RAP-peptide functionalized biomimetic nanoformulation with pathological ROS/pH-responsive drug release for target immunotherapy in glioma

Abstract

Glioblastoma (GBM) is one of the most aggressive malignancies of the central nervous system. Gemcitabine (GEM), a pyrimidine analogue with broad-spectrum anticancer activity, can activate the cGAS-STING pathway and alleviate the immunosuppressive microenvironment of GBM. However, its clinical application is hampered by the formidable challenge of crossing the blood-brain barrier (BBB) and accumulating at the tumor lesion. Herein, a dual-responsive biomimetic nanoprodrug (RMM@GEM NPs) was exploited to enhance the efficient BBB penetration and target cargo delivery by functionalization of glioblastoma cell membranes (MM) camouflaging and further targeting peptide RAP modification. After its selective accumulation at glioma lesion, RMM@GEM NPs accelerates GEM release under the tumor pathological stimuli of reactive oxygen species (ROS) and acidic microenvironment to robustly activate the STING signaling cascades (increased p-STING, p-TBK1, p-IRF3, and p-NF-κB). Simultaneously, cyclodextrin-mediated cholesterol depletion further suppresses PD-L1 expression and alleviates T-cell exhaustion. These findings highlight RMM@GEM NPs as a promising strategy to enhance immune responses in “cold” tumor, providing a potential candidate for efficient and safe immunotherapy in GBM.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04055-4.

Introduction

Glioblastoma multiforme (GBM) is among the most aggressive and difficult-to-treat brain tumors [1]. As the most common primary brain tumor, gliomas make up roughly 30% of all primary brain neoplasms, with over 80% being malignant [2]. According to the World Health Organization (WHO) classification [3–5], GBM is classified as grade IV, marked by high invasiveness, rapid growth, and notable treatment resistance. Current clinical approaches mainly involve surgical removal, but their success is limited because tumor cells extensively infiltrate white matter tracts, making complete excision nearly impossible [5–7]. Additionally, long-term chemotherapy frequently leads to drug resistance [10]. Recently, targeted therapies and immunotherapies have attracted growing interest [8]. However, GBM is often considered a classic “cold tumor” due to the blood-brain barrier (BBB) and its unique immune environment. It employs dual immune resistance mechanisms, i.e., intrinsic immune evasion by tumor cells and intense adaptive immunosuppression [9, 10]. These interconnected processes create a complex immunosuppressive network that significantly reduces the efficacy of immunotherapy.

Gemcitabine (GEM) has been reported to induce DNA damage, activate p53, and enhance its transcriptional accessibility to apoptosis-related genes, thereby triggering programmed cell death [11]. In addition, DNA fragments generated by chemotherapy activate the STING-dependent type I interferon pathway, which promotes the expression of primary histocompatibility complex class I (MHC I) and interferon-γ (IFN-γ) in tumor cells, thereby initiating an immune response [12]. These findings highlight the preliminary therapeutic potential of GEM for GBM. However, its clinical efficacy is severely hampered by its low accumulation in brain tissue [13, 14]. Moreover, STING activation paradoxically induces PD-L1 upregulation through the IFN-β/IFN-γ-JAK-STAT signaling cascade, thereby facilitating immune evasion by tumor cells [15]. This “double-edged sword” phenomenon markedly compromises the recognition and cytotoxicity of immune cells.

Recently, cholesterol was reported to play a pivotal role in immune evasion by tumor cells, as it binds to the CRAC domain of PD-L1, forming a stable complex that prevents PD-L1 degradation and thereby enables tumors to evade immune surveillance. Meanwhile [16], tumor growth demands high cholesterol, and cholesterol-enriched tumor microenvironments (TMEs) sustain hyperactivation of the AKT-mTORC1-SREBP1 axis in CD8⁺ T cells, ultimately driving T cell exhaustion [17]. β-CD a class of polysaccharide-based supramolecular compounds with unique cavity structures, has been widely employed in drug delivery owing to its excellent inclusion capacity [18]. CDs not only function as carriers to enhance solubility, stability, and bioavailability of drugs but also serve as cholesterol scavengers, offering novel opportunities for tumor immunotherapy [19]. In addition, CDs can selectively bind and extract cholesterol, effectively reducing cholesterol content in tumor cell membranes [19]. This dual functionality could suppress STING-mediated PD-L1 upregulation, thereby blocking immune escape, while simultaneously reversing T cell exhaustion, leading to enhanced immune recognition and tumor eradication. Consequently, CDs could not only serve as efficient GEM carriers but also mitigate GEM-induced immunological side effects, yielding a synergistic therapeutic benefit.

To further improve the target function and mitigate potential side effects of β-CD or β-CD-based nanocarriers [20], cell membrane-camouflaged nanoformulation can significantly enhance biocompatibility and prolong blood retention time, inheriting the natural biological functions of the source cells [21, 22]. By leveraging natural membranes (e.g., erythrocytes, macrophages, neutrophils, or tumor cells), these biomimetic systems acquire cellular functionalities such as homologous targeting and immune evasion [23]. Glioma cell membrane-coated nanoformulation displays unique cytokine receptors (e.g., integrin αV, CD44, CD47, EpCAM), endowing their payloads with homotypic targeting and immune evasion capabilities. However, the intrinsic asymmetry and biological complexity of membranes limit the directional “right-side-out” assembly of membrane proteins when using conventional coating techniques [23, 24]. Based on our previous studies, a PS-targeting peptide (PSP, CLIKKPF)-modified nanoformulation was engineered to bind phosphatidylserine (PS) in glioma membranes with high affinity, thereby guiding the membrane assembly in the desired right orientation [25, 26]. This strategy effectively overcomes the limitations of traditional co-extrusion or sonication methods, harnessing the homing capacity of glioma membranes. The spontaneous right-side-out assembly generates a biomimetic coating for nanocarriers, thereby enhancing the targeting and safety of drug delivery to glioma lesions [27].

Herein, as shown in Scheme 1, we developed a dual-responsive drug delivery system in which β-CD served as the supramolecular host, conjugated with gemcitabine prodrugs via a linker responsive to reactive oxygen species (ROS). A benzimidazole-grafted PSP acted as a pH-sensitive switch, co-assembling with β-cyclodextrin into supramolecular nanostructures. The PSP selectively recognized phosphatidylserine, enabling glioma membranes to spontaneously coat on the surface of the nanoformulation, while RAP was further embedded on the outer shell to provide the enhanced targeting activity to glioma lesion [28, 29]. This design ensured that GEM was released exclusively within ROS-enriched tumor microenvironments. In contrast, the pH-sensitivity of benzimidazole prevented nonspecific cholesterol extraction by β-cyclodextrin [30], thereby reducing hemolytic activity and renal toxicity. Importantly, this strategy localized β-cyclodextrin release to tumor cell lysosomes, allowing effective cholesterol depletion and enhancing immune cell recognition. Furthermore, the target functionalization of glioma membrane camouflaging and RAP modification not only facilitated the enhanced BBB penetration and lesion-specific homing.

Scheme 1.

Schematic illustration of the preparation and applications of RMM@GEM NPs for glioma treatment

Results and discussions

Preparation and characterization of tumor cell membrane-camouflaged ROS/pH Dual-Responsive nanotherapeutics

The synthetic route involved an amidation reaction between amino-modified β-CD and thiohydroxyacetic anhydride, followed by the covalent attachment of GEM to the activated carboxyl groups, yielding CD-GEM prodrugs (Figure S1, S2, Supporting Information). Subsequently, the benzimidazole group coupled with PSP targeting peptide (BM-PEG-PSP) was introduced into the CD-GEM cavity to self-assemble to form GEM NPs (Fig. 1A). Fourier verified the chemical structures using transform infrared spectroscopy (FTIR) and hydrogen nuclear magnetic resonance (¹H NMR) spectroscopy (Figure S3-S7, Supporting Information). The C = O stretching vibration peak at 1715 cm⁻¹ diminished after GEM conjugation, while characteristic GEM skeletal vibrations (1630 cm⁻¹) and β-CD C-O-C peaks (1080 cm⁻¹) were retained. Formation of amide bonds was further supported by distinct absorption at 1719 cm⁻¹ (C = O stretching) and 1550 cm⁻¹ (N-H bending). Notably, upon BM insertion into the β-CD cavity, the N-H stretching peak shifted and broadened from 3400 cm⁻¹, confirming host-guest interactions between BM-PEG-PSP and GEM-CD. ¹H NMR spectra further revealed pyrimidine ring protons of GEM at 8.17–8.18 ppm and 7.73–7.75 ppm, while β-CD resonances appeared at 5.63–5.79 ppm and 3.42–3.78 ppm (Figure S6, S7, Supporting Information). The NOESY spectrum further confirmed intermolecular interactions, with cross-peaks observed between the aromatic region (7.5–9.0.5.0 ppm) and the inner sugar protons of β-CD (3.0–4.0 ppm) (Fig. 1F). In addition, successful synthesis of DSPE-PEG-RAP was confirmed by the disappearance of maleimide signals and accurate mass determination by MALDI-TOF (Figure S8, S9, Supporting Information).

Fig. 1.

Synthesis and characterization of RMM@GEM NPs. (A) Schematic representation of the synthesis of RMM@GEM NPs. (B) Particle size distribution of RMM@GEM NPs. (C) Average particle size of GEM NPs, MM@GEM NPs, and RMM@GEM NPs. (D) Zeta potentials of GEM NPs, MM@GEM NPs, and RMM@GEM NPs. (E) Particle size and PDI stability of RMM@GEM NPs over 7 days. (F) NOESY NMR spectrum of GEM-CD. (G) Western blot analysis of key membrane proteins (Integrin αV, CD44, CD47, EpCAM) derived from glioblastoma cell membranes. (H) Representative TEM images of RMM@GEM NPs (scale bar = 200 nm). (I) CLSM images showing colocalization of Rhodamine B-labelled NPs with DiO-labeled RMMs in cells (scale bar = 10 μm). (J) The cumulative release profiles of GEM under acidic conditions (pH 5.5), ROS (0.5 mM H₂O₂), and their synergistic stimulation. Data are presented as mean ± standard deviation (n = 3)

To obtain RMM@GEM NPs, freshly prepared GL261 cell membrane (MM) was co-incubated with GEM NPs and DSPE-PEG-RAP. Fluorescence imaging analysis revealed a pronounced colocalization of RB-labelled GEM NPs with DiO-labelled vesicles (Fig. 1I). Transmission electron microscopy (TEM) imaging revealed a coronal layer surrounding RMM@GEM NPs (Fig. 1H), indicating the successful cell membrane coating mediated by the spontaneous PSP coupling. Dynamic light scattering measurements showed that GEM NPs possessed a mean hydrodynamic diameter of 186.55 ± 8.3 nm with a PDI index of 0.16 (Fig. 1B, C), which increased to 230.18 ± 2.9 nm with a PDI index of 0.18 after MM coating (mass ratio 1 : 2) (Figure S10, Supporting Information). Zeta potential measurements further demonstrated the stepwise surface charge shifts: GEM NPs (−3.6 ± 0.4 mV), MM@GEM NPs (−13.65 ± 0.8 mV), and RMM@GEM NPs (−13.98 ± 0.7 mV) (Fig. 1D). The pronounced negative shift not only confirmed membrane incorporation but also suggested the improved colloidal stability because of the enhanced electrostatic repulsion (Fig. 1E). Western blot analysis confirmed the retention of tumor-homing adhesion proteins (Integrin αV, CD44, EpCAM) and immune-evasive marker CD47 of MM@GEM NPs and RMM@GEM NPs (Fig. 1G).

The dual-responsive drug release profile was subsequently evaluated. As shown in Figure S12, GEM release remained minimal under physiological pH (7.4), where benzimidazole (BM) retained its hydrophobic state, exhibiting strong affinity toward β-cyclodextrin (β-CD) and remaining stably accommodated within its hydrophobic cavity. Under acidic conditions (pH 5.5), protonation enhanced the hydrophilicity of BM, thereby weakening its binding with β-CD and increasing the cumulative release from 10.7% to 21.5%, indicative of pH responsiveness [30]. Meanwhile, thioether bonds, intrinsically unstable under oxidative stress, were readily oxidized by ROS into sulfoxide/sulfone intermediates or cleaved through radical-mediated homolysis, thereby accelerating drug release. Accordingly, in the presence of H₂O₂, the cumulative release increased from 19.8% (0.1 mM H₂O₂) to 47.3% (0.5 mM H₂O₂). Notably, the combination of acidic pH and ROS stress elicited the most pronounced release (Fig. 1J), significantly surpassing that triggered by either stimulus alone. Consistently, TEM images also revealed distinct disassembly of the nanoformulation (Figure S13, Supporting Information). These results highlight the synergistic dual-responsiveness of the system to ROS and pH, ensuring rapid and efficient payload release within the tumor microenvironment.

In vitro antitumor activity of RMM@GEM NPs

In clinical practice, a hallmark feature of gliomas is their aggressive invasiveness. To assess the impact of nanoparticle formulations on glioma progression, we performed Transwell migration assays. In this experiment, viable cells treated with various drugs were transferred to the upper chamber of the Transwell, and the number of cells in the lower chamber was observed at 6, 12, and 24 h. As shown in Fig. 2A, RMM@GEM NPs significantly inhibited the invasion of GL261 cells, with the inhibitory effect increasing over time with drug pre-treatment. This result indicateed that gemcitabine effectively suppressed cell invasion. Additionally, Western blot analysis revealed a significant reduction in the expression levels of two key proteins associated with tumor cell invasion, MMP-9 and VEGF, following gemcitabine treatment, with the lowest contents observed in the RMM@GEM NPs group (Fig. 2B). This effect is primarily attributed to the incorporation of RAP, which specifically recognizes the overexpressed LRP1 receptor on tumor cell surfaces (Fig. 2C).

Fig. 2.

In vitro antitumor efficacy of RMM@GEM NPs. (A) Cell invasion and migration images after incubation with different formulations for 6, 12, and 24 h (scale bar = 50 μm). (B) Western blot analysis of key intracellular proteins (MMP-9, VEGF) in glioblastoma cells after treatment with different formulations. (C) Schematic illustration showing RAP-mediated specific recognition of the LRP1 receptor on tumor cell surfaces, facilitating the target binding of RMM@GEM NPs. (D) Live/dead staining of cells after 48 h incubation with the different formulations. Live cells were labelled with Calcein-AM (green) and dead cells with PI (red) (scale bar = 200 μm). (E) Cell viability of GL261 cells after incubation with the different formulations for 48 h. (F) Representative images of colony formation following treatment with the different formulations. (G) Semi-quantitative analysis of colony formation in plate assays. (H) Flow cytometric analysis of apoptosis in GL261 cells treated with the different formulations. Statistical significance was determined as *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001

To evaluate the pharmacological activity of the nanoformulation, free GEM was used as a control, and the cytotoxicity of RMM@GEM NPs was assessed against a panel of glioma cells (GL261, U251, and U87) as well as normal cells (BV2 microglia and bEnd.3 brain microvascular endothelial cells) using the CCK-8 assay. As shown in Fig. 2E, all four formulations, including Free GEM, GEM NPs, MM@GEM NPs, and RMM@GEM NPs, exhibited pronounced dose-dependent cytotoxicity toward glioma cells, with no significant differences among groups (p > 0.05), Consistent with the results of live death staining (Fig. 2D). Remarkably, in assays using normal cells, RMM@GEM NPs demonstrated a favorable safety performance, with IC50 values significantly higher than those of the other groups (p < 0.001). Particularly at low concentrations, the cytotoxicity of RMM@GEM NPs toward BV2 and bEnd.3 cells were substantially weaker than those of the controls (Figure S14, Supporting Information), indicating that RAP modification confers the selective cytotoxicity resulting from the enhanced target antitumor activity while the minimized off-target toxicity of RMM@GEM NPs. Subsequently, such selectivity likely arises from tumor-specific uptake mediated by the targeting peptide and the dual regulation of drug release by ROS- and pH-responsiveness.

Apoptosis induction was subsequently examined by Annexin V-FITC/PI staining and flow cytometry. After 12 h incubation, RMM@GEM NPs significantly increased the late apoptosis to 18.44%, outperforming Free GEM, GEM NPs, and MM@GEM NPs (Fig. 2H). Live/dead staining further confirmed the highest level of cell death in the RMM@GEM NPs group, in agreement with flow cytometric analysis (Fig. 2D). Moreover, colony formation assays revealed that RMM@GEM NPs markedly reduced clonogenic growth, highlighting their long-term tumor-suppressive effect (Fig. 2F, G). Collectively, these results demonstrate that RMM@GEM NPs exert a robust and selective cytotoxicity against glioma cells while sparing normal cells, thereby enhancing the therapeutic safety and efficacy.

RMM@GEM NPs penetration across the in vitro BBB model and lysosomal escape

Previous studies have reported that the RAP exhibits a high affinity for the LRP1 receptor, enabling modified nanoformulation to more effectively traverse the blood-brain barrier and accumulate at tumor lesions [29]. To evaluate the BBB permeability of the different formulations, a monolayer of murine brain microvascular endothelial cells (bEnd.3 cells were seeded in the top chamber of a transwell system (Fig. 3A), maintaining a transendothelial electrical resistance (TEER) above 300 Ω·cm² throughout the study (Figure S15, Supporting Information). As shown in Fig. 3B, amount of the penetrated nanoformulation in the bottom chamber increased over time, especially the BBB permeability of RAP-modified nanoformulation, RMM@GEM NPs, was significantly higher than that of GEM NPs and MM@GEM NPs at 12 h, rising by 2.14- and 2.07-fold, respectively. In contrast, GL261 cells were introduced into the bottom chamber to establish an in vitro BBB model (Fig. 3C). RB-labelled GEM NPs, MM@GEM NPs, and RMM@GEM NPs were added to the upper chamber, and samples were collected at the different time interval to investigate the BBB permeability based on the concentration of nanoformulation detected in the bottom chamber. In parallel, Confocal laser scanning microscopy (CLSM) and flow cytometry were employed to evaluate the uptake of nanoformulation by tumor cells in the bottom chamber (Figure S16, Supporting Information). CLSM images revealed that GL261 cells incubated with RMM@GEM NPs exhibited the strongest fluorescence, whereas MM@GEM NPs (without RAP modification) showed weaker fluorescence, and GEM NPs displayed the lowest signal (Fig. 3D). Consistently, flow cytometry results confirmed that tumor cell uptake of RMM@GEM NPs was significantly higher than that of the other two groups after 12 h of co-incubation (Fig. 3E-F). Subsequently, RB-labelled RMM@GEM NPs were co-localized with LysoTracker to investigate lysosomal escape behavior. The clear colocalization of RMM@GEM NPs (red) with LysoTracker-labelled lysosomes (green) was observed at 2 h, indicating that the nanoformulation was internalized into the acidic lysosomes. Moreover, the colocalization of nanoformulation with lysosomes decreased after 8 h of incubation, suggesting drug escape from lysosomes (Fig. 3G-H). In addition, the cell uptake of RMM@GEM NPs was significantly inhibited after treatment with the LRP1 inhibitor 11H4 (Figure S17, Supporting Information). Consistent with the previous findings, in the acidic lysosomal environment (pH ≈ 5.5), benzimidazole groups been protonated and dissociated from the β-cyclodextrin cavity, leading to nanoformulation disassembly and subsequently the enhanced release of active drug components, thereby promoting lysosomal escape. Overall, RMM@GEM NPs demonstrated the superior BBB penetration, tumor-targeting ability, and lysosomal escape for enhancing the anti-tumor therapeutic efficacy.

Fig. 3.

In vitro investigation of BBB penetration and lysosomal escape of RMM@GEM NPs. (A) Schematic illustration of the in vitro BBB model. (B) Quantification of RB fluorescence in the bottom chamber after incubation with RB-labelled GEM NPs, MM@GEM NPs, and RMM@GEM NPs for 3, 6, and 12 h. (C) Schematic illustration of the in vitro BBB model in the GL261 cellular uptake assay. (D) CLSM images of GL261 cells in the bottom chamber (scale bar = 20 μm). (E) Flow cytometry histograms of GL261 cells after the different treatments. (F) Quantification of RB fluorescence by flow cytometry. (G) CLSM images of GL261 cells after incubation with RB-labelled RMM@GEM NPs for 2, 4, 6 and 8 h. RB-labelled NPs (red) and Lyso-tracker (green) labelling lysosomes (scale bar = 10 μm). (H) Colocalization analysis of RB-labelled NPs (red) with Lyso-tracker-labelled lysosomes (green). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 were considered statistically significant

RMM@GEM NPs crossing the BBB, prolonged Circulation, and glioblastoma targeting in vivo

An in vivo imaging system (IVIS) spectrum was employed to evaluate the biodistribution of nanomedicines in glioma-bearing mice, thereby further verifying their BBB penetration capability and tumor-targeting efficiency. Free RB, RB-labelled GEM, RB-labelled MM@GEM NPs, and RB-labelled RMM@GEM NPs were intravenously administered into mice bearing orthotopic gliomas (Figure S18, Supporting Information). As shown in Fig. 4A, among them, RB-labelled RMM@GEM NPs exhibited the highest brain accumulation at 6 h, reached the strongest fluorescence signal at 12 h, and remained the detectable fluorescence for up to 36 h. By contrast, because of the insufficient targeting, RB-labeled GEM and RB-labeled MM@GEM NPs displayed significantly weaker brain fluorescence signals at 12 h, only 0.24- and 0.29-fold of that observed with RMM@GEM NPs (Fig. 4C). Consistent with the in vivo imaging results, analysis of brain distribution revealed that RMM@GEM NPs exhibited the most vigorous fluorescence intensity within the brain tissue of glioma-bearing mice (Fig. 4E), showing a 1.6-fold enhancement compared with that in the other groups (Fig. 4F). After 24 h post-injection, immunofluorescence staining of whole-brain slices confirmed the intra-tumoral distribution of the formulations. Notably, RMM@GEM NPs, which are modified by both cell membrane coating and the RAP, demonstrated the pronounced BBB penetration and achieved the specific enrichment within the tumor lesion (Fig. 4I), confirming the enhanced delivery efficiency resulting from the RAP-mediated specifically recognized LRP1 receptors highly expressed on both neovascular endothelial cells and glioma cells (Fig. 4G). Such spatially elevated expression provided multiple binding sites for RAP-modified nanoformulation, mediating the active targeting and the efficient intra-tumoral accumulation. Furthermore, the cell membrane coating endowed the nanoformulation with favorable biocompatibility and prolonged circulation, thus providing additional advantages for the target cargo delivery. These results highlight the synergistic contribution of the target peptide and tumor cell membrane homing properties in enhancing glioma-specific drug delivery.

Fig. 4.

In vivo BBB penetration, prolonged circulation, and glioblastoma targeting of RMM@GEM NPs. (A) Representative fluorescence images of mice bearing orthotopic GL261 tumors at 6, 12, 24, and 36 h after the intravenous injection of the different treatment groups (n = 3). (B) Ex vivo fluorescence images of the heart, liver, spleen, lungs, and kidneys were collected from GL261 tumor-bearing mice after treatment with the different treatment groups. (C) Quantification of RB fluorescence intensity in mice from the different treatment groups. (D) Quantification of RB fluorescence intensity in major organs. (E) Fluorescence images of mice brain tissues treated with the different treatment groups. (F) Quantification of RB fluorescence intensity in brain tissues. (G) CLSM images showing LRP1 protein expression in RAW264.7, bEnd.3, and GL261 cells (scale bar = 20 μm). (H) Fluorescence intensity of RB in blood samples was measured to evaluate the circulation profile of the formulations in mice. (I) Immunofluorescence images of brain Sect. 36 h after treatment, showing the distribution of RB (red), LRP1 (green), and nuclei stained with DAPI (blue) (scale bar = 20 μm, n = 3). *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 were considered statistically significant

Biodistribution analysis in major organs (heart, liver, spleen, lung, and kidney) was further validated by the in vivo imaging results, showing the most vigorous fluorescence intensities in the liver and kidneys (Fig. 4B, D), indicating that RMM@GEM NPs could be metabolized by the renal and excreted. To further assess the circulation profile, blood samples were collected from the tail veins of mice at different time intervals after post-injection, and fluorescence intensities were quantified. As shown in Figure S19, fluorescence signals gradually decreased within 24 h across all groups. Interestingly, RMM@GEM NPs and MM@GEM NPs maintained the strong fluorescence signals after post-injection for 6 h, whereas signals from free RB and GEM NPs declined rapidly after post-injection for 2 h (Fig. 4H). Blood circulation analysis further confirmed that RMM@GEM NPs significantly prolonged the blood circulation time compared with that of the free RB and GEM NPs. In addition, hemocompatibility was also evaluated using a hemolysis assay in vitro. The hemolysis rate of RMM@GEM NPs remained below 5% at all tested concentrations, which was substantially lower than that of free β-CD (Figure S20, Supporting Information). Therefore, given the synergetic effects of the targeting peptide and biomimetic cell membrane coating, RMM@GEM NPs significantly improved drug delivery efficiency across the BBB and enhanced the precise target delivery to glioma.

In vitro assessment of STING activation and PD-L1 Inhibition

To evaluate the effect of GEM on STING pathway activation in vitro, Western blotting was performed to analyze STING-related protein expression in GL261 cells. As shown in Fig. 5A, the phosphorylation levels of STING, TBK1, IRF3, and NF-κB were increased after GEM treatment, whereas β-CD treatment did not show a significant effect, but STING activation occurs upon GEM treatment alone or in combination with β-CD, suggesting that β-CD does not directly activate the STING pathway. Previous studies have demonstrated that chemotherapeutic agents such as GEM can induce DNA double-strand breaks (DSBs), thereby activating the cGAS-STING pathway [31]. As a pivotal molecular switch of innate immunity, cGAS-STING recognizes cytosolic DNA and triggers downstream IRF3-type I interferon (IFN-I) and NF-κB signal, which promote antigen cross-presentation by dendritic cells (DCs) to CD8⁺ T cells and enhance T-cell infiltration. Moreover, GEM treatment alone upregulated PD-L1 expression on the cell membrane, whereas the RMM@GEM NPs group showed decreased PD-L1 levels, indicating that the addition of β-CD markedly suppressed PD-L1 expression (Fig. 5B and C). Mechanistically, β-CD has been reported to suppress PD-L1 expression on tumor cell membranes by promoting cholesterol efflux, thereby counteracting the PD-L1 upregulation induced by STING activation during antitumor immune responses (Fig. 5G) [31, 32]. Flow cytometry and confocal microscopy further confirmed that RMM@GEM NPs significantly reduced the intracellular cholesterol levels in GL261 cells, attributed to β-cyclodextrin release from the nanoformulation (Figure S21, S22, Supporting Information), which facilitated cholesterol efflux (Fig. 5D). Furthermore, the immune regulation of RMM@GEM NPs was further assessed in a glioma mice model. Immunohistochemical staining of brain tumor tissues revealed that RMM@GEM NPs treatment significantly increased p-STING expression while reducing PD-L1 expression to the lowest level among all groups compared with the saline group (Fig. 5F), indicating that GEM and the released β-CD synergistically activated the STING pathway and suppressed PD-L1 expression.

Fig. 5.

Regulation of the cGAS-STING pathway and PD-L1 expression by RMM@GEM NPs. (A) Western blot analysis of cGAS-STING pathway-related proteins in GL261 cells treated with GEM, β-CD, or a combination of GEM and β-CD. (B) CLSM images of BODIPY and PD-L1 in GL261 cells under different treatments (scale bar = 20 μm). (C) Western blot analysis of PD-L1 protein in GL261 cells treated with PBS, Free GEM, GEM NPs, MM@GEM NPs, or RMM@GEM NPs. (D) CLSM images of BODIPY in GL261 cells under different treatments (scale bar = 20 μm) (n = 3). (E) Schematic diagram of mechanism. (F). Immunohistochemical staining of tumor paraffin sections for p-STING and PD-L1 after 21 days of treatment in mice (scale bar = 100 μm). (G) Semi-quantitative analysis of stained tumor tissues. (H-L) ELISA analysis of IFN-γ, IFN-1β, IL-6, IL-10, and TNF-α expression levels in tumor tissues from the different treatment groups (n = 3). Data are presented as mean±SD.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 were considered statistically significant

Furthermore, ELISA analysis demonstrated that, different from the free GEM limited by its poor BBB penetration, RMM@GEM NPs treatment substantially enhanced IFN-1β and IL-6 secretion in tumor tissues, with the elevated expression levels up to 2.19- and 1.44-fold compared with controls (Fig. 5I, J). This effect should be attributed to phosphorylation and nuclear translocation of IRF3, which promoted IFN-1β expression, and STING-mediated NF-κB activation, which enhanced IL-6 and other pro-inflammatory cytokines. Conversely, IL-10, an anti-inflammatory cytokine, was significantly reduced in the RMM@GEM NPs group, decreasing to 0.68-fold compared with the saline group (Fig. 5K). In addition, RMM@GEM NPs promoted immune cell-mediated secretion of TNF-α and IFN-γ in the tumor microenvironment, with expression levels elevated to 1.76- and 1.93-fold, respectively (Fig. 5H, L), ELISA analysis of tumor tissues treated with the different formulations showed a significant increase in ATP release and HMGB1 ratio, especially in the RMM@GEM NPs group, thereby inducing immunogenic cell death of tumor cells (Figure S23, Supporting Information). The results from Figure S24 further supported these findings, where Western blot analysis of cGAS-STING pathway-related proteins in tumor cell lysates after treatment with different formulations showed consistent activation of the STING pathway across the groups. Moreover, resulting from the target functional RAP peptide mediated tumor cell endocytosis, the markers of the cGAS-STING pathway in tumor tissues were further significantly enhanced. In summary, the local release of GEM and β-CD within tumor cells directly kills tumor cells while simultaneously activating the STING pathway to enhance antitumor immune responses. Meanwhile, β-CD reduces the intracellular cholesterol via efflux, leading to downregulation of PD-L1 expression on tumor cell membranes. Together, these effects synergistically improve the therapeutic efficacy of GEM against glioma.

In vivoTreatmentand safety

To evaluate the intracranial antitumor efficacy of RMM@GEM NPs, an orthotopic GL261-Luc glioblastoma model was established in C57BL/6 mice. The successful tumor implantation was verified on day 7 post-inoculation using an in vivo imaging system (IVIS) spectrum. The mice were subsequently randomized into 5 groups, including Saline, Free GEM, GEM NPs, MM@GEM NPs, and RMM@GEM NPs, and treated with the corresponding formulations via tail vein injection every three days starting on day 8. The antitumor effects of the nanoformulation were assessed on days 11, 14, 18, and 21 by the intraperitoneal injection of D-luciferin followed by bioluminescence imaging (Fig. 6A). As the results shown in Fig. 6B, the survival of mice in the Saline group did not exceed 20 days, while bioluminescence signals increased significantly in the Free GEM and GEM NPs groups. The MM@GEM NPs group exhibited only a moderate increase in bioluminescence because of the limited targeting efficiency. In contrast, the RMM@GEM NPs group showed the lowest signal intensity. The tumor burden accounted for 3.53% of the saline group and 38.22% of the free GEM group. Respectively, demonstrating that the incorporation of targeting peptides markedly enhanced the tumor targeting and effectively suppressed GL261-Luc tumor growth (Fig. 6C, E). The survival analysis further confirmed that RMM@GEM NPs significantly extended lifespan, with a median survival time notably longer than that of the Saline (14 days), Free GEM (26 days), GEM NPs (34 days), and MM@GEM NPs (38 days) groups (Fig. 6B). Additionally, while tumor burden led to the progressive weight loss in most groups, mice treated with RMM@GEM NPs exhibited relatively stable body weight at later stages (Fig. 6D).

Fig. 6.

In vivo therapeutic efficacy of RMM@GEM NPs in an orthotopic GL261-Luc glioblastoma mice model. (A) Schematic illustration of the experimental protocol for glioblastoma treatment. (B) Survival curves of mice under different treatments (n = 10). (C) Representative bioluminescence images of mice bearing intracranial luciferase-labeled tumors treated with the different treatments at day 7, 11, 14, 18, and 21 (n = 6). (D) Changes in body weight of mice following the different treatments. (E) Quantification of total photon flux from intracranial tumor sites corresponding to (C). (F) H&E-stained images and semi-quantitative analysis of brain tumor sections from mice on day 21 after various treatments (scale bar = 1 mm), along with immunohistochemical staining and semi-quantitative analysis of GFAP, TUNEL, and Ki-67 (scale bar = 100 μm). (G) Semi-quantitative analysis of GFAP, TUNEL, and Ki-67 staining in tumor tissues. Data are presented as mean±SD. Statistical significance compared with the saline group is indicated (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001)

At the experimental endpoint on day 22, the mice were euthanized, and whole-brain tissues were collected for histological analysis using H&E, GFAP (glial fibrillary acidic protein), TUNEL (apoptosis marker), and Ki-67 (proliferation marker) staining [33]. The RMM@GEM NPs group exhibited the most pronounced antitumor effects, including a substantial reduction in tumor lesion size (Fig. 6F) and significantly improved the histopathological features, consistent with dynamic in vivo bioluminescence monitoring. By contrast, tumors in the Saline and Free GEM groups greatly preserved their architecture, displaying prominent multinucleation, giant cells, hyperchromatic nuclei, and pathological mitoses, indicating the limitations of the untreated or monotherapy approaches. GFAP and TUNEL staining revealed the pronounced necrosis and fibrosis at various stages, along with the extensive apoptosis and necrosis in the RMM@GEM NPs group, with the highest proportions observed among all groups (GFAP: 6.8%, TUNEL: 3.1%) (Fig. 6F-G). Immunohistochemical analysis of Ki-67 further demonstrated that tumor cell proliferation was markedly inhibited in the RMM@GEM NPs group, with a Ki-67 positivity rate of 8.5%, significantly lower than those in the Saline (25.1%), Free GEM (18.9%), GEM NPs (13.7%), and MM@GEM NPs (9.3%) groups (Fig. 6F-G). Collectively, these results indicate that RMM@GEM NPs can effectively inhibit tumor growth and prolong the survival in glioblastoma-bearing mice. This chemo-immunotherapeutic synergistic therapeutic efficacy could be attributed to the efficient BBB penetration of the delivery system, enabling the tumor-targeting drug accumulation at the tumor lesion, subsequently improving antitumor immune responses within the tumor microenvironment via the activated STING and the inhibited PD-L1 expression.

To further evaluate the biocompatibility and safety of RMM@GEM NPs, histopathological analysis of major organs (heart, liver, spleen, lung, and kidney) using H&E staining showed no significant pathological alterations or toxic damage in any treatment group (Figure S25, Supporting Information). Moreover, the complete blood count analysis further confirmed that key haematological parameters, including red blood cell count, white blood cell count, platelet count, and haemoglobin levels, remained within the normal physiological range, with no statistically significant differences compared with the control group (Figure S26, Supporting Information). Serum biochemical analyses showed that the levels of ALT, AST, and ALP in all treatment groups remained within the normal physiological range, with no significant differences compared to the saline control. These results indicate that the administration of Free NPs, GEM NPs, MM@GEM NPs, and RMM@GEM NPs did not cause observable hepatic dysfunction, suggesting normal liver function was maintained during treatment (Figure S27, Supporting Information). Collectively, these findings indicate that RMM@GEM NPs exhibit an excellent biosafety, supporting their potential as promising candidates for glioma therapy.

RMM@GEM NPs remodel the glioblastoma immune microenvironment

Glioblastoma, characterized by the low immunogenicity and a profoundly immunosuppressive microenvironment [34, 35], remains a significant challenge for immunotherapy. To investigate the specific effects of RMM@GEM NPs treatment on the tumor immune microenvironment (TME), a glioblastoma mice model was established and administered the corresponding treatments (Fig. 7A). Following the intravenous administration of the different formulations, mice were sacrificed on day 14 post-treatment. Tumor tissues were harvested, dissociated into single-cell suspensions, and analyzed using CyTOF technology for high-dimensional immune profiling (Figure S28, Supporting Information) [36]. Marker-based annotation identified several key immune cell subsets, including neutrophils, dendritic cells (DCs), NK cells, CD4⁺ T cells, CD8⁺ T cells, and monocytes (Fig. 7C). t-SNE-based dimensionality reduction further revealed the distinct spatial distribution patterns of these immune populations across the different treatment groups (Fig. 7D) (Figures S29, S30, Supporting Information). The results demonstrated that RMM@GEM NPs treatment significantly remodeled the TME. Within T-cell populations, the RMM@GEM NPs group showed much higher infiltration of CD4⁺ and CD8⁺ T cells compared with the saline control (Fig. 7E-F), suggesting the enhanced adaptive immune responses. Although free GEM could activate the STING pathway to promote T-cell enrichment, it was simultaneously associated with elevated PD-L1 expression, leading to T-cell exhaustion [32]. Notably, compared with free GEM, the RMM@GEM NPs, MM@GEM NP, and GEM NP groups exhibited a significant reduction in CD8⁺ T_ex cells (Fig. 7I), consistent with the findings in Fig. 5C. This improvement in T-cell exhaustion can be attributed to the introduction of β-CD. In terms of the innate immunity, RMM@GEM NPs group exhibited significantly higher levels of mature DC and NK cell infiltration (Fig. 7G-H). The increase in mature DCs suggests that nanoformulation-induced immunogenic cell death (ICD) promotes antigen presentation and activates downstream T-cell immune response [37, 38]. Interestingly, the proportion of regulatory T cells (Tregs, CD4⁺, CD25⁺, CD127⁻) was found to be the lowest in the RMM@GEM NPs group. Since Tregs are known to exert immunosuppressive functions and contribute to tumor immune evasion, the reduced frequency of Tregs suggests that RMM@GEM NPs may attenuate immunosuppression within the tumor microenvironment and thereby facilitate a more effective antitumor immune response (Figure S31, Supporting Information).

Fig. 7.

Regulation of the glioblastoma immune microenvironment by RMM@GEM NPs. (A) Schematic illustration of the experimental procedure. (B) Proportional distribution of immune cell clusters, including neutrophils, M-MDSCs, PMN-MDSCs, CD4⁺ T cells, CD8⁺ T cells, NK cells, monocytes, and B cells. (C) Heatmap showing the relative expression of lineage markers in tumor single-cell suspensions. (D) t-SNE plots revealing the spatial distribution of immune cells within the tumor microenvironment of glioma mice models. (E-I) The distribution and abundance of CD4⁺ T cells, CD8⁺ T cells, DCs, NK cells, and CD8⁺ T memory cells under the different treatment conditions, with quantitative statistical analyses presented in the box plots on the right. Data are presented as mean±SD.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 were considered statistically significant

Meanwhile, the elevated NK cell frequency was closely correlated with the enhanced IFN-γ secretion, which suppresses Treg expansion [39], thereby further reinforcing the immune-activated state of the TME. Collectively, these findings indicate that RMM@GEM NPs not only enhanced the infiltration of effector immune cells (CD4⁺ T, CD8⁺ T, NK cells, mature DCs, and M1-type macrophages) but also effectively suppressed the accumulation of immunosuppressive cells (CD8⁺ T_ex and Tregs). This bidirectional regulation remodeled the TME, strengthened antitumor immune responses, and alleviated the common challenge of T-cell exhaustion in immunotherapy (Fig. 8).

Fig. 8.

Schematic of RMM@GEM NPs chemotherapy combined with immunotherapy for synergistic treatment of glioma

Conclusions

This study designed a dual-responsive biomimetic drug delivery system, RMM@GEM NPs, incorporating β-cyclodextrin and gemcitabine for positively regulating the tumor immune microenvironment, aiming to enhance the therapeutic efficacy against glioblastoma (GBM) [40–42]. In this system, GEM was conjugated to β-CD via a reactive oxygen species (ROS)-responsive linker to form a prodrug. Meanwhile [43, 44], a pH-sensitive PSP directed the right orientation of glioblastoma cell membranes, spontaneously coating them. Moreover, the target agent, DSPE-PEG-RAP, could also improve the target cargo delivery for promoting the precise accumulation at glioblastoma tumor lesions [45]. Moreover, the dual pH and ROS responsiveness of this system promoted the enhanced GEM and β-CD release in the pathological local. Given the synergetic therapeutic efficacy for β-CD-mediated cholesterol-depleting functions, RMM@GEM NPs could not only suppress the STING pathway-mediated PD-L1 upregulation but also reverse T-cell exhaustion, ultimately strengthening antitumor immune responses. However, this study is defect of the long-term toxicity and immunogenicity assessments, which could be requisite in further clinic application investigation. The comparison in Figure S32 highlights the uniqueness of the tumor membrane biomimetic GEM-NP approach, particularly its combination of chemo-immunotherapy and STING pathway activation. Different from other strategies that primarily focus on delivery or immune modulation, this study provides a comprehensive platform that integrates the target drug delivery with immune system activation, offering a novel synergistic approach for glioma treatment [46–51].

Materials and methods

Materials

Mono-(6-amino-6-deoxy)-β-cyclodextrin (NH₂-β-CD), Thiodiglycolic anhydride (HS), Gemcitabine (GEM), N-Hydroxysuccinimide (NHS), and 4-Dimethylaminopyridine (DMAP) were purchased from Aladdin (Shanghai, China). n-(3-dimethylaminopropyl)-n’-ethylcarbodiimide hydrochloride (EDC), 5-Benzimidazolecarboxylic acid, Mal-PEG-NH₂, DSPE-PEG-NHS, Alexa Fluor 488, and Rhodamine B (RB) dyes were obtained from Macklin (Shanghai, China). Rap-12 targeting peptide (EAKIEKHNHYQK) and PSP targeting peptide (CLIKKPF) were purchased from Nanjing Taiye Co., Ltd. (Nanjing, China). Annexin V-FITC/PI apoptosis detection kit, Calcein-AM/PI live/dead cell double-staining kit, Crystal Violet Methanolic Solution (2.5%), and CCK8 kit were purchased from Solarbio (Beijing, China). Cell culture was performed using high-glucose DMEM, fetal bovine serum (FBS), penicillin-streptomycin, and trypsin, all of which were obtained from Gibco (Thermo Fisher Scientific, USA).

Methods

Synthesis of ROS- and pH-responsive nano-prodrug (GEM-CD@BM-PEG-PS)

GEM grafted onto NH₂-β-CD was obtained through a two-step reaction process. First, GEM (26 mg, 0.1 mmol) and HS (13 mg, 0.1 mmol) were dissolved separately in 3 mL of DMSO, followed by the addition of DMAP (0.1 mmol) as a catalyst. After heating to 40 °C and stirring continuously for 6 h, EDC (19 mg) and NHS (11 mg) were added to the reaction solution and stirred magnetically for an additional 6 h. This was followed by the addition of NH₂-β-CD (14 mg, 1.2 mmol) and continued magnetic stirring overnight. Upon completion of the reaction, the product was dialyzed with deionized water (MWCO: 1500 Da) to remove the unreacted materials and DMSO. Finally, the dialyzed sample was lyophilized for 36 h to obtain GEM-CD (Figure S1, Supporting Information). Similarly, RB was grafted onto NH₂-β-CD through the same reaction to yield RB-CD. BM (16 mg, one mmol), EDC (19 mg), and NHS (11 mg) were dissolved in 10 mL of DMSO [41, 42], stirred at 40 °C for 6 h, followed by the addition of NH₂-PEG-Mal (580 mg) (Figure S2, Supporting Information), and the reaction was stirred overnight. PSP (34 mg) was then introduced into the reaction solution and allowed to react at 40 °C for an additional 12 h. The resulting product was purified by dialysis with deionized water (MWCO: 6000 Da). Finally, the dialyzed sample was lyophilized for 36 h to obtain BM-PEG-PSP. GEM-CD was added to BM-PEG-PSP (160 mg, 0.2 mmol) dissolved in 5 mL of DMSO and stirred overnight [44]. The solution was then injected into 10 mL of Dulbecco’s phosphate-buffered saline (DPBS, pH 7.4) under sonication to promote self-assembly. Before subsequent use, the mixture was dialyzed against 1 L of DPBS (MWCO: 8000 Da) for 24 h, with fresh DPBS added every 3 h to remove DMSO and the unreacted residues. The final solid nano-prodrug GEM-CD@BM-PEG-PSP was obtained by lyophilization. Similarly, RB-CD@BM-PEG-PSP (GEM NPs) was obtained by replacing GEM with RB during the self-assembly process.

Synthesis of DSPE-PEG-RAP

Mal-PEG₂₀₀₀-DSPE (250 mg) and the RAP (60 mg, 1.2 mmol) were separately dissolved in DMSO and DPNS buffer (0.2 M, pH 7.4), followed by stirring overnight at 45 °C. After the reaction was completed, the unreacted residues were removed by dialysis (MWCO: 3500 Da). The product DSPE-PEG-RAP was obtained by lyophilization and stored at −20 °C.

Preparation of biomimetic nanoformulation (RMM@GEM NPs)

GL261 cell membranes (MM) were extracted following previously reported procedures [45]. Freshly prepared DSPE-PEG-RAP and MM were mixed at a 1:6 (w/w) ratio and subjected to sonication for 5 min to obtain RMM. The fresh prepared RMM was co-incubated with RB-CD@BM-PEG-PSP for 3 h, during which the PSP specifically recognized and bound to the inner leaflet phosphatidylserine on the tumor cell membrane, yielding the final product RMM@GEM NPs.

Characterization of RMM@GEM NPs

For FTIR characterization, more than 1 mg of GEM-CD, BM-PEG-PSP, and GEM NPs powder were prepared using the KBr pellet method, and infrared spectra were recorded in the wavenumber range of 4000–400 cm⁻¹ [45]. For ¹H NMR characterization, 5 mg of GEM-CD, BM-PEG-PSP, and GEM NPs were dissolved in DMSO-d₆, respectively. The Structural characterization was performed using ¹H NMR spectroscopy, and two-dimensional NMR spectra were also acquired. The hydrodynamic diameter (Dh) and zeta potential of GEM-CD, BM-PEG-PSP, and GEM NPs were measured using a Malvern Zetasizer Nano ZS instrument (Nano ZS 90, Malvern, UK) equipped with a He-Ne laser (λ = 633 nm) at 37 °C with a scattering angle of 90°. The stability of GEM NPs was evaluated by monitoring particle size changes over 7 days at room temperature [25]. GL261 cell membranes were extracted using a membrane protein extraction kit, and glioma, MM@GEM NPs, and RMM@GEM NPs were analyzed by SDS-PAGE. Western blotting (WB) was performed to detect the expression of Integrin alpha V, CD44, CD47, and EpCAM. Transmission electron microscopy (TEM) characterization: suspensions of GEM NPs and RMM@GEM NPs (100 µg mL⁻¹) were dropped onto 200-mesh carbon-coated copper grids and allowed to stand for 2 min, followed by negative staining with 2% (w/v) phosphotungstic acid. Samples were observed using a JEM-2100 F field-emission TEM (JEOL, Japan) operated at 200 kV to obtain high-resolution images.

Drug loading and in vitro release study

The drug release behavior of RMM@GEM NPs was evaluated using the dialysis method. Lyophilized RMM@GEM NPs were reconstituted into a 1 mg/mL solution and transferred into disposable dialysis bags (MWCO: 1000 Da). The dialysis bags were separately immersed in the following PBS solutions: (1) 0.1 mmol PBS, pH 7.4; (2) 1 mmol PBS, pH 7.4; (3) 0.1 mmol PBS, pH 5.5; and (4) 1 mmol PBS, pH 5.5. At predetermined time intervals, 1 mL of release medium was collected, and the absorbance at 275 nm was measured using a UV-Vis spectrophotometer (DU730, Beckman Coulter) to quantify the release of GEM.

Cytotoxicity evaluation

GL261, U251, U87, bEnd. 3, and BV2 cells were seeded into 96-well plates and cultured overnight. The medium was then replaced with serum-free medium, followed by treatment with Free GEM, GEM NPs, MM@GEM NPs, or RMM@GEM NPs for 24 h, and cell viability was subsequently assessed using the CCK-8 assay.

Evaluation of cell migration ability

A total of 5 × 10⁴ suspended GL261 cells were seeded into the upper chamber of a 24-well Transwell insert with an 8 μm pore polycarbonate membrane, with serum-free medium in the top chamber and complete medium containing 10% FBS in the bottom chamber. Following treatment with the respective formulations for 6、12 and 24 h, the cells were further incubated at 37 °C for an additional 48 h. The cells that migrated to the bottom chamber were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and subsequently imaged under a microscope.

Apoptosis assay

To further evaluate the effect of nanoformulation on cell apoptosis, GL261 cells (1 × 10⁵) were seeded into 12-well plates and incubated with different drug formulations for 12 h. The cells were then harvested, stained with Annexin V-FITC/PI according to the manufacturer’s protocol, and analyzed using a flow cytometer (CytoFlex LX, Beckman) [45].

Plate colony formation assay

GL261 cells in the logarithmic growth phase were seeded into 6-well plates at a density of 1,500 cells per well. On the following day, the medium was replaced with complete culture medium, and the cells were incubated at 37 °C in a humidified incubator with 5% CO₂ for 2 days. Subsequently, drug treatments were applied, and the cultures were maintained for 14 days. When colonies containing more than 60 cells became visible to the naked eye, the cells were washed twice with pre-cooled PBS, fixed with 4% paraformaldehyde at room temperature for 15 min, stained with crystal violet, air-dried, and photographed.

Establishment of an in vitro BBB transwell model

Matrigel was diluted with culture medium at a ratio of 8:1 (v/v), and 100 µL of the diluted solution was evenly coated onto the basement membrane of the top chamber of a 0.4 μm Transwell insert. In comparison, 600 µL of DMEM was added to the bottom chamber to balance the hydrostatic pressure between the two chambers. After the Matrigel solidified, 1 × 10⁵ cells were seeded into the top chamber. The transendothelial electrical resistance (TEER) of the cells in the top chamber was monitored daily. When the TEER value exceeded 300 Ω [52], the model was considered suitable for subsequent in vitro BBB permeability experiments.

Establishment and treatment of an orthotopic tumor implantation model

All animal experiments were conducted in accordance with the animal care and use guidelines of the Chongqing University Cancer Hospital Animal Welfare Ethics Committee (CQCH-LAE-A0000202046).

To establish mice bearing GL261-Luc orthotopic tumors, 6-week-old male C57BL/6 mice were injected with 1.5 × 10⁵ GL261-Luc cells into the caudate nucleus (1.8 mm anterior to the bregma, 0.6 mm lateral to the right of the sagittal suture, and 2 mm below the dura), with 10 mice assigned to each group(Shanghai Yuyan Instruments Co.,Ltd，SA-150). After tumor implantation, the mice were randomly divided into 5 groups and treated with Saline, Free GEM, GEM NPs, MM@GEM NPs, or RMM@GEM NPs (equivalent GEM dose: 10 mg/kg, administered via tail vein injection every 3 days). Tumor growth was monitored using an in vivo imaging system (Lumina IVIS III, D-luciferin bioluminescence). At the end of treatment, 6 mice from each group were randomly selected for euthanasia, and brain tissues were harvested to prepare paraffin and frozen sections for hematoxylin and eosin (H&E) staining, immunofluorescence staining, and immunohistochemistry.

Measurement of tumor cytokine expression

According to the manufacturer’s instructions, ELISA kits were used to determine the levels of inflammatory cytokines IL-6 (ThermoFisher, MAN0017508), IFN-γ (ThermoFisher, KMC4021), IFN-1β (ThermoFisher, MAN0017504), IL-10 (ThermoFisher, MAN0017316), and TNF-α (ThermoFisher, MAN0017423) in the brains of tumor-bearing mice established via orthotopic injection. Following euthanasia, brain tissue samples were immediately collected from the mice. The tissues were mixed with pre-cooled homogenization medium (0.86%−0.9% saline recommended) at a ratio of 1:9 (w/v) and thoroughly homogenized on ice using a mechanical homogenizer to prepare a 10% tissue homogenate. The homogenates were centrifuged at 2500–3000 rpm for 10 min, and the supernatants were carefully collected for subsequent ELISA assays. All procedures were conducted at 4 °C to ensure sample stability.

In vivo biodistribution and Long-Circulation monitoring

A total of 100 µL of RB or RB-equivalent labeled GEM, MM@GEM NPs, and RMM@GEM NPs was intravenously injected via the tail vein into mice bearing orthotopic gliomas. In vivo biodistribution was monitored using a small-animal imaging system at 6, 12, 24, and 36 h post-injection (excitation wavelength: 550 nm, emission wavelength: 580 nm). After 36 h post-injection, mice were euthanized, and brain tissues, along with major organs, were harvested. Ex vivo imaging was subsequently performed using the An in vivo imaging system (IVIS) spectrum to compare RB distribution across the different tissues.

In vivo blood circulation measurement

To investigate the blood circulation profile of the nanoformulation, 100 µL of RB, RB-labeled GEM, MM@GEM NPs, or RMM@GEM NPs was intravenously injected via the tail vein of mice. Blood samples were collected from the tail tip at 0.5, 1, 2, 4, 8, 12, and 24 h post-injection, diluted with 50 µL of PBS, and transferred into 96-well plates [53]. Fluorescence intensity was measured using a microplate reader and An in vivo imaging system (IVIS) spectrum.

Western blotting

For the evaluation of protein expression in vitro, 2 × 10⁶ GL261 cells were seeded into each well of a 6-well plate. Four groups were established: GEM treatment alone, β-CD treatment alone, GEM combined with β-CD, and PBS control. After treatment for 24 h, cells were harvested, lysed, and the proteins were quantified. Proteins were subjected to SDS-PAGE, transferred onto PVDF membranes, blocked, and incubated with primary and secondary antibodies, followed by detection using chemiluminescent substrate. A complete list of antibodies used is provided in the supplementary table.

Immunofluorescence staining

To investigate intracellular protein expression, drug-treated cells were washed three times with 1 × PBS to remove the residual culture medium and then fixed with 4% PFA at 37 °C for 15 min to preserve proteins and cellular structures. To block the non-specific binding, cells were incubated with BSA for 30 min. Cells were then incubated with primary antibodies at 4 °C for 12 h, followed by thorough washing and incubation with secondary antibodies at room temperature. After removing the secondary antibody, cells were counterstained with DAPI in the dark for 10 min and imaged using a confocal microscope (ZEISS LSM 900). The primary antibodies used are listed in the supplementary table.

In vivo biosafety evaluation

Six-week-old male C57BL/6 mice were randomly divided into 5 groups with 3 mice for each group and intravenously injected with Free GEM, GEM NPs, MM@GEM NPs, or RMM@GEM NPs (GEM equivalent: 10 mg/kg), with saline as the control. After treatment for 22 days, the mice were sacrificed, and major organs (heart, liver, spleen, lung, and kidney) were collected, fixed in 4% paraformaldehyde, embedded in paraffin, and subjected to H&E staining for histopathological analysis. EDTA-anticoagulated whole blood was simultaneously collected for hematological analysis, serum biochemical assays, and in vitro hemolysis evaluation.

Mass cytometry sample Preparation

Male C57BL/6 mice bearing gliomas were established and divided into 4 treatment groups and one control group, with 3 mice in each group. After treatment for 14 days, tumor tissues were isolated from the mice and processed using a tumor tissue dissociation kit (Mogengel Biotechnology, MB-0818L05S) according to the manufacturer’s instructions [53]. Tumor tissues were cut into 1–3 mm³ fragments using a scalpel, added to a pre-prepared digestion solution, and digested at 37 °C in a shaking incubator for 2 h. The supernatant was collected, digestion was terminated with FBS, and a single-cell suspension was obtained by passing the mixture through a 70 μm cell strainer. Red blood cells were lysed with ACK lysis buffer, and after washing, cell number and viability were assessed using the Muse Cell Analyzer (Merck) with the Muse Count & Viability Kit (Merck). Finally, cells were subjected to cisplatin fixation, FcR-blocking, membrane protein antibody incubation, intracellular protein staining, nuclear intercalator Ir staining, and data acquisition by mass cytometry.

CyTOF data analysis

After normalization of the FCS files, debris, dead cells, and doublets were excluded from the manually gated data using FCS Express and RStudio software. Clustering algorithms were applied to classify cells into distinct phenotypes based on expression levels, and heatmaps were generated [54–56]. The dimensionality reduction algorithm, t-distributed stochastic neighbor embedding (t-SNE), was employed to visualize high-dimensional data in two dimensions, thereby revealing the distribution of each cluster and marker expression, as well as differences among groups or across different sample types [56].

Data expression and statistical analysis

All data were expressed as mean ± standard deviation (SD), and statistical analysis was performed using one-way analysis of variance (ANOVA).

Author contributions

CRediT StatementYunfan li: Methodology, Formal analysis, Investigation, Writing original draft. Kaiwen Bao: Conceptualization, Writing - Review & Editing, Supervision.Renzheng Huan: Methodology, Investigation, Validation.Tian Wang: Methodology, Investigation, Validation.Ya Wang: Methodology, Investigation, Data curation.Shuai Wu: Methodology, Investigation, Resources. Xing Cheng: Methodology, Investigation, Data curation.Jiashang Huang: Methodology, Investigation, Data curation.Li Zhu: Methodology, Investigation, Data curation.Jianshu Li: Conceptualization, Writing - Review & Editing, Supervision, Funding acquisition.Haifeng Yang: Conceptualization, Writing - Review & Editing, Supervision, Funding acquisitionWei Wu: Conceptualization, Writing - Review & Editing, Supervision, Projectadministration, Funding acquisition.

Funding

This work was supported by grants from the Fundamental Research Funds for the National Key R&D Project (2022YFF0710700), Chongqing medical scientific research project (Joint project of Chongqing Health Commission and Science and Technology Bureau) (2024ZDXM006), JinFeng Laboratory of Chongqing (JFLKYXM202303AZ-204), Natural Science Foundation of Chongqing (CSTB2024NSCQ-MSX0137, cstc2021jcyj-cxttX0002), Fundamental Research Funds for Central Universities (2024CDJXY017). In addition, we would like to thank Mr. Zhang Bin at the Analytical and Testing Center of Chongqing University for their assistance with TEM.

Data availability

Data available on request from the authors.

Declarations

Competing interests

The authors declare no competing interests.