Potassium-scavenging train-style nanorobots potentiate radiotherapy and reverse immunosuppression in glioblastoma

Abstract

Radiotherapy (RT) is essential for glioblastoma treatment but is compromised by two key challenges: hypoxic radio-resistance and the post-radiotherapy leakage of intracellular potassium ions (K⁺) that creates a profound immunosuppressive microenvironment, impairing T cell function and driving macrophage M2 polarization. While existing strategies largely focus on radiosensitizers to overcome hypoxia, few address the downstream consequence of K⁺-mediated immunosuppression. Herein, we engineer a train-style nanorobot (M@Ang-SiO₂–Au–Pd) that concurrently overcomes both obstacles. The nanorobot features an angiopep-2-modified SiO₂ head for targeted glioblastoma delivery and an asymmetric Pd tail with catalase-like and peroxidase-like nanozyme activities. At the tumor site, Pd catalyzes H₂O₂ conversion into O₂ to provide bubble propulsion for deep penetration, while Au–Pd amplify X-ray energy deposition and generate lethal hydroxyl radicals (·OH) to synergistically enhance radiotherapy-induced apoptosis. The nanorobot is coated with a K⁺-selective membrane (M). This design enables its SiO₂ head to specifically capture K⁺ leaked from dying tumor cells. By maintaining low intratumoral K⁺ levels, it prevents macrophages from adopting the immunosuppressive M2 phenotype and alleviates T cell exhaustion. Together, these effects effectively reverse tumor immunosuppression. In an orthotopic glioblastoma model, the nanorobot demonstrated precise tumor targeting, enhanced radio-sensitization, and potent K⁺ regulation. This combination triggered a robust antitumor immune response and significantly prolonged survival. This work establishes a novel paradigm employing nanorobots as ion-adsorbing carriers to modulate the post-radiotherapy ionic microenvironment for synergistic therapy.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04696-5.

Introduction

Glioblastoma multiforme (GBM) is the most common and lethal primary malignant tumor of the central nervous system, characterized by aggressive proliferation, diffuse infiltration, and a dismal prognosis. Despite maximal surgical resection followed by adjuvant chemoradiotherapy, the median survival of GBM patients remains below 15 months, with a five-year survival rate of less than 5% [1, 2]. The intrinsic resistance of GBM to conventional therapies is largely attributed to the blood–brain barrier (BBB), which severely restricts drug delivery, and a profoundly immunosuppressive tumor microenvironment (TME) that fosters immune evasion and disease recurrence [3–5].

RT remains a cornerstone of GBM treatment, providing localized tumor control through the induction of DNA damage via ionizing radiation [6, 7]. However, the therapeutic efficacy of RT is substantially compromised by two major challenges inherent to the GBM microenvironment. The first challenge is tumor hypoxia [8–10]. The radiosensitivity of tumor cells is critically dependent on the presence of molecular oxygen, which stabilizes radiation-induced DNA lesions by forming irreversible oxidative damage. In hypoxic regions, which are common in GBM, the generation of reactive oxygen species (ROS) is markedly reduced, leading to radio-resistance and treatment failure. The second, and less appreciated, challenge is the massive efflux of intracellular K⁺ induced by RT [11, 12]. Tumor cells maintain an exceptionally steep K⁺ gradient, with an intracellular concentration of approximately 140 mM compared to only 4–5 mM in the extracellular space. Upon RT-induced apoptosis and necrosis, the disruption of cell membranes leads to a sudden and substantial release of K⁺ into the TME. Recent studies have revealed that elevated extracellular K⁺ (>40 mM) acts as a potent immunosuppressive signal. It dampens T cell receptor (TCR)-driven signaling and effector functions by impairing the activity of the Kir2.1 channel, thereby promoting T cell exhaustion [13, 14]. Furthermore, high K⁺ levels polarize tumor-associated macrophages from a pro-inflammatory (M1) to an anti-inflammatory (M2) phenotype, exacerbating local immunosuppression and facilitating tumor immune escape [15, 16]. Consequently, while RT effectively kills tumor cells, it simultaneously fosters a K⁺-rich milieu that counteracts sustained antitumor immunity.

To overcome RT resistance, considerable efforts have been directed toward developing radiosensitizers, including high-Z-element nanoparticles (e.g., Au, Pd, Pt) that enhance local ROS generation, and catalase-like nanozymes that relieve hypoxia by catalyzing the decomposition of endogenous H₂O₂ to O₂ [17–19]. Radiotherapy alone often fails to sufficiently mobilize antitumor immunity. To compensate for this shortfall, immune checkpoint inhibitors (e.g., anti-PD-1/PD-L1) are frequently combined with radiation. However, this strategy has achieved limited success in glioblastoma (GBM). The primary barriers include poor drug penetration across the blood-brain barrier and the highly complex, multifaceted nature of the tumor microenvironment [20–22]. While recent advances have begun to explore K⁺-targeted immunomodulation—such as crown ether-mediated extracellular K⁺ capture to repolarize tumor-associated macrophages [23] and intracellular K⁺/Ca²⁺ homeostasis perturbation to trigger immunogenic cell death [24]—these strategies primarily address baseline or chemotherapy-induced ionic imbalances. Notably, few studies have focused on directly mitigating RT-evoked K⁺ efflux as a strategy to reverse immunosuppression [25]. Moreover, existing therapeutic platforms rarely integrate radio-sensitization, TME modulation, and BBB-crossing capability into a single system capable of deep tumor penetration.

Herein, we introduce a train-style nanorobot M@Ang-SiO₂-Au-Pd (M@Ang-APSi) engineered to resolve these clinical challenges through a dual-mechanism strategy (Fig. 1). This nanorobot is rationally designed to first actively traverse the BBB and penetrate into deep glioblastoma tissues via angiopep-2-mediated targeting and Pd-catalyzed bubble propulsion. Once localized within the tumor, the platform enhances radiotherapy from two distinct aspects: first, the Au-Pd heterojunction effectively prolongs the local action time of X-rays, while the Pd nanozyme component exhibits catalase-like activity to catalyze the conversion of endogenous H₂O₂ into O₂ within the hypoxic tumor microenvironment, providing the essential substrate to amplify radiosensitivity, and concurrently processes peroxidase-like activity to convert intracellular H₂O₂ into highly toxic·OH, synergistically inducing tumor cell apoptosis with radiation; second, the nanorobot is designed to specifically adsorb the large amounts of K⁺ that are released post-radiotherapy. By encapsulating the silica head with a K⁺-selective membrane (M), the nanorobot acts as an ion-adsorbing carrier to continuously maintain K⁺ concentrations at a globally low level throughout the entire tumor mass. This specific K⁺ adsorption prevents macrophage polarization to the M2 state and alleviates T cell exhaustion, thereby reversing the K⁺-mediated immunosuppressive microenvironment and activating robust antitumor immunity to synergize with radiotherapy for the clearance of residual tumor cells and the prevention of relapse. Our work represents a novel paradigm that employs nanorobots not just for enhanced drug delivery, but as active modulators of the ionic microenvironment to amplify the effects of conventional treatment.

Fig. 1.

(a) Schematic representation of the fabrication process for nanorobots. (b) Illustrative depiction of nanorobots traversing the blood–brain barrier, exerting antitumor effects, and mobilizing immune responses

Methods

Synthesis and basic physicochemical characterization of APSi

First, we synthesized Au–SiO₂ nanoparticles following our previously reported protocol. Briefly, 0.5 mL of HAuCl₄ solution (10 mg/mL) was mixed with 100 mL of water, heated to 150 °C, and stirred for 10 min to generate 17 nm gold nanoparticles. Subsequently, 4 mL of the 17 nm AuNP suspension and 0.4 mL of sodium citrate dihydrate (1% w/w) were thoroughly mixed and reacted for 15 min to produce 50 nm gold nanoparticles. Then, 20 mL of AuNPs was concentrated to 5 mL by centrifugation. Under stirring, 25 mL of isopropanol, 200 µL of 4-MPAA, and 720 µL of PAA were sequentially added to the concentrated solution. After 45 min, 9 mL of TEOS and 0.9 mL of NH₃·H₂O were added, and the reaction continued for 4 h. Subsequently, a PVP solution (1 mL, 4 wt%) was added to the Au–SiO₂ mixture, which was then heated to 100 °C for 1 h to obtain Au–SiO₂ nanoparticles. CTAB, H₂PdCl₄, and AA were sequentially added to the Au–SiO₂ heterojunction solution. After stirring at room temperature for 30 min, the Pt–Au–SiO₂ heterojunction was collected by centrifugation and named APSi. For comparison, control materials Au–SiO₂ nanoparticles and Au–Pt nanoparticles were synthesized using the above methods and designated AS and AP, respectively. To enhance the active targeting ability of the system, the as-synthesized APSi nanoparticles were modified with PEG-Angiopep-2 targeting peptide. Unloaded PEG-Angiopep-2 was removed by ultrafiltration, and the resulting Ang-APSi nanoparticles were collected, dispersed in deionized water, and stored at 4 °C.

Evaluation of ROS and oxygen generation capacity of APSi

To evaluate the·OH generation capacity of APSi, the as-synthesized APSi nanoparticles were diluted with deionized water to different concentrations (0.05, 0.1, 0.2, 0.5, and 1.0 mg/mL). TMB solution (Concentration: 1.6 mM) and H₂O₂ solution (final concentration 1 mM) were then added. After thorough mixing, the APSi samples at different concentrations were irradiated by radiotherapy (4 Gy). At different time points after irradiation (0, 5, 10, 15, 20, 25, and 30 min), the APSi samples were subjected to spectral analysis, and images in the range of 400–800 nm were collected to analyze the ·OH generation capacity. Each sample was measured in triplicate to exclude errors. In addition, electron spin resonance (ESR) trapping experiments were performed to identify the free radical species generated by APSi under different conditions. To evaluate the oxygen generation capacity of APSi, a portable dissolved oxygen meter was employed to assess the oxygen production, and the dissolved oxygen level was continuously monitored for 600 s.

In vitro verification of Ang-APSi deep tumor penetration capacity

First, the heterojunction nanorobots (ASi, APSi, and Ang-APSi) were labeled with ICG. After labeling, the materials were collected by centrifugation, and the concentration of ICG-labeled materials was quantified by UV spectroscopy. The materials were then incubated with GL261 cells and 3D tumor spheroids at a concentration of Au = 100 µg/mL for 12 h, respectively. For GL261 cell culture, 5 × 10⁴ cells per well were seeded in 24-well plates and cultured for 24 h. For GL261 3D tumor spheroids, 500 cells per well were seeded in 1% agarose-coated 96-well plates and incubated for 4 days, with the culture medium replaced every 48 h. After incubation of the fluorescently labeled materials with cells or spheroids, confocal fluorescence microscopy was used to perform tomographic scanning of the cells or spheroids, and the fluorescence signals were analyzed and images were acquired (Ex = 785 nm, Em = 830–860 nm).

In vivo verification of Ang-APSi deep tumor penetration capacity

To evaluate the actual tumor-targeting performance of the Ang-APSi Janus nanorobot against glioblastoma, a mouse model of brain glioma was first established. Briefly, 5 µL of GL261 cell suspension (cell concentration: 2 × 10⁸ cells/mL) was injected into the ventricle (ML = 1 mm, AP = − 2 mm, DV = − 3 mm) of each C57BL/6 mouse using a stereotaxic apparatus. 7 days after tumor cell injection, the growth of glioblastoma was evaluated using a small animal in vivo fluorescence imaging system. The mice were randomly divided into four groups: Free ICG, ASi, APSi, and Ang-APSi. On day 10 post-injection, the mice received intravenous administration according to their respective groups. At different time points after administration (0, 4, 8, 12, 24, and 36 h), the accumulation of the drug in the brain was observed using the small animal in vivo fluorescence imaging system. At 12 h post-administration, the mice were dissected, and major organs were collected to evaluate the in vivo distribution of the drug.

Evaluation of the killing effect of Ang-APSi combined with radiotherapy on GL261 cells

To evaluate the cytotoxicity of Ang-APSi against tumor cells and the tumor-killing effect after radiotherapy, GL261 cells were seeded in 96-well plates and cultured for 24 h. Subsequently, different concentrations of ASi, AP, and Ang-APSi were co-incubated with GL261 cells for 24 h with or without 6 Gy radiation. Cell viability was assessed by CCK-8 assay according to the manufacturer’s instructions. In addition, a Transwell model of the BBB was established using bEnd.3 cells and GL261 cells. When the resistance of the bEnd.3 cell layer reached 300 Ω, Ang-APSi and APSi at different concentrations were added to the bEnd.3 cell layer. After 12 h of co-incubation, the GL261 cells were irradiated with 6 Gy, followed by an additional 12 h of incubation, and cell viability was assessed by CCK-8 assay according to the manufacturer’s instructions.

Analysis of the action mechanism of Ang-APSi combined with radiotherapy on GL261 cells

To evaluate the specific effects of Ang-APSi combined with radiotherapy on GL261 cells, at 30 min after radiotherapy, the intracellular ROS production and DNA damage of GL261 cells were analyzed using a ROS probe (DCFH-DA) and γ-H2AX immunofluorescence staining. Flow cytometry and fluorescence microscopy were employed to analyze intracellular ROS and DNA damage and to acquire images. Subsequently, the collected images were subjected to fluorescence quantification using ImageJ.

Analysis of the synergistic killing effect of Ang-APSi combined with radiotherapy on GL261 tumor spheroids

First, 3D tumor spheroids of GL261 cells were cultured according to the previously described method. Subsequently, ASi, AP, and APSi were co-incubated with the tumor spheroids at a concentration of Au = 50 µg/mL for 24 h, followed by radiotherapy at 6 Gy. Twenty-four hours after treatment, the tumor spheroids from different treatment groups were collected and washed twice with physiological saline. AM/PI staining solution was added at a ratio of 1 µL per 1 mL of PBS, mixed thoroughly, and incubated at 37 °C in the dark for 15 min. Finally, fluorescence signals were analyzed and images were acquired under a fluorescence microscope. In addition, the tumor spheroids were digested with 0.5% collagenase and DNase at 37 °C for 30 min to prepare a single-cell suspension. The cells were then seeded at 5 × 10⁴ cells per well and stained with Annexin V-FITC and PI at room temperature in the dark for 30 min. After washing with physiological saline, the fluorescence signals were immediately collected by flow cytometry.

Evaluation of DC cell activation by Ang-APSi combined with radiotherapy

First, mononuclear cells were collected from mouse femur bone marrow by gradient density centrifugation, and immature DCs were induced using GM-CSF and IL-4. Subsequently, the supernatant of tumor spheroids from different treatment groups in the previous step was collected (centrifuged at 6000 rpm, 4℃, for 20 min) and co-incubated with the immature DCs at 37℃ for 24 h. The DCs were then collected (centrifuged at 1500 rpm, 4℃, for 5 min), washed three times with physiological saline, and stained with APC-A700-CD45, FITC-CD11c, APC-CD80, and PE-CD86 antibodies for 1 h. After incubation, the cells were washed three times with physiological saline, and DC maturation was detected by flow cytometry.

In vivo evaluation of antitumor efficacy and safety of Ang-APSi combined with radiotherapy

Following the glioma model construction method described in the previous steps, C57BL/6 mice received intracranial cell injection. On day 5 post-injection, the tumors were observed using IVIS system, and this time point was designated as day 0. The mice were randomly divided into five groups: PBS group, RT group, APSi group, APSi + RT group, and Ang-APSi + RT group. APSi was administered via intravenous injection at a dose of 2 mg/kg. Twelve hours after injection, the mice received 6 Gy radiotherapy. On days 3, 9, and 14 thereafter, glioblastoma growth was monitored using the small animal in vivo fluorescence imaging system, and mouse body weight was recorded every two days. On day 15, the mice in each group were euthanized, and the brain, heart, liver, spleen, lungs, and kidneys were collected as major organs for H&E staining to evaluate biosafety.

Synthesis of M@Ang-APSi

1 mL acetonitrile solution containing 1 mg of the membrane precursor (M) was added to a 5 mL acetonitrile suspension of Ang-APSi (5 mg) under ultrasonication with vigorous stirring. The mixture was allowed to react at 50 °C for 10 min. After annealing at room temperature for 12 h, the resulting M@Ang-APSi was washed once with methanol, twice with water (centrifuged at 6000 rpm for 10 min), and finally dispersed in 10 mL of deionized water and stored at 4 °C for subsequent experiments.

Evaluation of K⁺ adsorption capacity and specificity of M@Ang-APSi

To evaluate the specific K⁺ adsorption performance, the material was diluted to 1000 µg/mL. A mixed ion solution (10 mM K⁺, 10 mM Mg²⁺, 10 mM Ca²⁺, and 10 mM Na²⁺), the material, and water were mixed at a ratio of 1:2:7. After co-incubation for 30 min, the mixture was centrifuged at 8000 rpm, and the supernatant was collected for quantitative analysis of cation concentrations by ICP-MS.

Effect of M@Ang-APSi on macrophage function via K⁺adsorption

First, 1 × 10⁷ GL261 cells were irradiated at 20 Gy and incubated at 37 °C for 24 h. The culture medium was then collected by centrifugation (1500 rpm, 4 °C, for 5 min). Subsequently, the collected medium was incubated with M@Ang-APSi at 100 µg/mL at 37 °C for an additional 2 h. After incubation, the medium treated with different materials was collected by centrifugation (1500 rpm, 4 °C, for 5 min). This medium was then co-incubated with macrophages for 48 h. The macrophages were subsequently collected (centrifuged at 1500 rpm, 4 °C, for 5 min), washed three times with physiological saline, and stained with FITC-CD11b, PE-CD86, and APC-CD206 antibodies for 1 h. After incubation, the cells were washed three times with physiological saline, and macrophage polarization was detected by flow cytometry.

Effect of M@Ang-APSi on T cell function via K⁺adsorption

First, 1 × 10⁷ GL261 cells were irradiated at 20 Gy and incubated at 37 °C for 24 h. The culture medium was then collected by centrifugation (1500 rpm, 4 °C, for 5 min). Subsequently, the collected medium was incubated with M@Ang-APSi at 100 µg/mL at 37 °C for an additional 2 h. After incubation, the medium treated with different materials was collected by centrifugation (1500 rpm, 4 °C, for 5 min). This medium was then co-incubated with freshly isolated T cells for 72 h. The T cells were subsequently collected (centrifuged at 1500 rpm, 4 °C, for 5 min), washed three times with physiological saline, and stained with FITC-CD3, APC-CD4, PC5.5-CD8a, PE-IFN-γ, and PB450-TNF-α antibodies for 1 h. After incubation, the cells were washed three times with physiological saline, and the changes in intracellular cytokine levels of T cells were detected by flow cytometry.

In vivo improvement of the immune microenvironment by M@Ang-APSi combined with radiotherapy

Following the glioma model construction method described in the previous steps, C57BL/6 mice received intracranial cell injection. On day 5 post-injection, the tumors were observed using IVIS system, and this time point was designated as day 0. The mice were randomly divided into three groups: PBS group, Ang-APSi + RT group, and M@Ang-APSi + RT group. APSi was administered via intravenous injection of 2 mg/kg. Twelve hours after injection, the mice received 6 Gy radiotherapy. On days 7, 14, and 21 thereafter, glioma growth was monitored using the IVIS system, and mouse survival was recorded. On day 14 of treatment, the brain glioma tissues were dissected from the brains of mice in each group. Subsequently, the superficial cervical lymph nodes, spleen, and glioblastoma tissues were separately prepared into single-cell suspensions. The expression of DCs, macrophage, and T cell maturation within the glioblastoma tissue, the maturation of T cells and cytokine expression levels in the spleen, and the proportion of mature DCs in the lymph nodes were detected.

In vivo RNA-seq transcriptomic analysis of M@Ang-APSi

Following the glioma treatment described in the previous step (Ang-APSi + RT group and M@Ang-APSi + RT group), on day 14 after treatment, the tumors were collected and snap-frozen in liquid nitrogen. RNA transcriptomic analysis was then performed on the Ang-APSi + RT group and the M@Ang-APSi + RT group to determine the differentially expressed genes between the two groups after treatment.

Statistical analysis

The two-tailed Student’s t-test was employed for comparisons between two groups, whereas one-way analysis of variance (ANOVA) was used for comparisons among multiple groups. A p value less than 0.05 was deemed to signify a significant difference. The significance levels were denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Quantitative results are reported as mean ± standard deviation (SD). The bar charts and curve plots were generated using GraphPad Prism version 9 software.

Results and Discussion

Synthesis, dual nanozyme activities, and catalytic mechanism of the train-style nanorobot

The train-style APSi nanorobot was designed with a spatially asymmetric architecture that structurally segregates distinct functional domains, thereby enabling the independent execution of catalytic, targeting, and ion-adsorption functions without mutual interference [26–28]. The synthesis proceeded through a stepwise bottom-up strategy that sequentially integrated three functional modules: a ~ 50 nm Au nanoparticle served as the central core, with a ~ 40 nm SiO₂ shell coated on one hemisphere and a ~ 30 nm Pd nanoparticle grown asymmetrically on the opposite side (Fig. 2a). Elemental mapping validated the heterogeneous spatial distribution of Pd, Au, Si, and O within the nanorobot, confirming the intended asymmetry that structurally underpins its propulsion capability (Fig. 2b). Hydrodynamic diameter measurements revealed that the particle sizes of AP, ASi, and APSi were approximately 55.3 nm, 66.8 nm, and 89.1 nm, respectively, and all formulations exhibited comparable surface zeta potentials of approximately − 10 mV (Fig. 2c, d). Subsequently, we evaluated the stability of APSi in both ddH₂O and DMEM culture medium. Over a 7-day period post-synthesis, the hydrodynamic diameter of APSi exhibited negligible variation (Fig.S1).

Fig. 2.

(a) TEM images of AP, ASi, and APSi nanorobots. (b) The element mapping of APSi. Hydrodynamic diameters (c) and zeta poteintial (d) of various nanorobots. (e) Absorbance of TMB at 652 nm following various treatments. (f) and (g) Absorbance of TMB at 652 nm following various concentration with or without RT. (h) ESR spectrum of ·OH after different treatments. (i) O₂ concentration of PBS and APSi without/with RT. (j) Fluorescence images of GL-261 cells after co-incubation of APSi and Ang-APSi with or without free Ang peptide. Fluorescence images (k) and fluorescence intensity (l) of 3D GL-261 cell spheroids after co-incubation of ASi, APSi and Ang-APSi. (m) Fluorescence images and fluorescence intensity (n) of tumor sites in orthotopic glioma-bearing mice following intravenous administration. Fluorescence images (o) and quantitative fluorescence intensity analysis (p and q) of the heart, liver, spleen, lungs, kidneys and brain. Data are presented as means ± standard deviation (n = 3)

The APSi nanorobot harbors dual enzyme-mimetic activities that operate in a spatially programmed manner within the tumor microenvironment. The Pd component functions as a catalase-like nanozyme, catalyzing the decomposition of endogenous H₂O₂ into O₂. This reaction serves two critical purposes aligned with the therapeutic rationale of our design: first, it increases the local oxygen concentration to serve as a radio-sensitization substrate, directly addressing the hypoxia-induced radiotherapy resistance that plagues glioblastoma treatment; second, the gaseous O₂ bubbles generated provide a physical propulsive force that drives the nanorobot through the dense tumor stroma into deep, otherwise inaccessible regions. Simultaneously, APSi exhibits POD-like activity, converting H₂O₂ into highly cytotoxic ·OH that synergize with the DNA-damaging effects of ionizing radiation to amplify tumor cell killing. This dual-enzyme cascade is strategically orchestrated: the CAT-like activity consumes H₂O₂ in the extracellular tumor microenvironment to generate O₂ for radio-sensitization, while the POD-like activity exploits intracellular H₂O₂ to produce ·OH for direct oxidative damage.

We systematically evaluated these catalytic capabilities [29–31]. The absorbance of oxidized TMB at 652 nm progressively increased over time in the APSi + TMB + H₂O₂ group, confirming that the ·OH generation activity of APSi was both stable and sustained (Fig. 2e and S2). Importantly, when radiotherapy was applied, the ·OH production capacity of APSi was markedly enhanced across all tested concentrations (Fig. 2f, g and S3). This radiation responsive augmentation is mechanistically attributed to the ability of high Z elements (Au and Pd) to deposit more X ray energy within the tumor cells through enhanced photoelectric and Compton scattering effects, thereby boosting the catalytic efficiency of the nanozyme and generating higher yields of cytotoxic radicals. Specifically, The Au component serves not only as a high-Z radiosensitizer but also as an electronic bridge that facilitates interfacial electron transfer to Pd active sites, optimizing the adsorption energy of reaction intermediates and lowering the activation barrier for H₂O₂ decomposition. Furthermore, X-ray-induced water radiolysis products may synergize with Pd-catalyzed radical generation, creating a self-amplifying oxidative cascade [23, 24, 32]. ESR spectroscopy further identified the radical species generated, providing direct evidence that APSi produced elevated levels of ·OH specifically under irradiation conditions (Fig. 2h). The oxygen-generating capacity, which is central to the strategy of overcoming hypoxia-mediated radio-resistance, was also confirmed. In the presence of H₂O₂, APSi efficiently catalyzed its conversion into O₂, as continuously monitored by a dissolved oxygen meter, with sustained oxygen production observed over the entire 600 s measurement period (Fig. 2i). This on-demand, tumor-microenvironment responsive O₂ generation ensures that when radiotherapy is administered, tumor cells that were previously protected by hypoxia become sensitized to radiation-induced DNA damage, directly contributing to the enhanced radiosensitivity observed in subsequent experiments.

BBB penetration and targeted tumor accumulation through synergistic navigation

Effective glioblastoma therapy requires overcoming two sequential delivery barriers: the BBB, which prevents most therapeutic agents from entering the brain parenchyma, and the dense tumor stroma, which limits intratumoral penetration. Our nanorobot was engineered with a dual-mechanism navigation system that addresses both barriers. Ang, a peptide that specifically binds to low-density lipoprotein receptor-related protein 1 (LRP1) overexpressed on both brain endothelial cells and glioblastoma cells, was covalently conjugated to the SiO₂ head of APSi, yielding Ang-APSi. This modification confers BBB-transcytosis capability and tumor-cell-specific recognition. Simultaneously, the Pd-catalyzed bubble propulsion provides active motility that drives the nanorobot through the tumor interstitial space, a form of autonomous navigation that is particularly advantageous in the elevated H₂O₂ microenvironment characteristic of glioblastoma.

We first evaluated the receptor-mediated cellular internalization enabled by Ang modification. In 2D GL261 cell cultures, Ang‑APSi exhibited significantly higher cellular internalization compared to unmodified APSi (Fig. 2j). To explicitly validate the LRP1‑mediated targeting mechanism, we performed a competitive inhibition assay by pre‑incubating cells with excess free Ang peptide. This treatment markedly attenuated the uptake of Ang‑APSi, whereas APSi internalization remained unaffected (Fig. 2j and S4). These findings definitively confirm that the enhanced cellular accumulation of Ang‑APSi is specifically driven by LRP1‑dependent endocytosis, which can be competitively blocked by the free Ang ligand. To assess the synergistic effect of ligand-directed targeting and bubble-driven propulsion in a physiologically relevant context, we employed a 3D tumor spheroid model that mimics the penetration barriers present in solid tumors. The results revealed a stepwise enhancement in penetration depth. The ASi nanorobot, lacking both the Ang ligand and the Pd propulsion unit, exhibited limited penetration confined primarily to the spheroid periphery. APSi, which possesses the Pd propulsion unit but lacks Ang modification, achieved deeper penetration than ASi owing to O₂ bubble-driven propulsion generated by the catalase-like activity of Pd in the H₂O₂-containing tumor microenvironment. Most strikingly, Ang-APSi, which integrates both Ang-mediated LRP1 targeting and Pd-driven propulsion, penetrated most deeply into the tumor spheroids, with bright fluorescence signals readily detected at a depth of 60 μm (Fig. 2k, l). This result mechanistically demonstrates that Ang targeting and bubble propulsion act synergistically: Ang binding to LRP1 initiates receptor-mediated endocytosis at the tumor cell surface, while the continuous O₂ bubble generation propels the nanorobot further through the extracellular matrix, effectively overcoming the passive diffusion limitations.

We further validated the in vivo tumor-targeting performance in an orthotopic GL261 glioblastoma mouse model. 2 days prior to the commencement of the experiment, in vivo tumor imaging was performed on the mice to eliminate the potential confounding effect of tumor size on drug accumulation. Tumors across all experimental groups were localized to the same lateral region, with fluorescence intensities consistently maintained at approximately 5 × 10⁶ (Fig. S5). Fluorescence imaging revealed a distinct gradient of tumor accumulation across the different formulations. While ASi showed modest tumor retention and APSi exhibited improved accumulation attributable to its self-propulsion, Ang-APSi achieved the highest fluorescence intensity at the tumor site across all time points examined (Fig. 2m, n). Quantitative analysis at 12 h post-administration demonstrated that the fluorescence intensity of APSi was 3.4-fold that of Free ICG, whereas that of Ang-APSi reached 4.8-fold that of Free ICG (Fig. 2o, p, q). Ex vivo imaging of major organs further corroborated the predominant brain tumor accumulation of Ang-APSi, with substantially higher fluorescence in brain tissue compared with liver and spleen, indicating favorable biodistribution for glioblastoma therapy.

Ang-APSi enables deep tumor penetration and catalytic radio-sensitization to amplify radiotherapy-induced oxidative stress and DNA damage

We next investigated whether the targeted tumor penetration capability of Ang‑APSi translates into potentiated radiotherapy‑mediated tumor cell killing, and dissected the underlying mechanisms linking catalytic activity, radio-sensitization, and synergistic cytotoxicity. The experimental strategy was designed to decouple the contributions of individual nanorobot components and to evaluate their performance under conditions that recapitulate the biological barriers encountered in glioblastoma therapy.

We first assessed the cytotoxicity of the nanorobots and their capacity to enhance radiation‑induced cell death. In the absence of radiotherapy, co‑incubation with ASi, AP, or APSi at concentrations up to 200 µg/mL for 24 h resulted in cell viabilities remaining above 75%, indicating that all formulations exhibited minimal inherent toxicity under non‑irradiated conditions (Fig. 3a). This low baseline toxicity is a prerequisite for a tumor‑selective therapeutic window. When combined with 6 Gy X‑ray irradiation, the cell‑killing profiles diverged markedly according to nanorobot composition. ASi, which contains the Au core and SiO₂ shell but lacks the Pd nanozyme, modestly reduced cell viability relative to radiation alone, attributable primarily to the radio-sensitizing effect of the Au core through enhanced photoelectric absorption. AP, comprising Au and Pd without the SiO₂ head, exhibited more pronounced cytotoxicity, reflecting the combined radio-sensitization from both high‑Z elements and the additional ·OH generated by the Pd peroxidase‑like activity. APSi, which integrates all three functional modules, demonstrated the most potent radio-sensitization, reducing cell viability to below 20% at both 100 µg/mL and 200 µg/mL under 6 Gy irradiation (Fig. 3b). The superior performance of APSi over AP suggests that while Pd‑mediated ·OH production is a major contributor, the silica head may also play an ancillary role by increasing the overall hydrodynamic size and colloidal stability, thereby enhancing cellular association. These results establish that the Au–Pd heterojunction core of APSi functions as an effective radiosensitizer, and that the Pd nanozyme provides an additional catalytic layer of ROS generation that cooperates with radiation to induce cell death.

Fig. 3.

(a) Viability of GL-261 cell after treated with ASi, AP and APSi. (b) Viability of GL-261 cell after treated with different nanorobots and 6 Gy X-ray irradiation. (c) Schematic diagram of trans-well system. (d) Viability of GL-261 cell after different treatments. (e) The internal ROS of cells after different treatments. (f) The proportion of positive cells after different treatments. Representative confocal fluorescence images of γ-H2AX (g) and quantitative analysis (h) of γ-H2AX in 3D GL-261 cell spheroids. (i and j) Fluorescence images of post-treatment GL-261 3D cell spheroids following AM/PI staining, along with corresponding quantitative analysis. Flow cytometric plot (k) and quantification analysis (l) of GL-261 cells stained with Annexin V-fluorescein isothiocyanate/propidium iodide (V-FITC/PI) after different treatments. (m) Schematic illustration of the experimental procedure. Representative confocal fluorescence images (n) and quantitative analysis (o) of HMGB1 in 3D GL-261 cell spheroids. Flow cytometric plot (p) and quantification analysis (q) of BMDCs after different treatments. Data are presented as means ± standard deviation (n = 3)

In glioblastoma, therapeutic agents must first traverse the BBB before gaining access to tumor cells. To evaluate whether Ang‑mediated targeting improves the therapeutic outcome in a physiologically relevant context, we employed a Transwell BBB‑tumor co‑culture model in which a confluent bEnd.3 endothelial monolayer was established on the insert with a transendothelial electrical resistance exceeding 300 Ω·cm², and GL261 cells were cultured in the lower compartment (Fig. 3c). Nanorobots were added to the apical chamber, and after crossing the endothelial barrier, they encountered the tumor cells in the basolateral compartment. Following 12 h of incubation, the lower chamber GL261 cells were irradiated with 6 Gy, and viability was assessed after an additional 12 h. Unmodified APSi, lacking the Ang ligand, produced only a modest reduction in tumor cell viability at 50 µg/mL and 100 µg/mL, consistent with limited trans-endothelial passage. In striking contrast, Ang‑APSi, equipped with the Angiopep‑2 targeting ligand that binds LRP1 on brain endothelial cells, significantly reduced tumor cell viability at both concentrations tested (Fig. 3d). This targeting‑dependent enhancement demonstrates that the Ang modification is indispensable for efficient BBB traversal and subsequent tumor cell engagement, and that the active transcytosis conferred by LRP1 binding is a critical determinant of therapeutic efficacy in this model. The ability of Ang‑APSi to cross the BBB and retain its radio-sensitizing activity validates the design rationale of integrating a targeting ligand with a catalytic radiosensitizer in a single nanorobotic platform.

Tumor spheroids recapitulate the penetration barriers, oxygen gradients, and cell–cell interactions of solid tumors more faithfully than monolayer cultures. We therefore employed GL261 3D tumor spheroids to interrogate the mechanistic sequence through which Ang‑APSi enhances radiotherapy. Spheroids were incubated with different nanorobots in Transwell inserts for 12 h to allow diffusion or active penetration, then irradiated with 6 Gy, and intracellular ROS levels were quantified immediately after radiotherapy. In spheroids treated with PBS or APSi, the ROS fluorescence signal was barely detectable. The radiotherapy‑alone group showed a moderate increase, reflecting the ·OH and other reactive species generated directly. The ASi + RT group exhibited a slightly higher ROS signal, consistent with Au‑mediated radio-sensitization. The APSi + RT group, benefiting from self‑propulsion that enables deeper intratumoral penetration than ASi, produced a further elevation in ROS. Critically, the Ang‑APSi + RT group displayed the most intense ROS fluorescence throughout the spheroid volume (Fig. 3e, f). This pronounced ROS burst is mechanistically attributed to three converging factors: (i) enhanced spheroid penetration and cellular internalization driven by the combination of Ang‑mediated LRP1 endocytosis and Pd‑catalyzed bubble propulsion, ensuring a high intracellular nanorobot concentration; (ii) Pd catalase‑like activity that converts endogenous H₂O₂ into O₂, partially reversing the hypoxic status of the spheroid core and thereby augmenting the oxygen‑dependent fixation of radiation‑induced DNA damage; and (iii) direct generation of ·OH from H₂O₂ via Pd peroxidase‑like activity, amplified by the increased X‑ray energy deposition from Au and Pd.

We next examined DNA double‑strand breaks, the most lethal form of radiation‑induced damage, by immunofluorescence staining for phosphorylated histone γ‑H2AX at 2 h post‑irradiation. The pattern of γ‑H2AX foci mirrored that of ROS generation. The PBS and APSi‑only groups showed minimal nuclear foci, while the RT group displayed a moderate number of foci. The Ang‑APSi + RT group exhibited the highest density of γ‑H2AX foci per nucleus, indicative of extensive and unrepaired DNA double‑strand breaks (Fig. 3g, h). The parallelism between ROS levels and DNA damage severity supports the mechanistic model in which ·OH radicals, whether generated by water radiolysis or by Pd nanozyme catalysis, directly attack the sugar‑phosphate backbone of DNA, producing strand breaks that are stabilized and rendered irreparable by the concomitant presence of molecular oxygen supplied through the catalase‑mimetic function.

Ang-APSi-mediated radio-sensitization promotes synergistic apoptosis and immunogenic cell death to activate dendritic cells

To determine the ultimate cell fate following this combinatorial treatment, we performed live/dead staining and flow cytometric apoptosis analysis on 3D tumor spheroids at 24 h post‑irradiation. Live/dead imaging revealed a progressive increase in the proportion of dead cells across the treatment groups. The PBS and APSi groups showed negligible cell death. Radiotherapy alone produced a discernible dead cell population. The ASi + RT and APSi + RT groups each yielded greater cell death than RT alone, consistent with their respective penetration and radio-sensitization capabilities. The Ang‑APSi + RT group, however, exhibited the most extensive cell death throughout the spheroid, with the dead cell signal penetrating into regions that were clearly beyond the reach of the other formulations (Fig. 3i, j).

Quantitative apoptosis analysis by flow cytometry corroborated these observations. In the PBS group, 8.6% of cells were in early apoptosis (Annexin V⁺/PI⁻) and 2.7% in late apoptosis (Annexin V⁺/PI⁺). RT reflecting the rapid progression of radiation‑induced cell death. In the Ang‑APSi + RT group, the early apoptotic fraction dramatically rose to 65.3%, while the late apoptotic fraction increased to 17.9% (Fig. 3k, l). This redistribution toward early apoptosis, coupled with the overall increase in total apoptotic cells, indicates that the combined treatment engages a distinct and amplified apoptotic program. Mechanistically, we attribute this to the synergistic action of radiation‑induced DNA damage and Pd‑generated ·OH.

Radiotherapy can induce immunogenic cell death (ICD), a form of apoptosis characterized by the release of damage‑associated molecular patterns (DAMPs) that activate innate immune cells. So we evaluated whether Ang‑APSi‑enhanced radiotherapy promotes ICD. High mobility group box 1 (HMGB1) was assessed by immunofluorescence in 3D tumor spheroids at 48 h post‑treatment (Fig. 3m). The Ang‑APSi + RT group exhibited the most pronounced loss of nuclear HMGB1 signal, with a fluorescence intensity of only 6.9, compared with 29.2 and 28.7 in the RT and APSi groups, respectively (Fig. 3n, o). This substantial HMGB1 release indicates that the synergistic cell death induced by Ang‑APSi + RT fulfills the criteria of ICD, generating a repertoire of DAMPs capable of engaging the innate immune system.

To functionally validate the immunological consequence of this ICD induction, we collected conditioned medium from the treated spheroids and incubated it with immature bone marrow‑derived dendritic cells (BMDCs). Flow cytometric analysis of DC maturation markers (CD80 and CD86) revealed a stepwise increase in the proportion of mature DCs. The RT group induced 33.5% maturation, consistent with the basal immunogenicity of radiation‑induced cell death. The APSi + RT group achieved 41.6% maturation, attributable to the deeper tumor cell killing and consequently greater DAMP release. The Ang‑APSi + RT group drove DC maturation to 46.7%, the highest among all groups (Fig. 3p, q). These findings demonstrate that Ang‑APSi‑mediated radio-sensitization not only directly eliminates tumor cells but also converts radiotherapy into an in situ vaccination event, creating an immune‑stimulatory context that, in later experiments, will be further potentiated by the removal of K⁺‑mediated immunosuppression.

In summary, these results establish a mechanistic cascade wherein the targeted penetration capability of Ang‑APSi enables deep intratumoral delivery of the Au–Pd radiosensitizer, the Pd nanozyme alleviates hypoxia through O₂ generation to sensitize tumor cells to radiation, and simultaneously produces ·OH to synergistically induce DNA damage and apoptosis. This enhanced tumor killing is inherently immunogenic, as reflected by HMGB1 release and subsequent DC maturation, providing a foundation for the subsequent modulation of the immunosuppressive K⁺ microenvironment.

Evaluation of the inhibitory effect of Ang-APSi on glioblastoma progression

An orthotopic GL261 glioblastoma model was established in C57BL/6 mice to evaluate the therapeutic efficacy of Ang‑APSi under conditions that recapitulate the anatomical barriers and microenvironmental constraints of the clinical disease. Tumor engraftment was confirmed by in vivo bioluminescence imaging on day 5 post‑implantation, after which mice were randomly divided into five groups—PBS, RT, APSi, APSi + RT, and Ang‑APSi + RT—and treatment was initiated according to the schedule shown in Fig. 4a. Nanorobot formulations were administered intravenously at a dose of 2 mg/kg on day 0 and day 2, with radiotherapy (6 Gy) delivered 12 h after each injection, a time point selected to coincide with the peak intratumoral accumulation of Ang‑APSi as determined by the preceding pharmacokinetic studies (Fig. 4a). On day 14, bioluminescence imaging revealed markedly divergent tumor burdens across the treatment groups (Fig. 4b, c). Mice receiving PBS or APSi alone exhibited aggressive glioblastoma growth, with a mean luminescence intensity of 2.5 × 10⁸, confirming that the nanorobot per se does not restrain tumor progression in the absence of radiation. Radiotherapy alone reduced the tumor signal to 6.8 × 10⁷, demonstrating the radiosensitivity of this model but leaving substantial residual disease. The APSi + RT group, benefiting from the radio-sensitizing capacity of the Au–Pd heterojunction and the intratumoral penetration afforded by Pd‑driven bubble propulsion, achieved a significantly lower tumor burden of 1.5 × 10⁷. Most notably, the Ang‑APSi + RT group exhibited the most profound tumor suppression, with a luminescence intensity of only 2.7 × 10⁶, representing an approximately 25‑fold reduction relative to APSi + RT and an approximately 100‑fold reduction compared with PBS. This substantial superiority can be mechanistically attributed to the synergistic integration of Ang‑mediated LRP1 targeting and Pd‑catalyzed bubble propulsion, which together enable the nanorobot to cross the blood–brain barrier, accumulate selectively at the glioblastoma site, and penetrate deep into the tumor parenchyma; once localized, the Au–Pd core functions as a high‑Z radiosensitizer and the Pd nanozyme simultaneously generates cytotoxic ·OH.

Fig. 4.

(a) Schematic diagram of the treatment procedure. IVIS images (b) and quantitative analysis (c) of orthotopic brain glioma mice post-treatment. (d) Body weight of mice in each group after different treatments. (e) Survival rate of mice from different group. (f) H&E staining of tumor bearing brain from mice in different treatment group. (g) Fluorescence images of TUNEL staining at the tumor site. (h) H&E staining of liver, lung, spleen, kidney, heart and brain from mice in different group

The impact of each treatment on overall animal health was reflected in body weight trajectories over the course of therapy (Fig. 4d). Mice in the PBS and APSi groups experienced progressive and severe body weight loss, consistent with unchecked glioblastoma growth and progressive neurological decline. Radiotherapy alone partially stabilized body weight, while both APSi + RT and Ang‑APSi + RT groups maintained relatively stable body weights throughout the treatment period, with the Ang‑APSi + RT group showing the best preservation, suggesting that effective local tumor control mitigates tumor‑associated morbidity without causing overt systemic toxicity. Survival analysis over a 40‑day follow‑up period confirmed that the observed tumor control directly translated into a survival benefit (Fig. 4e). All mice in the PBS and APSi groups died within 20 days, yielding a survival rate of 0%. Radiotherapy alone extended median survival and produced a 40% long‑term survival rate, while APSi + RT increased the survival rate to 60%. Strikingly, the Ang‑APSi + RT group achieved an 80% survival rate, with the majority of mice surviving beyond 40 days. This stepwise improvement in survival—from RT through APSi + RT to Ang‑APSi + RT—directly mirrors the increasing capacity of each combination to deliver and retain the radiosensitizing nanorobots within the glioblastoma, underscoring the principle that therapeutic efficacy is governed by tumor‑specific accumulation and deep penetration.

Histological analyses of brain tissues collected on day 14 provided direct evidence of the differential treatment effects at the cellular level. Hematoxylin and eosin (H&E) staining revealed large, hypercellular tumor masses with invasive borders in the PBS and APSi groups. Radiotherapy alone produced a moderate reduction in tumor area, whereas the APSi + RT group showed further tumor shrinkage. In agreement with the bioluminescence data, the Ang‑APSi + RT group displayed the smallest residual tumor area, with a notable reduction in tumor cellularity and the presence of extensive necrotic regions (Fig. 4f). TUNEL staining quantitatively corroborated these findings: only a low basal level of apoptosis was detected in the PBS and APSi groups, while radiotherapy moderately increased TUNEL‑positive cells; the Ang‑APSi + RT group exhibited the most intense and widespread TUNEL signal throughout the tumor bed, consistent with the massive induction of apoptosis by the synergistic effects of radiation, O₂‑mediated radio-sensitization, and ·OH‑mediated oxidative damage. The spatial distribution of TUNEL‑positive cells was not confined to perivascular regions but extended into deep tumor compartments, confirming that the deep penetration capability of Ang‑APSi translates into homogeneously distributed therapeutic effects (Fig. 4g and S6). Finally, to evaluate the systemic safety of this combination therapy, major organs including the heart, liver, spleen, lungs, and kidneys were collected at the end of the treatment period and examined by H&E staining (Fig. 4h). No histopathological abnormalities—such as necrosis, fibrosis, or inflammatory infiltration—were observed in any group; the normal architecture of hepatocytes, renal glomeruli, cardiac myocytes, splenic follicles, and pulmonary alveoli was fully preserved. Collectively, these in vivo findings establish that Ang‑APSi, through its integrated functions of active BBB‑crossing, deep tumor penetration, catalytic oxygen generation, and radio-sensitization, achieves potent inhibition of orthotopic glioblastoma growth and significantly prolongs survival. Nevertheless, the observation that not all mice were cured suggests that additional barriers—most notably the immunosuppressive microenvironment that fosters tumor recurrence—remain to be addressed, providing the rationale for equipping the nanorobot with K⁺‑scavenging capability, as investigated in the subsequent section.

Construction of M@Ang-APSi and investigation of K⁺regulation performance

The robust tumor suppression achieved by Ang-APSi combined with radiotherapy was accompanied by the observation that a subset of tumors eventually recurred, implying that the immunosuppressive microenvironment—exacerbated by radiotherapy-induced K⁺ efflux—remained a significant therapeutic barrier. To address this, we endowed the nanorobot with the ability to specifically scavenge excess K⁺ by encapsulating Ang-APSi with a K⁺-selective membrane (M), yielding M@Ang-APSi, a single platform that integrates radio-sensitization, deep tumor penetration, and ion-mediated immune modulation (Fig. 5a). The M coating was deposited onto the nanorobot surface through a self-assembly and annealing process, forming a thin conformal layer that increased the hydrodynamic diameter by approximately 10 nm and shifted the surface zeta potential from − 7.9 mV to − 2.2 mV, consistent with the charge-screening effect of the neutral membrane (Fig. 5b, c). This membrane serves as a molecular gate that selectively permits K⁺ permeation into the mesoporous SiO₂ reservoir while excluding other physiologically abundant cations.

Fig. 5.

(a) Schematic illustration of the in situ formation of nanorobots and immune activation. Hydrodynamic diameters (b) and surface zeta potentials (c) of Ang-APSi and M@Ang-APSi. (d) K⁺ concentration after different treatments. (e) Concentrations of various ions following different treatments. Fluorescence images (f) and fluorescence intensity (g) at the tumor site following separate intravenous bolus injections of Ang-APSi or M@Ang-APSi. (h) The polarization status of macrophages after different treatments. (i-n) Flow cytometric plot and quantification analysis of CD4⁺TNFα⁺ T cells, CD4⁺IFNγ⁺ T cells, CD8⁺TNFα⁺ T cells and CD8⁺IFNγ⁺ T cells in tumor after different treatments

We first characterized the K⁺ adsorption capacity and ionic selectivity of M@Ang-APSi. In a solution containing only K⁺, both M@Ang-APSi and uncoated Ang-APSi reduced the K⁺ concentration from 87.3 to 39.4 mM, confirming that the underlying SiO₂ head provides ample adsorption capacity (Fig. 5d). The key advantage of the M coating emerged under competitive ionic conditions. When a mixed ion solution containing K⁺, Na⁺, Mg²⁺, and Ca²⁺ was used, uncoated Ang-APSi lost most of its K⁺ adsorption capacity, reducing the K⁺ concentration by only 11.8%, whereas M@Ang-APSi maintained its full K⁺-adsorbing performance, achieving removal comparable to that in the pure K⁺ solution (Fig. 5e and S7). This demonstrates that the membrane effectively discriminates against Na⁺, Mg²⁺, and Ca²⁺ while permitting selective K⁺ permeation, a selectivity attributable to the precisely defined coordination cavity within the membrane’s benzylcarbamoyl-functionalized architecture. In summary, when K⁺ ions are located at the pore channels of the membrane, the calculated K-O bond length is approximately 2.65 Å, which is highly consistent with the pore diameter radius (about 2.85 Å). K⁺ ions can precisely insert into the membrane pore channels in a way that achieves equilibrium between energy cost and benefit. However, when other hydrated ions attach to the membrane surface and remove the coordinating water molecules, the distance between the ions and the carbonyl oxygen atoms on the other side of the pore channel will significantly exceed the distance that should be present in the optimal coordination state (Fig.S8). Critically, the addition of the M coating did not compromise tumor targeting: IVIS system in orthotopic glioblastoma-bearing mice showed virtually identical tumor accumulation and organ distribution for M@Ang-APSi and Ang-APSi, confirming that the ultrathin membrane neither occludes the Ang ligand nor impedes Pd-driven propulsion (Fig. 5f, g and S9). Similarly, 2 days before the experiment, we also conducted IVIS system on glioma mice. The fluorescence intensities of the tumors in all experimental groups remained consistently at approximately 1.3 × 10⁷ (Fig. S10). To establish the pathophysiological relevance of K⁺ scavenging, we first verified the ionic consequences of radiotherapy-induced cell death. High-dose irradiation (20 Gy) of GL261 cells resulted in the release of approximately 2 mM K⁺ into the culture medium, whereas the concentrations of Na⁺, Mg²⁺, and Ca²⁺ remained essentially unchanged (Fig. S11), confirming that K⁺ is the dominant ionic species perturbed by tumor cell lysis. We then tested whether M@Ang-APSi-mediated K⁺ adsorption could mitigate the immunosuppressive effects of this high-K⁺ environment. Conditioned medium from irradiated tumor cells was incubated with M@Ang-APSi or Ang-APSi, and the treated medium was subsequently applied to primary macrophages and freshly isolated T cells. For macrophages, the high-K⁺ conditioned medium drove polarization toward an M2-like immunosuppressive phenotype. Treatment with uncoated Ang-APSi did not alter this outcome, consistent with its inability to selectively adsorb K⁺ under ion-competitive conditions. In contrast, macrophages cultured with M@Ang-APSi-treated medium exhibited a polarization profile closely resembling that of cells maintained in normal medium, with a restored proportion of M1-type cells (Fig. 5h and S12). For T cells, both CD4⁺ and CD8⁺ subsets cultured in high-K⁺ medium displayed markedly suppressed production of TNF-α and IFN-γ, recapitulating the established capacity of elevated extracellular K⁺ to impair T cell effector function. This suppression was not alleviated by Ang-APSi treatment but was significantly reversed by M@Ang-APSi, which restored intracellular cytokine levels in both T cell subsets to near-normal levels (Fig. 5i–n). Collectively, these findings establish that M@Ang-APSi, through its membrane-enabled selective K⁺ adsorption, effectively counteracts the dual immunosuppressive axes triggered by post-radiotherapy K⁺ efflux—preserving macrophage antitumor polarization and T cell effector function—while fully retaining its tumor-targeting and radio-sensitizing capabilities.

M@Ang-APSi alleviated the immunosuppressive microenvironment of glioblastoma by modulating K⁺

To determine whether the K⁺-scavenging capability of M@Ang-APSi translates into meaningful remodeling of the immunosuppressive tumor microenvironment in vivo, we established an orthotopic GL261 glioblastoma model and compared three treatment regimens: PBS, Ang-APSi + RT, and M@Ang-APSi + RT. Mice received intravenous nanorobot administration on day 0 and day 2, with 6 Gy radiotherapy delivered 12 h after each injection. Bioluminescence imaging on day 14 revealed that the tumor burden in the M@Ang-APSi + RT group was modestly but consistently lower than that in the Ang-APSi + RT group (Fig. 6a, b). Mice in the PBS group all died within 20 days, whereas both nanorobot-treated groups exhibited significantly prolonged survival. Critically, the M@Ang-APSi + RT group achieved superior long-term survival compared with the Ang-APSi + RT group (Fig. 6c), demonstrating that the addition of K⁺-scavenging function provides a survival benefit beyond that afforded by radio-sensitization alone. Histological examination of brain tissues further supported these findings: H&E and TUNEL staining showed that the M@Ang-APSi + RT group exhibited both a smaller residual tumor area and a higher density of apoptotic cells relative to the Ang-APSi + RT group (Fig. 6d and S13), confirming that K⁺ adsorption augments tumor cell elimination in the post-radiotherapy period.

Fig. 6.

IVIS images (a) and quantitative analysis (b) of orthotopic glioma mice post-treatment. (c) Survival rate of mice from different group. (d) H&E and TUNEL staining of tumor bearing brain from mice in different treatment group. (e-i) Flow cytometric plot and quantification analysis of mature DCs, CD3⁺CD4⁺ T cells, CD3⁺CD8⁺ T cells and macrophage in spleen or tumor. (j) Quantification analysis of CD4⁺TNFα⁺ T cells, CD4⁺IFNγ⁺ T cells, CD8⁺TNFα⁺ T cells and CD8⁺IFNγ⁺ T cells in spleen after different treatments. (k) Quantification analysis of CD4⁺TNFα⁺ T cells, CD4⁺IFNγ⁺ T cells, CD8⁺TNFα⁺ T cells and CD8⁺IFNγ⁺ T cells in tumor from mice after different treatments

To mechanistically dissect how M@Ang-APSi-mediated K⁺ scavenging reshapes the antitumor immune response, we collected tumors, spleens, and superficial cervical lymph nodes at the end of the treatment period and performed comprehensive immune profiling by flow cytometry. The proportion of DCs (CD80⁺CD86⁺) in the tumor-draining lymph nodes was comparable between the Ang-APSi + RT and M@Ang-APSi + RT groups (Fig. 6e, f), indicating that K⁺ adsorption does not further enhance the initial antigen presentation event, which is primarily driven by the immunogenic cell death induced by radiotherapy. This is consistent with the fact that DC maturation is mainly triggered by DAMPs released during radiation-induced apoptosis and is not directly influenced by extracellular K⁺ levels.

In contrast, striking differences emerged in the effector compartments of the immune response within the tumor microenvironment. Analysis of tumor-infiltrating leukocytes revealed that the M@Ang-APSi + RT group exhibited a distinct shift in macrophage polarization compared with the Ang-APSi + RT group: the proportion of M1-like macrophages (CD86⁺) increased from 36.0% to 41.5%, while the population of immunosuppressive M2-like macrophages (CD206⁺) was drastically reduced by approximately 4.8-fold (from 36.6% to 7.58%) (Fig. 6g). Furthermore, the M@Ang-APSi + RT treatment promoted robust T cell infiltration into the glioblastoma tissue. Specifically, the frequency of CD3⁺CD8⁺ T cells nearly doubled (increasing from 24.2% to 45.1%), and CD3⁺CD4⁺ T cells rose from 28.4% to 32.4% relative to the Ang-APSi + RT group (Fig. 6h, i), suggesting that the alleviation of K⁺-mediated immunosuppression facilitates T cell recruitment and/or intratumoral persistence.

The functional quality of the T cell response was further quantified by intracellular cytokine staining. In the tumor microenvironment, the M@Ang-APSi + RT group demonstrated superior effector function compared to the Ang-APSi + RT group. For instance, the frequency of IFN-γ-producing CD8⁺ T cells increased by approximately 1.7-fold (from ~ 15% to ~ 25%), and TNF-α-producing CD8⁺ T cells rose from ~ 15% to ~ 20% (Fig. 6k). Similarly, within the spleen, the proportions of cytokine-secreting T cells were elevated, with IFN-γ⁺CD8⁺ T cells increasing from ~ 35% to ~ 50% (Fig. 6j). These quantitative improvements indicate that the functional competence of tumor-infiltrating lymphocytes is restored when K⁺ is actively removed. By globally reducing K⁺ concentrations, M@Ang-APSi relieves suppression and allows T cells to execute their full cytotoxic and helper functions, extending systemic antitumor immunity beyond the local tumor site.

Taken together, these in vivo immunological analyses establish that M@Ang-APSi, through its M-enabled selective K⁺ adsorption, effectively remodels the immunosuppressive microenvironment of glioblastoma following radiotherapy. The nanorobot retains the potent tumor-targeting and radio-sensitizing properties of Ang-APSi while adding a critical immune-modulatory function: by scavenging K⁺ leaked from dying tumor cells, it prevents macrophage M2 polarization, enhances T cell infiltration, and restores T cell effector cytokine production. This dual action—direct radio-sensitization and indirect immune reactivation—provides a mechanistic basis for the superior survival outcomes observed in the M@Ang-APSi + RT group and validates the therapeutic rationale of integrating ion-scavenging capability within a targeted nanorobotic platform.

RNA transcriptomic analysis of the M@Ang-APSi groupR

To further elucidate the molecular mechanisms through which M@Ang-APSi-mediated K⁺ scavenging reshapes the glioblastoma immune microenvironment, we performed a comprehensive transcriptomic analysis. We first verified that the K⁺ concentration in the tumor tissue of M@Ang-APSi-treated mice was indeed reduced to near-physiological levels. At 12 h after radiotherapy, intratumoral K⁺ in the M@Ang-APSi group was approximately 5 mM, comparable to that of normal brain tissue, whereas the Ang-APSi group maintained markedly elevated K⁺ concentrations (Fig. 7a). This confirms that the M-enabled selective adsorption effectively normalizes the ionic microenvironment in vivo. Subsequently, tumors were collected at the end of the treatment period from the Ang-APSi + RT and M@Ang-APSi + RT groups and subjected to whole-genome RNA sequencing. Partial least squares discriminant analysis (PLS-DA) revealed a clear separation of the transcriptomic profiles between the two groups, indicating that K⁺ scavenging induces a global reprogramming of gene expression within the tumor tissue (Fig. 7b). Compared with the Ang-APSi + RT group, the M@Ang-APSi + RT group exhibited 2800 significantly upregulated genes and 694 significantly downregulated genes (Fig. 7c).

Fig. 7.

(a) K⁺ ion concentration at the tumor site. (b) Partial least squares discriminant analysis of Ang-APSi and M@Ang-APSi. (c) Volcano plot displaying the differentially expressed genes. (d) KEGG enrichment scatter diagram of Ang-APSi and M@Ang-APSi. (e) Reactome functional enrichment analysis of Ang-APSi and M@Ang-APSi. (f) GO analysis of Ang-APSi and M@Ang-APSi. (g) Heatmap evaluating gene expression in different pathways. (n = 3)

Pathway enrichment analyses using KEGG and Reactome databases identified the biological processes most prominently altered by K⁺ scavenging. The most significantly enriched pathways centered on ion channel function and regulation, with a particular emphasis on K⁺ channel activity, as well as on the inhibition of protein synthesis and translation represented by the ribosomal pathway (Fig. 7d, e). This dual signature is mechanistically consistent with the canonical cellular response to ionic stress: elevated intracellular K⁺ is known to regulate the activity of multiple ion channels and transporters, and the removal of excess extracellular K⁺ relieves this regulatory constraint. Concurrently, the suppression of ribosomal protein expression and translation-related genes suggests that the high-K⁺ microenvironment imposes a metabolic and translational burden on tumor and immune cells that is alleviated upon K⁺ normalization. In addition, the cAMP signaling pathway, the Wnt signaling pathway, and the G protein-gated inwardly rectifying K⁺ (GIRK/Kir3) channel pathway were among the significantly altered signaling cascades, indicating that K⁺ scavenging reverberates through multiple interconnected signaling networks that govern cell survival, proliferation, and immune function.

Gene Ontology (GO) enrichment analysis across cellular component, biological process, and molecular function categories further refined this picture (Fig. 7f). Beyond the expected associations with ion channel complexes and ribosomal subunits, we observed significant enrichment in terms related to GABAergic synaptic signaling, AMPA and NMDA receptor activity regulation, and microtubule- and kinesin-mediated cell migration and guidance. The involvement of neurotransmitter receptor pathways likely reflects the unique neural microenvironment of glioblastoma and suggests that K⁺ dysregulation may interfere with neuron–glioma synaptic communication, while the enrichment of cytoskeletal and cell migration terms indicates that the ionic microenvironment influences the invasive behavior of tumor cells and the motility of infiltrating immune cells.

Analysis of the significantly altered individual genes across key pathways provided a coherent mechanistic narrative (Fig. 7g). Genes encoding K⁺ channels, including multiple inwardly rectifying and voltage-gated K⁺ channel subunits, were prominently upregulated in the M@Ang-APSi group, reflecting the activation of transcriptional programs that restore ion homeostasis. In contrast, genes involved in ribosomal assembly and protein translation were broadly downregulated, consistent with a release from a stress-induced translational state. Notably, components of the MAPK signaling pathway were modulated, with a shift toward profiles associated with reduced proliferation and increased differentiation. Concurrently, transcripts encoding pro-inflammatory mediators were reduced, while those associated with macrophage M1 polarization and T cell activation—including genes linked to antigen presentation, co-stimulation, and effector cytokine production—were significantly enriched in the M@Ang-APSi group. This transcriptional signature directly aligns with the observed in vivo phenotypes of increased M1 macrophage and activated T cell infiltration, and provides the molecular basis for the functional immune restoration achieved by K⁺ scavenging.

Conclusion

In summary, we have developed a train‑style nanorobot, M@Ang‑SiO₂–Au–Pd, that concurrently addresses two major barriers to effective glioblastoma radiotherapy-hypoxia‑driven radio-resistance and post‑irradiation K⁺‑mediated immunosuppression-through the rational integration of catalytic radio-sensitization and selective ion scavenging within a single platform. The nanorobot employs an asymmetric architecture in which an Angiopep‑2‑modified mesoporous SiO₂ head enables LRP1‑mediated blood–brain barrier crossing and tumor targeting, while a Pd nanozyme tail catalyzes the decomposition of endogenous H₂O₂ into O₂ to relieve tumor hypoxia and provide bubble propulsion for deep intratumoral penetration [33]. Under X‑ray irradiation, the Au–Pd heterojunction functions as a high‑Z radiosensitizer that amplifies local energy deposition and synergizes with Pd‑generated ·OH to maximize DNA damage and apoptosis [34]. This targeted radio-sensitization achieved potent tumor growth inhibition and significantly prolonged survival in an orthotopic glioblastoma model. To overcome the immunosuppressive consequence of radiotherapy—the abrupt leakage of intracellular K⁺ that drives macrophage M2 polarization and T cell exhaustion—the nanorobot was encapsulated with a K⁺‑selective membrane that endows the SiO₂ reservoir with the capacity to specifically adsorb excess K⁺ from the tumor microenvironment. This ion‑scavenging function effectively maintained intratumoral K⁺ at physiological levels, preserved macrophage antitumor polarization, restored T cell effector cytokine production, and promoted immune infiltration, thereby reversing the immunosuppressive microenvironment and further extending survival beyond that achieved by radio-sensitization alone. Transcriptomic analysis confirmed that K⁺ normalization triggered broad reprogramming of ion channel, ribosomal, and immune effector pathways, providing a molecular basis for the observed functional immune restoration. This work establishes a new paradigm in which nanorobots are employed not merely as drug delivery vehicles but as active modulators of the ionic tumor microenvironment, offering a comprehensive strategy that integrates radio-sensitization with immunomodulation to improve treatment outcomes for glioblastoma and potentially other malignancies where ionic dysregulation contributes to immune evasion [35–38].

Author contributions

X.J. and W.Z. conceptualized and designed the study. P.L., R.C. and M.D. performed the experiments, and drafted the manuscript. P.L. X.X. and X.Y. contributed to data collection, analysis, and interpretation. Y.Z. B.Z. and X.L. validated the experimental results, and helped with data visualization. P.L. and M.D. contributed to software development and data validation. X.J. oversaw the project, secured funding, and reviewed and edited the manuscript for final approval.

Funding

Tianjin Science and Technology Project (Grant No. 25JCLMJC00090, X.J.), National Natural Science Foundation of China (Grant No. 32471388 and 32071322, X.J.), and National Natural Science Funds for Excellent Young Scholars (Grant No. 32122044, X.J.), and Natural Science Foundation of Shandong Province (Grant No. ZR2021QH045, W.Z.).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The experimental protocols were approved by the Animal Ethics Committee of the Tianjin University Laboratory Animal Center (Tianjin, China) under Approval No. TJUE-2024-314.

Consent for publication

The author confirm that the work described has not been published before; that it is not under consideration for publication elsewhere; that its publication has been approved by all co-authors.

Competing interests

The authors declare no competing interests.