Skip to main content
Premium Trial:

Request an Annual Quote

Single-cell Analysis Underscores Heterogeneity within Brain Tumors

NEW YORK (GenomeWeb) — Cells that make up brain tumors are more heterogeneous than previously thought, according to a new single-cell transcriptomic analysis appearing in Science today.

Researchers from the Broad Institute and Massachusetts General Hospital performed single-cell RNA-seq on some 430 cells isolated from five different primary glioblastoma tumors. From this, they found that the tumors harbor cells belonging to a number of tumor subtypes and that the distribution of these subtypes varies from tumor to tumor.

"To focus on the aspects of heterogeneity that we thought could be clinically relevant, we looked primarily for distinct cell states and found multiple sub-populations in each tumor," co-senior author Aviv Regev, a core member at the Broad and an associate professor at the Massachusetts Institute of Technology, said in a statement. "Clinically, what this means is that we might need to treat each tumor based on the complement of cellular subtypes it contains — not just the most prevalent one."

To examine cellular subtypes contained within a single tumor, Regev and her colleagues isolated individual cells from five resected glioblastomas. For each, they generated full-length transcriptomes using SMART-seq. At the same time, they generated bulk RNA-seq profiles from the tumor samples as well as from normal brain tissue.

Cells from the same tumors had gene expression patterns that were more similar to each other than to cells from other tumors, though the researchers noted a range of expression patterns in cells from the same tumors. Using multi-dimensional scaling of global transcription, the researchers also noted that cells grouped together based on the tumor they came from, though some cells overlapped with the transcriptional space of tumors from which they did not originate.

The researchers also uncovered cell-to-cell variability in the expression patterns of signaling molecules, including receptor tyrosine kinases. Mosaic expression of receptor tyrosine kinases, a common therapeutic target, throughout a tumor could affect treatment, the researchers noted.

Other surface receptors and ligands — such as EGFR, NOTCH2, JAG1, and others — that are expressed in tumor-related pathways also exhibited mosaic expression.

Three of the tumors included in the researchers' analysis had high EGFR expression levels, and one of those tumors, dubbed MHG30, harbored the oncogenic mutant form of the gene, EGFRvIII, which is a putative immunotherapy target.

In MGH30, they found that some 7 percent of single cells expressed wild-type EGFR, 19 percent expressed EGFRvIII, and 25 percent of the cells expressed a second oncogenic variant, de4. Just a small percentage of cells co-expressed EGFR and EGFRvIII. Additionally, a number of cells lacked EGFR, but expressed other tyrosine kinase receptors.

"These findings suggest that heterogeneous expression and/or mutational status of RTKs and other signaling molecules across individual glioblastoma tumor cells may compromise therapies targeting receptor immunogenicity or RTK signaling," the researchers wrote in their paper.

Using hierarchical clustering and principal components analysis, Regev and her colleagues uncovered four meta-signatures that defined clusters of cells. These signatures, they noted, were enriched for genes involved in cell cycle, hypoxia, complement and immune response, and oligodendrocyte function, respectively.

Only a small percentage of cells from each tumor had cell cycle program activity, in contrast to in vitro models of glioblastoma where most cells scored for the cell cycle program, they reported. Genes previously linked to quiescence, such as DM5B, were found in 10 percent to 20 percent of individual cells across all tumors and were negatively correlated with the cell-cycle meta-signature.

The expression of a dozen other genes, a number of which belonged to hypoxia meta-signature, were also anti-correlated to the cell cycle meta-signature, the researchers said. By ordering cells according to their hypoxia score, the researchers revealed a gradient in each sample, possibly mirroring tumor microenvironment variations in oxygen levels, blood supply, or nutrition.

"[I]n vivo microenvironment and genes linked to quiescence may affect dormant and possibly refractory compartments in glioblastoma," they noted.

Through cell cultures of stem-like and more differentiated glioblastoma cells, Regev and her colleagues derived a stemness signature based on genes differentially expressed between those types of cells.

By applying that signature to their single-cell transcriptome profiles, the researchers found gradients of stemness throughout all five tumors, and that stemness was negatively correlated with the cell cycle meta-signature.

Based on bulk analysis, the tumors the researchers examined fell into different categories of The Cancer Genome Atlas classification scheme. For instance, MGH26 was classified as a proneural tumor and MGH30 as classical tumor.

At the single-cell level, though, they were all mixes of the different glioblastoma subtypes. Single-cell qPCR of 30 classifier genes from more than 160 additional MGH26 and MGH30 cells confirmed the presence of multiple subtypes within those tumors, though the bulk analysis reflected the dominant transcriptional program.

"Thus, although population-level data detect the dominant transcriptional program, they do not capture the true diversity of transcriptional subtypes within a tumor," the researchers said.

The different classification schemes further seem to be associated with certain transcriptional signatures. For instance, the stemness signature is strongest in proneural and classical tumor cells, but lacking in mesenchymal tumor cells.

Heterogeneity, the researchers further noted, appears to affect clinical outcomes. By binning tumors belonging to the proneural TCGA subtype based upon the level of the signal they had for other subtypes, they found that increased heterogeneity is linked to decreased survival.

"Understanding the cellular landscape can provide a blueprint for identifying new therapies that target each of the various sub-populations of cancer cells, and ultimately for tailoring such therapies to individual patient tumors," said co-senior author Bradley Bernstein from the Broad and MGH in a statement.