Targeting redox metabolism and low-dose radiotherapy synergistically activate cGAS-STING pathway to improve NK cell therapy efficacy in hepatocellular carcinoma

Abstract

Adoptive natural killer (NK) cell therapy (ANKCT) is a promising strategy for hepatocellular carcinoma (HCC); however, its efficacy is hampered by insufficient NK cell homing and the immunosuppressive activity of M2-polarized tumor-associated macrophages (TAMs). Activation of the cyclic guanosine monophosphate (GMP)–adenosine monophosphate synthase (AMP) synthase–stimulator of interferon genes (cGAS–STING) pathway enhances NK-cell recruitment and reprograms TAMs toward a proinflammatory M1 phenotype. Radiation therapy (RT) activates the cGAS–STING pathway by inducing reactive oxygen species (ROS)-mediated DNA damage. Compared with high-dose irradiation, low-dose radiotherapy (LDRT) offers advantages, including reduced toxicity and enhanced antitumor immunity. However, the robust thioredoxin (Trx) and glutathione (GSH) antioxidant systems in HCC inhibit LDRT-induced ROS accumulation, thereby limiting immune activation. These limitations highlight the need for auxiliary strategies to complement LDRT-induced immunogenicity. Here, we developed a biomimetic nanoparticle in which auranofin-loaded MOF-199 is cloaked with tumor cell membranes (A@MMOF). A@MMOF disrupts tumor redox homeostasis by irreversibly inhibiting the GSH and Trx antioxidant systems while inducing GSH-dependent Cu²⁺ release from the MOF framework. The subsequent reduction of Cu²⁺ to Cu⁺ catalyzes Fenton-like reactions, markedly amplifying the intracellular ROS levels. By providing a sustained redox-driven stimulus, A@MMOF compensates for the insufficient oxidative stress induced by LDRT, leading to robust activation of the cGAS–STING pathway. Thus, A@MMOF synergizes with LDRT to remodel the tumor microenvironment, enhance NK cell infiltration and activation, and ultimately improve the efficacy of ANKCT in HCC.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04171-1.

Introduction

Hepatocellular carcinoma (HCC) is one of the most prevalent and lethal cancers in the world, ranking as the third leading cause of cancer-related mortality globally [1]. In China, over 50% of patients are diagnosed at an advanced stage [2]. Despite ongoing advances in surgical techniques and comprehensive therapies, the prognosis of HCC remains poor, with a 5-year survival rate of only 12.1% [3, 4], underscoring the urgent need for more effective and safer therapeutic strategies.

Recently, immunotherapy has emerged as a major focus in HCC research, with strategies such as immune checkpoint inhibitors, dendritic cell vaccines, and adoptive natural killer cell therapy (ANKCT) under active investigation [3]. NK cells are pivotal effectors of the innate immune system and, unlike T lymphocytes, possess the inherent ability to recognize and eliminate tumor stem cells in a human leukocyte antigen-independent manner, thereby bypassing the requirement for antigen processing and presentation [5]. ANKCT also confers unique benefits, including “off-the-shelf” applicability, reduced toxicity, and simplified manufacturing compared with T cell–based approaches [6]. ANKCT has demonstrated substantial clinical activity in hematologic malignancies, achieving high remission rates with an excellent safety profile [6]. In advanced HCC, NK cells exhibit reduced frequency and impaired functional capacity, deficits that are strongly associated with poor clinical outcomes [7]. This immunological impairment provides a strong rationale for ANKCT to restore NK cell–mediated antitumor immunity. Consistently, early-phase clinical studies have demonstrated that ANKCT is safe and well-tolerated in patients with HCC, supporting its therapeutic potential [8]. Nevertheless, despite these encouraging findings, the clinical efficacy of ANKCT as monotherapy remains limited, with postoperative recurrence rates approaching 40% within 1 year [9]. A major limitation is the inefficient homing of adoptively transferred NK cells to tumor sites, which significantly limits therapeutic benefit [8, 10]. Moreover, the tumor microenvironment (TME) in HCC is enriched with M2-polarized tumor-associated macrophages (TAMs), a dominant immunosuppressive population that promotes NK cell dysfunction and exhaustion [5, 11, 12]. Therefore, therapeutic strategies that enhance NK cell homing and reprogram TAMs toward an M1 phenotype are essential for improving the efficacy of ANKCT in HCC.

The cyclic GMP–AMP (cGAMP) synthase–stimulator of interferon genes (cGAS–STING) pathway, initially identified for its role in antiviral defense and innate immunity, has recently emerged as a key regulator of antitumor immunity [13]. Under conditions such as DNA damage or mitochondrial stress, cytosolic double-stranded DNA (dsDNA) activates cGAS, leading to cGAMP synthesis, which in turn triggers the STING–interferon regulatory factor 3 (STING–IRF3) signaling axis. Notably, activation of this pathway promotes NK cell–mediated innate antitumor immunity [13] by inducing chemokines such as C-X-C motif chemokine ligand 10 (CXCL10) and C-C motif chemokine ligand 5 (CCL5), which are essential for recruiting NK cells [14]. Beyond its intracellular signaling role, extracellular cGAMP functions as a “second messenger” by entering TAMs through solute carrier family 19 member 1 (SLC19A1) or volume-regulated anion channels (VRAC), promoting M1 polarization and further amplifying innate immune-mediated antitumor responses [13, 15, 16]. Nevertheless, most HCC cells were insufficiently activated in the cGAS–STING pathway [17]. Radiotherapy (RT) is a cost-effective and widely accessible local therapy in HCC [3, 4]. Mechanistically, it induces oxidative stress and DNA damage through ionizing radiation, generating cytosolic dsDNA that activates the cGAS–STING pathway [18]. However, HCC cells possess intrinsic resistance mechanisms that attenuate RT-induced cGAS–STING activation [19]. Although dose escalation may enhance tumoricidal effects, it also increases the risk of radiation-induced liver disease and immunosuppression, restricting its clinical application [20, 21]. In contrast, low-dose radiotherapy (LDRT) has gained increasing attention owing to its favorable safety profile and immunomodulatory potential, making it a promising adjunct to ANKCT [20]. However, previous studies have shown that LDRT plus ANKCT often produces suboptimal results in HCC, typically requiring increased radiation doses to achieve efficacy [22, 23]. This limitation may arise from LDRT’s inability to generate sufficient oxidative stress to produce robust cGAS–STING activation, highlighting the need for auxiliary strategies to amplify its immunomodulatory effects [24].

The limited ability of LDRT to induce sufficient oxidative stress is primarily attributable to the robust antioxidant defense systems of tumor cells, which are dominated by the glutathione (GSH) and thioredoxin (Trx) metabolic systems [25]. Mechanistically, GSH detoxifies reactive oxygen species (ROS), such as H₂O₂ and lipid peroxides, through glutathione peroxidase (GPx)-mediated catalysis, converting reduced GSH into glutathione disulfide (GSSG). In parallel, the Trx system depends on thioredoxin reductase (TrxR) to regenerate reduced Trx [Trx-(SH)₂] from its oxidized form, thereby collaboratively maintaining intracellular redox homeostasis [26]. Importantly, these two systems engage in extensive crosstalk and synergistic interactions, together forming a robust antioxidant defense in tumor cells [26]. Consequently, their combined targeting alongside LDRT emerges as a promising strategy to effectively activate the cGAS–STING pathway and potentiate NK cell-mediated antitumor responses. Among potential agents, auranofin (AUR), an FDA-approved drug traditionally used to treat rheumatoid arthritis, has demonstrated potent antitumor activity across multiple cancer types by irreversibly inhibiting TrxR and promoting ROS accumulation [27]. Nevertheless, the clinical utility of AUR is constrained by dose-limiting toxicities, including nephrotoxicity and myelosuppression, as well as variable pharmacokinetic profiles [27]. In addition, divalent copper ions (Cu²⁺) provide an additional strategy for disrupting redox balance. Cu²⁺ reacts with GSH to generate GSSG while being reduced to Cu⁺, a process that simultaneously depletes intracellular GSH and promotes hydroxyl radical (•OH) generation via Fenton-like reactions [28]. Metal–organic framework 199 (MOF-199), a copper-based nanomaterial, provides an ideal platform to exploit this mechanism through GSH-responsive Cu²⁺ release [28]. Moreover, MOF-199 functions as a versatile nanocarrier with high surface area and efficient drug loading–release capacity while passively targeting tumors through the enhanced permeability and retention effect [28]. However, its in vivo circulation stability is often compromised by nonspecific uptake and rapid clearance mediated by phagocytic cells of the reticuloendothelial system (RES) [29]. To address this limitation, biomimetic membrane-coating strategies have been developed to confer immune evasion, reduce off-target clearance, and markedly improve tumor accumulation [29].

Here, we introduce a synergistic nanoparticle system designed to overcome the redox barriers that LDRT alone cannot effectively address, thereby enhancing the therapeutic efficacy of ANKCT in HCC. In this strategy, AUR is encapsulated into MOF-199 to generate A@MOF nanoparticles, which are subsequently cloaked with homologous tumor cell membranes to form A@MMOF. A@MMOF enables coordinated disruption of the GSH and Trx antioxidant systems, providing the additional redox-driven stimulus required to potentiate the immunological effects of LDRT. By amplifying ROS generation and robustly activating the cGAS–STING pathway, A@MMOF synergizes with LDRT to remodel the TME into a state more permissive for NK cell recruitment and effector activation. Consequently, this integrated approach strengthens the overall therapeutic efficacy of ANKCT in HCC (Fig. 1).

Fig. 1.

Schematic illustration of A@MMOF synthesis and the mechanism underlying its synergistic antitumor activity. (A) A@MOF is coated with a tumor cell membrane to form A@MMOF nanoparticles. (B) In the tumor microenvironment,GSH–triggered release of AUR and Cu²⁺ disrupts redox homeostasis and enhances ROS generation under LDRT. This process subsequently activates cGAS–STING signaling, facilitating NK cell infiltration and M1 polarization of TAMs, thereby potentiating ANKCT

Materials and methods

Synthesis of MOF-199 and A@MOF

To provide a clear overview of the fabrication process, a schematic diagram illustrating the step-by-step preparation of MOF-199, A@MOF, and A@MMOF, as well as the mechanistic role of each component, is shown in Fig. S1.

As depicted in this figure, MOF-199 was synthesized following a previously reported method [28, 30]. Trimesic acid (1.25 mmol) and copper nitrate trihydrate (2.85 mmol) were separately added into a mixture of ethanol and deionized water (10 mL, v/v, 1:1). After mixing the above solutions, triethylamine (250 µL) was slowly added, and the mixture was stirred for 30 min at room temperature. After centrifugation, the product was washed twice with water and three times with ethanol. The final product was stored in anhydrous ethanol at room temperature.

To prepare A@MOF, different weight ratios of MOF-199 to AUR (1:0.1, 1:0.5, 1:1, and 1:2, w/w) were prepared by mixing 1 mg of MOF-199 with the corresponding amount of AUR in 1 mL of ethanol. The mixture was stirred at room temperature for 24 h to ensure complete drug loading. After solvent removal via vacuum freeze-drying, the resulting product was redispersed in ultrapure water by sonication. The dispersion was then subjected to three rounds of centrifugation (10,000 rpm, 5 min) and washing to remove any unencapsulated drug. Residual free AUR was eliminated by dialysis (MWCO 1000 Da) for 24 h.

Preparation of A@MMOF

Cell membranes were extracted from Hepa1-6 cells as previously described [30], then lysed in ice-cold buffer and manually homogenized (50 strokes). After centrifugation at 700 × g for 10 min at 4 °C to remove nuclei and debris, the supernatant was collected and further centrifuged at 14,000 × g for 30 min at 4 °C to isolate membrane fragments. The membrane concentration was quantified using a BCA protein assay kit (Beyotime Biotechnology, China).

Referring to protocols used in drug-loaded MOF-199 systems [30], A@MMOF nanoparticles were prepared by mixing 1 mg of A@MOF with predetermined amounts of Hepa1-6 cell membrane at various weight ratios (1:0.5, 1:1, and 1:2; A@MOF: tumor cell membrane, w/w). The mixture was gently agitated and incubated at room temperature for 2 h to facilitate membrane–core interaction. Subsequently, the mixture was extruded 10 times through a 200 nm polycarbonate membrane using an Avanti mini-extruder to promote uniform membrane coating. The resulting A@MMOF nanoparticles were collected by centrifugation at 10,000 rpm for 15 min and washed to remove excess free membranes.

Characterization of MOF-199, A@MOF, and A@MMOF

The morphology and elemental distribution of MOF-199, A@MOF, and A@MMOF were examined using transmission electron microscopy (TEM; FEI Tecnai G2 12, USA) along with energy-dispersive spectroscopy (EDS) mapping. Particle size and zeta potential were measured by dynamic light scattering (DLS; Nano ZS, Malvern, UK). The colloidal stability of A@MMOF was evaluated by monitoring the hydrodynamic size daily for 7 days. The drug loading efficiency (DLE) and encapsulation efficiency (EE) of A@MOF and A@MMOF were determined by dissolving 20 mg of lyophilized nanoparticles in PBS. DLE and EE were calculated using the following equations: DLE (%) = (weight of loaded AUR / weight of nanoparticles) × 100,EE (%) = (weight of loaded AUR / weight of initially added AUR) × 100.The content of AUR was quantified using high-performance liquid chromatography (HPLC; Waters e2695) equipped with a diode array detector and a Phenomenex Gemini C18 column (250 × 4.6 mm, 5 μm). The mobile phase consisted of methanol and water (7:3, v/v) containing 0.1% (v/v) acetic acid. The detection wavelength was set at 220 nm. The flow rate was 0.6 mL/min, and the injection volume was 10 µL.The surface elemental composition of Cu and Au in A@MMOF was determined by X-ray photoelectron spectroscopy (XPS; Thermo Scientific K-Alpha), and the crystal structure was examined by X-ray diffraction (XRD; X’Pert PRO).

SDS-PAGE analysis of cell membrane coating

To characterize the membrane protein composition of the coated nanoparticles, Hepa1-6 cell membrane, MOF-199, and A@MMOF were mixed with loading buffer and heated at 100 °C for 10 min. Subsequently, 10 µL of each sample was loaded onto a 10% precast Tris–glycine gel and subjected to electrophoresis at 110 V for 60 min. The gel was then stained with Coomassie Brilliant Blue fast staining solution and imaged.

Cell culture and animal

Hepa1-6 and alpha mouse liver 12 (AML12) cells were obtained from the American Type Culture Collection (ATCC). The cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin. All cells were incubated at 37 °C in a humidified atmosphere with 5% CO₂.

Male BALB/c mice (4 weeks old) were purchased from SLAC Laboratory Animal Co., Ltd. (Shanghai, China). All animals were housed under specific pathogen-free conditions at 22 °C with a relative humidity of 30–70% and a 12 h light/dark cycle, with free access to food and water. Humane endpoints were defined as any of the following: (1) tumor burden exceeding 10% of the animal’s normal body weight, (2) body weight loss > 20%, (3) ulceration at the tumor site, or (4) persistent self-mutilation. All animal experiments were conducted in accordance with the ARRIVE guidelines and the U.K. Animals (Scientific Procedures) Act 1986 [31]. Experimental procedures were approved by the Ethics Committee for Animal Experimentation at our institution (Approval No. 2024 − 639).

Intracellular redox biomarker analysis

To analyze redox-related intracellular changes, Hepa1-6 cells (3 × 10⁶ cells) were seeded into 10 cm dishes and incubated for 24 h. The cells were then treated with PBS, MMOF, or A@MMOF (20 µg/mL) for 6 h, followed by X-ray irradiation at 0–2 Gy. After a further 24 h, the cells were washed with cold PBS, lysed by sonication (4 °C, 5 min), and centrifuged at 10,000 × g for 10 min. The supernatants were collected for analysis of redox biomarkers. TrxR activity was measured using a commercial thioredoxin reductase assay kit (Beijing Solarbio Life Science Co., Ltd.) and expressed as a percentage of the control group. Total thiol content was quantified using the 5,5′-dithiobis (2-nitrobenzoic acid) (DTNB) titration method, with absorbance measured at 412 nm and GSH used as the calibration standard. Reduced thioredoxin (Trx) levels were also evaluated using the DTNB method after ultracentrifugation (4,000 × g, 30 min) to remove insoluble fractions. Intracellular GSH levels were assessed using a commercial GSH assay kit (Solarbio) and normalized to the control. The NADPH/NADP⁺ ratio was determined using a NADP⁺/NADPH quantification kit (Beyotime Biotechnology, China) and expressed as a percentage of the control group [31]. Protein concentration was measured using the Bradford protein assay (Beyotime Biotechnology, China) for normalization. All assays were performed according to the respective manufacturers’ protocols.

NK cell migration assay

The ability of Hepa1-6 cells to promote NK cell migration under different treatment conditions was evaluated as previously described [10]. Briefly, 600 µL of serum-free conditioned medium (CM) collected from treated Hepa1-6 cells was added to the lower chambers of 24-well Transwell plates (pore size: 5 μm; Thermo Fisher Scientific, USA). NK cells were pre-labeled with the CellTrace CFSE probe (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions and resuspended at a density of 2.5 × 10⁵ cells in 100 µL of serum-free medium, then seeded into the upper chambers. After incubation at 37 °C for 4 h, the migrated NK cells in the lower chambers were harvested and quantified under a fluorescence microscope.

Polarization of bone Marrow–Derived macrophages (BMDMs)

Bone marrow cells were harvested from male BALB/c mice and differentiated into bone marrow–derived macrophages (BMDMs) as previously reported [12]. Mature BMDMs were cultured in complete medium supplemented with 50% conditioned medium (CM) from treated Hepa1-6 cells and incubated for 24 h. After co-incubation, the cells were washed three times with PBS and stained with the following antibodies: APC–anti-mouse CD11b (macrophage marker), FITC–anti-mouse F4/80 (macrophage marker), PerCP-Cy5.5–anti-mouse CD86 (M1 marker), and PE-Cy7–anti-mouse CD206 (M2 marker). Flow cytometry was performed to assess the proportions of M1- and M2-polarized macrophages.

X-ray irradiation and antitumor efficacy studies in subcutaneous HCC models

For all irradiation procedures, an RT250 orthovoltage X-ray machine (Philips, USA) equipped with a fixed setting of 250 kVp, 15 mA, and a built-in 1 mm copper filter was used. Localized X-ray irradiation (2 Gy per dose) was applied to tumor-bearing regions on days 1, 3, and 5, with the remainder of the mouse body shielded using custom-designed lead plates to minimize off-target radiation exposure [24].

To evaluate the synergistic antitumor efficacy of A@MMOF combined with LDRT, subcutaneous HCC models were established by injecting 1 × 10⁶ Hepa1-6 cells into the right flanks of BALB/c mice. Once tumors reached approximately 100 mm³, the mice were randomized into five groups (n = 10 per group): (G1) PBS (Control), (G2) PBS + LDRT (PBS(+)), (G3) LDRT + MMOF (MMOF(+)), (G4) LDRT + A@MMOF (A@MMOF(+)), and (G5) A@MMOF (A@MMOF(–)). In this context, “(+)” and “(–)” indicate whether the treatment was combined with LDRT or not, respectively. On day 0, the mice received the corresponding treatment (5 mg/kg, 200 µL) via tail vein injection, and localized LDRT was administered as described above. Tumor volumes and body weights were recorded every 5 days during the 25-day treatment period. Tumor volume was calculated using the formula V = (L × W²)/2, where L and W represent the longest and shortest tumor diameters, respectively. On day 25, five mice from each group were euthanized for tumor collection, weighing, histological analysis, and flow cytometric evaluation. The remaining animals were monitored for survival analysis.

To further investigate the involvement of NK cells and the potential benefit of ANKCT in the A@MMOF(+) regimen, an additional cohort of mice was established using the same tumor model and grouped as follows (n = 10 per group): (G1) PBS (Control), (G2) LDRT + A@MMOF (A@MMOF(+)), (G3) LDRT + A@MMOF plus anti-asialo GM1 antibodies for NK cell depletion (A@MMOF(+) + anti-ASGM1), (G4) ANKCT alone (NK), and (G5) ANKCT + LDRT + A@MMOF (NK + A@MMOF(+)). All mice received the designated formulation (5 mg/kg, 200 µL) via tail vein injection on day 0, and localized LDRT was administered as described above. For NK cell depletion, mice were intraperitoneally injected with 100 µL of anti-asialo GM1 antibodies or control rabbit sera (Wako Pure Chemical Industries, Japan), diluted 1:10 in PBS, starting 3 days before treatment initiation, according to the manufacturer’s instructions and previously reported studies [32, 33]. The injections were repeated twice per week for a total of six doses to ensure sustained NK cell depletion throughout the experimental period. For NK adoptive transfer, 1 × 10⁶ primary murine NK cells stimulated ex vivo with interleukin-15 (10 ng/mL; PeproTech) were intravenously infused on day 6, according to previously established protocols [10]. Tumor growth and body weight were monitored every 5 days. On day 25, five mice per group were euthanized for tumor analysis, and the remaining animals were followed for survival evaluation.

Statistical analysis

All data are expressed as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 9. Differences between two groups were evaluated using a two-tailed Student’s t-test. For comparisons among multiple groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was applied. Statistical significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. n.s.=not significant.

Results and discussion

Preparation and characterization of the A@MMOF nanoparticle

MOF-199 was synthesized following established protocols, and AUR was loaded to obtain A@MOF nanoparticles. Encapsulation of A@MOF with tumor cell membranes yielded A@MMOF. A schematic of the stepwise fabrication of MOF-199, A@MOF, and A@MMOF, along with the mechanistic roles of each component, is presented in Figure S1.

We first systematically examined the effect of varying the AUR-to-MOF-199 weight ratio on DLE% and EE%. A 1:1 MOF-199-to-AUR ratio yielded a DLE% of 32.91% and a moderate EE% of 49.05% (Table S1). Increasing the AUR ratio to 1:2 led to only a slight increase in DLE% (35.10%), whereas EE% decreased dramatically (27.04%). These findings suggest that further increases in AUR content do not significantly improve drug loading, possibly due to saturation of the drug delivery framework [34]. Therefore, the 1:1 ratio was selected as it provides an optimal balance between DLE and EE, consistent with previous MOF-based drug-loading studies [28, 30]. The polydispersity index (PDI) is a critical parameter for assessing nanoparticle size distribution and uniformity, which are essential for effective targeting and circulation stability [34, 35]. The PDI values of MOF-199 and A@MOF were 0.156 and 0.149, respectively. We then investigated the effect of varying the A@MOF-to-tumor cell membrane weight ratio (1:0.5, 1:1, and 1:2, w/w) on the PDI of A@MMOF. All formulations exhibited low PDI values: 0.147 (1:0.5, w/w), 0.131 (1:1, w/w), and 0.137 (1:2, w/w), indicating uniform particle size and excellent monodispersity (Fig. S2A). Notably, incomplete membrane coverage on the nanoparticles (1:0.5) likely contributed to their increased PDI values [35]. Consequently, a 1:1 weight ratio of A@MOF to tumor membrane was selected as the final formulation for A@MMOF in subsequent experiments. The DLE of the final A@MMOF formulation was 24.76%, indicating a reasonable payload capacity.

Morphological analysis by TEM showed that MOF-199, A@MOF, and A@MMOF exhibited uniform polyhedral structures (Figs. 2A–C) [28]. Following AUR loading, the morphology of the MOF-199 framework remained unchanged (Fig. 2B). Encapsulation with tumor cell membranes yielded A@MMOF particles with a distinct core–shell structure and a lightly electron-dense outer membrane layer (Fig. 2C), confirming successful surface coating while maintaining structural uniformity and crystallinity. These morphological changes were accompanied by a progressive increase in hydrodynamic diameter, as measured via DLS, from 123.4 ± 4.58 nm for MOF-199 to 154.3 ± 4.45 nm for A@MOF and 175.7 ± 5.33 nm for A@MMOF (Fig. 2D), indicating effective drug encapsulation and membrane cloaking [30]. Zeta potential measurements revealed a progressive decrease in surface charge from MOF-199 to A@MMOF (Fig. 2E), with the more negative potential of A@MMOF primarily attributable to the anionic properties of the tumor cell membrane coating [30]. Additionally, A@MMOF displayed excellent colloidal stability, with no significant changes in particle size after incubation in PBS, mouse plasma, or 10% fetal bovine serum for up to 7 days (Fig. S2B).

Fig. 2.

Preparation and characterization of A@MMOF. (A–C) TEM images of MOF-199 (A), A@MOF (B), and A@MMOF (C); scale bar = 100 nm. (D, E) Dynamic light scattering (DLS) analysis of hydrodynamic diameter (D) and zeta potential (E) of MOF-199, A@MOF, and A@MMOF (n = 3). (F) XRD patterns of simulated MOF-199, MOF-199, and A@MMOF.(G, H) High-resolution XPS spectra of Au 4f (G) and Cu 2p (H) for A@MMOF.(I) SDS–PAGE analysis of membrane protein profiles of tumor cell membrane, A@MMOF, and MOF-199. (J) EDS elemental mapping of A@MMOF; scale bar = 100 nm.Results are presented as mean ± SD

XRD analysis confirmed that the characteristic diffraction peaks of MOF-199 were preserved in A@MMOF (Fig. 2F), indicating that neither AUR loading nor membrane coating disrupted the framework’s intrinsic crystalline structure. XPS confirmed the successful incorporation of Au(I) and Cu(II) species in A@MMOF, supporting incorporation of AUR into the copper-based MOF-199 structure (Figs. 2G–H). SDS-PAGE analysis verified the successful membrane coating of A@MMOF (Fig. 2I), on the nanoparticle surface, further supporting the biomimetic identity of A@MMOF. Elemental mapping by energy-dispersive spectroscopy revealed a uniform distribution of key elements throughout the nanoparticles (Fig. 2J). The presence of Cu and Au validated the incorporation of the MOF-199 framework and AUR, respectively [27, 28]. In addition, the uniform distribution of nitrogen (N), sulfur (S), and phosphorus (P) was attributable to membrane-derived proteins and phospholipids, indicating effective tumor cell membrane cloaking. Collectively, these findings confirm the successful construction of A@MMOF with preserved structural integrity and efficient cancer cell membrane camouflage.

A@MMOF simultaneously targets GSH and Trx systems to collapse tumor redox defense

Cu-based MOFs, such as MOF-199, have emerged as promising platforms for redox modulation in tumor therapy owing to their GSH-responsive degradation and ability to catalyze Fenton-like reactions [28]. Compared with normal cells, tumor cells exhibit elevated intracellular GSH levels, rendering them particularly susceptible to redox-based therapeutic strategies [31]. GSH acts as a reducing agent that converts Cu²⁺ within MOF-199 to Cu⁺ while being oxidized to GSSG (Fig. 3A). The generated Cu⁺ then reacts with H₂O₂ through a Fenton-like reaction to produce highly cytotoxic •OH. This redox cascade accelerates GSH depletion and shifts the TME toward a prooxidant state [31].

Fig. 3.

GSH-responsive degradation, copper redox cycling, and redox defense collapse induced by A@MMOF. (A) Schematic illustration of the GSH-mediated copper redox process in MOF-199. (B) Digital photographs and TEM images of A@MMOF incubated in PBS or 10 mM GSH for 0 and 24 h (scale bar = 100 nm). (C) GSH consumption assay of A@MMOF at different doses. (D) Cu⁺ generation assay in different groups. (E) ·OH detection assay in different groups. (F–H) Confocal fluorescence microscopy images of HCC and normal liver cells incubated with FITC-labeled indicated formulations (F; scale bar = 25 μm), flow cytometry histograms (G), and quantitative analysis of mean fluorescence intensity (MFI) (H) for different groups (n = 3). (I–M) Quantitative analysis of thioredoxin reductase (TrxR) activity (I), total thiol levels (J), reduced thioredoxin (Trx) (K), GSH levels (L), and the NADPH/NADP⁺ ratio (M) in different groups (n = 3). (N) Schematic illustration of the mechanism by which A@MMOF induces collapse of the redox defense system. “(+)” and “(–)” indicate whether the treatment was combined with LDRT or not, respectively. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

To evaluate the GSH-responsive degradation of A@MMOF, nanoparticles were incubated in GSH-containing solutions to mimic the reductive TME, and GSH-induced disassembly of the MOF-199 framework was monitored. TEM imaging showed that after 24 h of GSH exposure, the polyhedral structure of A@MMOF was completely collapsed, and the solution gradually changed from blue to colorless (Fig. 3B). The disappearance of the characteristic blue color, which reflects Cu²⁺ coordination within the MOF framework, together with the loss of structural integrity, confirms complete GSH-induced disassembly of MOF-199. We selected the 24-h time point because it corresponds to the plateau phase of GSH-induced degradation, at which maximal structural degradation typically occurs, consistent with previous studies on GSH-responsive MOF-199 systems [28, 36]. Consistent with this rationale, GSH-responsive release assays revealed that both AUR and Cu exhibited rapid and concentration-dependent release profiles (Fig. S3). In the presence of 10 mM GSH, cumulative AUR release increased sharply during the initial phase and reached approximately 80% within 24 h, whereas lower GSH concentrations produced proportionally slower release profiles. A similar trend was observed for Cu, with approximately 50% of total Cu released within 24 h under 10 mM GSH, whereas negligible release occurred in PBS. Statistical analysis confirmed significant differences among release groups, except between PBS and 0.1 mM GSH, which did not differ significantly. These release kinetics align with previously reported GSH-responsive MOF-199 systems [28, 36].

Furthermore, we employed DTNB, a classical thiol-reactive probe, to evaluate GSH to generate 2-nitro-5-thiobenzoate (TNB), which exhibits an absorbance peak at 412 nm. Upon incubation with A@MMOF, a dose-dependent reduction in absorbance at 412 nm was observed, indicating efficient and concentration-dependent GSH depletion (Fig. 3C). Given that GSH facilitates the reduction of Cu²⁺ to Cu⁺, we next employed neocuproine, a chromogenic probe that selectively complexes with Cu⁺. Upon incubation with GSH, a pronounced increase in absorbance at 457 nm was observed, confirming Cu⁺ generation as a reduction product (Fig. 3D). To evaluate the GSH-triggered Fenton-like activity of A@MMOF, methylene blue (MB) was used as an indicator of •OH activity. The addition of GSH substantially reduced MB absorbance at 665 nm in the A@MMOF + H₂O₂ + GSH group compared with the A@MMOF + H₂O₂ + PBS group, indicating that GSH-mediated reduction of Cu²⁺ to Cu⁺ accelerates •OH generation via Fenton-like reactions (Fig. 3E) [28]. In contrast, treatment with H₂O₂ alone or H₂O₂ + GSH resulted in negligible MB degradation. Collectively, these results demonstrate that A@MMOF effectively depletes GSH and initiates copper redox cycling, ultimately catalyzing •OH production.

To further elucidate the intracellular behavior of the nanoparticles, we assessed the cellular uptake of A@MMOF using cross-group comparison analysis. Confocal imaging and flow cytometry were performed after 4 h of incubation. A@MMOF exhibited significantly higher internalization in tumor cells (Hepa1–6) compared with free AUR and A@MOF (Figs. 3F–H). This enhanced uptake is likely due to the tumor cell membrane coating, which facilitates homotypic recognition via native adhesion molecules and membrane proteins, thereby promoting receptor-mediated endocytosis and overcoming the reduced nonspecific adsorption typically associated with negatively charged nanoparticles [28, 30]. As a supplementary analysis, we conducted a kinetic uptake study to monitor A@MMOF internalization at 0, 1, 2, and 4 h (Fig. S4). Fluorescence intensity increased steadily over time and peaked at 4 h, as confirmed by confocal microscopy and flow cytometry (Figs. S4A–C). To evaluate the tumor specificity of the nanoparticles, we also compared A@MMOF uptake in normal hepatocytes (AML12). A@MMOF exhibited significantly lower internalization in AML12 cells than in tumor cells (Figs. 3F–H). This reduced uptake is likely attributable to the nonhomologous nature of the tumor cell membrane coating and the negatively charged membrane surface, which may mitigate nonspecific cellular adsorption [29, 30].

LDRT offers several advantages and represents a potential adjunctive therapy to enhance the efficacy of ANKCT [20, 22]. However, accumulating evidence suggests that LDRT alone induces only limited oxidative stress and cytotoxicity in tumor cells, which may be insufficient elicit robust immune-stimulatory effects [20, 37]. This limitation is primarily attributed to the strong antioxidant capacity of tumor cells, which is mainly regulated by the GSH and Trx systems [26]. Building on our previous findings that A@MMOF effectively depletes intracellular GSH, we next investigated its additional impact on cellular antioxidant defenses [26]. AUR, a clinically approved gold-containing compound, has been extensively studied for its potent inhibitory effect on TrxR [27]. Mechanistically, AUR disrupts redox-homeostasis by irreversibly binding to the selenocysteine active site of intracellular TrxR, thereby inhibiting Trx reduction and impairing ROS detoxification [38]. Notably, our cellular uptake studies demonstrated that A@MMOF achieved significantly higher internalization than both A@MOF or free AUR. Therefore, A@MMOF serves as a holistic nanoplatform strategy to enhance LDRT’s efficacy. The subsequent experimental treatment groups were designed to explore the combined effects based on the LDRT framework, with the incorporation of A@MMOF components. The A@MMOF(+) group, which received A@MMOF and LDRT (2 Gy), exhibited the strongest disruption of redox homeostasis compared with all other groups, including PBS (control), LDRT alone [PBS(+)], membrane-coated MOF-199 plus LDRT [MMOF(+)], and A@MMOF without LDRT [A@MMOF(–)] (Figs. 3I–M). In this context, “(+)” and “(–)” denote the presence or absence of LDRT, respectively. Specifically, TrxR activity was markedly reduced in A@MMOF(+) group (Fig. 3I), impairing the recycling of reduced Trx, a key electron donor involved in repairing oxidized proteins and eliminating peroxides (Fig. 3K) [26]. Intracellular total thiol levels were also markedly reduced (Fig. 3J), reflecting the depletion of redox-active sulfhydryl groups, including GSH and Trx [39]. Moreover, A@MMOF(+) induced pronounced GSH depletion (Fig. 3L), compromising hydrogen peroxide detoxification via the GPx–GSH axis. This was accompanied by a substantial decline in the NADPH/NADP⁺ ratio (Fig. 3M), further limiting the regeneration of GSH and Trx, which are critically dependent on NADPH availability [39]. Notably, the A@MMOF(-) group induced only moderate oxidative perturbation, whereas this effect was markedly amplified when combined with radiotherapy. This synergistic effect is likely attributable to LDRT-induced intracellular generation of H₂O₂ through water radiolysis, with the elevated H₂O₂ serving as a key substrate for Cu⁺-mediated Fenton-like reactions. This process accelerates •OH production and amplifies oxidative damage [20, 31]. Overall, the A@MMOF nanoparticle facilitates the codelivery of AUR and MOF-199 to concurrently inhibit TrxR activity and deplete GSH, thereby collapsing intracellular antioxidant defenses and shifting the redox equilibrium toward a prooxidant state (Fig. 3N).

In vitro synergistic amplification of LDRT-Induced cytotoxic and oxidative stress by A@MMOF

To evaluate whether A@MMOF could augment LDRT’s therapeutic efficacy, we investigated its antitumor effects in HCC cells. CCK-8 assays demonstrated that A@MMOF(–) induced a clear dose-dependent cytotoxicity. Among the irradiated groups, cell viability decreased progressively from PBS(+) to MMOF(+), reaching the lowest level in the A@MMOF(+) group and suggesting an improved cytotoxic effect when dual redox modulation was combined with LDRT. Notably, the cytotoxicity observed with A@MMOF(–) was notably weaker than that of A@MMOF(+), highlighting LDRT’s essential role in synergizing with redox disruption (Fig. 4A). Consistently, colony formation assays revealed that A@MMOF(+) treatment greatly reduced clonogenic survival (p < 0.05) (Fig. 4B; Fig. S5A). Flow cytometry analysis further confirmed a substantial increase in apoptotic cells in the combination group (Figs. 4C; Fig. S5B), consistent with CCK-8 results and highlighting the markedly enhanced proapoptotic and radiosensitizing efficacy of A@MMOF under LDRT. In contrast, PBS(+) induced only modest cytotoxicity, and A@MMOF(–) exhibited limited antitumor activity compared to the combination treatment (p < 0.05). These observations are consistent with the aforementioned redox analysis, which demonstrated that LDRT alone induces only limited oxidative stress due to compensatory intracellular antioxidant systems [24]. In contrast, LDRT plus A@MMOF effectively disrupts redox homeostasis, thereby exerting synergistic cytotoxic effects. To further elucidate the mechanisms underlying the enhanced cytotoxicity observed with A@MMOF(+), we next assessed intracellular oxidative stress. Ionizing radiation exerts antitumor effects primarily through ROS generation [20]. Intracellular ROS levels were quantified using DCFH-DA, a widely used fluorescent probe for monitoring oxidative stress [40]. Compared with untreated controls, the PBS(+) and MMOF(+) groups showed modest increases in ROS levels. A@MMOF(–) treatment also induced mild ROS accumulation, likely due to partial redox disruption mediated by AUR and Cu. Notably, the A@MMOF(+) group exhibited the highest level of ROS generation (p < 0.05), indicating a synergistic enhancement of oxidative stress (Fig. 4D; Fig. S5C). Because excessive ROS can induce DNA double-strand breaks, one of the most detrimental outcomes of DNA damage leading to apoptosis and downstream pathway activation [40, 41], we further examined DNA damage using γ-H2AX immunofluorescence staining, a well-established marker of DNA double-strand breaks [40]. Consistent with the ROS findings, the PBS(+), MMOF(+), and A@MMOF(–) groups showed only moderate increases in γ-H2AX foci. In contrast, A@MMOF(+) treatment led to a substantial accumulation of γ-H2AX-positive nuclei (p < 0.05) (Fig. 4D; Fig. S5D). These findings suggest that A@MMOF(+) promotes oxidative stress and DNA damage, thereby improving tumor radiosensitivity.

Fig. 4.

In vitro radiosensitization, cGAS–STING pathway activation, and modulation of macrophage and NK cell responses by A@MMOF combined with low-dose radiotherapy (LDRT). (A) Cell viability of Hepa1-6 cells treated with different formulations at varying concentrations (n = 3). (B) Representative images of colony formation assays in different groups (n = 3). (C) Flow cytometric analysis of apoptosis in different groups (n = 3). (D) Fluorescence microscopy images of DCFH-DA staining and γ-H2AX immunofluorescence in different groups (scale bar = 25 μm, n = 3). (E) Western blot images of phosphorylated STING (p-STING), total STING, phosphorylated IRF3 (p-IRF3), and total IRF3 in different groups (n = 3). (F) Schematic illustration of the experimental workflow for evaluating NK cell chemotaxis and macrophage polarization using tumor-conditioned medium (TCM). Quantitative ELISA analysis of relative extracellular cGAMP levels (G) (n = 3), CXCL10 concentrations (H) (n = 3), and CCL5 concentrations (I) (n = 3). (J) Flow cytometric analysis of BMDMs and (K) quantitative assessment of M1-like macrophage polarization (CD86⁺CD206⁻) induced by TCM from different treatment groups (n = 3). (L) Immunofluorescence images showing NK cell chemotaxis (scale bar = 50 μm) and (M) quantification of the NK cell migration index in Transwell assays using TCM from different treatment groups (n = 3). (N) Schematic illustration of the mechanism by which A@MMOF activates the cGAS–STING pathway and reprograms macrophage and NK cell responses. “(+)” and “(–)” indicate whether the treatment was combined with LDRT or not, respectively. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

In vitro A@MMOF combined with LDRT activates the cGAS–STING pathway and reprograms macrophage and NK cell responses

Recent studies have highlighted the pivotal role of ROS and DNA damage in activating the cGAS-STING pathway, a key regulator of innate immunity [13]. Elevated ROS generation induced by stressors such as RT and redox-active compounds leads to DNA damage, which is accompanied by the nuclear leakage of dsDNA. The dsDNA is detected by cGAS, triggering cGAMP synthesis and cGAS–STING pathway activation [13, 24]. This endogenous second messenger activates STING localized on the endoplasmic reticulum, promoting IRF3 phosphorylation and transcription of downstream inflammatory mediators [24]. cGAMP can also be transferred to adjacent immune or stromal cells, amplifying immune signaling throughout the TME [42]. To investigate whether A@MMOF combined with LDRT activates this pathway, we evaluated the expression of key proteins of the cGAS–STING cascade. Western blot analysis revealed markedly increased phosphorylation of STING and IRF3 in the A@MMOF(+) group, whereas PBS(+), MMOF(+), and A@MMOF(–) exhibited only weak-to-moderate activation (Fig. 4E). Quantitative densitometry further demonstrated that the A@MMOF(+) group displayed markedly higher p-STING/STING and p-IRF3/IRF3 ratios than other groups (p < 0.05; Figs. S5E–F). cGAS–STING pathway activation was further supported by ELISA. A@MMOF(+) treatment resulted in the highest extracellular accumulation of cGAMP and substantially increased secretion of the STING-dependent chemokines CXCL10 and CCL5 in culture supernatants (Figs. 4G–I). The expression trends of these immune mediators closely mirrored those of ROS elevation and γ-H2AX expression described earlier. Consistent with our findings, previous studies have demonstrated that nanoradiosensitizers promote ROS generation, leading to DNA damage and subsequent activation of the cGAS–STING pathway [18, 19, 24]. However, we acknowledge that the present evidence is primarily associative rather than strictly causal. Incorporation of ROS scavenger–based rescue experiments would provide more direct causal evidence, representing a limitation of the current study.

The clinical efficacy of ANKCT in HCC is limited by insufficient NK cell trafficking and the immunosuppressive activity of M2-polarized TAMs [5, 11]. The cGAS–STING pathway represents a pivotal innate immune regulator, and its activation enhances NK cell recruitment while promoting TAM reprogramming toward an M1 phenotype [13, 14]. Building on our observed activation of this pathway, we next examined how A@MMOF(+) influences these innate immune cell populations. Tumor-conditioned medium (TCM) collected from Hepa1–6 cells subjected to different treatments was used to treat primary BMDMs (Fig. 4F). Flow cytometry revealed that BMDMs exposed to A@MMOF(+) TCM underwent a pronounced shift toward an M1-like phenotype (CD86⁺CD206⁻), whereas PBS(+) TCM favored M2 polarization (CD86⁻CD206⁺) (Figs. 4J–K; Fig. S6). Only limited changes were observed in the MMOF(+) and A@MMOF(–) groups. Notably, the degree of M1 polarization positively correlated with extracellular cGAMP levels, consistent with previous studies demonstrating that tumor-derived cGAMP can be transferred to macrophages via SLC19A1 or VRAC to promote M1 conversion [13, 15, 16]. In parallel, transwell migration assays demonstrated that TCM from the A@MMOF(+) group induced the strongest NK-cell chemotaxis (Figs. 4L–M). This enhanced migration is likely mediated by elevated secretion of STING-dependent chemokines, including CXCL10 and CCL5, which recruit NK cells via the CXCR3 and CCR5 axes [14]. A@MMOF treatment alone did not elicit similar chemotactic responses, underscoring the critical role of LDRT in amplifying tumor-derived immunostimulatory signals. Together, these findings indicate that A@MMOF complements LDRT to activate cGAS–STING signaling. The resulting paracrine mediators, including extracellular cGAMP and NK cell chemokines, promote macrophage M1 polarization and facilitate NK cell recruitment (Fig. 4N).

In vivo biodistribution and biocompatibility of A@MMOF

Before evaluating the in vivo therapeutic efficacy of A@MMOF, we examined its biodistribution and biocompatibility to confirm suitability for systemic administration and effective tumor delivery. Hemolysis assays revealed that A@MMOF at 25–100 µg/mL induced hemolysis rates below the 5% safety threshold, comparable to the PBS negative control and markedly lower than the positive control (Fig. S7), demonstrating good hemocompatibility.

The tumor-targeting capability of A@MMOF was investigated using in vivo fluorescence imaging at sequential time points. A@MMOF rapidly accumulated at the tumor site, with fluorescence intensity peaking at 24 h (Figs. 5A–B). A@MMOF exhibited stronger and more sustained intratumoral retention than A@MOF. Ex vivo fluorescence imaging at 48 h further confirmed preferential tumor accumulation and relatively low distribution in off-target organs such as the liver and spleen (Fig. 5C), suggesting improved tumor selectivity mediated by homotypic tumor-cell membrane coating [28, 29]. These results are consistent with the cellular uptake findings in vitro. Biocompatibility was further assessed via histological and biochemical analyses. Hematoxylin and eosin (H&E) staining of major organs (heart, liver, spleen, lungs, and kidneys) revealed no detectable histopathological abnormalities following administration of 10 or 20 mg/kg A@MMOF (Fig. 5D). Additionally, routine blood parameters and serum biochemical markers remained within normal physiological ranges and showed no significant differences compared with untreated controls (Fig. 5E), indicating an acceptable biosafety profile. In this study, A@MMOF preferentially accumulated in tumors, and its GSH-responsive activation mechanism suggests that redox-disruptive effects are more likely to occur within the high-GSH TME than in healthy tissues with lower GSH levels [28, 29]. These characteristics may help minimize unintended redox disturbance outside the tumor. Nonetheless, this strategy confers preferential rather than absolute specificity, and further long-term studies are required to comprehensively evaluate potential off-target effects.

Fig. 5.

In vivo biodistribution and biocompatibility evaluation of A@MMOF. (A) Representative in vivo fluorescence images of tumor-bearing mice at 0–48 h after intravenous injection of A@MMOF or A@MOF, and ex vivo fluorescence images of major organs and tumors collected at 48 h (n = 3). (B, C) Quantitative analysis of fluorescence intensity at the tumor site over time (B) and in excised organs and tumors at 48 h (C) for A@MMOF and A@MOF (n = 3). (D) Representative H&E staining images of heart, liver, spleen, lung, and kidney from mice treated with control or A@MMOF (10 or 20 mg/kg) (scale bar = 100 μm). (E) Hematological and serum biochemical parameters in control and A@MMOF-treated mice (10 or 20 mg/kg) (n = 3). Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

In vivo synergistic antitumor effects of A@MMOF combined with LDRT

Building on the radiosensitizing potential and tumor-targeting capacity of A@MMOF, we next evaluated its therapeutic efficacy in vivo. Tumor-bearing mice were randomized into five groups: Control, PBS(+), MMOF(+), A@MMOF(+), and A@MMOF(–). As shown in the treatment schedule (Fig. 6A), all formulations were administered intravenously on Day 0, followed by LDRT. As previously stated, “(+)” and “(–)” indicate the presence or absence of concurrent LDRT, respectively. To maintain consistency with commonly used preclinical LDRT schemes in antitumor research [43] and nanomaterial-assisted radiosensitization [44] studies, we adopted a fractionated 2 Gy × 3 regimen delivered on Days 1, 3, and 5.

Fig. 6.

In vivo evaluation of the combined therapeutic effects of A@MMOF and LDRT in the HCC mouse model. (A) Schematic illustration of the experimental timeline. (B) Representative photographs of excised tumors from different treatment groups (n = 5). (C–F) Statistical analysis of tumor growth (C), tumor weights (D), body weights of tumor-bearing mice (E), and survival percentage during the 60-day treatment period (F) (n = 5). (G) Individual tumor growth curves for each treatment group (n = 5). (H) Representative H&E staining, Ki-67 immunohistochemistry, and TUNEL immunofluorescence staining of tumor tissues from each group (scale bar = 100 μm). “(+)” and “(–)” indicate whether the treatment was combined with LDRT or not, respectively. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

Gross examination on Day 25 revealed a marked reduction in tumor burden in the A@MMOF(+) group (Figs. 6B–D). In contrast, moderate suppression was observed in the MMOF(+) and A@MMOF(–) groups, whereas the PBS(+) and control groups exhibited progressive tumor growth (Figs. 6B–D). Notably, tumor weight in the PBS(+) group showed only a marginal reduction compared with the control group and failed to reach statistical significance (Fig. 6D), indicating that LDRT at this dose exerts limited antitumor efficacy as monotherapy. Longitudinal tumor growth curves reinforced this finding, demonstrating sustained tumor suppression in the A@MMOF(+) group, with minimal effects in all other arms (Fig. 6G). Survival outcomes were evaluated using Kaplan–Meier analysis, which revealed a significant survival benefit in the A@MMOF(+) relative to other groups (Fig. 6F). Meanwhile, body weights remained stable across all groups (Fig. 6E), supporting the systemic tolerability of A@MMOF and aligning with its previously demonstrated biocompatibility profile. Histopathology further validated the superior efficacy of the combination strategy. H&E staining revealed extensive necrosis and loss of tumor architecture in the A@MMOF(+) group, focal necrosis in the MMOF(+) and A@MMOF(–) groups, and intact morphology in the PBS(+) and control groups (Fig. 6H). Ki-67 immunostaining showed the lowest proliferation in the A@MMOF(+) group, consistent with strong antiproliferative activity. Moreover, terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assays revealed widespread apoptosis exclusively in the A@MMOF(+) group, further supporting synergistic cytotoxicity. Collectively, these results demonstrate that neither A@MMOF nor LDRT alone is sufficient to elicit a robust antitumor effect, whereas their combination substantially suppresses tumor proliferation, enhances apoptosis, and markedly prolongs survival.

To further elucidate the mechanisms underlying the enhanced in vivo efficacy of A@MMOF, we investigated whether these effects were associated with disruption of cellular redox homeostasis and consequent DNA damage. PBS(+) exerted negligible effects on TrxR activity, GSH levels, and other redox parameters, reflecting both the limited capacity induced by LDRT alone and the strong intrinsic antioxidant buffering capacity of tumor cells (Figs. 7A–E) [24, 26]. In contrast, MMOF(+) induced moderate disruption of redox balance, as evidenced by significant depletion of intracellular GSH and total thiols. This effect is attributable to GSH consumption during the reduction of Cu²⁺ to Cu⁺, which subsequently catalyzes Fenton-like reactions and promotes ROS accumulation [28]. When AUR was incorporated, inhibition of TrxR further suppressed the Trx system, thereby compounding redox disruption [27]. Consequently, the A@MMOF(+) group exhibited the most profound collapse of redox homeostasis, characterized by marked inhibition of TrxR activity, marked depletion of intracellular GSH, and pronounced loss of reduced Trx. In addition, total thiol content and the NADPH/NADP⁺ ratio were sharply reduced, indicating severe exhaustion of reducing equivalents essential to sustain antioxidant defenses. By comparison, A@MMOF(–) induced only moderate oxidative stress through TrxR inhibition and partial GSH depletion; however, in the absence of LDRT, it lacked the additional ROS amplification, resulting in less pronounced redox disruption [19, 31]. The redox imbalance was accompanied by significant DNA damage, as evidenced by markedly elevated γ-H2AX expression in tumor tissues [15, 39]. This progressive redox imbalance across groups was paralleled by the extent of DNA damage, as assessed by γ-H2AX expression (Fig. 7F). The A@MMOF(+) group displayed the highest γ-H2AX levels, consistent with pronounced redox disruption and radiosensitization. By contrast, PBS(+) treatment induced only minimal γ-H2AX expression, underscoring that LDRT monotherapy is insufficient to overcome the intrinsic antioxidant defenses and genomic repair mechanisms [24, 26]. Collectively, these data demonstrate that A@MMOF effectively potentiates LDRT-induced redox destabilization and DNA damage, mechanistically accounting for its enhanced antitumor efficacy in vivo.

Fig. 7.

In vivo evaluation of redox homeostasis disruption, cGAS–STING pathway activation, and immune responses induced by A@MMOF combined with LDRT in the HCC mouse model. (A–E) Quantitative analysis of TrxR activity (A), total thiol levels (B), Trx (C), GSH content (D), and the NADPH/NADP⁺ ratio (E) in different groups (n = 5). (F) Representative immunohistochemical staining for γ-H2AX in tumor sections from different groups. (G) Western blot analysis of p-STING, total STING, p-IRF3, and total IRF3 in tumor tissues (n = 5). (H–J) Quantitative ELISA analysis of cGAMP levels (H), CXCL10 (I), and CCL5 (J) in tumor lysates (n = 5). (K–N) Flow cytometric TAM phenotypes (CD86⁺ or CD206⁺, K), NK cell populations (CD45⁺NKP46⁺, L), and NK cell effector molecule expression, including IFN-γ (M) and GZMB (N), in different groups (n = 5). (O–R) Quantification of M1-like TAMs (CD86⁺CD206⁻) among total TAMs (O), intratumoral NK cell frequency (P), and NK cells positive for IFN-γ (Q) or GZMB (R) (n = 5). “(+)” and “(–)” indicate whether the treatment was combined with LDRT or not, respectively. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

In vivo A@MMOF-Mediated cGAS–STING activation promotes NK Cell–Mediated antitumor immunity

To investigate the impact of A@MMOF on immune activation, we assessed its effect on the cGAS–STING pathway and NK cell-mediated immunity within the TME. Western blot analysis revealed enhanced STING and IRF3 phosphorylation in the A@MMOF(+) group compared with all other treatment arms (Fig. 7G). Consistently, densitometric quantification confirmed robust pathway activation in the A@MMOF(+) group (p < 0.05) (Fig. S8). In parallel, ELISA showed a marked increase in extracellular cGAMP levels after A@MMOF(+) treatment (Fig. 7H). Similarly, the expression of STING-related chemokines, including CCL5 and CXCL10, was notably increased in the A@MMOF(+) group than in the PBS(+), MMOF(+), and A@MMOF(–) groups (Figs. 7I–J). These findings indicate that A@MMOF(+) achieves the strongest cGAS–STING activation among LDRT-based regimens, whereas A@MMOF monotherapy remains insufficient, underscoring the necessity of LDRT co-administration [24].

Given the central role of the cGAS–STING pathway in orchestrating innate immune responses, we further examined the phenotypic status of TAMs and NK cells within tumors [13]. Flow cytometric analysis revealed that A@MMOF(+) treatment significantly reprogrammed TAMs, with a notable increase in proinflammatory M1 TAMs (CD86⁺CD206⁻) and a concurrent decrease in immunosuppressive M2 TAMs (CD86⁻CD206⁺) (Figs. 7K, O, Fig. S9), consistent with cGAMP-mediated M1 polarization [42, 45]. In parallel, A@MMOF(+) treatment markedly enhanced intratumoral NK cell infiltration and activation. The proportion of NKP46⁺ cells increased significantly compared to all other groups, corresponding with the elevated intratumoral expression of the key NK-recruiting chemokines CCL5 and CXCL10 [5, 14]. Furthermore, NK cells isolated from the A@MMOF(+) group exhibited strong upregulation of effector molecules, including IFN-γ and GZMB (Figs. 7L–N, P–R; Fig. S10), indicating robust cytotoxic capacity. This enhanced NK functionality is likely potentiated by the proinflammatory reprogramming of TAMs, establishing a feed-forward loop of innate immune activation [11]. These findings demonstrate that A@MMOF synergizes with LDRT to enhance NK cell–mediated antitumor immunity by activating the cGAS–STING pathway, accompanied by M1 polarization and NK cell activation.

In vivo synergistic antitumor efficacy of A@MMOF-Enhanced LDRT combined with ANKCT

In advanced HCC, NK cells exhibit markedly reduced frequency and impaired cytotoxic function, leading to weakened innate antitumor immunity [5]. ANKCT represents a rational strategy to restore NK cell activity; however, its therapeutic efficacy is often limited by the profoundly immunosuppressive TME [8]. Previous studies have shown that although LDRT offers significant advantages in activating antitumor immunity, it does not effectively enhance ANKCT in HCC [18, 22]. Additionally, achieving significant NK-cell activation in combination with ANKCT often requires radiation dose escalation, which inevitably increases the risk of radiation-induced toxicity [22, 23]. Compared with the monotherapy groups (LDRT or A@MMOF), we found that A@MMOF complements LDRT by optimally disrupting redox homeostasis and activating the cGAS–STING pathway, thereby driving robust NK cell–based innate immune remodeling. We next evaluated whether the synergistic therapeutic strategy of A@MMOF(+) could further enhance ANKCT’s efficacy in vivo. HCC tumor–bearing mice were randomly assigned to five groups: (1) PBS (control); (2) A@MMOF(+) alone; (3) A@MMOF(+) combined with anti-asialo GM1 antibodies (A@MMOF(+) + anti-ASGM1); (4) ANKCT monotherapy (NK); and (5) ANKCT combined with A@MMOF(+) (NK + A@MMOF(+)). A@MMOF was administered intravenously on Day 0, followed by fractionated LDRT (2 Gy on Days 1, 3, and 5). For ANKCT, ex vivo–activated NK cells were injected via the tail vein on Day 6. In the NK cell–depletion group, the anti- ASGM1 antibodies were used to eliminate endogenous NK cells (Fig. 8A) [32, 33].

Fig. 8.

In vivo evaluation of the synergistic antitumor efficacy of A@MMOF-enhanced LDRT combined with ANKCT in the HCC mouse model. (A) Schematic illustration of the experimental timeline. (B) Representative photographs of excised tumors from different treatment groups (n = 5). (C–F) Statistical analysis of tumor growth (C), tumor weights (D), body weights of tumor-bearing mice (E), and survival percentage during the 60-day treatment period (F) in different groups (n = 5). (G) Representative H&E staining, Ki-67 immunohistochemistry, and TUNEL immunofluorescence staining of tumor tissues from each group (scale bar = 100 μm). (H–M) Flow cytometric analysis (H, J, L) and corresponding quantitative analysis (I, K, M) of NK cell populations (CD45⁺NKP46⁺, H–I) and NK cell effector molecule expression, including IFN-γ⁺ (J–K) and GZMB⁺ (L–M), in different groups (n = 5). “(+)” indicates that the treatment was combined with LDRT. Data are presented as mean ± SD. Statistical analysis was performed using one-way ANOVA. Statistical significance was defined as p < 0.05, p < 0.01, p < 0.001, and p < 0.0001. All significance markers indicate comparisons with the corresponding group as shown in the figure

Compared with A@MMOF(+) alone, the addition of ANKCT significantly suppressed tumor progression and prolonged survival (Figs. 8B–F). In contrast, NK cell depletion markedly abolished these therapeutic benefits, as evidenced by the loss of NKP46⁺ cells detected by flow cytometry (Fig. 8H), confirming the essential role of NK cells in mediating antitumor response. ANKCT monotherapy provided modest tumor control, suggesting that an unmodified TME is insufficient to support effective NK cell activity [5, 8]. In addition, body weight remained stable across all groups, suggesting acceptable systemic tolerance (Fig. 8D). Histopathological analysis further validated the superior efficacy of the combination strategy. The NK + A@MMOF(+) group exhibited widespread necrosis (H&E staining), reduced proliferation (Ki-67), and extensive apoptosis (TUNEL) (Fig. 8G). In contrast, other groups exhibited comparatively mild pathological alterations, consistent with their greater tumor burden (Figs. 8H–I). Flow cytometry confirmed a significant increase in intratumoral infiltration of NKP46⁺ cells, accompanied by elevated expression of functional cytotoxic markers, including IFN-γ (Figs. 8J–K) and GZMB (Figs. 8L–M). These immunological enhancements were notably diminished in the NK monotherapy group and nearly absent in the NK-depleted cohort (Figs. 8H–M). These results show that A@MMOF plus LDRT promotes a favorable immune remodeling for NK cell immunity, which is essential for enhancing the therapeutic efficacy of ANKCT in HCC.

Notably, because the efficacy of ANKCT relies on the immune status of the transferred NK cells [6, 10], when designing the experimental groups based on ANKCT, we inferred that LDRT alone could serve as a negative control, compared to the more effective A@MMOF(+) strategy for enhancing NK cell immunity. Based on this rationale, we included only the NK + A@MMOF(+) group but not the NK(+) group. Although previous studies have shown that LDRT alone exhibits only marginal efficacy in enhancing ANKCT [18, 22], we acknowledge that this gap in our experimental design may affect the rigor of the evidence chain. Therefore, further comprehensive investigations are essential to refine this novel approach and explore novel and more effective strategies for optimizing LDRT-based enhancement of ANKCT.

Conclusions

This study demonstrates that A@MMOF functions as an effective biomimetic nanoparticle designed to complement the limited immunological activation induced by LDRT. By simultaneously disrupting the GSH and Trx antioxidant systems and overcoming the intrinsic redox buffering capacity of HCC cells, A@MMOF provides an additional redox-driven stimulus that amplifies LDRT-induced activation of the cGAS–STING pathway. Acting as an essential auxiliary to LDRT, A@MMOF synergistically drives TAM reprogramming toward an M1 phenotype and enhances NK cell recruitment and cytotoxic activation. Together, these effects remodel the innate immune microenvironment and substantially improve the therapeutic efficacy of ANKCT in HCC. Despite these findings, further investigation is needed to support clinical translation. Key challenges include species-specific immune differences, long-term biosafety, and formulation scalability. Future studies should also investigate additional antitumor mechanisms of A@MMOF and its individual components, particularly their potential roles in modulating adaptive immunity and enhancing NK cell–based immunotherapy.

Author contributions

Conceptualization: B.L., W.K., X.Z., Y.C.,W.Z., Y.F.; Methodology: B.L., W.K., X.Z., Y.F., Y.C., W.Z. Investigation: B.L., W.K., X.Z., B.R., Y.C., Y.S., R.W.; Data Curation: B.L., W.K., X.Z., B.R.; Formal Analysis: B.L., W.K., X.Z., B.R.; Resources: W.K., Y.F., W.Z.; Writing – Original Draft: B.L., W.K., X.Z., B.R.; Writing – Review & Editing: B.L., W.K., X.Z., Y.C., Y.F., W.Z.; Visualization: B.L., W.K., X.Z., B.R.; Supervision: Y.F., W.Z.; Project Administration: Y.F., W.Z.; Funding Acquisition: W.K.,W.Z.

Funding

This work was supported by grants from the Natural Science Foundation of China (grant numbers: 82260555, 42407101), Major Science and Technology Projects of Gansu Province (grant numbers: 22ZD6FA021-4), Joint Research Fund General Projects of Gansu Province (grant numbers: 23JRRA1508), Shandong Provincial Natural Science Foundation(grant numbers: ZR2022QB250),Gansu Provincial Top-notch Talent Program(grant numbers: (2023)9), Cuiying Science and Technology Innovation Project of the Second Hospital of Lanzhou University(grant numbers: CY2024-CQ-01), Public Hospital High-quality Development of Scientific Research Project of Chinese Health Promotion Foundation(grant numbers: GL-C032) and the matching funds for National Natural Science Foundation projects of Lanzhou University Second Hospital (grant numbers: 2025-23-zdfy-014 and 2025-25-zdfy-001).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All animal experiments were conducted in accordance with the ARRIVE guidelines and the U.K. Animals (Scientific Procedures) Act, 1986. Experimental procedures were approved by the Ethics Committee for Animal Experimentation at our institution (Approval No. 2024 − 639).

Competing interests

The authors declare no competing interests.