Background: Glioblastoma survival remains unchanged despite continuing therapeutic innovation. Herein, we aim to (i) develop chimeric antigen receptor (CAR) T cells with a specificity to a unique antigen, carbonic anhydrase IX (CAIX), which is expressed in the hypoxic microenvironment characteristic of glioblastoma, and (ii) demonstrate its efficacy with limited off-target effects. Methods: First we demonstrated expression of CAIX in patient-derived glioblastoma samples and available databases. CAR T cells were generated against CAIX and efficacy was assessed in 4 glioblastoma cell lines and 2 glioblastoma stem cell lines. Cytotoxicity of anti-CAIX CAR T cells was assessed via interferon gamma, tumor necrosis factor alpha, and interleukin-2 levels when co-cultured with tumor cells. Finally, we assessed efficacy of direct intratumoral injection of the anti-CAIX CAR T cells on an in vivo xenograft mouse model using the U251 luciferase cell line. Tumor infiltrating lymphocyte analyses were performed. Results: We confirm that CAIX is highly expressed in glioblastoma from patients. We demonstrate that CAIX is a suitable target for CAR T-cell therapy using anti-CAIX CAR T cells against glioblastoma in vitro and in vivo. In our mouse model, a 20% cure rate was observed without detectable systemic effects. Conclusions: By establishing the specificity of CAIX under hypoxic conditions in glioblastoma and highlighting its efficacy as a target for CAR T-cell therapy, our data suggest that anti-CAIX CAR T may be a promising strategy to treat glioblastoma. Direct intratumoral injection increases anti-CAIX CAR T-cell potency while limiting its off-target effects.

Background: The tumor immune microenvironment can influence the prognosis and treatment response to immunotherapy. We aimed to develop a non-invasive radiomic signature in high-grade glioma (HGG) to predict the absolute density of tumor-associated macrophages (TAMs), the preponderant immune cells in the microenvironment of HGG. We also aimed to evaluate the association between the signature, and tumor immune phenotype as well as response to immunotherapy. Methods: In this retrospective setting, total of 379 patients with HGG from three independent cohorts were included to construct a radiomic model named Radiomics Immunological Biomarker (RIB) for predicting the absolute density of M2-like TAM using the mRMR feature ranking method and LASSO classifier. Among them, 145 patients from the TCGA microarray cohort were randomly allocated into a training set (N=101) and an internal validation set (N=44), while the immune-phenotype cohort (N=203) and the immunotherapy-treated cohort (N=31, patients from a prospective clinical trial treated with DC vaccine) recruited from Huashan Hospital were used as two external validation sets. The immunotherapy-treated cohort was also used to evaluate the relationship between RIB and immunotherapy response. Radiogenomic analysis was performed to find functional annotations using RNA sequencing data from TAM cells. Results: An 11-feature radiomic model for M2-like TAM was developed and validated in four datasets of HGG patients (area under the curve = 0.849, 0.719, 0.674, and 0.671) using MRI images of post contrast enhanced T1-weighted (T1CE). Patients with high RIB scores had a strong inflammatory response. Four hub-genes (SLC7A7, RNASE6, HLA-DRB1 and CD300A) expressed by TAM were identified to be closely related to the RIB, providing important evidence for biological interpretation. Only individuals with a high RIB score were shown to have survival benefits from DC vaccine [DC vaccine vs. Placebo: median progression-free survival (mPFS), 10.0 mos vs. 4.5 mos, HR=0.17, P=0.0056, 95%CI=0.041-0.68; median overall survival (mOS), 15.0 mos vs. 7.0 mos, HR=0.17, P =0.0076, 95%CI=0.04-0.68]. Multivariate analyses also confirmed that treatment by DC vaccine was an independent factor for improved survival in the high RIB score group. However, in the low RIB score group, DC vaccine was not associated with improved survival. Furthermore, a radiomic nomogram based on the RIB score and clinical factors could efficiently predict the 1-, 2-, and 3-year survival rates, as confirmed by ROC curve analysis (AUC for 1-, 2- and 3-year survival: 0.705, 0.729 and 0.684, respectively). Conclusions: The radiomic model could allow for non-invasive assessment of the absolute density of TAM from MRI images in HGG patients. Of note, our RIB model is the first immunological radiomic model confirmed to have the ability to predict survival benefits from DC vaccine in gliomas, thereby providing a novel tool to inform treatment decisions and monitor patient treatment course by radiomics.

Glioma, with its heterogeneous microenvironments and genetic subtypes, presents substantial challenges for treatment prediction and development. We integrated 3D bioprinting and multi-algorithm machine learning as a novel approach to enhance the assessment and understanding of glioma treatment responses and microenvironment characteristics. The bioprinted patient-derived glioma tissues successfully recapitulated molecular properties and drug responses of native tumors. We then developed GlioML, a machine learning workflow incorporating nine distinct algorithms and a weighted ensemble model that generated robust gene expression-based predictors, each reflecting the diverse action mechanisms of various compounds and drugs. The ensemble model superseded the performance of all individual algorithms across diverse in vitro systems, including sphere cultures, complex 3D bioprinted multicellular models, and 3D patient-derived tissues. By integrating bioprinting, the evaluative scope of the treatment expanded to T cell-related therapy and anti-angiogenesis targeted therapy. We identified promising compounds and drugs for glioma treatment and revealed distinct immunosuppressive or angiogenic myeloid-infiltrated tumor microenvironments. These insights pave the way for enhanced therapeutic development for glioma and potentially for other cancers, highlighting the broad application potential of this integrative and translational approach.

Background Characterizing expression profiles of different immune checkpoint molecules are promising for personalized checkpoint inhibitory immunotherapy. Gliomas have been shown as potential targets for immune checkpoint inhibitors recently. Our study was performed to determine coexpression levels of two major B7 immune regulatory molecules programmed death ligand 1 (PD-L1) and B7-H4, both of which have been demonstrated to inhibit antitumor host immunity in gliomas. Methods We assessed tumor tissues from stage II–IV primary gliomas (n=505) by immunohistochemistry (IHC) for protein levels of both PD-L1 and B7-H4. Gene coexpression analysis assessing clusters based on extent of PD-L1/B7-H4 classifier genes expression were investigated in two transcriptome datasets (The Cancer Genome Atlas and Chinese Glioma Genome Atlas). In addition, levels of immune cell infiltrates were estimated with IHC and RNA-seq data for assessing the tumor immune microenvironment of PD-L1/B7-H4 subgroups. Results High expression of PD-L1 and B7-H4 in gliomas was 23% and 20%, respectively, whereas coexpression of two proteins at high levels was limited to 2% of the cases. Comparable results were seen in RNA-seq datasets where PD-L1 mRNA expression levels negatively correlated with that of B7-H4. Gene coexpression modules clustered within each grade of gliomas demonstrated lack of double-high modules (cluster with high expression of both PD-L1 and B7-H4 classifier genes). B7-H4 mRNA expression levels showed negative correlation with extent of immune cell infiltration and High-B7-H4 module gliomas (high B7-H4 but low PD-L1 classifier genes expression) had less tumor-infiltrating lymphocytes (TILs) and tumor-associated macrophages (TAMs). IHC assessment also showed few TILs and TAMs in High-B7-H4 subgroup gliomas. Conclusions The majority of gliomas express PD-L1 or B7-H4, however, coexpression of both at high levels is minimal. The high-B7-H4 patients could be considered as ‘super-cold’ gliomas with significantly deficient in TILs, suggesting that B7-H4 might inhibit T-cell trafficking into the central nervous system. This study demonstrated that PD-L1 and B7-H4 may serve as mutually compensatory immune checkpoint molecules in gliomas for immune targeted or active-specific immunotherapy. The distinct B7-H4 pathways modulating T-cell function and immune evasion in glioma patients deserved to be further explored in the future during immunotherapy.