HMGB3 expression is finely regulated at the post-transcriptional level by a network of ncRNAs [miRNAs, lncRNAs and circular (circ)RNAs] through competitive endogenous RNA (ceRNA) mechanisms. It may eventually affect tumor progression and therapy resistance significantly, as summarized in Table I.

Multiple miRNAs (e.g., miR-101-5p, miR-142-3p, miR-205 and miR-758) may inhibit HMGB3 expression by directly targeting the corresponding 3′UTR. In this regard, it may effectively suppress cancer cell proliferation, migration and invasion in various tumor types, such as breast cancer, cervical cancer and non-small cell lung cancer (NSCLC) (12,28–30). In particular, beyond targeting HMGB3 directly, miR-142-3p can also induce mitochondrial dysfunction and promote apoptosis by inactivating the mTOR/STAT3 pathway (29).

In addition to direct regulation, lncRNAs (e.g., SNHG5, HOTTIP) and circRNAs (e.g., circFOXO3, circIGF1R, circN4BP2L2) can sequester corresponding miRNAs to upregulate HMGB3 expression directly, which may further promote the progression of malignancies such as nasopharyngeal carcinoma (NPC), NSCLC and CRC (31–37). In CRC, circIGF1R and circN4BP2L2 can remarkably stimulate tumor cell proliferation, invasion and glycolysis while inhibiting apoptosis through the activation of the HMGB3/Wnt/β-catenin signaling pathway (36,37). Under hypoxic conditions, downregulation of miR-200b-3p in cancer-associated fibroblast-derived exosomes can attenuate the suppression of HMGB3, in turn weakening the sensitivity of CRC cells to 5-fluorouracil (5-FU)-based chemotherapy (38).

Altogether, the multi-level regulation of HMGB3 expression by ncRNAs through ceRNA networks is essential in malignant progression and therapy resistance. Additional representative ncRNA-mediated regulatory axes involving HMGB3 across different cancers, including NSCLC, breast cancer, colorectal cancer, gastric cancer, and cervical cancer, are summarized in Table I (10,29,33,38–50).

Critically, despite less advanced research compared to that on HMGB1 and HMGB2, the function of HMGB3 from the HMG family may be dynamically regulated by PTMs (2,51,52). The current understanding of the PTMs of HMGB3 relies largely on homology-based inference. In particular, these modifications may play a role in cancer therapy resistance by affecting subcellular localization, DNA damage response (DDR) and extracellular release under stress conditions.

In HMGB1, acetylation and phosphorylation act as crucial regulators of nuclear-cytoplasmic trafficking and participation in DNA repair. Following these modifications, there may be reduced chromatin-binding stability, which may further boost the exposure of DNA damage sites and recruitment of repair factors, while driving nuclear export and extracellular release under conditions of severe stress, thereby influencing inflammatory signaling and cell survival (53–63). Similarly, the PTM status of HMGB3 may modulate DNA damage tolerance, replication stress response and post-chemotherapy or post-radiotherapy cell survival, considering that it retains highly conserved DNA-binding domains and regulatory regions (2). Furthermore, HMGB proteins may have a relationship with therapy resistance, given the redox regulation. To be specific, the oxidation state of cysteine residues in HMGB1 can determine the functional switch from a nuclear DNA chaperone to an extracellular signaling molecule, a process that is highly relevant under radiotherapy- or chemotherapy-induced oxidative stress (53,64–67). With the presence of conserved cysteine residues as well, HMGB3 may also have similar stress-induced functional reprogramming, potentially affecting DNA repair efficiency and TME-associated inflammatory responses, both of which are recognized to be key components of acquired therapy resistance (68). Additionally and significantly, there is so far limited research on the methylation and glycosylation of HMGB3. Furthermore, existing evidence indicates a negative correlation of HMGB3 DNA methylation with its expression in tumors (24). Despite the scarcity of direct experimental evidence, at the protein level, HMGB3 activity may be further modulated by lysine methylation and potential glycosylation sites within the B-box domain (69–72).

Collectively, even with limited direct experimental evidence for HMGB3 PTMs, PTMs may modulate HMGB3 activity during DDR and therapy-induced stress, as evidenced by homology- and structure-based functional insights. Consequently, it may participate in the modulation of tumor cell tolerance to radiotherapy, chemotherapy and targeted therapies potentially (2,51–53).

HMGB3 is a chromatin-associated protein overexpressed in multiple cancers, establishing an intimate association with CSC characteristics. CSCs represent a subpopulation of tumor cells that can undergo self-renewal and differentiation, exhibiting close relationships with tumor recurrence, metastasis and intrinsic therapeutic resistance (73–75).

Initially, HMGB3 was found to be preferentially expressed in hematopoietic stem cells, where it balances self-renewal and differentiation. For example, there are reduced common lymphoid progenitors and common myeloid progenitors in HMGB3-deficient mice, suggesting its potential role in maintaining primitive stem cell properties (76). Importantly, the proposed stemness-maintaining function appears to be conserved in malignancies. In solid tumors, HMGB3 can enhance spheroid formation, colony formation and the expression of pluripotency-associated transcription factors (e.g., NANOG, SOX2 and OCT4) to promote the formation of CSC-like phenotypes (27,77–79). In ovarian and breast cancer models, HMGB3 overexpression may indicate increased expression of stemness markers and unfavorable prognosis (77–80). Conversely, HMGB3 silencing can reduce self-renewal capacity and impair tumorigenicity. Its mechanisms may be related to the regulation of multiple stemness-associated signaling pathways. It can improve the expression of downstream targets [e.g., c-Myc and MMP7] by boosting β-catenin nuclear accumulation and transcriptional activity. Its deficiency may also increase the level of E-cadherin while reducing that of mesenchymal markers (e.g., Snail and vimentin), indicating its role in driving epithelial-mesenchymal transition (EMT) and reinforcing invasive phenotypes (21,27,81,82). In addition, HMGB3 may also activate MAPK/ERK signaling and cooperate with transcriptional regulators (e.g., SOX9) to enhance the transcription of NANOG, thereby stabilizing the pluripotency network (80,83–85).

HMGB3 can activate Wnt/β-catenin, MAPK/ERK and related regulatory axes collaboratively, thereby sustaining the CSC pool and enhancing tumor adaptability under therapeutic stress.

Genomic instability is a hallmark of cancer. Sustained activation of DDR pathways may enable tumor cells to survive under endogenous and exogenous genotoxic stress. HMGB3 is a chromatin-associated architectural protein that may contribute to DDR regulation at multiple functional levels, including damage sensing, chromatin remodeling, transcriptional control of repair genes and coordination of complex DNA lesion repair.

At the early stage of the DDR, rapid lesion recognition and chromatin relaxation constitute the premise of efficient repair to permit the recruitment of repair complexes. Depending on its HMG box domains, HMGB3 can bind distorted DNA structures and modulate chromatin conformation, thereby facilitating access of repair machinery to damaged sites. HMGB3 can recognize specific DNA adducts, such as N²-alkyl-2′-deoxyguanosine (dG) lesions induced by benzo[a]pyrene-7,8-dihydrodiol-9,10-epoxide (BPDE), and, together with SUB1 regulator of transcription, functions as a damage sensor exhibiting stereoselective repair activity toward trans-N²-BPDE-dG adducts (86). Thus, HMGB3 exerts potential roles in early lesion recognition and damaged DNA stabilization. Following the recognition of damage, HMGB3 can further promote repair progression by interacting with key repair enzymes. Through a direct interaction with poly(ADP-ribose) polymerase 1 (PARP1), it can enhance its PARylation activity and potentially influence its retention at DNA damage sites (9,87). By modulating PARP1 activity, HMGB3 can support efficient repair of single-strand breaks and alkylation-induced lesions. Beyond its structural role, HMGB3 can transcriptionally regulate major DDR components. By binding to its promoter, it can activate human telomerase reverse transcriptase (hTERT), resulting in telomere maintenance and genomic stability under stress (88,89). HMGB3 can also activate the transcription of ATR checkpoint kinase (ATR) and checkpoint kinase 1 (CHK1) (90–92), which may further strengthen checkpoint signaling, replication fork stabilization and homologous recombination repair. Consistently, the depletion of HMGB3 may attenuate ATR/CHK1 signaling and impair the efficiency of interstrand crosslink repair (8). These coordinated functions converge on the repair of interstrand crosslinks (ICLs) and DSBs, two complex DNA lesions. However, its deficiency may impair the deficiency of ICL repair (3,93) and DSB repair (3,94,95). Accordingly, HMGB3 can facilitate the recruitment or stabilization of core repair factors involved in homologous recombination and non-homologous end joining.

Collectively, HMGB3 works to maintain genomic stability by integrating chromatin remodeling, enzymatic activation and transcriptional amplification within the DDR network.

HMGB3 can modulate the balance between apoptosis and autophagy, serving as a critical indicator in tumor cell survival. There is growing evidence that HMGB3 can suppress apoptotic signaling while sustaining pro-survival autophagic activity, thereby promoting cellular adaptation under stress.

In breast cancer, there is an increased expression of Beclin-1, ATG family proteins and light chain 3-II, supporting the role of HMGB3 in maintaining protective autophagy (29). Suppression of HMGB3 by miR-142-3p can disrupt this autophagic program, leading to mitochondrial dysfunction, featured by the accumulation of reactive oxygen species (ROS) and the loss of mitochondrial membrane potential. These alterations may further activate the caspase cascade and shift cellular fate toward apoptosis, suggesting that HMGB3 can regulate autophagy through the preservation of mitochondrial integrity and prevention of cell death. Consistently, downregulation of HMGB3 can promote apoptotic signaling in multiple tumor types. For instance, in gastric cancer, reduced HMGB3 expression can induce G0/G1 cell cycle arrest and modulate the p53/p21 pathway, while decreasing the Bcl-2/Bax ratio to favor apoptosis (1). Similar HMGB3 silencing-induced pro-apoptotic effects have been reported in NSCLC (10,40), cervical cancer (96), ovarian cancer (47), CRC (37,44), thyroid cancer (97) and esophageal cancer (50), highlighting its broad role in sustaining tumor cell viability. Despite the scarcity of direct evidence linking HMGB3 to autophagy regulation, its structural similarity to HMGB1 suggests potential mechanistic parallels. HMGB1 can regulate autophagy by interacting with Beclin-1 (98), highlighting potential roles of HMGB3 in autophagosome formation or autophagy flux via related pathways. Of note, context-dependent effects were also noted. HMGB3 overexpression combined with selinexor can enhance apoptosis in myelodysplastic syndromes, possibly through the activation of cytoplasmic DNA-sensing pathways and interferon-related innate immune signaling (99). It underlines the complexity of HMGB3-mediated regulation of cell fate across different cellular contexts.

In addition to intrinsic genetic alterations, tumor progression may also be determined by dynamic interactions between malignant cells and their surrounding microenvironment. Therefore, HMGB3 functions as a critical regulator linking tumor cell plasticity to microenvironmental remodeling. HMGB3 can regulate cellular behavior, intercellular communication and oncogenic signaling networks coordinately, thereby benefiting the reshaping of the structural and functional landscape of the TME.

HMGB3 can enhance malignant cell plasticity, serving as a primary force in TME remodeling. Its overexpression may promote the proliferation and cell cycle progression across multiple tumor types. In leukemia, HMGB3 may enhance cell proliferation by activating MAPK/ERK signaling (13). In ovarian cancer, its overexpression may accelerate cell growth, whereas its silencing can induce G2/M arrest (80). In prostate cancer, depletion of HMGB3 may cause G0/G1 arrest by regulating cyclin D1, p21 and p27 (46). More importantly, HMGB3 may trigger EMT, a process that fundamentally alters tumor-stroma interactions. Its upregulation can escalate the expression of mesenchymal markers (e.g., N-cadherin, vimentin, β-catenin, snail and slug) (80), while its silencing may hinder cell migration and invasion in gastric cancer (14) and CRC (24). HMGB3 can strengthen EMT to disrupt tissue architecture, thereby facilitating dynamic cellular redistribution within the TME. Additionally, HMGB3 can also support CSC phenotypes. In breast cancer, it can increase mammosphere formation, while upregulating the expression of Nanog, SOX2 and OCT-4 (100), with similar effects observed in ovarian cancer (80). Maintenance of stem-like subpopulations can contribute to intratumoral heterogeneity and continuous microenvironmental adaptation. With respect to the above, HMGB3-driven cellular plasticity may offer a possible biological foundation for TME restructuring.

Beyond tumor cell-intrinsic changes, HMGB3 may also play a role in non-malignant components of the TME, among which angiogenesis represents a well-defined mechanism. In NPC, HMGB3, secreted via nuclear exosomes, may be internalized by endothelial cells to promote proliferation and tube formation (101). HMGB3-containing exosomes may increase microvascular density to facilitate the expansion of the vascular network within the TME. Furthermore, HMGB3 can modulate the immune microenvironment. Even with an insufficiency of the direct mechanistic data, its structural homology to HMGB1-an established damage-associated molecular pattern-may suggest potential immunoregulatory functions (18,102,103). In glioblastoma, the overexpression of HMGB3 may be related to reduced immune cell infiltration and an immunosuppressive microenvironment (104). Dysregulation of HMGB3 expression can also amplify ROS generation and activate the NF-κB signaling to mediate the production of cytokines such as VEGF and IL-6, thereby shaping the status of inflammation within the TME (1). The remodeling of extracellular matrix can further induce microenvironmental restructuring. HMGB3 can regulate MMPs, as evidenced by reduced MMP2 expression following HMGB3 knockdown in bladder cancer and decreased MMP2/MMP9 activity in gastric cancer, ultimately altering the physical and biochemical properties of the TME (1,105).

Diverse effects of HMGB3 on TME architecture are mediated through interconnected oncogenic signaling pathways. HMGB3 can activate the MAPK/ERK signaling in leukemia and ovarian cancer (13,80), promote β-catenin nuclear accumulation and downstream target (e.g., MMP7 and c-Myc) transcription (44,46,80,105), and modulate PI3K/AKT signaling, as revealed by miR-93-mediated suppression of HMGB3 in CRC (44). The interaction between HMGB3 and hypoxia-inducible factor (HIF)-1α in breast cancer also establish a relationship of HMGB3 to hypoxia-associated microenvironmental adaptation (100).

By integrating these signaling cascades, HMGB3 can coordinate malignant cell plasticity, stromal reprogramming and microenvironmental adaptation, thereby reshaping the organization and function of TME.

Chemotherapeutic agents may produce cytotoxic effects through diverse mechanisms and HMGB3 has been reported to modulate resistance in a drug-specific manner. Platinum compounds, such as cisplatin, can induce cytotoxicity primarily through the formation of DNA adducts that distort the DNA helix and activate DNA damage signaling pathways (1,4). HMGB3 is predominantly involved in DNA damage processing, supporting its implication in platinum resistance. Through potential binding to cisplatin-DNA adducts, it may facilitate lesion recognition and subsequent repair (1). In cisplatin-resistant ovarian cancer models, the inhibition of HMGB3 may enhance drug sensitivity and attenuate the activation of the ATR/CHK1/p-CHK1 axis, suggesting its role in supporting sustained DDR signaling under platinum-induced genotoxic stress (8). HMGB3 may also influence the clearance rate of cisplatin-DNA adducts by interacting with the cisplatin resistance-associated protein (also known as LUC7L3). HMGB3 knockdown may reduce the efficiency of adduct removal, thereby altering cellular responses to cisplatin exposure (93). Of note, in gastric cancer cells, HMGB3 silencing may amplify the sensitivity to cisplatin and paclitaxel but reduce the sensitivity to oxaliplatin (14,106), indicating the possible compound- and cellular context-dependent HMGB3-mediated modulation of platinum response. Microtubule-targeting agents represent another major class of chemotherapeutics. For example, paclitaxel can stabilize microtubules and prevent their depolymerization (107), while vincristine can inhibit tubulin polymerization and disrupt spindle formation (108). Alterations in cell cycle regulation and apoptotic signaling thresholds have been proven to be strongly associated with resistance to these agents. In gastric cancer, HMGB3 knockdown can suppress cell proliferation, induce G0/G1 arrest and enhance paclitaxel sensitivity by modulating p53, p21 and the Bcl-2/Bax ratio (14). Similarly, in cervical cancer, given reduced IC50 values, its depletion may enhance the sensitivity to both paclitaxel and vincristine (88). Collectively, HMGB3 enables the modulation of cell cycle progression and apoptosis-related pathways to promote cell survival under mitotic stress eventually. HMGB3 has been proposed to be associated with resistance to antimetabolites and endocrine therapies, in addition to the aforementioned agents. In CRC, by activating Wnt/β-catenin signaling and EMT-associated transcriptional programs, HMGB3 may mediate the resistance to 5-FU (109). In breast cancer, HMGB3 expression is negatively regulated by miR-27b, and its overexpression is associated with resistance to tamoxifen (41). Currently, although there is still an incomplete definition of the mechanistic details, there is reason to believe that HMGB3 exerts a broader role in shaping chemotherapeutic responsiveness through transcriptional and signaling reprogramming.

Radiotherapy induces cytotoxicity primarily through the induction of DSBs. Therefore, the efficiency of DNA repair and the propensity to undergo post-damage apoptosis remain the major determinants for cellular radiosensitivity. Current evidence suggests that HMGB3 may induce radioresistance by regulating the HMGB3/hTERT axis. HMGB3, a transcriptional regulator, can bind to the hTERT promoter and enhance its expression. Elevated hTERT levels may be related to increased DNA repair capacity and reduced radiation-induced apoptosis. Conversely, HMGB3 knockdown may contribute to accumulated γH2AX foci, impaired DSB repair, and enhanced radiosensitivity in both in vitro and in vivo models (88). As a result, HMGB3 may possibly promote repair competence and restrict apoptotic execution to sustain radiation tolerance.

Targeted therapies frequently encounter resistance owing to pathway reactivation, compensatory signaling or DNA repair dynamic alterations. Via both protein-protein interactions and oncogenic signaling network modulation, HMGB3 appears to participate in these adaptive processes. Notably, HMGB3 can interact with PARP1 to affect its functional activity. Loss of HMGB3 can induce PARP1 ‘trapping’ at sites of DNA damage and weaken the activity of PARylation, highlighting the regulatory role of HMGB3 in PARP1 DNA-binding kinetics. Such modulation can potentially attenuate the cytotoxic effect of olaparib and other PARP inhibitors (PARPi) (9,110). Besides, given the common function of the MAPK/ERK pathway as a bypass pathway in targeted therapy failure, HMGB3-mediated activation of the pathway-reported in multiple tumor types-may induce adaptive resistance (80,83–85). However, there is still an insufficiency of direct causal evidence linking HMGB3 to resistance against specific kinase inhibitors, necessitating further mechanistic studies to clarify this relationship.

Immune checkpoint inhibitors (ICIs) may disrupt inhibitory signaling pathways that restrain T-cell activity to restore antitumor immunity. Emerging evidence indicates that, on the basis of both tumor-intrinsic signaling alterations and microenvironmental modulation, HMGB3 may boost immune evasion and reduce responsiveness to ICIs. In triple-negative breast cancer, HMGB3 can suppress interferon (IFN)-γ-induced STAT1 phosphorylation and IFN regulatory factor 1 expression while enhancing STAT3 activation. Concurrently, it can also upregulate ferroptosis-inhibitory proteins [e.g., solute carrier family 7 member 11 (SLC7A11), glutathione peroxidase 4 and SLC3A2] to accumulate lipid ROS and restrict IFN-γ-mediated ferroptotic cell death. These molecular changes are associated with the resistance to anti-programmed cell death 1 (PD-1) therapy (111). In glioblastoma, elevated expression of HMGB3 is associated with a non-inflammatory, immune-excluded microenvironment characterized by reduced immune cell infiltration (104). Such an immunologically ‘cold’ tumor phenotype may compromise the efficacy of ICIs, further implicating HMGB3 in immunotherapy resistance.

Overall, HMGB3 can mediate the resistance to chemotherapy, radiotherapy, targeted therapy and immunotherapy in a context-dependent manner. Instead of functioning through a single dominant mechanism, HMGB3 may enhance tumor survival under therapeutic pressure by integrating DNA repair regulation, cell cycle control, apoptotic modulation, signaling pathway activation and immune adaptation. Nonetheless, in order to determine whether HMGB3 can serve as a predictive biomarker or actionable therapeutic target for overcoming treatment resistance, further mechanistic validation and clinical correlation are required. Representative HMGB3-mediated mechanisms contributing to resistance to chemotherapy, radiotherapy, targeted therapy and immunotherapy are summarized in Table II.

Targeting HMGB3 directly is the best approach for precision therapy. Through antisense oligonucleotides (ASOs) or small interfering RNA (siRNA), the specific silencing of HMGB3 has been documented to possess significant antitumor activity in multiple cancer models, which can hinder the proliferation and invasion while enhancing chemosensitivity (24,80,88). Such findings provide proof-of-concept for developing ASO- or siRNA-based HMGB3-targeted therapeutics.

The function of HMGB3, a DNA-binding protein and transcription factor, depends on specific domains and protein interactions. Small-molecular compounds developed by targeting its key functional sites (e.g., the HMGB domain or protein-protein interaction interfaces) may block its pro-cancer activities. For example, HMGB3can promote PARPi resistance by interacting with PARP1 in ovarian cancer (9), enhance radioresistance by binding the hTERT promoter in cervical cancer (88) and promote breast cancer growth by interacting with HIF-1α (100). Developing inhibitors that specifically block these interactions can precisely suppress specific oncogenic functions of HMGB3.

HMGB3 may exist extracellularly, although it is primarily nuclear, offering antibody targets. In NPC, HMGB3 is secreted via nuclear exosomes, and circulating nuclear exosomes HMGB3 may link to angiogenesis and metastasis (101), suggesting the potential of neutralizing antibodies in impeding extracellular HMGB3 functions. The high expression of HMGB3, even without extracellular activity, in specific cancers warrants the development of antibody-drug conjugates targeting cell surface markers associated with HMGB3 expression, thereby enabling precise drug delivery.

miRNAs are recognized as post-transcriptional regulators that may participate in HMGB3 regulation, which can be modulated to indirectly influence HMGB3 levels. In prostate cancer, HMGB3 can be negatively regulated by miR-205-5p and its elevated expression is associated with poor outcomes (112). In breast cancer, by targeting HMGB3, miR-27b is linked to tamoxifen resistance. Thus, targeting miR-27b or HMGB3 may reverse the resistance to tamoxifen (109), indicating the miR/HMGB3 axis serving as a potential therapeutic target (113).

HMGB3 may activate or participate in multiple oncogenic signaling pathways. In ovarian cancer (92) and leukemia (13), HMGB3 can promote malignancy and stemness via MAPK/ERK. Significantly, MEK/ERK inhibitors (e.g., AZD6244, PD0325901) can effectively reverse HMGB3-induced pro-cancer effects (80). It may also promote disease progression by activating pathways such as Wnt/β-catenin, PI3K/AKT, hypoxia/HIF-1α (44,100,114,115). Inhibiting these pathways represents an effective indirect strategy targeting HMGB3.

Given the central role in resistance and stemness, HMGB3 may be an ideal combination therapy target. Targeted HMGB3 inhibition combined with PARPi can restore sensitivity to PARPi, thereby overcoming resistance (9). When combined with chemotherapy, targeting HMGB3 can enhance the sensitivity to multiple agents (e.g., paclitaxel, cisplatin) (9,109). While combined with radiotherapy, it can enhance radiation response by blocking the HMGB3/hTERT axis. When used jointly with immunotherapy, it may overcome HMGB3-mediated anti-PD-1 resistance by inhibiting IFN-γ-driven ferroptosis in triple-negative breast cancer. In addition, HMGB3 can promote the stemness (80) and CSC-associated EMT in ovarian cancer via modulating the Wnt/β-catenin pathway (24). Accordingly, targeting HMGB3 may suppress or eliminate CSCs, reducing the risks of recurrence and metastasis.

HMGB3 can also be regarded as a promising biomarker for diagnosis, prognosis and treatment response prediction, given its abnormal expression in multiple cancers and strong association with clinical outcomes. HMGB3 is overexpressed in various tumors (e.g., breast cancer, NSCLC, CRC, bladder cancer, prostate cancer), with elevated levels being associated with poor prognosis. Furthermore, its expression also exhibits a positive association with advanced clinicopathological features, such as tumor grade, size, stage and lymph node metastasis (105,109,114,116). In breast cancer, HMGB3 shows excellent diagnostic potential, with an area under the receiver operating characteristic curve of 0.932 (109). In NPC and thyroid cancer, circulating HMGB3 in exosomes or serum may imply metastasis, suggesting the potential of liquid biopsy (101,115). Elevated HMGB3 may predict resistance to radiotherapy, chemotherapy, targeted therapy and ICIs, as well as unfavorable survival, underscoring its role in treatment response prediction and prognostic assessment.