HMGB1 in Cancer: Good, Bad, or Both?

M.T. Lotze received a commercial research grant from Prometheus and is a consultant/advisory board member of Neumedicine, Celgene, and Immunocellular Therapeutics, Ltd. No potential conflicts of interest were disclosed by the other authors.

The members of the planning committee have no real or apparent conflict of interest to disclose.

Upon completion of this activity, the participant should have a better understanding of the role of high mobility group box 1 (HMGB1) in cancer and the new therapeutic development targeting HMGB1.

This activity does not receive commercial support.

High mobility group box 1 (HMGB1) was first isolated and characterized in calf thymus in 1973 and is named for its electrophoretic mobility in polyacrylamide gels. HMGB1 contains two DNA-binding HMG-box domains (N-terminal A and central B) and an acidic C-terminal tail (Fig. 1A). In most cells, HMGB1 is located in the nucleus, where it acts as a DNA chaperone to help maintain nuclear homeostasis. HMGB1 was later discovered to express on cell surface membranes, cytosol, and mitochondria and release into the extracellular space. HMGB1 has many biologic functions inside as well as outside the cell (Fig. 1B) and plays a significant role in many diseases, especially inflammatory diseases and cancer (1–3).

Cancer development is a multistep process. As cells become more abnormal, they gain new capabilities. In 2011, Hanahan and Weinberg described 10 functional capabilities of cancers that they called “the hallmarks of cancer” (Fig. 2A; ref. 4). Evidence that HMGB1 dysfunction is associated with each hallmark of cancer and contributes to cancer development and therapy is increasing (1). However, HMGB1 has paradoxically been reported to promote both cell survival and cell death by regulating multiple cancer signaling pathways (Fig. 2B). This review describes recent advances in our understanding of HMGB1 regulation and function that affect cancer biology and influence the strategies that target HMGB1 for the prevention and treatment of cancer.

HMGB1 is stored in the nucleus as a result of the presence of two lysine-rich nuclear localization sequences (NLS) located in the A box and in the B box (Fig. 1A). Hyperacetylation of NLS promotes HMGB1 translocation from the nucleus to the cytosol and the subsequent release of HMGB1. The HMG boxes enable HMGB1 to bind different DNA structures without sequence specificity and act as a DNA chaperone. HMGB1 is the structural protein of chromatin and regulates nuclear homeostasis and genome stability in several ways (Fig. 1B).

Nucleosome is the basic unit of chromatin, consisting of a short length of DNA wrapped around a core of histone proteins. HMGB1 binds to nucleosomes at the dyad axis, promotes nucleosome sliding, relaxes nucleosome structure, and makes chromatin more accessible by its ability to bend DNA (5).

HMGB1 knockout mice show a defect in the transcriptional enhancement of the glucocorticoid receptor and die shortly after birth. HMGB1 has been found to increase the binding affinity of many sequence-specific transcription factors to their cognate DNA, such as p53, p73, the retinoblastoma protein (Rb), NF-κB, and the estrogen receptor.

Loss of HMGB1 increases DNA damage and decreases DNA repair efficiency in response to chemotherapy, irradiation, and oxidative stress. HMGB1 directly binds to a variety of bulky DNA lesions and allows it to participate in DNA repair pathways, including nucleotide excision repair, base excision repair, mismatch repair, and double-strand break repair via nonhomologous end-joining (6).

V(D)J recombination is a mechanism of genetic recombination in the early stages of immunoglobulin and T-cell receptor production of the immune system. HMGB1 is required for efficient V(D)J recombination by enhancing recombination-activating gene (RAG) binding to the recombination signal sequences (RSS).

A telomere is a region of repetitive nucleotide sequences at the end of a chromosome. Loss of HMGB1 reduces telomerase activity, decreases telomere length, and increases chromosomal instability.

In addition to its role in the nucleus, HMGB1 also functions as a damage-associated molecular pattern (DAMP) when passively released from dead, dying, or injured cells or when actively secreted from immune cells or cancer cells in response to exogenous and endogenous stimuli, such as endotoxin, CpG DNA, double-stranded RNA (dsRNA), TNF-α, interleukin (IL)-1, IFN-γ, hydrogen peroxide, ATP, and hypoxia (Fig. 3). In addition, macrophage engulfment of apoptotic cells may induce significant active HMGB1 release, suggesting a direct interplay between dying cells and immune cells, which also induces HMGB1 release (7).

Depending on the inducing stimulus, the mechanism of HMGB1 secretion and release can vary. HMGB1 lacks a leader signal sequence, and its secretion needs a nonclassical, secretory lysosome-mediated exocytosis (8). Several mechanisms have been recently proposed to explain the mechanism of HMGB1 secretion and release.

Following stimuli, the HMGB1 protein is modified by different posttranscriptional modifications (PTM), such as acetylation, ADP-ribosylation, methylation, phosphorylation, and oxidation, which regulate HMGB1 secretion. However, we still do not know whether these PTMs are competitively, cooperatively, or independently regulated (Fig. 4A).

Several of the secondary messengers, such as cytosolic-free calcium, reactive oxygen species (ROS), and nitric oxide, regulate HMGB1 secretion (Fig. 4A).

Chromosome-region maintenance 1 (CRM1) directly mediates HMGB1 export from the nucleus (ref. 9; Fig. 4A).

Pyroptosis is an inflammatory cell death and is typically triggered by caspase-1 after its activation by various inflammasomes. dsRNA-dependent protein kinase (PKR) is implicated in inflammation and immune dysfunction by interfering with many signaling pathways (10). PKR-mediated inflammasome activation is required for DAMP and pathogen-associated molecular pattern (PAMP)–induced HMGB1, IL-1β, and IL-18 release in macrophages (Fig. 4B).

An early study suggests that the chromatin of apoptotic cells sequesters HMGB1 and prevents inflammation (11). However, HMGB1 also can be released by apoptotic cells at a late stage (12). It has been shown that nuclear DNA and histones are released during apoptosis, and they are well-known binding partners of HMGB1 in the nucleus. The mechanism for HMGB1 release undergoing apoptosis partly involves caspase-3/7–mediated mitochondrial ROS production (ref. 13; Fig. 4C).

HMGB1 release occurs passively as cell permeability breaks down upon necrosis (11). In addition, PARP1-mediated ATP deletion regulates HMGB1 release in necrosis (14), suggesting that PARP1 is a potential therapy target in inflammatory disease (Fig. 4D).

Autophagy is a lysosome-mediated, self-degrading process and plays an important role in sustaining cellular homeostasis under stress. The ATG5-mediated classical autophagy pathway has been reported to promote HMGB1 release in starvation and lipopolysaccharide (LPS) treatment (15, 16). This process requires ROS signaling (ref. 15; Fig. 4E).

Once released, extracellular HMGB1 binds to several cell surface receptors to activate the downstream signaling pathway [e.g., NF-κB, IFN regulatory factor-3 (IRF3), and phosphoinositide 3-kinase (PI3K)] to produce a functional response, such as activation of innate immune cells, induction of proinflammatory cytokines and type I IFNs, stimulation of cell adhesion and migration, inhibition of phagocytosis, promotion of cell proliferation and angiogenesis, and induction of autophagy (17, 18). These receptors include the receptor for advanced glycation end products (RAGE), Toll-like receptors (TLRs; such as TLR2, TLR4, and TLR9), Mac-1, syndecan-1 (CD138), phosphacan protein-tyrosine phosphatase (PPTP)-ζ/β, CD24, chemokine (C–X–C motif) ligand 4 (CXCL4), T-cell immunoglobulin mucin-3 (TIM-3), and possibly others. Of these receptors, CD24 and TIM-3 act as negative receptors and inhibit immune activity of HMGB1 in macrophages and tumor-associated dendritic cells (TADC), respectively (19, 20). Apart from a direct receptor interaction, HMGB1 may form heterocomplexes with other immune coactivators such as IL-1, CXCL12, DNA, nucleosome, or LPS that generate synergistic responses in inflammation and immunity.

In 1995, RAGE was the first receptor shown to bind HMGB1 (21). At the time, the functions of RAGE in HMGB1 signaling were unknown, but it was later discovered that HMGB1 signaling through RAGE mediates chemotaxis and migration, proliferation, and differentiation of immune and cancer cells and upregulation of cell-surface receptors and autophagy. The cytoplasmic tail is believed to be essential for intracellular signaling of RAGE, including RAGE-mediated energy metabolism (22). In addition, RAGE provides a functional platform for cross-talk with other HMGB1 receptors. For example, interplay between RAGE and TLR9 is critical for HMGB1-DNA complex-activated immune responses in dendritic cells (23). Interplay between Mac-1 and RAGE is required for HMGB1-mediated adhesive and migratory neutrophil functions (24). Interplay between syndecan-1, PPTP-ζ/β, and RAGE is required for neurite outgrowth.

Apart from RAGE, HMGB1 binding of TLR2 and TLR4 also results in NF-κB activation. TLR4 may be more important for HMGB1-induced macrophage activation and proinflammatory cytokine release (25). TLR4-deficient animals are significantly protected from ischemia–reperfusion injury to the liver, kidney, and heart, suggesting that TLR4 plays a critical role in sterile inflammation (26). Determining the functional relationship between RAGE and TLR4 in inflammation and immunity is still of great interest.

Recent studies underscore the importance of redox modification in the regulation of HMGB1 translocation, release, and activity in disease (27). Three cysteines are encoded within the HMGB1, two vicinal cysteines in box A (C23 and C45), and a single one in box B (C106; Fig. 1A). Replacement of Cys23 and/or 45 with serines did not affect the nuclear distribution of the mutant proteins, whereas C106S and triple cysteine mutations impaired nuclear localization of HMGB1, allowing entry of some of the protein into the cytosol. Moreover, increased endogenous and exogenous ROS promotes HMGB1 translocation and release (ref. 28; Fig. 4).

Of note, the redox status of HMGB1 distinguishes its cytokine-inducing and chemokine activity (ref. 29; Fig. 3). Initial studies suggest that reduced C106 is necessary for the binding of HMGB1 to TLR4 to stimulate cytokine release and inflammation (25). A recent study suggests that a disulfide bond between C23 and C45 is also required for HMGB1 cytokine activity (29). Mutations of C45 or C23 abolish the cytokine activity of HMGB1. In contrast, all-cysteine–reduced HMGB1 does not have TLR4-dependent cytokine activity but binds to CXCL4 to induce inflammatory cell recruitment and chemotaxis by the CXCL12 receptor (30). ROS oxidizes the HMGB1 at C106 released from apoptotic cells, thereby neutralizing its cytokine-inducing activity and promoting tolerance in dendritic cells (13). Finally, all-cysteine–oxidized HMGB1 prevents HMGB1′s cytokine or chemotactic activity (29). Thus, redox modifications are crucial for HMGB1 functionality as a mediator during infection and sterile inflammation. As described below, redox also regulates HMGB1 function and activity in autophagy.

It is clear that autophagy, literally meaning “self-eating,” is not only a crucial biologic process from physiology to pathology, but is also a promising and daunting target for several diseases, including inflammatory disease, neurodegeneration, and cancer. The role of autophagy in cancer is complex and is likely dependent on tumor type, stage, genetic context, and tumor microenvironment (TME; ref. 31). On one hand, autophagy acts as a tumor suppressor by preventing genome instability, limiting oxidative stress, reducing intratumoral necrosis-dependent inflammation, and inhibiting angiogenesis (Fig. 5A). On the other hand, autophagy acts as a survival mechanism that can promote the growth of established tumors and resistance to anticancer treatment by providing substrates for metabolism, removing damaged organelles and proteins, and diminishing apoptosis (Fig. 5A). Thus, defining the context-specific role for autophagy in cancer and the mechanisms involved will be important to guide autophagy-based therapeutic intervention.

During autophagy, double-membrane structures called autophagosomes engulf cytosol or organelles and deliver them to lysosomes (called autolysosomes) to be degraded and recycled (Fig. 5A). It is now known that the complex and fine-tuned process of autophagy is regulated by many proteins, including HMGB1. HMGB1 has transcription-dependent and -independent proautophagic activities (Fig. 5B). HSPβ1, a regulator of actin cytoskeleton dynamics, is a direct transcriptional target of HMGB1 in autophagy (32). Suppression of expression of HSPβ1 and HMGB1 inhibits the dynamics of autophagy as well as mitophagy, a form of selective autophagy to remove damaged mitochondria (32). During stress, including starvation, oxidative stress, and chemotherapy, HMGB1 translocates from the nucleus to the cytoplasm and binds to Beclin 1, which initiates autophagosome formation (15). Furthermore, the intramolecular disulfide bridge (C23–C45) of HMGB1 is required to bind to Beclin 1 and sustain autophagy (15). Once released, reduced HMGB1 triggers autophagy in a RAGE-dependent manner in cancer cells, which promotes tumor cell proliferation (33). In contrast, oxidized HMGB1 induces apoptosis (33). Thus, both intracellular and extracellular HMGB1 are critical in regulating autophagy (Fig. 5B). Given the dual role of autophagy in cancer, HMGB1-mediated autophagy may have a similar role in cancer biology that needs further investigation.

Tumor-infiltrating leukocytes, as well as cytokine-related signaling pathways, are critical components in the development of the inflammatory TME. The inflammatory TME can promote neoplastic transformation and support tumor growth, invasion, and metastases. Infiltrating leukocytes, as well as the cancer cells themselves, have the ability to secrete HMGB1 under hypoxia, injury, inflammatory stimuli, or environmental factors (34). In turn, extracellular HMGB1 can activate proinflammatory signaling pathways, such as the NF-κB and inflammasome pathways, to induce proinflammatory cytokine release. This loop will accelerate inflammatory responses, tumor formation, and metastasis. Both RAGE and TLR4 are involved in the HMGB1-mediated inflammatory response in tumorigenesis (35, 36), but the cross-talk between these receptors is not well known.

Cancer requires more energy that allows for rapid, invasive, and metastatic growth. Although normal cells generate ATP through a combination of oxidative and glycolytic metabolism, cancer cells dramatically accelerate and reprogram their metabolism to better fit the energy requirement. This is one of cancer's most common phenotypes, called the Warburg effect. HMGB1 has been implicated in tumor energy metabolism (22, 32). Recombinant exogenous HMGB1 or endogenous HMGB1 in necrotic tumor cell lysates increased ATP production and pancreatic tumor cell proliferation, providing a direct link between inflammation and energy metabolism with the TME (22). Extracellular HMGB1 increases mitochondrial RAGE expression and translocation, which in turn increases mitochondrial complex I activity and ATP production (22). Loss of RAGE in mice prevents tumorigenesis in the skin, pancreas, and intestine (35, 37, 38) and inhibits ATP production for tumor growth (37). Loss of HMGB1 increases mitochondrial injury and decreases ATP production (32). The cross-talk between the HMGB1–RAGE signaling pathway and other key cancer metabolic signaling pathways is still unknown.

Most cancer deaths are caused by tumor invasion and metastasis rather than the primary tumor itself. In the clinic, expression of RAGE is closely associated with the invasive and metastatic activity of cancers such as gastric cancer (39) and colorectal cancer (40). In animals, inhibition of RAGE–HMGB1 interaction suppresses tumor growth and metastasis by activation of mitogen-activated protein kinases and the NF-κB pathway, which results in the expression of matrix metalloproteinases (MMP), such as MMP2 and MMP9 (41). MMPs degrade extracellular matrix proteins and play a major role in tumor invasion and metastasis. In vitro, suppression of RAGE and HMGB1 by antisense S-oligodeoxynucleotide or HMGB1 150–183 peptide (RAGE-binding motif) inhibits growth, migration, and invasion in cancer cells (42). These studies suggest that HMGB1–RAGE signaling plays a major role in tumor invasion and metastasis.

Cancer immunity surveillance is considered to be an important host defense process to inhibit carcinogenesis and to maintain cellular homeostasis. As a multifunctional cytokine, HMGB1 has been characterized with both immunosuppressive and immune-activation properties, which depend on receptors, targeted cells, and redox state (27, 43). Dendritic cells are professional antigen-presenting cells and constitute several subsets with distinct phenotypes and crucial roles in promoting both immune tolerance and immunity. HMGB1 has the ability to induce apoptosis in macrophage-derived dendritic cells, which diminish host anticancer immunity (44). In addition, tumor cell–derived HMGB1 suppresses naturally acquired CD8⁺ T-cell–dependent antitumor immunity, partly by enhancing tumor-associated regulatory T cells (Treg) to produce IL-10 (45). More recently, HMGB1 was found to promote tumor-infiltrating T-cell–expressed lymphotoxinα1β2, which leads to the recruitment of CD11b⁺F4/80⁺ macrophages into the tumor for tumor promotion by providing growth factors and supporting angiogenesis (46). The downstream signaling pathway of HMGB1-mediated tumor immune suppression needs to be elucidated in future studies.

A complex balance between pro- and antiangiogenic factors exists in TMEs and regulates vessel formation. In a fast-growing tumor, the diffusion distance from the existing vascular supply increases, resulting in hypoxia. The regulation of angiogenesis by hypoxia is well studied in cancer. Hypoxia causes translocation and release of HMGB1 and increases expression of RAGE in TME (47, 48). Once released, HMGB1 binds to RAGE and activates the NF-κB pathway to induce expression of proangiogenic growth factors (e.g., VEGF) and their receptors (49). Thus, using HMGB1 antibody or knockdown of RAGE inhibits tumor angiogenesis and metastasis in vitro and in vivo (49, 50).

Intracellular HMGB1 functions as a tumor suppressor in breast cancer by directly binding to Rb (ref. 51; Fig. 2B). Rb is a well-known tumor suppressor protein that is dysfunctional in many cancers. Rb has three functionally distinct binding domains and interacts with its regulatory proteins with a conserved LXCXE motif. HMGB1 binds to Rb by the LXCXE motif, which in turn enhances Rb-mediated transcription repression such as E2F and cyclin A1 (51). HMGB1 causes Rb-dependent G₁ arrest and apoptosis induction, and overexpression of HMGB1 inhibits Rb-positive breast cancer growth in vitro and prevents tumorigenicity in subcutaneous tumor models in vivo (51). These findings suggest that endogenous HMGB1, as a Rb-associated protein, suppresses breast tumorigenesis. Whether HMGB1 interacts with other tumor suppressors (e.g., p53) or oncogenes (e.g., KRAS) to regulate tumorigenesis is of great interest.

As a major driving force for tumorigenesis, genomic instability, which includes both numerical (e.g., aneuploidy) and structural chromosomal abnormalities, occurs early in the development of most cancers. HMGB1 has an important role in modulating genome stability. HMGB1 deficiency in cells results in genome instability (52, 53). Telomeres are essential for chromosome stability by protecting them from recombination and degradation. Loss of HMGB1 displays telomere shortening (52, 54). The mechanisms by which HMGB1 maintains telomere lengthening are involved in regulation of telomerase activity by binding to telomerase (54). However, the cooperative action of HMGB1 and telomerase in regulating telomere length and function still remains unknown. Topoisomerase IIα (Topo IIα), an enzyme that controls and alters the topologic states of DNA during transcription, is implicated in chromosome replication, segregation, and recombination. HMGB1 not only binds Topo IIα to stimulate its enzymatic activity but also promotes its expression (55). Moreover, the upregulation of Topo IIα by HMGB1 is inhibited by Rb (55), suggesting that interplay between HMGB1 and Rb regulates Topo IIα expression and subsequently, genome stability. In addition, HMGB1-mediated DNA damage repair contributes genome stability (6). These findings suggest that loss of HMGB1 in tissue may cause genome instability and tumorigenesis (Fig. 2B).

Several studies have indicated that defective autophagy-associated genes (e.g., Beclin 1, ATG5, UVRAG, and Bif-1) in mice increase genome instability, inflammation, oxidative stress, and mitochondrial injury, which contribute to tumorigenesis (56–59). It is well shown that autophagy is a negative regulator of inflammasome activation by promoting its degradation or diminishing ROS production. Autophagy-deficient mice lose their ability to remove damaged mitochondria, which in turn induces ROS production and inflammasome activation (60). However, the mechanism by which autophagy deficiency compromises genomic stability is still unclear. One possibility is that autophagy has the ability to remove damaged DNA, which in turn prevents chromosomal instability. Given that HMGB1 is a critical regulator of autophagy and mitophagy (15, 32), loss of HMGB1 leading to autophagy deficiency may cause genome instability and inflammation, which lead to tumorigenesis (Fig. 2B).

HMGB1-mediated immunogenic cell death (ICD) contributes to immune-mediated eradication of tumors during chemotherapy (e.g., anthracyclines) or radiotherapy (61–63). ICD is characterized by the release of dying cancer cells or cell surface exposure of DAMPs (e.g., calreticulin, HSPs, ATP, and HMGB1), which are helpful for the maturation, antigen uptake, and presentation of dendritic cells and serve as powerful immunologic adjuvants to active CTL response. Autophagy-mediated ATP release also contributes to ICD development (64). Blocking the HMGB1–TLR4 pathway inhibits ICD-associated anticancer immune responses upon chemotherapy both in vitro and in vivo (ref. 63; Fig. 2B). However, HMGB1 released from necrotic cancer cells treated with chemotherapy enhances regrowth and metastasis of remnant cancer cells in a RAGE-dependent way (65). Thus, blocking HMGB1–RAGE signaling increases the effectiveness of chemotherapy (66). These studies suggest that TLR4 in dendritic cell is critical for HMGB1-mediated ICD and tumor clearance, whereas RAGE in cancer cells is important for HMGB1-mediated survival after chemotherapy. Although both apoptotic and necrotic cells have the ability to release HMGB1, HMGB1 released from apoptotic cells is immunity tolerant (13). Moreover, HMGB1 binding to TIM-3 in TADCs is important for evasion of the immune system by tumor cells in response to chemotherapy and vaccines (20), suggesting that a balance mechanism between TLR4 and TIM-3 exists in dendritic cells (Fig. 2B). Of note, a recent study suggests that ICD-mediated tumor clearance works only in tumor cell line transplantation models or the immunogenic 3-methylcholanthrene fibrosarcoma model, but not in spontaneous mammary tumor models (67). In addition, TLR2, but not TLR4 in TADCs, mediates T-cell–dependent brain tumor regression in antibrain cancer immunotherapy (68). Thus, defining the context-specific role for extracellular HMGB1 in chemotherapy and immunotherapy, including ICD, and the mechanisms involved will be important to optimal therapeutic outcome.

Induction of cell death and inhibition of cell growth are the main targets of cancer therapy. A number of studies have shown that suppression of HMGB1 expression by RNA interference increased the anticancer activity of cytotoxic agents, whereas overexpression of HMGB1 expression by gene transfection increased drug resistance (69–71). HMGB1 expression regulates chemotherapeutic response and resistance by interfering with autophagy and the apoptotic pathway (Fig. 2B). HMGB1 increases prosurvival autophagy in a Beclin 1–dependent way in chemotherapy, whereas HMGB1 inhibits both intrinsic and extrinsic programmed cell death/apoptosis in a caspase-dependent way in cancer cells. In some cases, knockout HMGB1 in fibroblasts inhibits antimetabolite drug-induced apoptosis (72), suggesting that HMGB1 plays distinct roles in apoptosis depending on cell types. The cross-talk between apoptosis and autophagy regulates cell death and determines cell fate in anticancer therapy. Upregulation of apoptosis inhibits autophagy, whereas upregulation of autophagy inhibits apoptosis during chemotherapy. HMGB1 and p53 are capable of physical interaction, and a region of inducible structure in the p53 transactivation domain (residues 38–61) is the essential element for binding to the A box (residues 7–74; Fig. 1A; ref. 73). Interplay between HMGB1 and p53 regulates apoptosis and autophagy in clone cancer cells after treatment with DNA-damaging anticancer drugs (70). DNA damage promotes interactions between p53 and HMGB1 in the nucleus and cytoplasm. Loss of p53 increases cytosolic HMGB1, leading to increased binding to Beclin 1, thereby promoting autophagy and decreasing apoptosis. In contrast, loss of HMGB1 increases cytosolic p53 and apoptosis and decreases autophagy (70). These findings provide new insights into the HMGB1–p53 signaling and cancer cell response to DNA damage.

Apart from genetic inhibition or overexpression of HMGB1 expression in cancer cells, several HMGB1-targeting agents have been used in experimental cancer research. These agents include soluble RAGE (sRAGE), HMGB1-neutralizing antibody, A box protein, platinating agent, ethyl pyruvate, quercetin, and glycyrrhizin. sRAGE acts as a decoy to prevent RAGE signaling and has been used successfully in blocking the HMGB1–RAGE signaling pathway in animal tumor models (41). HMGB1-neutralizing antibody and A box protein can block activity of extracellular HMGB1 in tumor therapy (71). Platinating agents such as cisplatin and oxaliplatin have the ability to retain HMGB1 within the nucleus by conformational changes in the double helix to which HMGB1 binds quite stably (74). Ethyl pyruvate, the first HMGB1 inhibitor used in animal models of sepsis by inhibition of the NF-κB pathway, inhibits liver tumor growth (75). In addition, glycyrrhizin and quercetin, potential HMGB1 inhibitors by directly binding to HMGB1 or inhibition of PI3K, improve the effectiveness of anticancer agents in several different tumor models (33, 71). Further investigation is needed to evaluate these therapies and their possible role in clinical practice.

We have described how HMGB1 seemingly plays both oncogenic and tumor-suppressing roles during tumor development and therapy (Fig. 2B). In many cases, extracellular HMGB1 acts as a protumor protein due to its cytokine, chemokine, and growth factor activity, whereas intracellular HMGB1 acts as an antitumor protein due to its ability to sustain genome stability and autophagy activity during tumor growth. In contrast, extracellular HMGB1 may contribute to ICD-associated antitumor immunity in the early stages of chemotherapy but promote residual tumor cell survival in the late stages of chemotherapy. Moreover, suppression of intracellular HMGB1 expression inhibits autophagy and increases apoptosis, which boosts the effectiveness of anticancer treatment. Thus, the strategies that target HMGB1 for the prevention and treatment of cancer may vary at different stages. Perhaps we should be thinking about the design of tumor-selective modulators of HMGB1. Given the major role of HMGB1 in the nucleus, HMGB1 could be a sensible target for hereditary cancers such as breast cancer that are characterized by the presence of genomic instability (51). Knowledge of the signal transduction and regulation of HMGB1 release is one of the hot topics in immunity and cancer research. HMGB1 can induce immune responses by itself, as well as in association with other endogenous and exogenous molecules. In addition, dynamic PTMs of HMGB1, as well as receptor selection, might shift their activity from immunity activation to immunity tolerance. However, the roles of PTMs of HMGB1 in tumorigenicity and anticancer therapies still remain largely unknown. In addition, it remains unclear whether HMGB1 mutations exist. Further work is needed to understand the molecular basis and context-dependent role of HMGB1 as an anti- or protumor protein in cancer. Also, it is important to establish HMGB1-associated spontaneous tumor models, although many xenograph or transplantable tumor models have been used.

Conception and design: H.J. Zeh III, M.T. Lotze

Writing, review, and/or revision of the manuscript: R. Kang, Q. Zhang, H.J. Zeh III, M.T. Lotze, D. Tang

Study supervision: M.T. Lotze

The authors thank Christine Heiner (Department of Surgery, University of Pittsburgh) for her critical reading of the article. The authors apologize to the researchers whose work was not cited because of space limitations.

This work was supported by the NIH (R01CA160417 to D. Tang) and internal support from the University of Pittsburgh Department of Surgery (to D. Tang).