Paradoxical Role of HMGB1 in Pancreatic Cancer: Tumor Suppressor or Tumor Promoter?

HMGB1: A Double-edged Sword

HMGB1 is a nuclear DNA-binding protein that belongs to the High Mobility Group-box superfamily with a dual function: transcription and organization of DNA and extracellular signaling, being one of the main extracellular DAMP molecules. HMGB1 binds DNA through its two domains, A and B boxes. Aspartate and glutamate form the end-string with an acidic carboxyl terminus (8).

HMBG1 is essentially expressed in the nucleus of all eukaryotic cells and continuously transferred between the nucleus and cytoplasm. Acetylation and phosphorylation induce translocation of HMGB1 to the cytoplasm and increase the secretion from macrophages and cancer cells. Post-translational methylation of HMGB1 causes its cytoplasmic localization in neutrophils. Overexpression, as well as cytoplasmic localization of HMGB1, has been observed in most tumor cells. Moreover, HMGB1 plays a critical role in DNA repair and is related to a number of transcription factors implicated in cancer development, stimulating their activity. For instance, HMGB1 interacts directly with P53 (tumor suppressor) to enhance its DNA binding (3, 9).

HMGB1 can also function as a negative signaling molecule, mediating the response to infection and injury, thereby promoting inflammation. To act as a “danger signal” and inflammatory mediator, HMGB1 must be transported extracellularly. This occurs in two fundamentally different manners: active secretion from living inflammatory cells or passive release from necrotic or stressed cells. Once in the extracellular space, it binds to different receptors, such as RAGE and toll-like receptors (TLR)-2, TLR-4, TLR-9 (3).

Nevertheless, even if the mechanisms are not very well described yet, oxidative stress appears as a central regulator of HMGB1's translocation, release and activity in inflammation and cell death (10). Calcium- or reactive oxygen species (ROS)-dependent mechanisms were also confirmed to induce HMGB1 release (3).

Cells undergoing necrosis or tumor cells treated with chemotherapy release HMGB1, which possesses proinflammatory properties. HMGB1, released from dying tumor cells, stimulates mature dendritic cells (mDC) tumor antigen processing, favoring anticancer immune responses. Furthermore, activation of HMGB1 and its receptor RAGE results in the activation of nuclear transcription factor NF-kappa B (NF-κB), which up-regulates leukocyte adhesion molecules and the production of proinflammatory cytokines and angiogenic factors (3).

In many cases, extracellular HMGB1 acts as a pro-tumor protein due to its cytokine, chemokine and growth factor activity. It has been found that certain aminoacids in the B box domain are essential for the protein's proinflammatory cytokine function, especially the cysteine aminoacid in the position 106 (C106) in the B box domain that is required for HMGB1 to bind to TLR-4 activating macrophages (11).

Dendritic cell activation by HMGB1 involves the reduced form of C106. The inflammatory capacity of HMGB1 is prevented when cysteines are oxidized to sulfonic acids, suggesting that the function of HMGB1 can be modulated through cysteine redox modifications (12). Fully reduced HMGB1 has no post-translational modifications and it is the predominant form in nucleus and cytoplasm in basal conditions and interacting with several receptors, such as CXCR4, to promote cell migration. HMGB1 is oxidized in the cytoplasm after ROS production (13). The oxidation of HMGB1 acts as a physiological negative feedback mechanism. Because the extracellular environment is mainly oxidative during inflammatory processes, the oxidation restricts the proinflammatory activity of HMGB1 preventing excessive inflammatory damage (12). After oxidation, sulfonyl HMGB1 has no activity for cell migration or cytokine induction (13).

Moreover, C23 and C45, two additional cysteine residues in HMGB1, form an intramolecular disulfide bond that stabilize the protein and may be implicated in HMGB1's inflammatory signaling pathway (12). Disulfide HMGB1 interacts with TLR4 to induce cytokine and chemokine transcription and secretion (13). The role of C23 and C45 is so crucial that, when fully reduced HMGB1 (with no post-translational modifications) is moderately oxidized, even if the cytokine function would be inhibited, the present disulfide bond is still able – and needed – to induce its cytokine capacity. Additionally, when mutated, C23 and C45 avoid the initiation of autophagy by HMGB1 inhibiting the relation to Beclin1 (12, 13).

On the other hand, intracellular HMGB1 acts as an anti-tumor protein due to its ability to sustain genome stability and autophagy activity during tumor growth (3). Importantly, it has been described that when HMGB1 is overexpressed, oncoproteins regulating cell proliferation increase, whereas tumor suppressor proteins decrease (14).

With regard to pancreatic cancer, the HMGB1/RAGE pathway has a crucial function in the regulation of apoptosis. Targeted knockdown of RAGE in pancreatic tumor cells, leads to increased apoptosis, diminished autophagy and decreased tumor cell survival. Furthermore, depletion of HMGB1 in pancreatic tumor cell renders them significantly more sensitive to induced apoptotic cell death (3).

Moreover, RAGE and HMGB1 jointly enhance tumor cell mitochondrial complex I activity and, therefore, ATP production, needed for supporting anabolism and tumor cell proliferation and migration (4).

Increasing evidence, thus, suggests that cellular localizations of RAGE and HMGB1 affect their function and interaction. RAGE and HMGB1 appear in their soluble form in the normal tissues, whereas they are insoluble and membrane-bound in cancer cells where they form a stable complex (15). Of note, soluble forms of these proteins are detectable in the bloodstream of cancer patients, representing potential diagnostic biomarkers, as well as therapeutic targets (16).

HMGB1, Autophagy and Apoptosis in Pancreatic Cancer

Autophagy is an ubiquitous process in mammalian cells that results in the breakdown of cytoplasmic components within the lysosome in response to stress conditions, such as starvation, allowing the cell to adapt to environmental or developmental changes (17). Autophagy is generally recognized as programmed cell survival mechanism maintaining miscellaneous physiological processes. On the other hand, however, autophagy is also implicated in the pathogenesis of several diseases, including cancer (18).

The role of autophagy in pancreatic cancer is complex as it is associated with both tumor suppression and increased tumorigenesis (19, 20). Pancreatic cancer cells exhibit constitutive autophagy under basal conditions. When autophagy is inhibited in pancreatic cancer cells, ROS increase, causing DNA damage and a reduction of mitochondrial oxidative phosphorylation. This results in significant growth suppression of pancreatic cancer cells in vitro leading to tumor regression (21).

Moreover, ROS trigger the translocation and release of HMGB1 in pancreatic cancer cells. Disulfide HMGB1 binds to RAGE, induces Beclin 1-dependent autophagy (the intramolecular disulfide bridge in HMGB1 is needed to bind to Beclin1 and sustain autophagy) and promotes tumor cell resistance to chemotherapeutic agents or ionizing radiation. In contrast, oxidized HMGB1 increases the cytotoxicity of these agents and induces apoptosis via the mitochondrial pathway (22, 23).

In a hypoxic microenvironment, hypoxia-induced autophagy mediates survival of pancreatic tumor-initiating cells. Physiological enhanced levels of autophagy render pancreatic cancer stem-like cells resistant to induction of apoptosis by hypoxia and starvation, while inhibition of autophagy induces death of pancreatic cancer stem-like cells and inhibits self-renewal potential (18).

The HMGB1-RAGE pathway plays critical roles in the induction of autophagy and inhibition of HMGB1-RAGE leads to increased apoptosis and decreased autophagy in pancreatic cancer cells. RAGE inhibits apoptosis through P53 in response to chemotherapy. P53 has a dual function. Nuclear P53 stimulates autophagy in a transcription-dependent manner, whereas cytoplasmic P53 inhibits the formation of autophagosomes in a transcription-independent fashion (24, 25). Furthermore, studies have shown that p53-deficient mice form markedly fewer autophagosomes (26). Moreover, RAGE sustains autophagy by relating to certain molecules involved in the formation of the autophagosome as it decreases phosphorylation of the mammalian target of rapamycin (mTOR) or increases Beclin 1-Vps34 interaction (24, 27). These results can be extrapolated to HMGB1, as it has been found that endogenous HMGB1 is a critical pro-autophagic protein that enhances cell survival and limits programmed apoptotic cell death (22).

Murine pancreas containing an activated oncogenic allele of Kras, the most common mutational event in pancreatic cancer, develops a small number of pre-cancerous lesions that stochastically advance into pancreatic cancer over time. In mice containing oncogenic Kras and lacking p53, loss of autophagy no longer blocks tumor progression, but actually accelerates tumor onset (26).

RAGE is required for extracellular HMGB1-mediated autophagy. Antioxidant enzymes (e.g., superoxide dismutases) and small-molecule antioxidants (e.g., N-acetyl-L-cysteine) inhibit HMGB1 activation, as well as autophagy in pancreatic cancer cells. In the nucleus, HMGB1 regulates expression of heat shock protein beta-1 (HSPB1). Recent data suggest that HSPB1 inhibits apoptosis and its phosphorylation promotes autophagy (25).

When analyzing the relationship between autophagy and pancreatic cancer risk factors, only one connection has been outlined so far. It has been found that pancreatitis stimulates autophagy induction and, at the same time, impairs late stages of autophagy (18). A defective autophagy could cause vacuole accumulation and trypsinogen activation, contributing to the development of pancreatitis (28).

Seaweed polyphenols suppress the transcription of all investigated autophagy regulators, including HMGB1, in human and mice pancreatic cancer cell lines in vitro and in vivo. Moreover, the data collected strongly suggest that the selected polyphenols could serve as effective adjuvants for current pancreatic cancer treatment modalities and may inhibit tumor relapse (29).

HMGB1 and Immunogenic Cell Death

Immunogenic cell death (ICD) is a form of cell death that stimulates an anti-tumor immune response against dead-cell antigens, derived from cancer cells (30). Multiple stimuli can trigger ICD, which involves changes in the composition of the cell surface, as well as release of soluble mediators, occurring in a defined temporal sequence. In fact, ICD is preceded or accompanied by the emission of a series of immunostimulatory DAMPs in a precise spatiotemporal configuration. ICD-associated biomarkers include, amongst others, HMGB1, the endoplasmic reticulum (ER) chaperone calreticulin (CALR) and ATP. Cell death that is not accompanied by CALR exposure, ATP secretion and HMGB1 release is generally not perceived as immunogenic, resulting in a partial or even total loss of immune system stimulation (31).

Such signals operate on a series of receptors expressed by dendritic cells to stimulate the presentation of tumor antigens to T cells (32).

The notion that cancer cell apoptosis is detected by the immune system involves the possibility of a self anticancer response and also implies a critical role for immunostimulation in therapies against malignant cells (31).

As it has been previously mentioned, HMGB1 is considered to be an important DAMP and its release to the extracellular space seems to be essential for the optimal presentation of tumor antigens to dendritic cells and efficient T cell stimulation (19, 20).

Nevertheless, it should be taken into consideration that a high release of nucleosomes and HMGB1, as well as low release of soluble RAGE (sRAGE), have been found to be associated with poorer prognosis (33). Soluble RAGE has a different role than cellular RAGE, acting as a decoy receptor that binds to sHMGB1, inhibiting its functions. In late tumor stages, high levels of sHMGB1 are associated with poor prognosis due to chronic adaptation. In fact, this would be occurring due to the reason that chronic ICD could lead to adaptation and down-regulation of the immune response, while acute ICD – also induced by some chemotherapeutic agents – leads to immune stimulation and better patient outcome (34).

Immunogenic cell death can be stimulated by retinoic acid-inducible gene I (RIG-I)-like helicase (RLH), an ubiquitously expressed cytosolic helicase that recognizes viral RNA species and can be activated by synthetic RNA. When treated with RIG-I ligands, efficient tumor control is observed, as well as the activation of dendritic cells and CD8+ T cells as effector, connecting RLH to immunogenic cell death triggering adaptive immunity. Markers of ICD, HMGB1 included, are highly released when pancreatic cancer cells are treated with RLH ligands (20).

The pathway involves type I interferon (IFN) (7, 20) with a double action; it alerts the immune system to danger leading to activation of dendritic and effector cells, while it stops inhibition of immunosuppressive cells, such as regulatory T cells (20). Local treatment of potentially resectable pancreatic cancers with RLH ligands that would carry a highly immunogenic form of cell death before surgery may facilitate the induction of a T-cell-mediated antitumor immune response with the potential to control outgrowth of residual tumor cells (7).

As it has been previously described, the cure of particularly aggressive malignancies requires induction of ICD, coupling oncolysis with immune responses via calreticulin, ATP and HMGB1 release from dying tumor cells.

Novel therapies employing oncolytic viruses have emerged as promising anticancer approaches.

One interesting strategy involves the oncolytic parvovirus H-1 (H-1PV), which activated multiple interconnected death pathways when targeting human pancreatic cells and elevated extracellular HMGB1 in all infected cultures, whether non-dying, necrotic or apoptotic. HMGB1 release was not promoted by additional oncolysis upon co-treatment with gemcitabine. This chemotherapeutic alone failed to stimulate HMGB1 release despite strong cytotoxic effects. HMGB1 release in pancreatic cancer cells appears to follow the alternative cytokine secretory pathway (using, for instance, molecular transporters). Secreted HMGB1, therefore, functions as an alarm and might represent a general mechanism by which viral infection is signaled to the immune system. The oncolytic virus H-1PV, which is not pathogenic in humans, could be then regarded as a multimodal anticancer treatment (35).

HMGB1 and Pancreatic Cancer Growth and Metastasis

The response of the host stroma to the invasive tumor is remarkable in pancreatic cancer and important to recognize in order to understand tumor growth. This dynamic process is known as a desmoplastic reaction and involves a complex interaction between the normal host epithelial cells, invading tumor cells, stromal fibroblasts, inflammatory cells, proliferating endothelial cells, an altered extracellular matrix and growth factors activating oncogenic signaling pathways by autocrine and paracrine mechanisms (36).

Several key molecules have been identified, such as collagen type I, fibronectin, laminin, matrix metalloproteinases (MMP) and tissue inhibitors of MMP, as well as growth factors, such as transforming growth factor-beta (TGFβ), platelet-derived growth factor (PDGF), connective tissue growth factor (CTGF) and hepatocyte growth factor (HGF), and also chemokines and integrins (36).

The tumor necrosis factor (TNF) superfamily ligand-receptor – of which the fibroblast growth factor (FGF)-inducible 14 (Fn14) is a part of, is as well involved in many cellular processes including proliferation, migration, differentiation, inflammation and angiogenesis.

The Fn14 receptor is expressed at relatively low levels in normal tissues but dramatically elevated in a wide variety of tumor types, especially in pancreatic cancer. Pancreatic cancer cells treated with a compound composed of anti-Fn14 monoclonal antibody (ITEM-4) conjugated to recombinant gelonin (rGel), a highly cytotoxic ribosome-inactivating N-glycosidase, showed the highest release of HMGB1 compared to different cancer cell lines (e.g., breast, lung and bladder) (37).

The dismal prognosis of pancreatic cancer is partly due to its high risk of local invasion and metastasis. Moreover, metastasis is the major cause of mortality and morbidity in patients with cancer. HMGB1 has a role in the transcription of many genes involved in different steps in the metastatic cascade and has been linked to different types of cancer in human and animal models (38). The HMGB1 protein was proposed to be directly involved in tumor cell metastasis through its ability to promote cell migration, modulate the adhesive properties of cells and modify components of the extracellular matrix (15).

HMGB1 has been identified as a gene involved in the metastatic phenotype of pancreatic cancer in several studies. An up-regulation of this gene with a subsequent overexpression of HMGB1 was shown in cell lines derived from the metastatic lesions in patients with pancreatic cancer (39). Therefore, HMGB1 has been closely associated with tumorigenesis in the pancreas and suggested as a diagnostic marker for pancreatic cancer (40).

The crucial role of HMGB1 for the invasive properties of pancreatic cancer was further proven when transfecting antisense-HMGB1 expression vector into a human pancreatic cancer cell line. The antisense-HMGB1 inhibited the expression of HMGB1, as well as that of metalloproteinases MMP-2 and MMP-9 mRNA expression, observing then a significant reduction in cell migration (41).

Moreover, strong expression of the HMGB1 receptor, RAGE, was detected in human pancreatic carcinoma cells with high metastatic ability. On the contrary, low levels of RAGE were observed in cells with low ability. Similarly, expression of MMP-9 showed nearly the same tendency. RAGE and MMP-9 are expressed concordant with the metastatic capacity of the human pancreatic cancer cells (42), although the connection between them has not yet been found.

RAGE, and its ability to regulate gene expression through NF-κB, constitutively activated in pancreatic cancer, was proposed as one of the possible mechanisms involved in pancreas cancer metastasis (43).

Most importantly, systemic in vivo administration of small RAGE antagonist peptide (RAP) blocking receptor activation by multiple ligands (S100P, S100A4 and HMGB-1) reduced the growth and metastasis of pancreatic tumors. Besides, RAP inhibited the interaction of S100P, S100A4 and HMGB-1 with RAGE at micromolar concentrations. Finally, RAP also reduced the ability of the ligands to stimulate RAGE activation of NF-κB in cancer cells in vitro and in vivo. (44).

Therefore, targeting RAGE may represent a promising tool for future treatment of pancreatic cancer.

Gemcitabine Resistance

The standard treatment for advanced pancreatic cancer is gemcitabine (Gem). However, the clinical effect of Gem is very limited due to an inherent and acquired chemoresistance, becoming, therefore, a hindrance to effective treatment. ICD is one of the keys to improve the efficacy of anticancer treatment; however, the capacity of Gem to induce ICD is currently debated (45).

Firstly, considering that hypoxia is one of the hallmarks of pancreatic cancer, the impact of hypoxia inducible factor-1 inhibitor, PX-478, has been investigated showing that, when combined with Gem, the anti-tumor effect is significantly enhanced as it stimulates ICD (45).

Secondly, studies on the relationship between miRNAs and the chemosensitivity of tumors have received increasing attention in recent years. miR-218 expression, one of the main miRNAs involved in cancer, is down-regulated in human pancreatic cancer tissues and cell lines. Additionally, miR-218 expression is significantly lower in Gem-resistant cells compared to the sensitive ones, thus indicating that the resistance of pancreatic cancer cells to Gem might be related to miR-218. It has also been demonstrated that overexpression of miR-218 enhances the sensitivity of pancreatic cancer cells to Gem (5). HMGB1 is also related to the sensitivity of pancreatic cancer cells to Gem-based chemotherapy (35) and miR-218 negatively regulates the expression of HMGB1, inhibiting the proliferation and invasion of pancreatic cancer cells (6). Consequently, the mechanism of the miR-218 effect may depend on the regulation of HMGB1 expression (5) and stimulating miR-218 might show positive results in pancreatic cancer patients (6).

Finally, up-regulated HMGB1 was found in Gem-resistant pancreatic cancer cells, supporting, once again, the association of high HMGB1 and poor prognosis. HMGB1, as a nuclear transcription factor, is similarly recognized to regulate the expression of heat-shock protein 27 (HSP27), a key player in gemcitabine-resistance, implicating HMGB1 as the activator of the HSP27 pathway in Gem-resistance (46, 47).

HMGB1: A Potential Diagnostic and Prognostic Biomarker for Pancreatic Cancer?

It has been found that increased quantities of HMGB1, as well as low serum levels of sRAGE, are associated with an unsatisfactory response to therapy and worse prognosis. HMGB1 could, thus, be regarded as a potential diagnostic and prognostic/predictive biomarker for pancreatic cancer (48), although cancer antigen (CA) 19-9 (the only established serum biomarker for pancreatic cancer to date) and cytokeratin-19 fragments (CYFRA 21-1) were proven remarkably predictive in a similar patient setting (49). Even if immunogenic biomarkers are not as powerful as these molecules, they should still be taken into consideration with regard to chemotherapy response evaluation and patient prognosis (48).

However, HMGB1 has its limitations. Serum HMGB1 can be elevated in other cancers or in pancreatic inflammatory diseases, such as acute pancreatitis (50), although, in these inflammatory conditions, levels of HMGB1 are distinctively lower. To solve the previous problem, HMGB1 can be measured alongside with CA 19-9, increasing the specificity and sensitivity (33, 51).

Tissue samples of pancreatic cancer patients after surgical tumor resection were analyzed to investigate the prognostic role of immunohistochemical HMGB1 and HMGB2 expression. The study showed that nuclear HMGB1-positive and HMGB2-negative expression was a powerful and potentially independent negative indicator of disease-specific survival in operated patients and, by extension, a novel prognostic marker, especially within the first 2 years after surgery (52). Even though it is believed that HMGB2 has similar effects and roles as HMGB1 in cancer development, it functions reportedly as a potential tumor suppressor gene. It has consequently a key role in preventing malignant transformation by inducing cell-cycle arrest and/or apoptosis, especially in the setting of pancreatic cancer in a very similar way as P53, although it is not clearly described (52).

Regarding the HMGB1 receptor, RAGE levels measured in blood samples of pancreatic cancer patients undergoing systemic chemotherapy were found to be significantly higher in patients with stable or remissive disease (53). This fact correlates to the previous information as soluble RAGE acts differently from cellular RAGE preventing immunostimulatory effects rather than promoting them, i.e. functioning as a decoy receptor (34).