Membrane proteomics of human glioblastoma stem-like cells

Glioblastoma multiforme (GBM) is a poorly treated human brain cancer with few established clinically useful molecular prognostic markers and tools for detection and isolation of their stem-like cell (GSC) populations. We used three approaches to better characterize the membrane proteome of GSCs: 1) differential gene expression profiling; 2) immunoarray subclassification with validation on a clinically annotated tissue microarray; and 3) mining a human naive yeast antibody library for GSC-preferential antibodies.;We report that membrane proteomic characterization of patient-derived GSCs reveals different subclasses using neural lineage markers showing association with differential xenograft invasiveness, and also demonstrate significant correlation to survival in both the mouse model and human patients. Orthotopic implantation into immunodeficient mice demonstrated Ki-67 proliferative index independent xenograft infiltration: class I GSCs (oligodendrocyte and neural progenitor cell marker positive) established focal lesions, class II GSCs (neural progenitor cell marker positive) formed minimally invasive lesions, and class III GSCs (astrocyte progenitor cell marker positive) established highly infiltrative lesions. The oligodendrocyte progenitor cell marker, CNP also exhibited high expression in focal xenografts versus low expression in invasive xenografts. Differential CNP expression correlated with mouse model survival, and CNP immunoassay of a large GBM tissue microarray (n=115) also showed significant differential patient survival. Additionally, mining with a yeast antibody library has generated novel immunoreagents to identify and isolate GSCs from each class as well as GSC-selective antibodies across multiple patient samples.