Reprogramming PD-L1 endosomal recycling and lysosomal degradation via engineered exosome-based nanocomposites overcomes immune resistance in lung cancer brain metastasis Authors: Xiujuan Hong, Xiaoqi Wang, Wankun Wang, Weiye Ren, Wenjia Ding, Dashan Yin, Chenli Wang, Chuan Sun, Jun Chen, Ji-Gang Piao, Zhibing Wu Published: 2026 ABSTRACT While anti-PD-L1 antibody (αPD-L1) therapy holds promise, its efficacy against lung cancer brain metastasis (LCBM) is severely limited by the blood-brain barrier (BBB), the immunosuppressive tumor microenvironment and adaptive immune resistance. To overcome these barriers, we engineered a pH-responsive nanocomposite (VP-αPD-L1@REB) by functionalizing brain-metastatic tumor cell-derived exosomes (EB) with RGD peptides (REB) for targeted co-delivery of verteporfin (VP) and αPD-L1. Benefiting from homotypic affinity and integrin-mediated transcytosis, VP-αPD-L1@REB efficiently crosses the BBB, accumulates within intracranial tumors, and undergoes pH-responsive cargo release. Mechanistically, VP induces a lethal reactive oxygen species (ROS) storm for direct tumor ablation. Simultaneously, VP downregulates the chaperone protein CMTM6 and activates cellular autophagy, forcibly driving internalized PD-L1 toward degradation via dual “endosome-lysosome” and “autophagy-lysosome” pathways. Driven by the synergy of VP's robust intracellular clearance and αPD-L1's surface blockade, this targeted nanoplatform successfully remodels the intracranial immunosuppressive microenvironment and triggers potent systemic anti-tumor immunity. This study provides a highly promising translational paradigm for overcoming adaptive immune resistance in central nervous system (CNS) malignancies. Introduction Lung cancer is one of the most common malignancies worldwide in terms of both occurrence rate and mortality [1,2]. Among its most severe complications, brain metastases occur in 30%-50% of patients, a diagnosis that portends a dismal prognosis with a median overall survival of less than 12 months [3]. While targeted therapies benefit the patients with lung cancer brain metastasis (LCBM) harboring driver gene mutations [4,5], they remain largely ineffective for the substantial subset (40%-50%) of patients who are driver-gene negative [6,7]. Immune checkpoint inhibition (ICI) has emerged as a promising therapeutic option for this challenging population [8]. Although monoclonal antibody therapies targeting programmed cell death ligand 1 (PD-L1) have revolutionized oncological care, their efficacy against central nervous system (CNS) metastases remains significantly limited. Clinical evidence indicates an intracranial objective response rate (icORR) of only 10%–15%, with minimal improvement in progression-free survival and overall survival [9]. This suboptimal intracranial efficacy is primarily hindered by poor blood-brain barrier (BBB) penetration of systemic antibodies [10,11], the immunosuppressive tumor microenvironment [12,13] and adaptive immune resistance driven by PD-L1 endosomal recycling [14,15]. Overcoming this tripartite physiological and immunological barrier is essential to improving the comprehensive therapeutic outcome for patients with driver gene-negative LCBM [[16], [17], [18]]. A significant impediment to single-agent anti-PD-L1 antibody (αPD-L1) therapy is adaptive immune resistance, which is predominantly driven by the continuous endosomal recycling of PD-L1. This dynamic process is orchestrated by the chaperone protein CMTM6 [15,19]. By protecting internalized PD-L1 from degradation, CMTM6 facilitates its redistribution to the cell membrane, thereby restoring the tumor's immune evasion capacity and contributing to resistance against ICI therapy. Conversely, depletion of CMTM6 effectively directs PD-L1 to the lysosome for permanent degradation [20]. Therefore, targeting the lysosomal degradation axis of PD-L1 to achieve permanent clearance of this protein, rather than relying solely on transient blockade of its surface expression, represents a highly promising strategy for overcoming ICI resistance. Verteporfin (VP), classically identified as a YAP inhibitor, has recently emerged as a pleiotropic immunomodulator capable of disrupting these intracellular trafficking networks. VP robustly suppresses PD-L1 transcription while activating autophagic pathways [21,22], forcibly redirecting PD-L1 toward the autophagy-lysosome degradation axis [23]. Furthermore, independent of photo-activation, the intracellular accumulation of VP induces intense oxidative stress and a burst of reactive oxygen species (ROS). This lethal oxidative damage not only triggers direct cytolysis but also promotes the release of tumor-associated antigens and damage-associated molecular patterns (DAMPs) [[24], [25], [26]]. Ultimately, this mechanism bridges local cytotoxicity with robust systemic anti-tumor immunity by remodeling the immunosuppressive microenvironment [27]. Therefore, combining VP with ICI therapy, potentially through a dual synergistic mechanism of downregulating PD-L1 and comprehensively activating the immune microenvironment, represents a highly promising strategy for improving treatment outcomes in patients with driver gene-negative LCBM. However, realizing this synergistic potential in vivo requires bridging the formidable BBB. To overcome this critical delivery hurdle, natural extracellular vesicles, exosomes, have emerged as an ideal biomimetic nanocarrier owing to their high biocompatibility, low immunogenicity, and inherent capacity to traverse the BBB [28,29]. In particular, exosomes derived from homologous tumor cell lines inherit the complex lipidomic and proteomic profiles of their parental lung cancer cells, thereby exhibiting an inherent homing capability [30,31]. Furthermore, modification of these biomembranes with integrin-binding RGD (arginine-glycine-aspartic acid) peptides introduces an active targeting ability by specifically recognizing and binding to αvβ3 integrins, which are aberrantly overexpressed on both the tumor neovasculature and cancer cells [32], thereby optimizing initial transport across the BBB and subsequent binding affinity to malignant cells within the CNS [33,34]. Thus, constructing an RGD-modified, exosome-based targeted delivery platform represents an ideal strategy to translate this multi-target mechanism into profound intracranial efficacy. In this study, we developed VP-αPD-L1@REB, an integrin-targeted biomimetic nanoplatform that efficiently traverses the blood-brain barrier and executes pH-responsive cargo release within intracranial lesions. Mechanistically, beyond its known YAP-inhibitory role, we uncovered that VP orchestrates a novel dual-clearance axis for PD-L1. VP forcefully abrogates CMTM6-mediated endosomal recycling while triggering autophagy, thereby driving PD-L1 toward permanent degradation via endosomal and autophagic lysosomal pathways. Furthermore, VP amplifies this clearance by upregulating lysosomal biogenesis. By synergizing VP's deep intracellular PD-L1 exhaustion with αPD-L1's surface blockade, our nanocomposite successfully remodels the immunosuppressive tumor microenvironment and activates systemic peripheral immunity (Scheme 1). This substantially sensitizes LCBM to immunotherapy, providing a highly promising translational paradigm against adaptive immune resistance.Scheme 1Schematic illustration of targeted co-delivery via VP-αPD-L1@REB for synergistic LCBM immunotherapy through driving PD-L1 dual-clearance and reprogramming the tumor microenvironment. Schematic illustration of targeted co-delivery via VP-αPD-L1@REB for synergistic LCBM immunotherapy through driving PD-L1 dual-clearance and reprogramming the tumor microenvironment. Results and discussion To preliminarily validate the synergistic effect of VP and αPD-L1 in vivo, a subcutaneous tumor model was established using a murine lung adenocarcinoma tumor-derived cancer cell line (TDCL) allografts in C57BL/6 mice (Fig. S1A). Following the 21-day treatment course, the combination therapy (VP-αPD-L1) exhibited superior tumor suppression compared to both the IgG control and monotherapy groups, as evidenced by the excised tumor morphology (Fig. S1B) and changes in tumor volume in each group during treatment (Fig. S1C). These preliminary findings substantiate that the VP and αPD-L1 combination effectively restrains tumor progression. Preparation and characterization of VP-αPD-L1@REB Nevertheless, translating this synergistic pharmacological regimen to LCBM is hindered by the BBB. To overcome this anatomical barrier and achieve precise intracranial combination drug delivery, we designed an integrin-targeting, exosome-based biomimetic nanocomposite (VP-αPD-L1@REB). Specifically, brain-metastatic tumor cell-derived exosomes (EB) were co-extruded with self-assembled VP-αPD-L1, followed by surface functionalization with RGD peptide (REB) (Fig. 1A). The spontaneous self-assembly of VP and αPD-L1 into nanoscale cores is governed by multivalent non-covalent interactions, involving localized electrostatic complexation between the anionic carboxylates of VP and cationic surface domains of αPD-L1, further stabilized by hydrophobic aromatic stacking and intermolecular hydrogen bonding. The morphology and ultrastructure of the nanocomposite were first examined using transmission electron microscopy (TEM). As shown in Fig. 1B, the native exosomes exhibited a cup-shaped morphology, while the complexes of VP and αPD-L1 presented a spherical shape. Following the co-extrusion process, VP-αPD-L1@REB maintained a well-defined structure without obvious aggregation, indicating the integrity of the exosomal membrane coating. The particle size distribution of all formulations was analyzed by dynamic light scattering (DLS) (Fig. 1C). As detailed in Table S1, the final nanocomposite (VP-αPD-L1@REB) demonstrated an expanded diameter of 234.4 ± 68.11 nm compared to the natural EB, indicating successful drug encapsulation and RGD surface modification. Consistent with the size expansion, zeta potential measurements (Fig. 1D) revealed a clear charge shielding effect. Specifically, encapsulating the highly negative VP-αPD-L1 core (−27.91 ± 1.18 mV) into the weakly negative EB (−3.72 ± 0.31 mV) yielded an intermediate potential of −17.79 ± 0.91 mV, confirming effective biomimetic lipid cloaking. The final RGD-functionalized formulation stabilized at −15.16 ± 0.77 mV. Together, these optimized physicochemical properties, an appropriate nanoscale size and a moderately negative surface charge, satisfy the essential prerequisites for avoiding rapid immune clearance and ensuring prolonged systemic circulation. Furthermore, SDS-PAGE protein profiling (Fig. 1E) and western blot analysis (Fig. 1F) demonstrated that the engineered VP-αPD-L1@REB successfully retained the protein composition of the exosomes, including the classical exosome-specific markers TSG101 and CD9. The successful loading of VP was also verified via UV-Vis spectrophotometry, where VP-αPD-L1@REB exhibited the characteristic absorption peak of VP at 689 nm (Fig. 1G). The drug loading capacity (DL%) and encapsulation efficiency (EE%) of VP-αPD-L1@REB were quantitatively analyzed. As shown in Fig. S2, VP demonstrated a robust drug loading of 38.67%, while the co-loaded protein payload, αPD-L1, reached a capacity of 12.89%. Accordingly, by quantifying the drug loss in the filtrate, the exact encapsulation efficiencies of VP and αPD-L1 were calculated to be 62.0% and 41.3%, respectively.Fig. 1Preparation and characterization of VP-αPD-L1@REB. (A) Schematic illustration of the fabrication process of VP-αPD-L1@REB. (B-D) TEM images (Scale bar: 100 nm), hydrodynamic size distributions, and zeta potentials of EB, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB. (E-F) The respective protein composition of EB, VP-αPD-L1@EB, and VP-αPD-L1@REB was examined using SDS-PAGE and Western blot. (G) UV-Vis absorption spectra of VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB. (H) TEM images of VP-αPD-L1@REB incubated under physiological (pH 7.4) and acidic (pH 6.3) conditions. Scale bars: 200 nm (Left) and 100 nm (Right). (I) Stability evaluation of VP-αPD-L1@REB in PBS and FBS. (J) Hemolysis assays for free VP, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB at various concentrations. PC represents the positive control. (K) Quantitative evaluation of the hemolysis ratio corresponding to the assay in J. Preparation and characterization of VP-αPD-L1@REB. (A) Schematic illustration of the fabrication process of VP-αPD-L1@REB. (B-D) TEM images (Scale bar: 100 nm), hydrodynamic size distributions, and zeta potentials of EB, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB. (E-F) The respective protein composition of EB, VP-αPD-L1@EB, and VP-αPD-L1@REB was examined using SDS-PAGE and Western blot. (G) UV-Vis absorption spectra of VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB. (H) TEM images of VP-αPD-L1@REB incubated under physiological (pH 7.4) and acidic (pH 6.3) conditions. Scale bars: 200 nm (Left) and 100 nm (Right). (I) Stability evaluation of VP-αPD-L1@REB in PBS and FBS. (J) Hemolysis assays for free VP, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB at various concentrations. PC represents the positive control. (K) Quantitative evaluation of the hemolysis ratio corresponding to the assay in J. Considering the acidic nature of the tumor microenvironment (TME) [35], we further investigated the structural responsiveness of VP-αPD-L1@REB under different pH conditions using TEM (Fig. 1H). At a physiological pH 7.4, the nanocomposites retained their intact, dense spherical structures. When incubated in an acidic buffer (pH 6.3) mimicking the TME, the nanocomposites underwent significant structural dissociation and deformation, driven by proton-mediated disruption of both the outer lipid bilayer and the inner electrostatic core. To quantitatively validate this pH-responsive disassembly, in vitro release profiles were evaluated (Fig. S3). At physiological pH 7.4, the nanocomposite remained stable, with restricted leakage of VP (41.2%) and αPD-L1 (15.7%) over 72 h. Conversely, the acidic condition (pH 6.3) triggered rapid, size-dependent cargo liberation. The small-molecule VP exhibited an initial burst (49.2% within 2 h) and plateaued at 77.1% by 72 h. Meanwhile, owing to its macromolecular steric hindrance, αPD-L1 displayed a more controlled, gradual release, reaching 60.5% at 72 h. These quantitative release kinetics confirm that tumor-relevant acidic conditions effectively disrupt the non-covalent interactions of the inner VP-αPD-L1 core, facilitating the localized disassembly and release of both payloads within the intracranial tumor site. Beyond stimuli-responsiveness, physiological stability and systemic biosafety are fundamental prerequisites for in vivo translation. During a 5-day observation period in PBS and medium containing 10% fetal bovine serum (FBS), the hydrodynamic diameter of VP-αPD-L1@REB remained stable without significant aggregation (Fig. 1I), demonstrating excellent colloidal stability. Although the potential adsorption of serum proteins onto the exosomal surface can contribute to a moderate increase in hydrodynamic diameter in FBS compared to protein-free PBS, the particle size remains relatively stable over time without significant continuous aggregation. Subsequently, in vitro hemolysis testing was conducted using mammalian red blood cells incubated with varying concentrations of the formulations (0-100 μM). In contrast to the severe, dose-dependent erythrocyte rupture induced by free VP, the biomimetic VP-αPD-L1@REB nanocomposite exhibited a lower hemolysis rate, highlighting the protective encapsulation by the REB membrane (Fig. 1J–K). Finally, confirming endothelial biosafety is a strict requirement for non-destructive BBB transcytosis. Assays on murine brain microvascular endothelial cells (bEnd.3) demonstrated that while VP-αPD-L1 caused severe concentration-dependent cytotoxicity, VP-αPD-L1@EB and VP-αPD-L1@REB effectively preserved cellular viability at equivalent low concentrations (Fig. S4). By shielding the vascular endothelium from free VP toxicity, the engineered exosomes demonstrate a highly favorable safety profile, ensuring non-disruptive intracranial delivery. In vitro BBB penetration and anti-tumor efficacy of VP-αPD-L1@REB The BBB represents a major bottleneck for the effective CNS delivery of therapeutic antibodies and chemotherapeutics [11]. To evaluate the transcytosis capacity of the engineered nanoplatforms, an in vitro BBB model was established using a bEnd.3 cell monolayer cultured on the upper chamber of a Transwell insert, with TDCL cells seeded in the lower chamber (Fig. 2A). To rule out passive leakage, we monitored the transendothelial electrical resistance (TEER) of the bEnd.3 monolayer. Post-incubation (4 h), TEER values across all groups remained stable (>200 Ω cm2) with no significant deviation from the PBS control (Fig. S5). Following a 4-h co-incubation, the cellular uptake by the underlying TDCL cells was observed via confocal laser scanning microscopy (CLSM), utilizing the intrinsic red fluorescence of VP. As depicted in Fig. 2B, the control group exhibited negligible background signal, while the VP-αPD-L1 group showed limited red fluorescence, indicating poor BBB penetration. Conversely, the exosome-coated VP-αPD-L1@EB group displayed markedly enhanced intracellular fluorescence, suggesting that the TDCL exosomal membrane effectively facilitates endothelial transcytosis via homotypic affinity and natural vesicular transport pathways. Notably, the RGD-functionalized VP-αPD-L1@REB group exhibited the most intense red fluorescence. Quantitative analysis confirmed that the uptake in the VP-αPD-L1@REB group was significantly higher than in other groups (p < 0.0001; Fig. 2C). These results indicate that the synergistic combination of biomimetic exosomal coating and RGD-mediated integrin targeting substantially enhances both BBB traversal and subsequent active internalization by tumor cells.Fig. 2In vitro BBB penetration and anti-tumor efficacy of VP-αPD-L1@REB. (A) Schematic illustration of the in vitro BBB model. (B) CLSM images of TDCL cells in the lower chamber following a 4 h incubation with different formulations. Blue: DAPI-stained nuclei; Red: intrinsic fluorescence of VP. Scale bar: 50 μm. (C) Corresponding fluorescence quantification of VP. (D-E) ROS levels in TDCL cells from the lower chamber after respective treatments and corresponding fluorescence intensity analysis. Scale bar: 100 μm. (F-G) Flow cytometry analysis evaluating ROS levels in TDCL cells, along with the corresponding fluorescence quantification. (H) Live/dead cell imaging of TDCL cells in the lower chamber after being subjected to different treatments. Scale bars: 100 μm. (I-J) Flow cytometry plots and the corresponding quantitative analysis of apoptosis in TDCL cells following various treatments. Data are presented as mean ± SD (n = 3). Statistical significance was evaluated using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. In vitro BBB penetration and anti-tumor efficacy of VP-αPD-L1@REB. (A) Schematic illustration of the in vitro BBB model. (B) CLSM images of TDCL cells in the lower chamber following a 4 h incubation with different formulations. Blue: DAPI-stained nuclei; Red: intrinsic fluorescence of VP. Scale bar: 50 μm. (C) Corresponding fluorescence quantification of VP. (D-E) ROS levels in TDCL cells from the lower chamber after respective treatments and corresponding fluorescence intensity analysis. Scale bar: 100 μm. (F-G) Flow cytometry analysis evaluating ROS levels in TDCL cells, along with the corresponding fluorescence quantification. (H) Live/dead cell imaging of TDCL cells in the lower chamber after being subjected to different treatments. Scale bars: 100 μm. (I-J) Flow cytometry plots and the corresponding quantitative analysis of apoptosis in TDCL cells following various treatments. Data are presented as mean ± SD (n = 3). Statistical significance was evaluated using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. The intracellular accumulation of VP is known to trigger severe oxidative stress, even in the absence of light irradiation, which serves as a critical driver for subsequent cytotoxicity [36]. Mechanistically, under dark conditions, accumulated VP interacts with and disrupts mitochondrial membrane proteins, impairing the electron transport chain. This severe mitochondrial dysfunction causes continuous electron leakage, resulting in an intense burst of endogenous ROS, massive lipid peroxidation, and subsequent primary necrosis independent of photo-activation. To verify this, the intracellular ROS levels in the lower-chamber TDCL cells were evaluated using a DCFH-DA probe. CLSM images and the corresponding quantitative analysis indicated that while the control and VP-αPD-L1 groups generated minimal to mild ROS, the VP-αPD-L1@REB treatment provoked a massive ROS burst (Fig. 2D–E). This phenomenal oxidative stress was further corroborated at the precise, single-cell level via flow cytometry (Fig. 2F–G). Consistent with the imaging data, the VP-αPD-L1@REB group demonstrated a drastically elevated ROS median fluorescence intensity. These results confirm that the superior BBB-penetrating ability of the REB system translates directly into a high intracellular drug concentration, thereby effectively exacerbating the lethal oxidative stress within the tumor cells. Driven by the enhanced cellular uptake and subsequent ROS storm, we further investigated the in vitro anti-tumor efficacy of the nanocomposites. A Live/Dead cell staining assay was conducted to visually assess cytotoxicity (Fig. 2H). The control cells displayed robust green fluorescence (Calcein AM), indicating excellent viability. In contrast, the VP-αPD-L1@REB treated group exhibited extensive red fluorescence (PI), representing severe membrane damage and catastrophic cell death, visibly superior to the VP-αPD-L1 and VP-αPD-L1@EB groups. To precisely quantify this multifactorial cell death, an Annexin V-FITC/PI dual-staining assay was performed via flow cytometry (Fig. 2I). The quantitative analysis revealed a distinct formulation-dependent cytotoxicity profile (Fig. 2J). While the VP-αPD-L1 complex induced a modest apoptotic rate of 8.85%, VP-αPD-L1@EB elevated this to 14.45%, suggesting enhanced cellular internalization facilitated by the homotypic membrane. Strikingly, treatment with VP-αPD-L1@REB resulted in a massive induction of overall cell death. Beyond the robust apoptotic population (27.18%, FITC+), the targeted VP delivery induced a substantial accumulation of cells in the upper-left quadrant (36.1%, FITC-/PE+). This indicates that the intense intracellular ROS burst not only triggers classical apoptosis but also causes direct, severe membrane lipid peroxidation [37]. Given that apoptosis is an energy-dependent process, the concomitant bioenergetic crisis and acute structural damage to the lipid bilayer rapidly drive tumor cells into primary necrosis and complete cytolysis before the apoptotic cascade can be fully executed [38]. This dual-mode killing (apoptosis and necrosis) underscores the critical role of RGD-integrin binding in maximizing tumor ablation. Importantly, ROS overproduction and severe lipid peroxidation are classic triggers for immunogenic cell death (ICD), which acts as a vital bridge between local cytotoxicity and systemic immune activation. To verify this, we evaluated the emission of DAMPs in vitro. As shown in Fig. S6, compared to the control group, VP-αPD-L1@REB treatment induced massive exposure of calreticulin (CRT) on the tumor cell membrane, accompanied by the substantial nuclear export of HMGB1. Consistently, quantitative analyses revealed a significant extracellular release of ATP into the culture supernatant. By driving the emission of these classical DAMPs, targeted VP delivery effectively transforms ablated tumor cells into an autologous in situ vaccine. These emitted molecules act autonomously as potent endogenous in situ adjuvants, thereby establishing a robust immunological foundation for systemic immune priming without the need for synthetic adjuvant co-delivery. VP-mediated intracellular clearance of PD-L1 via dual lysosomal degradation pathways In addition to its robust ROS-mediated cytotoxicity, VP exerts a profound impact on the intracellular trafficking network of the immune checkpoint PD-L1 (Fig. 3A). We found that by disrupting PD-L1 dynamics, VP actively dismantles the tumor's defensive shield at the cellular level. Specifically, this interference forcefully drives internalized PD-L1 toward permanent degradation via dual “endosome-lysosome” and “autophagy-lysosome” pathways, thereby establishing a critical molecular foundation for subsequent immune activation. Given that VP-αPD-L1@REB nanocomposites were engineered to disassemble and release their respective cargos within the acidic tumor microenvironment, we utilized free VP and the unassembled VP-αPD-L1 complex to investigate the fundamental intracellular clearance mechanisms.Fig. 3VP-mediated intracellular clearance of PD-L1 via dual lysosomal degradation pathways. (A) Schematic diagram illustrating the reduction of PD-L1 by VP. (B-C) Flow cytometry histograms and corresponding mean fluorescence intensity of cell surface PD-L1 on TDCL cells treated with Control, VP, or VP-αPD-L1 for 24 h. (D-E) Flow cytometry analysis and corresponding fluorescence quantification of surface PD-L1 expression following 24 h treatment with varying VP concentrations. (F-H) Western blot analysis and relative protein quantification of total PD-L1 and CMTM6 in treated TDCL cells. (I-J) Representative protein expression and corresponding quantitative analysis of p62 and LC3B II. (K) Representative images of autophagy detected by TEM. Scale bars: 5 μm, top; 1 μm, bottom. (L-M) Representative immunofluorescence images and corresponding quantification of LC3B puncta (red; nuclei, blue; scale bar: 20 μm). (N) Relative mRNA expression of lysosome biogenesis-associated genes via RT-qPCR. (O-Q) Western blot analysis and relative quantification of PD-L1 and LAMP1 in TDCL cells co-incubated with or without the autophagy inhibitor Baf A1. (R-T) Corresponding protein expression and quantitative analysis of p62 and LC3B II under identical experimental conditions. Data are presented as mean ± SD (n = 3). Statistical significance was determined using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001; ns indicates not significant. VP-mediated intracellular clearance of PD-L1 via dual lysosomal degradation pathways. (A) Schematic diagram illustrating the reduction of PD-L1 by VP. (B-C) Flow cytometry histograms and corresponding mean fluorescence intensity of cell surface PD-L1 on TDCL cells treated with Control, VP, or VP-αPD-L1 for 24 h. (D-E) Flow cytometry analysis and corresponding fluorescence quantification of surface PD-L1 expression following 24 h treatment with varying VP concentrations. (F-H) Western blot analysis and relative protein quantification of total PD-L1 and CMTM6 in treated TDCL cells. (I-J) Representative protein expression and corresponding quantitative analysis of p62 and LC3B II. (K) Representative images of autophagy detected by TEM. Scale bars: 5 μm, top; 1 μm, bottom. (L-M) Representative immunofluorescence images and corresponding quantification of LC3B puncta (red; nuclei, blue; scale bar: 20 μm). (N) Relative mRNA expression of lysosome biogenesis-associated genes via RT-qPCR. (O-Q) Western blot analysis and relative quantification of PD-L1 and LAMP1 in TDCL cells co-incubated with or without the autophagy inhibitor Baf A1. (R-T) Corresponding protein expression and quantitative analysis of p62 and LC3B II under identical experimental conditions. Data are presented as mean ± SD (n = 3). Statistical significance was determined using one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001; ns indicates not significant. We first evaluated surface PD-L1 expression on TDCL cells using flow cytometry. TDCL cells were pre-stimulated with IFN-γ (100 ng/mL) to upregulate baseline PD-L1 levels, mimicking the inflammatory TME. As shown in Fig. 3B–C, both free VP and VP-αPD-L1 significantly reduced the expression of PD-L1 on the cell membrane compared to the control group. Notably, the VP-αPD-L1 complex achieved an even more robust reduction. This indicates a powerful synergy: the antibody's surface blockade accelerates the endocytosis of the PD-L1 receptor, while VP's intracellular activity prevents its return to the membrane. Furthermore, this downregulation was highly dose-dependent, as increasing concentrations of VP led to a progressive decline in surface PD-L1 mean fluorescent intensity (Fig. 3D–E). A major obstacle to αPD-L1 monotherapy is acquired immune resistance, which may be driven by the internalization and subsequent recycling of PD-L1 back to the tumor cell membrane [19]. Next, we focused our investigation on CMTM6, the key chaperone protein responsible for continuous endosomal recycling of PD-L1. Our results demonstrated that both VP and VP-αPD-L1 drastically suppressed the expression of CMTM6, accompanied by a massive reduction in total PD-L1 protein levels (Fig. 3F–H). To elucidate the mechanism underlying this CMTM6 downregulation, we investigated the role of YAP, the primary target of VP. YAP is also a core coactivator driving epithelial-mesenchymal transition (EMT) [39,40]. Specifically, YAP directly promotes the transcription of PD-L1 [21,22], while the YAP-driven EMT program endogenously upregulates CMTM6 to stabilize PD-L1 [41]. By functioning as a YAP inhibitor, our nanocomposite fundamentally disrupts this axis. As validated in TDCL cells, VP treatment suppressed YAP expression and reversed the EMT phenotype, evidenced by the upregulation of the epithelial marker E-cadherin and the downregulation of the mesenchymal marker Vimentin (Fig. S7A–B). Consequently, this EMT reversal stifles the downstream expression of CMTM6. Stripped of the CMTM6 protective shield, internalized PD-L1 is destabilized and redirected toward intracellular degradation pathways. Interestingly, overexpressing PD-L1 (PD-L1 OE) in TDCL cells directly induced an EMT phenotype, downregulating E-cadherin while upregulating Vimentin (Fig. S7C–D), and significantly accelerated tumor cell migration in wound healing assays (Fig. S7E). This confirms a reciprocal link where PD-L1 actively drives EMT and metastatic potential, a pathophysiological synchronization that is effectively disrupted by VP-based treatment. To further visualize this intracellular trafficking, immunofluorescence co-localization studies were performed (Fig. S8). VP and VP-αPD-L1 treatments increased the spatial overlap of PD-L1 with both the early endosome marker Rab5 and the late endosome marker Rab7 compared to the control group. These imaging results suggest that both VP and VP-αPD-L1 redirect PD-L1 from the recycling pathway toward the endosome-lysosome degradation route. Simultaneously, we investigated the “autophagy-lysosome” axis. Western blot analysis revealed that VP and VP-αPD-L1 treatments upregulated the expression of LC3B II while accelerating the degradation of p62 (Fig. 3I–J), indicating the robust activation of the cellular autophagy pathway [42]. This was visually corroborated by TEM, which captured the extensive formation of double-membrane autophagosomes and autolysosomes engulfing cytoplasmic contents (Fig. 3K). Immunofluorescence staining further confirmed a significant increase in LC3B puncta accumulation following treatment (Fig. 3L–M). Autophagic degradation is ultimately executed by lysosomes. Given that transcription factor EB (TFEB)-mediated lysosomal biogenesis is often coupled with enhanced autophagy [15], we examined the transcriptional regulation of lysosome-related genes. To accommodate the clearance of these massive endosomal and autophagosomal cargos, VP significantly upregulated the expression of lysosomal biogenesis-related genes, including Lamp1, Gla, Ctsb, Ctsd, Atp6v0d1, and Atp6v0b2 (Fig. 3N), thereby amplifying the cellular lysosomal degradation capacity. Furthermore, to definitively confirm that PD-L1 is permanently degraded via these lysosomal pathways, we utilized Bafilomycin A1 (Baf A1), a potent inhibitor of lysosomal acidification and fusion. Upon co-incubation with Baf A1 (100 nM), the profound degradation of total PD-L1 induced by VP and VP-αPD-L1 was successfully rescued (Fig. 3O–P). Concurrently, a massive accumulation of LAMP1 was observed across all Baf A1-treated groups, confirming that the inhibitor successfully paralyzed global lysosomal degradation (Fig. 3Q). Similarly, the blockade of lysosomal degradation led to a significant intracellular accumulation of p62 and LC3B II (Fig. 3R–T), validating that the previously observed autophagic flux was effectively truncated at the lysosome. These results demonstrate that the synergistic action of VP-mediated intracellular clearance and αPD-L1 surface blockade can drive PD-L1 into a dual lysosomal pathway for degradation. Finally, to resolve whether VP-mediated PD-L1 clearance is merely a secondary consequence of general oxidative cytotoxicity or an independent pharmacological mechanism, we performed functional decoupling assays utilizing the ROS scavenger N-acetylcysteine (NAC) (Fig. S9). Flow cytometry and cytotoxicity analyses demonstrated that co-incubation with 10 mM NAC effectively attenuated the VP-induced ROS burst, reducing the ROS-positive population from 75.3% to 49.1% (Fig. S9A), which consequently rescued TDCL cells from acute cell death across a broad concentration range (Fig. S9B). Strikingly, despite this substantial neutralization of oxidative stress and preservation of cell viability, Western blot profiling revealed that NAC co-treatment exerted no significant impact on VP-mediated downregulation of either CMTM6 or total PD-L1 (Fig. S9C–D). Consistent with these protein alterations, flow cytometry evaluation further verified that the VP-induced clearance of cell-surface PD-L1 also remained unaffected by NAC treatment (Fig. S9E–F). Both the VP and VP + NAC combination cohorts achieved suppression of CMTM6 and PD-L1 protein levels compared to the control and NAC-only groups (p > 0.05). Collectively, these decoupling findings demonstrate that VP-mediated PD-L1/CMTM6 degradation and ROS-induced cytotoxicity operate as parallel, independent mechanisms. This dual-action profile ensures that our delivery platform simultaneously achieves direct tumor ablation and robust elimination of adaptive immune resistance, establishing a powerful mechanistic foundation for subsequent TME remodeling in vivo. In vivo biodistribution and brain targeting evaluation To evaluate the BBB penetration and active targeting capability of the engineered nanocomposites in vivo, an intracranial LCBM model was established (Fig. 4A). Following intravenous administration, the real-time biodistribution of the formulations was dynamically monitored over an 8-h period utilizing the intrinsic fluorescence of VP (Fig. 4B). The non-targeted VP-αPD-L1 group exhibited minimal fluorescence in the cranial region, highlighting the restrictive nature of the BBB against conventional nanoparticles. VP-αPD-L1@EB yielded an observable increase in brain accumulation, attributable to the homologous targeting properties and natural vesicular transport mechanisms of the parent cell membrane. In sharp contrast, the VP-αPD-L1@REB group demonstrated rapid penetration and robust, sustained accumulation at the intracranial tumor site, maintaining a distinctly higher fluorescence signal than the other cohorts throughout the observation window. Quantitative tracking of the brain region (Fig. 4C) confirmed this substantial enhancement. Notably, at 8 h post-injection, the intracranial fluorescence intensity of the VP-αPD-L1@REB cohort was approximately 10.2-fold and 4.1-fold higher than that of the free VP-αPD-L1 and VP-αPD-L1@EB groups, respectively.Fig. 4In vivo biodistribution and brain targeting evaluation. (A) Schematic illustration of in vivo BBB traversal and brain tumor targeting. (B-C) In vivo fluorescence images and quantification of brain VP accumulation in LCBM-bearing mice 0.5-8 h post-injection of PBS, VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB. (D-G) Ex vivo fluorescence imaging and corresponding quantification of (D-E) major organs and (F-G) excised brains at 8 h post-injection. Data are mean ± SD (n = 3). In vivo biodistribution and brain targeting evaluation. (A) Schematic illustration of in vivo BBB traversal and brain tumor targeting. (B-C) In vivo fluorescence images and quantification of brain VP accumulation in LCBM-bearing mice 0.5-8 h post-injection of PBS, VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB. (D-G) Ex vivo fluorescence imaging and corresponding quantification of (D-E) major organs and (F-G) excised brains at 8 h post-injection. Data are mean ± SD (n = 3). To assess the targeting specificity and systemic off-target distribution, the major organs and whole brains were harvested for ex vivo imaging at the 8-h endpoint. As depicted in Fig. 4D and E, fluorescence accumulation in peripheral organs (heart, liver, spleen, lung, and kidney) was relatively comparable across the three treatment cohorts. This indicates that RGD surface modification did not exacerbate the non-specific aggregation phenomenon induced by the reticuloendothelial system [43]. The ex vivo fluorescence imaging of the isolated brains (Fig. 4F) and the corresponding quantitative analysis (Fig. 4G) directly corroborated the in vivo observations. The brains from the VP-αPD-L1@REB group exhibited intensely localized fluorescence within the tumor loci, with a signal intensity substantially surpassing the others, representing a 1.9-fold increase over the non-targeted VP-αPD-L1 group. Collectively, these biodistribution profiles validate that integrating the homotypic affinity of the exosomal membrane with the integrin-targeting precision of the RGD peptide efficiently facilitates BBB traversal and maximizes drug retention in LCBM lesions. In vivo anti-tumor efficacy in the LCBM model To evaluate the in vivo therapeutic potential of the nanocomposites against LCBM, an orthotopic intracranial tumor model was established utilizing luciferase-expressing TDCL cells (TDCL-Luc). Following a 14-day tumor maturation period, mice were randomly assigned to receive intravenous injections of PBS, VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB (Fig. 5A). Intracranial tumor burden was continuously monitored via in vivo bioluminescence imaging. As depicted in Fig. 5B and C, the control group exhibited aggressive and unrestrained tumor proliferation. Conversely, while the VP-αPD-L1 complex afforded limited tumor inhibition due to its poor BBB penetrability, the VP-αPD-L1@REB demonstrated the most profound anti-tumor efficacy, maintaining the lowest bioluminescence signal throughout the monitoring period. Quantitative analysis on day 8 post-treatment revealed that the tumor burden in the VP-αPD-L1@REB group was suppressed by approximately 3.1-fold, 2.3-fold and 1.6-fold compared to the control, VP-αPD-L1 and VP-αPD-L1@EB groups, respectively. This marked suppression highlights the critical advantage of the RGD-engineered biomimetic surface in facilitating active BBB transcytosis and subsequent tumor deep penetration. This tumor inhibition directly translated into a substantial survival advantage. Kaplan-Meier survival analysis (Fig. 5D) demonstrated that mice treated with VP-αPD-L1@REB exhibited the most significant prolongation of median survival time, with efficacy far surpassing that of the VP-αPD-L1 and VP-αPD-L1@EB groups. Furthermore, body weight monitoring throughout the 10-day therapeutic window showed negligible fluctuations across all treatment groups (Fig. 5E), indicating the physiological tolerance and systemic biosafety of the delivery platform.Fig. 5In vivo anti-tumor efficacy in the LCBM model. (A) Experimental timeline. (B-C) Representative in vivo bioluminescence images and the corresponding bioluminescence intensity of intracranial LCBM tumors following the indicated treatments. (D) Kaplan-Meier survival curves of the tumor-bearing mice. (E) Body weight profiles are monitored throughout the therapeutic regimen. (F-H) Representative images of H&E, TUNEL, and Ki67 staining in excised brain tumor sections (F), alongside the relative quantification of TUNEL (G) and Ki67 (H) signals (n = 3). Scale bars: 100 μm. Data are presented as mean ± SD. Statistical significance was determined via one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates not significant. In vivo anti-tumor efficacy in the LCBM model. (A) Experimental timeline. (B-C) Representative in vivo bioluminescence images and the corresponding bioluminescence intensity of intracranial LCBM tumors following the indicated treatments. (D) Kaplan-Meier survival curves of the tumor-bearing mice. (E) Body weight profiles are monitored throughout the therapeutic regimen. (F-H) Representative images of H&E, TUNEL, and Ki67 staining in excised brain tumor sections (F), alongside the relative quantification of TUNEL (G) and Ki67 (H) signals (n = 3). Scale bars: 100 μm. Data are presented as mean ± SD. Statistical significance was determined via one-way ANOVA. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001; ns indicates not significant. To elucidate the microscopic mechanisms underlying this therapeutic efficacy, excised brain tumor tissues were subjected to histological evaluations. Hematoxylin and eosin (H&E) staining (Fig. 5F, top) revealed extensive regions of structural disintegration, nuclear pyknosis, and massive necrosis within the tumor tissues of the VP-αPD-L1@REB group, standing in stark contrast to the dense, hypercellular morphology observed in the control group. Consistently, the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay demonstrated that the VP-αPD-L1@REB induced the most intense cellular apoptosis (Fig. 5F), achieving a remarkable 15.5-fold increase in apoptotic fluorescence intensity compared to the VP-αPD-L1 group (Fig. 5G). Conversely, the expression of Ki67, a universal marker for cellular proliferation, was most profoundly downregulated in the VP-αPD-L1@REB-treated tissues (Fig. 5F), representing a significant 4.2-fold decrease relative to the control (Fig. 5H). Collectively, these in vivo and ex vivo findings demonstrate that the VP-αPD-L1@REB nanocomposite possesses a remarkable ability to safely traverse the BBB and exert potent, targeted clearance of established brain metastases. Lysosomal clearance of PD-L1 facilitates TME remodeling and systemic anti-tumor immunity Having elucidated the dual intracellular clearance mechanisms in vitro, we sought to verify whether this targeted degradation of PD-L1 could be successfully translated into the complex orthotopic TME. Immunofluorescence staining was performed on excised brain tumor sections (Fig. 6A). The in vivo data revealed a massive upregulation of LC3B (7.5-fold increase) and LAMP1 (9.8-fold increase), coupled with a drastic 90.9% reduction in PD-L1 expression in the VP-αPD-L1@REB cohort compared to the control (Fig. 6B–D). These robust in vivo findings confirm that the RGD-based biomimetic delivery strategy effectively drives the dual “endosome-lysosome” and “autophagy-lysosome” clearance of PD-L1 in brain metastases. Notably, the massive accumulation of LAMP1 serves as the ultimate convergence point for both degradation axes, mirroring our in vitro mechanistic models.Fig. 6Lysosomal clearance of PD-L1 facilitates TME remodeling and systemic anti-tumor immunity. (A-D) Representative immunofluorescence images (A) and corresponding relative quantitative analyses of PD-L1 (B), LAMP1 (C), and LC3B (D) in excised tumor sections following the indicated treatments. Scale bars: 50 μm. (E-G) Immunofluorescence evaluation (E) and relative quantification of CD4+ (F) and CD8+ (G) T cell infiltration within the tumor microenvironment. Scale bars: 50 μm. (H-I) Representative flow cytometry plots (H) and statistical quantification (I) of matured DCs (CD80+CD86+) following different treatments. (J-K) Flow cytometry analysis (J) and the calculated ratio of CD8+ to CD4+ T cells (K) in the spleen. Note: The color in all bar graphs corresponds to the specific treatment cohorts indicated at the bottom of the figure. Data are presented as mean ± SD (n = 3). Statistical significance was assessed via one-way ANOVA (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns indicates not significant). Lysosomal clearance of PD-L1 facilitates TME remodeling and systemic anti-tumor immunity. (A-D) Representative immunofluorescence images (A) and corresponding relative quantitative analyses of PD-L1 (B), LAMP1 (C), and LC3B (D) in excised tumor sections following the indicated treatments. Scale bars: 50 μm. (E-G) Immunofluorescence evaluation (E) and relative quantification of CD4+ (F) and CD8+ (G) T cell infiltration within the tumor microenvironment. Scale bars: 50 μm. (H-I) Representative flow cytometry plots (H) and statistical quantification (I) of matured DCs (CD80+CD86+) following different treatments. (J-K) Flow cytometry analysis (J) and the calculated ratio of CD8+ to CD4+ T cells (K) in the spleen. Note: The color in all bar graphs corresponds to the specific treatment cohorts indicated at the bottom of the figure. Data are presented as mean ± SD (n = 3). Statistical significance was assessed via one-way ANOVA (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns indicates not significant). Following the successful dismantling of the immunosuppressive PD-L1 shield, we evaluated the consequent immunological remodeling within the local TME. Immunofluorescence staining revealed that the VP-αPD-L1@REB group exhibited a massive infiltration of effector lymphocytes, achieving a remarkable 17.0-fold and 6.5-fold increase in the recruitment of CD4+ helper T cells and CD8+ cytotoxic T cells, respectively (Fig. 6E–G). To elucidate the systemic origin of this robust local T cell infiltration, we further analyzed the immune status in the spleen, a major peripheral lymphoid organ [44]. The local tumor destruction and oxidative stress induced by the nanocomposites effectively released tumor-associated antigens, serving as a potent trigger for a systemic immune response. Flow cytometry analysis revealed that the VP-αPD-L1@REB treatment induced the most pronounced maturation of dendritic cells (DCs), yielding 31.9% CD80+/CD86+ mature DCs compared to merely 18.7% in the PBS control (Fig. 6H–I), indicative of significantly enhanced systemic antigen presentation. Consequently, this peripheral antigen presentation drove a prominent systemic expansion of effector T lymphocytes, evidenced by the significantly elevated proportion of CD8+ T cells (peaking at 36.2% in the VP-αPD-L1@REB cohort, compared to 23.6% in the control) and the highest CD8+/CD4+ T cell ratio within the splenic tissues (Fig. 6J–K). Together, these results demonstrate a dual-synergistic mechanism: while ICD-derived endogenous adjuvants systemically prime and expand peripheral T cells, the targeted lysosomal degradation of PD-L1 removes the primary immunosuppressive checkpoint in brain metastases, permitting effector T cells to freely infiltrate and convert “cold” intracranial lesions into inflamed, T-cell-infiltrated “hot” tumors. In vivo decoupling of ROS-induced tumor ablation and PD-L1 clearance To further validate the in vitro decoupling findings within the complex orthotopic TME, an in vivo rescue experiment was performed using the LCBM model (Fig. S10A). Immunofluorescence analysis of the excised brain tumors confirmed that pretreatment with NAC effectively scavenged the intratumoral ROS generated by the nanocomposite (Fig. S10B–C). Consequently, this neutralization of oxidative stress attenuated cellular apoptosis (TUNEL) and partially restored cellular proliferation (Ki67) (Fig. S10D–F), which corresponded to a partial recovery of tumor growth (Fig. S10G–H), while body weights remained stable across all groups (Fig. S10I). Crucially, despite the reduction in ROS levels, the targeted in vivo degradation of CMTM6 and PD-L1 within the brain tumor tissues remained unhindered (Fig. S11A–C), mirroring the in vitro results. Benefiting from this sustained immune checkpoint clearance, local TME analysis revealed that the combination treatment maintained a robust infiltration of CD4+ and CD8+ T cells in the brain lesions (Fig. S11D–F). Furthermore, we evaluated the impact on systemic immunity. Flow cytometry of the spleen indicated that NAC-mediated ROS scavenging led to a moderate attenuation in systemic DC maturation (Fig. S11G–H) and splenic T-cell priming (Fig. S11I–J), likely due to a decrease in ROS-dependent immunogenic cell death. Collectively, these in vivo findings corroborate that ROS-induced tumor ablation and ROS-independent PD-L1 clearance operate synergistically, with the latter mitigating local adaptive immune resistance to facilitate effector T cell infiltration. In vivo biosafety evaluation Excellent biosafety and biocompatibility are fundamental prerequisites for the potential clinical translation of any nanomedicine. Therefore, comprehensively evaluating the systemic toxicity of the nanocomposites is imperative. At the end of the therapeutic regimen, whole blood and serum samples were collected to assess hematological and biochemical parameters. As shown in Fig. 7A, standard complete blood count (CBC) markers, including white blood cells (WBC), red blood cells (RBC), platelets (PLT), hemoglobin (HGB), and other related erythrocyte indices (e.g., HCT, MCV, MCH, and MCHC), remained within normal physiological ranges across all treatment groups, indicating an absence of severe systemic inflammation or myelosuppression. Furthermore, serum biochemical analyses (Fig. 7B) revealed no significant fluctuations in the levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) for liver function, nor in blood urea nitrogen (BUN) and creatinine (CREA) for kidney function, ruling out potential hepato-renal toxicity.Fig. 7In vivo biosafety evaluation. (A) CBC analysis of the treated mice. (B) Serum biochemical parameters in mice following designated treatments. (C) Representative H&E-stained tissue sections of the major peripheral organs excised from the respective experimental cohorts. Scale bar: 100 μm. Data are presented as mean ± SD (n = 5). No statistically significant differences were observed among the groups. In vivo biosafety evaluation. (A) CBC analysis of the treated mice. (B) Serum biochemical parameters in mice following designated treatments. (C) Representative H&E-stained tissue sections of the major peripheral organs excised from the respective experimental cohorts. Scale bar: 100 μm. Data are presented as mean ± SD (n = 5). No statistically significant differences were observed among the groups. Consistent with the blood test results, histological examinations further validated the in vivo safety of the formulations. H&E staining of major organs (heart, liver, spleen, lung, and kidney) (Fig. 7C) demonstrated intact, normal tissue architectures. No significant signs of pathological damage, such as cellular necrosis, substantial hemorrhage, or structural disorganization, were observed in any of the experimental groups. Notably, despite the robust systemic immune activation observed previously, the spleen maintained its healthy histological morphology without showing signs of immunotoxicity or structural collapse. Taken together, these comprehensive biosafety evaluations confirm that the VP-αPD-L1@REB nanocomposites possess a broad therapeutic window, successfully achieving an optimal balance between potent anti-tumor efficacy and excellent systemic tolerability. Current delivery strategies for brain tumors often rely on attaching antibodies to the surface of nanocarriers. However, this approach exposes the therapeutic proteins to the systemic circulation, increasing the risk of premature clearance and unintended binding to peripheral tissues [45]. In contrast, our VP-αPD-L1@REB platform utilizes an internal encapsulation strategy. By shielding the antibody within the biomimetic vesicle, we minimize systemic exposure while reserving the carrier's surface for targeting peptides. This structural design significantly enhances the stability of the payload and maximizes its accumulation within the intracranial tumor site. Mechanistically, the co-delivery system effectively bridges targeted therapy and immunotherapy. VP facilitates the intracellular degradation of tumor PD-L1, thereby sensitizing the cancer cells to the co-delivered αPD-L1 antibody. Despite these promising synergistic effects, several translational challenges remain. First, the inherent light-sensitivity of VP requires strict light-protective measures during formulation and clinical administration. Second, ensuring the batch-to-batch consistency of cell-derived vesicles remains a significant hurdle for large-scale manufacturing. Additionally, while our results demonstrate a strong initial T cell response, further long-term studies are needed to confirm whether this therapy can establish durable immune memory to prevent tumor recurrence. Conclusion In summary, the integrin-targeting engineered exosome-based nanocomposite VP-αPD-L1@REB successfully overcomes the BBB and circumvents adaptive immune resistance in LCBM. By synergistically coupling ROS-induced tumor ablation with the dual “endosome-lysosome” and “autophagy-lysosome” clearance of PD-L1, this platform triggers robust systemic anti-tumor immunity and converts highly immunosuppressive “cold” brain metastases into “hot” tumors. Exhibiting profound therapeutic efficacy, targeted biodistribution, and excellent biosafety in orthotopic models, VP-αPD-L1@REB represents a highly promising, mechanistically distinct translational strategy for the treatment of driver gene-negative CNS malignancies.List of abbreviationsAbbreviationsFull nameBaf A1Bafilomycin A1BBBBlood-brain barrierCBCComplete blood countCLSMConfocal laser scanning microscopyCNSCentral nervous systemDCDendritic cellDLSDynamic light scatteringEBBrain-metastatic tumor cell-derived exosomesEMTEpithelial-mesenchymal transitionFBSFetal bovine serumH&EHematoxylin and eosinICIImmune checkpoint inhibitionLCBMLung cancer brain metastasisPD-L1Programmed cell death ligand 1αPD-L1Anti-PD-L1 antibodyRGDArginine-glycine-aspartic acidREBEB with RGD peptidesROSReactive oxygen speciesRT-qPCRReverse transcription-quantitative polymerase chain reactionTDCLTumor-derived cancer cell lineTEMTransmission electron microscopyTMETumor microenvironmentTUNELTerminal deoxynucleotidyl transferase dUTP nick end labelingVPVerteporfin List of abbreviationsAbbreviationsFull nameBaf A1Bafilomycin A1BBBBlood-brain barrierCBCComplete blood countCLSMConfocal laser scanning microscopyCNSCentral nervous systemDCDendritic cellDLSDynamic light scatteringEBBrain-metastatic tumor cell-derived exosomesEMTEpithelial-mesenchymal transitionFBSFetal bovine serumH&EHematoxylin and eosinICIImmune checkpoint inhibitionLCBMLung cancer brain metastasisPD-L1Programmed cell death ligand 1αPD-L1Anti-PD-L1 antibodyRGDArginine-glycine-aspartic acidREBEB with RGD peptidesROSReactive oxygen speciesRT-qPCRReverse transcription-quantitative polymerase chain reactionTDCLTumor-derived cancer cell lineTEMTransmission electron microscopyTMETumor microenvironmentTUNELTerminal deoxynucleotidyl transferase dUTP nick end labelingVPVerteporfin List of abbreviations Experimental section Materials and methods Verteporfin (VP) was purchased from Aladdin (Shanghai, China). The anti-mouse PD-L1 antibody (Clone: 10F.9G2, Cat# BE0101) and its corresponding isotype control (rat IgG2b, Cat# BE0090) were obtained from BioXCell (Lebanon, NH, USA). DSPE-PEG-RGD was purchased from Yare Biotech (Shanghai, China). Mouse IFN-γ (Cat# 315-05) was sourced from PeproTech (Cranbury, NJ, USA). Given the inherent photosensitivity of VP, strict light-avoidance protocols were implemented throughout all experimental procedures to completely exclude the interference of ambient light and prevent unintentional photo-activation. Specifically, all formulation preparations, in vitro cell cultures, compound dosing, and functional assays were conducted under dark conditions. Furthermore, all drug-containing tubes, cell culture dishes, and animal housing cages were strictly wrapped in aluminum foil during incubation, transportation, and in vivo housing. The murine lung adenocarcinoma tumor-derived cancer cell line (TDCL) was kindly provided by Prof. Shuguo Sun (Huazhong University of Science and Technology, Wuhan, China). The cells were maintained in RPMI 1640 medium (Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS; Sigma-Aldrich, St. Louis, MO, USA) in a humidified incubator at 37 °C with 5% CO2. Male C57BL/6 mice (6-8 weeks old, 16-20 g) were purchased from Hangzhou Qizhen Laboratory Animal Co., Ltd. (Hangzhou, China). The mice were housed under specific pathogen-free (SPF) conditions (23 ± 1 °C; 50% ± 10% humidity) with a 12 h light/dark cycle and free access to a standard rodent diet and water. Exosome extraction To isolate brain metastatic TDCL-derived exosomes (EB), cells were cultured in medium supplemented with exosome-depleted FBS. The conditioned medium was collected and concentrated using a 100 kDa molecular weight cut-off (MWCO) ultrafiltration centrifugal filter at 4000 rpm and 4 °C for 15 min. Subsequently, differential centrifugation was performed to remove dead cells and debris: the medium was centrifuged at 500 × g for 10 min, 2000 × g for 10 min, and 10000 × g for 30 min. The resulting supernatant was filtered through a 0.22 μm syringe filter and then ultracentrifuged using a Beckman SW 32 Ti rotor at 100000 × g and 4 °C for 70 min. The obtained exosome pellets were resuspended in 30 mL of ice-cold PBS and ultracentrifuged again at 100000 × g for 70 min for washing. Finally, the purified EBs were resuspended in cold PBS and stored at −80 °C for downstream applications. Preparation of VP-αPD-L1@REB The VP-αPD-L1 core was first self-assembled by stirring VP (4 mg) and αPD-L1 antibody (2 mg) in PBS at 4 °C for 1 h. Then the core was mixed with sonicated TDCL-derived exosomes. To achieve homogeneous membrane coating, the mixture was extruded through a 200 nm polycarbonate membrane for 50 cycles, yielding VP-αPD-L1@EB. Finally, 2 mg of RGD peptide was added and incubated overnight at 4 °C to obtain the target VP-αPD-L1@REB nanocomposites. Characterization of VP-αPD-L1@REB The morphological observations of EB, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB were conducted using a TEM (Hitachi H-7650, Japan). Briefly, the samples were dropped onto carbon-coated copper grids, negatively stained with 2% phosphotungstic acid, and dried before imaging. The hydrodynamic size distribution and zeta potential of the formulations were determined in PBS utilizing a Nano ZS90 particle size analyzer (Malvern Panalytical, UK). The encapsulation of VP was qualitatively verified by recording the UV-Vis absorption spectra over a range of 600-800 nm using a spectrophotometer, focusing on the characteristic peak of VP at 689 nm. To confirm the retention of the exosomal membrane proteins, protein concentrations were determined using a BCA protein assay kit (Cat# 23225, Thermo Fisher Scientific, USA). For SDS-PAGE, the protein lysates were separated on polyacrylamide gels and stained with Coomassie Brilliant Blue to visualize total protein profiles. For Western blotting, the separated proteins were transferred onto PVDF membranes (Millipore, USA). The membranes were blocked with 5% non-fat milk in TBST and sequentially incubated with primary antibodies against TSG101 (Cat# 72312, CST, USA) and CD9 (Cat# 98327, CST, USA) overnight at 4 °C. After incubation with HRP-conjugated secondary antibodies (Cat# 7074, CST, USA), the protein bands were visualized using a ChemiDoc XRS + Imaging System (Bio-Rad, USA) and analyzed with Image Lab software. To accurately determine the drug loading capacity (DL%) and encapsulation efficiency (EE%), the assembled nanocomposites were first purified by ultrafiltration centrifugation to separate the unencapsulated free VP and αPD-L1. The filtrate was collected to evaluate the loss of payloads during assembly, which was calculated by directly quantifying the free drug concentrations in the filtrate (Loss = Cfiltrate × Vfiltrate, where Cfiltrate is the drug concentration in the filtrate and Vfiltrate is the total volume of the filtrate). Subsequently, a measured quantity of the purified VP-αPD-L1@REB was completely lysed in a 1% Triton X-100 solution, accompanied by brief sonication to ensure complete release of the payloads. The concentration of the encapsulated VP was quantified using a fluorescence spectrophotometer (λex = 400 nm, λem = 690 nm), while the encapsulated αPD-L1 content was determined using a specific Rat IgG ELISA kit (Cat# ab189578, Abcam) according to the manufacturer's instructions. The DL% and EE% were calculated using the following exact formulas:(1)DL(%)=WencapsulatedWtotal×100(2)EE(%)=WencapsulatedWinitial×100Wencapsulated represents the mass of the encapsulated payload and Wtotal represents the total dry mass of the purified nanocomposites. Winitial represents the initial input mass of VP or αPD-L1 added to the assembly system. pH-responsive dissociation and stability To evaluate structural responsiveness, VP-αPD-L1@REB was incubated in different buffers simulating normal physiological conditions (PBS, pH 7.4) and the acidic tumor microenvironment (PBS, pH 6.3) for 6 h. The morphological alterations and structural dissociation were subsequently captured via TEM imaging. The in vitro release profiles of VP and αPD-L1 from VP-αPD-L1@REB were evaluated using a dialysis method. Briefly, 1 mg of lyophilized VP-αPD-L1@REB nanoparticles was dispersed in 1 mL of PBS and transferred into a dialysis bag (MWCO 300 kDa). The dialysis bag was then immersed in 10 mL of release medium (PBS, pH 6.3 or 7.4) supplemented with 0.1% (v/v) Tween-80, and maintained at 37 °C with continuous agitation. At predetermined time intervals (0, 0.5, 1, 2, 4, 8, 12, 24, 48, and 72 h), 1 mL of the external release medium was withdrawn for analysis and immediately replenished with an equal volume of fresh buffer. The concentration of VP was determined using UV-Vis spectrophotometry at 689 nm. The concentration of αPD-L1 was quantified utilizing a Rat IgG ELISA kit (Cat# ab189578, Abcam). All release experiments were performed in triplicate. Furthermore, the stability of VP-αPD-L1@REB was evaluated by incubating the nanocomposites in PBS and cell culture medium containing 10% FBS at 37 °C. The hydrodynamic diameter was monitored dynamically using DLS over a continuous 5-day period. Hemolysis assay The hemocompatibility of the nanoplatform was assessed using a standard hemolysis assay. Fresh mammalian red blood cells (RBCs) were collected, washed repeatedly with PBS, and diluted to a 2% (v/v) suspension. The RBC suspensions were incubated with various concentrations (0-100 μM) of free VP, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB at 37 °C for 2 h. Deionized water and PBS were utilized as the positive control (PC, representing 100% hemolysis) and negative control, respectively. Following incubation, the mixtures were centrifuged, and the absorbance of the hemoglobin released in the supernatant was measured at 540 nm using a microplate reader. The hemolysis ratio was calculated using the following formula:(3)Hemolysisratio(%)=Asample‐AnegativeApositive‐Anegative×100Asample represents the absorbance of the sample group; Anegative represents the absorbance of the negative control group; Apositive represents the absorbance of the positive control group. In vitro endothelial cytotoxicity bEnd.3 cells (5 × 103 cells/well) were seeded in 96-well plates and incubated overnight. The cells were then treated with various formulations (ranging from 0 to 10 μM VP-equivalent) for 24 h. Subsequently, CCK-8 solution (10 μL per well) was added, and the cells were incubated at 37 °C for another 1-4 h. The absorbance at 450 nm was measured using a Spark multimode microplate reader (Tecan, Switzerland). Establishment of the in vitro BBB model and cellular uptake assay The in vitro BBB model was established using Transwell inserts (0.4 μm pore size, LABSELECT, China). Briefly, bEnd.3 cells were seeded in the upper chamber and cultured to form a confluent monolayer. The model integrity was confirmed by transendothelial electrical resistance (TEER), with values exceeding 200 Ω cm2 before use. TDCL cells were seeded in the lower chamber. For cellular uptake, the upper medium was replaced with fresh medium containing PBS (Control), VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB (0.5 μΜ VP-equivalent). To thoroughly assess the barrier integrity during the treatment, TEER values of the bEnd.3 monolayer were measured again immediately after the 4 h incubation period. After 4 h of incubation at 37 °C, TDCL cells were washed, fixed with 4% paraformaldehyde, and counterstained with DAPI. The intracellular distribution of VP was visualized using a CLSM (LSM 900, Zeiss, Germany) and quantified by mean fluorescence intensity using ImageJ software. Intracellular ROS detection To evaluate intracellular ROS generation after crossing the BBB, TDCL cells treated with 0.5 μM VP-equivalent were collected from the lower chamber following the 4-h Transwell treatment and incubated with DCFH-DA (Cat# S0033, Beyotime, China) for 30 min at 37 °C in the dark. For qualitative observation, cells were washed with PBS and immediately imaged via CLSM. For quantitative analysis, the cells were harvested, resuspended in PBS, and analyzed using a flow cytometer (BD FACSCanto, USA). The mean fluorescence intensity was determined using FlowJo software to reflect intracellular ROS levels. Furthermore, to assess the ROS-scavenging effect of N-acetylcysteine (NAC, Cat# HY-B0215, MCE) in the mechanistic decoupling assays, TDCL cells were seeded in 6-well plates and allowed to adhere overnight. The cells were then treated with PBS (Control), NAC (10 mM), VP (0.5 μM), or the combination of VP and NAC (0.5 μM VP + 10 mM NAC) for 24 h. Following treatment, the cells were harvested, incubated with DCFH-DA for 30 min at 37 °C in the dark, and analyzed via flow cytometry as described above. The percentage of ROS-positive cells was quantified using FlowJo software. In vitro cytotoxicity and apoptosis assays The anti-tumor efficacy after BBB penetration was first evaluated via Live/Dead staining. Following the Transwell treatments, TDCL cells treated with 0.5 μM VP-equivalent were stained with a Calcein-AM/PI kit (Cat# C2015, Beyotime, China) for 30 min at 37 °C, followed by CLSM imaging to visualize live (green) and dead (red) cells. To quantify the cell death mode, an apoptosis assay was performed using an Annexin V-FITC/PI kit (Cat# A211-02, Vazyme, China). Briefly, treated cells were harvested, resuspended in binding buffer, and dual-stained for 15 min at room temperature in the dark. The apoptotic populations were acquired via flow cytometry and analyzed using FlowJo software. Furthermore, to assess the cytoprotective effect of NAC against VP-induced cell death in the mechanistic decoupling assays, TDCL cells were seeded into 96-well plates and allowed to adhere overnight. The cells were then treated with varying concentrations of VP (ranging from 0 to 5 μM) in the presence or absence of 10 mM NAC for 24 h. Subsequently, CCK-8 solution (10 μL per well) was added, and the cells were incubated at 37 °C for another 1-2 h. The absorbance at 450 nm was measured using a microplate reader to calculate relative cell viability. In vitro immunogenic cell death (ICD) assay To evaluate the emission of damage-associated molecular patterns (DAMPs), an in vitro blood-brain barrier model was established. Subsequently, the samples were treated with PBS, VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB for 24 h (with a VP concentration of 0.5 μM for all). For calreticulin (CRT) detection, the treated cells were gently washed with cold PBS and fixed with 4% paraformaldehyde (without Triton X-100 permeabilization to preserve intact surface antigens). For HMGB1 detection, the fixed cells were permeabilized with 0.1% Triton X-100 for 10 min. Following blocking with 3% BSA for 1 h, the cells were incubated overnight at 4 °C with primary antibodies against CRT (Cat# 27298-1-AP, Proteintech, China) or HMGB1 (Cat# 10829-1-AP, Proteintech, China). Subsequently, the cells were incubated with an Alexa Fluor 488-conjugated secondary antibody for 1 h in the dark. Nuclei were counterstained with DAPI. The fluorescence images were acquired utilizing a CLSM (LSM 900, Zeiss, Germany). The extracellular release of ATP was quantitatively assessed utilizing an Enhanced ATP Assay Kit (Cat# S0027, Beyotime, China) according to the manufacturer's protocol. Flow cytometry analysis To evaluate surface PD-L1 expression, TDCL cells were pre-stimulated with 100 ng/mL IFN-γ for 12 h. Following treatment with different formulations or VP with different concentrations (0.5 to 10 μM) for 24 h, cells were harvested and stained with ABflo 488-anti-PD-L1 (Cat# A24568, ABclonal, China) or a corresponding IgG isotype control (Cat# A22069, ABclonal) at 4 °C for 30 min in the dark. To further assess the impact of ROS scavenging on surface PD-L1 clearance in the decoupling assays, IFN-γ-primed TDCL cells were treated with PBS, NAC (10 mM), VP (0.5 μM), or VP + NAC (0.5 μM VP + 10 mM NAC) for 24 h, and subsequently stained and analyzed for surface PD-L1 expression following the same protocol. Splenocytes were isolated by passing spleens through a 70 μm cell strainer, followed by treatment with ACK lysis buffer to remove erythrocytes. To assess DC maturation, single-cell suspensions were stained with FITC-anti-CD11c (Cat# 117306), PE-anti-CD80 (Cat# 104707), and PE/Cy7-anti-CD86 (Cat# 105013). For T cell profiling, cells were labeled with FITC-anti-CD3 (Cat# 100204), PE/Cy7-anti-CD4 (Cat# 100422), and PE-anti-CD8a (Cat# 100708). All immune cell antibodies were purchased from BioLegend (San Diego, CA, USA) and incubated at 4 °C for 30 min in the dark. All samples were acquired using a BD FACSCanto flow cytometer (BD Biosciences, USA) and analyzed with FlowJo software. The results were expressed as the percentage of positive cells or mean fluorescence intensity. Establishment of PD-L1 overexpressing cells To establish the PD-L1 overexpressing (PD-L1 OE) cell model, TDCL cells were transduced with lentiviral vectors encoding the murine Cd274 (PD-L1) gene. Cells transduced with empty lentiviral vectors served as the negative control (Control). Following infection, the cells were cultured in medium containing puromycin for stable selection. The successful overexpression of PD-L1 and its subsequent impact on epithelial-mesenchymal transition (EMT) markers were verified via Western blot analysis. Wound-healing migration assay The migratory capacity of the tumor cells was evaluated using a wound-healing assay. Control and PD-L1 OE TDCL cells were seeded into 6-well plates and cultured until reaching 90%–100% confluence. A linear scratch was subsequently created across the center of the cell monolayer using a sterile 200 μL pipette tip. After washing gently with PBS to remove detached debris, the cells were cultured in low-serum medium (containing 1% FBS) to minimize the confounding influence of cell proliferation. Wound closure was monitored, and representative images were captured at 0, 24, 48, and 72 h post-scratch using an inverted phase-contrast microscope. The relative cell migration was qualitatively assessed by observing the gap closure over time. Western blotting analysis TDCL cells were pre-stimulated with 100 ng/mL IFN-γ for 12 h and then treated with 0.5 μM VP or VP-αPD-L1 for 24 h. Total proteins were then extracted using RIPA lysis buffer (Cat# P0038, Beyotime, China) supplemented with protease inhibitors (Cat# HY-K0010, MCE) and phosphatase inhibitors (Cat# HY-K0022, MCE). Equal amounts of protein were resolved via SDS-PAGE and transferred onto PVDF membranes. After blocking with 5% non-fat milk for 1 h at room temperature, the membranes were incubated overnight at 4 °C with primary antibodies against PD-L1 (Cat# ab213480, Abcam), CMTM6 (Cat# 90329, CST), E-cadherin (Cat# 20874-1-AP, Proteintech), Vimentin (Cat# 60330-1-Ig, Proteintech), YAP (Cat# 14074, CST), LAMP1 (Cat# 99437, CST), p62 (Cat# 5114, CST), LC3 (Cat# 14600-1-AP, Proteintech) and GAPDH (Cat# HY-P80137, MCE). Subsequently, the membranes were incubated with HRP-linked secondary antibodies (Cat# 7074, CST, USA) for 1 h. Protein bands were visualized using ECL substrate (Cat# RM00021P, ABclonal) on a ChemiDoc XRS + system (Bio-Rad, USA). GAPDH served as the internal loading control for semi-quantification via ImageJ. To investigate the role of the autolysosomal pathway in VP-induced PD-L1 degradation, a rescue experiment was performed using Bafilomycin A1 (Baf A1). TDCL cells were pre-treated with 100 nM Baf A1 (Cat# S1413, Selleck Chemicals, USA) for 8 h to inhibit autophagosome-lysosome fusion. Cells were then treated with PBS, VP, or VP-αPD-L1 for an additional 24 h before being subjected to Western blot analysis as described above. To evaluate the impact of ROS scavenging on checkpoint expression in the decoupling assays, TDCL cells pre-stimulated with IFN-γ (100 ng/mL) were treated with PBS, NAC (10 mM), VP (0.5 μM), or VP + NAC (0.5 μM VP + 10 mM NAC) for 24 h, followed by protein extraction and Western blot analysis for PD-L1 and CMTM6 as described above. Transmission electron microscopy (TEM) To examine the ultrastructural formation of autolysosomes, TDCL cells treated with PBS, 0.5 μM VP, or 0.5 μM VP-αPD-L1 for 24 h were harvested, pelleted by centrifugation, and immediately fixed in 2.5% glutaraldehyde at 4 °C overnight. The cell pellets were post-fixed with 1% osmium tetroxide, dehydrated through a graded ethanol series, and embedded in epoxy resin. Ultrathin sections (approximately 70 nm) were sliced, double-stained with uranyl acetate and lead citrate, and observed under a TEM (HT7800, Hitachi, Japan). Immunofluorescence (IF) staining To visualize intracellular PD-L1 trafficking and autophagy induction, TDCL cells were seeded in confocal dishes. For the PD-L1 co-localization assay, TDCL cells expressing a PD-L1-SNAP-tag were constructed. These cells were incubated with 5 μM SNAP-Surface Alexa Fluor 488 (Cat# S9129S, New England Biolabs, USA) for 30 min at 37 °C to specifically label the cell-surface and internalized PD-L1, followed by washing to remove unbound dye. Subsequently, cells for both the co-localization and autophagy assays were subjected to the designated treatments (PBS, 0.5 μM VP, or 0.5 μM VP-αPD-L1) for 24 h. After washing with PBS, cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X-100 for 10 min. Following a blocking step with 3% BSA for 1 h, the cells were incubated with respective primary antibodies against LC3B (Cat# 83506, CST, USA), Rab5 (Cat# 3547, CST, USA), or Rab7 (Cat# 9367, CST, USA) in a humidified chamber at 4 °C overnight. Subsequently, the cells were incubated with an Alexa Fluor 568-conjugated secondary antibody (Cat# A-11004, Invitrogen, USA) or Alexa Fluor 546-conjugated secondary antibody (Cat# A-11035, Invitrogen, USA) for 1 h in the dark. Nuclei were counterstained with DAPI. Fluorescence imaging was performed using a CLSM (LSM 900, Zeiss, Germany). ImageJ software was utilized to quantify the average number of LC3B puncta per cell. RNA extraction and RT-qPCR analysis Total RNA was extracted from treated TDCL cells using the RNA-Quick Purification Kit (Cat# ES-RN001, Yishan, China). The concentration and purity of the isolated RNA were determined using a NanoDrop spectrophotometer. Subsequently, complementary DNA (cDNA) was synthesized from equal amounts of RNA utilizing the Evo M-MLV RT Premix (Cat# AG11706, Accurate Biology) on a T100 Thermal Cycler (Bio-Rad, Hercules, CA, USA). Quantitative PCR was then performed on a LightCycler 96 Real-Time PCR System (Roche Diagnostics, Basel, Switzerland) utilizing the SYBR Green assay (Cat# RK21203, ABclonal). The mRNA levels of target genes associated with lysosomal biogenesis, including Lamp1, Gla, Ctsb, Ctsd, Atp6v0d1, and Atp6v0b2, were evaluated. The gene expression levels were normalized to the internal control GAPDH. All primer sequences are listed in Supplementary Table S2. In vivo subcutaneous tumor model and treatments To establish the subcutaneous tumor model, 2 × 106 TDCL cells were subcutaneously injected into the right flank of the mice. On day 3 post-implantation, mice were randomly assigned into four groups (n = 3 per group) and received the following treatments via intraperitoneal (i.p.) injection every three days: (1) IgG control (10 mg/kg), (2) VP (5 mg/kg), (3) αPD-L1(10 mg/kg), and (4) VP-αPD-L1(5 mg/kg VP + 10 mg/kg αPD-L1). Tumor dimensions were monitored every three days using a digital caliper, and tumor volume (V) was calculated according to the formula: (L, length; W, width).(4)V=L×W22 On day 21, the mice were humanely euthanized; the tumors were then excised, photographed, and weighed for macroscopic evaluation. Establishment of the intracranial LCBM model The luciferase-expressing murine lung adenocarcinoma cell line (TDCL-Luc) was generated in-house via lentiviral transfection. C57BL/6 mice were induced with 3% isoflurane and secured in a stereotaxic frame. A burr hole was drilled, and 5 μL of TDCL-Luc cell suspension (1 × 108 cells/mL) was injected into the right striatum (coordinates: lateral 1.8 mm, longitudinal 4 mm, depth 3 mm) using a Hamilton microsyringe at a rate of 2 μL/min. To prevent cell reflux, the needle was retained in situ for 3 min before slow retraction. The incision was sutured, and mice were housed under standard SPF conditions. Successful tumor engraftment was verified by bioluminescence imaging using an IVIS Spectrum system (Caliper, USA) 14 days post-implantation following the intraperitoneal injection of D-luciferin (150 mg/kg; Cat# MB1834-2, Meilunbio, China). In vivo and ex vivo biodistribution Upon validation of tumor growth via bioluminescence, the LCBM-bearing mice were randomized into four treatment cohorts (n = 3) and intravenously administered with PBS, VP-αPD-L1, VP-αPD-L1@EB, or VP-αPD-L1@REB at an equivalent VP dose of 5 mg/kg. For real-time biodistribution profiling, the mice were anesthetized and sequentially imaged at 0.5, 1, 2, 3, 4, and 8 h post-injection utilizing an IVIS Spectrum system. At the 8 h endpoint, mice were euthanized, and major organs (heart, liver, spleen, lung, and kidney), along with the excised brains, were rinsed with cold saline for ex vivo fluorescence imaging under identical optical settings. The fluorescence intensities at the cranial sites and within isolated tissues were quantified by calculating the average radiant efficiency within defined regions of interest (ROIs) using Living Image software. In vivo anti-tumor efficacy On day 14 post-tumor inoculation, mice exhibiting successful orthotopic LCBM, as verified via bioluminescence imaging, were randomly assigned to four treatment groups (n = 5 per group). Intravenous (i.v.) therapeutic interventions, including PBS, VP-αPD-L1, VP-αPD-L1@EB, and VP-αPD-L1@REB (at an equivalent VP dose of 5 mg/kg), were administered every other day from day 15 to day 25, totaling six doses. Tumor progression was monitored via bioluminescence imaging concurrently with each treatment session. Systemic toxicity and therapeutic efficacy were further evaluated by tracking body weight fluctuations and documenting daily mortality to construct Kaplan-Meier survival curves. At the experimental endpoint, the tumor-bearing brains were harvested, fixed in 4% paraformaldehyde, embedded in paraffin, and coronally sectioned. Standard H&E staining was performed for routine histopathological evaluation. The stained sections were digitized using a NanoZoomer S360 digital slide scanner (Hamamatsu Photonics, Japan). To evaluate intratumoral apoptosis, consecutive tumor sections were subjected to a TUNEL assay using a commercial kit (Cat# 11684817910, Roche) strictly following the manufacturer's instructions. For the assessment of cell proliferation, sections were incubated with an anti-Ki67 primary antibody (Cat# 9129, CST), followed by incubation with a corresponding HRP-conjugated secondary antibody (Cat# 7074, CST). Fluorescent signals were subsequently developed using the iFluor 488 tyramide reagent (Cat# 11060, AAT Bioquest). Nuclei were counterstained with DAPI. The stained sections were visualized using a CLSM (LSM 900, Zeiss, Germany). The relative fluorescence intensities of TUNEL and Ki67 were quantitatively analyzed utilizing ImageJ software. To evaluate the relative contributions of ROS-induced cytotoxicity and PD-L1 clearance in vivo, a decoupling assay was conducted in the orthotopic LCBM model. Tumor-bearing mice were randomly assigned to four groups: Control (PBS, i.v.), NAC (i.p.), VP-αPD-L1@REB (i.v.), and VP-αPD-L1@REB + NAC. For the combination treatment, NAC was administered via intraperitoneal injection 2 h prior to the intravenous administration of the nanocomposite to ensure sufficient in vivo ROS scavenging. Tumor progression (via bioluminescence) and body weight were monitored as previously described. To evaluate intratumoral ROS levels, freshly excised brain tumor tissues were immediately embedded in OCT compound and cryosectioned. The frozen sections were incubated with a reactive oxygen species fluorescent probe, Dihydroethidium (DHE, Cat# D7008, Sigma-Aldrich) at 37 °C for 30 min in the dark, followed by DAPI counterstaining. Immunofluorescence analysis of brain tumor tissues To evaluate the tumor immune microenvironment and autophagy-related markers, brain tumor sections were deparaffinized, rehydrated, and subjected to heat-induced antigen retrieval in citrate buffer. After blocking non-specific binding with 5% BSA for 1 h at room temperature, the sections were incubated overnight at 4 °C with specific primary antibodies targeting PD-L1 (Cat# 64988, CST), LAMP1 (Cat# ab208943, Abcam), LC3B (Cat# 18725-1-AP, Proteintech), CMTM6 (Cat# 90329, CST), CD4 (Cat# ab183685, Abcam), and CD8 (Cat# ab209775, Abcam). Following thorough washing with PBS, the sections were incubated with the corresponding HRP-conjugated secondary antibodies (Cat# 7074, CST). Fluorescent signals were subsequently developed utilizing Tyramide Signal Amplification (TSA) technology with iFluor 488, iFluor 555, and iFluor 647 tyramide reagents (Cat# 11060, 11065, and 11066, respectively; AAT Bioquest) for 1 h at room temperature in the dark. Nuclei were counterstained with DAPI. Fluorescence images were acquired using a CLSM (LSM 900, Zeiss, Germany). The relative fluorescence intensities were quantitatively analyzed across multiple randomly selected high-power fields using ImageJ software. In vivo biosafety assessment On Day 28 post-inoculation, the mice were euthanized for systemic toxicity evaluation. Collected blood samples were divided into two aliquots: one was directly analyzed utilizing an automated hematology analyzer, while the other was centrifuged (1000 rpm, 10 min) to isolate serum. Serum hepatic and renal function markers were quantified via an automated biochemical analyzer. Concurrently, major organs (heart, liver, spleen, lung, and kidney) were harvested, fixed in 4% neutral formaldehyde, paraffin-embedded, and sectioned. Standard H&E staining was performed for microscopic histopathological evaluation. Statistical analysis All quantitative experimental data were presented as the mean ± standard deviation (SD) from at least three independent biological replicates (n ≥ 3). Statistical significance for comparisons between multiple groups at a single time point was evaluated using one-way analysis of variance (ANOVA) followed by Tukey's post hoc test. For data involving multiple time points, two-way ANOVA was employed. All statistical analyses were performed using GraphPad Prism software (version 9.0). A p-value of less than 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). CRediT authorship contribution statement Xiujuan Hong: Writing – original draft, Methodology, Investigation, Conceptualization. Xiaoqi Wang: Writing – original draft, Validation, Data curation. Wankun Wang: Methodology, Investigation. Weiye Ren: Validation, Investigation. Wenjia Ding: Resources, Investigation. Dashan Yin: Investigation, Data curation. Chenli Wang: Software, Data curation. Chuan Sun: Writing – review & editing, Validation. Jun Chen: Writing – review & editing, Formal analysis. Ji-Gang Piao: Writing – review & editing, Supervision, Conceptualization. Zhibing Wu: Writing – review & editing, Supervision, Project administration, Funding acquisition. Ethics approval and consent to participate All animal experiments and welfare protocols were strictly conducted in accordance with the guidelines and approval of the Institutional Animal Care and Use Committee (IACUC) of Zhejiang University (ZJU20250175). Funding This work was supported by the 10.13039/501100018684National Health Commission Scientific Research Fund of China (grant number WKJ-ZJ-2501), the 10.13039/501100004731Natural Science Foundation of Zhejiang Province (grant numbers LMS25H160022 and LY23H300002), the independently deployed project of 10.13039/501100002367Chinese Academy of Sciences Hangzhou Institute of Medicine (grant number 2024ZZBS20), and the Hangzhou Natural Science Foundation (grant numbers 2024SZRYBH160002 and 2025SZRJJ1206). Declaration of competing Interest The authors declare that they have no competing interests