Functional phenotyping of stem cells, including cancer stem cells, has relied on the use of robust markers that indicate the phenotypic state of the cell. These markers are typically interrogated using antibodies targeting an extracellular protein product. Accumulating evidence suggests that expression of the putative stem cell markers CD133 and CD44 specify molecular subtype in Glioblastoma multiforme (GBM). We have conducted a targeted gene coexpression analysis of the Cancer Genome Atlas (TCGA) GBM dataset to compare several genes purported to be markers of the Glioblastoma stem-progenitor cell (GSPC) phenotype.
Pearson correlation was used to determine the suite of genes that are coexpressed with candidate stem cell markers. The positively coexpressed genes were used to build a gene signature that classifies patients into a CD133 coexpression module score (CD133-M) or CD44-M subtype of GBM. The CD133-M subtype was enriched for the Proneural (PN) subtype of GBM compared to CD44-M tumors which was significantly enriched for the Mesenchymal (MES) subtype. Gene set enrichment identified DNA replication and cell cycle progression as enriched in the CD133-M and invasion and migration as enriched in the CD44-M. Functional experiments confirmed cellular growth as increased in cells expressing CD133 and invasion as increased in cells expressing CD44.
Similar to the 4 major molecular subtypes of GBM there was no long-term survival difference between CD44-M and CD133-M patients, however CD44-M patients obtained greater benefit from temozolomide and CD133-M patients benefited more from radiotherapy. The use of a targeted coexpression approach to predict functional properties of surface marker expressing cells is novel and in the context of GBM supports the accumulating evidence of CD133 and CD44 marker expression indicating molecular subtype. The approach described in this study may assist in the choice of targets for future characterization of cancer stem cells.