Peptide-drug conjugates repolarize glioblastoma-associated macrophages to resensitize chemo-immunotherapy of glioblastoma

Abstract

The prevalent tumor-supporting glioblastoma-associated macrophages (GAMs) promote glioblastoma multiforme (GBM) progression and resistance to multiple therapies. Repolarizing GAMs from tumor-supporting to tumor-inhibiting phenotype may troubleshoot. However, sufficient accumulation of drugs at the GBM site is restricted by blood-brain barrier (BBB). Herein, we designed peptide-drug conjugates (PDCs) by conjugating camptothecin or resiquimod to a tandem peptide composed of matrix metalloproteinase 2–responsive peptide and angiopep-2 via disulfonyl-ethyl carbonate/carbamate (^MAPDCs). The mixed self-assembly ^MAPDCs could recognize low-density lipoprotein receptor–related protein 1 (LRP1) to facilitate BBB transport. Once reaching the GBM site, the responsive peptide would be cleaved to shed the angiopep-2, blocking abluminal LRP1-mediated brain-to-blood efflux and enhancing drug retention. Sequentially, drugs are released under the high level of intracellular glutathione. In vivo studies demonstrated that ^MAPDCs repolarized GAMs, boosted immune response, and resensitized chemotherapeutic toxicity, offering a much-improved anti-GBM effect. The effectiveness of ^MAPDCs validates GAMs as therapeutic target and PDCs as versatile brain delivery system with high design flexibility.

A peptide-drug conjugates with enhanced camptothecin and resiquimod retention effectively combats glioblastoma.

INTRODUCTION

Glioblastoma multiforme (GBM), a highly aggressive and lethal intracranial malignancy in adult, remains invariably incurable. Despite many clinical trials, the standard of care therapy for over 20 years has been maximum resection combined with radiotherapy and adjuvant chemotherapy. The median lifespan from the first diagnosis to death is about 15 months, and the 5-year survival is less than 10% (1–3). The unsatisfied treatment outcome has stimulated ongoing efforts to reveal notable insights into GBM and furthermore facilitate the development of therapeutic regimens that can combine different therapeutic targets.

Accumulating studies indicate that an important feature of GBM cells is that they closely affect and manipulate surrounding cells, inducing them to support tumor progression and drug resistance (4, 5). For example, GBM cells can recruit peripheral circulating macrophages or monocytes and brain-resident microglia to infiltrate the GBM microenvironment (6). These macrophages and microglia collectively compose glioblastoma-associated macrophages (GAMs), which are the most prevalent immune cells and make up 30 to 50% percentage of stromal cells within GBM mass (7). In particular, a predominant proportion of GAMs represents the anti-inflammatory M2 phenotype (M2-GAMs) (8). It contributes to immunosuppressive tumor microenvironment (TME) by compromising antigen presentation, secreting anti-inflammatory cytokines, directly inhibiting effector T cells, as well as promoting regulatory T (T_reg) cells, myeloid-derived suppressive cells differentiation, etc. (8–10). M2-GAMs also secret vascular endothelial growth factor and express matrix metalloprotease-2 (MMP2) to support the angiogenesis and degrade the extracellular matrix, respectively, which helps GBM growth and invasion (6, 11). In addition to their immunosuppressive and tumor-supporting function, increasing evidence reveals that M2-GAMs actively influence glioblastoma stem cells, which are highly resistant to chemotherapy, suggesting that M2-GAMs play a critical role in GBM resistance to chemotherapy (12–14). These facts establish that M2-GAMs have emerged as a potent therapeutic target to be used in combination with chemotherapy. Fortunately, GAMs are characterized by high phenotypic plasticity, allowing the flexible conversion between proinflammatory M1 phenotype and M2 phenotype (15, 16). Successfully repolarizing M2-GAMs to proinflammatory M1-phenotypic GAMs (M1-GAMs) not only damages the microenvironment favorable to tumor growth and invasion, unlocks GAMs’ innate antitumor power, and ameliorates immunosuppressive TME but also resensitizes chemotherapy through reducing GBM resistance to chemotherapy while simultaneously strengthening chemotherapy-triggered immunogenic cell death (ICD) effect, which is inhibited in the immunosuppressive TME (17, 18).

Despite the combination strategy being promising, their therapeutic efficiency is greatly restricted by the poor delivery efficiency to the GBM site due to the existence of a series of delivery barriers, of which blood-brain barrier (BBB), a specialized neurovascular unit evolved to maintain brain hemostasis, represents the most formidable delivery barrier (19). Because of the ineffective transport across BBB, a plethora of therapeutic drugs that treat brain diseases fail to receive approval for clinical application, even promising treatment outcomes are demonstrated in preclinical studies (20–22). Over the past decades, brain-targeting drug delivery systems leveraging the endogenous receptor–mediated transporting, including transferrin receptor, insulin receptor, and low-density lipoprotein receptor–related protein 1 (LRP1), have been largely explored to assist BBB transport (23). However, many receptors are expressed on both luminal and abluminal site of brain endothelial cells (BECs) of BBB endothelium (24). For example, LRP1 expressed on luminal (blood) site mediates the blood-to-brain influx, whereas LRP1 expressed on abluminal (brain) site simultaneously mediates brain-to-blood efflux, presenting a bidirectional transport (24, 25). Similarly, the conventional LRP1-targeted delivery system may also subject to this bidirectional transport, leading to brain-to-blood clearance and reduced drug retention in brain parenchyma. Therefore, in the design of brain-targeted delivery systems using LRP1, blocking the abluminal LRP1-mediated brain-to-blood efflux to enhance drug retention in brain is of great importance. Another important issue is that nanocarrier-based delivery systems generally require additional materials or ingredients to assemble nanocarriers, which may raise the issues including complicated manufacturing processes, material-induced toxicity or immunogenicity, and potential drug leakage (26, 27). Recently, peptide-drug conjugates (PDCs), composed of peptide, drug, and linker, that can self-assemble into supramolecular nanostructures have emerged as the next generation of drug delivery system after antibody-drug conjugates (28). PDCs exhibit the superiority in the biocompatibility, targeting specificity, cellular internalization, especially for on-demand functionalities due to the utilization of stimulus-responsive linker, such as site-specific drug release, alternation of physiological properties and surface modification (28, 29). Inspired by this, we concept that PDCs may hold great promise to precisely deliver drug to the GBM site while circumventing LRP1-mediated brain-to-blood efflux by using TME responsive linker and brain-targeting peptides.

Herein, we proposed a dual-responsive coupling brain-targeting PDCs by conjugating camptothecin (CPT) or resiquimod (R848) to a tandem peptide composing of MMP-2–responsive peptide (GPLGLAG, M pep) and angiopep-2 peptide (Ang2) via disulfonyl-ethyl carbonate/carbamate (pre–^MAPDCs-C and pre–^MAPDCs-R). M pep and disulfonyl-ethyl carbonate/carbamate were MMP-2 and glutathione (GSH)–responsive linker, respectively, and Ang2 can bind to LRP1. We further mixed two pre-^MAPDCs to form supramolecular nanostructure (^MAPDCs). After systemic administration, ^MAPDCs could specifically recognize LRP1 expressed on luminal site of BBB to trigger blood-to-brain transport. Once crossing the BBB into extracellular GBM region, which was enriched with MMP-2, M pep of ^MAPDCs would be cleaved by MMP-2, leading to the shedding of outer Ang2 from ^MAPDCs. Therefore, the remaining PDCs without Ang2 modification was unable to bind LRP1 on the abluminal site of BBB to undergo brain-to-blood efflux, followed by being internalized by either GBM cells or M2-GAMs via a receptor-independent endocytosis predominantly, finally resulting in the enhanced drug retention with high selectivity (30). Then, drugs would undergo a GSH-triggering selective release from PDCs upon internalizing into GBM cells or M2-GAMs, where the intracellular GSH has a much higher concentration than extracellular GSH. CPT could exert chemotherapeutic killing on GBM cells and initiate anti-GBM immunity via ICD effect, while R848 could repolarize M2-GAMs to M1-GAMs to strengthen both innate and adaptive immune response and reduce chemotherapy resistance. ^MAPDCs also had chance to cross BBB endothelial cells around normal brain parenchyma where the MMP2 concentration was relatively low. However, ^MAPDCs at normal brain parenchyma would remain intact and underwent abluminal LRP1-mediated brain-to-blood efflux or targeted at the GBM site, leading to lower retention at normal brain region and less effect to normal brain cells. Both in vitro and in vivo studies confirmed the superiority of ^MAPDCs in enhancing drug retention at the GBM site, and, consequently, successful repolarization of M2-GAMs to M1-GAMs and chemotherapy-induced killing on GBM cells were observed. The effectiveness of ^MAPDCs provides an idea for developing therapies targeting GAMs or other stromal cells to combine with standard therapy. Meanwhile, the versatile PDCs may be a promising brain-delivery platform with higher design flexibility and translational potential (Fig. 1 and fig. S1).

Fig. 1. Design and preparation of ^MAPDCs and its action patterns for GBM.

(A) The mechanisms underlying the recovery of two pre-^MAPDCs activities in response to MMP-2 and GSH. (B) The preparation of ^MAPDCs that were able to cross BBB with the help of luminal LRP1 and to shed outside LRP1 ligand (Ang2) via cleavage of MMP-sensitive peptide, thus blocking abluminal LRP1–mediated efflux and increasing the drug concentration in brain. (C) Under high level of GSH, ^MAPDCs released drugs in their prototypes to affect tumor cells and M2-GAMs, leading to improved chemotherapeutic response and anti-GBM immune response. DAMP, damage-associated molecular patterns; iDC, immature dendritic cell; mDC, mature dendritic cell; ECM, extracellular matrix. The illustrations of cells were generated using Servier Medical Art, provided by Servier, licensed under CC BY 4.0.

RESULTS

Experimental design validation

We firstly determined MMP-2 and GSH content in normal brain and the GBM site after 2 weeks of intracranial implantation with GL261 cells, a murine GBM cell line. In contrast to normal brain, MMP-2 and GSH content in the GBM site were significantly increased by 4.4 and 26.1 times, respectively (Fig. 2A). The results indicated that MMP-2 and GSH were closely relevant with GBM progression, and their much-higher expression qualified them as effective and specialized stimuli. We also determined the GSH level inside two main GAMs populations, namely circulating bone marrow–derived macrophages (BMDMs) and microglia. The intracellular GSH level was 1.05 and 1.26 μmol/million cells for BMDMs and microglia, respectively, which was completely sufficient to cleave disulfide bond and release drug (fig. S2). In addition, we verified the affinity of Ang2 to recombinant mouse LRP1 using surface plasmon resonance (SPR)–based molecular interaction system. The result exhibited a distinct association of Ang2 to LRP1, whereas negligible association was detected in poly(ethylene glycol) (PEG), indicating the specific and strong affinity between Ang2 and LRP1 (Fig. 2B). Together, these results served as a proof of concept that ^MAPDCs might have targeting specificity toward LRP1, stimulus-responsive on-demand delivery, and drug release capabilities.

Fig. 2. Experimental design validation and characterization of ^MAPDCs.

(A) Expression of MMP-2 and GSH at the GBM site and normal brain zone from GL261-bearing mice (n = 5). (B) SPR analysis of Ang2 (n = 3) and PEG (n = 1) binding to LRP1. The shaded area indicated the error bar. (C) The CD spectrum of M pep, Ang2, and M-A tandem peptide (n = 3). The shaded area indicated the error bar. (D) MALDI-TOF MS analysis of pre-^MAPDCs-C and pre-^MAPDCs-R. (E) Critical micelle concentration measurement of ^MAPDCs using a Nile red method and the corresponding fitting curve (n = 3). (F) Hydrodynamic sizes of ^MAPDCs at different time points determined by DLS. The inset was TEM image of ^MAPDCs. Scale bar, 200 nm. (G) SPR analysis of ^MAPDCs and ^NAPDCs binding to LRP1 (n = 3). (H and I) HPLC analysis of ^MAPDCs treated with MMP-2 or GSH. Data in (H) and (I) were related to CPT and R848, respectively. (J) Hydrodynamic sizes of ^MAPDCs after incubation with MMP-2 or GSH. The insets were TEM images of ^MAPDCs after incubation with MMP-2 or GSH. Scale bar, 200 nm. (K and L) The release profile of CPT (K) and R848 (L) from ^MAPDCs in the presence or absence of GSH (n = 3). Data in (A) to (C), (E), (G), and (K) to (L) were given as means ± SD. Data in (A) were analyzed by two-sided, independent-samples t test.

Fabrication and characterization of ^MAPDCs

In the design of ^MAPDCs, M pep was directly linked to Ang2 via amido bond to form a tandem peptide (M-A). M-A demonstrated the similar circular dichroism (CD) spectrum as the individual M pep and Ang2, providing evidence that the secondary structures of two peptides were correctly maintained even after covalent conjugation (Fig. 2C). M-A was furthermore conjugated onto CPT- or R848–disulfonyl-ethyl carbonate/carbamate (figs. S3 to S7) through sulfhydryl-disulfhydryl exchange reaction to synthesize pre–^MAPDCs-C and pre–^MAPDCs-R, confirmed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis (Fig. 2D) (31, 32). The ^MAPDCs was prepared by co-nanoprecipitation of pre–^MAPDCs-C and pre–^MAPDCs-R with a critical micelle concentration (CMC) as low as 1.56 μM, suggesting that pre–^MAPDCs-C and pre–^MAPDCs-R could easily self-assemble into ^MAPDCs with high thermodynamic stability even after injection into blood circulation (Fig. 2E). Other precursor PDCs were also synthesized to prepare control PDCs with similar procedures, including ^NPPDCs (without M pep and Ang2 conjugation), ^MPPDCs (with M pep but without Ang2 conjugation), and ^NAPDCs (without M pep but with Ang2 conjugation). Dynamic light scattering (DLS) analysis showed that the hydrodynamic diameter of ^MAPDCs was approximately 95 nm, with a polydispersity index of 0.145 (Fig. 2F). Transmission electron microscopy (TEM) image confirmed the size of ^MAPDCs with a uniform dispersion and spherical morphology (Fig. 2F). Meanwhile, no obvious size change was observed within 7 days when ^MAPDCs was stored at room temperature, indicating the high stability of ^MAPDCs. To determine whether the self-assembled ^MAPDCs was in an expected spatial orientation, we next examined the affinity of ^MAPDCs to recombinant LRP1. SPR analysis showed that ^MAPDCs had a strong affinity to LRP1, with a dissociation constant (K_d) as low as 0.645 mM, while ^MPPDCs had a negligible affinity to LRP1 even at a high concentration, with a K_d of 2340 mM (Fig. 2G). Similar results were also observed that ^NAPDCs had a much higher affinity than ^NPPDCs (fig. S8). These results directly verified that ^MAPDCs was assembled in a correct spatial orientation that hydrophobic drugs were located inside while hydrophilic Ang2 was localized outside, which can specifically interact with LRP1. We next determined the responsiveness of ^MAPDCs to MMP-2 and GSH by assessing the cleavability of disulfide linker and M pep using high-performance liquid chromatography (HPLC). After incubation with MMP-2 for 4 hours, a new peak denoting the PDCs without Ang2 (LAG-Ang2) was found in the chromatography of ^MAPDCs (Fig. 2, H and I). For the case of incubation with GSH, the peak denoting CPT or R848 was found while the peak denoting ^MAPDCs completely disappeared. TEM images and DLS analysis further exhibited smaller ^MAPDCs particles after incubation with MMP-2, which was mainly owing to the whole structure being maintained by hydrophobic CPT/R848 and the rest hydrophilic peptide fragment although part of hydrophilic peptide fragment was cleaved by MMP-2 (Fig. 2J and fig. S9). However, incubation with GSH led to larger irregular nanotube instead of spherical particles, with an increased size distribution, likely due to the separation of hydrophilic and hydrophobic part (Fig. 2J). These results collectively confirmed the responsiveness of ^MAPDCs to both MMP-2 and GSH. Last, the release profiles of drugs from ^MAPDCs with or without GSH were determined. The cumulative release ratio of either CPT or R848 from ^MAPDCs without GSH treatment was less than 10% even after 48 hours of incubation (Fig. 2, K and L). By comparison, the release of either CPT or R848 from ^MAPDCs with GSH treatment was much quicker, with a cumulative release ratio reaching approximately 70 and 60% only after 2 hours of incubation. These results strongly confirmed the GSH-responsive drug release behavior of ^MAPDCs with high effectiveness and specificity (fig. S10). This was beneficial for realizing on-demand drug release at the GBM site where it is enriched with GSH.

Cellular uptake, endosomal escape, and cytotoxicity

Next, we evaluated the internalization efficiency of ^MAPDCs by bEnd.3 cell, a murine BEC line, and GL261 cell, a murine GBM cell line. Flow cytometry analysis showed a much higher cellular uptake of ^MAPDCs by either bEnd.3 or GL261 cells compared with ^MPPDCs and ^NPPDCs, while similar to ^NAPDCs (Fig. 3, A to D). These results indicated that Ang2 modification could promote the cellular uptake of PDCs by either bEnd.3 or GL261 cells via binding the LRP1 expressed on both of them. To dig the underlying mechanisms involved in cellular uptake, we pretreated bEnd.3 cells with different inhibitors and then incubated them with ^MAPDCs. Pretreatment with chlorpromazine (CPZ, an inhibitor of clathrin), amiloride (AMI, an inhibitor of macropinocytosis), and genistein (Geni, an inhibitor of caveolae) reduced the cellular uptake of ^MAPDCs to 47, 61, and 61%, respectively. (Fig. 3, E and F). These results implied that multiple mechanisms were involved in the internalization of ^MAPDCs, including clathrin-mediated endocytosis, macropinocytosis, and caveolae-mediated endocytosis. Moreover, pretreatment with free Ang2 at a low concentration scarcely affected cellular uptake while at a high concentration of 100 μg/ml significantly reduced the cellular uptake to nearly 32%, suggesting that the receptor-mediated endocytosis was a predominant mechanism. However, it was well known that the formed cargo-loaded endocytic vesicles during receptor-mediated endocytosis may undergo three main pathways: (i) trafficking to different plasma membrane, namely transcytosis; (ii) recycling back to initial plasma membrane; and (iii) trafficking to lysosomes for fusion and degradation. Whether these endocytic vesicles sorted into lysosomes could escape from the degradative compartment was also crucial for crossing BBB (23, 33, 34). To address this question, we performed the fluorescence distribution experiments of ^MAPDCs within bEnd.3 cells using confocal laser scanning microscope (CLSM). The images revealed that ^MAPDCs colocalized with lysosomes at 4 hours after incubation, with a colocalization coefficient of 0.62 (Fig. 3, G and H). By contrast, the fluorescence colocalization was apparently weakened at 12 hours, accompanied by a reduced colocalization coefficient of 0.43, underscoring the excellent lysosomal escape capability of ^MAPDCs. In addition, CLSM images also showed a higher internalization of ^MAPDCs and ^NAPDCs by bEnd.3 cells compared with PEGylated PDCs, which was in line with flow cytometry analysis (Fig. 3I and fig. S11). To prove that ^MAPDCs remained intact after lysosomal escape, we established a transwell model where bEnd.3 cells were cultured in the upper chamber and introduce ^MAPDCs in the upper chamber. After 12 hours of incubation, we collected the medium in the bottom chamber, which contained ^MAPDCs. TEM images of collected ^MAPDCs demonstrated the spherical morphology with similar size to untreated ^MAPDCs (fig. S12). Meanwhile, HPLC analysis also showed similar retention time of collected ^MAPDCs to untreated ^MAPDCs (Fig. 3J). These results collectively suggested that even after lysosomal escape, ^MAPDCs remained intact. The efficient internalization of ^MAPDCs encouraged us to determine its cytotoxicity to GL261 cells. After 48 hours of incubation, the cell proliferation of GL261 cells treated with ^MAPDCs was notably inhibited in a concentration-dependent manner, with a half-maximal inhibitory concentration (IC₅₀) of 351.4 nM (Fig. 3K). A much higher IC₅₀ of 736.5 nM, almost two times higher than that of ^MAPDCs, was obtained after treatment with free CPT, indicating that ^MAPDCs were able to completely release two CPT molecules to exert cytotoxic effect (35). Together, these results indicated that ^MAPDCs could be efficiently internalized by both BECs and GBM cells, without compromising the cytotoxicity of CPT toward GBM cells.

Fig. 3. ^MAPDCs efficiently internalized into bEnd.3 and GL261 cells and exerted cytotoxicity to GL261 cells.

(A and B) Representative flow cytometry histogram of cellular uptake of ^MAPDCs by bEnd.3 cells (A) and GL261 cells (B) at 1, 4, and 12 hours. (C and D) Mean fluorescence intensity (MFI) quantification of (A) and (B) (n = 3). (E) Representative flow cytometry histogram of cellular uptake of ^MAPDCs by bEnd.3 cells when inhibitors were introduced or cells were incubated at 4°C. (F) The relative uptake noted by normalized MFI in (E) (n = 3). (G) Representative CLSM images of bEnd.3 cells after incubation with ^MAPDCs for 4 and 12 hours. The cell nucleus and lysosome were stained with DRAQ5 (blue) and lysotracker red (red), respectively. Scale bar, 20 μm. The right scatter plots were obtained by plotting intensity of red channel against that of green channel. (H) Pearson’s correlation coefficient between red channel intensity and green channel intensity in (G). (I) MFI quantification of green channel per cell in (G). Data were combined from three independent experiments. (J) HPLC analysis of the untreated ^MAPDCs (before) and the collected ^MAPDCs (after) in the bottom chamber of the transwell model. (K) Cytotoxicity of free CPT and PDCs toward GL261 cells (n = 3). Data in (C) and (D), (F), (I), and (K) were given as means ± SD. Data in (F) were analyzed by one-way analysis of variance (ANOVA) followed by Games-Hoswell test. Data in (H) were analyzed by two-sided, independent-samples t test.

Assessment of transport using in vitro BBB model

The favorable effect at monolayer cell level prompted us to explore in vitro BBB transport efficiency of ^MAPDCs. To address this, we constructed a transwell model (i.e., a hanging cell culture system) to mimic in vitro BBB structure and functionalities. In this system, bEnd.3 cells were cultured in the upper chamber, and GL261 cells or BMDMs/microglia were cultured in the bottom chamber (Fig. 4A). CLSM images showed that ZO-1, a classical tight junction protein, abundantly expressed on bEnd.3 cells, indicating that the tight junction formed between bEnd.3 cells and in vitro BBB model could somehow mimic the real BBB in vivo (fig. S13). Before determining the transport efficiency, we first measured the cellular uptake of ^MAPDCs and control PDCs by the upper bEnd.3 monolayer cells. In line with the experiment result at monolayer cell level, ^MAPDCs had the highest cellular uptake at 4 hours after incubation and this was maintained for up to 12 hours (Fig. 4B). As for GL261 cells in the bottom chamber, the cellular uptake of ^MAPDCs was also significantly higher than that of control PDCs at either 4 or 12 hours, as supported by CLSM images and semiquantitative analysis (Fig. 4C). These results collectively denoted that ^MAPDCs not only efficiently internalized into bEnd.3 cells but also crossed the bEnd.3 monolayer efficiently and retained at GL261 cells in the bottom chamber. We speculated that this result could be attributed to the canonical LRP1-mediated transport once ligation with Ang2 and MMP-2–triggered Ang2 detachment after reaching the bottom chamber circumvented reversal efflux. To prove our hypothesis, we delicately verified the effect of Ang2 detachment on cellular uptake through adding ^MAPDCs or ^NAPDCs into the bottom chamber (Fig. 4A). The results showed that the uptake of ^NAPDCs by bEnd.3 cells was consistently stronger than that of ^MAPDCs, which further supported our hypothesis that ^NAPDCs could undergo LRP1-mediated retrograde reflux, resulting in higher cellular uptake by bEnd.3 cells than that of ^MAPDCs. As for GL261 cells, it was unexpected that at early time point (0.5 and 2 hours), more ^NAPDCs were internalized by GL261 cells than ^MAPDCs. With increasing time, the uptake of ^MAPDCs by GL261 cells gradually exceeded that of ^NAPDCs. It was ascribed that ^NAPDCs took advantage of the targeting groups to be internalized faster into GL261 cells, leading to higher cellular uptake at early time point. However, the overall amount of ^NAPDCs in the bottom chamber was lower than that of ^MAPDCs because of their simultaneously faster reflux. In this case, the amount of PDCs in the bottom chamber was a dominant factor to cellular uptake, particularly in the later period (Fig. 4D and fig. S14). These results supported our hypothesis that the Ang2 detachment induced by MMP-2 circumvented LRP1-mediated bottom-to-upper transport of ^MAPDCs, thereby increasing drug retention in the bottom chamber. In contrast, ^NAPDCs was unable to shed Ang2 since it did not respond to MMP-2 and therefore subjected to this reverse transport to the upper chamber.

Fig. 4. ^MAPDCs crossed in vitro BBB to induce GL261 cells apoptosis and macrophages repolarization.

(A) Schematic illustration of in vitro BBB model. (B) Representative flow cytometry histogram of cellular uptake by bEnd.3 cells and GL261 cells, and the corresponding MFI quantification (n = 3). (C) Representative CLSM images of GL261 cells and the MFI quantification of green channel. The cell nucleus was stained with DRAQ5 (blue). Scale bar, 25 μm. (D) The cellular uptake by bEnd.3 and GL261 cells when ^MAPDCs or ^NAPDCs was added into the bottom chamber (n = 3). (E) Representative flow cytometry plots and the quantification of apoptotic GL261 cells (n = 3). (F) Representative CLSM images of the bottom GL261 cells with ^NPPDC or ^MAPDC treatment and the corresponding fluorescence intensity quantification. The staining included CRT (green), F-actin (red), and 4′,6-diamidino-2-phenylindole (DAPI, blue). Scale bar, 100 μm. (G) Representative CLSM images of the bottom GL261 cells with ^NPPDC or ^MAPDC treatment. The staining included HMGB1 (green), F-actin (red) and DAPI (blue). Scale bar, 100 μm. (H) ELISA analysis of the extracellular HMGB1 concentration in the bottom supernatant collected from differently treated GL261 cells (n = 3). (I) Representative flow cytometry plots and the analysis of the bottom M1 (CD86⁺CD206⁻) and M2 (CD86⁻CD206⁺) within primary microglia. (J) Representative CLSM images of BMDMs. The staining included CD86 (green), CD206 (red), and DAPI (blue). The fluorescence ratio images were obtained through dividing CD86 intensity by CD206 intensity. Scale bar, 100 μm. (K) The fluorescence ratio quantification of each individual cell in (J). (L) Detection of cytokines secreted by differently treated BMDMs (n = 3). Data were given as means ± SD. Data in (D), (F), (H), (I), and (K) were analyzed by two-sided, independent-samples t test, one-way ANOVA followed by Turkey test and Games-Hoswell test, respectively.

Since ^MAPDCs has the ability to efficiently cross bEnd.3 monolayers and retain in GL261 cells in higher quantity, we determined the ability of ^MAPDCs to induce GL261 cells apoptosis and BMDMs/microglia polarization. Flow cytometry analysis displayed that ^MAPDCs remarkably induced GL261 cells apoptosis, with a 1.4-, 1.6-, 2.4-, and 2.5-fold increase compared with ^NAPDCs, ^MPPDCs, ^NPPDCs, and free drugs, respectively (Fig. 4E). The ICD effect induced by CPT was also visualized through immunofluorescence staining for calreticulin (CRT) and high-mobility group box 1 protein (HMGB1). ^MAPDCs caused the highest expression of CRT with a 5.2-fold increase in fluorescence intensity, as opposed to ^NPPDC treatment (Fig. 4F and fig. S15A). Meanwhile, the nucleus-to-cytoplasm translocation of HMGB1 was the most apparent in ^MAPDCs group (Fig. 4G and fig. S15B). Notably, the supernatant of GL261 cells treated with ^MAPDCs showed the highest HMGB1 concentration, which further demonstrated the nucleus-to-cytoplasm translocation of HMGB1 and its secretion to into extracellular space induced by ^MAPDCs (Fig. 4H). Regarding primary microglia polarization ability, ^MAPDC treatment notably increased the expression of CD86 (a typical M1 marker) of microglia while decreased the expression of CD206 (a typical M2 marker), with a high M1/M2 ratio of 1.42, indicating that M2-like microglia were successfully repolarized to M1-like microglia (Fig. 4I). In terms of BMDMs polarization ability, we used CD86 to represent M1-like BMDMs while both CD163 and CD206 to characterize M2-like BMDMs. Similar to the microglia polarization results, ^MAPDC treatment led to a high M1/M2 ratio of 2.45, indicating that M2-like BMDMs were successfully repolarized to M1-like BMDMs (fig. S16). In comparison, treatment with control PDCs or free R848 resulted in a much lower M1/M2 ratio of approximately 1.0 or less than 1.0, suggesting that a large proportion of BMDMs remained M2 phenotype. CLSM images of differently treated BMDMs shared a common result that ^MAPDC treatment most effectively induced the repolarization from M2-like BMDMs to M1-like BMDMs (Fig. 4J, and fig. S17). To directly compare the repolarizing ability, we divided the fluorescence signal of CD86 by CD206 to reconstruct new fluorescence images. The images and quantitative analysis of fluorescence signal clearly demonstrated that ^MAPDCs had the strongest ability to induce repolarization (Fig. 4, J and K). The successful repolarization was confirmed by the significantly decreased secretion of anti-inflammatory cytokines in ^MAPDCs group, such as transforming growth factor–β1 (TGF-β1) and interleukin-10 (IL-10), and an increased secretion of proinflammatory cytokines, such as IL-6, IL-12 (p70), tumor necrosis factor–α (TNF-α), and interferon-inducible protein 10 (IP-10) (Fig. 4L and fig. S18). Together, we concluded that benefiting from the design superiority that Ang2 facilitated upper-to-bottom transport while cleavage of M peptide blocked bottom-to-upper transport, ^MAPDCs had the much-enhanced ability to cross in vitro bEnd.3 monolayer and retained in the bottom chamber and, consequently, to induce more GL261 cells apoptosis and BMDMs/microglia repolarization.

Biodistribution and pharmacokinetics

To extend our in vitro findings to in vivo performance, we assessed biodistribution and GBM-targeting specificity by labeling ^MAPDCs and control PDCs with DiD, a near-infrared red cell membrane fluorophore. ^MAPDCs were shown to efficiently encapsulate DiD with no apparent leakage through the DLS measurement and release experiment (fig. S19). After intravenous injection, compared with the faint fluorescence signal in the brains of control PDC-treated mice, especially for PEGylated ^MPPDCs and ^NPPDCs, the fluorescence signal in the brains of ^MAPDCs group mice was evident as early as 2 hours and peaked at 24 hours after injection (Fig. 5A). Semiquantitative analysis of fluorescence signal confirmed these findings that Ang2-modified PDCs had a higher distribution in the brain than PEGylated PDCs (Fig. 5B), indicating the excellent brain-targeting ability of Ang2-modified PDCs. However, it was still hard to determine the GBM-targeting specificity based on in vivo imaging, so we harvested brains at 48 hours and examined the fluorescence signal via ex vivo imaging. Excitingly, the fluorescence signal of ^MAPDCs was intensively localized to the GBM site with a much higher intensity as opposed to control PDCs, supported by the semiquantitative analysis (Fig. 5, C and D). To directly compare the GBM-targeting specificity, a G/B ratio calculated by dividing the fluorescence signal at the GBM site by that at normal brain was introduced (Fig. 5E). The G/B ratio of ^MAPDCs group was 2.0, which was higher than that of ^NAPDCs (1.5), ^MPPDCs (1.2), and ^NPPDCs (1.3). Ex vivo images of other organs showed lower accumulation of ^MAPDCs in the liver than ^NPPDCs (fig. S20). Consistent with ex vivo imaging, immunofluorescence staining of brain sections displayed a better accumulation of ^MAPDCs within GBM zone and less accumulation in other organs than those of control PDCs (Fig. 5, F and G, and fig. S21 and S22). These results collectively validated that ^MAPDCs could cross the BBB effectively and targeted GBM zone with higher specificity and retention. Similar to in vitro study, the enhanced GBM-targeting specificity and accumulation should be triggered by the design superiority that Ang2 facilitated BBB transport and cleavage of M pep blocked brain-to-blood efflux. To further confirm the superiority of M pep in enhancing drug retention at the GBM site, the fluorescence distribution of ^MAPDCs and ^NAPDCs in brains was visualized by light-sheet fluorescence microscopy (LSFM, Fig. 5, H and I, and movies S1 and S2). The tomography images of ^MAPDCs group expressed a stronger DiD fluorescence intensity at the GBM site than that of ^NAPDCs, confirming the introduction of M pep reduced drugs efflux and improved retention. The tomography images of ^MAPDCs group simultaneously revealed that the fluorescence signal of GAMs colocalized well with DiD signals, suggesting that ^MAPDCs could target GAMs with improved specificity.

Fig. 5. The optimized biodistribution and pharmacokinetics in ^MAPDCs.

(A) Representative in vivo bioluminescence imaging of orthotopic GL261 tumor (left) and fluorescence imaging of GL261-bearing mice at the indicated time points after intravenous administration of various DiD-labeled PDCs (right). (B) Average radiant efficiency (ARE) quantification of brain zone in (A) (n = 3). (C) Representative ex vivo fluorescence imaging of brains at 48 hours after administration. (D) ARE quantification of normal brain and the GBM site in (C) (n = 3). (E) Schematic illustration of the normal brain and the GBM site. (F) Representative frozen brain sections after staining with LRP1 (green) or CD31 (green) and DAPI (blue) at 48 hours after administration. Scale bar, 200 μm. (G) DiD intensity analysis of frozen brain sections in (F) (n = 3). (H and I) Mice brains cleared with iDISCO protocol and imaged in dibenzyl ether using LSFM. ^NAPDCs-treated mouse (H); ^MAPDCs-treated mouse (I). Scale bar, 1000 μm. (J and K) The plasma drug concentration as a function of time for CPT (J) and R848 (K) after intravenous administration of free drugs, ^MAPDCs, or ^NAPDCs (n = 3). The light blue and light yellow part represent the distribution phase and the elimination phase in the two-compartment model, respectively. (L and M) The pharmacokinetic parameters of free drugs, ^MAPDCs, or ^NAPDCs. Half-time (L); AUC, area under concentration-time curve (M) (n = 3). All data were given as means ± SD. Data in (D) were analyzed by one-way ANOVA followed by Turkey test. Data in (G) and (L) and (M) were analyzed by one-way ANOVA followed by LSD test.

To comprehensively understand the in vivo fate of ^MAPDCs, we next evaluated its pharmacokinetic profiles. The plasma drug concentration of ^MAPDCs group was higher than that of free drugs group for both CPT and R848 (Fig. 5, J and K). The half-life of ^MAPDCs group for CPT (37.4 min) and R848 (60.2 min) were almost two and four times longer than those of free CPT (19.6 min) and free R848 (16.4 min), respectively (Fig. 5L). The area under the plasma drug concentration-time curve (AUC) of ^MAPDCs group for CPT and R848 were increased from 5.39 to 30.3 μg/ml·min and from 70.6 to 254.5 μg/ml·min, respectively, compared with free drugs group (Fig. 5M). One of reasons contributing to the improved pharmacokinetic properties of ^MAPDCs was the self-assembling nanostructures of ^MAPDCs with hydrophobic drugs inside and hydrophilic peptide outside not only prevented drug degradation during circulation but also prolonged circulation time by reducing opsonization and subsequent clearance by mononuclear phagocytosis system (36). The other reason is that ^MAPDCs preferentially distributed to brain and retained here, which resulted in slower drugs clearance and consequently, prolonged half-life and increased AUC (Fig. 5, L and M, and fig. S23). The favorable biodistribution and pharmacokinetics of ^MAPDCs suggested its potential for the treatment of GBM.

Pharmacodynamics in GBM-bearing mouse model and anti-GBM immunity

To evaluate the anti-GBM effect, we treated orthotopic GL261-bearing mice with ^MAPDCs and control formulations at day 7 after tumor inoculation, once every 2 days, and harvested biological samples at day 20 (Fig. 6A). Hematoxylin and eosin (H&E) staining presented no noticeable tissue damages in the main organs after ^MAPDC treatment, supported by serum biochemistry tests showing that all biochemical indicators were almost within the normal range (figs. S24 and S25). Monitoring of body weight displayed that ^MAPDCs outperformed other groups in reversing body weight loss caused by GBM progression and prognosis (Fig. 6B). The survival curve of mice was also recorded, manifesting a significantly prolonged survival time after ^MAPDC treatment (Fig. 6C). The median and mean survival time of mice in ^MAPDCs group increased from 31 and 30 days to 51 and 52 days, respectively, compared with the saline group (Fig. 6D). By comparison, ^NAPDC treatment resulted in shorter median and mean survival time than ^MAPDCs, and free drugs treatment failed to improve survival benefit. This result suggested that the survival benefit was closely related to drug concentration at the GBM site. Next, we used H&E staining to directly determine the anti-GBM effect of ^MAPDCs (Fig. 6E). We unexpectedly found that ^MAPDC treatment remarkably inhibited the growth of GL261 cells, with only 0.47% of the brain was invaded, whereas 23% of brain was invaded in saline group. Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining showed a large number of apoptosis cells within the GBM site of ^MAPDCs group (1733 cells/mm²), which was much higher than that of control groups (Fig. 6F). Collectively, these results proved that ^MAPDCs had good biocompatibility and strong anti-GBM ability benefiting from the greatly increased drug concentration.

Fig. 6. ^MAPDCs exhibited robust anti-GBM effect and reprogrammed immunosuppressive TME.

(A) Schematic illustration of experimental design. Mice were orthotopically inoculated with GL261 cells at day 0, followed by five times of intravenous administrations starting on day 10. The bio-samples were collected at day 20 and subjected to corresponding analysis. (B and C) Changes in body weight (B) and survival curves (C) of differently treated mice after tumor inoculation. (n = 9). (D) The calculated mean survival day and estimated median survival day of differently treated mice after tumor inoculation (n = 9). (E) H&E staining images of brains and the percentage of GBM zone in the whole brain. (F) Immunohistochemistry sections of brains for TUNEL, CD4, CD8, and Foxp3, indicating the apoptotic cells, CD4⁺ T cells, CD8⁺ T cells, and T_reg cells, respectively. Scale bar, 50 μm (left). Quantification analysis of the number of positive cells per square millimeter and the ratio of T_reg cells to CD4⁺ T cells (right). (G) Brain immunofluorescence sections after staining with anti-CD86, anti-CD206 antibodies, and DAPI, indicating M1-GAMs (red) and M2-GAMs (green), respectively. Scale bar, 150 μm. (H) Volumetric tissue exploration and analysis (VTEA) scatterplots of CD86 versus CD206 in (G). (I) Normalized cytokines level in GBM homogenates (n = 3). Data in (B) and (D) were given as means ± SD and mean/median ± 95% confidence interval, respectively. Data in (C) and (D) were analyzed by log-rank (Mantel-Cox) test.

Encouraged by the exciting treatment outcome, we were curious about the anti-GBM immunity and GAMs-repolarizing state within TME. In ^MAPDCs group, immunohistochemical (IHC) staining analysis presented a much higher population of CD4⁺ T cells (2040 cells/mm²) and CD8⁺ T cells (1310 cells/mm²) while a much lower population of Foxp3⁺CD4⁺ T_reg (19 cells/mm²) within the GBM site, than those of control groups (Fig. 6F). The augmented CD4⁺ T cells and decreased Foxp3⁺CD4⁺ T_reg resulted in an extremely low percentage of T_reg within CD4⁺ T cells (0.1%, Fig. 6F). These results showed that ^MAPDC treatment greatly improved the infiltration of effector T cells and helper T cells, while reduced the differentiation of T_reg cells. We assumed that ICD effect induced by CPT partly contributed to this strong anti-GBM immunity; therefore, we detected the expression of CRT and HMGB1 in GBM cells. As expected, immunofluorescence staining of GBM-bearing brain slices showed that ^MAPDCs strongly induced the expression of CRT and the nucleus-to-cytoplasm translocation of HMGB1, which was consistent with that of GL261 cells in vitro (fig. S26). On the other hand, R848-induced M2- to M1-GAMs repolarization may also be associated with this strong anti-GBM immunity (37). This repolarization reversed the immune-desert TME to immune-stimulating TME, which was favorable for T lymphocytes infiltration. Meanwhile, GAMs phenotype is closely related to the function of T helper (T_H) cells, particularly for the T_H1, T_H2, and T_reg subpopulations. M1-GAMs are mainly responsive for promoting T_H1 cells differentiation and maintaining T_H1 immune response by secreting proinflammatory cytokines, whereas M2-GAMs are involved in promoting T_H2 cells and T_reg differentiation by secreting anti-inflammatory cytokines. Therefore, M2-to-M1 repolarization could promote the T_H2-to-T_H1 transformation and inhibit T_reg differentiation. To confirm this repolarization, we examined M1/M2-GAMs populations within GBM using immunofluorescence staining. Fluorescence images clearly showed that ^MAPDC treatment reduced M2-GAMs population while increased M1-GAMs population, with an M1/M2 ratio of 1.08, which was dramatically higher than other control groups (Fig. 6, G and H). The M2-to-M1 repolarization brought an increased secretion of proinflammatory cytokines (IP-10, IL-12, and IL-6) and reduced secretion of anti-inflammatory cytokines (IL-4) in the GBM homogenates after ^MAPDC treatment (Fig. 6I and fig. S27). These results potently verified that enhanced accumulation of R848 within the GBM site benefited from ^MAPDCs potentiated local anti-GBM immunity by T_H2-to-T_H1 transformation and T_reg cells inhibition along with repolarizing M2-GAMs to M1-GAMs. In addition, we performed immunofluorescence staining of the brain cryosections from differently treated mice with anti-CD31 and anti-fibronectin antibodies, respectively. The fluorescence signal of both CD31 (a typical marker for blood vessels) and fibronectin (a typical marker for extracellular matrix) in ^MAPDCs group were much weaker than that in other groups, suggesting that repolarization from M2-GAMs to M1-GAMs not only sufficiently inhibited angiogenesis but also reduced the expression of fibronectin, which promoted tumor cell invasion within TME (fig. S28).

To determine the extent the therapeutic effect of CPT is strengthened by R848, we conducted an additional animal experiment to investigate the survival time of orthotopic GL261-bearing mice treated with alone ^MAPDCs-C and ^MAPDCs (fig. S29A). The decrease in body weight of ^MAPDCs-treated mice was obviously reversed. (fig. S29B). More unexpectedly, all mice treated by alone ^MAPDCs-C died after 40 days, while all the mice in the ^MAPDCs group still survived (fig. S29C). This result clearly manifested that R848 resensitized chemotherapeutic toxicity and brought a better therapeutic effect toward GL261-bearing mice.

Adaptive anti-GBM immunity in peripheral lymphoid organs

Given that CPT could induce the release of tumor-associated antigens (TAA) and damage-associated molecular patterns (DAMP), both of which would be captured by antigen-presenting cells to initiate anti-GBM adaptive immunity. Meanwhile, R848 has long been used as an immune adjuvant to facilitate the maturation of dendritic cells (DCs) in addition to its role as an immunomodulator (38, 39). Therefore, we next determined the activation of adaptive immunity in peripheral lymphoid organs, including spleen and tumor-draining lymph nodes (TDLNs). ^MAPDC treatment led to the highest proportion of CD80⁺CD86⁺ dendritic cells in either spleen or TDLNs, validating that ^MAPDCs could induce the release of TAA and DAMP to a higher level than control PDCs and thus triggered more DCs maturation (Fig. 7, A and B). The mature DCs with stronger antigen-presenting ability might activate the T lymphocytes more efficiently. As expected, ^MAPDC treatment increased the proportion of cytotoxic CD8⁺ T cells and helper CD4⁺ T cells in either spleen or TDLNs (Fig. 7, C to F). Although the increase is moderate, ^MAPDC treatment remarkably reduced the proportion of Foxp3⁺ T_reg cells while increased the expression of granzyme B (GrB) of CD8⁺ T cells (Fig. 7, G to J). Specifically, in TDLNs, the proportion of Foxp3⁺CD4⁺ T_reg was six times lower and GrB⁺CD8⁺ cytotoxic T cells was five times higher in ^MAPDCs group than those of free drugs treatment. We concluded that ^MAPDCs could maximumly induce ICD effect and elicit the peripheral anti-GBM immunity in both spleen and TDLNs after treatment, which cooperated with local anti-GBM immunity to combat GBM (40, 41).

Fig. 7. ^MAPDCs elicited systemic anti-GBM immune response.

(A and B) Representative flow cytometry plots and quantification analysis of mature DCs (noted by CD80⁺ CD86⁺) within CD11c⁺ cells for spleen (A) and TDLNs (B). (C to F) Representative flow cytometry plots and quantification analysis of CD4⁺ (C) and CD8⁺ (E) in CD45⁺ cells for spleen, and of CD4⁺ (D) and CD8⁺ (F) in CD45⁺ cells for TDLNs. (G to J) Representative flow cytometry plots and quantification analysis of Foxp3⁺ in CD4⁺ T cells (G) and GrB⁺ in CD8⁺ T cells (I) for spleen, and of Foxp3⁺ in CD4⁺ T cells (H) and GrB⁺ in CD8⁺ T cells (J) for TDLNs. All data were given as means ± SD (n = 5) and analyzed by one-way ANOVA followed by Turkey test or Games-Hoswell test (Foxp3⁺ and GrB⁺ analysis in TDLNs).

DISCUSSION

Achieving an effective drug concentration at the tumor site is the most crucial factor for suppressing GBM development. Over the past few decades, therapeutic agents or regimens have been rapidly developed and their therapeutic effects on GBM have also been extensively explored. However, almost all of them fail to receive approval for clinical application, which was largely due to their very limited BBB transport as well as insufficient drug concentration. Effective delivery of therapeutic drugs across the BBB and to the GBM site at a sufficient concentration using brain-targeting drug delivery system is urgently demanded to unlock their therapeutic potential and enrich treatment regimens (42). PDCs are essentially a class of prodrugs consisting of peptide, drug, and linker through the covalent conjugation, similar to well-known antibody-drug conjugates. Compared to conventional nanocarrier-based drug delivery systems, PDCs-based drug delivery systems exhibit several advantages in terms of preparation, drug loading efficiency, biocompatibility, drug release, customizable application, etc. (i) Unlike nanocarrier-based drug delivery systems that rely on synthesized or naturally derived materials, PDCs can easily self-assemble into supramolecular nanostructure or hydrogel without additional materials, which is beneficial for scale-up preparation and potential clinical translation. (ii) PDCs have much higher biocompatibility since they omit materials that may induce potential toxicity or immunogenicity, particularly, these synthesized materials. (iii) The simple compositions endow PDCs with a higher drug encapsulating and loading efficiency. (iv) The covalent conjugation of PDCs can greatly prevent the drug from leakage before reaching target organs or cells, which can reduce side effect, improve bioavailability, as well as improve the pharmacodynamic effect. Meanwhile, by introducing stimulus-responsive linker, PDCs may acquire on-demand and controllable drug release behavior at specific site where the heterogenicity can be used as potential stimulus trigger. (v) On the basis of combinatorial chemistry principle, it is flexible and easy to customize PDCs with desirable targeting specificity, pharmacologic function, and release profile for precise disease treatment (28). In this study, the designed that ^MAPDCs can substantially cross BBB and remain in the GBM site, taking advantage of LRP-1–mediated transport and M pep–induced brain-to-blood efflux inhibition, providing a robust antitumor capability against GBM.

In terms of drug release, PDCs generally require endogenous or exogenous stimulus to trigger linker cleavage and drug release. However, stimulus-triggered drug release behavior may be questionable as some drugs cannot be released sufficiently as expected. The other question is that some drugs cannot be released as prototypes given to the introduction of additional groups, leading to compromised bioactivity. Using specific linkers that have excellent stimulus responsiveness and enable drug release in prototype form is of importance. In our study, we used disulfonyl-ethyl carbonate/carbamate (a reducible linker cleaved by GSH) to link CPT or R848 onto peptide. Once reaching the GBM site, M pep would be firstly cleaved, and then the cleaved ^MAPDCs without Ang2 modification were internalized by GBM cells and M2-GAMs. Subsequently, the disulfonyl-ethyl carbonate/carbamate would be cleaved by high-level GSH inside GBM cells and M2-GAMs to release CPT or R848 in their prototype form, which preserved their pharmacodynamic functions. To further ensure the targeting reliability and therapeutic effect, the stability of targeting peptide should also be taken into consideration. If tumor-targeting peptide is subjected to easy degradation and short half-life in the blood, the period leaving for the PDCs to be internalized into tumor cells is limited, which will affect the in vivo distribution of PDCs. Strategies to prolong the half-life of peptides while preserving their binding affinity have been widely developed, such as head-to-tail cyclization, disulfide bond cyclization, replacement of unnatural amino acids, peptidomimetics, and stapled peptides. Thus, in the future study, we will set out to optimize the stability of peptide to obtain more stable and reliable PDCs.

Our purpose of this design is to modulate GBM cells and GAMs, but the targeting specificity of ^MAPDCs toward both of two types of cells is questionable since the cleaved ^MAPDCs by MMP2 has not targeted peptide toward GBM cells and GAMs. Modulating the GBM cells and GAMs more precisely through introducing an additional peptide that targeted GBM cells or GAMs is the step that needs to be taken to optimize ^MAPDCs. For example, epidermal growth factor receptor is a promising target for GBM cells (43). M2 peptide identified from peptide library has been demonstrated to have a high affinity for macrophages with M2 phenotype (44). PDCs modified with M2 peptide can directly target M2-GAMs. Another peptide, α-peptide, has high and specific affinity to scavenge receptor B type 1, which can also be integrated into PDCs to target M2-GAMs (45).

In this study, ^MAPDCs is finely engineered on the basis of BBB and GBM microenvironment characteristics while retaining the bioactivity of its own components. Thanks to the bioactive peptides, it performs outstandingly in several experiments, suggesting a proof of concept of our initial design and that similar design aiming at other diseases may also work well. Results from in vitro BBB model indicate that ^MAPDCs can kill GL261 cells and revert BMDMs at cellular level, and in vivo experiments indicate the preferable biodistribution to GBM and improved pharmacokinetics of ^MAPDCs. These two results imply that ^MAPDCs could elicit GL261 tumor regression, as evidenced by the most robust survival benefit of ^MAPDCs. Meanwhile, we observed a robust local antitumor immune response in the TME as well as systemic immune perturbations in peripheral spleen and TDLNs. We speculate that substantial drugs accumulation induces robust local immune response, which further initial systemic immune response because immunity is coordinated across tissues (40). Given the delicate design and extensive characterization, this platform may serve as a paradigm for designing similar deliver systems.

MATERIALS AND METHODS

Experimental design

This study developed a self-assembled nanostructure consisted of two PDC supramolecules that are able to cross BBB efficiently, accumulate more in the GBM site, release drugs only in the presence of GSH, and finally to repolarize M2-GAMs into M1-GAMs for enhanced GBM chemo-immunotherapy. In vitro characterizations included ¹H nuclear magnetic resonance, MALDI-TOF, CD, DLS, TEM, SPR, HPLC, cytokine analysis, flow cytometry, and CLSM. The biodistribution, pharmacokinetics, and antitumor efficacy were evaluated in orthotopic GL261 tumor model. Biodistribution experiments were performed using in vivo image system (IVIS), and blood sample were subjected to HPLC. Body weight and survival time were monitored every day, and, at predetermined time point, the mice were euthanized to collect bio-samples, which were analyzed by flow cytometry, CLSM, and enzyme-linked immunosorbent assay (ELISA). The determination of sample size was based on previous experimental experience. All experiments were repeated at least three times except for brain slices evaluation experiments whose sample size was two.

Animals and cell lines

The animal study was carried out under the supervision of the Animal Care and Use Committee of Beijing Institute of Technology. C57BL/6J mice and BALB/c-Nude mice (8 weeks old, specific pathogen free) were purchased from Jiangsu GemPharmatech Ltd. (China). All animals were housed in an animal facility at Beijing Institute of Technology, and animal studies were performed in compliance with the guidelines outlined by Institutional Animal Care and Use Committee of the Beijing Institute of Technology. Unless otherwise stated here, bEnd.3 cells and BMDMs were cultured in Dulbecco’s modified Eagle’s medium (DMEM, PAN Biotech, Germany). GL261 and GL261-luc cells were cultured in DMEM containing 1% Hepes (PAN Biotech, Germany) and 1% l-glutamine (PAN Biotech, Germany). Microglia were cultured in minimum essential medium (Thermo Fisher Scientific Inc., USA) containing 1% l-glutamine (PAN Biotech, Germany). All media were supplemented with 10% fetal bovine serum (PAN Biotech, Germany) and 1% penicillin-streptomycin (Thermo Fisher Scientific Inc., USA).

GSH and MMP-2 detection

For the orthotopic GL261 tumor model, 5 × 10⁵ GL261 cells were inoculated in the right striatum of anesthetized male C57BL/6J mice using a brain stereotactic fixation device with a mouse adapter (46). The injection site was 1.8 mm lateral, 0.6 mm longitudinal, and 3 mm depth (day 0). On day 10, the mice were euthanized, and then normal and tumor zone of brain were dividually collected and homogenized in 1× phosphate-buffered saline (PBS) containing 0.5% Triton X-100. The GSH and MMP-2 content in homogenates were determined by respective ELISA kits (Beijing Solarbio Science & Technology Ltd., China) under the manufacturer’s instructions.

Microglia and BMDMs were isolated from C57BL/6J mice and were cultured for over 6 days. At the 7th day, IL-4 and IL-10 were introduced to the culture medium at a final concentration of 20 and 10 ng/ml, respectively. After 24 hours, the microglia and BMDMs were collected and subjected to sonication. The GSH content was then determined by ELISA kits under the manufacturer’s instructions.

SPR assays of Ang2

Recombinant mouse LRP1 (LMAI Bio Ltd., China) was immobilized on a CM5 sensor chip surface at a level of 3000 response units using a solution of LRP1 (20 μg/ml) in 10 mM sodium acetate under the manufacturer’s instructions (Biacore 8K, GE Healthcare, Sweden). The Ang2 (Nanjing Top-Peptide Biotechnology Ltd., China) and PEG2000 (Shanghai Ponsure Biological Ltd., China) were dissolved in PBS containing 0.005% Tween 20 and trace of dimethyl sulfoxide (DMSO), and then, the solution was allowed to flow through the CM5 sensor chip surface at a flow rate of 30 μl/min and 25°C. Throughout, the angle shift of minimum reflected intensity was recorded. The sensor chip’s regeneration was conducted by 15 s contact of PBS containing 0.005% Tween 20 at a flow rate of 30 μl/min.

CD spectrum

The secondary structure of Ang2, M pep (Nanjing Top-Peptide Biotechnology Ltd., China), and M-A (Nanjing Top-Peptide Biotechnology Ltd., China) were evaluated by J-1500 CD spectropolarimeter (JASCO, Japan) at 10-mm path length under a scanning speed of 200 nm/min at 25°C. The spectra were recorded from 190 to 350 nm with a bandwidth of 1 nm.

Critical micelle concentration

Nile red solution (Beyotime Biotechnology, China) was prepared at 2 mM in methanol. Then, 5 μl of Nile red solution was allowed to evaporate in a microcentrifuge tube under dark environment. Pre–^MAPDCs-C and pre–^MAPDCs-R were together dissolved in DMSO at various concentrations, after which 10 μl of solution was added into microcentrifuge tube containing Nile red. The solution was then dropped into 0.5 ml of ultrapure water with stirring and allowed to age overnight. The excitation wavelength was set at 550 nm, and the spectrum from 560 to 700 nm was recorded on microplate system (47).

^MAPDCs preparation and characterization

A DMSO mixture of pre–^MAPDCs-C (25 μl, 10 mg/ml) and pre–^MAPDCs-R (75 μl, 10 mg/ml) was added dropwise to stirred ultrapure water (5 ml). The resulting ^MAPDCs were washed with ultrapure water and collected by ultrafiltration for 15 min at 4000g (Ultracel membrane with 10 kDa nominal molecular weight limit; Millipore, Germany). Other control PDCs including ^NPPDCs, ^MPPDCs, and ^NAPDCs were prepared following the same procedures. DiD-labeled PDCs were also prepared by the above procedures but adding an additional 12.5 μl of DiD (1 mg/ml in DMSO) to the DMSO mixture. Hydrodynamic size measurement was performed using DLS with a Zetasizer Nano ZS90 (Malvern Instruments Ltd., UK) at a scattering light angle of 173°. The surface zeta-potential was measured with Zetasizer Nano ZS90 by using disposable folded capillary cells (DTS1070, Malvern, UK) under 10-V voltage. The structure was then characterized by JEM 1200EX TEM (JEOL Ltd., Japan).

Interaction of PDCs with LRP1

Various concentrations of PDCs were allowed to flow through CM5 sensor chip with immobilized LRP1, and the interaction strength was reflected by the angle shift of sensor chip. The brief procedures were same as described above.

GSH/MMP-2 responsiveness and release

Equal volumes of 20 mM GSH (Shanghai Aladdin Biochemical Technology Ltd., China) and 100 μM ^MAPDCs were mixed and incubated at 37°C and 100 rpm. At 0, 0.5, 1, 2, 4, 8, 12, 24, and 48 hours, 100 μl of the sample was taken and subjected to Brave HPLC (Changzhou Panna Instruments Ltd., China). The mobile phase was H₂O-acetonitrile (60:40, v/v) and H₂O-methanol (70:30, v/v) for CPT and R848, respectively. The DiD release from DiD-labeled ^MAPDCs was determined under GSH-free condition using Microplate Reader (BioTek Cytation 5 MF, USA) 0.5, 1, 2, 4, 8, 12, 24, 48, and 72 hours after dialysis.

For responsiveness experiments, 1 mM p-aminophenylmercuric acetate (MedChemExpress, China) was used to activate MMP-2 (10 μg/ml, Beijing Solarbio Science & Technology Ltd., China), which was subsequently mixed and incubated with equal volume of 100 μM ^MAPDCs at 37°C and 100 rpm. Then, free CPT and R848, and ^MAPDCs treated with/without GSH or activated MMP-2 were characterized using DLS, TEM, and HPLC, as described above.

Cellular uptake and endosomal escape

bEnd.3 or GL261 cells were incubated into a six-well plate at 1.5 × 10⁵ per well overnight. The cells were treated with 1 μM free drugs and PDCs. At 1, 4, and 12 hours, the cells were collected for cellular uptake determination using flow cytometry (LSRFortessa, BD Biosciences, USA).

To investigate the cellular uptake mechanism, bEnd.3 cells were respectively preincubated with a variety of uptake inhibitor including chlorpromazine, genistein, amiloride, and three concentrations of Ang2 for 1.5 hours followed by ^MAPDC treatment for 4 hours. One more group that bend.3 cells were cultured with ^MAPDCs at 4°C was set. The cells were collected and analyzed using flow cytometry without additional staining.

bEnd.3 cells treated with free drugs and PDCs were imaged by CLSM (LSM 980, Carl Zeiss, Germany) for cellular uptake and endosomal escape study. Briefly, lysotracker red was added into cell culture 15 min before collecting cells that were washed with 1 × PBS, fixed with cold 4% paraformaldehyde (PFA), and stained with DRAQ5 (BioLegend, USA), in that order. Last, the coverslips covered with cells were transferred onto glass microscope slides with a drop of antifade mounting media and imaged.

BBB was mimicked using a transwell cell culture system. bEnd.3 cells were cultured in the upper chamber at 3 × 10⁵ per well for 5 days. Twenty-four hours after the introduction of ^MAPDCs to the upper chamber, we collected and concentrated the bottom medium to obtain ^MAPDCs, which was analyzed by TEM and LC-3020 HPLC SYSTEM (Chin-Fine Technology, China) as described above.

Cytotoxicity

GL261 cells were incubated in 96-well plate at 3.5 × 10³ per well overnight. Various concentrations of free drugs and PDCs were added into 96-well plate. At 36 hours, cell viability was tested using CCK-8 assay (Shanghai Yeasen Biotechnology Ltd., China). Cells without treatment served as a control.

Cellular uptake with an in vitro BBB model

BBB was mimicked using a transwell cell culture system. bEnd.3 cells were cultured in the upper chamber at 3 × 10⁵ per well for 5 days, while GL261 cells were cultured in the lower chamber at 1.5 × 10⁵ per well overnight. The upper chambers with bEnd.3 monolayers, containing free drugs or PDCs, were then embedded in the lower chambers containing GL261 cells. At 4 and 12 hours, both of bEnd.3 and GL261 cells were collected and processed as described above for flow cytometry. In addition, GL261 cells were imaged by CLSM after washing, fixing, and staining with DRAQ5. To prove the integrity of bEnd.3 monolayers, they were incubated with anti–ZO-1 (1:500, Proteintech Group Inc., USA) at 4°C overnight followed by incubation with CL594-labeled goat antirabbit (1:200, Proteintech Group Inc., USA) at room temperature for 1 hour. Last, the coverslips covered with cells were transferred onto glass microscope slides with a drop of antifade mounting media containing 4′,6-diamidino-2-phenylindole (DAPI) and imaged.

In vitro BBB model was constructed, and cells were cultured as described above. Then, the upper chambers with bEnd.3 monolayers were embedded in the lower chambers containing GL261 cells. ^NAPDCs and ^MAPDCs were added into the lower chambers. At 0.5, 2, 4, and 12 hours, the cells were collected and processed for flow cytometry.

Apoptosis with an in vitro BBB model

In vitro BBB model was constructed, and cells were cultured as described above. The upper chambers with bEnd.3 monolayers were embedded in the lower chambers containing GL261 cells. Then, free drugs or PDCs (2 μM) were added into the upper chambers. At 24 hours, the culture medium and GL261 cells in the lower chamber were collected followed by centrifuging. The cells were reserved and processed with annexin V–fluorescein isothiocyanate (FITC)/propidium iodide apoptosis detection kit according to the manufacturer’s instructions (Shanghai Yeasen Biotechnology Ltd., China).

ICD effect with an in vitro BBB model

In vitro BBB model was constructed, and cells were cultured as described above. After the same treatment as described above, the GL261 cells were imaged by CLSM after washing, fixing, permeabilization, and staining with DAPI (Beyotime Biotechnology, China), FITC-phalloidin (1:1000; Sigma-Aldrich, USA), and CRT or HMGB1 (1:200; Proteintech Group Inc., USA). The secondary antibody was Coralite Plus 594–labeled goat antirabbit (1:500; Proteintech Group, Inc., USA). The HMGB1 concentration of supernatant of the bottom GL261 cells was also measured by ELISA kits (Elabscience Bionovation Inc., China) according to the manufacturer’s instructions.

Macrophages polarization with an in vitro BBB model

Microglia were cultured as previously described (48). BMDMs were cultured by standard macrophage colony-stimulating factor (MedChemExpress, China) culture over 6 days, as previously described (2 × 10⁶ per well) (49). On the 7th day, IL-4 and IL-10 were introduced to the culture medium at a final concentration of 20 and 10 ng/ml, respectively. After 24 hours, the culture medium was removed and fresh standard culture medium containing activated MMP-2 (50 ng/ml) was added. Then, the upper chambers with bEnd.3 monolayers were embedded in the lower chambers containing M2-like BMDMs or microglia. After this, bEnd.3 cells were incubated with free drugs or PDCs for 24 hours. The BMDMs or microglia in the lower chambers were washed, fixed, blocked, stained, and finally resuspended in 1× PBS for flow cytometry (FACSAriaII, BD Biosciences, USA). The antibodies for BMDMs included FITC anti-F4/80, phycoerythrin (PE)/Cy7 anti-CD11b, allophycocyanin (APC) anti-CD86, and PE anti-CD163 (BioLegend, USA). The antibodies for microglia included AF488 TMEM119 (Thermo Fisher Scientific Inc., USA), APC anti-CD86, and PE anti-CD206 (BioLegend, USA). M2-like BMDMs and microglia without treatment served as a control.

The BMDMs were also imaged by CLSM after a series of processing. Specifically, the BMDMs were incubated with anti-CD86 (1:200; BioLegend, USA) and anti-CD206 (1:50; Santa Cruz Biotechnology Inc., USA) at 4°C overnight followed by incubation with FITC-labeled donkey antirabbit (1:200; BioLegend, USA) and AF647-labeled goat antimouse (1:50; BioLegend, USA) at room temperature for 1 hour. Last, the coverslips covered with cells were transferred onto glass microscope slides with a drop of antifade mounting media containing DAPI and imaged. M2-like BMDMs without treatment served as a control.

Cytokine secretion with an in vitro BBB model

As described above, the upper chambers with bEnd.3 monolayers, containing free drugs or PDCs, were embedded in the lower chambers containing M2-like BMDMs. At 24 and 48 hours, 100 μl of culture medium from the lower chamber was taken and analyzed with TGF-β1, IL-10, IL-6, IL-12(p70), TNF-α, and IP-10 ELISA kits (Beijing Solarbio Science & Technology Ltd., China). M2-like BMDMs without treatment served as a control.

Biodistribution

The orthotopic GL261-luc tumor model was established in BALB/c-Nude mice as described above. After 10 days, DiD-labeled PDCs were intravenously injected through the tail vein at a dose of DiD (2 mg/kg), whose signals were measured with IVIS (IVIS Spectrum, PerkinElmer, USA) at predetermined time points after mice were anesthetized. Before measuring bioluminescence signals from GL261-luc tumor, d-luciferin was administrated intraperitoneally at a dose of 150 mg/kg. Images were analyzed with living image software.

All mice were euthanized followed by transcardial perfusion with 20 ml of 1 × PBS and 4% PFA after 48 hours. The main organs were harvested, and DiD signals in these organs were determined using IVIS. After this, the organs were fixed in 4% PFA, dehydrated in 30% sucrose, and embedded in optimal cutting temperature compound (Sakura Finetek USA Inc., USA). Cryosections were cut at a thickness of 20 μm using a freezing microtome (FS800A, RWD Life Science Ltd., China). Brain sections were then blocked with phosphate-buffered saline with Tween 20 (PBST) containing 10% goat serum and permeabilized with 0.5% Triton X-100 in PBST. Primary antibodies were added overnight at a dilution of 1:100 for LRP1 or 1:100 for CD31 (Abcam Ltd., USA), at 4°C. Cy3-labeled goat antirabbit (Proteintech Group, Inc., USA) as secondary antibody was added at a dilution of 1:500 for 2 hours at room temperature. Coverslips were mounted with antifade mountant containing DAPI. Other organs sections were processed with the same procedures but only stained with DAPI. Images were taken using CLSM.

Whole-brain clearing and imaging

DiD-labeled ^NAPDCs and ^MAPDCs were intravenously administrated to GL261-bearing C57BL/6J mice at an equivalent dose of DiD (2 mg/kg). At 24 hours, the mice were euthanized to collect brains, which were processed with a standard iDISCO protocol that helped to immunolabel large tissue samples for volume imaging (50). Briefly, fixed brain samples were dehydrated and rehydrated with methanol-H₂O mixture, immunolabeled in PBST, heparin, DMSO, and bull serum albumin, and cleared in dibenzyl ether after the second dehydration. These processes would lead to a clear brain, which was subsequently volume-imaged in the microscope chamber (FSLight) filled with dibenzyl ether.

Pharmacokinetics

Free drugs, ^NAPDCs, and ^MAPDCs were intravenously administrated to GL261-bearing C57BL/6J mice at an equivalent dose of CPT (1.5 mg/kg) and R848 (4.5 mg/kg). At prescheduled time points, 20 μl of venous blood was collected through infraorbital venous plexus strictly following “Decision Tree for Blood Sampling (Mice)” casted by the National Center for the Replacement, Refinement, and Reduction of Animal in Research (51). All mice were intragastrically supplemented with normal saline after each blood collection. Blood samples were centrifuged for 5 min at 1500g to obtain the upper plasma that was added to 1 ml of 0.1% acetic acid in ethyl acetate. Completely vortexing before the samples were further centrifuged at 3000g for 5 min. The supernatant was reserved and evaporated under N₂ atmosphere at 37°C to obtain the residue, which was redissolved with 50 μl of methanol. After vortexing for 1 min, the samples were centrifuged for 10 min at 10,000g, and 15 μl of supernatant was analyzed with corresponding LC-MS method (TSQ Fortis, Thermo Fisher Scientific Inc., USA). The drug plasma concentration-time data were fitted with two-compartment model by PKSolver to calculate pharmacokinetic parameters (52).

Pharmacodynamics and immune profiling in tumor models

At an equivalent dose of CPT (1.5 mg/kg) and R848 (4.5 mg/kg), free drugs and PDCs were intravenously injected to GL261-bearing mice via tail vein according to the indicated schedule. The saline-treated group served as a control. Meanwhile, alone ^MAPDCs-C (1.5 mg/kg CPT) and ^MAPDCs (0.75 mg/kg CPT and 2.25 mg/kg R848) were also intravenously administrated to access the extent the therapeutic effect of CPT is strengthened by R848. The mouse body weight and survival were monitored. On the 2nd day of completing the treatment, blood was collected before five mice of each group were euthanized. The heart, liver, spleen, lung, kidney, TDLNs, and brain were harvested. Two brains of each group were sectioned and stained with H&E, TUNEL, IHC (CD4, CD8, and Foxp3), and immunofluorescence (CD86, CD206, CRT, HMGB1, CD31, and fibronectin). The sections were visualized by CLSM or digital pathology slide scanners (VS200, Olympus, Japan). Tumor zone of three brains of each group were homogenized in 1× PBS containing 0.5% Triton X-100. IP-10, IL-12, IL-6, and IL-4 content in homogenates were determined by ELISA kits (Beijing Solarbio Science & Technology Ltd., China). Single cells from spleen and TDLNs were divided into three groups and stained with corresponding antibodies: T_reg lymphocyte (1:200 for CD45, 1:200 for CD3, 1:80 for CD4, and 1:50 for Foxp3), cytotoxic T lymphocyte (1:200 for CD45, 1:200 for CD3, 1:80 for CD8, and 1:50 for GrB), and mature DCs (1:200 for CD45, 1:80 for CD11c, 1:80 for CD80, and 1:80 for CD86). All antibodies were purchased from BioLegend (USA). Flow cytometry was then performed to analyze the stained cell suspensions.

Biocompatibility

The harvested heart, liver, spleen, lung and kidney as mentioned above were sectioned and stained with H&E. The serum was isolated from collected blood followed by biochemical analysis.

Software and statistics analysis

All CLSM images were quantified by Fiji 2.14.0. The colocalization analysis in Fig. 3G was performed with Coloc 2 plugin, and the data were transferred to MATLAB (MathWorks, R2020b, USA) for plotting. The divide function embedded in Fiji was used to process Fig. 4J. The VTEA 1.1.8 plugin was applied to quantify the M1- and M2-GAMs fraction in Fig. 6G (53). The H&E and IHC images in Fig. 6 were quantified with QuPath v0.5.1. All gate strategies and data process for flow cytometry were conducted in FlowJo (Becton Dickinson & Company, 10.8.1, USA; figs. S30 to S33). For other data, Origin 2023 (OriginLab Corporation, USA) was used for plotting.

All statistical analyses were performed using IBM SPSS Statistics 26 (USA). All data were assumed to be subject to normal distribution. Homogeneity test of variance were primarily performed for comparison between multiple groups. For equal variances assumed, the normal post hoc multiple comparison is applied; otherwise, Games-Hoswell analysis method is used. The significance level was set at 0.05. Detailed statistical analysis methods were included in figure captions.

Supplementary Materials

The PDF file includes:

Supplementary Methods

Figs. S1 to S33

Legends for movies S1 and S2

Other Supplementary Material for this manuscript includes the following:

Movies S1 and S2