Therapeutic resistance remains a persistent challenge for patients with malignant tumors. Here, we reveal that endothelial cells (ECs) acquire transformation into mesenchymal stem cell (MSC)–like cells in glioblastoma (GBM), driving tumor resistance to cytotoxic treatment. Transcriptome analysis by RNA sequencing (RNA-seq) revealed that ECs undergo mesenchymal transformation and stemness-like activation in GBM microenvironment. Furthermore, we identified a c-Met–mediated axis that induces β-catenin phosphorylation at Ser675 and Wnt signaling activation, inducing multidrug resistance–associated protein-1(MRP-1) expression and leading to EC stemness-like activation and chemoresistance. Last, genetic ablation of β-catenin in ECs overcome GBM tumor resistance to temozolomide (TMZ) chemotherapy in vivo. Combination of Wnt inhibition and TMZ chemotherapy eliminated tumor-associated ECs, inhibited GBM growth, and increased mouse survival. These findings identified a cell plasticity–based, microenvironment-dependent mechanism that controls tumor chemoresistance, and suggest that targeting Wnt/β-catenin–mediated EC transformation and stemness activation may overcome therapeutic resistance in GBM.

Gliomas are intrinsic brain tumors that originate from glial cells. Glioblastoma (GBM) is the most aggressive glioma type and resistant to immunotherapy, mainly due to its unique immune environment. Dimensional data analysis reveals that the intra-tumoral heterogeneity of immune cell populations in the glioma microenvironment is largely made up of cells of myeloid lineage. Conventional therapies of combined surgery, chemotherapy and radiotherapy have achieved limited improvements in the prognosis of glioma patients, as myeloid cells are prominent mediators of immune and therapeutic responses—like immunotherapy resistance—in glioma. Myeloid cells are frequently seen in the tumor microenvironment (TME), and they are polarized to promote tumorigenesis and immunosuppression. Reprogramming myeloid cells has emerged as revolutionary, new types of immunotherapies for glioma treatment. Here we detail the current advances in classifying epigenetic, metabolic, and phenotypic characteristics and functions of different populations of myeloid cells in glioma TME, including myeloid-derived suppressor cells (MDSCs), glioma-associated microglia/macrophages (GAMs), glioma-associated neutrophils (GANs), and glioma-associated dendritic cells (GADCs), as well as the mechanisms underlying promotion of tumorigenesis. The final goal of this review will be to provide new insights into novel therapeutic approaches for specific targeting of myeloid cells to improve the efficacy of current treatments in glioma patients.

Background: Electronic health records represent a large data source for outcomes research, but the majority of EHR data is unstructured (e.g. free text of clinical notes) and not conducive to computational methods. While there are currently approaches to handle unstructured data, such as manual abstraction, structured proxy variables, and model-assisted abstraction, these methods are time-consuming, not scalable, and require clinical domain expertise. This paper aims to determine whether selective prediction, which gives a model the option to abstain from generating a prediction, can improve the accuracy and efficiency of unstructured clinical data abstraction. Methods: We trained selective prediction models to identify the presence of four distinct clinical variables in free-text pathology reports: primary cancer diagnosis of glioblastoma (GBM, n = 659), resection of rectal adenocarcinoma (RRA, n = 601), and two procedures for resection of rectal adenocarcinoma: abdominoperineal resection (APR, n = 601) and low anterior resection (LAR, n = 601). Data were manually abstracted from pathology reports and used to train L1-regularized logistic regression models using term-frequency-inverse-document-frequency features. Data points that the model was unable to predict with high certainty were manually abstracted. Findings: All four selective prediction models achieved a test-set sensitivity, specificity, positive predictive value, and negative predictive value above 0.91. The use of selective prediction led to sizable gains in automation (anywhere from 57% to 95% reduction in manual abstraction of charts across the four outcomes). For our GBM classifier, the selective prediction model saw improvements to sensitivity (0.94 to 0.96), specificity (0.79 to 0.96), PPV (0.89 to 0.98), and NPV (0.88 to 0.91) when compared to a non-selective classifier. Interpretation: Selective prediction using utility-based probability thresholds can facilitate unstructured data extraction by giving "easy" charts to a model and "hard" charts to human abstractors, thus increasing efficiency while maintaining or improving accuracy.

Immunotherapy is a promising therapeutic domain for the treatment of gliomas. However, clinical trials of various immunotherapeutic modalities have not yielded significant improvements in patient survival. Preclinical models for glioma research should faithfully represent clinically observed features regarding glioma behavior, mutational load, tumor interactions with stromal cells, and immunosuppressive mechanisms. In this review, we dive into the common preclinical models used in glioma immunology, discuss their advantages and disadvantages, and highlight examples of their utilization in translational research.