Immunotherapy Drives Mesenchymal Transition and Tumor Microenvironment Immune Response in Glioblastoma

Introduction

Glioblastoma is a highly malignant brain tumor with no curative treatment currently available. Although immunotherapy has shown efficacy in other cancer types, the response in glioblastoma patients is limited. Investigating the effects of immunotherapy on tumor cells and the tumor microenvironment at single-cell resolution can help unravel potential mechanisms of treatment resistance and design new therapeutic strategies.

Paper Source

This study was conducted by Joachim Weischenfeldt and colleagues from the Copenhagen University Hospital, the Center for Genomic Medicine, the Danish Comprehensive Cancer Center – Brain Tumor Unit, and other institutions in Denmark. The research findings were published as a preprint in the journal Neuro-Oncology.

Methods and Results

Study Subjects and Workflow

The study subjects were 8 glioblastoma patients participating in a clinical trial of the PD-1 inhibitor nivolumab. The researchers collected tumor tissue samples from these patients at initial surgery and one week after receiving immunotherapy before recurrence surgery.

a) For each time point, the researchers performed single-cell RNA sequencing (scRNA-seq) on the tumor samples. Cell type and state identification included using public marker genes, copy number analysis with the copyKAT software, and inference of cell states from single-cell data.

b) For the transcriptional state analysis of tumor cells, the researchers used three cell state signatures: the three subtypes of glioblastoma, four single-cell resolved cell states, and pan-genomic state signatures. The Stream software was used for pseudotime analysis to trace cell developmental trajectories.

c) For the tumor microenvironment cells, the researchers integrated their single-cell data with other public single-cell datasets to classify myeloid and T cell subtypes. The CellChat software was used to evaluate cell-cell interactions.

d) Based on single-cell discoveries, the team identified a “potential immune” signature in a large external dataset of 298 glioblastoma samples and associated it with clinical outcomes. Finally, a random forest model was built to predict this signature using machine learning.

Key Findings

1. After immunotherapy, tumor cells in some patients underwent a transition toward a mesenchymal-like cell state, exhibiting a more invasive phenotype. Pseudotime analysis suggested that these mesenchymal-like cells may originate from other types of tumor stem cells and became enriched after immunotherapy.

2. After immunotherapy, some patients showed an enrichment of tumor-associated macrophages (TAMs) and proliferative T cells in the tumor microenvironment, accompanied by upregulation of T cell exhaustion markers.

3. In 298 glioblastoma samples, the researchers identified an 18% “potential immune” subgroup characterized by mesenchymal tumor cell states, activated immune pathways, TAM and proliferative/exhausted T cell infiltration. Patients in this subgroup had shorter overall survival when treated with immune checkpoint inhibitors.

4. The researchers built a random forest classifier that can identify this “potential immune” subgroup from tumor transcriptomic data, which could aid in clinical patient stratification and treatment decision-making.

Research Value

This study comprehensively revealed the cellular dynamics of glioblastoma after immunotherapy, finding that a subset of patients exhibited mesenchymal tumor cell transition, TAM and T cell activation/exhaustion, suggesting potential resistance to immunotherapy in this subgroup. The established “potential immune” molecular signature can be used for patient stratification and to guide the selection of new adjuvant therapies targeting mesenchymal cells for potential responders. This work contributes to understanding the mechanisms of immunotherapy in glioblastoma and provides new molecular targets for improving treatment efficacy.