The roles of the mtDNA-cGAS-STING axis in tumor immunity: from immune activation to immune evasion.

1
Introduction
The TME is a dynamic ecosystem surrounding tumor cells, comprising immune cells (such as TAMs and regulatory T cells), stromal cells, vascular systems, and cytokines. It plays a central role in tumor initiation, progression, and immune evasion (1–3). As a complex ecosystem, the TME relies on DNA-sensing mechanisms to act as a key link between cellular stress and immune responses (4). mtDNA, a unique immunostimulant, leaks into the cytoplasm from damaged mitochondria when cells are subjected to genomic instability, metabolic stress, or therapy-induced damage (5). This cytoplasmic mtDNA can be recognized by cGAS, which in turn activates the STING-IRF3 signaling pathway, ultimately triggering an inflammatory response (6). Notably, this DNA-sensing mechanism not only participates in antitumor immune responses but is also closely associated with tumor immune evasion (5).
As a circular double stranded DNA (dsDNA) molecule, mtDNA is effectively recognized by DNA sensors due to its structural characteristics and plays a particularly important role in activating the cGAS-STING pathway (7). It is worth noting that the cGAS-STING pathway activated by mtDNA plays a dual role in tumor immunity. On the one hand, this pathway exerts anti-tumor effects by activating host immunity (8, 9); On the other hand, certain tumors may use this pathway to evade immune surveillance (5, 10, 11).
This article focuses on the core scientific question of the dual role of mtDNA-cGAS-STING axis in tumor immunity. We discussed the potential mechanisms by which this pathway participates in immune activation and immune escape, and summarized the drugs currently under research targeting this pathway.

2
Mechanisms of mtDNA release and cGAS-STING pathway activation
2.1
Pathways of mtDNA release induced by mitochondrial stress
mtDNA is typically sequestered within the mitochondrial matrix and enclosed by mitochondrial nucleoids (12). Under stress conditions, however, it can be released from damaged mitochondria into the cytoplasm and extracellular space, serving as a key immunostimulatory signal. This release is triggered by various stressors, including oxidative stress, viral infection, and drug toxicity (9), that activate distinct but often overlapping mechanisms. These mechanisms can be classified into four major pathways based on the underlying cellular process: (i) Membrane permeabilization pathways represent primary routes for mtDNA release. Mitochondrial outer membrane permeabilization (MOMP), mediated by BAX/BAK-dependent membrane depolarization (13, 14) and promoted by PGAM5-mediated Bax translocation (15), is prominently induced by severe oxidative stress, viral infection, and apoptosis-inducing drugs. Alternatively, mitochondrial inner membrane permeability transition (MIMP) can independently mediate mtDNA release through progressive widening of outer membrane pores coupled with increased inner membrane permeability, a process particularly associated with calcium overload and metabolic toxicity (13, 16). (ii) Vesicular transport provides a membrane-preserving mechanism for mtDNA export. Mitochondrial-derived vesicles (MDVs) selectively transport mtDNA to the cytoplasm, serving as an alternative release pathway activated by metabolic stress conditions (e.g., fumarate accumulation) and during viral infection (16). (iii) Quality control failure leads to catastrophic mtDNA release. Defective mitophagy causes accumulation of damaged mitochondria, resulting in spontaneous mtDNA release through organelle rupture under conditions of prolonged stress, aging, and drug toxicity (17). (iv) Direct molecular destabilization of the mtDNA-nucleoid complex facilitates leakage. Downregulation of mitochondrial transcription factor A (TFAM), triggered by nutrient stress and viral infection, impairs mtDNA stability and promotes its release (7, 18). Overexpression of Drp1 induces mitochondrial dysfunction and stress-induced leakage, particularly under bioenergetic stress conditions (19). Additionally, excessive oxygen species (ROS) production, a hallmark of oxidative stress, oxidizes mtDNA and directly facilitates its translocation into the cytoplasm and extracellular environment (20, 21).

2.2
Involvement of mtDNA in the activation of the cGAS-STING pathway
The cGAS- STING pathway is a core component of the innate immune system that specifically recognizes cytoplasmic dsDNA to initiate host immune responses (22). This pathway plays a crucial role in tumor immunity, antiviral responses, and autoimmune diseases (23). mtDNA in the cytosol is specifically recognized by cGAS, an enzyme that possesses unique structural features as a dsDNA sensor (24). cGAS forms a complex with negatively charged mtDNA via its positively charged DNA-binding domain; this binding induces a conformational change in cGAS and exposes its catalytic pocket (25). Experimental evidence has shown that either digestion of mtDNA by DNase I or reduction of cytosolic mtDNA using ethidium bromide (EtBr) can significantly inhibit cGAS activation (7, 26, 27). Notably, oxidatively modified mtDNA exhibits stronger cGAS-binding capacity, suggesting that oxidative stress may amplify this signaling pathway (21, 28).
Upon activation, cGAS produces the second messenger cyclic GMP-AMP (cGAMP), which binds to endoplasmic reticulum (ER)-localized STING protein and induces its conformational change (29). Activated STING translocates to the Golgi apparatus region, where it recruits and phosphorylates TANK-binding kinase 1 (TBK1) (25). Subsequently, TBK1 phosphorylates the transcription factor IRF3, promoting the dimerization of IRF3 and its translocation to the nucleus (30, 31). This cascade reaction can be affected by multiple regulatory mechanisms: epigenetic silencing of STING blocks signal transduction (5), while overexpression of mitochondrial TFAM reduces STING activation by stabilizing mtDNA (32, 33). Super-resolution imaging technology has confirmed a clear spatiotemporal correlation between mtDNA release and STING activation (13).
After translocating into the nucleus, activated IRF3 initiates the transcription of type I interferon (IFN-α/β) genes (25). These interferons activate the JAK-STAT pathway through autocrine and paracrine effects, inducing the expression of hundreds of interferon-stimulated genes (ISGs) (5, 34, 35). Overall, upon binding cytosolic mtDNA, cGAS synthesizes the second messenger 2’3’-cGAMP, activating STING on the ER membrane. STING then recruits TBK1 and IRF3, triggering IRF3 phosphorylation and nuclear translocation to drive IFN-I transcription. IFN-I is a pleiotropic cytokine with antiviral, antiproliferative, and immunomodulatory functions. In tumors, IFN-I can activate anti-tumor immune responses by recruiting natural killer cells and expanding CD4+ and CD8+ T cells; conversely, mild and persistent IFN-I signaling can lead to immunosuppression by inducing mediators involved in T cell exhaustion, such as PD-L1, IDO, and IL-10 (36). Studies have shown that STING-deficient mice exhibit a significantly attenuated inflammatory response (37, 38), whereas the introduction of exogenous mtDNA can restore IFN-β production (39, 40). Notably, this pathway can also synergistically activate the NLRP3 inflammasome, facilitating the release of pro-inflammatory cytokines such as IL-1β (21, 41).

3
mtDNA-cGAS-STING axis promotes tumor immune activation (Figure 1)
3.1
Mechanisms of NK cell and T cells activation
NK cells are cytotoxic lymphocytes with the ability to kill tumor cells and secrete pro-inflammatory cytokines (42). Owing to their role in tumor suppression, NK cells play a critical function in tumor immune surveillance, particularly in preventing tumor metastasis (43). The mtDNA-cGAS-STING pathway activates cytotoxic immune cells through multiple mechanisms. In NK cells, mtDNA released by tumor cells partially triggers intrinsic STING activation in NK cells via recognition by cGAS, thereby maintaining the antitumor activity of the TCF-1+ NK cell subset (a subset with long-term memory potential) (8).
T cells exhibit multiple mechanisms of action in tumor therapy and provide robust support for tumor treatment through various approaches, including directly killing tumor cells, regulating immune responses, enhancing immune memory, and participating in immune checkpoint inhibition (44, 45). In particular, CD8+ T cells can directly recognize and kill tumor cells: they identify complexes of tumor-specific antigens (TSAs) or tumor-associated antigens (TAAs) with MHC class I molecules on the surface of tumor cells via T cell receptors (TCRs). Once recognition occurs, cytotoxic T lymphocytes (CTLs) release cytotoxic granules, such as perforin and granzyme, which can induce tumor cell apoptosis (46). In CD8+ T cells, IFN-I produced after STING pathway activation can promote their proliferation and the expression of cytotoxic granules (47). Experimental evidence shows that direct activation of STING using cyclic GMP-AMP (cGAMP) significantly enhances the IFN-γ production capacity of NK cells; meanwhile, the ferroptosis-induced mtDNA-releasing can effectively recruit CD8+ T cells infiltration into tumors by activating the STING pathway (48). Additionally, the cGAS-STING axis activated by cytoplasmic mtDNA can promote the occurrence of pyroptosis, which in turn activates CD8+ T cells in a paracrine manner (49).

3.2
Maturation of DCs and enhancement of antigen presentation
DCs play a crucial role as a link between innate and adaptive immunity in the immune system and serve as important hubs for immune responses. As a bridge connecting innate and adaptive immunity, DCs have attracted significant attention due to their excellent antigen-presenting ability (50). In the process of antitumor immunity, DCs perform an indispensable and vital function; their roles are closely associated with the cancer-immune cycle, ultimately facilitating the elimination of tumor cells by effector T cells (51). Studies have revealed that DCs are key effector cells for the activation of the cGAS-STING pathway (52, 53). Acute STING activation promotes DCs function through two mechanisms: first, it directly induces the upregulated expression of DCs maturation markers (e.g., CD80, CD86); second, it enhances DCs’ antigen-presenting ability, which mainly depends on the autocrine loop of IFN-I triggered by mtDNA (47, 54, 55). Immunogenic cell death (ICD) induced by photothermal therapy can synergize with STING pathway activation to promote the uptake and processing of tumor antigens by DCs, forming a positive feedback loop (55).

3.3
Neutrophil extracellular traps and TME
NETs are mesh-like structures released by neutrophils, mainly composed of DNA, histones, and related proteases. The release of NETs is usually triggered by reactive ROS, which subsequently activates peptidylarginine deiminase 4 (PAD4), leading to the citrullination of histone 3 (H3) and further causing DNA unwinding and nuclear membrane rupture (56). Immediately after, gasdermin D (GSDMD)-mediated cell perforation further results in the release of DNA, histones, and related proteases, a process termed NETosis (57). During the release of NETs, mtDNA leakage often occurs, and the interaction between mtDNA and the cGAS-STING signaling pathway can sometimes promote the release of NETs (24, 58). In addition, various cell types can capture NETs through phagocytosis by recipient cells; intracellular cGAS recognizes NETs-derived DNA, thereby activating the cGAS-STING signaling pathway and increasing the expression levels of interleukin-6 (IL-6) and IFN-I (59). Most myeloid cells, such as macrophages and DCs, have been reported to recognize NETs-DNA intracellularly (60). Other epithelial-derived cells and tumor cells can also take up NETs and recognize them via intracellular cGAS (61). For example, hepatic tumor cells possess the ability to take up NETs, thereby activating the cGAS-STING pathway and inhibiting tumor migration (62). In summary, the release of NETs can amplify inflammation mediated by the mtDNA-cGAS-STING pathway, while activation of the mtDNA-cGAS-STING pathway can in turn promote the release of NETs—yet the specific mechanisms remain unclear.

3.4
Activation of macrophages
Tumor-associated macrophages (TAMs) are among the most abundant immune cells in the TME. Their plasticity allows them to differentiate into antitumor M1-like phenotypes or protumor M2-like phenotypes. Classically activated M1 macrophages possess potent tumor-killing, phagocytic, and antigen-presenting capacities; however, TAMs in the TME are typically educated into M2-like phenotypes, which assist tumor progression by supporting angiogenesis, promoting tumor proliferation, and establishing an immunosuppressive network (63–65). Therefore, targeting and reprogramming the activation status of TAMs has emerged as a crucial strategy to enhance antitumor efficacy. Recent studies have shown that cytoplasmic mtDNA induces inflammatory cascades by activating the cGAS-STING-IRF3 signaling axis, effectively reshaping macrophage function—this not only inhibits their protumor activities but also significantly enhances their antigen-presenting capacity (66, 67). For instance, when exogenous mtDNA is taken up by TAMs, it can directly drive their conversion from a protumor to an antitumor phenotype, providing a novel intervention target for cancer immunotherapy (68).

4
The mtDNA-cGAS-STING axis involved in tumor immune evasion (Figure 1)
4.1
Immunosuppressive polarization of TAMs
The mtDNA-cGAS-STING axis exerts dual roles in regulating the polarization of TAMs. In the hepatocellular carcinoma model, mtDNA induces the polarization of TAMs toward the M2 subtype through the TLR9- nuclear factor κB (NF-κB) signaling, forming an immunosuppressive microenvironment (69, 70); however, whether the cGAS-STING signaling is involved remains to be further studied. There is evidence that cGAS-STING activation induced by mtDNA can activate JAK–STAT3 signaling (71), and this signaling plays an important role in the differentiation of TAMs into the M2 subtype (72–74). Interestingly, oxidatively modified mtDNA can escape from tumor cells and act as an immunogenic damage-associated molecular pattern to induce the polarization of TAMs toward the M1 subtype, thereby reactivating the immune response of macrophages against cancer cells (75, 76). These findings reveal the complex role of mtDNA-cGAS-STING axis in regulating macrophage polarization in shaping the TME.

4.2
Coordinated regulation of the mtDNA-cGAS-STING axis and PD-1/PD-L1 pathway in tumor immune evasion
The mtDNA-cGAS-STING axis critically regulates tumor immune evasion through coordinated interaction with the PD-1/PD-L1 pathway. Activation of this axis promotes IFN-I secretion, which upregulates PD-L1/PD-1 expression and drives T cell exhaustion (77–79). Vesicle-mediated mechanisms further integrate these pathways. Necrotic tumor cells release extracellular vesicles (EVs) enriched in mtDNA and PD-L1 that induce macrophage IFN/IL-6 production to weaken T cell responses while directly triggering T cell apoptosis (80, 81). Similarly, IL-6-induced EVs promote mtDNA leakage in endometrial cancer, an effect reversible by anti-PD-L1 therapy (80). Although the precise role of cGAS-STING activation in these vesicle-mediated processes requires further investigation, these findings establish a strong rationale for combination therapies targeting both the mtDNA-cGAS-STING pathway and immune checkpoints.

4.3
mtDNA-cGAS-STING axis and MDSCs in tumor immune escape
The core function of MDSCs is to construct an immunosuppressive microenvironment in tumors, which inhibits anti-tumor immune responses through various mechanisms, while directly or indirectly promoting tumor growth, metastasis, and drug resistance (82). mtDNA released by senescent cells can be packaged in extracellular vesicles, which are selectively transferred to polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) in TEM. This process enhances the immunosuppressive activity of PMN-MDSCs via the cGAS-STING-NF-κB signaling (83). Additionally, exogenous mtDNA can also activate the STING pathway, which in turn creates an immunosuppressive microenvironment in MDSCs. This ultimately provides favorable conditions for the survival and proliferation of tumor cells (83, 84).

5
The potential mechanism of double sided immunoregulation of mtDNA-cGAS-STING axis
Acute and chronic activation of the mtDNA-cGAS-STING signaling pathway exhibit distinct differences in their underlying mechanisms and biological effects, with core disparities manifested in the duration of activation, intensity of signal transduction, and the ultimate outcomes of the mediated immune responses (77). Nevertheless, these two activation modes share a common feature: both can effectively activate IRF3 and NF-κB signaling pathways, thereby inducing dual effects of immune activation and immunosuppression in the organism (85–87).
Mitochondrial damage induced by short-term chemotherapy, or acute oxidative stress triggers massive acute release of mtDNA into the cytoplasm. The leaked mtDNA potently activates the cGAS-STING pathway, which in turn induces the phosphorylation of TBK1 (5). Phosphorylated TBK1 further activates IRF3 and NF-κB signaling, leading to the secretion of large amounts of type I interferons (IFN-α/β) and inflammatory factors (31). This immune activation effect can enhance antitumor immune responses, specifically characterized by strengthened T cell-mediated tumor cell killing and promoted maturation of DCs as well as their antigen-presenting function (77).
In the context of chronic oxidative stress, sustained elevation of reactive ROS results in persistent mtDNA leakage, which also activates the cGAS-STING pathway but with significantly lower signal intensity compared to acute activation. The low-intensity and sustained signal stimulation ultimately contributes to the formation of an immunosuppressive microenvironment and promotes tumor progression (77). Typical biological effects include sustained high expression of PD-L1, recruitment of MDSCs with enhanced immunosuppressive activity, accumulation of regulatory T cells (Tregs), and impairment of effector T cell infiltration (88, 89).
These observations suggest that the combination of radiotherapy/chemotherapy with STING agonists may exert a synergistic antitumor effect, but chronic activation of the pathway should be strictly avoided. In clinical practice, pulsatile stimulation is superior to sustained stimulation as an administration strategy, as it can enhance immune activation while minimizing the risk of immunosuppression.
Notably, the STING signaling pathway can activate NF-κB through a redundant mechanism involving TBK1 and IκB kinase ϵ (IKKϵ) (90). This mechanistic characteristic determines that a single TBK1 inhibitor cannot fully block NF-κB activation but can completely abrogate IRF3 signaling. This finding implies that specific signal blockers may serve as molecular “switches” for the precise regulation of IRF3 and NF-κB pathways, providing a novel direction for the optimization of immunotherapeutic strategies.

6
Clinical trials targeting the mtDNA-cGAS-STING axis
Clinical trials related to STING agonists and ENPP1 inhibitors in tumors are showing a pattern of active exploration and gradual advancement. According to data from ClinicalTrials.gov (as of December 1, 2025), 15 oncology-related studies have been registered for STING agonists, mostly in Phase I; research on ENPP1 inhibitors in tumors is still in the early exploratory stage, with a total of 4 studies all in Phase I, among which 3 are actively recruiting participants and 1 has not yet initiated recruitment (Table 1). Currently, no clinical data on safety and efficacy have been accumulated for ENPP1 inhibitors. Among STING agonist studies, most were terminated due to adjustments in corporate business policies, while TAK-500 and MIW815 were discontinued due to the lack of observed definite antitumor activity, reflecting the complexity of clinical development for this class of drugs.
Notably, the STING agonist MK-1454 has preliminarily validated its antitumor potential and safety through two clinical studies: Phase I (NCT03010176) and Phase II (NCT04220866) (91). In terms of antitumor activity, MK-1454 has demonstrated clear target-binding ability and synergistic therapeutic effects: in the Phase I study (N = 156), plasma drug concentrations increased in a dose-dependent manner, and key STING pathway-related cytokines in the circulation, such as CXCL10, IFNγ, and IL-6, began to increase 2–4 hours after administration, peaked at 6–8 hours, and partially declined by 24 hours, directly confirming that the drug can effectively activate the STING pathway; the exploratory combination therapy for head and neck squamous cell carcinoma (HNSCC) and triple-negative breast cancer in the expansion phase of this study provided an important direction for the expansion of its clinical application scenarios. In the Phase II randomized controlled study, for treatment-naïve patients with metastatic or unresectable recurrent HNSCC, the objective response rate (ORR) of MK-1454 combined with pembrolizumab reached 50% (4/8), which was significantly higher than the 10% (1/10) of pembrolizumab monotherapy, clearly demonstrating the synergistic antitumor advantages of the combination regimen.
In terms of safety, the toxicity profile of MK-1454 is controllable and manageable. The most common adverse event in the Phase I study was pyrexia (incidence of 70%), and only 10 patients experienced dose-limiting toxicities (DLTs), based on which 540 μg was identified as the recommended Phase II dose; in the Phase II study, pyrexia remained the main adverse event (n=5), with no serious adverse events endangering patients’ safety. Neither monotherapy nor combination therapy with pembrolizumab showed an intolerable toxicity profile. In summary, existing data preliminarily confirm that drugs targeting the mtDNA-cGAS-STING pathway (such as MK-1454) have considerable antitumor potential. However, current studies have limitations such as small sample sizes and short follow-up periods. Efficacy heterogeneity and long-term safety still need to be further clarified through larger-scale and longer-cycle clinical studies. Nevertheless, existing explorations have provided key references for the subsequent clinical translation and optimization of treatment regimens for this class of drugs.

7
Conclusion and outlook
The mtDNA–cGAS–STING axis plays a Janus-faced role in tumor immunity: acute engagement ignites type-I interferon signaling that empowers immune cells and antitumor responses, whereas chronic activation sculpts an immunosuppressive niche that enables immune evasion. Its functional output is shaped by tumor subtype, microenvironment milieu, and signal intensity.
While existing studies have achieved considerable progress, several key limitations remain unresolved. First, in terms of mtDNA quantification, existing methods have a technical constraint in differentiating DNA sources (nuclear vs. mitochondrial DNA) during leakage, which may compromise the accuracy of functional interpretations related to mtDNA. Super-resolution imaging, as a cutting-edge tool with high spatial resolution, can effectively visualize the subcellular localization of DNA and distinguish the distribution of nuclear DNA from mitochondrial DNA. The application of this technology in future studies will help address the current limitation and improve the reliability of mtDNA leakage detection. Second, there is the limitation of model systems. Immune-deficient models cannot fully recapitulate STING-dependent immune crosstalk in the physiological tumor microenvironment, as they lack a functional adaptive immune system. Therefore, the findings derived from these models should be interpreted with caution when extrapolated to clinical settings. To enhance the generalizability of existing research findings, we suggest further validating them in humanized mouse models.
To date, most insights derive from in-vitro or animal models, with limited clinical corroboration; precise tools to titrate mtDNA release and robust STING-targeted delivery platforms are still missing. Furthermore, cross-talk with other immune-regulatory circuits remains poorly charted, constraining the design of optimal combination regimens. Future efforts must map spatiotemporal control of this axis across human cancers, engineer accurate strategies to modulate mtDNA leakage and STING activity, integrate nanotechnology-based delivery systems with combinatorial immunotherapies, and rigorously validate safety and efficacy, thereby accelerating the clinical translation of personalized cancer immunotherapy.