Intratumoral heterogeneity is a hallmark of glioblastoma multiforme and thought to negatively affect treatment efficacy. Here, we establish libraries of glioma-initiating cell (GIC) clones from patient samples and find extensive molecular and phenotypic variability among clones, including a range of responses to radiation and drugs. This widespread variability was observed as a continuum of multitherapy resistance phenotypes linked to a proneural-mesenchymal shift in the transcriptome. Multitherapy resistance was associated with a semi-stable cell state that was characterized by an altered DNA methylation pattern at promoter regions of mesenchymal master regulators and enhancers. The gradient of cell states within the GIC compartment constitutes a distinct form of heterogeneity. Our findings may open an avenue toward the development of new therapeutic rationales designed to reverse resistant cell states.

The identity of the glioblastoma (GBM) cell of origin and its contributions to disease progression and treatment response remain largely unknown. We have analyzed how the phenotypic state of the initially transformed cell affects mouse GBM development and essential GBM cell (GC) properties. We find that GBM induced in neural stem-cell-like glial fibrillary acidic protein (GFAP)-expressing cells in the subventricular zone of adult mice shows accelerated tumor development and produces more malignant GCs (mGC1) that are less resistant to cancer drugs, compared with those originating from more differentiated nestin- (mGC2) or 2,'3'-cyclic nucleotide 3'-phosphodiesterase (mGC3)-expressing cells. Transcriptome analysis of mouse GCs identified a 196 mouse cell origin (MCO) gene signature that was used to partition 61 patient-derived GC lines. Human GC lines that clustered with the mGC1 cells were also significantly more self-renewing, tumorigenic, and sensitive to cancer drugs compared with those that clustered with mouse GCs of more differentiated origin.

No abstract

Cell-based anticancer drug screening generally utilizes rapidly proliferating tumour cells grown as monolayer cultures. Hit compounds from such screens are not necessarily effective on hypoxic and slowly proliferating cells in 3-D tumour tissue. The aim of this study was to examine the potential usefulness of 3-D cultured tumour cells for anticancer drug screening. We used colon carcinoma multicellular spheroids containing hypoxic and quiescent cells in core areas for this purpose. Three libraries (∼11 000 compounds) were screened using antiproliferative activity and/or apoptosis as end-points. Screening of monolayer and spheroid cultures was found to identify different sets of hit compounds. Spheroid screening enriched for hydrophobic compounds: median XLogP values of 4.3 and 4.4 were observed for the hits in two independent screening campaigns. Mechanistic analysis revealed that the majority of spheroid screening hits were microtubuli inhibitors. One of these inhibitors was examined in detail and found to be effective against non-dividing cells in the hypoxic centres of spheroids. Spheroid screening represents a conceptually new strategy for anticancer drug discovery. Our findings have implications for drug library design and hit selection in projects aimed to develop drugs for the treatment of solid tumours.

In conclusion, the ex vivo profile suggests that further evaluation of J1 as the alkylating agent in for example aggressive breast cancer might be of particular interest, preferentially in combination with DNA-topoisomerase II inhibitors like etoposide.