HCQ inhibitor – Chk2 Inhibitori

Aggregable Nanoparticles-Enabled Chemotherapy and Autophagy Inhibition Combined with Anti-PD-L1 Antibody for Improved Glioma Treatment

ABSTRACT: Glioma treatment using targeted chemotherapy is still far from satisfactory due to not only the limited accumulation but also the multiple survival mechanisms of glioma cells, including up-regulation of both autophagy and programmed cell death ligand 1 (PD-L1) expression. Herein, we proposed a combined therapeutic regimen based on functional gold nanoparticles (AuNPs)-enabled chemotherapy, autophagy inhibition, and blockade of PD-L1 immune checkpoint. Speciﬁcally, the legumain-responsive AuNPs (D&H-A- A&C) could passively target the glioma site and form in situ aggregates in response to legumain, leading to enhanced accumulation of doXorubicin (DOX) and hydroXychloroquine (HCQ) at the glioma site. HCQ could inhibit the DOX-induced cytoprotective autophagy and thus resensitize glioma cells to DOX. Parallelly, inhibiting autophagy could also inhibit the formation of autophagy-related vasculogenic mimicry (VM) by glioma stem cells. In vivo studies demonstrated that D&H-A-A&C possessed promising antiglioma eﬀect.Moreover, cotreatment with anti-PD-L1 antibody was able to neutralize immunosuppressed glioma microenvironment and thus unleash antiglioma immune response. In vivo studies showed D&H-A-A&C plus anti-PD-L1 antibody could further enhance antiglioma eﬀect and eﬃciently prevent recurrence. The eﬀectiveness of this strategy presents a potential avenue to develop a more eﬀective and more personalized combination therapeutic

Glioma is the most common primary tumor of the central nervous system (CNS) in adults, accounting for 15% of all brain tumors.1 Despite the standard treatment of maximal surgical resection followed by radiotherapy and adjuvant chemotherapy, the overall survival time is only approXimately 15−17 months.2 Because of the blood-brain barrier (BBB), most chemotherapeutic agents are subjected to poor delivery eﬃciency.3 The advancement of nanotechnology makes it easy to overcome BBB and transport chemotherapeutic agents to the glioma site.4,5 However, most glioma-targeted nano- medicines still face challenges of heterogeneous distribution and insuﬃcient accumulation.6 Currently, numerous stimulus- responsive nanoparticle delivery systems capable of releasing therapeutic cargos at sites of interest in an on-demand manner are explored.7 Our previous study validated that legumain-chemotherapy. Therefore, rationally designing a combination regimen will be promising to achieve better therapeutic eﬀect or even complete glioma regression.Emerging studies revealed that autophagy serves as a cytoprotective mechanism that allow cancer cells to survive from chemotherapy.9,10 Inhibiting cytoprotective autophagy using chloroquine was proven to resensitize the glioma cells to Temozolomide (TMZ) and improve the treatment outcome of glioma.11,12 Thus, it can be expected that a combination of hydrochloroquine (HCQ) and DOX can augment the chemotherapeutic eﬀect on glioma. In addition, autophagy has also been reported to be involved in the formation of vasculogenic mimicry (VM), an alternative microvascular circulation independent of vascular endothelial growth factor (VEGF)-driven angiogenesis by glioma stem cells (GSC).13−15responsive aggregable gold nanoparticles (AuNPs) were able to increase doXorubicin (DOX) accumulation at the glioma site.8 Despite the enhanced chemotherapeutic eﬀect, glioma cells can still develop several mechanisms to survive

Autophagy-associated VM drives glioma cells to be more adaptive to the alternation in the tumor microenvironment. Thus, we assumed that inhibiting autophagy may simulta- neously inhibit the formation of VM, thus accelerating glioma cells apoptosis.Given the diﬀusely inﬁltrating growth of glioma cells into surrounding CNS parenchyma, the precision with which immune cells home in on and eradicate glioma cells makes immunotherapy an attractive treatment modality.16,17 As is the case with many other tumors, glioma cells express or secrete several immunosuppressive molecules to suppress the anti- glioma immunity, especially the programmed death ligand 1 (PD-L1).18,19 PD-L1 is able to speciﬁcally bind to programmed death 1 (PD-1), a T-cell coinhibitory receptor with distinct biological function and ligand speciﬁcity, mediating an immune escape mechanism.20 Antibodies blocking the PD-1/PD-L1 pathway demonstrated potential antitumor immunity and durable cancer regression.21 Accu- mulating evidence revealed that cancer chemotherapy resulted in up-regulated expression of PD-L1.22−24 Thus, it can be putative that chemotherapy combined with PD-L1 checkpoint blockade will unleash powerful glioma curative potential.

In this study, we proposed a combination regime for glioma treatment, which combined legumain-responsive nanoparticle- enabled chemotherapy and autophagic interference with immune checkpoint blockade (Scheme 1). The functional drug delivery system, D&H-A-A&C, consisted of two nano-particles. One was Ala-Ala-Asn-Cys-Lys-polyethylene glycol- thiol (AK-PEG-SH) modiﬁed citrate-stable AuNPs coloaded with pH-sensitive DOX and pH-sensitive HCQ prodrug (D&H-A-AK) through the “SH-Au” chelation. The other was 2-cyano-6-amino-benzothiazole-polyethylene glycol-thiol (CABT-PEG-SH) modiﬁed AuNPs coloaded with DOX and HCQ (D&H-A-CABT). D&H-A-A&C was able to passively target the glioma site via enhanced permeability and retention (EPR) eﬀect after systemic injection and form in situ aggregates in response to the overexpressed legumain, resulting in enhanced accumlation of DOX and HCQ. As a consequence, DOX and HCQ could exert a synergistic eﬀect in inhibiting glioma cell proliferation. The in vivo survival study demonstrated that D&H-A-A&C possessed an excellent treatment outcome to glioma-bearing mice. More importantly, D&H-A-A&C plus anti-PD-L1 antibody further enhanced antiglioma eﬀect compared to D&H-A-A&C or anti-PD-L1 antibody treatment alone. Besides, the mice that survived after treatment also showed an improved memory immunity. The eﬀectiveness of this strategy will open the mind to design more eﬀective combinational regimens for glioma treatment.In Vitro Charaterization of AuNPs-D&H-A&C. DOX and HCQ ﬁrst underwent modiﬁcation to conjugate onto AuNPs together with pH-responsive release property. pH- responsive DOX was synthesized according to the procedures described in our previous study.25 The synthetic route of HCQ and characterization were described in Supporting Information.

Figure 1. In vitro characterization and evaluation of D&H-A-A&C. (A) Size distribution by number of D&H-A-A&C and control nanoparticles; inner graphs indicate the corresponding TEM images. (B) UV−vis spectral of D&H-A-A&C, A-A&C, free DOX, and free HCQ. OL1*, OL2*, and OL3* indicate the characteristic absorption peaks of D&H-A-A&C overlaid with A-A&C, free DOX, and free HCQ, respectively. (C) Fluorescent recovery of DOX from D&H-A-A&C in PBS (pH 5.0) at diﬀerent time intervals. (D) Hydrodynamic size of D&H-A-A&C and control nanoparticles in HEPES (pH 5.0) after incubation with legumain for diﬀerent time intervals, *p < 0.05 and **p < 0.01 (n = 3). (E) Confocal images of C6 cells incubated with D&H-A-A&C and control nanoparticles for 24 h and labeled with Lysotracker Red; bar represents 10 μm. (F) Lysotracker Red positive puncta (red) are calculated from three high-powered pictures, ***p < 0.001. (G) Confocal images of C6 cells after incubation with D&H-A-A&C, D-A-A&C, and D&H-A-P for 4 and 24 h; bar indicates 50 μm. (H) Semiquantitative mean ﬂuorescent intensity (MFI) of DOX calculated from G, ***p < 0.001 indicates the statistic diﬀerence versus D&H-A-A&C (n = 3). (I) Flow cytometry analysis of cellular uptake of D&H-A-A&C, D-A-A&C, and D&H-A-P by C6 cells, *p < 0.05, **p < 0.01, and ***p < 0.001 indicate the statistic diﬀerence versus D&H-A-A&C; N.S. indicates no statistic diﬀerence (n = 3).(Figures S1 and S2). After conjugating with DOX, HCQ, and functional fragments, the hydrodynamic size of D&H-A-AK and D&H-A-CABT displayed an increase compared with bare AuNPs, suggesting that prodrugs and functional fragments were successfully conjugated onto AuNPs (Table S1). After centrifugation and resuspension together, the size of D&H-A- A&C was 38.10 ± 1.6 nm with a negative surface charge. Besides, the transmission electron microscope (TEM) image showed a uniform morphology and consistent size (Figure 1A). Similarly, size distribution and morphologies of PEGylated DOX&HCQ-co-loaded AuNPs (D&H-A-P), AK&- CABT-modiﬁed HCQ-loaded AuNPs (H-A-A&C), and AK&CABT-modiﬁed DOX-loaded AuNPs (D-A-A&C) were close to D&H-A-A&C (Table S2). To further identify that DOX and HCQ was successfully conjugated onto AuNPs, the UV−vis scanning spectrum was performed. The characteristic absorption peaks of D&H-A-A&C that contained these individual peaks belong to DOX, HCQ, AK&CABT-modiﬁed AuNPs (A-A&C), respectively (Figure 1B). Also, the characteristic absorption peaks of control nanoparticles matched well with each component (Figure S3). The results indicated that both DOX and HCQ were successfully conjugated onto the surface of functional nanoparticles. As demonstrated in our previous study, the ﬂuorescent intensity of DOX could be quenched after conjugating onto AuNPs due to the nanosurface energy transfer (NSET) eﬀect.25 Therefore, we further determine the ﬂuorescence recovery of DOX to evaluate pH-responsive release. After incubation in pH 5.0 PBS, the ﬂuorescent intensity of DOX was low at the beginning whereas it increased over time, suggesting DOX could be released in an acidic condition (Figure 1C). The release of DOX from D-A-A&C and D&H-A-P showed a similar behavior, which was in a time-dependent manner (Figure S4). When incubated in pH 6.0 PBS, all nanoparticles showed relatively slower ﬂuorescence recovery of DOX, indicating that the release of DOX was also in a pH-dependent manner (Figure S5). Legumain-Responsiveness. Although the responsiveness of AK&CABT-modiﬁed nanoparticles to legumain was evaluated in our previous work,8 the legumain-responsiveness of D&H-A-A&C remained unknown. After incubation with legumain for 12 h, the size of D&H-A-A&C increased from 44.51 ± 0.97 nm to 310.17 ± 12.48 nm (Figure 1D), whereas the size of D&H-A-P increased slightly from 41.67 ± 1.26 nm to 63.89 ± 3.38 nm. In comparison, the size of D-A-A&C increased from 42.43 ± 1.19 nm to 345.47 ± 5.34 nm and H- A-A&C increased from 40.11 ± 1.19 nm to 347.27 ± 6.33 nm. The size increase of D&H-A-A&C was close to D-A-A&C and H-A-A&C, indicating that D&H-A-A&C still possessed the legumain-responsiveness. To further conﬁrm the legumain- responsiveness, we determined the size change of D&H-A- A&C after incubation with 10% fetal bovine serum (FBS). The size of D&H-A-A&C had a slight increase over time and was close to control nanoparticles, which may be attributed to protein absorption (Figure S6A). A similar trend was observed in the turbility of D&H-A-A&C and control nanoparticles incubated with 10% FBS (Figure S6B). Even when the concentration of FBS increased to 50% and 90%, the turbilities of D&H-A-A&C and control nanoparticles demonstrated a slight and close increase (Figure S7). All the results attested that D&H-A-A&C had negligible responsiveness to FBS and possessed good serum stability.Evaluation of Lysosomotropic Effect of HCQ. HCQ is a lysosomal tropic agent capable of neutralizing lysosomal acidic microenvironment, resulting in lysosomal dysfunction.26 To evaluate the lysosomal neutralizing capacity of D&H-A-A&C, lysosomal labeling was determined after incubation for 24 h. Plenty of lysosome tagged ﬂuorescent puncta were observed in C6 cells treated with D-A-A&C or free DOX, which was close to that in untreated cells, suggesting DOX may be not able to neutralize lysosomal pH condition (Figure 1E). However, less puncta were found in cells treated with free HCQ and H-A- A&C, suggesting HCQ could be released from AuNPs and prevent lysosomal labeling. In addition, the ability of HCQ to prevent lysosomal labeling was in a concentration-dependent manner (Figure S8). In comparison, cells treated with D&H-A- A&C also showed less puncta than those treated with D-A- A&C, indicating D&H-A-A&C also possessed the ability to prevent lysosomal labeling. Also, the puncta were also more than those treated with D&H-A-P, owing to the enhanced intracellular HCQ accumulation by D&H-A-A&C. Semi- quantitative analysis of the number of puncta further supported that D&H-A-A&C could eﬃciently prevent lysosomal labeling (Figure 1F). As described in our previous study, the enhanced intracellular accumulation was attributed to the fact that AK&CABT-modiﬁed AuNPs could form aggregates in response to legumain within lysosomes, which in turn restricted the exocytosis of AuNPs by cells.8 In Vitro Cellular Uptake. The increased pH condition may impede the pH-sensitive release of DOX and HCQ. Therefore, it was necessary to investigate whether the increase of pH aﬀected the release of DOX and HCQ from D&H-A- A&C. Confocal images showed no signiﬁcant diﬀerence of ﬂuorescent intensity between cells treated with D&H-A-A&C and D-A-A&C after 4 and 24 h incubation. However, the ﬂuorescent intensity of both was stronger than that treated with D&H-A-P, especially after 24 h incubation. The results implied that DOX could be eﬃciently released from D&H-A- A&C despite of the increased pH condition (Figure 1G,H). Additionally, ﬂow cytometry analysis also demonstrated that the ﬂuorescent intensity in cells treated with D&H-A-A&C showed the highest ﬂuorescent intensity after incubation for 24 h, further supporting that HCQ-induced pH increase has no inﬂuence on the release of DOX (Figure 1I). As such, it could be inferred that the release of HCQ itself may not be aﬀected by the increased pH condition. After systemic administration, D&H-A-A&C immediately interacted with a variety of speciﬁc and nonspeciﬁc blood plasma protein, leading to the formation of protein corona. Therefore, to evaluate whether protein corona inﬂuence the legumain responsiveness of D&H-A- A&C, we next investigated the cellular uptake after preincubation with 20% mouse plasma for 24 h. The ﬂuorescence of D&H-A-A&C in C6 cells after incubation for another 24 h was close to that of D-A-A&C but was higher than that of D&H-A-P (Figure S9). Similar uptake behaivor was demonstrated even after preincubated with 50% mouse plasma. The results indicated that D&H-A-A&C after coating with protein corona still possessed the legumain responsive- ness.Evaluation of Autophagic Manipulation. To investigate the ability of D&H-A-A&C to manipulate autophagy, we transfected C6 cells with microtube-associated protein 1 light chain 3 (LC3) fused eRFP to acquire eRFP-LC3-expressing C6 cells (C6 eRFP-LC3 cells). Figure 2. In vitro evaluation of autophagy manipulation. (A) Confocal images of eRFP-LC3-expressing C6 cells treated with D&H-A-A&C and control formulations for 24 h, red puncta represent autophagosomes, bar represents 20 μm. (B) Autophagosome positive puncta are calculated from three high-powered pictures, *p < 0.05. (C) Western blotting analysis of LC3 expression in C6 cells treated with D&H-A-A&C and control nanoparticles for 4 and 24 h. (D) Semiquantitative ratio of LC3-II/LC3-I calculated from C using ImageJ, ***p < 0.001 (n = 3) which LC3-I and LC3-II are two main forms.27,28 When autophagy is in a low level, LC3-I is distributed homoge- neously in the cytoplasm and can be processed to LC3-II by speciﬁc intracellular enzymes upon initiating autophagy. In comparison, LC3-II is capable of binding to autophagosome membrane, leading to a concentrated localization in autophagosome.29,30 Therefore, the ﬂuorescence of LC3-I is more likely to be a dispersed form while the ﬂuorescence of LC3-II is more likely to be a vacuolar form after post- transcriptional translation. In untreated cells, the ﬂuorescence was mainly dispersed, suggesting low autophagy level and LC3- I was the main form in normal conditions (Figure 2A, Figure S10). After treatement with D-A-A&C or free DOX, the numbers of ﬂuorescent puncta in C6 cells increased, suggesting chemotherapy could induce autophagy and thus increase LC3- II formation (Figure 2B). However, the numbers were less than that in cells treated with H-A-A&C or free HCQ, mainly because that HCQ could eﬃciently inhibit autophagy by blocking autophagosomes degradation. In comparison, cells treated with D&H-A-A&C displayed much more ﬂuorescent puncta than that treated with D-A-A&C and H-A-A&C, probably owing to a synergistic eﬀect in increasing LC3-II formation by DOX and HCQ. Additionally, the number was also more than that treated with D&H-A-P, further supporting that D&H-A-A&C had the superiority in increasing intra- cellular accumulation of DOX and HCQ. LC3 Expression. As described above, LC3-I and LC3-II are two important indicators of autophagy, which could be used for reﬂecting the autophagic modulation. To further conﬁrm the autophagic modulation eﬀect of D&H-A-A&C, we next determined the LC3 expression using Western blotting analysis. After 4 h of incubation, the LC3-II expression in C6 cells treated with D&H-A-A&C demonstrated a relatively higher expression compared with control nanoparticles (Figure 2C). With time extended to 24 h, the LC3-II expression in cells treated with D&H-A-A&C further increased and was much higher than that treated with control nanoparticles. The results suggested D&H-A-A&C could induce more LC3-II formation, largely because DOX induces more LC3-I transforming to LC3-II by inducing autophagy, whereas HCQ simultaneously blocks LC3-II recycling to LC3-I by inhibiting autophago- somes degradation. Here we introduced a ratio calculated by the intensity of LC3-II to LC3-I (LC3-II/LC3-I ratio) to directly evaluate the autophagy interference (Figure 2D). After 4 h of incubation, cells treated with D&H-A-A&C showed the highest LC3-II/LC3-I ratio compared to control nanoparticles, indicating more LC3-I transformed into LC3-II. After 24 h of incubation, the LC3-II/LC3-I ratio in cells treated with D&H- A-A&C further increased and much higher than that of control nanoparticles. The results further conﬁrmed that D&H-A-A&C possessed a stronger ability to manipulate autophagy in C6 cells.Autophagesome Formation. To directly observe the intracellular accumulation of autophagosomes that resulted from D&H-A-A&C, we applied TEM to observe the number and morphology of autophagic vacuoles. TEM images showed no obvious autophagic vacuoles in untreated cells. After treatment with D&H-A-A&C for 24 h, obvious autophagic vacuoles could be observed, which was more than that treated with D&H-A-P, D-A-A&C, or H-A-A&C (Figure 3A). The Figure 3. In vitro evaluation of autophagosome and legumain-responsiveness. (A) TEM images of C6 cells after incubation with D&H-A-A&C and control formulations for 24 h, red arrows represent AuNPs aggregates, blue arrows represent autophagic vacuoles, and bars represent 1 μm. (B) Percentage incubated dose (% ID) of cellular uptake of D&H-A-A&C and control nanoparticles after incubation with C6 cells for 24 h, *p < 0.05 (n = 6). % ID is calculated by the gold content in cells divided by the incubated gold content. (C) Flow cytometry analysis of C6 cells double- stained with Annexin V-FITC/PI after incubation with diﬀerent formulations for 24 h; untreated cells were used as negative control. (D) Percentage of apoptosis and necrosis of C6 cells incubated with diﬀerent formulations for 24 h, **p < 0.01 (n = 3)results further conﬁrmed that D&H-A-A&C could induce more autophagosomes accumulation. In addition, obvious AuNPs aggregates could be observed in cells treated with D&H-A-A&C, similar to that treated with D-A-A&C or H-A- A&C. However, cells incubated with D&H-A-P showed no obvious AuNPs aggregates inside. The results validated that AK&CABT-modiﬁed nanoparticles were able to form intra- cellular aggregates in response to legumain.ICP-OES Analysis. Inductively coupled plasma−optical emission spectrometer (ICP-OES) analysis demonstrated the percentage incubated dose (% ID) of cellular uptake of D&H- A-A&C was 9.29 ± 0.56% after 24 h incubation (Figure 3B). The result was close to that of D-A-A&C (9.54 ± 0.33%) and H-A-A&C (8.78 ± 0.34%) whereas higher than D&H-A-P (6.99 ± 0.19%). The results further supported that legumain- responsive intracellular aggregates could lead to enhanced cellular accumulation and coloading of DOX and HCQ has no signiﬁcant inﬂuence on legumain-responsiveness.Apoptosis. Inhibiting autophagesome degradation will accelerate the cell apoptosis because the ability to address the cytotoXicity of DOX was deﬁcient. Therefore, the cytotoXicity of C6 cells incubated with D&H-A-A&C and control formulations was then evaluated by MTT assay. After 24 h incubation, cells treated with D&H-A-A&C showed the lowest cell viability compared with control nanoparticles (Figure S11), suggesting D&H-A-A&C could eﬀectively inhibit cells proliferation. To further evaluate the cells apoptosis induced by D&H-A-A&C, the Annexin V-FITC/PI double staining assay was performed (Figure 3C). After 24 h incubation, the early and late apoptosis percentage of cells treated with D&H-A-A&C was 42.3% and 18.3%, which was much higher than those treated with D&H-A-P (20.0%, 10.8%), D-A-A&C (24.6%, 14.4%), and H-A-A&C (26.1%, Figure 4. In vitro evaluation of VM destruction. (A) In vitro tube formation of GSC after incubation with diﬀerent formulations for 12 h, bar represents 20 μm. (B) Semiquantitative analysis of tubular channels captured in A using Image-pro plus software (n = 4). (C) Western blotting analysis of MMP-2, MMP-9, and VE-Cadherin expression in GSC treated with diﬀerent formulations for 24 h. (D−F) Semiquantitative analysis of the ratio of these three protein expressions to GAPDH expression calculated from C using ImageJ (n = 3)11.4%), respectively. The results indicated that D&H-A-A&C could induce more cell apoptosis due to synergistic eﬀect of DOX and HCQ, as well as the aggregation-induced accumulation. Moreover, the total percentage of apoptosis and necrosis of D&H-A-P, H-A-A&C, and D-A-A&C was 36.3 ± 0.9%, 39.2 ± 0.1% and 41.5 ± 0.7%, respectively (Figure 3D). In contrast, D&H-A-A&C demonstrated the highest percentage of 61.4 ± 0.7%, supporting the enhanced cytotoXicity of D&H-A-A&C.Destruction of VM. To determine the ability of D&H-A- A&C to inhibit VM formation by GSC, we further assessed the in vitro tube formation using a reported matrigel-model.31,32 The GSC without any treatment was prone to aggregate rapidly, twist, and ﬁnally form an obvious tubular web within 12 h (Figure 4A). However, after treatment with D&H-A-A&C for 12 h, GSC only formed a tiny and local tubular web, suggesting the formation of tubelike channels was signiﬁcantly inhibited. In comparison, GSC treated with D&H-A-P and D- A-A&C were able to form a more integrated tubular web compared with D&H-A-A&C. However, GSC treated with H- A-A&C showed a similar tubular web to D&H-A-A&C, indicating HCQ played a key role in inhibiting the formation of tubelike channels. Similar phenomenon was observed in GSC treated with free HCQ, whereas tubelike channels were observed after treatment with free DOX. The reason why HCQ possessed stronger tube destruction was likely because HCQ can inhibit autophagy, thus inhibit the autophagy-related formation of VM.13 Moreover, the quantitative percentage of tube formation was determined to further conﬁrm the eﬀectiveness of D&H-A-A&C in inhibiting VM formation (Figure 4B).Molecular Mechanism. To investigate the molecular mechanism involved in the VM destruction, VM-related indicators such as matriX metalloproteinase-2 (MMP-2), matriX metalloproteinase-9 (MMP-9), and vascular endothelial cadherin (VE-Cadherin) expression levels were determined by Western blotting (Figure 4C). Figure 5. In vivo evaluation of antiglioma eﬀect. (A) Three-dimensional volume-rendering images of C6 glioma-bearing mice brains after intravenous injection with D&H-A-A&C and D&H-A-P for 24 h, red dash line represents the mock gold morphology using the collected signal at glioma site, bars represent 1 mm. (B) Kaplan−Meier survival analysis of C6 glioma-bearing mice after intravenous injection with diﬀerent formulations (n = 10). (C) TUNEL staining of C6 glioma-bearing mice brains collected at the day after last administration, brown puncta represent the apoptotic body and bar represents 20 μm. (D) Semiquantitative percentage of positive TUNEL staining was calculated using IHC Prolifer analysis, *p < 0.05 (n = 3). (E) CD34-PAS double staining of C6 glioma-bearing mice brains collected on the day after the last administration with diﬀerent formulations, pink staining represent the VM, and bar represents 10 μm.MMP-2, MMP-9, and VE-Cadherin were down-regulated in GSC compared with untreated cells. Especially, GSC treated with D&H-A-A&C showed the lowest expression of the three proteins, indicating D&H-A-A&C could eﬃciently down- regulate their expression in GSC. The expression of these three proteins in GSC treated with free HCQ was lower than free DOX, further identifying the crucial role of HCQ in suppressing VM signaling pathway. However, the VM destructing eﬀect of DOX was probably due to its cytotoXicity. Moreover, the semiquantitative analysis further conﬁrmed the eﬀectiveness of D&H-A-A&C in down-regulating the ex- pression of VM-related MMP-2, MMP-9, and VE-Cadherin (Figure 4D−F). Glioma Targeting. Computed tomography (CT) is one of the most useful diagnostic tools applied in biological imaging, which is characterized by high spatiotemporal resolution.33 Increasing studies reported that AuNPs are optical contrast agents for CT imaging due to their high atomic number, electron density, and high X-ray photon cross-section capture of gold.34 Although our previous studies validated that this legumain-responsive nanoplatform could selectively accumu- late at glioma site,8 in vivo distribution of D&H-A-A&C remained unknown especially when coloaded with DOX and HCQ. Therefore, we further evaluated the glioma-targeting eﬃciency of D&H-A-A&C by three dimension (3D) whole head volume rendering CT imaging (Figure 5A). After 24 h of intravenous injection, the CT signal of D&H-A-A&C could be observed in the brain and was precisely localized at glioma site, validating their good glioma-targeting eﬃciency of D&H-A- A&C. In contrast, there was no obvious signal in glioma site after treatment with D&H-A-P, suggesting that D&H-A-A&C could selectively deliver to glioma site. Figure 6. Evaluation of antiglioma immunity of combination therapy. (A) Kaplan−Meier survival analysis of C6 glioma-bearing mice treated with D&H-A-A&C plus anti-PD-L1 antibody, anti-PD-L1 antibody, D&H-A-A&C, N.S. (n = 10). (B) Flow cytometry analysis of mature DCs (CD83+CD11c+) in spleen after treatment, *p < 0.05, **p < 0.01, ***p < 0.001. (C) Flow cytometry analysis of mature DCs (CD86+CD11c+) in spleen after treatment, *p < 0.05, **p < 0.01, ***p < 0.001. (D) IHC staining of mice brains collected the day after last administration, CD4, CD8, and FoXp3 were staining brown, bars represent 40 μm. (E) Semiquantitative analysis of percentage of CD4+ T cells, CD8+ T cells, and FoXp3+ Treg of total T cells using ImageJ, *p < 0.05, **p < 0.01, ***p < 0.001 represent statistical signiﬁcance versus N.S. group (n = 3). (F) Ratio of CD4+ T cells to FoXp3+ Treg and CD8+ T cells to FoXp3+ Treg; *p < 0.05, **p < 0.01, ***p < 0.001 represent statistical signiﬁcance versus N.S. group (n = 3). In Vivo Antiglioma Effect. Encouraged by in vitro and in vivo investigation, we further evaluated the antiglioma eﬀect of D&H-A-A&C by monitoring survival time (Figure 5B, Tables S3 and S4). Compared with normal saline (N.S.) group, C6 glioma-bearing mice treated with free DOX or HCQ showed a slightly prolonged survival time. The median survival time after treatment with free DOX and free HCQ was 36 days and 30 days, respectively, mainly attributed to that BBB restricted transcytosis of free DOX or HCQ. After treatment with D&H-bodies in the glioma region than that treated with control formulations. Semiquantitative analysis of positive staining further supported this ﬁnding, suggesting that D&H-A-A&C could induce more glioma cells apoptosis (Figue 5D). All the results indicated that D&H-A-A&C could eﬃciently improve the antiglioma eﬀect, largely because in situ legumain-induced aggregation could enhance accumulation of DOX and HCQ at glioma site. Besides, DOX and HCQ could further function synergistically to induce more glioma cell apoptosis. To A-A &C, the median survival time was proloned to 56 days,investigate the potential systemic toXicity of D&H-A-A&C,suggesting this functional nanoplatform could eﬃciently overcome BBB and deliver more drug to glioma site. In comparison, the median survival time is longer than that treated with D&H-A-P (42 days), D-A-A&C (44 days), and H- A-A&C (38 days). Moreover, hematoXylin-eosin (H&E) staining showed the number of glioma cells nuclei in these groups treated with drug-loaded nanoparticles was much lower than that treated with free HCQ or DOX (Figure S12A), indicating the enhanced therapeutic eﬀect of nanoparticle- based formulations. In contrast, the number of glioma cells nuclei in D&H-A-A&C group was lowest, suggesting D&H-A- A&C possessed a better eﬃcacy to inhibit glioma cell proliferation. The results were further supported by the percentage of healthy brain tissue occupied by glioma cells (Figure S12B). To further reveal the underlying mechanism involved in the inhibition of glioma cells proliferation, terminal deoXynucleotidyl transferase-mediated dUTP-biotin nick end labeling (TUNEL) assay was performed (Figure 5C). After treatment with D&H-A-A&C, there were more apoptotic H&E staining assay on major organs was carried out. Typical myocardial injury in the heart treated with free DOX was observed, attesting to the severe cardiotoXicity of DOX (Figure S13). However, there was no obvious myocardial injury in mice treated with D&H-A-A&C, mainly owing to the reduced distribution in heart and slower DOX release. Additionally, there were also no pathological aberrances in other organs after treatment, indicating that D&H-A-A&C could decrease systemic toXicity of DOX and HCQ.CD34-PAS Staining. Moreover, the ability of D&H-A- A&C to inhibit VM formation in vivo was investigated by CD34-periodic acid-Schiﬀ (PAS) double staining (Figure 5E). CD34 staining is used for identifying the endothelium in glioma tissue sections and PAS staining is used for determining the basement membrane of tumor blood vessels.35 CD34-PAS double staining demonstrated obvious pink tubular structures with a dispersed distribution at glioma site in the N.S. group. After treatment with D-A-A&C and free DOX, the pink tubular structures decreased compared to N.S. treatment, probably because chemotherapy could prevent GSC proliferation and thus prevent the VM formation. Additionally, the tubular structures in H-A-A&C and HCQ groups also decreased, largely because HCQ could inhibit cytoprotective autophagy and thus inhibit VM formation. In contrast, the pink tubular structures in D&H-A-A&C group further decreased with a much lower density compared to control groups. The results indicated that D&H-A-A&C could not only prevent glioma cells proliferation but also inhibit the VM formation by GSC. Combination Therapy with Anti-PD-L1 Antibody. Immune escape mediated by up-regulating PD-L1 expression on tumor cell surface is one of the major obstacles that restrict the eﬀectiveness of antitumor chemotherapy.36 To determine whether PD-L1 checkpoint blockade could improve treatment eﬀect of D&H-A-A&C and inhibit glioma recurrence, we thus combined two therapeutic agents. Glioma-bearing mice treated with anti-PD-L1 showed a signiﬁcantly prolonged survival time compared to the N.S. group, implying anti-PD-L1 antibody possessed excellent antiglioma eﬀect due to the improved antiglioma immunity (Figure 6A). In comparison, D&H-A- A&C showed a similar therapeutic outcome to anti-PD-L1 antibody, whereas D&H-A-A&C plus anti-PD-L1 antibody showed a higher survival rate. Furthermore, the mean survival time of glioma-bearing mice treated with D&H-A-A&C plus anti-PD-L1 antibody was 60.9 days (Tables S5 and S6), which is much longer than anti-PD-L1 antibody group (51.3 days), D&H-A-A&C groups (47.0 days), and N.S. group (22.9 days). We next assess the antiglioma eﬀect of combination therapy using H&E staining. After treatment with anti-PD-L1 antibody or D&H-A-A&C, the number of glioma cells nuclei decreased obviously, denoting their excellent therapeutic potential (Figure S14A). In comparison, the number of nuclei in brain parenchyma treated with D&H-A-A&C plus anti-PD-L1 antibody group further decreased, which was supported by the percentage of healthy brain tissue occupied by glioma cells (Figure S14B). The results indicated that D&H-A-A&C plus anti-PD-L1 possessed an enhanced antiglioma eﬀect. Evaluation of DCs Maturation. To further evaluate immunological proﬁle after treatment, the population of mature dendritic cells (DCs) in spleen was determined. CD11c, CD83, and CD86 as a biomarker of mature DCs were chosen to analyze the mature proportion. Flow cytometry analysis demonstrated that treatment with D&H-A-A&C could induce higher CD83+CD11c+ and CD86+CD11c+ DCs proportion than that treated with N.S. (Figure 6B, C and S15). The results were probably attributed to that D&H-A- A&C could induce glioma apoptosis and neoantigen release, which might stimulate the DCs maturation. In contrast, treatment with anti-PD-L1 antibody led to relatively higher proportion of mature DCs, considering that blocking PD-L1 checkpoint on DCs could promote the DCs maturation and recruitment. As expected, when combining two therapeutic agents together, D&H-A-A&C plus anti-PD-L1 induced the highest mature DCs proportion. More importantly, the improved mature DCs proportion indicated improved antigen presenting ability, probably resulting in more CD4+ T helper cells and CD8+ cytotoXic T cells inﬁltrating into the glioma site to unleash the antiglioma immunity. Immunohistochemistry Staining. Next, we move on to the determination of the T cells population implicated in antiglioma immune response. Immunohistochemistry (IHC) staining showed a higher population of both CD4+ T cells and CD8+ T cells inﬁltrating into the glioma site after treatment with D&H-A-A&C compared with N.S. treatment. Because D&H-A-A&C could induce more DCs maturation, these mature DCs then entered into the glioma site to recruit more T cells (Figure 6D). However, IHC staining also showed the highest population of FoXp3+ regulatory T cells (Treg), conﬁrmed by the semiquantitative analysis of positive staining (Figure 6E). The contradictory results indicated chemotherapy could induce more DCs maturation, whereas it simultaneously promoted immunological tolerance by up-regulating PD-L1 expression on glioma cells. As a consequence, the up-regulated PD-L1 not only led to T cells dysfunction but also induced more regulatory T cell (Treg) diﬀerentiation. After treatment with anti-PD-L1, the population of both CD4+ T cells and CD8+ T cells also increased compared with N.S. treatment, whereas the proportion of FoXp3+ Treg decreased compared with D&H-A-A&C treatment. The result further validated that PD-L1 checkpoint blockade could increase the population of T cells and inhibit Treg diﬀerentiation. In comparison, treatment with D&H-A-A&C plus anti-PD-L1 antibody led to a higher population of both CD4+ T cells and CD8+ T cells but lower population of FoXp3+ Treg than that treated with anti-PD-L1 antibody or D&H-A-A&C alone. To directly evaluate the immunomodulatory eﬀect of this combination regimen, the ratio of CD4+ T cells to FoXp3+ Treg and ratio of CD8+ T cells to FoXp3+ Treg were determined (Figure 6F). Treatment with D&H-A-A&C plus anti-PD-L1 antibody demonstrated the highest ratio of both CD4+ T cells/FoXp3+ Treg and CD8+ T cells/FoXp3+ Treg, supporting D&H-A-A&C plus anti-PD-L1 could induce more T cells inﬁltrating into the glioma site and simultaneously inhibit Treg diﬀerentiation. Moreover, the pro- inﬂammatory cytokines, such as interferon-γ (IFN-γ) and tumor necrosis factor-α (TNF-α), are essentially regarded as antitumor cytokines. The release of IFN-γ is suppressed within immunosuppressed glioma, especially resulting from the PD- L1/PD-1 interaction. Accumulating evidence indicated that immune checkpoint blockade led to the up-regulation of IFN- γ.20 Thus, we further measured the secretion of IFN-γ and TNF-α in the serum after treatment. In D&H-A-A&C plus anti-PD-L1 antibody group, IFN-γ and TNF-α in serum demonstrated the highest concentration (Figure S16). However, interleukin-10 (IL-10), an immunosuppressive molecule, demonstrated a lower concentration than that in D&H-A-A&C and N.S. group. Together, all these results validated that combination therapy could improve antiglioma immunity and neutralize the immunosuppressive glioma microenvironment. Glioma Rechallenge. Memory immunity is important in preventing tumor recurrence and depleting neoantigen with similar phenotype once exposed to the immune system.37 To investigate memory immunity, the surviving glioma-bearing mice after treatment (survival time >65 days) were reimplanted with C6 cells in the contralateral brain. After glioma rechallenge, the surviving mice treated with D&H-A- A&C plus anti-PD-L1 antibody showed a higher survival rate compared to that treated with D&H-A-A&C or anti-PD-L1 antibody alone (Figure S17). In D&H-A-A&C plus anti-PD-L1 antibody group, H&E staining demonstrated no obvious glioma cells in brain parenchyma. The results suggested that treatment with D&H-A-A&C plus anti-PD-L1 could eﬃciently prevent glioma recurrence, supported by the percentage of healthy brain tissue occupied by glioma cells that was lowest (Figure S18). Despite relatively poor eﬃciency of anti-PD-L1 antibody in preventing glioma recurrence

Figure 7. Immunologic memory evaluation. (A) Flow cytometry analysis of immune cells in peripheral blood, draining lymph nodes and spleen at day 11 after glioma rechallenge. The number of CD3+ T lymphocytes, T memory cells (CD3+CD44[high] CD62L[low]), T helper cells (CD3+CD4+), and cytotoXic T cells (CD3+CD8+) were processed by Flowjo software (n = 3). (B) IFN-γ in serum determined by ELISA assay at day 11 after glioma rechallenge (n = 3)combination therapy, whereas the number of glioma cells nuclei was fewer than that in D&H-A-A&C group. The results indicated the ability of anti-PD-L1 antibody to induce memory immunity whereas D&H-A-A&C had limited potential in inducing memory immunity.T Cells Population Analysis. Although IHC staining showed a lower ratio of CD4+ T cells to FoXp3+ Treg and CD8+ T cells to FoXp3+ Treg in D&H-A-A&C plus anti-PD-L1 antibody group than that in anti-PD-L1 antibody group, whereas it was much higher than that in D&H-A-A&C group (Figure S19). The results conﬁrmed that combination therapy could not only induce memory immunity but also eﬀectively inhibited the FoXp3+ Treg diﬀerentiation. Moreover, the immune cells in peripheral blood, spleen, and draining lymph CD3+ CD44[high]CD62L[low] memory T cells in peripheral blood and DLNs. Although the percentage of CD3+CD44[high]CD62L[low] memory T cells was relatively lower in spleen than that treated with anti-PD-L1 antibody, it was much higher than that in D&H-A-A&C group. The results validated the eﬀectiveness of combination therapy in inducing more memory T cells diﬀerentiation. To further quantitatively evaluate the T cells recruited by memory T cells after the rechallenge, the percentage of CD4+ helper T cells and CD8+ cytotoXic T cells were determined. In D&H-A-A&C plus anti- PD-L1 antibody group, the CD3+CD4+ helper T cells and CD3+CD8+ cytotoXic T cells showed the higher percentage in DLNs compared to anti-PD-L1 antibody and D&H-A-A&C treatment alone, suggesting that more CD4+ helper T cells and nodes (DLNs) at day 11 after being rechallenged were CD8+ cytotoXic T cells were recruited.

To evaluate the analyzed by ﬂow cytometry (Figures 7A, S20−S22). Treatment with D&H-A-A&C plus anti-PD-L1 antibody exhibited the highest percentage of CD3+ T cells in peripheral blood, spleen, and DLNs compared to anti-PD-L1 antibody or D&H-A-A&C treatment alone. Because memory T cells play an important role in maintaining the memory immune response, we next determined the percentage of memory T cells in the previous listed tissues. Treatment with D&H-A-A&C plus anti-PD-L1 ant i bod y s h ow ed the h ighest p e rc enta ge o f peripheral memory immunity, we continued to detect the concentration of IFN-γ in serum (Figure 7B). In anti-PD-L1 antibody group, the IFN-γ concentration in serum was much higher than that in D&H-A-A&C group, further identifying that PD-L1 checkpoint blockade led to increasing the release of IFN-γ. In comparison, in D&H-A-A&C plus anti-PD-L1 antibody group the concentration of IFN-γ was highest in serum. Additionally, the concentration of TNF-α was also highest in serum in D&H-A-A&C plus anti-PD-L1 antibody group (Figure S23). These results suggested combination therapy could not only induce stronger antiglioma immunity but also maintain a higher level of memory immunity to prevent glioma recurrence.

In summary, we developed a combination therapy that combined nanoparticle-enabled chemotherapy and autophagy inhibition (D&H-A-A&C) with PD-L1 checkpoint blockade (anti-PD-L1 antibody). D&H-A-A&C could form aggregates within glioma site in response to legumain after systemic injection, resulting in enhanced accumlation of DOX and HCQ. As a consequence, DOX and HCQ exert a synergistic eﬀect in inhibiting glioma cells proliferation. In vitro studies validated the responsiveness of D&H-A-A&C to legumain and the eﬀectiveness in inhibiting autophagy. In vivo survival studies demonstrated that D&H-A-A&C signiﬁcantly pro- longed the survival of C6 glioma-bearing mice. More importantly, combining D&H-A-A&C with anti-PD-L1 anti- body further enhanced the antiglioma eﬀect compared with D&H-A-A&C or anti-PD-L1 treatment alone. Moreover, combination therapy also induced glioma-speciﬁc memory immunity to prevent recurrence. The eﬀectiveness of this strategy may provide insight into developing more eﬀective combination therapy for glioma treatment.

Discussion. The development of traditional glioma-targeted delivery system based on EPR eﬀect-mediated passive that high PEG modiﬁcation and small AuNPs size, which could reduce the plasma protein absorption.45 As for the functional molecules, diﬀerent molecules modiﬁcation will induce diﬀer- ent contents and compositions of protein corona, which is positively correlated with the amount and molecular weight of molecules. The formation of protein corona is critically determined by the physicochemical properties of nanoparticle (i.e., nanoparticle composition, size, shape, surface property) and environmental factor (i.e., gradients of plasma, kinetic equilibrium constants, circulation time, and temperature).46−48 Therefore, understanding and regulating the protein corona formation is essential for minimizing the adverse eﬀects by protein corona and maintaining or even promoting the speciﬁc functionalities of nanoparticles.Accumulating evidence has demonstrated that various anticancer therapies, including chemotherapy, radiotherapy, and targeted therapy, up regulate autophagy in diﬀerent cancer cell types.49 Similar to the function in tumor progression, autophagy also plays a paradoXical role in either increasing or diminishing their anticancer activity.50 On the one hand, autophagy functions as a prodeath mechanism by inducing autophagic cell death which is distinctive with type I programmed cell death (apoptosis).

On the other hand, autophagy is activated as a shared prosurvival mechanism to protect some cancer cells against anticancer activity. DOX-delivery or receptor-mediated active delivery is still challenging induced genotoXic stress activated tumor suppressor p53 due to not only the biological complexity of glioma but also the physiochemical properties of nanoparticles.38 Stimulus-respon- sive drug delivery system based on speciﬁc tumor micro- environment stimuli (acidic pH, up-regulated enzyme, reductive, and hypoXia) has been widely explored.39 This class of stimulus-responsive nanoparticles is inactivated under normal physiological conditions during systemic circulation but is activated under the trigger of these stimuli once they are delivered into tumor microenvironment. The activated nano- particles are accompanied by change of their physiochemical properties, such as size, morphology, surface charge, or chemistry, which is more favorable for site-speciﬁc “on- demand” targeting, penetration, endocytosis, retention, or drug release.40 Glioma also hold these stimuli, which can be used as therapeutic targets. Legumain is a well-conserved lysosomal cysteine protease and has a highly restricted speciﬁcity for hydrolysis requiring an asparagine at the P1 family, which is a critical checkpoint protein mediating apoptosis.51 However, the activation of the p53 family contributes to induction of autophagy by suppressing the mammalian target of rapamycin (mTOR).51,52 Autophagy ﬂuX begins with isolation with double-membrane structures or a phagephore. These membrane structures elongate and expand, which simultaneously sequester or engulf organelles and cytoplasmic proteins. At this step, LC-3 is recruited to conjugate onto the phagephore membrane to complete the formation of autophagosome. The autophagosomes mature with acidiﬁcation by the H+-ATPase and fuse with lysosomes to become autolysosomes.9 Eventually, the sequestered contents and inner membrane of the autolysosome are degraded by lysosomal hydrolases for recycling, leading to the increased resistance. Therefore, adequate modulation of cytoprotective autophagy has been renewed as reliable approaches to augment DOX-induced cytotoXicity. Chlor-site of substrate.41 Emerging studies reveal that legumain expression is positively associated with the invasiveness of malignancy and is up-regulated in glioma.42 In our previous study, we validated the legumain-responsive AuNPs-DOX- A&C could form in situ aggregates to enhance DOX accumulation in glioma site after systemic administration.8 Therefore, in this study we applied this AK&CABT-function- alized AuNPs to systemically deliver chemotherapy with autophagy inhibitor together.
A major issue faced by injected nanoparticles is the immediate interaction with blood plasma once entered into blood, leading to the formation of protein corona on nanoparticles surface.43 As a consequence, the coating of protein corona can alter the physicochemical properties of nanoparticles and thus may aﬀect their speciﬁc functionality.44 Our previous study validated AuNPs-A&C still possessed the in vitro legumain-responsiveness after preincubation with mouse plasma protein.8

In this study, we also conﬁrmed the legumain responsiveness of D&H-A-A&C even after preincu- bation with mouse plasma. These results were probably due to oquine (CQ) and its derivative HCQ are lysosomotropic agents that can raise lysosomal pH, thereby blocking autophagesome fusion with lysosome. Increasing preclinical studies and ongoing clinical trials have been evaluated in the combination of CQ or HCQ with DOX to improve antiglioma eﬀect.53 Moreover, Wu et al. has reported that the formation of VM was promoted by bevacizumab-induced autophagy in GSC, which was induced by reactive oXygen species (ROS)- mediated kinase domain insert receptor (KDR) activation.13 Although it is still unclear whether DOX-induced autophagy could induce VM formationt, the results indicated that inhibiting autophagy could aﬀect VM formation.
The success of immune checkpoints such as PD-1, PD-L1, and cytotoXic T lymphocyte antigen 4 (CTLA-4) blockade has led to a paradigm shift in cancer treatment.54 Studies have revealed an upregulated expression of immune checkpoints on glioma cells, especially PD-L1, and the PD-1/PD-L1 pathway play a key role in glioma progression.55 Strategy using PD-1/ PD-L1 inhibitors for glioma immunotherapy is attracting increasing attention. Nevertheless, many patients do not have a good response to monotherapy approaches and alternative strategies are thereby required to achieve optimal therapeutic beneﬁt. Combination therapies with PD-1/PD-L1 blockade will be needed to overcome resistance and broaden the clinical utility of immunotherapy.56 PD-1/PD-L1 inhibitors are now actively being investigated in combination with an ever- widening spectrum of anticancer therapies, such as chemo- therapy, radiotherapy, photothermal therapy, and photo- dynamic therapy.57 Chemotherapy has been considered to play a role in the treatment of almost all newly diagnosed diﬀuse glioma.58 However, systemic chemotherapy to glioma remains unsatisﬁed due to the presence of BBB which restricts their distribution to glioma cells.

Nanoparticle delivery system can eﬃciently overcome BBB and mediate the targeting delivery of chemotherapy to glioma cells, resulting in enhanced accumulation.5 Nanoparticle-based chemotherapy demonstra- ted improved treatment outcome to glioma whereas it reduced systemic toXicity. Despite the enhanced accumulation, chemo- therapy can further induce PD-L1 up-regulation on glioma cells, leading to the chemoresistance.36 Therefore, the combination of PD-L1 inhibitors with chemotherapy will hold promising antiglioma eﬃcacy. Several clinical trials are studying the combination of checkpoint inhibitors with Temozolomide, which strengthen the rationale for combina- tion regimens.59 However, the cross-link between PD-L1 expression and diﬀerent chemotherapeutics are still far from being explored.36 Therefore, an improved understanding of how chemotherapy aﬀects the antiglioma immunity is crucial for designing more powerful and individual combination therapy. Moreover, rational design of new immunomodulatory HCQ inhibitor nanomedicine provides opportunities for overcoming the physiological and pathological barriers that are intrinsic to glioma microenvironment and improves delivery eﬃciency.