Nanoengineered STING agonists for cancer therapy: toward immunodynamic control

Abstract

Activation of the cyclic GMP-AMP synthase–stimulator of interferon genes (cGAS–STING) pathway is a mechanistically compelling strategy for cancer immunotherapy, but first-generation STING agonists have shown limited clinical activity despite clear pharmacodynamic induction of interferon-related signals. This translational gap suggests that therapeutic benefit depends not only on pathway engagement but also on where, in which cells, at what amplitude, and for how long STING is activated. In this review, we summarize the cellular context of cGAS–STING signaling, the pharmacological evolution and clinical experience of STING agonists, and the major barriers imposed by route, species differences, biological heterogeneity, and systemic inflammatory toxicity. We then examine how nanoengineering may improve agonist protection, cytosolic access, tissue and cell-type selectivity, and temporal exposure. Importantly, we distinguish improved delivery or tumor accumulation from direct evidence of altered signaling kinetics or a wider therapeutic window. Representative lipid, polymeric, inorganic, biogenic, biomimetic, antibody-directed, and prodrug-based platforms are evaluated together with mechanism-based combination sequencing, hematologic malignancy-specific considerations, biomarker-guided patient selection, repeated-dose safety, and manufacturability. Overall, nano-STING therapy should be developed toward immunodynamic control rather than simple delivery enhancement, with benefit judged by the relationship among delivery, cell-resolved pharmacodynamics, antitumor efficacy, and toxicity.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13045-026-01860-0.

Introduction

The biological case for activating stimulator of interferon genes (STING) in cancer is compelling, yet the clinical record remains modest. STING agonists can induce interferon-related pharmacodynamic signals in patients, but objective responses have generally been infrequent, short-lived, or confined to selected lesions; several development programs have been discontinued, and no STING agonist has been approved for cancer therapy [1]. This gap between strong mechanistic promise and limited clinical efficacy is the central problem for the field [2, 3]. It indicates that simply producing more innate immune activation is not enough. Clinical benefit depends on where the signal is generated, which cells receive it, how long it persists, whether the tumor retains a functional pathway, and whether the resulting inflammation can be converted into durable adaptive immunity without unacceptable toxicity.

The problem is particularly relevant in immunologically “cold tumors”. These tumors commonly show weak antigen presentation, sparse cytotoxic T-cell infiltration, suppressive myeloid populations, hypoxia, abnormal vasculature, and metabolic checkpoints within the tumor microenvironment (TME). Because the cGAS–STING axis links cytosolic DNA sensing to type I interferon production, dendritic-cell maturation, antigen cross-presentation, and effector-cell recruitment, it offers a mechanistically attractive route for converting cold tumors into inflamed tumors. However, the same pathway can also recruit suppressive myeloid cells, induce counter-regulatory programs, or sustain tumor-promoting inflammation when activation is excessive or prolonged [4, 5]. The therapeutic objective is therefore not maximal STING activation, but a controlled immune pulse of the appropriate amplitude, duration, location, and cellular distribution.

Free cyclic dinucleotides (CDNs) illustrate this translational tension. They can be potent ligands but are highly polar, poorly membrane permeable, vulnerable to extracellular degradation, readily trapped in endolysosomal compartments, and difficult to distribute to deep or disseminated lesions. Systemic exposure may activate peripheral immune organs more readily than the tumor, whereas repeated intratumoral injection restricts treatment to accessible lesions and does not reliably solve tumor heterogeneity. Non-CDN small molecules and immune-stimulating conjugates broaden the available pharmacology, but they do not eliminate the need to control exposure or select patients whose tumors and host immune compartments can respond.

Nanotechnology provides a versatile platform to address these pharmaceutical limitations. A useful formulation should protect the agonist, direct it to the relevant tissue and cell population, enable cytosolic access, program a reversible exposure profile, and remain manufacturable at a clinically meaningful scale. Rather than simply cataloguing nanocarrier platforms, this review uses the limited clinical efficacy of first-generation STING agonists as a starting point to discuss how nanoengineering may reshape the therapeutic window of STING activation. We focus on how nanotechnology controls the amplitude, duration, location, and cellular compartment of STING signaling, and how these parameters should be integrated with route selection, combination sequencing, biomarker-guided patient selection, and clinical manufacturability.

For this narrative review, literature available up to 16 September 2026 was identified through PubMed, Web of Science, ClinicalTrials.gov, and the reference lists of selected articles. Search terms included “cGAS-STING”, “STING agonist”, “cyclic dinucleotide”, “nanoparticle”, “nanomedicine”, “cancer immunotherapy”, “clinical trial”, “immune checkpoint blockade”, “CAR-T”, “biomarker”, and “clinical translation”. Priority was given to original mechanistic studies, published clinical trial reports, translational nanomedicine studies, and recent authoritative reviews. Clinical trial stage, registration, and reported outcomes were cross-checked against published reports and ClinicalTrials.gov where applicable.

Several recent reviews have summarized STING agonists, nanocarrier classes, responsive delivery systems, and combination formulations [6–8]. Qu et al. provided a particularly comprehensive engineering-oriented analysis of STING-centered combination nanomedicine, including physicochemical optimization, co-encapsulation, surface targeting, programmable release, formulation development, and clinical translation [8].

Against this background, the present review focuses on a narrower question: when does nanoengineering merely improve delivery, and when does it demonstrably alter the immunodynamics of STING activation? We therefore examine representative platforms according to three linked levels of evidence: delivery to the intended tissue and cellular compartment; direct measurements of signaling amplitude or duration in responsive cells; and improvement in efficacy relative to toxicity under interpretable dosing conditions. Where direct kinetic measurements or appropriate comparators are unavailable, immunodynamic control is treated as a design objective rather than an established property. This framework is integrated with clinical trial outcomes, disease-specific biology, treatment sequencing, biomarker-guided patient selection, and translational constraints.

cGAS-STING pathway: cellular context and therapeutic window

Overview of cGAS-STING pathway

In 2008, STING (also known as MITA, MPYS, ERIS, or TMEM173) was identified as an adaptor protein linking cytosolic nucleic-acid sensing to type I interferon signaling [9, 10]. STING is primarily an endoplasmic reticulum (ER)-resident transmembrane adaptor that traffics through the ER-Golgi axis after activation and can also functionally interact with mitochondria-associated membrane compartments. In 2013, cyclic GMP-AMP synthase (cGAS) was identified as the major cytosolic DNA sensor that produces cyclic GMP-AMP (cGAMP), thereby completing the canonical cGAS-cGAMP-STING signaling axis [11] (Fig. 1).

Fig. 1.

Canonical cGAS–STING signaling and intercellular propagation. Schematic overview of canonical cGAS–STING signaling and major routes of intercellular cGAMP propagation. Cytosolic double-stranded DNA (dsDNA) derived from tumor DNA, damaged micronuclei, or pathogens is sensed by cGAS, which catalyzes the generation of 2′,3′-cGAMP from ATP and GTP. cGAMP binds and activates STING at the endoplasmic reticulum (ER), followed by STING trafficking through the ER–Golgi intermediate compartment (ERGIC) to the Golgi apparatus, where downstream signaling activates TBK1, IRF3, and NF-κB, ultimately inducing transcription of innate immune genes. Major outputs include type I interferons, chemokines such as CXCL9/CXCL10, and inflammatory cytokines. In parallel, cGAMP can propagate between cells through gap junctions, extracellular release and uptake, or extracellular vesicles (EVs)/exosomes, thereby extending STING signaling beyond the initially activated cell. This schematic was prepared with AI-assisted drawing and subsequently reviewed and edited by the authors

During pathogen infection, cellular stress, mitochondrial damage, micronuclear rupture, or therapy-induced DNA damage, double-stranded DNA can accumulate in the cytosol. cGAS binds cytosolic DNA through its nucleotidyltransferase domain and undergoes conformational rearrangement to catalyze the synthesis of 2′,3′-cGAMP from ATP and GTP [12–14]. cGAMP then binds the cytosolic ligand-binding domain of ER-localized STING, inducing a conformational transition from an inactive open state to an active closed dimeric state [15]. Activated STING exits the ER and traffics to the Golgi apparatus, where it recruits and activates TANK-binding kinase 1 (TBK1) [16]. TBK1 phosphorylates STING and interferon regulatory factor 3 (IRF3), enabling IRF3 recruitment, phosphorylation, and dimerization [17, 18]. Phosphorylated IRF3 then translocates into the nucleus and induces the transcription of type I interferons, particularly IFN-β. In parallel, STING signaling can activate the IKK-NF-κB axis, promoting the production of inflammatory cytokines and chemokines such as TNF-α, IL-6, IL-1β, CXCL9, and CXCL10 [19]. Therefore, the cGAS-STING pathway links innate DNA sensing to adaptive antitumor immunity, but its biological output depends strongly on cell type, activation intensity, duration, and tissue distribution.

cGAS–STING signaling can propagate between cells. Tumor-derived DNA, including DNA contained in dying-cell material or extracellular vesicles, can be internalized by dendritic cells and other myeloid cells and activate cGAS after gaining access to the cytosol [20]. cGAMP produced in one cell can reach neighboring cells through gap junctions or through extracellular export followed by uptake via cell-type-dependent channels and transporters [21, 22]. Extracellular vesicles can additionally transfer tumor-derived DNA and, in some settings, activated STING [23]. Thus, cGAMP production and STING activation need not occur in the same cell. In several tumor types, tumor-cell-intrinsic cGAS–STING signaling is reduced or epigenetically silenced, whereas host antigen-presenting cells may retain functional signaling and mediate antitumor priming [24]. Conversely, systemic activation of circulating myeloid cells and nonmalignant endothelial or stromal compartments can contribute to cytokine-mediated and organ-specific toxicity [25]. Therapeutic design should therefore define whether the intended target is tumor cells, host hematopoietic antigen-presenting cells, or nonhematopoietic stromal cells, rather than treating the tumor microenvironment as a single signaling compartment [26].

The physiological function of the cGAS-STING pathway

Pathogen-sensing function

The most primitive and fundamental function of the cGAS-STING pathway is anti-pathogen immunity. As a universal sensor for cytoplasmic dsDNA, cGAS can recognize almost all DNA viruses (herpesviruses, adenoviruses, hepatitis B virus), some RNA viruses (retroviruses) that are produced during infection, some bacterial and parasite [27–29]. Pathogens can evade this response by shielding nucleic acids or directly antagonizing cGAS–STING signaling [30, 31]. For cancer therapy, the relevant lesson is not the taxonomy of viruses, bacteria, or parasites, but the operating logic of the pathway: a rapid danger signal is beneficial when it is spatially restricted and self-limited, whereas persistent or systemic activation can damage host tissues [25, 32].

Antitumor immune surveillance and TME remodeling

Tumors generate several sources of cytosolic DNA. Chromosomal instability, replication stress, and segregation errors produce micronuclei whose fragile envelopes rupture and expose genomic DNA to cGAS [33–35]. Oxidative stress and mitochondrial dysfunction release mitochondrial DNA, while apoptosis, necrosis, pyroptosis, radiotherapy, chemotherapy, and DNA-repair inhibition increase the availability of nuclear and mitochondrial DNA in tumor cells and antigen-presenting cells [36–38]. These substrates can activate STING within tumor cells, but host dendritic cells and macrophages are often the more important therapeutic compartments.

In dendritic cells, STING activation promotes maturation, increases MHC and co-stimulatory molecules, supports cross-presentation, and induces chemokines that recruit CD8⁺ T cells and natural killer cells (Fig. 2) [1, 39]. CCR7 upregulation can facilitate migration to tumor-draining lymph nodes, where tumor-specific T-cell priming occurs. Macrophage responses are context dependent, but appropriately timed STING signaling can favor an inflammatory, antigen-presenting state over an immunosuppressive phenotype [40]. NK-cell cytotoxicity and IFN-γ production may be enhanced directly or through cytokines released by dendritic cells and macrophages [41–44]. The integrated outcome can convert a poorly infiltrated tumor into an inflamed lesion that is more responsive to immune checkpoint blockade.

Fig. 2.

STING-mediated orchestration of antitumor immunity. Schematic illustration of how STING activation in antigen-presenting cells coordinates antitumor immunity. Tumor-derived extracellular vesicles (EVs) and released antigens are taken up by dendritic cells, where STING activation promotes type I interferon production, chemokine secretion, and antigen cross-presentation. These processes enhance CD8 + T-cell priming and activation, support NK-cell activation, and contribute to tumor-cell killing. STING signaling also affects the myeloid compartment, including macrophages, and can promote myeloid reprogramming toward a more immunostimulatory state. Together, these interconnected effects link innate sensing to adaptive antitumor immune responses. This schematic was prepared with AI-assisted drawing and subsequently reviewed and edited by the authors

Under the influence of cGAS-STING, these immune cells work together to transform “cold tumor” into “hot tumor”, thereby achieving the antitumor effect. However, some articles also suggest that the pathway is related to the accelerated growth of tumors and the process of metastasis [45]. For example, previous research found that mice with the STING gene knocked out are almost completely resistant to the skin carcinogenicity caused by mutagens [46]. And this is the first report that controlling STING may have beneficial effect to cancer. Researchers suggest that this might be due to long-term chronic inflammation, which has been proven to be a cause of cancer. Furthermore, the activation of the cGAS-STING pathway can lead to the activation of the NF-κB signaling, which in turn drives the IL6/STAT3 signaling and the EMT process, respectively promoting tumor growth and metastasis (Fig. 3) [47]. In addition, there are also studies that show cGAS-STING pathway enhances suppressive inflammation in tumors by recruiting myeloid cells in part via the CCR2 pathway [48]. However, after the combined application of CCR2 inhibitors, this tumor-promoting effect was reversed. Therefore, some researchers believe that persistent or excessive engagement of STING results an unfavorable outcome of cancer patients.

Fig. 3.

Context-appropriate STING activation versus chronic/systemic activation. Conceptual comparison of beneficial versus deleterious modes of STING activation. On the left, acute, localized, and controlled STING activation in dendritic cells promotes type I interferon production, dendritic-cell maturation, antigen cross-presentation, chemokine secretion, recruitment and activation of CD8 + T cells and NK cells, and ultimately durable antitumor immunity. On the right, chronic, systemic, or dysregulated STING activation may drive persistent NF-κB signaling, IL-6/STAT3-associated immunosuppressive programs, suppressive myeloid-cell recruitment, T-cell dysfunction, chronic inflammation, tissue injury, and systemic toxicity. The central concept is that the therapeutic window depends on the amplitude, duration, location, and cellular context of STING activation rather than on a simple “short versus long” dichotomy. This schematic was prepared with AI-assisted drawing and subsequently reviewed and edited by the authors

Although some studies suggest that the activation of cGAS-STING pathway has a promoting effect on tumors, more extensive researches indicate that the activation of this pathway plays a crucial role in fighting against tumors and can become an important target for future anti-tumor treatments.

A double-edged pathway: duration, cellular context, and safety

Human genetics and inflammatory biology provide direct evidence that excessive STING signaling can be pathogenic (Fig. 3). Gain-of-function signaling or failure to clear self-DNA can sustain IFN-I and inflammatory cytokines, disrupt immune tolerance, and injure vascular, pulmonary, hepatic, renal, and other tissues [5, 49]. For oncology, these observations define a mechanism-based safety boundary. Systemic cytokine release, delayed inflammatory injury, and autoimmune-like toxicity are not incidental formulation problems; they are predictable consequences of broad or persistent pathway engagement.

The same temporal distinction applies within tumors. Acute, appropriately located STING activation can support immune surveillance, whereas chronic low-level signaling may activate NF-κB, IL-6/STAT3, epithelial–mesenchymal transition, and suppressive myeloid recruitment [45, 50]. STING signaling can also induce counter-regulatory molecules such as PD-L1 and IDO, creating adaptive resistance even after successful immune priming [4, 51]. Accordingly, efficacy and safety cannot be optimized independently: a platform that prolongs exposure may improve drug retention yet shift the response toward chronic inflammation or negative feedback [32].

A rational formulation should therefore be judged by the shape of the exposure–response curve. Relevant attributes include onset, peak amplitude, residence time, reversibility, tissue distribution, and recovery between doses. Safety testing should include repeated-dose pharmacodynamics and tissue histopathology, not only acute cytokines. This double-edged biology also supports triggerable or degradable systems that create a finite immune pulse and allows dose interruption when pharmacodynamic thresholds are exceeded.

STING agonists: pharmacological evolution and translational barriers

Therapeutic manipulation of the cGAS–STING axis in cancer can occur through three mechanistically distinct routes. First, direct STING agonists bind and activate STING independently of cGAS; these include cyclic dinucleotides, synthetic CDN analogues, and non-CDN small molecules. Second, cGAS-sensitizing strategies, such as Mn²⁺-based systems, enhance upstream DNA sensing and therefore require functional cGAS as well as STING. Third, chemotherapy, radiotherapy, oxidative stress, and selected nanomaterials can activate cGAS–STING indirectly by generating cytosolic nuclear or mitochondrial DNA. These routes should be distinguished because they differ in cGAS dependence, target-cell requirements, and relevant biomarkers. Table 1 summarizes representative direct STING agonists, whereas indirect and cGAS-sensitizing strategies are discussed separately in Sect.  3.4 and 4.

Table 1.

Representative STING agonists: pharmacological classes, development status, and translational features

Class	No.	Representative agent	Route	Development stage / trial	Main tumor type/model	Advantages	Disadvantages / limitations	Key ref.
CDNs	1	c-di-GMP	i.v.	Preclinical	Murine melanoma / vaccine models	Established bacterial CDN; immune-adjuvant activity in murine models.	Weak activity toward human STING; poor membrane permeability and metabolic stability.	[52]
CDNs	2	c-di-AMP	i.t.	Preclinical	Mechanistic and murine tumor studies	Activates STING and can engage additional innate pathways.	Affected by human STING polymorphism; also engages RECON–NF-κB and ERAdP–TAK1.	[53]
CDNs	3	3′,3′-cGAMP	i.p.	Preclinical	Malignant B-cell and murine tumor models	Direct CDN STING agonist with antitumor activity in preclinical models.	Weaker activation of human STING than endogenous 2′,3′-cGAMP; poor cellular uptake.	[54, 55]
CDNs	4	2′,3′-cGAMP	i.v./i.t. in preclinical studies	Preclinical	Melanoma, colon and breast tumor models	Endogenous STING ligand; activates most common human STING variants.	Rapid extracellular degradation and poor membrane permeability limit systemic use.	[56]
CDN analogues	5	ADU-S100 (MIW815]	i.t.	Phase I–II; NCT02675439; NCT03172936	Advanced/metastatic solid tumors and lymphoma	Pan-genotypic human STING agonism; clear local and systemic pharmacodynamic activity.	Repeated intratumoral dosing; monotherapy activity low; combination study was single-arm.	[57]
CDN analogues	6	MK-1454 (ulevostinag)	i.t.; with i.v. pembrolizumab	Phase I/II; NCT03010176; NCT04220866	Advanced solid tumors/lymphoma; recurrent/metastatic HNSCC	Published phase I pharmacodynamics; randomized phase II showed responses in 4/8 combination vs. 1/10 pembrolizumab-alone participants.	Very small randomized phase II cohort; does not establish definitive superiority or broad clinical benefit.	[58]
CDN analogues	7	TAK-676 (dazostinag)	i.v.	Phase I/II; NCT04420884; NCT04879849	Advanced solid tumors; radiotherapy/pembrolizumab combinations	Systemic administration; published pharmacodynamic activation and early combination data.	Modest clinical activity; peripheral cytokine exposure and combination attribution remain concerns.	[59]
CDN analogues	8	E7766	i.t./intravesical	Phase I; NCT04144140	Advanced solid tumors; bladder-cancer program	Macrocycle-bridged CDN with pan-genotypic activity and published early clinical pharmacodynamics.	Antitumor efficacy remains unproven; development was route- and program-limited.	[60]
CDN analogues	9	BMS-986,301	i.m./i.t./i.v.	Phase I; NCT03956680	Selected advanced solid tumors	Multiple administration routes evaluated clinically.	Public efficacy data remain limited.	-
CDN analogues	10	IACS-8779	i.t.	Preclinical	B16 melanoma	Potent preclinical STING agonism with systemic antitumor effects after local dosing.	Preclinical evidence only; intratumoral administration limits accessibility.	[61]
CDN analogues	11	BI-1,387,446	i.t.	Phase I; NCT04147234	Advanced/metastatic solid tumors	Clinical evaluation with checkpoint blockade.	Objective clinical efficacy not established.	-
CDN analogues	12	IMSA101	i.t.	Phase I/II; NCT04020185	Treatment-refractory solid tumors	Published pharmacodynamic immune activation with generally manageable safety.	Very limited objective antitumor activity in published phase I data.	[62]
CDN analogues	13	SB-11,285	i.v.	Phase I; NCT04096638	Lymphoma and advanced solid tumors	Synthetic CDN evaluated by systemic administration.	Published clinical efficacy and dose–response evidence remain limited.	-
Non-CDN agonists	14	MK-2118	i.t./s.c.	Phase I; NCT03249792; published	Refractory advanced solid tumors and lymphoma	Intratumoral dosing produced dose-dependent systemic STING pharmacodynamics.	Limited antitumor activity; ORR 0%, 6%, and 4% across reported treatment arms; s.c. dosing lacked dose-related immune responses.	[63]
Non-CDN agonists	15	DMXAA (ASA404)	i.v.	Phase III; NCT00662597	Advanced NSCLC; multiple murine models	Potent murine STING activation and vascular-disrupting activity.	Activates mouse but not human STING; clinical development failed.	[64]
Non-CDN agonists	16	diABZIs	i.v.	Preclinical	CT26 and other syngeneic tumor models	Potent systemic activation of human and mouse STING in preclinical models.	Clinical therapeutic window and inflammatory toxicity remain unestablished.	[65]
Non-CDN agonists	17	MSA-2	p.o./s.c./i.t. in preclinical studies	Preclinical	MC38 and CT26 colorectal models	Orally bioavailable; pH-dependent tumor-associated pharmacology; synergizes with PD-1 blockade preclinically.	Clinical safety and efficacy remain unestablished; systemic exposure and inflammatory effects require regimen-specific characterization.	[66]
Non-CDN agonists	18	SR-717	i.p./i.v. in preclinical studies	Preclinical	Melanoma and colorectal models	Activates human and mouse STING; systemic preclinical activity.	Preclinical evidence only; human safety and therapeutic window are unknown.	[67]
Non-CDN agonists	19	CRD3874-SI	i.v.	Phase I; NCT06021626; NCT06626633	Advanced solid tumors; relapsed/refractory AML	Systemic clinical development includes a hematologic-malignancy program.	Mature safety and efficacy data are not yet available.	-
Non-CDN agonists	20	GSK3745417	i.v.	Phase I; NCT03843359	Advanced solid tumors	Systemic non-CDN STING agonist evaluated clinically.	Definitive antitumor efficacy has not been established in published reports.	-
Non-CDN agonists	21	SNX281	i.v.	Phase I; NCT04609579	Advanced solid tumors/lymphoma	Acid-activated design intended to favor tumor STING activation.	Program terminated; public clinical efficacy data are limited.	-
Non-CDN agonists	22	KL340399	i.t.	Phase I; NCT05387928; NCT05549804	Advanced solid tumors	Direct intratumoral STING activation.	Intratumoral route restricts eligible lesions; mature results unavailable.	-
ISAC / antibody-directed	23	TAK-500	i.v.	Phase I/II; NCT05070247; terminated	Selected locally advanced/metastatic solid tumors	CCR2-directed delivery to intratumoral myeloid cells.	Trial terminated for clinical futility; target-cell delivery did not establish clinical benefit.	[68]
ISAC / antibody-directed	24	XMT-2056	i.v.	Phase I; NCT05514717; terminated	HER2-expressing advanced/recurrent solid tumors	HER2-directed STING-agonist ADC designed for tumor and tumor-resident immune-cell activation.	Clinical hold after a grade 5 SAE was lifted on 31 Oct 2023; the trial later terminated. Available cited sources do not establish that the later termination was caused by the earlier SAE.	[69]

Cyclic dinucleotides (CDNs) and analogues

CDNs

CDN, as a natural STING stimulant, mainly originates from bacteria. The naturally occurring CDNs found in bacteria mainly consist of three types: c-di-GMP, c-di-AMP and 3’,3’-cGMP-AMP (cGAMP) [70].

c-di-GMP, as the earliest discovered CDN, did not function as a second messenger when it was first discovered in Gram-negative bacteria [71]. Instead, it was proven to be the true activator of cellulose synthase. It was not until 2004 that its role as a second messenger was discovered. It is connected by two symmetrical 3’-5’ phosphodiester bonds between two GMP molecules [72]. c-di-GMP can activate mouse STING, but has a lower affinity for human STING [52].

c-di-AMP, connected by two 3’-5’ phosphodiester bonds to two AMP molecules, and is mainly found in Gram-positive bacteria [73]. It, like c-di-GMP, is also significantly affected by the STING polymorphism. In addition to activating STING, c-di-AMP can also activate other immune pathways such as RECON-NF-κB and ERAdP-TAK1 [53, 74]. The immune activation effect is more complex.

3’,3’-cGAMP, which presents in some bacteria, is composed of two nucleotides that are linked by two 3’-5’ phosphodiester bonds [75]. 3’,3’-cGAMP is the earliest discovered cGAMP isomer that can activate STING. It participates in the host’s immune recognition of intracellular bacterial infections, but its activation activity on human STING is significantly weaker than that of the endogenous 2’,3’-cGAMP [54].

In addition to bacterial-derived CDNs, the mammalian endogenous CDN, 2’,3’-cGAMP, which is connected to GMP and AMP through unique 2’-5’ and 3’-5’ mixed phosphodiester bonds, can effectively activate the majority of human STING genotypes. Some studies have showed that the conventional (3′-5′) cyclic-di-nucleotides produced by bacteria, have a weaker ability to activate the human STING variants in comparison with the non-canonical cyclic-di-nucleotide produced by mammalian cGAS [76]. This might be the reason why endogenous 2’,3’-cGAMP in mammals has superior activity compared to c-di-GMP and c-di-AMP.

Natural CDNs have several drawbacks, such as poor membrane permeability, susceptibility to degradation by serum phosphodiesterase, and significant differences in the activity of different human STING alleles [77, 78]. These limitations have hindered their clinical application. Nevertheless, natural CDNs still have an indelible effect and research value.

CDN analogues

In recent years, in order to better achieve the activation effect of the pathway, researchers have developed a large number of synthetic CDN derivatives through chemical modification strategies, significantly enhancing their stability, cellular uptake efficiency, and STING agonistic activity.

ADU-S100 is the world’s first STING agonist to enter clinical trials. Its chemical nature is a R, R dithio-modified cyclic adenosine dimer analogue, which can strongly activate all the known five human STING genotypes [79]. Its potent tumor elimination effect has been widely verified in various mouse models [80]. In the phase I clinical trial published in 2022, when 50 to 6,400 micrograms of the ADU-S100 was injected weekly and administered according to the plan of 3 weeks of use followed by 1 week of discontinuation, most of the injected tumors showed shrinkage, and 2.1% of the patients achieved PR in patients with advanced or metastatic solid tumors or lymphomas [81]. Although the single drug ADU-S100 showed limited clinical activity in this first-in-human study, the research did uncover evidence of systemic immune activation. Plasma cytokines, including IFNβ, MIP-1β, IL6, MCP-1 and CXCL10, significantly increased after the injection. This may support the theory of combining STING activation with PD-L1 inhibition. Immediately following that, in 2023, a multi-center study released its research findings. The combination of ADU-S100 with the PD-1 inhibitor Spartalizumab in advanced or metastatic solid tumors or lymphomas resulted in an overall objective response rate (ORR) of 10.4%. Because this phase Ib study was single-arm, the contribution of MIW815 beyond spartalizumab cannot be isolated from this dataset. The reported ORR of 10.4% therefore indicates limited activity of the combination in an unselected population rather than a direct comparison with PD-1 monotherapy [57].

Ulevostinag (MK-1454) is a synthetic CDN STING agonist evaluated by intratumoral administration, alone or with pembrolizumab. The 2025 report included a phase I study of 156 participants and a small randomized phase II study in untreated recurrent or metastatic HNSCC [58]. In phase I, circulating CXCL10, IFN-γ, and IL-6 increased after dosing, generally peaking at approximately 6–8 h and showing plateau or partial resolution by 24 h; above 540 µg, a clear additional cytokine dose–effect relationship was not observed. In the randomized phase II study, objective responses occurred in 4 of 8 participants receiving ulevostinag plus pembrolizumab and in 1 of 10 receiving pembrolizumab alone. These findings represent a preliminary activity signal, but the very small sample does not establish definitive superiority or broadly generalizable clinical benefit.

TAK-676, also called Dazostinag, a CDN derivative developed by Takeda Oncology, employs a novel chemical modification strategy [82]. Intravenous administration of TAK-676 can lead to a significant STING-dependent anti-tumor activity in mouse models, without tumor regeneration, indicating a persistent inhibitory effect on tumors. TAK-676 is currently undergoing some clinical studies for patients with locally advanced or metastatic solid tumors with the PD-1 antibody, Pembrolizumab, and with several other anti-cancer drugs. In 2025, the results of a phase Ib study on some patients with advanced solid tumors showed that after receiving hypofractionated radiotherapy, TAK-676 combined with pembrolizumab could bring clinical benefits to some patients, with an ORR of 7.1% [59]. For patients who received TAK-676 via intravenous injection throughout the body, the levels of IFN-γ and pro-inflammatory cytokines such as IL-6 and IP-10 in their blood plasma increased. These studies indicate that TAK-676 may benefit patients with solid tumors who have already undergone extensive pre-treatment and have experienced disease progression.

E-7766 is the third-generation CDN, a type of large-ring bridge connection or multi-modification optimized CDN [83, 84]. It retains the core framework of the CDN while innovatively introducing a cross-ring macrocycle bridged by a trans-alkene structure. By directly binding through covalent bridges to the biologically active U-shaped conformation of the CDN when combined with STING, the affinity for human STING is significantly enhanced. Currently, this drug has entered the Phase I clinical trial. The available results indicate that E-7766 can induce immune activation and has acceptable safety, but its anti-tumor efficacy still requires further study.

SB-11,285 is a synthetic cyclic dinucleotide STING agonist evaluated by intravenous administration in an early-phase clinical program (NCT04096638). It should therefore be classified with CDN-type STING agonists rather than as a separate “small molecule–nucleic acid hybrid” pharmacological class. Published clinical efficacy and dose–response data remain limited.

Apart from ADU-S100, MK-1454, TAK-676, E7766 and SB-11,285, other CDN analogues are also under further investigation, such as BMS-986,301, IACS-8779 and BI-1,387,446 [61, 85].

Small molecule agonists

Dimethyloxoxanthenyl Acetic Acid (DMXAA) is the first small molecule discovered to have STING agonistic activity. It was once used as a vascular disruptor in Phase III clinical trials [86]. Subsequent studies confirmed that it only specifically activates STING in mice and has almost no agonistic activity on human STING [64]. The clinical Phase III trials failed, and currently, it is only used as a research tool molecule [87]. Despite the clinical failure, these valuable efforts have driven the development of new non-nucleotide analogues with higher human STING affinity.

Small molecule drugs with a non-cyclic dinucleotide backbone have diverse chemical structures and can activate the pathway by binding to the allosteric site of STING or mimicking the binding mode of CDN. Compared with CDN-like drugs, they have the potential for simple synthesis, better pharmacokinetics, and the ability to be administered orally or systemically, making them a research hotspot in recent years. Apart from DMXAA, an increasing number of small molecule drugs have emerged, such as dimeric amidobenzimidazole (diABZIs), MSA-2 and SR-717.

diABZIs are the first small molecule agonists that have been proven to efficiently activate human STING [65]. They were obtained through high-throughput screening and structural optimization. The core structure of these compounds is a dimerized 2-aminobenzimidazole, which can mimic the U-shaped biological activity conformation of the natural CDN ligand and bind to STING, thereby addressing the species-specific defect of DMXAA. Among this kind of drugs, the most representative one is diABZI-1. Intravenous injection of diABZI-1 can induce significant tumor regression and systemic anti-tumor immune responses in the mouse colon cancer CT26 model. Based on diABZI-1, the research team further enhanced the STING affinity, metabolic stability and in vivo exposure of the molecule by adjusting the structure, and obtained optimized derivatives such as diABZI-2 and diABZI-3 [88, 89]. In addition, GSK3745417 is a clinical development candidate molecule derived from the diABZIs after a systematic screening and has now entered phase I clinical trial [78].

MSA-2 is the first orally bioavailable non-nucleotide STING agonist discovered by Merck & Co. through high-throughput screening [66]. MSA-2 exists in solution in the form of a balance between monomers and non-covalent dimers [90]. Only the dimer form can bind to STING. Furthermore, this affinity also exhibits remarkable specificity, highlighting its potential as a targeted therapeutic agent. More importantly, MSA-2 exhibits pH-dependent cell-penetrating activity: in the circulatory system with a physiological pH of 7.4, the molecule carries a negative charge and has a relatively weak ability to penetrate the cell membrane; while in the acidic tumor microenvironment (with a pH of around 6.5), the molecule is in a neutral state and is more likely to penetrate the cell membrane and enter the cytoplasm to activate the STING pathway. This characteristic enables MSA-2 to preferentially activate the STING pathway at the tumor site, thereby exerting anti-tumor activity throughout the body while reducing systemic side effects. Preclinical studies have confirmed that both oral administration and subcutaneous injection of MSA-2 have good bioavailability [91]. MSA-2 alone can effectively activate the STING pathway in mice and significantly inhibit tumor growth. When used in combination with anti-PD-1 antibodies, the anti-tumor effect is further enhanced. Although MSA-2 shows tumor-acidity-dependent pharmacology and favorable preclinical antitumor activity, its clinical safety and efficacy remain unestablished. Systemic exposure and inflammatory effects therefore require regimen-specific characterization. Therefore, an increasing number of researchers have designed MSA-2-related nanodelivery systems to enhance its pharmacokinetics and anti-tumor efficacy [92].

SR-717 is a non-nucleotide agonist obtained by Scripps Research through structure-based drug design [67]. SR-717 can mimic the binding pattern of the natural ligand 2’,3’-cGAMP and bind to the cGAMP binding pocket of STING. SR-717 is a neutral small molecule with better cell membrane permeability, and it can effectively activate human and mouse STING. Systemic administration of SR-717 demonstrated anti-tumor activity in various tumor models, and no significant systemic toxicity was observed. What is more significant is that SR-717 not only activates the STING pathway, but also can induce the expression of PD-L1 in a STING-dependent manner.

In conclusion, an increasing number of small molecule STING activators have been discovered through high-throughput screening or drug design. These drugs have distinct characteristics in terms of their chemical structure, mechanism of action, and administration route. However, further research is still needed to reduce their systemic side effects, increase targeting and anti-tumor efficacy, and achieve broader clinical translation.

Immune-stimulating antibody conjugate (ISAC)

The traditional Antibody-Drug Conjugate (ADC) is usually composed of a monoclonal antibody (mAb) covalently linked to a cytotoxic drug through a chemical linker [93]. Immune-Stimulating Antibody Conjugate (ISAC) is an extension of ADC technology in the field of immunotherapy, but they replace the cytotoxic payload with immune stimulants [94]. It can target and deliver STING agonists to the tumor site through antibodies, achieving local immune activation and reducing systemic toxicity.

XMT-2056 is a HER2-directed STING-agonist antibody–drug conjugate designed to activate STING in HER2-expressing tumor cells and tumor-resident immune cells [69]. Its phase I trial (NCT05514717) was placed on clinical hold in March 2023 after a treatment-related grade 5 serious adverse event. The FDA lifted the hold on 31 October 2023, and enrollment subsequently resumed with a lower starting dose. The trial is now listed as terminated, and XMT-2056 was removed from GSK’s phase I pipeline in the first quarter of 2026. The available cited sources do not establish that the later termination was caused by the earlier fatal event.

Another well-known ISAC is called TAK-500, a CCR2-directed immune-stimulating antibody conjugate designed to deliver a STING agonist to CCR2-positive intratumoral myeloid cells. Preclinical studies showed enhanced antitumor activity in combination with PD-1 blockade or radiotherapy [68]. The phase I/II clinical study (NCT05070247) was subsequently terminated for clinical futility, indicating that myeloid-cell targeting alone did not establish sufficient clinical benefit.

In addition, Jacobio Pharma has developed the world’s first CD73-targeted STING agonist immune-stimulating ADC, JAB-X1800. This drug has also entered the preclinical research stage. Some of the results presented at the 2023 American Association for Cancer Research (AACR) Annual Meeting showed that the drug demonstrated strong anti-tumor activity and good tolerance in mouse models, but complete preclinical research results and further clinical studies still need to be made public.

Traditional antitumor therapeutic drugs

In addition to direct agonists, many traditional anti-tumor therapeutic drugs can indirectly activate the cGAS-STING pathway by inducing DNA damage and releasing substances.

Chemotherapeutic drugs such as anthracyclines (doxorubicin, daunorubicin), docetaxel, and 5-FU can induce immunogenic cell death (ICD) in tumor cells, releasing large amounts of nuclear DNA and mitochondrial DNA into the cytoplasm and the tumor microenvironment, thereby activating the cGAS-STING pathway [95–97].

Radiotherapy induces double-strand DNA breaks, leading to DNA release and micronucleus formation, which is a potent inducer of the cGAS-STING pathway [98]. Studies have shown that the anti-tumor effect of radiotherapy is weakened in mice with STING knockout [99]. Moreover, the combination of radiation and the STING agonist cGAMP can reduce radiation resistance and synergistically enhance the anti-tumor host response.

In addition, some targeted drugs can also indirectly activate the cGAS-STING pathway. poly-(ADP-ribose) polymerase (PARP) inhibitors, by inhibiting DNA damage repair, cause cytoplasmic DNA accumulation and activate the cGAS-STING pathway [100]. After Olaparib treatment, the cytoplasmic DNA levels of BRCA-deficient tumor cells and IFN-β expression increased. In recent years, anti-HER2 ADC drugs used for treating breast cancer and gastric cancer, T-DXd, have also been proven to activate the cGAS-STING pathway, thereby activating DCs and CD8⁺ T cells and exerting a strong anti-tumor effect [101]. CDK4/6 inhibitors can also induce DNA damage or cell death and indirectly activate the STING pathway [102]. These findings provide a theoretical basis for the combination of targeted therapy and immunotherapy.

Clinical trials of STING agonists: lessons from limited efficacy

By 2026, multiple STING agonists had entered clinical development as monotherapies or in combination with immune checkpoint blockade, radiotherapy, chemotherapy, or other immunomodulatory strategies (Table 2). Published trials provide a consistent lesson: pharmacodynamic activation of STING does not necessarily translate into durable tumor regression. Intratumoral CDN agonists can generate local and systemic cytokine signals but remain restricted by lesion accessibility and repeated injection. Systemic agonists broaden anatomical reach but increase the importance of peripheral cytokine exposure, human STING pharmacology, target-cell selection, and safety.

Table 2.

Clinical lessons from STING agonist trials and implications for nanoengineering

Agent	Route / setting	Published clinical evidence	Main clinical lesson	Implication for nano-STING
ADU-S100 (MIW815)	i.t.; monotherapy or + spartalizumab	First-in-human monotherapy: PR 2.1%. Phase Ib MIW815 + spartalizumab: ORR 10.4%; single-arm study.	Clear cytokine/pharmacodynamic activation did not translate into high response rates; the combination study cannot be interpreted as a direct comparison with PD-1 monotherapy.	Improve cell-selective delivery and adaptive immune conversion rather than simply increasing local exposure.
MK-1454 (ulevostinag)	i.t.; ± pembrolizumab	Phase I: N = 156 with time-resolved cytokine pharmacodynamics. Randomized phase II HNSCC: responses in 4/8 combination vs. 1/10 pembrolizumab-alone participants.	Small randomized cohort shows an activity signal but is insufficient for a definitive efficacy conclusion.	Use pharmacodynamic kinetics and randomized benefit–risk comparisons rather than cytokine induction alone.
TAK-676 (dazostinag)	i.v.; after hypofractionated RT + pembrolizumab	Published phase Ib combination study reported ORR 7.1% with increases in IFN-γ, IL-6, and IP-10.	Systemic pathway activation is feasible, but clinical activity remains modest and attribution among RT, pembrolizumab, and STING agonism is difficult.	Restrict systemic exposure and align STING delivery with treatment-induced antigen/DNA release.
E7766	i.t. phase I	Published early clinical study demonstrated pharmacodynamic STING activation with manageable safety; objective antitumor activity remained limited.	Target engagement alone is insufficient to establish clinical benefit.	Select disease setting and route first; require tumor immune conversion and efficacy endpoints.
IMSA101	i.t. phase I	Published phase I study showed pharmacodynamic immune activation with generally manageable safety but very limited objective antitumor activity.	Local delivery and immune activation did not overcome low efficacy in an unselected population.	Add biomarker enrichment and predefined pharmacodynamic/efficacy go–no-go criteria.
MK-2118	i.t. or s.c.; ± pembrolizumab	Published phase I study, N = 140. Objective responses: 0% with i.t. monotherapy, 6% with i.t. + pembrolizumab, 4% with s.c. + pembrolizumab. Dose-related systemic immune effects were seen after i.t., not s.c., dosing.	Route changed pharmacodynamics; manageable toxicity and pathway activation did not translate into strong antitumor activity.	Route and tissue compartment are part of the drug mechanism and should be prospectively optimized.
TAK-500	i.v.	Phase I/II trial NCT05070247 terminated for clinical futility.	Myeloid-cell targeting alone did not establish sufficient clinical efficacy.	Cell targeting must be linked to target-cell state, payload release, and adaptive immune conversion.
XMT-2056	i.v.	Phase I trial placed on hold after a treatment-related grade 5 SAE in 2023; hold lifted 31 Oct 2023 and enrollment resumed; trial later terminated. XMT-2056 removed from GSK phase I pipeline in Q1 2026.	Precision targeting does not eliminate inflammatory safety risk. Later termination should not be attributed to the 2023 SAE without direct evidence.	Pair tumor targeting with controllable release, dose fractionation, and organ/cytokine stopping rules.
DMXAA (ASA404)	i.v.; phase III oncology development	Clinical development failed; later mechanistic work showed strong species selectivity for mouse rather than human STING.	Mouse efficacy can be misleading when human STING pharmacology differs.	Human STING variant activity and human-relevant models must precede formulation optimization.
exoSTING (CDK-002)	i.t.; phase I/II	First-in-human study completed/terminated without publicly posted efficacy results.	Manufacturing, program continuity, and lack of public clinical outcome data limit translational interpretation.	Bio-derived platforms require scalable CMC, potency assays, and clinically interpretable pharmacodynamic endpoints.

The available clinical evidence should therefore be interpreted at three separate levels: target engagement, objective antitumor activity, and benefit–risk. ADU-S100, ulevostinag, E7766, IMSA101, MK-2118, and dazostinag have all demonstrated some degree of pharmacodynamic pathway activation, yet objective responses have generally been limited, regimen-dependent, or observed in small cohorts. These findings support development strategies that measure tumor and immune-cell pharmacodynamics, select biologically responsive patients, and compare efficacy with toxicity rather than using cytokine induction alone as evidence of a widened therapeutic window.

Why STING agonists underperform clinically

Despite strong mechanistic rationale, the clinical translation of STING agonists remains constrained by three interconnected barriers: inadequate delivery to the appropriate cellular compartment, biological heterogeneity in STING responsiveness, and a narrow safety window. For CDN-like agonists, large molecular weight, strong polarity, negative charge, extracellular enzymatic degradation, and endosomal trapping all reduce cytosolic availability [103, 104]. Intratumoral injection can partially overcome local exposure limitations, but it is unsuitable as a universal solution for deep, multifocal, or disseminated disease. Conversely, systemic administration can activate peripheral immune organs and circulating myeloid cells before sufficient tumor or APC exposure is achieved [26].

The second barrier is biological resistance. Human STING is encoded by TMEM173, and common functional variants differ in ligand responsiveness [77]. Tumor cells may also downregulate cGAS or STING through epigenetic silencing or other immune-evasion mechanisms [105–107]. In addition, the ENPP1-CD73-adenosine axis can simultaneously degrade cGAMP/CDN-like molecules and suppress antitumor immunity through adenosine signaling [108–110]. These mechanisms explain why tumor accumulation or cytokine induction alone cannot be used as reliable surrogates for clinical efficacy.

The third barrier is safety. Excessive or persistent STING activation can induce systemic inflammation, cytokine release, autoimmune-like injury, and organ-specific toxicity [103]. Chronic low-level activation may also promote NF-κB/IL-6/STAT3 signaling, suppressive myeloid recruitment, angiogenesis, and adaptive resistance through PD-L1 or IDO upregulation [111–113]. Therefore, the central translational challenge is not simply to strengthen STING activation, but to generate a finite, reversible, cell-selective immune pulse that is strong enough to prime adaptive immunity yet sufficiently controlled to avoid systemic inflammatory toxicity.

Apart from the aforementioned issues, there are still many unresolved problems from the experimental stage to clinical translation. The ORR of single STING agonist treatment for advanced solid tumors is generally low. Although combination with PD-1/PD-L1 inhibitors is currently the mainstream research strategy, the synergistic mechanism, optimal dosing sequence, and dose ratio of the combined regimen still lack clinical validation. Many early combined studies have not achieved the expected improvement in efficacy. Moreover, the combination strategies with chemotherapy, radiotherapy, PARP inhibitors, etc. are still in the exploration stage, and there is no mature clinical protocol. Additionally, there is currently a lack of clinically validated predictive biomarkers, making it impossible to precisely screen the beneficiary population for STING agonists. This leads to a large heterogeneity in the clinical trial enrollment population, making it difficult to observe significant group benefits, which is a key issue restricting precise treatment.

Nanoengineering strategies to widen the STING therapeutic window

Nanoengineering can modify drug exposure, intracellular access, or the signaling machinery itself (Fig. 4), but these effects should not be treated as interchangeable evidence of immunodynamic control. In this review, representative platforms are evaluated at three levels: (i) delivery evidence, including stability, biodistribution, cellular uptake, and cytosolic access; (ii) signaling-dynamics evidence, requiring time-resolved measurements of pathway activity in defined cells or tissues; and (iii) therapeutic-window evidence, requiring interpretable comparison of efficacy and toxicity across dose, route, or schedule.

Fig. 4.

Design principles by which nanoengineering may improve the STING therapeutic window. Overview of major nanoengineering strategies that may improve the STING therapeutic window. These include protecting agonists from degradation and premature clearance, enhancing tumor or antigen-presenting-cell targeting, promoting endosomal escape and cytosolic access, enabling tumor- or stimulus-responsive release, tuning signaling duration in a context-appropriate manner, optimizing combination-treatment sequencing, and incorporating biomarker-guided patient or route selection. Together, these strategies aim to improve exposure in the intended compartment while balancing efficacy and toxicity, thereby broadening the therapeutic window of nano-STING approaches. This schematic was prepared with AI-assisted drawing and subsequently reviewed and edited by the authors

A prolonged plasma half-life does not establish prolonged intracellular STING signaling, and carrier retention does not establish continued release of bioactive agonist. Likewise, stronger tumor inhibition at one dose demonstrates efficacy under that regimen but does not by itself prove a wider therapeutic window. Table 3 therefore summarizes representative studies according to release/exposure pattern, measured signaling dynamics, efficacy, and toxicity. The broader formulation-level comparison is moved to Supplementary Table S1.

Table 3.

Representative evidence linking nano-STING platform design to signaling dynamics and therapeutic-window outcomes

Representative platform	Release / exposure pattern	Measured STING dynamics	Efficacy comparison	Toxicity / therapeutic-window evidence	Conclusion supported	Ref.
STING-NP polymersome	i.v. nanoencapsulation extended cGAMP half-life ~ 40-fold and increased tumor, liver, and spleen exposure.	Tumor and organ STING activation measured by gene expression; organ responses differed in magnitude and timing.	Systemic STING-NP reduced tumor burden and improved survival in melanoma and breast-cancer models.	10 µg cGAMP regimen tolerated; 20 µg caused > 15% weight loss, deaths, hepatic necrosis, and splenic apoptosis; liver and spleen were dose-limiting.	Direct PK–PD and dose-toxicity evidence; improved delivery opened a systemic dosing window but did not remove organ toxicity.	[26]
PLGA pulsatile cGAMP microparticles	Predetermined discrete cGAMP pulses from microfabricated PLGA microparticles, designed to mimic repeated injections.	No direct cell-resolved time course of downstream STING signaling sufficient to equate release pulses with signaling pulses.	Single administration reproduced antitumor effects of repeated soluble dosing in tested models.	Mainly local-delivery evidence; full systemic benefit–risk relationship was not established.	Evidence for programmed drug-release timing, not direct proof of altered intracellular STING kinetics; system is microparticulate, not a representative nanoparticle.	[114]
PC7A polymer	PC7A directly binds a noncanonical STING site and delays trafficking/degradation rather than simply releasing a payload slowly.	Slower and more durable STING-associated signaling than cGAMP, including persistent p-IRF3/IRF7/CXCL10-related responses in APC-centered models.	Enhanced antigen cross-presentation, CTL priming, and antitumor immunity in preclinical models.	Limited systemic cytokine exposure and no overt major-organ histopathology under the tested conditions; no clinical therapeutic-window data.	Direct evidence that a carrier/agonist can alter STING signaling duration, but benefit is context- and dose-dependent rather than evidence that longer is always better.	[115, 116]
D-SAM dual-STING micelle	Rapid cGAMP component combined with slower PC7A-mediated STING activation.	Designed and measured as sequential STING activation rather than simple sustained drug release.	Improved dendritic-cell activation, CTL priming, memory, and tumor control compared with component regimens in tested models.	Preclinical only; local dosing and formulation complexity limit generalization.	Direct preclinical example linking programmed temporal signaling to immune outcomes.	[117]
exoSTING	Engineered EVs increased intratumoral retention and APC-focused delivery of CDN agonist.	At equivalent low doses, higher intratumoral IFN-β/CXCL9/CXCL10 at 4 h than free CDN; minimal IFN-β remained at 24 h, showing that retention is not equivalent to persistent signaling.	Therapeutically active exoSTING achieved strong antitumor activity at much lower agonist dose than free CDN in B16F10 models.	Efficacious exoSTING dose produced little systemic cytokine induction and less local tissue/immune-cell ablation than high-dose free CDN.	Evidence for improved local benefit–risk and cell-focused pharmacodynamics, while also showing that prolonged carrier retention need not mean sustained IFN output.	[118]

Liposome nanocarriers

Liposomes are vesicular structures formed by a phospholipid bilayer membrane. They are one of the most mature and widely used nanocarriers in clinical applications, and are one of the preferred systems for delivering STING agonists.

From a mechanistic perspective, lipid carriers combine extracellular protection with pH-dependent intracellular delivery. Encapsulation shields CDNs from phosphodiesterases, while particle size and surface ligands support tumor accumulation and uptake by dendritic cells or macrophages. After endocytosis, ionizable lipids become protonated and interact with anionic endosomal phospholipids, promoting membrane fusion or phase transition and releasing the agonist into the cytosol. Acid-labile bonds and charge-switching layers can further separate stable circulation from drug release in the weakly acidic tumor microenvironment or lysosom. Thus, stability, targeting, pH responsiveness, and endosomal escape are not independent advantages of liposomes but sequential steps in the same delivery process.

Traditional cationic liposomes can form complexes with negatively charged CDNs through electrostatic interactions, thereby increasing the encapsulation efficiency and cellular uptake. For example, research has proved that liposomal nanoparticle-delivered cGAMP, which combined by a mixture of hydrogenated (soy)L-α-phosphatidylcholine (Soy-PC) and 1,2-dioleoyl-3-trimethyl-ammonium-propane (DOTAP), can more effectively activate cGAS-STING pathway, compared to cGAMP, in mouse models [119].

The traditional single-layer liposomes have the problem that it is difficult to balance the circulation stability and the efficiency of cytoplasmic delivery: high membrane fluidity can lead to premature drug leakage, while overly rigid membranes are not conducive to escape from the endosome [120]. By constructing multi-level membrane structures for liposomes, they can maintain high stability in the blood circulation and avoid premature leakage of cGAMP; after entering the APC, they disintegrate layer by layer in the acidic endosome, simultaneously achieving drug release and efficient endosome escape. Compared with traditional single-layer liposomes, the activation level of the STING pathway can be increased by several times. And lipid nanoparticles (LNPs) apply this method, consisting of ionizable cationic lipids, auxiliary phospholipids, cholesterol and PEGylated lipids [121]. In acidic buffer solution, the ionizable lipids become protonated and carry a positive charge, which combines with the negatively charged CDNs through electrostatic interaction, and simultaneously self-assemble into dense nanoparticles. The STING agonist is completely encapsulated inside the particles, protected from enzymatic degradation. After entering the bloodstream, under physiological pH conditions, the ionizable lipids become electrically neutral, significantly prolonging the circulation half-life. At the same time, the outer PEG layer gradually peels off, exposing the lipid structure on the particle surface. Through passive or active targeted delivery, it is endocytosed and enters the cell. Under the acidic endosomal pH, the tertiary amine groups of the ionizable lipids are protonated in large quantities, and then the endosomal membrane is disrupted through the membrane phase transition mechanism and the “proton sponge” effect, allowing the endosome to escape into the cytoplasm to exert its function. Researchers have designed a novel pH-sensitive polycationic polymer-modified lipid nanoparticle (LNP-B) system based on the above mechanism to deliver CDNs [122]. This new nano-platform significantly prolonged the circulation of CDNs, enhanced their accumulation in tumors, spleens, and tumor-draining lymph nodes, and produced a potent anti-tumor immune response and tumor suppression effect in a colon cancer mouse model.

However, lipid nanomaterials also have certain drawbacks. The phospholipid component is prone to oxidation, which will affect the stability of the lipid particles during long-term storage and the efficiency of drug encapsulation [6]. Moreover, the encapsulation rate of lipid particles is relatively low, resulting in insufficient drug loading capacity.

Therefore, lipid-based systems are most valuable when the dominant barrier is cytosolic access of polar CDNs. Their translational potential should be evaluated not only by tumor accumulation, but also by cell-resolved STING activation in dendritic cells and macrophages.

Polymer nanocarriers

Polymer nanocarriers possess characteristics such as highly adjustable physicochemical properties, controllable degradation rate, and ease of functionalization modification. They mainly include polymer nanoparticles, polymer vesicles, polymeric micelles and antibody-targeted polymer–drug conjugates, among others. They are one of the core research directions for the delivery of STING agonists.

The chemical diversity of polymers enables several activation mechanisms to be embedded directly in the carrier. Weakly basic amines support proton-sponge or membrane-disruptive endosomal escape, whereas disulfide and thioketal linkers respond to intracellular glutathione or reactive oxygen species [123, 124]. Peptide or polysaccharide linkers can be cleaved by matrix metalloproteinases, cathepsins, hyaluronidase, or other tumor-associated enzymes [125, 126]. Nitroimidazole and azo groups provide hypoxia-dependent charge conversion, carrier disassembly, or drug release [127]. These functions permit a polymer nanoparticle to remain comparatively stable in blood and then change charge, size, solubility, or integrity after reaching the tumor or an intracellular compartment.

Polymer nanoparticles

PLGA-based systems illustrate why drug-release kinetics should not be equated with downstream STING-signaling duration. Kodamasimham et al. formulated CDA and Mn²⁺ in injectable PLGA milli-cylinder implants, which limited the initial burst and continuously co-released both agents over approximately 3–4 weeks after local administration [128]. These milli-cylinder implants are local depots rather than nanoparticles. Lu et al. used microfabricated PLGA microparticles to generate discrete cGAMP pulses at predetermined times, thereby mimicking repeated injections after a single administration [114]. These microparticles provide evidence for programmed drug-release timing, but they should not be presented as proof that intracellular STING signaling itself follows the same pulse pattern. In contrast, PLGA-PEG/DMXAA nanoparticles represent a nanoscale formulation, although the translational relevance of DMXAA is limited by its selectivity for murine rather than human STING [129].

Polymer vesicles

Polymer vesicles are hollow spherical carriers formed by amphiphilic block copolymers through hydrophobic interactions in a selective solvent, with a closed bilayer membrane structure [130]. Their topological structure is highly similar to that of liposomes, both enclosing a water-filled cavity within a hydrophobic bilayer. However, synthetic polymers replace natural lipid molecules, thus possessing stronger structural tunability, mechanical stability, and functional design space. They are the core and cutting-edge carrier platform in the fields of nanomedicine and synthetic biology. Some researchers successfully enhanced the biological efficacy of cGAMP by encapsulating it in polymer vesicles [26]. This design enables the STING agonist to be more concentrated in the tumor area, prolongs the elimination half-life of the STING agonist, and enhances the anti-tumor effect in the mouse model.

Polymeric micelles

Polymeric micelles can either deliver a STING agonist or participate directly in pathway activation. PC7A is an intrinsically active polymeric micelle that binds a noncanonical STING site through multivalent interactions and can delay trafficking and degradation of activated STING [115, 131]. This mechanism differs from conventional sustained drug release: PC7A modifies the lifetime of the activated signaling complex itself. In preclinical models, PC7A produced slower and more durable STING-associated signaling than cGAMP, with persistent p-IRF3/IRF7 and CXCL9/CXCL10-related responses, enhanced antigen cross-presentation, and stronger antigen-specific T-cell priming [115, 116]. The dual-STING system D-SAM further combined rapidly acting cGAMP with slower PC7A-mediated signaling to generate sequential activation rather than simple prolonged exposure [117].

These findings do not imply that longer STING activation is universally beneficial. Prolonged signaling is most plausibly advantageous when it remains spatially restricted to tumors, draining lymph nodes, or antigen-presenting cells and overlaps with antigen processing and T-cell priming. Excessive, systemic, or T-cell-intrinsic STING activation can instead promote cytokine toxicity, negative feedback, impaired T-cell proliferation, or chronic inflammatory programs. PC7A should therefore be presented as a preclinical example in which altered signaling duration was linked to productive immune priming under defined conditions, not as evidence that sustained STING activation is intrinsically superior.

In addition to the inherent active polymer micelles, encapsulated polymer micelles can also be applied in the research and development of STING agonist nanomedicines. For hydrophobic small molecule STING agonists, encapsulated polymer micelles can significantly enhance their water solubility and tumor targeting properties. A study applied polymer micelle-encapsulated small molecule STING agonist SR-717, combined with a classic indoleamine 2,3-dioxygenase (IDO) inhibitor 1-MT, and a photosensitizer 5,10,15,20-tetraphenylporphyrin (TPP), to strongly activate anti-tumor immunity in vivo [132]. In the 4T1 breast cancer model, an 88% suppression rate of solid tumors could be achieved.

In addition to the mentioned polymer nanocarriers above, antibody-targeted polymer–drug conjugates are increasingly being applied in the development of STING agonists. We will discuss this part within the context of nanotechnology-mediated ADC drugs. Due to the ease of functionalization and high drug loading capacity of polymers, polymer-based nanotechnologies for STING agonists have become a research hotspot in recent years. More and more relevant nanoplatforms have been developed. However, during the production process of these polymers, some impurities or residues may be introduced, which may affect biological safety and also restrict the development of polymer-based nanomaterials for STING agonists [6].

Inorganic nanocarriers

Inorganic nanocarriers possess unique advantages such as high chemical stability, precise and controllable structure, and the ability to achieve multimodal diagnosis and treatment integration. Some inorganic materials themselves also have STING pathway sensitization activity, enabling functional synergy between the carriers and the agonists. Based on their material properties, they can be classified into two major categories: metallic nanomaterials and non-metallic nanomaterials.

Inorganic materials can also activate STING by generating endogenous DNA substrates. Fenton and Fenton-like reactions, photodynamic processes, and metal-mediated oxidative stress can damage nuclear DNA and inhibit its repair [133, 134]. Mitochondrial membranes and respiratory-chain complexes are particularly sensitive to these reactions, and their disruption releases mitochondrial DNA into the cytosol, where it serves as a potent cGAS substrate [135]. Crystalline silica or titanium-containing particles may additionally damage lysosomal membranes and facilitate cytosolic leakage of nucleic acids [136, 137]. These carrier-intrinsic mechanisms amplify immunity without requiring a larger dose of exogenous agonist, but they also make control of off-target genotoxicity essential.

The optical, magnetic, and radiophysical properties of inorganic carriers further enable external spatiotemporal regulation. Gold, copper sulfide, and related photothermal materials convert near-infrared light into local heat, while photosensitizing systems generate reactive oxygen species and DNA damage [138, 139]. Magnetic particles can be guided to tumors or heated by an alternating magnetic field, and high-atomic-number materials can enhance local radiation deposition and radiotherapy-induced STING activation [140, 141]. Ultrasound-responsive inorganic or hybrid systems can provide deeper tissue penetration than light-based approaches [142]. Accordingly, externally triggered activation is most naturally integrated into inorganic nanocarriers rather than treated as a separate delivery category.

Many metal ions can serve as carriers for targeted delivery of STING agonists. Among them, Mn²⁺ can directly bind to cGAS, thereby enhancing the sensitivity of the STING pathway to dsDNA. As a result, it has become a research hotspot in the field of metal-anchored nanoparticles. Some researchers have assembled Mn²⁺ with CDN to form CDN-Mn²⁺ particles (CMP), which significantly enhanced the activation of cGAS-STING pathway [133]. And this drug has demonstrated significant therapeutic effects in a variety of difficult-to-treat tumor mouse models. Some researchers have also developed a structurally ordered metal-based nanomedicine [143]. It is a nanomaterial composed of CDN, Mn²⁺, and an 11-mers of histidine (His11, used as a metal coordinating ligand), encapsulated in 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) liposomes. This nanotechnology-mediated STING activator, compared to single CDN, significantly increased the drug distribution half-life by more than five times in the B16F10 tumor mouse model. Moreover, a low dose of the nanodrug can achieve good anti-tumor effects in melanoma models and head and neck cancer models. Apart from manganese metal, iron nanoparticles and gold nanoparticles are also widely used in drug development mediated by nanotechnology. Gold nanoparticles exhibit excellent photothermal conversion effects and facile surface modifiability. They can be loaded with STING activators through surface chemical bonding and combined with photothermal therapy and immunotherapy [144, 145]. While, iron-based nanomaterials can generate ROS through the Fenton reaction, indirectly activating the STING pathway [146, 147]. And the Fe₃O₄ magnetic nanoparticles can also be combined with an external magnetic field to achieve magnetic targeted delivery, further enhancing the tumor accumulation efficiency of the STING agonist [148].

Among non-metallic nanomaterials, mesoporous silica nanoparticles (MSNs) are regarded as a classic inorganic drug carrier due to their advantages such as high specific surface area, tunable pore size, and good biocompatibility. The surface of MSNs has shown that they can be modified. Combined with pH-responsive or enzyme-responsive nanotechnology, they can help drugs escape from endosomes and achieve controlled release of the drugs. The STING agonists mediated by MSNs technology has been proven to have excellent immune activation effects and anti-tumor effects in animal models [149]. Apart from MSNs, other non-metallic inorganic materials, such as calcium carbonate with pH-responsive degradation properties and black phosphorus nanomaterials with photothermal effects, have also been applied in the delivery of STING agonists [150, 151].

In addition, nanocoordination polymers (NCPs) are self-assembled through the coordination interaction between metal ions and organic ligands [152]. They can efficiently encapsulate drugs and protect them from enzymatic degradation. Metal-organic frameworks (MOFs), which also belong to the coordination assembly system, specifically refer to nano-complex polymers with long-range ordered crystalline structure and regular inherent pores. And it is a crystalline subclass of NCP. Due to the drug molecules can directly act as ligands or metal nodes to participate in the framework construction, there is no additional carrier redundancy. Moreover, the regular pores of MOFs can adsorb and encapsulate small molecule drugs, and the surface can adsorb molecules through electrostatic and hydrophobic interactions, while the inner cavity can encapsulate drugs. These characteristics enable this kind of nanoplatform to have a much higher drug loading capacity than traditional nanocarriers. In recent years, NCPs have been widely used in the research and development of sting agonists, and has achieved great anti-tumor effects in animal models [153, 154].

The main advantage of inorganic systems is that the carrier itself can participate in pathway activation or external-energy conversion. However, this also makes off-target genotoxicity, long-term biodistribution, and tissue persistence critical translational concerns.

Biogenic nanocarriers

Biogenic nanomaterials are directly produced, isolated or assembled from natural cells or biological systems, and possess extremely low immunogenicity, excellent biocompatibility, and natural cell-to-cell delivery capabilities. They are an indispensable part of the nanotechnology field. They mainly include exosomes and various biological vesicles.

Biogenic vesicles implement cell-selective delivery through membrane biology rather than synthetic targeting alone. Their adhesion proteins, lipids, and transport machinery can promote uptake by antigen-presenting cells or homologous tumor cells, while the vesicular membrane protects the agonist during circulation. This feature addresses the key distinction between tumor accumulation and delivery to a functionally responsive cellular compartment. At the same time, bacterial or tumor-derived vesicles may carry endogenous danger signals and antigens, so delivery, pathway activation, and in situ vaccination can occur within the same platform.

Exosomes are membrane-bound vesicles with a size ranging from 30 to 150 nanometers that are secreted by cells [155]. They naturally perform the functions of transporting substances and transmitting signals between cells, making them an ideal natural delivery vehicle for STING agonists. Previous studies have confirmed that exosomes can effectively deliver cGAMP to the tumor site [156]. Compared with free drugs, the cGAMP encapsulated in exosomes is more efficiently taken up by tumor cells, which can significantly inhibit tumor growth and enhance anti-tumor immunity, and has lower systemic toxicity. Through engineering modification, the delivery efficiency and function of exosomes can be further enhanced. For instance, exoSTING consists of engineered extracellular vesicles carrying STING agonists [118]. This platform prolongs intratumoral retention and preferentially delivers the agonist to antigen-presenting cells within the tumor microenvironment. A phase I/II study (NCT04592484) was conducted in patients with injectable advanced solid tumors; however, peer-reviewed clinical efficacy results remain unavailable, and the clinical development program was subsequently discontinued. In addition, by chemically coupling the CDN with the TLR1/2 agonist Pam₃CSK₄, they can be efficiently loaded onto tumor-derived exosomes through hydrophobic interactions [157]. The constructed EXO^{TLR1/2−STING} can simultaneously activate the two innate immune pathways of STING and TLR, cooperatively inducing the activation of APC and the death of tumor cells. The anti-tumor effect is significantly superior to that of single-agonist exosomes.

Apart from exosomes, Microvesicles (MVs) and bacterial outer membrane vesicles (OMVs) are also used for the delivery of STING agonists. For instance, the MVs (with a diameter of 100–1000 nm) extruded from the tumor cell membrane can deliver STING agonists through homotypic targeting properties, while carrying complete tumor antigens and exerting the role of in situ tumor vaccine [158]. In addition, the bacterial OMV itself has natural immunostimulatory activity [159]. After being engineered, it can activate the cGAS-STING pathway in cancer cells and TAMs, enhancing tumor selectivity and efficacy while reducing adverse reactions [160].

Biomimetic nanoparticles

Biomimetic nanoparticles is a promising STING agonist delivery platform, which imitate the structural and functional characteristics of biological systems (such as cell membrane encapsulation, pathogen simulation, etc.) to endow the nanocarriers with functions such as immune evasion, homologous targeting, and natural biological recognition [161]. Biomimetic nanoparticles and biogenic nanocarriers are partially overlap: natural extracellular vesicles are primarily biologically derived, fully synthetic lipoprotein-mimetic nanoparticles are biomimetic, and cell-membrane-coated nanoparticles represent biohybrid systems that combine biologically derived components with synthetic nanocores.

Cell membrane encapsulation technology is the most mature biomimetic strategy. It involves coating the surface of nanoparticles with natural cell membranes, enabling the carrier to inherit both the physicochemical properties of the nanoparticle core and the biological functions of the cell membrane. Previous studies have utilized this nanotechnology for individualized cancer immunotherapy [162]. First, IFN-γ is used to stimulate antigen presentation, and the resulting tumor cell membranes provide sufficient tumor response antigens for the PC7A nanodrug, forming Antigen-Enriched tumor Cell Membranes (AECM). Then, AECM is combined with the PC7A nanodrug to form AECM@PC7A. AECM@PC7A not only enhances the anti-tumor effect of the stimulator agonist, but also provides a promising and universal approach for the rapid development of personalized cancer vaccines.

In addition to mimicking the cell membrane, virus-biomimetic nanoparticles can also be applied to nanotechnology-mediated STING agonists. Nanoparticles designed by imitating the structure and infection characteristics of viruses can enhance the uptake and immune activation efficiency of antigen-presenting cells. A study has utilized the preparation of a herpesvirus-mimicking nanoparticle (named Vir-ZM@TD) for tumor treatment [163]. Eventually, this drug effectively activated the cGAS-STING pathway in the 4T1 mouse model, causing the primary tumors to completely disappear by 68%, and significantly prolonging the survival time of the model mice.

Antibody-directed and macromolecular delivery systems

Antibody-directed and macromolecular delivery systems provide a route to cell-selective STING activation, but they should not be described simply as conventional ADCs or as nanomedicines solely because of molecular size. In the context of STING agonism, the relevant platforms include immune-stimulating antibody conjugates (ISACs), antibody-conjugated nanoparticles, and polymeric or dendritic macromolecular conjugates that increase payload capacity or alter biodistribution. Their shared goal is to deliver a potent innate immune stimulus to tumor cells or intratumoral immune cells while limiting systemic exposure.

ISACs such as XMT-2056 and TAK-500 illustrate both the promise and the risk of this approach [69, 164]. Antibody targeting can enrich STING agonists in HER2-expressing tumor compartments or CCR2-positive myeloid cells, but available clinical experience indicates that targeting alone is insufficient [69]. Efficacy may depend on target density, internalization, linker stability, payload release kinetics, the activation state of the target cell population, and whether local innate activation is converted into adaptive antitumor immunity. Severe toxicity or futility in early clinical programs also shows that precision targeting does not eliminate the need for careful cytokine monitoring, dose escalation rules, and tissue-specific safety assessment.

Antibody-conjugated nanoparticles and macromolecular conjugates may increase payload loading beyond the drug-to-antibody ratios typical of classical ADCs and can incorporate responsive release, endosomal escape, or combination payloads. However, greater structural complexity may also increase CMC burden, batch variability, immunogenicity, and off-target reticuloendothelial uptake. Therefore, future antibody-directed nano-STING systems should be developed only when the platform provides a measurable advantage over simpler formulations. Key requirements include target-density biomarkers, cell-state biomarkers, quantitative payload-release assays, cell-resolved pharmacodynamics, and repeated-dose safety evaluation.

Nano-prodrugs

Nano-prodrugs represent a formulation strategy in which therapeutic activity and tumor-microenvironment (TME)-responsive activation are encoded into the drug molecule or its covalent linker. Chemical modification temporarily masks drug activity and may also enable the prodrug to self-assemble into nanoparticles or be incorporated into a nanocarrier. The resulting formulation protects the payload from premature degradation and systemic exposure, while promoting sustained and localized drug activation within tumors, thereby reducing off-target toxicity and widening the therapeutic window [165]. This combination of molecular-level activation and nanoscale delivery is particularly attractive for cancer immunotherapy, in which the magnitude, location, and duration of immune stimulation must be carefully controlled.

Nano-prodrugs can be activated by either endogenous TME characteristics or externally applied physical energy. Acid-labile groups release or unmask therapeutic payloads under acidic conditions; disulfide and thioketal linkers respond to intracellular glutathione and reactive oxygen species (ROS), respectively; enzyme-cleavable linkers exploit tumor-associated proteases or glycosidases; and nitroimidazole or azo structures undergo bioreduction under hypoxia [124, 127]. These mechanisms can directly liberate a STING agonist or release a cytotoxic agent that induces DNA damage and generates cytosolic double-stranded DNA, thereby activating the cGAS–STING pathway indirectly. External stimuli—including light, ultrasound, magnetic heating, and ionizing radiation—provide an additional layer of temporal and spatial control by cleaving sensitive bonds, destabilizing nanostructures, generating ROS, or inducing localized DNA damage within a defined treatment window [166].

A representative nano-prodrug example is MSA-NLG, an esterase-responsive bifunctional prodrug formed by covalently coupling the non-nucleotide STING agonist MSA-2 with the IDO inhibitor NLG919; after formulation with DSPE-mPEG₂K, esterase-mediated cleavage releases both active agents and couples STING activation with relief of IDO-mediated immunosuppression [167]. By contrast, caged-diBSP01 is a photoactivatable molecular prodrug rather than a nanoparticle by itself [168]. Coumarin-based photoremovable groups mask critical carboxylate pharmacophores, and irradiation at approximately 400 nm restores active diBSP01 and STING signaling. It is therefore included only as a molecular example of externally triggered temporal control; current evidence is limited to in vitro studies and a zebrafish xenograft model.

Nevertheless, responsive activation should not be presumed solely from the nominal characteristics of the TME. Acidity, redox status, enzyme abundance, and hypoxia vary across tumor types, patients, and spatial regions within the same lesion. Therefore, activation thresholds, prodrug-conversion kinetics, and the duration of downstream STING signaling should be quantitatively evaluated. For physically activated systems, clinical feasibility additionally depends on tissue penetration, lesion accessibility, and whether sufficient activating energy can be delivered to all relevant primary and metastatic sites. These factors will ultimately determine whether nano-prodrugs can translate their improved selectivity into reproducible therapeutic efficacy and reduced systemic cytokine toxicity.

Taken together, carrier selection should follow the activation mechanism required by the intended indication. Lipid and polymer systems are strongest when protection and cytosolic delivery are the principal barriers; inorganic and selected micellar systems are valuable when the material itself contributes pathway activation or external energy conversion; biogenic and biomimetic carriers emphasize cellular tropism and endogenous danger signaling; and nano-prodrugs or antibody-directed systems provide conditional or cell-specific activation. No platform is universally superior, and the most translatable design is the least complex system that achieves the required biodistribution, intracellular access, activation kinetics, and safety margin.

Mechanism-based combination and sequencing strategies

Although a single STING agonist treatment can activate the innate immune system, it often fails to completely eliminate established solid tumors and is prone to causing immune suppression and drug resistance. Combined therapy, through multi-mechanism synergy, can significantly enhance anti-tumor efficacy and is currently the mainstream research direction.

Combination efficacy depends not only on mechanistic complementarity but also on whether each intervention is delivered at the appropriate stage of the cancer–immunity cycle. For combinations involving an antigen-generating treatment, an idealized sequence may comprise: tumor-cell injury and antigen release; antigen uptake by dendritic cells; STING-mediated APC maturation and cross-presentation; recruitment and expansion of effector cells; and blockade of adaptive or metabolic feedback. These events do not necessarily occur in the same anatomical or cellular compartment. Therefore, co-encapsulation, simultaneous tumor accumulation, or improved overall tumor control should not by itself be interpreted as evidence of optimal temporal coordination. In the following sections, we distinguish between schedule-validated strategies, in which simultaneous or sequential regimens were directly compared; regimen-validated strategies, in which only one combination schedule was experimentally tested; and mechanism-based hypotheses, for which the proposed sequence has not yet been directly examined.

Surgery

Surgery is the main treatment method for solid tumors, but postoperative recurrence and micrometastasis are the main reasons for treatment failure. Surgical removal of the primary tumor can reduce the tumor burden. However, during the wound healing process after the surgery, factors such as PGE₂ are released, leading to a local and systemic immunosuppressive state [169, 170]. This, instead, may accelerate the growth and metastasis of the residual lesions. Nano STING agonists have unique advantages in combined surgical treatment. Under the background of surgical trauma, applying STING agonists at the tumor site can transform the immunosuppressive wound microenvironment into an immunostimulatory one, promoting tumor antigen presentation and inducing systemic anti-tumor immune memory. This has an “in situ vaccine” effect in terms of local residual micro-diseases and distant metastases of the tumor. Research has confirmed that by placing hydrogels containing sting agonists at the site of tumor removal, the slow release of the drugs can prevent local tumor recurrence, activate anti-tumor immunity, and eliminate existing spontaneous metastasis [171]. By combining surgical treatment with the therapeutic strategy of nano-STING agonists, not only local tumors can be eliminated, but also a persistent systemic immune protection can be activated, endowing radical surgery with the characteristics of preventive treatment. Another study combined cGAMP nanoparticles, doxorubicin, anti-PD-1 antibodies or OX40 ligand to prepare injectable hydrogel in situ vaccines [172]. Whether used alone or in combination with surgery as an adjuvant or neoadjuvant immunotherapy, it can significantly activate the immune system, inhibit tumors and recurrence models.

In addition to using the nano-STING agonist locally at the tumor site, intravenous injection of the nano-STING agonist can also be combined with surgery. It can also exert the effect of an “in situ vaccine”. In the context of surgical trauma, the use of nanocarriers to target and deliver STING agonists can transform the immunosuppressive wound microenvironment into an immunostimulatory one, promoting tumor antigen presentation and inducing systemic anti-tumor immune memory. This has an “in situ vaccine” effect on residual microlesions and distant metastases. A study has developed a hierarchical-responsive liposome which contains STING agonists and doxorubicin [173]. Intravenous administration of this drug can be used as a neoadjuvant therapy to reduce the size of tumors before surgery, and by simultaneously activating both innate and adaptive anti-tumor immune responses, it can prevent tumor recurrence and metastasis after surgery.

Perioperative nano-STING strategies should be viewed as locoregional immune-priming approaches rather than stand-alone curative therapy. Their clinical value will depend on controlled release at the surgical margin, systemic immune memory against micrometastatic disease, and safe integration with chemotherapy, radiotherapy, or immune checkpoint blockade.

Chemotherapy

Chemotherapy is the fundamental treatment method for solid tumors. Various chemotherapy drugs can induce DNA damage and ICD, release dsDNA, and are natural activators of the endogenous cGAS-STING pathway. However, the immune activation effect of simple chemotherapy is weak, often accompanied by immunosuppressive side effects. While nano-co-delivery of chemotherapy drugs and STING agonists can achieve multiple synergistic effects. It can not only maintain DNA damage, but also reshape the tumor microenvironment, reduce drug toxicity, and simultaneously improve anti-tumor efficacy. This combined use method has been explored and verified in various solid tumors. For example, Li’s team developed a nano-system with a specific content of cationic lipids, which efficiently co-delivered cGAMP and doxorubicin [174]. In the mouse melanoma model, the tumor growth inhibition rate of this system was significantly better than that of the free drug combination group, and the systemic immune-related adverse reactions were milder. In addition, manganese-based nanomaterials do not require the addition of exogenous STING agonists. They merely release Mn²⁺ to amplify the activation of the endogenous cGAS-STING pathway induced by chemotherapy, possessing the advantages of simple composition and high safety. For instance, Wang et al. developed a manganese-based nano-system, PEG-MnMOF@PTX [175]. In the TME, the nanoparticles rapidly degrade and release paclitaxel and Mn²⁺, significantly enhancing the activation efficiency of the cGAS-STING pathway, ameliorating the immunosuppressive microenvironment, and improving the efficacy of immunotherapy.

For chemotherapy combinations, the key question is not whether DNA damage can activate STING, but whether the timing and dose of cytotoxic therapy create an antigen-release window that can be amplified by a finite STING pulse. Nano-co-delivery should therefore be justified by improved sequencing, reduced systemic toxicity, or synchronized exposure, rather than by simple payload accumulation.

Direct experimental evidence supporting this sequence has recently emerged. Yang et al. first administered Doxil to induce immunogenic tumor-cell death and then delivered the lymph node-targeted STING nanoadjuvant Mn/MSA-2@Lipo after an optimized interval of 48 h [176]. This delay allowed dendritic cells to acquire chemotherapy-released antigens before STING-driven maturation, thereby enhancing cross-presentation and CD8⁺ T-cell priming. Importantly, this study compared different administration intervals and therefore provides schedule-level evidence that sequential delivery can outperform poorly timed administration. By contrast, most co-delivery studies demonstrate the efficacy of a predefined simultaneous regimen but do not establish that co-release is superior to sequential dosing.

Radiotherapy

Radiotherapy can cause DNA damage to tumor cells through ionizing radiation and is the core treatment method for local solid tumors. The “distant effect” induced by local radiotherapy highly depends on the systemic immune activation mediated by the cGAS-STING pathway. However, the immune activation effect of simple radiotherapy is weak and difficult to induce an effective distant effect. Therefore, the combination of STING agonists mediated by nanotechnology and radiotherapy can produce a synergistic mechanism, significantly strengthen the “distant effect”. In addition, nanotechnology may enable STING activation to overlap with, or follow, radiation-induced DNA damage and antigen release. However, temporal overlap should not automatically be regarded as optimal, because premature APC maturation may limit subsequent antigen uptake, whereas excessive delay may miss the period of maximal antigen availability. Most current studies have tested only one combined schedule and have not directly compared simultaneous with delayed STING activation. For example, some researchers have developed a HfO2@MnO2-PAH/PAA/HA nanoplatform to enhance the therapeutic effect of radiotherapy and improve the anti-tumor effect both locally and systemically [177]. Similarly, some researchers have developed a nano-platform that simultaneously carries the TLR9 agonist CpG and ADU-S100, exerting an “in situ tumor vaccine” effect, effectively reversing the immunosuppressive response induced by radiotherapy and achieving a better anti-tumor effect compared to radiotherapy alone [178].

Radiotherapy combinations are most rational when STING activation is synchronized with radiation-induced DNA damage, antigen release, and dendritic-cell priming. Future studies should define optimal fractionation, agonist timing, and whether nano-STING delivery can convert local irradiation into reproducible systemic antitumor immunity.

Phototherapy

Phototherapy includes photothermal therapy (PTT) and photodynamic therapy (PDT). PDT generates ROS through photodynamic agents under the excitation of specific wavelength lasers to kill tumor cells, while PTT converts light energy into heat energy through photothermal converters to induce thermal ablation of tumor cells. Both can induce ICD, causing tumor cells to release cytoplasmic dsDNA, and endogenously activating the cGAS-STING innate immune pathway. However, the activation intensity of the pathway induced solely by phototherapy is limited, and the local treatment attribute is difficult to cope with distant metastasis and postoperative recurrence. The combination of exogenous STING agonists can further amplify the immune activation effect and initiate systemic anti-tumor immunity. Nanotechnology provides an ideal integrated platform for the synergy of phototherapy and STING pathway activation. A single nanocarrier can simultaneously encapsulate photodynamic agents or photothermal agents and STING agonists, achieving tumor-targeted enrichment and precise intracellular delivery. Light-triggered systems can coordinate treatment initiation and local drug release, but this does not by itself establish the optimal relative timing between phototherapy-induced antigen release and STING-mediated APC maturation.

In addition, tumor hypoxia is the core bottleneck for the efficacy of PDT. Nanocarriers can integrate the function of hypoxia alleviation. For example, Dong et al. have developed a nano-platform called cDVPMA, which contained photosensitizer verteporfin-phospholipid (VL), cGAMP and thrombin inhibitor dabigatran etexilate [179]. CDVPMA releases VL and cGAMP in the acidic TME. Then, VL mediates PDT under near-infrared (NIR) irradiation, generating singlet oxygen (¹O₂) which directly kills cancer cells and induces ICD of the cancer cells, thereby promoting the release of tumor antigens throughout the entire tumor and synergistically activating the cGAS-STING pathway through cGAMP. Dabigatran etexilate improves tumor microcirculation by inhibiting intratumoral thrombosis, alleviates hypoxia and reduces the level of immunosuppressive factors, providing a “clearing path” for T cell infiltration.

The combination of PTT and nanotechnology mediated STING agonists has also been extensively studied. For instance, Qiu et al. designed a Cu-ZnO2@PDA nano-platform, with copper ions serving as the core and polydopamine (PDA) forming the shell [180]. In the TME, nanoparticles release copper particles, zinc ions and hydrogen peroxide. Cu²⁺ and toxic hydroxyl radicals damage mitochondrial integrity and induce mtDNA leakage. Zn²⁺ can enhance the recognition sensitivity of cGAS to mtDNA, synergistically activating the cGAS-STING pathway. Moreover, the excellent photothermal conversion ability of the PDA coating, under NIR light, can further amplify oxidative stress and DNA damage, further enhancing the activation intensity of the cGAS-STING pathway.

The main translational limitation of phototherapy-based nano-STING strategies is lesion accessibility. Light penetration, energy delivery, tissue heating, hypoxia heterogeneity, and treatment of occult metastases must be considered before these platforms are positioned as systemic immunotherapy strategies.

Immune checkpoint inhibitors (ICIs)

ICIs are the cornerstone of clinical solid tumor immunotherapy. However, they are only effective for “hot tumors” with sufficient T-cell infiltration. Most solid tumors exhibit a “cold tumor” phenotype due to antigen presentation defects and an immunosuppressive TME. The low response rate of ICIs monotherapy is an important factor limiting its clinical application. The activation of cGAS-STING can form a positive feedback synergy with ICIs, which is characterized by “immune activation plus brake release”. Furthermore, nanotechnology can solve the core problems of mismatch between the action sites of STING agonists and ICIs antibodies and the large toxic side effects of systemic administration. They can achieve precise temporal regulation of the release of antibodies in the tumor stroma and agonists in the intracellular compartments of immune cells through TME responsiveness design. At the same time, they can enhance drug tumor accumulation and reduce immune-related adverse reactions by leveraging the EPR effect and targeting ligand modification. Therefore, the combination of STING activation with PD-1/PD-L1 blockade is one of the most mechanistically rational strategies. The two components of this combination need not enter the same cell. STING agonists generally require cytosolic delivery to APCs or selected tumor cells, whereas checkpoint antibodies act extracellularly on T cells, tumor cells, or myeloid cells. Thus, co-accumulation within the tumor does not demonstrate appropriate cellular partitioning. A sequential rationale—STING-mediated priming and T-cell recruitment followed by checkpoint blockade—is mechanistically attractive, particularly when STING induces adaptive PD-L1 expression, but most available nanomedicine studies have not directly shown that this sequence is superior to concurrent treatment. And its value should be tested in biomarker-defined settings where STING-induced T-cell recruitment and adaptive PD-L1 upregulation are demonstrable.

Antibody of cytotoxic T lymphocyte-associated antigen 4 (CTLA4)

The anti-CTLA4 antibody mainly acts during the lymph node immune initiation stage, blocking the binding of CTLA4 on T cells to the B7 molecules on the surface of DCs, eliminating the co-inhibitory signals that inhibit T cell activation, and promoting the clonal expansion of tumor-specific T cells [181]. However, its efficacy highly depends on the antigen presentation function of DCs. Activation of the cGAS-STING pathway can directly enhance the maturation and antigen cross-presentation ability of DCs, forming a deep synergy with CTLA4 blockade at the lymph node level. Studies have confirmed that the use of lipid-based nanomaterial platforms carrying Mn²⁺ and c-di-AMP combined with anti-CTLA4 antibodies can significantly prolong the survival period of mice models with acute myeloid leukemia [182].

Antibody of programmed cell death ligand 1/ programmed cell death protein 1 (PD-L1/PD-1)

The anti-PD-1/PD-L1 antibodies mainly act on the TME, blocking the binding of PD-1 on the surface of effector T cells to PD-L1 on tumor cells, thereby relieving the functional inhibition of T cells. However, their efficacy highly depends on the already existing infiltration of effector T cells in the tumor tissue. The activation of the cGAS-STING pathway can, on the one hand, induce the secretion of chemokines such as CXCL9 and CXCL10, recruiting CD8⁺ effector T cells from the circulation to migrate to the tumor tissue, reprogramming TAM from the immunosuppressive M2 type to the anti-tumor M1 type, reducing the infiltration and function of myeloid-derived suppressor cells (MDSC) and Tregs, thereby multi-dimensionally transforming “cold tumors” into “hot tumors”, solving the core bottleneck of the lack of target points for PD-1/PD-L1; on the other hand, the IFN-I produced by the activation of the pathway will induce the compensatory upregulation of PD-L1 expression in tumor cells and immune cells, this adaptive immune escape precisely provides a target point for PD-1/PD-L1 inhibitors, and the two form a positive feedback loop of “immune initiation- brake release”. The combination of nano-STING activation with PD-1/PD-L1 blockade is widely studied preclinically and is mechanistically attractive when STING induces T-cell recruitment together with adaptive PD-L1 upregulation; its clinical value remains to be established. For instance, Li et al. have developed a dual STING activation micellar system (D-SAM). Compared with the use of the anti-PD-1 antibody alone, the combined treatment of the D-SAM and anti-PD-1 antibody further enhanced the inhibition rates of both primary and distant tumors in MC38 mouse colon cancer model [117].

Nanotechnology can co-localize STING agonists and checkpoint-directed agents within the same tumor, but co-accumulation does not demonstrate delivery to the same cellular compartment or establish that concurrent release is superior to sequential treatment [183]. For example, studies have shown that by applying lipid nanoparticles carrying diABZI, which possess dual anti-CD47/PD-L1 targeting capabilities, the immunosuppressive TME was successfully reversed, and more than 60% of the tumors were eliminated in the mouse glioma model [184].

Chimeric antigen receptor T-cell (CAR-T) immunotherapy

CAR-T immunotherapy has achieved remarkable therapeutic effects in hematological malignancies such as B-cell lymphoma and multiple myeloma, but it still shows poor efficacy in solid tumors with dense physical barriers and immunosuppressive TME. The activation of the cGAS-STING pathway can improve the immune response pattern of solid tumors from multiple aspects such as chemokine secretion, enhanced antigen presentation, and remodeling of the immunosuppressive microenvironment, providing a favorable environment for CAR-T cells. Preclinical studies support the potential benefit of spatially restricted STING activation before or during CAR-T-cell therapy. For instance, a study has developed a nanovesicle carrying cGAMP, namely NVs@cGAMP, which selectively delivers the drug to cells with overexpressed PD-L1 and reshapes the TME, thereby enhancing the activity of CAR-T cells [185]. At the same time, the anti-PD-L1 single-chain variable fragment on the surface of the nanovesicle blocks the upregulation of PD-L1 induced by STING agonists, preventing the exhaustion of CAR-T cells. In a mouse model of metastatic lung cancer, inhaled aPD-L1 NVs@cGAMP accumulated in lung tumors, increased local IFN-I and chemokine signaling, and simultaneously blocked STING-induced PD-L1 upregulation, thereby improving CAR-T-cell activity in murine lung cancer models, providing preclinical proof of concept that spatially restricted STING activation can improve CAR-T activity in selected solid-tumor models. However, this study validated a specific combination regimen rather than directly comparing STING priming before CAR-T infusion with concurrent or post-infusion administration.

However, the inflammatory environment that facilitates CAR-T-cell recruitment may not be intrinsically beneficial to the CAR-T cells themselves. Direct STING activation in primary human T cells has been shown to induce apoptosis, inhibit proliferation, and impair glycolytic and mitochondrial metabolism. These effects are particularly relevant when STING agonist exposure overlaps with TCR or CAR signaling, because activated effector cells may be more sensitive to cellular stress. Consistently, high-intensity STING activation can produce local tumor ablation while compromising durable CD8⁺ T-cell-mediated immunity. Therefore, improved whole-tumor control should not be interpreted as evidence that CAR-T-cell fitness, expansion, and persistence are simultaneously improved.

This concern is further supported by recent evidence in pancreatic cancer models. Piseddu et al. found that combining diABZI with conventional CAR-T cells was constrained by detrimental T-cell-intrinsic STING signaling, whereas STING-deficient CAR-T cells exhibited greater proliferation, reduced exhaustion, an expanded effector-memory phenotype, and superior tumor control [186]. These findings demonstrate that STING activation in tumor or innate immune compartments may benefit the combination while simultaneous STING activation within the therapeutic effector cells may be detrimental.

Accordingly, future CAR-T combinations should compare at least three schedules: STING priming before CAR-T infusion, concurrent treatment, and STING administration after CAR-T accumulation. Evaluation should include not only tumor regression but also CAR-T-cell biodistribution, expansion, persistence, apoptosis, metabolic fitness, exhaustion and memory phenotypes, together with systemic cytokines and markers of cytokine-release syndrome. APC-selective, tumor-cell-selective, or locally restricted STING delivery may be preferable to formulations that expose CAR-T cells directly to high agonist concentrations.

Box 1.

Special considerations for nano-STING therapy in hematologic malignancies

Although most nano-STING platforms have been developed in solid tumor models, hematologic malignancies deserve specific attention. Acute myeloid leukemia (AML), lymphoma, and multiple myeloma provide clinically relevant settings in which STING agonism could be explored as an immune-priming strategy, as a potential means of achieving durable post-remission or residual-disease control, and as a partner for established therapies or emerging immunotherapies [182, 187]. In selected malignant B-cell contexts, STING activation may also exert direct tumor-intrinsic effects, but the more general rationale is to activate antigen-presenting cells, remodel suppressive myeloid niches, and enhance T-cell or NK-cell antitumor responses [55, 188].

The biodistribution of nanoparticles creates both an opportunity and a risk in this disease group. Spleen, lymph nodes, and liver are major immune and reticuloendothelial compartments; they are also common sites of nanoparticle uptake [189, 190]. For hematologic cancers, this distribution may help deliver STING agonists to marrow or lymphoid immune niches where leukemia, lymphoma, or myeloma cells interact with dendritic cells, macrophages, stromal cells, and T cells [191, 192]. However, the same distribution may also increase the risk of systemic cytokine release, myeloid activation, vascular inflammation, hepatic or splenic toxicity, and damage to normal hematopoietic progenitors [25, 26]. Therefore, marrow exposure should not be treated as automatically beneficial; it must be linked to tumor or APC-selective activation and carefully measured with repeated-dose pharmacodynamics [193, 194].

a) AML. AML provides the most direct evidence supporting nano-STING therapy in a disseminated hematologic malignancy. Its bone-marrow microenvironment is characterized by defective innate immune activation, impaired antigen presentation, T-cell tolerance, and suppressive myeloid populations, providing a rationale for using STING agonism to restore immune recognition and establish antileukemic memory. In a disseminated C1498 AML model, intravenous administration of lipid-based cyclic dinucleotide–manganese particles carrying c-di-AMP and Mn²⁺ increased agonist delivery to the spleen, lymph nodes, and bone marrow [182]. Treatment activated innate and adaptive immune responses, reduced leukemia burden, and, when combined with CTLA-4 blockade, prolonged survival and generated protection against AML rechallenge. A separate bioinspired nanomedicine combined leukemia-directed treatment with amplification of cGAS–STING signaling and produced systemic antileukemic chemoimmunotherapy in preclinical models [195]. These studies provide direct support for nanocarrier-enabled STING activation in AML but do not yet establish which marrow-resident cell populations are primarily responsible for efficacy. The respective contributions of leukemia-cell-intrinsic STING, dendritic cells, macrophages, and lymphocytes therefore require further clarification. Clinical translation has begun with the phase I evaluation of the systemic STING agonist CRD3874-SI in relapsed or refractory AML (NCT06626633), although clinical efficacy data are not yet available.

b) Lymphomas and other malignant B-cell diseases. STING agonism may have a dual function in selected B-cell malignancies. In addition to activating host antigen-presenting cells, STING agonists can directly induce mitochondrial apoptosis in malignant B cells [55]. Prolonged STING signaling, associated with relatively inefficient STING degradation in B cells and modulation of the IRE1–XBP1 stress-response pathway, induced apoptosis and tumor regression in the Eμ-TCL1 model of chronic lymphocytic leukemia. In diffuse large B-cell lymphoma (DLBCL), SAMHD1 deficiency increased cytosolic DNA and STING signaling, leading to apoptosis, necroptosis, and pyroptosis-related PANoptosis [196]. STING activation also increased PD-L1 expression and enhanced the experimental activity of PD-L1 blockade. However, these observations should not be generalized to all lymphoma subtypes because STING expression, downstream signaling competence, cell-death machinery, and immune context vary substantially among B-cell and T-cell malignancies. Moreover, some DLBCL experiments used DMXAA, which preferentially activates murine rather than human STING and therefore has limited direct translational relevance. Several early-phase trials of STING agonists, including E7766, MK-1454, and SNX281, permitted the enrollment of patients with lymphoma, but subtype-specific clinical activity has not been established. Nano-STING strategies that target lymph nodes, spleen, malignant B cells, or lymphoma-associated antigen-presenting cells remain attractive, but direct evidence for lymphoma-specific nanodelivery is currently limited.

c) Multiple myeloma. STING activation may also be relevant to multiple myeloma through both direct plasma-cell killing and remodeling of the immunosuppressive bone-marrow niche. The STING agonist 3′,3′-cGAMP induced apoptosis and tumor regression in syngeneic and immunodeficient myeloma models, indicating that antimyeloma activity can occur independently of an intact adaptive immune response [55]. Additional studies showed that bortezomib-induced DNA damage and immunogenic cell death can engage cGAS–STING and type I interferon signaling [197]. Combining bortezomib with a STING agonist and immune-checkpoint blockade produced greater antimyeloma activity in preclinical models than the individual treatments [198]. Nevertheless, these findings do not yet establish the safety or clinical benefit of such combinations, particularly because immune-checkpoint inhibitors have encountered disease-specific safety and efficacy limitations in multiple myeloma. BCMA-, CD38-, or marrow-directed delivery of STING agonists could theoretically increase exposure of malignant plasma cells while reducing systemic inflammation, but these approaches should presently be described as design hypotheses rather than validated nano-STING strategies.

Combination strategies are particularly attractive but should be developed with modality-specific safeguards. Preclinical studies indicate that STING agonism can enhance CAR-T-cell trafficking, persistence, and antitumor activity by inducing type I interferons and chemokines, activating antigen-presenting cells, and establishing a locally inflammatory immune niche; nano- or biomaterial-based delivery has directly supported this concept in solid-tumor models [185, 199]. In lymphoma and multiple myeloma, combinations with PD-1/PD-L1 blockade have disease-specific preclinical support, whereas CD47–SIRPα blockade and macrophage reprogramming are mechanistically rational but remain less directly validated in hematologic nano-STING models [200]. Nevertheless, CAR-T cells and other immune-effector therapies are already associated with cytokine release syndrome, immune effector cell-associated neurotoxicity syndrome, and prolonged cytopenias. Additional STING-driven cytokine and myeloid activation could plausibly increase this inflammatory or hematologic burden, although additive toxicity has not yet been demonstrated clinically [26, 201]. Future studies should therefore incorporate serial complete blood counts, cytokine kinetics, immune-cell subset profiling, and predefined dose-hold and hematologic-recovery criteria; bone-marrow evaluation should be considered in preclinical studies and in patients with persistent or unexplained cytopenias [201, 202]. For hematologic malignancies, the preferred nano-STING product may ultimately be one that produces a brief, titratable, and cell-selective immune pulse within marrow or lymphoid tumor niches while preserving normal hematopoiesis and limiting systemic immune activation—a design objective that remains to be prospectively validated.

Others

In addition to the aforementioned treatment methods, the nanotechnology-mediated STING agonists can also be combined with treatments such as ferroptosis and indoleamine 2,3-dioxygenase (IDO) targeted therapy.

Ferroptosis-STING axis is an important anti-tumor mechanism discovered recently [203]. Ferroptosis is a type of iron-dependent programmed cell death characterized by excessive accumulation of lipid peroxides, distinct from apoptosis, necrosis, and pyroptosis. The immunogenic cell death induced by ferroptosis can release a large number of damage-related molecular patterns and cytoplasmic DNA, providing endogenous activating substrates for the cGAS-STING pathway. After activation of the cGAS-STING pathway, the I-type interferons and pro-inflammatory factors secreted can further amplify the ferroptosis effect by regulating tumor cell lipid metabolism and down-regulating the expression of glutathione peroxidase 4 (GPX4), etc. Together, they form a positive feedback loop of “ferroptosis-immune activation”, which is an important synergistic strategy for breaking through the immunosuppression of solid tumors and improving the treatment response rate. Some studies have utilized this strategy to design a nano-platform, which co-cloads ferroptosis inducers, STING agonists, and PD-L1 antibodies to synergistically activate ferroptosis and immune responses, achieving efficient anti-tumor treatment [204].

While, IDO is the core rate-limiting enzyme of the tumor immune-suppressive metabolic pathway, which can catalyze the conversion of the basic amino acid tryptophan (Trp) involved in T cell activation and proliferation into kynurenine (Kyn) [205]. Furthermore, the overexpression of IDO and Kyn in tumors can activate Tregs by binding to the aryl hydrocarbon receptor (AHR), thereby inhibiting the functions of effector T cells (Teffs) and natural killer cells (NKs). This is one of the core mechanisms underlying the resistance to immunotherapy in solid tumors. However, the use of drugs that inhibit IDO alone is ineffective for the majority of cancer patients. Meanwhile, studies have confirmed that STING agonists alone can upregulate the expression of IDO in the TME, forming a negative feedback loop of “immune activation-metabolic inhibition”, ultimately leading to impaired T cell function and immune escape, severely limiting the efficacy of STING agonists [4, 206]. Based on this mechanism, some nanoplatforms achieve efficient anti-tumor effects by co-loading stinging agent and IDO inhibitor [207, 208].

Overall, these combination strategies show that nano-STING therapy is most useful when it addresses a defined immunological bottleneck, such as insufficient antigen release, inadequate T-cell recruitment, metabolic feedback, or myeloid suppression. Future development should prioritize mechanism-based sequencing and measurable immune conversion over increasingly complex all-in-one formulations.

Repeated dosing and feedback-dependent scheduling

Evidence defining the optimal frequency of nano-STING administration remains limited. Repeated dosing may replenish local innate activation after transient pathway engagement, but it may also increase systemic cytokine exposure, directly impair infiltrating T cells, or reinforce adaptive and metabolic feedback pathways such as PD-L1, IDO, IL-6/STAT3, and myeloid suppression. Moreover, repeated drug administration should not be equated with repeated productive immune priming: later doses may encounter mature or refractory APCs, altered myeloid populations, or already activated effector cells with different sensitivity to STING signaling. Future studies should therefore compare single-pulse, fractionated, and repeated regimens using time-resolved measurements of tumor antigen availability, APC uptake and maturation, lymph-node cross-presentation, effector-cell recruitment, and systemic cytokines. At present, the superiority of repeated nano-STING dosing over a properly timed single immune pulse remains a mechanism-based hypothesis rather than an established principle.

Future perspectives: biomarker-guided and translation-by-design development

Future progress will depend less on adding new carrier classes than on linking delivery to patient selection, manufacturability, pharmacodynamic evidence, and safety. Early trials have demonstrated interferon-related activity, but modest response rates show that target engagement alone is insufficient. A clinically useful product must generate a controllable STING release in the intended human cell compartment, convert innate activation into durable adaptive immunity, and remain reproducible at clinical scale. The most valuable nano-platform will therefore be the simplest one that widens the therapeutic window and supports a clear clinical decision path.

Cell- and compartment-specific delivery

The first priority is to deliver the agonist to the cell population that can convert innate sensing into antitumor immunity. Tumor accumulation alone is an inadequate endpoint: a carrier may reach the lesion yet remain extracellular, become trapped in endolysosomes, or preferentially activate tumor cells, endothelial cells, or suppressive myeloid subsets. Future platforms should therefore be designed around cytosolic delivery to functionally competent dendritic cells and selected macrophage populations in tumors and tumor-draining lymph nodes, while limiting exposure in blood and healthy organs. Targeting ligands, biomimetic membranes, ionizable materials, and programmed endosomal escape are valuable only when cell-resolved biodistribution and pathway activation are demonstrated in human-relevant models.

Route- and disease-specific delivery

No single administration route is suitable for every clinical setting. Intravenous nanoparticles are most relevant to disseminated or inaccessible disease but require strong control of hepatic, splenic, and vascular exposure. Intratumoral formulations can establish local proof of mechanism, whereas hydrogels or implantable systems may be advantageous in the perioperative setting by treating residual disease at the resection margin. Inhaled or intravesical formulations may provide regional exposure for lung or bladder tumors. Development should therefore begin with the anatomy and treatment setting of the intended indication, then select the least complex carrier capable of reaching both the primary target compartment and clinically important metastatic sites.

Mechanism-based combination and sequencing

Combination therapy should address a defined bottleneck rather than simply co-load multiple active agents. Checkpoint blockade is most rational when STING activation produces T-cell infiltration together with adaptive PD-L1 upregulation; IDO or adenosine-pathway inhibition is appropriate when metabolic feedback is demonstrated; and radiotherapy or DNA-damaging therapy may be used when additional immunogenic DNA is required. Nanotechnology adds particular value by separating incompatible drugs, coordinating their distribution, and imposing a sequence—for example, innate priming before checkpoint blockade or CAR-T-cell exposure. Simpler modular regimens should be prioritized over highly complex all-in-one particles unless co-delivery or sequential release provides a measurable advantage.

Translation by design

The final priority is to treat manufacturability and clinical decision-making as design inputs. Platforms should use scalable components and processes, with defined critical quality attributes for particle size, loading, release, potency, sterility, storage stability, and batch consistency. Early trials should not rely on peripheral interferon-related signals alone. A practical go sequence is to demonstrate reproducible manufacture, route-appropriate pharmacokinetics, delivery to the intended human cell compartment, conversion to adaptive immunity, and an acceptable repeated-dose safety margin. Dose escalation should stop when systemic inflammation increases without corresponding tumor immune conversion.

Biomarker-guided patient selection

A clinically useful nano-STING strategy should not only improve delivery, but also identify patients whose tumors and host immune compartments are capable of responding to STING activation. The modest clinical activity observed with several first-generation STING agonists indicates that pharmacodynamic induction of interferon-related cytokines is not sufficient by itself. Patient heterogeneity in STING genotype, pathway expression, innate immune tone, myeloid composition, metabolic resistance, and baseline T-cell inflammation may dilute efficacy signals in unselected early-phase trials. Therefore, biomarker-guided patient selection should become an integral part of next-generation nano-STING development rather than a post hoc exploratory analysis. Candidate biomarker categories and their potential use in nano-STING clinical development are summarized in Table 4.

Table 4.

Candidate biomarkers for patient selection in nano-STING therapy

Biomarker	Rationale / what it measures	Implication for nano-STING development
TMEM173/STING genotype	Human STING variants can differ in ligand binding and signaling response.	Prefer agonists with broad human-variant activity; consider genotype when a ligand has variant-dependent potency.
Tumor-cell STING expression/function	Direct STING agonists require functional STING in the intended responding tumor cell.	Low or silenced tumor-cell STING argues against tumor-cell-directed agonism and may shift the target toward host APCs.
Tumor-cell cGAS expression/function	Indirect DNA-damage strategies and cGAS-sensitizing platforms require upstream cGAS competence; direct STING agonists can bypass cGAS if STING remains functional.	Use cGAS status to distinguish patients more likely to respond to indirect/cGAS-sensitizing approaches from those better suited to direct STING agonists.
ENPP1/CD73/adenosine axis	ENPP1 can degrade extracellular cGAMP/CDN-like agonists; CD73-derived adenosine suppresses antitumor immunity.	Use degradation-resistant agonists/protective carriers or combine with ENPP1/CD73/adenosine-pathway blockade when this resistance module is active.
IFN-I / ISG signature	Reflects baseline innate immune tone and inducibility versus chronic inflammatory activation.	Differentiate immune-silent tumors from chronically inflamed or feedback-dominated settings; use on-treatment change as a pharmacodynamic measure rather than a stand-alone efficacy surrogate.
cDC1 / APC abundance and state	Functional APCs are required to convert STING activation into cross-presentation and T-cell priming.	Support APC-targeted formulations and cell-resolved pharmacodynamic endpoints.
T-cell-inflamed phenotype / PD-L1	Indicates whether STING-induced priming may be consolidated by checkpoint blockade.	Guide use and timing of PD-1/PD-L1 blockade; lack of T-cell recruitment may require an additional antigen-releasing or myeloid-reprogramming step.

First, TMEM173/STING genotype should be considered because human STING variants can differ in ligand binding and downstream signaling intensity. Genotyping is most relevant when a candidate agonist shows variant-dependent activity.

Second, cGAS and STING should be assessed separately rather than treated as a single biomarker. Functional STING is required for direct STING agonism in the intended responding cell, whereas cGAS is additionally required for strategies that depend on endogenous DNA sensing or cGAS sensitization. Loss of tumor-cell cGAS may therefore reduce the effect of DNA-damage- or Mn²⁺-dependent activation without necessarily abolishing the response to a direct STING agonist if STING remains functional. Conversely, tumor-cell STING loss or silencing argues against tumor-cell-directed agonism and may shift the therapeutic target toward host antigen-presenting cells.

Third, the ENPP1/CD73/adenosine axis represents a key metabolic resistance module. ENPP1 can degrade extracellular cGAMP or CDN-like agonists, whereas CD73-mediated adenosine production suppresses T-cell, NK-cell, dendritic-cell, and macrophage function [109, 209, 210]. High ENPP1 or CD73 expression may therefore predict rapid agonist inactivation and immunosuppressive feedback, supporting combination strategies with ENPP1, CD73, or adenosine-receptor blockade.

Fourth, baseline IFN-I/ISG signatures may indicate whether the tumor is immunologically silent, inducible, or already chronically inflamed. Low baseline ISG expression may define tumors requiring innate immune priming, whereas persistently high inflammatory signatures may indicate negative feedback, immune exhaustion, or a higher risk of systemic inflammatory toxicity [211, 212].

Fifth, the abundance and functional state of cDC1s and other antigen-presenting cells should be assessed because STING activation must be converted into antigen cross-presentation and T-cell priming [213]. Multiplex immunohistochemistry, flow cytometry, RNA profiling, or single-cell sequencing can be used to quantify cDC1s, macrophage subsets, MDSCs, and tumor-draining lymph node immune states. Finally, T-cell-inflamed phenotype and PD-L1 expression should guide rational combination with immune checkpoint inhibitors. STING activation is most likely to synergize with PD-1/PD-L1 blockade when it recruits cytotoxic T cells and simultaneously induces adaptive PD-L1 upregulation [214]. By contrast, tumors lacking antigen presentation, APC infiltration, or T-cell recruitment may require additional antigen-releasing therapy, myeloid reprogramming, or regional delivery before checkpoint blockade is added.

In future clinical trials, biomarker panels should be incorporated into enrollment, dose escalation, and response evaluation. Pre- and on-treatment biopsies should measure tumor and immune-cell STING activation, cDC1 recruitment, CXCL9/CXCL10 induction, PD-L1 or IDO feedback, ENPP1/CD73 activity, and T-cell infiltration [20, 60]. Peripheral cytokines alone should be treated as incomplete pharmacodynamic markers. A practical biomarker framework would classify patients as pathway-competent, metabolically resistant, APC-deficient, T-cell-inflamed, or high-risk for systemic inflammation, and then match them to the appropriate nanoformulation, route, dose schedule, and combination partner.

Conclusions

Nanoengineering offers several ways to improve STING-directed cancer therapy, but improved delivery should not be equated automatically with immunodynamic control. The strongest evidence comes from platforms that connect formulation design to cell- or tissue-resolved pharmacodynamics, time-resolved signaling, efficacy, and toxicity. Some systems clearly improve pharmacokinetics or cytosolic access; a smaller subset directly alters the duration or sequence of STING signaling; and only limited preclinical studies provide sufficiently comparative data to argue for a wider therapeutic window.

Future development should therefore move toward, rather than assume, immunodynamic control. The key objective is to match agonist chemistry, carrier, route, target cell, signaling duration, biomarker-defined population, and combination schedule while maintaining an interpretable benefit–risk relationship. Clinical translation will require human-relevant pathway testing, prospective biomarkers, cell-resolved pharmacodynamics, repeated-dose safety, and manufacturable formulations. These criteria can distinguish nano-STING platforms that genuinely improve therapeutic control from those that primarily increase drug exposure.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADC

Antibody-drug conjugate

AE

Adverse event

AML

Acute myeloid leukemia

APC

Antigen-presenting cell

CAR-T

Chimeric antigen receptor T-cell

CDNs

Cyclic dinucleotides

cDC1

type 1 conventional dendritic cell

cGAMP

cyclic GMP-AMP

cGAS

cyclic GMP-AMP synthase

CMC

Chemistry, manufacturing and controls

CTLA-4

Cytotoxic T-lymphocyte-associated protein 4

DC

Dendritic cell

ENPP1

Ectonucleotide pyrophosphatase/phosphodiesterase 1

ER

Endoplasmic reticulum

EV

Extracellular vesicle

HNSCC

Head and neck squamous cell carcinoma

ICD

Immunogenic cell death

ICI

Immune checkpoint inhibitor

IDO

Indoleamine 2,3-dioxygenase

IFN

Interferon

ISAC

Immune-stimulating antibody conjugate

ISG

Interferon-stimulated gene

LNP

Lipid nanoparticle

MDSC

Myeloid-derived suppressor cell

MOF

Metal-organic framework

NF-κB

Nuclear factor kappa B

NK

Natural killer

ORR

Objective response rate

PARP

Poly(ADP-ribose) polymerase

PBMC

Peripheral blood mononuclear cell

PD-1

Programmed cell death protein 1

PD-L1

Programmed cell death ligand 1

PK/PD

Pharmacokinetics/pharmacodynamics

PLGA

Poly(lactic-co-glycolic acid)

ROS

Reactive oxygen species

RT

Radiotherapy

STING

Stimulator of interferon genes

TAM

Tumor-associated macrophage

TBK1

TANK-binding kinase 1

TME

Tumor microenvironment

TMEM173

Transmembrane protein 173

Treg

Regulatory T cell

Author contributions

NM.L. was the major contributors in writing the manuscript. XY.K. was the major contributors in preparing the tables and figures. XY.W. and WX.Z. were the main reviser of the manuscript. Y.K. and B.M.P. performed literature collection. J.W. and ZZ.W. made substantial contributions to the design of the work. All authors read and approved the final manuscript.

Funding

This work was supported by the Natural Science Foundation of China (No. 82371842), the Natural Science Foundation of China (No. 82203885), the CAMS Innovation Fund for Medical Sciences (CIFMS) (No. 2024-I2M-ZH-006), the CAMS Innovation Fund for Medical Sciences (CIFMS) (No. 2024-I2M-C&T-A-004), the CAMS Innovation Fund for Medical Sciences (CIFMS) (No. 2024-I2M-C&T-B-057), National High Level Hospital Clinical Research Funding (No. 80102022609), the Beijing Municipal Natural Science Foundation (No. 7252117), the Beijing Municipal Natural Science Foundation (No. L234043), the Natural Science Foundation of China (No. 82473205), CAMS Innovation Fund for Medical Sciences (CIFMS) (No.2025-I2M-XHXX-056), the National High Level Hospital Clinical Research Funding and the Cooperation Fund of CHCAMS (No. CFA202502003 and No. CFA202201010).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.