A barrier-crossing and immune checkpoint-blocking antibody fusion protein for enhanced glioma-targeted immunotherapy Authors: Yanning Bao, Shengmin Yang, Yuan Ding, Jianfen Zhou, Xudong Zheng, Ruohan Chen, Zhixuan Jiang, Nana Meng, Dongsheng Pan, Jiasheng Lu, Jiacheng Guan, Jing Ye, Hongjia Zhu, Nan Xu, Cao Xie, Lei Yu, Linwei Lu, Weiyue Lu Published: 2026 ABSTRACT AbstractWith low five-year survival estimates, poor prognosis, and high recurrence probabilities, glioma is considered one of the most intractable malignant tumors. Despite the discovery of lymphatic vascular system and immune system in the central nervous system (CNS), immune checkpoint blockade therapeutics, such as programmed death ligand 1 (PDL1, also called B7H1 or CD274) antibodies, are prevented from the CNS and glioma sites due to the existence of biological barriers including the blood-brain barrier (BBB) and blood-brain tumor barrier (BBTB). Herein, we constructed a BBB/BBTB-crossing recombinant antibody by fusing the PDL1 antibody (αPDL1) and the targeting moiety RAP22 peptide (RAP22) through a matrix metalloproteinase 2 (MMP2)-responsive cleavable linker, abbreviated as αPDL1-mRAP22. Not only was αPDL1-mRAP22 able to block PD1/PDL1 pathway, reduce T cell apoptosis, enhance T cell killing ability towards glioma cells in vitro, but also it showed higher accumulation in the glioma site, prolonged survival time, and potent immune responses as well as synergistic effects with temozolomide (TMZ) in vivo, offering a novel strategy for glioma immunotherapy.Keywords: Antibody fusion protein, PDL1 antibody, Immune checkpoint blockade, RAP22 peptide, Glioma targeting delivery Introduction Glioma, accounting for 81% of malignant brain tumors, is known as the most common and harmful primary intracranial tumor, with a median survival time of less than two years and a 5-year relative survival of approximately 5% [1,2]. Some major defects come along with current traditional treatments for glioma which remain to be overcome. For example, surgery is usually not sufficient to completely remove the infiltrating tumor while radiation therapy may be prone to damage the central nervous system (CNS), and chemotherapy with small-molecule drugs such as temozolomide (TMZ) are usually not capable of achieving effective therapeutic concentration in the brain due to their poor permeability through the blood-brain barrier (BBB). To address these clinical bottlenecks, researchers have turned toward investigating the innate immune mechanisms of the CNS. Historically, there was the long-standing belief that the brain was an immune-privileged site until Louveau and colleagues reported their discovery of the structure and function of lymphatic vascular system in the CNS [3,4]. It was found that functional lymphatic vessels existing within the dural sinuses are connected to the deep cervical lymph nodes, carrying out the function of transporting cerebrospinal fluid and immune cells to the deep cervical lymph nodes, which may also play an important role in antigen presentation [3,4]. Since then, the recent years have witnessed more and more discoveries related to the immune system in the CNS. Microglia, considered the major resident immune cell population in the CNS, react to inflammatory stimuli by recognizing specific antigens and presenting them to lymphocytes via the glial-lymphoid pathway, after which immune cells are recruited across the BBB to the brain parenchyma [[5], [6], [7]]. In addition, a variety of peripheral immune cells are also found in the glioma microenvironment, including myeloid-derived suppressor cells (MDSCs), natural killer (NK) cells, tumor-associated macrophages (TAMs), tumor-associated neutrophils (TANs), CD4+ T helper cells, CD8+ cytotoxic T cells, regulatory T cells (Tregs), etc. [8,9]. Based on the presence of immune cells within the CNS, tumor immunotherapy provides new insights for the treatment of glioma, which aims to destroy tumor cells and enhance immune responses by activating the immune system to recruit immune cells and avoid tumor immune escape. Currently, tumor immunotherapies are mainly categorized into immune checkpoint inhibitors, pro-lymphocytotic cytokines, engineered T cell therapies, tumor vaccines, oncolytic viruses, and bispecific antibodies, etc. [[9], [10], [11]]. Generating a kind of typical immunosuppressive microenvironment, glioma is characterized by upregulated expression of immune checkpoint molecules including programmed death ligand 1 (PDL1), which is prone to induce immune escape by binding to programmed cell death protein 1 (PD1) on immune cells [[12], [13], [14]]. Therefore, applying PDL1 antibodies to the blockade of PD1/PDL1 pathway can theoretically reactivate anti-tumor immune effects of T cells in the tumor sites, without affecting PD1/PDL2 pathway and thus forestalling undesirable potential systemic risks which may be brought about by PD1 antibodies [13,15,16]. However, the BBB, which is mainly formed by CNS capillary endothelial cells (ECs) intertwined with tight junctions (TJs) and the blood-brain tumor barrier (BBTB), which usually occurs in the later stages of glioma progression, both limit the therapeutic effect of biological macromolecules including PDL1 antibodies by preventing substances with molecular weight over 600 Da from permeating into the brain parenchyma [13,[17], [18], [19]]. Therefore, investigation on targeted delivery of antibodies to intracranial tumor sites is of great urgence and significance. To overcome the BBB and BBTB, endogenous transport pathways that naturally facilitate molecular exchange could be exploited. For example, varieties of receptors and transporters are expressed on the surface of CNS ECs which constitute the BBB, such as transferrin receptor, insulin receptor and insulin-like growth factor receptor, low-density lipoprotein receptor, glucose transporter, diphtheria toxin receptor, glutathione transporter, etc. [[18], [19], [20]]. Interactions between these receptors or transporters and their specific ligands can mediate transcytosis. Therefore, modifying drugs with ligands that can specifically bind to the surface receptors on the BBB/BBTB confers potentially active targeting ability towards glioma, which allows drugs to cross the biological barriers and exert therapeutic effect in the tumor sites [[21], [22], [23], [24], [25], [26], [27]]. Based on the above mechanism, current strategies for antibodies and other macromolecular therapeutic drugs to cross the BBB and BBTB are mainly categorized into three types: (i) molecule-mediated delivery strategies, including recombinant antibodies and small molecule/peptide-antibody conjugates which are modified with targeting molecules that can interact with the receptors on the BBB/BBTB; (ii) carrier-mediated delivery strategies in which carriers are modified with targeting moieties; (iii) other delivery strategies such as gene delivery, focused ultrasound or peptides transiently opening the BBB/BBTB [18]. However, carrier-mediated delivery strategies such as nanoparticles may suffer from the clearance of reticuloendothelial system, the immunogenicity and toxicity of the nano-based materials, and the obstacles of complex manufacturing and clinical translation. Disruption techniques such as focused ultrasound or transient BBB/BBTB-opening peptides may pose risks of uncontrolled solute entry. Also, gene delivery may bring about unpredictable neurotoxicity. In this case, the strategy of recombinant antibody fusion protein emerges as a promising alternative. The application of antibody fusion protein-based delivery is mainly based on efficient endogenous receptor-mediated transcytosis (RMT). In this context, low density lipoprotein receptor-related protein 1 (LRP1), which is highly expressed on both the BBB and BBTB, would be suitable for facilitating antibody fusion protein to cross the BBB and BBTB via RMT [20,21,28,29]. Receptor-associated protein (RAP) is an endoplasmic reticulum (ER) protein with a molecular weight of 39 kDa, which effectively binds to LRP1 and participates in its correct folding and transportation process. Pan and colleagues demonstrated that mediated by RAP, therapeutic peptides and proteins were likely to cross the BBB [30]. Based on the amino acid residues where LRP1 interacts with RAP, we previously miniaturized RAP into RAP22 peptide as an effective glioma targeting moiety [31]. Consequently, by fusing RAP22 to PDL1 antibody, LRP1-based RMT could be harnessed to enable the BBB and BBTB penetration. Meanwhile, since LRP1 is also reported to express on glioma cells, it may mediate the cellular uptake of the antibodies into tumor cells, thereby preventing them from blocking immune checkpoint pathways on the cell surface [28]. To mitigate the interference of RAP22 on the antibody’s surface-blocking activity, an MMP2-responsive cleavable linker was incorporated between the peptide and the antibody, based on the high expression of MMP2 within the glioma microenvironment [32,33]. Therefore, in this study, we constructed a barrier-crossing recombinant antibody by fusing the PDL1 antibody and the RAP22 peptide through an MMP2-responsive cleavable linker. After crossing the BBB and BBTB via RAP22 mediation, the antibody fusion protein αPDL1-mRAP22 was able to release PDL1 antibodies upon the cleavage by MMP2 existing in the glioma microenvironment, so as to block the immune checkpoint pathway, inhibit tumor immune escape, activate T cell immune responses, and finally achieve enhanced therapeutic effect for glioma immunotherapy as well as synergistic effects with TMZ (Fig. 1). Schematic illustration of preparation and functions of αPDL1-mRAP22. (A) Process of αPDL1-mRAP22 preparation; (B) Functions of αPDL1-mRAP22, including BBB/BBTB transcytosis via the mediation of RAP22 and αPDL1 release under the cleavage of MMP2, aiming for glioma-targeted immunotherapy as well as enhanced therapeutic effect with TMZ combination. The illustration was created with BioRender. Materials and methods Materials The heavy chain and light chain recombinant plasmids of antibodies were constructed and produced by Tsingke (Beijing, China). Expi293 expression kit, fixable dead cell stain kit and Foxp3/Transcription factor staining buffer were purchased from Thermo Fisher Scientific (MA, USA). Protein A affinity chromatography columns were purchased from Sepax (DE, USA). Recombinant mouse CD274 (i.e., PDL1) protein with His-tag was supplied by Sino Biological (Beijing, China). Recombinant mouse MMP2 protein with His-tag was purchased from ACRO Biosystems (Beijing, China). Recombinant murine interferon-gamma (IFN-γ) was obtained from PeproTech (NJ, USA). APC-conjugated anti-IgG antibody, mouse CD3 T cell isolation kit, buffer, anti-mouse CD3ε, anti-mouse CD28, and fluorescein-conjugated antibodies against mouse PDL1, CD3, CD8a, IFN-γ, tumor necrosis factor-alpha (TNF-α), CD44, CD62L, Ki67, PD1, CD4, forkhead box protein 3 (Foxp3), CD25 were purchased from BioLegend (CA, USA). Annexin V-FITC/PI apoptosis detection kit and lactate dehydrogenase (LDH) assay kit were purchased from Beyotime (Shanghai, China). Cy5 antibody conjugation kit was purchased from Bioss (Beijing, China). D-luciferin potassium salt and TMZ were obtained from MeilunBio (Dalian, China). Anti-MMP2, anti-PDL1 and anti-LRP1 antibodies for Western blot assay were provided by Abcam (Cambridge, UK). Horseradish peroxidase (HRP)-conjugated secondary antibodies for enzyme-linked immunosorbent assay (ELISA) and Western blot assay were purchased from ZSGB-BIO (Beijing, China). IFN-γ, TNF-α, interleukin-6 (IL-6), and interleukin-1β (IL-1β) ELISA kits were provided by Servicebio (Wuhan, China). Mouse glioma cells (GL261) and mouse monocyte macrophage leukemia cells (RAW264.7) were kindly provided by Prof. Zhan Changyou (Fudan University, China), and GL261 cells were stably transfected with luciferase-expressing genes to obtain GL261-luc cells. Mouse brain capillary endothelial cells (bEnd.3) and human umbilical vascular endothelial cells (HUVECs) were obtained from Shanghai Institute of Biochemistry and Cell Biology. All cell lines were cultured in Dulbecco’s modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. Primary CD3+ T cells extracted from mouse spleens were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium with 5 µg/mL CD28 antibody. Primary rat brain capillary endothelial cells (BCECs) were cultured in endothelial cell growth basal medium-2 (EBM-2). Cells were all cultured at 37 °C under a humidified atmosphere containing 5% CO2. Male C57BL/6 mice and SD rats were provided by Shanghai Sippr-BK Laboratory Animal Co., Ltd. (Shanghai, China) and kept under specific pathogen-free (SPF) conditions. All the animal experiments were conducted according to regulations and guidelines approved by the ethics committee of Fudan University. Western blot assay To characterize PDL1 expression on the cell surface of glioma tissues and MMP2 expression in the glioma microenvironment, GL261-luc glioma tissues were collected from glioma-bearing mice and then homogenized in phosphate buffered saline (PBS). Protein samples extracted from the tissues were loaded onto the gels for sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) (135 V, 1 h), and then the gels were placed in ice bath for blot transfer onto polyvinylidene fluoride (PVDF) membranes. After that, the membranes were incubated with 5% skimmed milk at room temperature for 1 h, washed with TBST (Tris–HCl buffered saline containing 0.1% Tween 20), and then incubated with anti-PDL1 antibody or anti-MMP2 antibody at 4 °C overnight. The primary antibodies were washed away with TBST, and the membranes were subsequently incubated with HRP-conjugated secondary antibodies at room temperature for 1 h. After adding enhanced chemiluminescence (ECL) solution, the membranes were visualized by gel imaging system (ProteinSimple, USA), and semi-quantification was conducted by ImageJ (version 1.53a). LRP1 expression on the surface of BCECs, bEnd.3 cells and HUVECs was also verified by Western blot assay as described above. Plasmid construction The sequence of brain tumor targeting moiety RAP22 peptide (EAKIEKHNHYQKQLEIAHEKLR) was fused to both N-terminus and C-terminus of the heavy chain sequence of αPDL1 as well as to mere N-terminus of the light chain sequence of αPDL1. The linker (GGGGSGPLGIAGQGGGGS) between the peptide and the antibody was designed to be responsively cleavable by MMP2. Next, the above fused heavy chain and light chain sequences were cloned into eukaryotic expression vectors to acquire recombinant plasmids of the antibody fusion protein αPDL1-mRAP22. Protein-protein interaction prediction The amino acid sequence of PDL1 was obtained from the UniProt database. Then, AlphaFold 3 was used to predict the 3D structure of the complex formed by the heavy and light chain variable regions of the antibodies and PDL1. The quality of the predicted complex models was evaluated based on the confidence scores including pLDDT and ipTM. The optimal complex model was energy-minimized to refine its geometry. After that, the molecular mechanics/generalized Born surface area (MM/GBSA) module on the HawkDock Server 2 was used to estimate the binding free energy (△G) between the antibodies and PDL1, which helped to assess their interaction [[34], [35], [36], [37], [38], [39]]. Finally, PyMOL (version 3.0.3) was used to visualize the binding interface and identify key amino acid residues involved in the interaction. Antibody expression and purification The heavy chain and light chain recombinant plasmids of antibodies (molar ratio = 1:1) were transfected into Expi293F cells (human embryonic kidney cells) according to the manufacturer’s protocol on Day 0, and then cell culture supernatant was harvested and purified by ӒKTA pure™ (Cytiva, USA) via protein A affinity chromatography on Day 7. Antibody characterization Molecular weight characterization MMP2 was diluted to certain concentrations with TCNB buffer (50 mM Tris, 10 mM CaCl2, 150 mM NaCl, 0.05% (w/v) Brij 35, pH 7.5) and added with 4-aminophenylmercuric acetate (APMA) at 37 °C for 2 h. Activated MMP2 at different concentrations (10, 5, 1 and 0.1 µg/ml) was incubated with αPDL1 or αPDL1-mRAP22 at 37 °C for 3 h to cleave the enzyme-responsive linker. After that, SDS-PAGE was carried out (120 V, 90 min), and the gels were then stained with Coomassie brilliant blue solution for 30 min, and decolorized with ultrapure water overnight. For matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS), 50% acetonitrile solution containing 0.1% trifluoroacetic acid was prepared, and 1 mg erucic acid was dissolved in 100 µl of this solution, which was used as the substrate for MALDI-TOF. After in vitro activation of MMP2 and in vitro cleavage of αPDL1-mRAP22 with above methods, the substrate was mixed with αPDL1, αPDL1-mRAP22, and αPDL1-mRAP22 after MMP2 cleavage (abbreviated as αPDL1-mRAP22/MMP2) at the volume ratio of 1:1, and 1 µl of the sample was dispensed on the plate for co-crystallization formation after evaporation drying. The accurate molecular weights of αPDL1, αPDL1-mRAP22, and αPDL1-mRAP22/MMP2 were detected by AB Sciex 5800 (AB SCIEX, USA) via high molecular weight linear cation mode. In vitro affinity characterization For ELISA, CD274 protein at a concentration of 0.5 µg/ml was first coated onto a high binding ELISA plate for overnight incubation at 4 °C. The plate was washed with PBST (phosphate buffered saline containing 0.05% Tween 20) and blocked by 2% skimmed milk at room temperature for 1 h. After PBST washing, αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 diluted to a series of gradient concentrations were added into the plate for 1 h incubation. HRP-conjugated secondary antibody was then added for 1 h incubation. Finally, the plate was stained with tetramethylbenzidine (TMB) for 10 min, and stopped with 2 M H2SO4 solution. The optical density (OD) values were detected by microplate reader (Biotek, USA) at 450 nm wavelength. The affinity curves were fitted by GraphPad Prism (version 9.0.2) via nonlinear regression analysis. The equilibrium dissociation constant (KD) values were further measured by surface plasmon resonance (SPR) assay via single-cycle kinetics. After SPR chip activation by NHS-EDC, CD274 protein was coupled onto the chips. αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 diluted to a series of gradient concentrations were injected and real-time response units were detected by WeSPR™ One (XLement, China). The association constant (ka), the dissociation constant (kd), and KD were calculated by WeSPR™ One analyzing software (version 5.3). In vitro PD1/PDL1 pathway blockade For PDL1 binding assay, GL261-luc cells were first stimulated with 100 ng/ml murine IFN-γ for 48 h to upregulate PDL1 expression on the surface. Cells were incubated with αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 of the same concentrations at 4 °C for 1 h, washed with PBS, and then incubated with APC-conjugated anti-IgG secondary antibody at 4 °C for another 1 h. For PDL1 blocking assay, GL261-luc cells with upregulated PDL1 expression were incubated with αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 of the same concentrations at 4 °C for 1 h, washed with PBS, and then incubated with APC-conjugated PDL1 antibody at 4 °C for another 1 h. Positive percentages and mean fluorescence intensity of GL261-luc cells in above assays were measured by flow cytometer (Cytek Biosciences, USA), and data were analyzed by FlowJo (version 10.4). Meanwhile, GL261-luc cells stimulated with murine IFN-γ were seeded in confocal dishes, fixed with 4% paraformaldehyde, nucleus-stained with diamidino-2-phenylindole (DAPI), and incubated with αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 for 1 h. After washing, cells were subsequently incubated with APC-conjugated PDL1 antibody for another 1 h and finally observed by laser scanning confocal microscope (Keyence, China). For PD1/PDL1 pathway blockade functioning assays, CD3+ T cells extracted from spleens of C57BL/6 mice according to the kit manufacturer’s protocol were co-cultured with GL261-luc cells with upregulated PDL1 expression, and then added with αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 for overnight incubation. Annexin V-FITC/PI apoptosis detection kit and LDH assay kit were used to quantify T cell apoptosis and killing efficiency towards GL261-luc cells, respectively. In vitro cellular uptake BCECs extracted in accordance with methods in previous literature [40] and HUVECs were seeded in 12-well plates or confocal dishes, and incubated with Cy5-conjugated αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 of the same concentrations at 37 °C for 2 h. Cellular uptake was then observed by confocal microscope and measured by flow cytometer. In vitro BBB/BBTB-crossing ability BCECs were seeded in the upper chambers of transwell plates to form the BBB in vitro. Transendothelial electrical resistance (TEER) was detected to evaluate monolayer integrity. BCECs with TEER of over 300 Ω⋅cm2 were used for further experiments. αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 of the same concentration were added into the upper chambers for incubation, and samples from the lower chambers were collected at 0.5, 1, 1.5, 2, 3 and 4 h after administration. The cumulative transport volumes at different timepoints were then quantified by ELISA. In vitro BBTB assay was carried out likewise, except that HUVECs and GL261-luc cells were seeded in the upper chambers and lower chambers of transwell plates respectively to form the BBTB in vitro. Apparent permeability coefficient (Papp) was calculated according to the equation below: Where dQ/dt refers to drug transport rate per unit time, A refers to the area of the transport membrane, and C0 refers to the initial drug concentration. In vivo targeting ability Male C57BL/6 mice about six weeks of age were used to establish the orthotopic GL261-luc glioma-bearing mouse model according to previous literature [41]. Briefly, 5 × 105 GL261-luc cells dispersed in 5 µl PBS were injected into the right striatum of each mouse brain with the help of stereotaxic apparatus. On Day 14 after implantation, mice were randomly divided into two groups receiving Cy5-conjugated αPDL1 (Cy5-αPDL1) or Cy5-conjugated αPDL1-mRAP22 (Cy5-αPDL1-mRAP22) via tail vein, and in vivo fluorescence images were captured by IVIS Spectrum (PerkinElmer, USA) at 2, 4, 6, 8, 12 and 24 h after single intravenous injection. At 24 h, mice were sacrificed followed by cardiac perfusion with PBS, and brains were harvested for ex vivo imaging. Semi-quantitative analysis of average radiance efficiency was conducted by Living Image (version 4.4). For in vivo biodistribution study, orthotopic GL261-luc glioma-bearing C57BL/6 mice were intravenously injected with 125I-αPDL1 or 125I-αPDL1-mRAP22 (15 µCi per mouse). At 6 and 24 h, mice were sacrificed followed by cardiac perfusion with PBS. Brain, heart, liver, spleen, lung and kidney of each mouse were harvested and weighed, and γ counts of the tissues were measured by gamma immunoassay counter (PerkinElmer, USA). To investigate in vivo targeting mechanism, 6 h after intravenous injection of Cy5-αPDL1 or Cy5-αPDL1-mRAP22 into orthotopic GL261-luc glioma-bearing C57BL/6 mice, brains were harvested and prepared for frozen slices to be stained with DAPI, along with CD31, LRP1 or PDL1. The frozen slices were then observed by confocal microscope. In vivo MMP2 cleavage and plasma stability Orthotopic GL261-luc glioma-bearing C57BL/6 mice were intravenously injected with αPDL1 or αPDL1-mRAP22 at a dose of 10 mg/kg. After 12 h, serum samples were collected, and after 24 h, brains were harvested and homogenized for protein extraction. Western blot assay was conducted to detect antibody status via molecular weight characterization. In vivo pharmacokinetics Healthy male C57BL/6 mice were randomly divided into two groups receiving αPDL1 or αPDL1-mRAP22 at a dose of 10 mg/kg via tail vein, and blood samples were collected from orbital venous plexus at 5, 15 and 30 min, and 1, 2, 4, 6, 8 and 12 h after administration. Plasma was collected after centrifugation (3000 rpm, 15 min, 4 °C), and plasma concentration of antibodies was determined by ELISA. Pharmacokinetic parameters were calculated by DAS 2.0. In vivo anti-glioma efficacy The orthotopic GL261-luc glioma-bearing mouse model was established as mentioned in Section 2.10 on Day 0, and mice received PBS, αPDL1 or αPDL1-mRAP22 treatment (single dose equals to 10 mg/kg αPDL1) via tail vein on Day 5, 7, 9 and 11. Body weight and survival time of mice were recorded, and bioluminescence images were captured on Day 5, 8, 12, 15, 19 and 22 by IVIS Spectrum. In addition, at the end of treatment, brains were harvested for Ki67 immunohistochemical staining and TUNEL immunofluorescence staining to assess glioma cell proliferation and apoptosis. In vivo immune responses At the end of treatment as described in Section 2.13, orthotopic GL261-luc glioma-bearing C57BL/6 mice were sacrificed to harvest glioma tissues, cervical lymph nodes and spleens for preparing single-cell suspension. After flow cytometry antibody staining, percentages of various immune cells in the aforementioned tissues were quantified by flow cytometry to evaluate immune responses triggered by different treatment regimens. In vivo safety At the end of treatment as described in Section 2.13, orthotopic GL261-luc glioma-bearing C57BL/6 mice were sacrificed to collect whole blood samples for complete blood count. Meanwhile, another portion of whole blood samples was centrifuged (3000 rpm, 15 min, 4 °C), and serum was obtained for detection of pro-inflammatory cytokines by ELISA as well as biochemical parameters concerning liver and kidney functioning. In addition, heart, liver, spleen, lung and kidney of mice were harvested for hematoxylin-eosin (H&E) staining. In vivo anti-glioma efficacy of combination treatment regimen The orthotopic GL261-luc glioma-bearing mouse model was established as mentioned in Section 2.10 on Day 0, and mice received PBS, αPDL1, αPDL1-mRAP22, TMZ, TMZ + αPDL1 or TMZ + αPDL1-mRAP22 treatment on Day 5, 7, 9 and 11. PBS or antibodies were intravenously administered at a single dose of 10 mg/kg, while TMZ was orally administered at a single dose of 10 mg/kg. Body weight and survival time of mice were recorded, and bioluminescence images were captured on Day 5, 8, 12, 15 and 19 by IVIS Spectrum. In addition, at the end of treatment, brains were harvested for Ki67 immunohistochemical staining as well as TUNEL, calreticulin (CRT), and high mobility group box 1 (HMGB1) immunofluorescence staining to assess glioma cell proliferation and apoptosis as well as immunogenic cell death (ICD). Heart, liver, spleen, lung and kidney of mice were harvested for H&E staining. Specifically, Jin’s formula was applied to evaluate whether combination of TMZ and αPDL1-mRAP22 achieved a synergistic effect [42]. q > 1.15 indicates synergy, q between 0.85 and 1.15 indicates additive effect, q < 0.85 indicates antagonism: Where EA, EB and EA+B refers to the prolongation rate of median survival time of drug A monotherapy group, drug B monotherapy group, and drug A & B combination therapy group. The prolongation rate (E) of median survival time was calculated according to the equation below: E (%) = (median survival time of the treatment group—median survival time of the control group)/median survival time of the control group × 100%. Statistical analysis Data were presented as mean ± standard deviation (SD). Statistical significance was calculated by GraphPad Prism 9, with unpaired Student’s t-test and one-way analysis of variance (ANOVA) for two-group and multiple-group comparison, respectively. P < 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, ns: no significance). Results and discussion Preparation and characterization of MMP2-responsive αPDL1-mRAP22 Since αPDL1-mRAP22 was designed to release αPDL1 upon MMP2 cleavage in the glioma site and target PDL1 on the cell surface of glioma tissues after crossing the BBB/BBTB by transcytosis triggered by the interaction between RAP22 and LRP1, we first detected PDL1 and MMP2 expression in the orthotopic GL261-luc glioma-bearing C57BL/6 mice (Fig. S1 and S2). As shown in Western blot analysis and immunohistochemical staining, there was no obvious PDL1 or MMP2 expression in normal brain tissues, but elevated PDL1 expression and continuous MMP2 excretion in GL261-luc glioma tissues, providing prerequisite for further investigation of αPDL1-mRAP22. The heavy chain sequence of αPDL1-mRAP22 was designed by combining the sequence of RAP22 to both N-terminus and C-terminus of the heavy chain of αPDL1, while the light chain sequence only comprised RAP22 sequence at the C-terminus of αPDL1 light chain (Fig. 2A). The heavy chain and light chain sequences were then cloned into eukaryotic expression vectors respectively. To evaluate whether RAP22 modification interfered with the αPDL1 binding sites, we conducted protein-protein interaction prediction. Structural analysis confirmed that RAP22 was located distantly from the binding pocket (Fig. 2B). Key CDR interactions with PDL1 residues (e.g., D73, Q66, R113 and D61) remained nearly identical across both models (Fig. 2B and 2C). The calculated △G for αPDL1-mRAP22-scFv and αPDL1-scFv were −69.95 kcal/mol and −65.91 kcal/mol, respectively. These results indicated that, theoretically, RAP22 modification did not compromise the thermodynamic stability of the antibody-antigen complex. Preparation and characterization of MMP2-responsive αPDL1-mRAP22. (A) Structure of αPDL1-mRAP22; Protein-protein interaction prediction of PDL1 (light pink) with (B) αPDL1-mRAP22-scFv (light blue) and (C) αPDL1-mRAP22-scFv (light blue), with interacting amino acid residues in stick model and hydrogen bonds in yellow dashed lines; (D) Coomassie brilliant blue staining of reduced αPDL1 and αPDL1-mRAP22 before and after MMP2 treatment with different concentrations; (E) MALDI-TOF MS of αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage; (F) Mouse PDL1 binding curves of αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage; (G) Mouse PDL1 kinetic binding curves and constants of αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage. Data were presented as mean ± SD (n = 3). As shown in SDS-PAGE, MMP2 cleaved the enzyme-responsive linker of αPDL1-mRAP22 in a dose-dependent manner which happened completely at a concentration of 10 µg/ml MMP2 (Fig. 2D). It was further confirmed by MALDI-TOF MS that the enzyme-responsive linker of αPDL1-mRAP22 could be cleaved by MMP2, restoring the molecular weight of αPDL1-mRAP22 to approximately the same as that of αPDL1 (Fig. 2E). Based on the structural confirmation of linker cleavage, we subsequently investigated whether the removal of RAP22 could effectively restore the pharmacological functionality of the antibody. As shown in ELISA, RAP22 cleavage by MMP2 improved antibody binding affinity to the degree that was comparable to unmodified αPDL1 (Fig. 2F). However, considering that ELISA may be methodologically susceptible to steric effects and is unable to calculate an absolute affinity constant, SPR assay was further carried out. αPDL1 bound to CD274 with a KD of 1.31E-10 M, which was slightly weaker than that of αPDL1-mRAP22 either before or after MMP2 cleavage, with a KD of 5.03E-11 M and 7.54E-11 M respectively (Fig. 2G). Generally, both assays indicated that modification of RAP22 had little negative effect on the interaction between αPDL1 and CD274 protein. In vitro PD1/PDL1 pathway blockade by αPDL1-mRAP22 Beyond the characterization of protein-protein interactions, we subsequently investigated the cellular binding dynamics of αPDL1-mRAP22. This allowed for a more comprehensive assessment of its targeting efficiency towards tumor cells. The affinity of antibodies to PDL1 on the surface of GL261-luc cells was assessed by flow cytometry. Although GL261-luc glioma tissues harvested from C57BL/6 mice was proved to express a high level of PDL1, during in vitro cell culture, 100 ng/ml murine IFN-γ was needed to stimulate GL261-luc cells for upregulating PDL1 expression [43]. Consistent with ELISA and SPR assay results, RAP22 scarcely affected binding of αPDL1 to PDL1 on the surface of GL261-luc cells, with positive percentage over 99% for either αPDL1 or αPDL1-mRAP22 before and after MMP2 cleavage (Fig. 3A). Next, PDL1 blockade was also assessed by flow cytometry. In this experiment, GL261-luc cells were pre-incubated with αPDL1 or αPDL1-mRAP22 (with/without MMP2 cleavage) to block surface PDL1, followed by incubation with APC-conjugated αPDL1. Quantitative results of flow cytometry showed that αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 significantly blocked PDL1 on the surface of GL261-luc cells compared to IgG as negative control, thus preventing binding of APC-conjugated αPDL1 (Fig. 3B). Consistent with flow cytometry results, little binding of APC-conjugated αPDL1 to GL261-luc cells pre-incubated with αPDL1, αPDL1-mRAP22 or αPDL1-mRAP22/MMP2 was observed under laser scanning confocal microscope (Fig. 3C). In vitro PD1/PDL1 pathway blockade by αPDL1-mRAP22. (A) Flow cytometry analysis of positive percentage and mean fluorescence intensity of GL261-luc cells binding to αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage; (B) Flow cytometry analysis of PDL1 blockade ability of αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage; (C) Images by laser scanning confocal microscope validating PDL1 blockade ability of αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage. Blue: DAPI. Red: APC-conjugated αPDL1. Scale bar: 50 µm; (D) CD3+ T cell apoptosis when co-cultured with GL261-luc cells and treated with IgG, αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage; (E) CD3+ T cell killing efficiency towards GL261-luc cells at 24 h after the treatment of IgG, αPDL1, and αPDL1-mRAP22 before and after MMP2 cleavage. Data were presented as mean ± SD, n = 3, *P < 0.05, ***P < 0.001, ns: no significance. Having confirmed the target-binding occupancy of αPDL1-mRAP22, we next explored whether this interaction could successfully lead to the functional inhibition of the immunosuppressive PD1/PDL1 pathway. To assess this, CD3+ T cell apoptosis was further detected after PD1/PDL1 pathway blockade. It turned out that the apoptosis rate of CD3+ T cells in IgG group was about 24.58%, while that was about 8.12% for αPDL1 group, 10.46% for αPDL1-mRAP22 group, and 9.29% for αPDL1-mRAP22/MMP2 group, respectively, confirming reduced CD3+ T cell apoptosis after αPDL1 or αPDL1-mRAP22 treatment (Fig. 3D). Besides, it was interesting to detect a higher proportion of CD3+PD1+ T cells in αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 group than in IgG group, which may be due to compensatory upregulation of PD1 on T cell surface after effective blockade of PD1/PDL1 pathway by αPDL1 or αPDL1-mRAP22 (Fig. S3). Killing efficiency, which is more direct to assess T cell tumor-killing ability, was further investigated. CD3+ T cells were co-cultured with GL261-luc cells at different ratios of 2.5:1 and 5:1, along with αPDL1, αPDL1-mRAP22 and αPDL1-mRAP22/MMP2 at the same time. LDH cytotoxicity assay was conducted to quantitatively measure T cell killing efficiency towards glioma cells. It was noted that increasing the CD3⁺ T cells to GL261-luc cells ratio from 2.5:1 to 5:1 enhanced killing efficiency across all groups, with values of killing efficiency rising from 11.02%−13.99% to 16.19%−17.23%. These results further confirmed that both αPDL1 and αPDL1-mRAP22 could effectively inhibit PD1/PDL1 interaction between T cells and GL261-luc cells in vitro. Moreover, RAP22 cleavage by MMP2 further restored antibody blocking ability and subsequently enhanced tumor-killing ability of T cells (Fig. 3E). In vitro cellular uptake and BBB/BBTB-crossing ability of αPDL1-mRAP22 Since the effective cellular internalization and subsequent barrier penetration of αPDL1-mRAP22 was highly dependent on LRP1-based RMT, the expression of LRP1 on the surface of BCECs and bEnd.3 cells (BBB-related cells) as well as HUVECs (BBTB-related cells) was initially validated by Western blot assay (Fig. S4). BCECs and HUVECs were incubated with Cy5-conjugated αPDL1 and αPDL1-mRAP22 for quantification of cellular uptake efficiency. With higher positive percentage and mean fluorescence intensity, RAP22 modification effectively mediated cellular uptake of αPDL1-mRAP22 by BCECs and bEnd.3 cells, while unmodified αPDL1 was barely taken up (Fig. 4A and 4B). Corroborated results were also visualized by confocal microscope, which likewise reflected the barrier-crossing potential of αPDL1-mRAP22 (Fig. 4C). In vitro cellular uptake and BBB/BBTB-crossing ability of αPDL1-mRAP22. (A) Flow cytometry analysis of αPDL1 and αPDL1-mRAP22 uptake by BCECs; (B) Flow cytometry analysis of αPDL1 and αPDL1-mRAP22 uptake by HUVECs; (C) Laser scanning confocal microscope images of αPDL1 and αPDL1-mRAP22 uptake by BCECs and HUVECs (orange-boxed areas representing the magnified regions). Blue: DAPI. Red: Cy5, Cy5-αPDL1 or Cy5-αPDL1-mRAP22. Scale bar: 10 µm; (D) Schematic illustration of the in vitro BBB/BBTB models; (E) Percentage of αPDL1 and αPDL1-mRAP22 crossing the BBB in vitro; (F) Percentage of αPDL1 and αPDL1-mRAP22 crossing the BBTB in vitro; (G) Papp of αPDL1 and αPDL1-mRAP22 crossing the BBB/BBTB in vitro. Data were presented as mean ± SD, n = 3, *P < 0.05, **P < 0.01, ***P < 0.001. The cellular uptake profiles provided the basis for evaluating the subsequent transcellular transport efficiency. To this end, the integrity of the barrier was simulated through the establishment of the in vitro BBB and BBTB models, and then the concentration of αPDL1 and αPDL1-mRAP22 in the lower chambers was quantified by ELISA (Fig. 4D). The BBB and BBTB transcytosis efficiency of αPDL1-mRAP22 was higher than that of αPDL1 at all detected timepoints, especially after 1.5 h (Fig. 4E and 4F). The Papp of αPDL1-mRAP22 crossing the BBB (5.46E-06 ± 2.03E-06 cm/s) increased by an order of magnitude when compared to that of αPDL1 (0.23E-06 ± 0.21E-06 cm/s), and the Papp of αPDL1-mRAP22 crossing the BBTB (9.21E-06 ± 0.15E-06 cm/s) exceeded that of αPDL1 (3.18E-06 ± 0.61E-06 cm/s) by approximately 1.9 times (Fig. 4G). The Papp in the BBTB was generally higher than that in the BBB for both antibodies, reasonably attributed to higher leakiness of the BBTB compared with the BBB. In vivo targeting ability and pharmacokinetics of αPDL1-mRAP22 Based on the cellular internalization and transcytosis observed in vitro, brain-targeting efficiency and pharmacokinetic behavior were then assessed in vivo. Cy5-αPDL1 and Cy5-αPDL1-mRAP22 were intravenously administered into orthotopic GL261-luc glioma-bearing C57BL/6 mice. Fluorescence images taken by IVIS Spectrum and related semi-quantitative analysis of in vivo radiant efficiency at observed timepoints after administration showed that Cy5-αPDL1-mRAP22 reflected much stronger fluorescent signal in the glioma region while no obvious signal was observed in Cy5-αPDL1 group at all imaging timepoints (Fig. 5A and 5B). Ex vivo images of brains and semi-quantitative analysis of radiant efficiency suggested that the accumulation of Cy5-αPDL1-mRAP22 in the glioma site was about 2.94 times of that of Cy5-αPDL1 (Fig. 5A and 5C). In vivo targeting ability of αPDL1-mRAP22. (A) In vivo fluorescence imaging of orthotopic GL261-luc glioma-bearing C57BL/6 mice at different timepoints and ex vivo fluorescence imaging of brains at 24 h after single intravenous injection of Cy5-αPDL1 or Cy5-αPDL1-mRAP22; Semi-quantitative analysis of in vivo radiant efficiency of orthotopic GL261-luc glioma-bearing C57BL/6 mice at (B) different timepoints and (C) ex vivo radiant efficiency of brains at 24 h after single intravenous injection of Cy5-αPDL1 or Cy5-αPDL1-mRAP22; (D) Distribution of 125I-αPDL1 and 125I-αPDL1-mRAP22 in glioma tissues of orthotopic GL261-luc glioma-bearing C57BL/6 mice at 6 and 24 h after single intravenous injection; Colocalization of Cy5-αPDL1 and Cy5-αPDL1-mRAP22 with (E) CD31 or (F) LRP1 in the glioma site of orthotopic GL261-luc glioma-bearing C57BL/6 mice at 6 h after intravenous injection. Blue: DAPI. Green: CD31 or LPR1. Yellow: Cy5-αPDL1 or Cy5-αPDL1-mRAP22. Scale bar: 50 µm; (G) Pharmacokinetic curves and (H) parameters of αPDL1 and αPDL1-mRAP22 in healthy C57BL/6 mice. Data were presented as mean ± SD (n = 3), *P < 0.05, **P < 0.01, ***P < 0.001. To accurately quantify drug accumulation in glioma and other tissues, antibodies were labeled with 125I to carry out biodistribution study in vivo. By substituting 125I for H atoms on the aromatic rings of tyrosine, histidine, and tryptophan residues in proteins, iodine labeling has less impact on the protein backbone compared to fluorescence labeling, which helps demonstrate distribution behavior in a way closer to that of the original protein [44]. Orthotopic GL261-luc glioma-bearing mice were intravenously injected with 125I-labeled antibodies, followed by γ-counting of glioma and major organs after 6 and 24 h. The results showed that the accumulation of αPDL1-mRAP22 in the glioma site was about 5 times of that of αPDL1 at both 6 and 24 h, whereas their accumulation was similar in other major organs (Figs. 5D and S5). Relatively high distribution of antibodies in the spleen might be attributed to their interaction with PDL1 expressed on the surface of a variety of lymphocytes, such as T cells, B cells, NK cells, macrophages, dendritic cells (DCs), etc., in the spleen [[45], [46], [47]]. The significant brain-targeting efficiency observed from glioma and organs prompted a deeper investigation into the cellular entry mechanism. In vivo targeting mechanism was further visualized by immunofluorescence. αPDL1-mRAP22 was proved to be colocalized with CD31 and LRP1 as well as exist in the nearby areas, indicating that it could target glioma neovasculature and then penetrate into the glioma microenvironment via LRP1 transport (Fig. 5E and 5F). Colocalization of αPDL1-mRAP22 with PDL1 reflected effective PDL1 pathway blockade while little trace of unmodified αPDL1 was found in the glioma site (Fig. S6). In vivo MMP2 cleavage of αPDL1-mRAP22 in the glioma microenvironment was also confirmed by Western blot assay (Fig. S7). Both αPDL1 and αPDL1-mRAP22 were defected in the tumor homogenates of the mouse brains from αPDL1-mRAP22 group, indicating that αPDL1-mRAP22 effectively reached brain tumor site where it underwent cleavage mediated by MMP2 within the tumor environment and subsequently released free αPDL1. The above results validated that αPDL1-mRAP22 had ideal glioma targeting and immune checkpoint blocking ability. Meanwhile, in vivo plasma stability assay showed that αPDL1-mRAP22 remained predominantly in its intact form in the plasma at 12 h after administration, indicating systemic stability prior to brain entry and suggesting that non-specific, MMP2-independent cleavage was negligible in the circulation (Fig. S8). These findings further elucidated that the MMP2-responsive cleavage of αPDL1-mRAP22 occurred primarily after brain entry, mediated by the RAP22 peptide. Furthermore, in vivo pharmacokinetic behavior of αPDL1-mRAP22 in healthy C57BL/6 mice were similar to that of αPDL1. No significant difference was found in the plasma concentration at any of the measured timepoints between αPDL1-mRAP22 and αPDL1 groups. Also, parameters including area under the curve (AUC), mean residence time (MRT), and clearance (CLz) were also similar between the two groups, indicating that modification of RAP22 barely affected major pharmacokinetic profile of αPDL1 in mice (Fig. 5G and 5H). In vivo anti-glioma efficacy of αPDL1-mRAP22 The superior brain-targeting capability and glioma site-specific accumulation of αPDL1-mRAP22 provided the basis to further evaluate its therapeutic potential. The anti-glioma efficacy of αPDL1-mRAP22 was assessed in orthotopic GL261-luc glioma-bearing C57BL/6 mice (Fig. 6A). The median survival time of PBS, αPDL1 and αPDL1-mRAP22 group was 27.5 d, 30 d and 37.5 d, respectively (Fig. 6B). Compared to PBS group, αPDL1 group exhibited a 4-d extension in terms of the median survival time, whereas αPDL1-mRAP22 group demonstrated a 10-d extension, representing a 4-fold greater prolongation over αPDL1 group. Meanwhile, mouse body weight changes were similar in all groups during treatment, indicating acceptable safety of αPDL1-mRAP22 treatment (Fig. 6C). In addition, in vivo bioluminescence signal in the glioma region of mice from αPDL1-mRAP22 group decreased significantly after treatment (Fig. 6D and 6E). Semi-quantification of Ki67 immunohistochemical staining and TUNEL immunofluorescence staining of brain sections from mice showed that αPDL1-mRAP22 group had the lowest Ki67-positive cell percentage while the highest TUNEL-positive cell percentage, indicating weaker proliferation whereas stronger apoptosis by αPDL1-mRAP22 treatment (Fig. 6F–6I). The above results indicated that αPDL1-mRAP22 could cross the BBB and BBTB, penetrate into the glioma microenvironment, block PD1/PDL1 pathway, and improve anti-tumor efficacy with preliminarily verified safety. In vivo anti-glioma efficacy of αPDL1-mRAP22 in orthotopic GL261-luc glioma-bearing C57BL/6 mice. (A) Treatment procedure including tumor inoculation, therapeutics administration, bioluminescence imaging, and survival analysis; (B) Kaplan-Meier survival curves of mice in different groups (n = 10). Data were analyzed by Log-rank test, ***P < 0.001; (C) Relative body weight of mice in different groups (n = 10); (D) In vivo bioluminescence imaging and (E) semi-quantitative analysis of radiant efficiency of the glioma site in mice of different groups (n = 4); (F) Ki67 immunohistochemical staining and (G) semi-quantitative analysis of brain sections from mice in different groups on Day 12 after treatment (n = 3). Scale bar: 50 µm; (H) TUNEL immunofluorescence staining and (I) semi-quantitative analysis of brain sections from mice in different groups on Day 12 after treatment (n = 3). Blue: DAPI. Green: TUNEL. Scale bar: 20 µm. Data were presented as mean ± SD, ***P < 0.001. In vivo immune responses of αPDL1-mRAP22 Based on the results that αPDL1-mRAP22 showed potent in vivo anti-tumor efficacy, whether αPDL1-mRAP22 could trigger CNS and peripheral immune responses were further investigated. Flow cytometry analysis of immune cells located in the glioma site showed that the percentages of CD3+ T cells and CD3+CD8+ T cells among CD45+ cells of αPDL1-mRAP22 group were 70.02% ± 4.96% and 37.46% ± 4.66%, with a certain degree of increase compared to those of αPDL1 group (57.66% ± 7.26% and 30.78% ± 4.14%, respectively) and significant increase compared to those of PBS group (43.76% ± 6.96% and 18.04% ± 4.98%, respectively) (Fig. 7A–7C). Previous studies usually assessed cytotoxicity of CD8+ T cells against tumor cells through measuring secreted pro-inflammatory cytokines including IFN-γ and TNF-α [[48], [49], [50]]. αPDL1-mRAP22 treatment markedly increased IFN-γ+ and TNF-α+CD3+CD8+ T cell proportions to 57.20% and 70.57%, surpassing both αPDL1 and PBS groups (Fig. 7D and 7E). Similarly, CD3+CD4+ T cells among CD45+ cells was significantly elevated in the αPDL1-mRAP22 group (34.18% ± 4.10%) relative to αPDL1 (23.08% ± 2.08%) and PBS (22.50% ± 6.35%) groups (Fig. 7H). αPDL1-mRAP22 also significantly reduced of the percentage of Foxp3+CD25+ cells among CD3+CD4+ T cells (Tregs playing an important role in immunosuppressive regulation) to 31.57% ± 7.25%, compared with αPDL1 (44.27% ± 4.71%) and PBS (51.60% ± 1.40%) (Fig. 7J). Additionally, Ki67 is an important intranuclear marker characterizing cell proliferation. By staining CD3+CD8+T cells and CD3+CD4+ T cells with Ki67 intranuclear antibody, we found the increase in the proportion of Ki67+ T cells after αPDL1-mRAP22 treatment, implying that it triggered more potent and persistent T cell proliferative activity in the glioma site (Fig. 7F and 7I). Finally, the increase in PD1 expression on the surface of CD3+CD8+ T cells in αPDL1-mRAP22 group was detected, which revealed that αPDL1-mRAP22 did effectively block PD1/PDL1 pathway and might have induced somewhat compensatory upregulation of PD1 (Fig. 7G). The above results provided strong evidence that αPDL1-mRAP22 effectively activated immune responses in the brain and improved immunosuppressive microenvironment of glioma after crossing the BBB and BBTB in vivo. In vivo immune responses of αPDL1-mRAP22 in orthotopic GL261-luc glioma-bearing C57BL/6 mice. (A) Gating strategy and (B) percentage of CD3+ T cells of CD45+ cells in glioma tissues from mice in different groups; Percentage of (C) CD3+CD8+ T cells of CD45+ cells, (D) IFN-γ+ cells of CD3+CD8+ T cells, (E) TNF-α+ cells of CD3+CD8+ T cells, (F) Ki67+ cells of CD3+CD8+ T cells, (G) PD1+ cells of CD3+CD8+ T cells, (H) CD3+CD4+ T cells of CD45+ cells, (I) Ki67+ cells of CD3+CD4+ T cells, and (J) Foxp3+CD25+ cells of CD3+CD4+ T cells in glioma tissues from mice in different groups. Data were presented as mean ± SD (n = 3–5), *P < 0.05, **P < 0.01, ***P < 0.001. Besides the fact that immune cells in the glioma site directly reflected the status of immune microenvironment of glioma, tumor-specific antigens (TSAs) produced by tumor cells in the brain could also be presented by drainage through meningeal lymphatic vessels (MLVs) to the cervical lymph nodes adjacent to the CNS, thus potentially activating the immune cells in the lymph nodes [51]. Therefore, the immune cells in the cervical lymph nodes were also analyzed. It was worth noticing that the proportion of CD3+ T cells and CD3+CD8+ T cells among CD45+ cells in αPDL1-mRAP22 group increased to 91.48% ± 5.18% and 27.40% ± 3.70% (Fig. S9A and S9B). Meanwhile, the proportion of CD44+CD62L− cells among CD3+CD8+ T cells, which represent effector T cells, was also significantly higher in αPDL1-mRAP22 group than that in αPDL1 and PBS group (Fig. S9E). In terms of CD3+CD4+ T cells among CD45+ cells, a statistically significant difference was also observed in αPDL1-mRAP22 group (57.30% ± 6.35%) compared to αPDL1 group (46.44% ± 6.23%) (Fig. S9F). In addition, Ki67 staining of CD3+CD8+ T cells and CD3+CD4+ T cells implied a more proliferative status of immune cells in cervical lymph nodes of αPDL1-mRAP22 group (Fig. S9C and S9G). Elevated PD1 expression was also found in αPDL1-mRAP22 group (Fig. S9D). To evaluate whether the observed activation within glioma tissues and draining lymph nodes could trigger sustained systemic immune effects, the distribution of functional T cell subsets was further characterized in the spleen, which allowed for a comprehensive evaluation of the overall immunological landscape. Increases of CD3+ T cells and CD3+CD8+ T cells among CD45+ cells, and CD44+CD62L− T cells among CD3+CD8+ T cells in the mouse spleens of αPDL1-mRAP22 group were observed, by reaching 47.34% ± 9.33%, 18.44% ± 3.25%, and 17.35% ± 2.36%, respectively (Fig. S10A, S10B and S10E). CD3+CD4+ T cells among CD45+ cells in the mouse spleens increased significantly to 28.48% ± 6.08% in αPDL1-mRAP22 group, compared with the control groups (Fig. S10F). Ki67 staining also demonstrated elevation in the proliferative status of both CD3+CD8+ T cells and CD3+CD4+ T cells in αPDL1-mRAP22 group (Fig. S10C and S10G). PD1 upregulation in αPDL1-mRAP22 group was also detected, similar to that in glioma tissue and cervical lymph nodes (Fig. S10D). Thus, αPDL1-mRAP22 was proved to improve systemic anti-tumor immunity to some extent, further strengthening therapeutic effect. Gating strategies of the above assays by flow cytometry were described in supporting information (Fig. S11-S13). To sum up, αPDL1-mRAP22 could block PD1/PDL1 pathway in the glioma site and activate T cell anti-tumor immune responses by increasing the number of CD4+ T helper cells and CD8+ cytotoxic T cells as well as upregulating the level of IFN-γ and TNF-α after crossing the BBB and BBTB. At the same time, TSAs were also drained to the cervical lymph nodes and spleen through MLVs to further stimulate peripheral immune responses and enhance therapeutic effect. In vivo safety of αPDL1-mRAP22 Common side effects of immunotherapy such as immune checkpoint inhibitors include systemic autoimmune reactions caused by increased levels of pro-inflammatory cytokines in serum [[52], [53], [54]], so it is of great importance to assess safety of αPDL1-mRAP22 in orthotopic GL261-luc glioma-bearing C57BL/6 mice. Quantitative results of ELISA showed that serum concentrations of pro-inflammatory cytokines including IFN-γ, TNF-α, IL-6 and IL-1β were not influenced by αPDL1-mRAP22 when compared with PBS, and were all lower than those in αPDL1 group, which demonstrated that αPDL1-mRAP22 could better reach the glioma site than unmodified αPDL1 and thus reduce the probability of immune-related adverse reactions in the periphery (Fig. S14A-S14D). Furthermore, since PD1/PDL1 inhibitors were also reported to induce pathological lesions in normal tissues such as liver and kidney, which may be caused by the aforementioned pro-inflammatory cytokines [53,55], we detected liver function parameters including alanine aminotransferase (ALT) and aspartate aminotransferase (AST) as well as kidney function parameters including blood urea nitrogen (BUN) and creatinine (CREA) in mouse serum. Despite some quantitative differences, the average levels of those parameters in αPDL1-mRAP22 group all lied within normal ranges, implying that the constructed antibody fusion protein had little impact on peripheral normal tissues (Fig. S14E–S14H). In addition to liver and kidney functions, no obvious pathological lesions were observed in H&E staining of major organ sections from mice treated with αPDL1-mRAP22 (Fig. S14I). Moreover, since some clinical studies pointed out that some high-grade adverse reactions such as lymphopenia and neutropenia sometimes occurred along with immune checkpoint blockade therapy [56,57], we conducted complete blood count test after administration. Heat map statistics showed that the levels of various blood cells in αPDL1 and αPDL1-mRAP22 group varied within normal ranges, verifying ideal in vivo safety of αPDL1-mRAP22 (Fig. S14J). Synergistic in vivo anti-glioma efficacy of αPDL1-mRAP22 combined with TMZ Based on the promising therapeutic efficacy of αPDL1-mRAP22 monotherapy, we further explored whether it could synergize with the clinical standard-of-care chemotherapeutic TMZ, so as to achieve a multi-targeted effect that could overcome the inherent heterogeneity of glioma and realize long-term remission. The anti-glioma efficacy of αPDL1-mRAP22 combined with TMZ was assessed in orthotopic GL261-luc glioma-bearing C57BL/6 mice (Fig. 8A). The median survival time was 37 d for αPDL1-mRAP22 group, 42.5 d for TMZ group, and 58 d for TMZ + αPDL1-mRAP22 combination group, respectively (Fig. 8B). Compared to PBS group, the number of days extended in median survival time of TMZ and αPDL1-mRAP22 combination therapy was 1.51-, 1.91- and 2.83-fold that of TMZ + αPDL1, TMZ and αPDL1-mRAP22, respectively. A q value of 1.56 (Jin’s formula) further demonstrated a strong synergistic effect (q > 1.15), suggesting the superiority of this combined regimen [42]. Although mice in all groups experienced varying degrees of body weight loss during treatment, body weight of αPDL1-mRAP22, TMZ and TMZ + αPDL1-mRAP22 groups recovered after treatment ceased, indicating good tolerance to the dosage (Fig. 8C). Synergistic effects brought about by combination with TMZ were further clarified by semi-quantitative results showing the lowest in vivo bioluminescence signal (Fig. 8D and 8E) and Ki67-positive signal representing glioma cell proliferation (Fig. 8F and 8G) as well as the highest TUNEL-positive signal representing glioma cell apoptosis (Fig. 8H and 8I) in the glioma region of TMZ + αPDL1-mRAP22 group. Synergistic in vivo anti-glioma efficacy of αPDL1-mRAP22 combined with TMZ in orthotopic GL261-luc glioma-bearing C57BL/6 mice. (A) Treatment procedure including tumor inoculation, therapeutic administration, bioluminescence imaging, and survival analysis; (B) Kaplan-Meier survival curve of mice (n = 10). Data were analyzed by Log-rank test, *P < 0.05, **P < 0.01, ***P < 0.001; (C) Relative body weight of mice (n = 10); (D) In vivo bioluminescence imaging and (E) semi-quantitative analysis of radiant efficiency of the glioma site in mice of different groups (n = 3); (F) Ki67 immunohistochemical staining and (G) semi-quantitative analysis of brain sections from mice in different groups on Day 12 after treatment. Scale bar: 100 µm (n = 3); (H) TUNEL immunofluorescence staining and (I) semi-quantitative analysis of brain sections from mice in different groups on Day 12 after treatment. Blue: DAPI. Green: TUNEL. Scale bar: 50 µm (n = 3). Data were presented as mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001, ns, no significance. The remarkable survival benefits and tumor regression observed from the combination therapy prompted us to explore the immunological mechanisms of this synergistic effect. It was reported in previous literature that beyond its classic role in DNA damage, TMZ has also been recognized for its capacity to ignite diverse regulated ICD modalities, including immunogenic apoptosis, necroptosis, ferroptosis, and pyroptosis [[58], [59], [60], [61], [62]]. This multifaceted cell death profile, termed mixed types of cell death (MTCD), is instrumental in liberating key damage-associated molecular patterns (DAMPs). During ICD process, DAMPs are released to stimulate DCs, and TSAs are then presented to lymphocytes, triggering enhanced immunotherapeutic effects [63]. Specifically, CRT and HMGB1 are classic markers of ICD, with the former translocating to tumor cell surface under ER stress and the latter being released from tumor cell nucleus to tumor microenvironment [64]. CRT and HMGB1 immunofluorescence staining of brain sections from mice of different groups in our study showed enhanced CRT translocation to tumor cell surface while lessened colocalization of HMGB1 with tumor cell nucleus (Fig. S15). These molecular danger signals are essential for the recruitment and activation of tumor-infiltrating lymphocytes in the CNS. Our results showed that by leveraging TMZ to initiate ICD, the subsequent immune checkpoint blockade provided by αPDL1-mRAP22 could more effectively induce a sustained anti-tumor immune response. This dual-action approach might provide a rational basis for utilizing frontline chemotherapeutics to convert the "cold'' glioma microenvironment into an immunologically ``hot'' state. Finally, no significant pathological lesions were observed in H&E staining of major organ sections from mice of all groups, demonstrating relatively good in vivo safety of αPDL1-mRAP22 combined with TMZ (Fig. S16). Conclusion In summary, we constructed αPDL1-mRAP22 as an antibody fusion protein for glioma-targeted therapy. With fusion of the PDL1 antibody and the BBB/BBTB-crossing peptide RAP22 via an MMP2-sensitive linker, αPDL1-mRAP22 was capable of efficiently crossing the BBB/BBTB, responsively releasing free αPDL1 upon MMP2 cleavage in the glioma microenvironment, and potently triggering anti-tumor immune responses of tumor-infiltrating T cells by blocking PD1/PDL1 pathway, which further achieved synergistic effect by combination therapy with TMZ. Our research provides a novel immunotherapeutic strategy for glioma treatment, which might be promising in future clinical translation. CRediT authorship contribution statement Yanning Bao: Writing – original draft, Visualization, Methodology, Investigation, Data curation. Shengmin Yang: Investigation. Yuan Ding: Investigation. Jianfen Zhou: Investigation. Xudong Zheng: Investigation. Ruohan Chen: Investigation. Zhixuan Jiang: Investigation. Nana Meng: Investigation. Dongsheng Pan: Investigation. Jiasheng Lu: Investigation. Jiacheng Guan: Investigation. Jing Ye: Investigation. Hongjia Zhu: Investigation. Nan Xu: Investigation. Cao Xie: Investigation. Lei Yu: Resources, Methodology. Linwei Lu: Writing – review & editing, Methodology. Weiyue Lu: Writing – review & editing, Funding acquisition, Conceptualization. Conflicts of interest The authors declare that there is no conflicts of interest. Acknowledgements This work was supported by the National Natural Science Foundation of China [grant number 81690263]. Footnotes Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.ajps.2026.101175. The figures and tables with “S” before the serial number are included in the Supplementary material. Contributor Information Lei Yu, Email: lei.yu@unicar-therapy.com. Linwei Lu, Email: lw_lu@fudan.edu.cn. Weiyue Lu, Email: wylu@shmu.edu.cn. Supplementary materials References Associated Data Supplementary Materials