subsequently mediate the host response to infection and injury (7, 10, 11), as well as exert an influence on the oncogenesis, progression, and metastases of various malignancies (5, 6).

miRNAs are a class of evolutionally conserved, single-stranded, small (approximately 19–23 nucleotides), endogenously expressed, and nonprotein-coding RNAs that act as posttranscriptional regulators of gene expression in a broad range of animals, plants, and viruses (12, 13). The alteration in miRNA expression profiles is a common characteristic of all human tumors (14, 15), suggesting their importance in the progression and prognosis of cancer (16). miR-21 is a well-described oncomir, overexpressed in a number of malignancies including neuroblastoma, glioblastoma, leukemia, lung, breast, pancreas, and colorectal cancers, HCC, and lymphoma (17–20). Like HMGB1, miR-21 has also been associated with multiple inflammatory pathways, including those regulated by TLRs. Specifically, knockout of TLR-4 has been linked to a reduction in IL6 and miR-21, which then protects mice from radiation-induced lymphoma (21). Given the importance of both miR-21 and HMGB1 in the inflammatory processes linked to cancer, we hypothesized that they may both be part of a common pathway.

In this study, we show that HMGB1 and miR-21 are overexpressed in human HCC tumors. We describe the role of IL6 and the transcription factor STAT3 in HMGB1-mediated induction of miR-21 expression. We also show that this pathway influences tumor progression through its targets RECK and TIMP3, inhibitors of matrix metalloproteinases (MMP), and demonstrate that the inhibition of miR-21 reduces HMGB1-induced tumor progression in a murine model of HCC. Taken together, our body of work defines a significant relationship between inflammation, miRNA, and HCC, which may provide novel targets for future therapeutic strategies.

Following an approved Institutional Review Board protocol and after informed consent of the tissue donors, samples of HCC tissue and nonadjacent liver tissue were collected from 20 patients with HCC undergoing curative surgery at the Liver Cancer Center of the University of Pittsburgh Medical Center (Pittsburgh, PA). Hepatocyte isolation was performed as previously described (22). Cell lines and reagents are described in the Supplementary Materials and Methods.

Total RNA and cellular protein were extracted as described previously (23).

Rabbit anti-Stat3, anti-phospho-STAT3 (P-STAT3) (tyr705), anti-RECK, and anti-TIMP3 were purchased from Cell Signaling Technology. Whole-cell protein was extracted with cell lysis buffer (Sigma-Aldrich), and anti-HMGB1 (Abcam) along with anti-actin antibody (Sigma) were used for Western blot analysis (22).

miRNA precursor, inhibitor, negative control, and HMGB1 interfering RNA (siRNA) were purchased from Ambion. Lipofectamine 2000 (Invitrogen) was used to transfect 50 nmol/L miRNA inhibitors, 100 nmol/L siRNAs, and 100 nmol/L miRNA precursor according to manufacturer instructions.

Reverse transcription and TaqMan real-time PCR for miRNAs were performed as described previously (23). Real-time PCR for mRNA detection was performed using SYBR Green PCR Master Mix (Ambion). Primers are listed in Supplementary Table S1.

To verify predicted targets of miR-21, the pMIR-REPORT system (Ambion) was used. Briefly, 55-mer reversion-inducing cysteine-rich protein with Kazal motifs precursor (RECK) and 57-mer metalloproteinase inhibitor 3 precursor (TIMP3) fragments of the wild-type 3′untranslated region (UTR) containing the putative binding site (Supplementary Table S2) were synthesized and ligated between the SpeI and HindIII restriction sites of the pMIR-REPORT Luciferase vector. Mutant 3′UTRs lacking the miR-21–binding site were also synthesized. The wild-type and mutant 3′UTRs were ligated between the SpeI and HindIII restriction sites of the pMIR-REPORT Luciferase vector to create the pLUC-targets and pLUC-mutTargets constructs. Huh7 cells were transfected with different reporter vectors (p-Luc-Empty, p-Luc-targets, or p-Luc-mutTargets) and cotransfected with negative control or miR-21 inhibitor. Huh7 cells were transfected with these reporter vectors in addition to a miR-21 inhibitor or negative control. pMIR-REPORT β-gal was also transfected for use in transfection normalization. At 36 hours after transfection, luciferase and β-galactosidase activities were measured using the Dual-Light System (Promega).

IL6 and HMGB1 levels in the supernatants from HCC cell lines or primary human hepatocytes were detected by ELISA (IBL), according to the manufacturer's instructions.

Cells were plated in 96-well plates at a density of 3,000 per well and incubated at 37°C. Cell proliferation was assessed at 24, 48, and 72 hours after transfection using the CellTiter 96 Aqueous One Solution Cell Proliferation Assay Kit (Promega).

The migration and invasion assays were performed by using Transwell insert chambers (6.5 mm diameter, 8 mm pore size, Corning) as previously described (24).

A total of 2×10⁵ Huh7 cells were seeded onto 6-well plates and allowed to recover for 24 hours. Transfection was accomplished using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instruction, and cells were then cultured in DMEM plus 700 μg/mL G418 (Sigma-Aldrich) to select stable transfectants. After 4 weeks, clones with G418 resistance were selected and expanded.

Athymic nude mice (NU/J, male, 4–6 weeks old, and 16–20 g) were purchased from Jackson ImmunoResearch Laboratories and bred in pathogen-free conditions. 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 NIH Guidelines for the Use of Laboratory Animals. To establish HCC xenograft model, 2×10⁶ Huh7-HMGB1 or Huh7-vector control were suspended in 100 μL serum free-DMEM and subcutaneously inoculated into the flanks of nude mice. Eight days after inoculation, a subgroup of these mice received either intratumoral injection (1 × 10⁸ copies in 50 μL PBS) of a lentiviral-based expression vector for the miR-21 inhibitor (anti-miR-21) or negative control (Lenti-GFP-NC). Every 4 days after inoculation, tumor length (L) and width (W) were measured with Vernier calipers, with volume (V) calculated by the formula V = (L×W²) × 0.5.

Three days after lentiviral injection, 150 μL of MMPsense 680 (PerkinElmer Inc.) was given to the mice intravenously. At 24 hours, the mice were imaged using IVIS Lumina. Relative intensities were determined using ImageJ software (NIH).

Quantitative data were presented as mean ± SD as analyzed by the Student t test or ANOVA, with P < 0.05 considered as statistically significant. Qualitative data were representative of three independent experiments.

Using qRT-PCR, HMGB1 and miR-21 mRNA expression were measured in three human HCC cell lines (Huh7, HepG2, Hep3B) and compared with primary human hepatocyte controls. HMGB1 and miR-21 mRNA levels were significantly increased in all three HCC cell lines (Fig. 1A and C). This observation was corroborated in vivo by comparing HMGB1 and miR-21 mRNA levels in HCC tumors and corresponding background liver tissue (Fig. 1B and D). Because the role of HMGB1 as a proinflammatory DAMP depends on its presence in the extracellular environment, we also confirmed the presence of elevated HMGB1 levels in the supernatant of HCC cell lines compared with primary hepatocytes (Fig. 1E).

To evaluate whether HMGB1 can regulate miR-21 expression, Huh7 and HepG2 HCC cells were treated with varying concentrations of recombinant human HMGB1 (rhHMGB1). HMGB1 stimulation resulted in a dose-dependent increase in miR-21 levels (Fig. 2A). HMGB1 was also found to amplify miR-21 expression in a time-dependent manner (Fig. 2B). We next established that the blockade of HMGB1 signaling with HMGB1-neutralizing antibody (anti-HMGB1) in both Huh7 and HepG2 cells prevented the expression of miR-21 (Fig. 2C). To further confirm the finding that HMGB1 mediates miR-21 expression, Huh7, HepG2, and Hep3B cells were each treated with siRNA to HMGB1 (si-HMGB1). HMGB1 knockdown was confirmed with decreased HMGB1 protein expression and release from all three cell lines (Fig. 2D). More importantly, knockdown of HMGB1 resulted in decreased miR-21 expression, suggesting that tumor-derived HMGB1 functions in a paracrine fashion to regulate miRNA expression in nearby cells (Fig. 2E).

IL6 and Stat3 have recently been described as mediators of miR-21 expression in lymphoma and glioma, respectively (17, 25). Taken with our initial observations, we hypothesized that the IL6/Stat3 signaling pathway serves as the bridge between HMGB1 and miR-21 expression. To first determine whether IL6 is upregulated in HCC, we stimulated Huh7 cells with rhHMGB1 and measured IL6 levels in the cell culture media. Media from cells treated with rhHMGB1 showed a significant increase in IL6 levels, but this effect was attenuated by addition of anti-HMGB1 antibodies (Fig. 3A). Similar results were found with HepG2 cells under the same experimental conditions (Supplementary Fig. S1A).

Stat3 not only is an important downstream target of IL6, but also has been implicated in the relationship between inflammation and HCC (26). Accordingly, Western blot analysis of HCC tumor samples revealed a significant increase in the active/total STAT3 ratio in tumor tissues compared with normal background tissue (Fig. 3B). In addition, treatment of Huh7 cells with HMGB1 induced Stat3 activation in vitro (Fig. 3C). This effect was abrogated with anti-HMGB1 treatment. HepG2 cells were used in the same experiments with similar results (Supplementary Fig. S1B). The importance of IL6 in the HMGB1–Stat3 pathway was further confirmed by treating Huh7 cells with anti-IL6, anti-IL6 receptor (anti-IL6R) antibodies, and WP1066 (an inhibitor of Stat3 phosphorylation), all of which inhibited Stat3 activation by HMGB1 (Fig. 3D).

Next, to determine whether HMGB1 induces miR-21 expression in HCC through the IL6/Stat3 signaling pathway, Huh7 and HepG2 cells were treated with HMGB1 in the presence of IL6 and Stat3 inhibitors. As anticipated, HMGB1 treatment alone resulted in a significant increase in miR-21 levels, while with inhibition of HMGB1, IL6, or Stat3 each limited miR-21 expression (Fig. 3E and Supplementary Fig. S1C). The effects of HMGB1 on the expression of pre-miR-21 (precursor mRNA for miR-21) were similar to that of the effects of HMGB1 on mature miR-21 (Supplementary Fig. S1C), which confirmed that our findings were related to alteration in miR-21 transcription and not merely generalized changes in miRNA processing.

To better evaluate the importance of the HMGB1–miR-21 signaling pathway in HCC, we sought to identify biologically relevant and measureable targets of miR-21. From the predicted targets of miR-21 determined by the TargetScan (27) algorithm, we chose to focus on the MMP inhibitors RECK and TIMP3 (Fig. 4A). These two proteins were first chosen because analysis of the HCC samples from our tissue bank show that not only are RECK and TIMP3 repressed in tumor tissue compared with matched background liver controls but, as expected, the abundance of miR-21 and HMGB1 is inversely correlated with levels of these proteins (Fig. 4B). RECK and TIMP3 were also selected because their expression has been linked to HCC prognosis (28), and they participate in regulating MMPs, which mediate invasion and metastases through degradation of the extracellular matrix (ECM; ref. 29). In addition to being predicted targets of miR-21, we also observed that the expression of RECK and TIMP3 is sensitive to HMGB1 signaling. HMGB1 stimulation represses these proteins while, in contrast, their expression is increased with siRNA knockdown of HMGB1 (Fig. 4C).

To confirm the validity of the predicted targets, the 3′UTRs containing the putative miR-21–binding sites of both RECK or TIMP3 were cloned into pLUC expression vectors (pLUC-target). After transfection of the expression vectors, Huh7 cells were treated with anti-miR-21 or an anti-miRNA negative control (anti-miR-NC), and luciferase expression was measured. Cells treated with the pLUC-target vector and anti-miR-NC showed basal repression of the luciferase activity due to the presence of endogenous miR-21, while addition of anti-miR-21 then reversed this repression (Fig. 4D, left). Luciferase activity was not suppressed when the vectors containing mutated target sequences (pLuc-mutTarget) were treated with pre-miR-21 but was inhibited as expected when vectors with the normal target sequence were treated with pre-miR-21 (Fig. 4D, right). These data confirm that translational repression is mediated by the specific binding of miR-21 to the identified target sequences within the 3′UTR of both RECK and TIMP3. These observations demonstrate that endogenous levels of miR-21 can repress expression of RECK or TIMP3 and that the repression is subsequently released with inhibition of miR-21.

The effect of miR-21 on RECK and TIMP3 expression was measured in vitro by transfection of normal hepatocytes with pre-miR-21. As expected, pre-miR-21 treatment resulted in a significant reduction of RECK and TIMP3 protein (Fig. 4E). Interestingly, RECK and TIMP3 mRNA were also downregulated by addition of pre-miR-21 in hepatocytes (Fig. 4F), suggesting that posttranscriptional regulation by miR-21 occurs through both mRNA destabilization and transcriptional blockade. Because miR-21 is overexpressed in HCC, we also investigated the effects of miR-21 knockdown in Huh7 cells. As expected, anti-miR-21 treatment resulted in significant increases in both RECK and TIMP3 protein (Fig. 4E) and mRNA levels (Fig. 4F). Together, these data show that the overexpression of miR-21 is associated with oncogenesis and provides a mechanism for modulation of RECK and TIMP3.

To evaluate the importance of the HMGB1–miR-21–RECK/TIMP3 pathway in HCC progression, we utilized the Transwell invasion/migration assay. Invasion of both Huh7 (Fig. 5A) and HepG2 (Supplementary Fig. S2A) cells were significantly increased after HMGB1 treatment, while anti-miR-21 treatment resulted in reduced invasive potential. Inhibition of Stat3 with WP1066 was also capable of abrogating HMGB1-mediated invasion. Similar effects on migration were observed in Huh7 (Fig. 5B) and HepG2 cells (Supplementary Fig. S2B) after either HMGB1 or anti-miR-21 treatment. In contrast with miR-21 inhibition, treatment of Huh7 cells with pre-miR-21 promoted invasion/migration (Supplementary Fig. S2C). In addition to its role in HCC invasion and migration, we also noted that miR-21 has a significant impact on HCC cell proliferation as measured by the MTT assay (Supplementary Fig. S2D).

To further confirm that the observed changes in invasion and migration were related to the differential expression of RECK and TIMP3, we performed phenotype rescue experiments. RECK and TIMP3 siRNAs were cotransfected into HMGB1-overexpressing Huh7 cells with anti-miR-21 or anti-miR-NC, and invasion/migration were measured using Transwell assays. In these cells, elevated miR-21 levels resulted in repression of RECK and TIMP3 and enhanced invasion and migration, while treatment with anti-miR-21 reduced their invasive potential. However, cells subjected to knockdown of RECK and TIMP3 after miR-21 inhibition exhibit restoration of their invasive potential (Fig. 5C and D). This finding is consistent with the conclusion that the regulation of RECK and TIMP3 is a critical element of miR-21's influence on the invasive potential of HCC.

The relevance of our in vitro findings was next confirmed in vivo using an HCC xenograft model. First, Huh7 cells with stable HMGB1 overexpression (Huh7-HMGB1) were generated through transfection and G418 selection. These cells were then injected into the flanks of nude mice. Eight days after tumor engraftment, one group of mice was treated with an intratumoral injection of a lentiviral-based miR-21 inhibitor (lenti-anti-miR-21) or a negative control (lenti-GFP-NC; Fig. 6A). Infection efficiency of lenti-anti-miR-21 and HMGB1 overexpression was confirmed by qRT-PCR and Western blot analysis, respectively (Supplementary Fig. S3D). MMPsense 680 and IVS lumina imaging were then used to analyze MMPs activity within the tumors. Consistent with our expectations, HMGB1-mediated MMP activity was suppressed by treatment with lenti-anti-miR21 (Fig. 6B). Two mice without any treatment were used as negative controls for the imaging agent (Supplementary Fig. S3C). After 6 weeks, Huh7-HMGB1 tumors were significantly larger than control, whereas lenti-anti-miR-21 mitigated the effect of HMGB1 overexpression on tumor growth (Fig. 6C and D and Supplementary Fig. S3A and S3B). To ensure that in vivo treatment with anti-miR-21 affects RECK and TIMP3 expression, we performed Western blot analysis of the explanted xenografts. As expected, anti-miR-21 treatment resulted in derepression of RECK and TIMP3 (Fig. 6E). Taken together, these findings suggest that HMGB1–miR-21 signaling leads to MMP activity through downregulation of RECK and TIMP3 and subsequently promotes HCC tumor progression in vivo.

Over the past decade, our understanding of the complex molecular mechanisms responsible for HCC tumorigenesis and progression has improved dramatically, implicating a number of classical cellular signaling pathways (30). However, this knowledge has not translated into improved chemotherapeutic strategies for advanced and recurrent disease, allowing HCC to remain the third most common cause of cancer deaths worldwide (31). In this study, we find that HMGB1 stimulates miR-21 overexpression through the IL6/Stat3 axis in HCC. Furthermore, we show that the HMGB1–miR-21 signaling cascade promotes tumor progression by repression of the MMP inhibitors, RECK and TIMP3, in a murine xenograft model.

The importance of inflammation in the initiation and progression of cancer is well established (4). With respect to HCC, over 90% of HCC cases occur in the setting of chronic liver inflammation related to underlying liver diseases such as viral hepatitis, NASH, and alcoholic liver disease. Thus, it is no surprise that HCC tumor biology is strongly influenced by mediators of the inflammatory response (32). Our previous work has focused on characterizing the role of DAMPs, including histones and HMGB1, in liver inflammation (8, 33). Importantly, we have shown that release of HMGB1 not only mediates the acute phase of liver injury associated with ischemia-reperfusion (25), but also promotes HCC invasion and metastasis by activating caspase-1 (6). Though not classified as a DAMP, miR-21 shares similarities with these molecules in that it bridges pathways linking inflammation to cancer (34). In breast and ovarian adenocarcinomas, hypoxia induces miR-21 expression through Akt-2, and confers resistance to apoptosis by targeting programmed cell death 4 (PDCD4) and Sprouty1 (35). With regard to HCC, miR-21 overexpression is associated with advanced tumor stage and poor prognosis (36), and this effect has previously been attributed to modulation of the tumor suppressors PTEN and PDCD4 (37, 38). Similar to recent findings that HMGB1 can stimulate miRNA expression in cardiac regeneration (39) and papillary thyroid cancer (40), our study demonstrates that HMGB1 and miR-21 participate in a common pathway linking inflammation to HCC progression.

The effects of HMGB1 release can be correlated with its ability to signal through the PRRs TLR2, TLR4 and the receptor for RAGE (41). Ligation of these cellular receptors activate the NF-κB, ERK, and p38 signaling cascades, ultimately resulting in the production of the prosurvival cytokines IL6, TNF, and IFNγ (42). Consistent with these previous observations, our work demonstrates that IL6 expression is induced by HMGB1 treatment of HCC cell lines. Furthermore, IL6 signaling is known to play an important role in the tumor microenvironment by promoting growth and inhibiting apoptosis (43, 44). This effect is dependent upon activation of the proinflammatory transcription factor STAT by Janus kinase (JAK), but the targets of STAT are less well characterized. Our observation that activation of the IL6/Stat3 axis by HMGB1 is not only sufficient, but also necessary to stimulate miR-21 expression therefore adds significant insight into the mechanism for inflammation-mediated growth of HCC.

As a classic gene family, MMPs are primarily involved in cancer development and progression via ECM degradation and/or the activation of pre-progrowth factors (45). Previous work from other groups has also shown that MMP activity can be used as a marker of disease progression in animal tumor models (46, 47). We corroborate these observations in HCC by showing that increased MMP activity correlates with the growth of tumor xenografts. Results from analysis of breast cancer cell lines have previously identified RECK and TIMP3 as targets of miR-21 in gastric (48) and breast (49) cancer, respectively. Our analysis of human HCC samples shows that RECK and TIMP3 expressions are significantly repressed when compared with background liver tissue, and we further confirm that posttranscriptional regulation by miR-21 contributes to this repression. Thus, we conclude that modulation of RECK and TIMP3 through the HMGB1–miR-21 axis plays a significant role in HCC tumor progression. It is important to note that the role of miR-21 in HCC is not limited to its effects on these two targets. The focus of our work, however, was not the identification of all downstream targets of miR-21 but rather the upstream pathway that links HMGB1 and miR-21 expression. RECK and TIMP3 therefore provided validated, biologically relevant, and experimentally measureable targets to assess the importance of miR-21 regulation by HMGB1.

In summary, we present a novel IL6/Stat3-dependent mechanism for miR-21 expression stimulated by HMGB1 in HCC, and further describe the importance of miR-21 in tumor growth and progression through its regulation of MMP activity. These observations provide insight into the pathways responsible for miRNA regulation, and further define the complex link between inflammation and cancer. Our results lend credibility to HMGB1 and miR-21 as potential targets for the development of novel therapeutic agents for HCC; and, future endeavors should work to translate these findings from animal models to the clinical setting.

No potential conflicts of interest were disclosed.

Conception and design: M. Chen, Y. Chang, X.-X. He, M.T. Lotze, J. Lin, A. Tsung

Development of methodology: M. Chen, Y. Chang, X.-X. He, M.T. Lotze, A. Tsung

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Chen, H. Huang, A. Tsung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Chen, Y. Liu, P. Varley, Y. Chang, D. Tang, A. Tsung

Writing, review, and/or revision of the manuscript: M. Chen, P. Varley, D. Tang, M.T. Lotze, A. Tsung

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Chen, M.T. Lotze, A. Tsung

Study supervision: M. Chen, H. Huang, M.T. Lotze, J. Lin, A. Tsung

Other (developed strategies for miR analysis at the UPSHS): M.T. Lotze

The authors thank Xinghua Liao, Kimberly Ferrero, and Nicole Martik-Hays for their expert technical assistance.

This work was supported by Howard Hughes Medical Institute Physician-Scientist Award (A. Tsung), R01-GM95566 (A. Tsung), The National Natural Science Foundation of China (No. 81101824), and by the Outstanding Youth Science Foundation of Tongji Hospital (No. YXQN005).

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.