Glioblastoma multiforme (GBM) is an extremely aggressive form of brain cancer and remains largely incurable. New treatments are desperately needed, but the high tumor heterogeneity makes this cancer especially hard to target. Recent research has found several subtypes of GBM based on gene expression, mutations, DNA methylation, and DNA copy number. However identification of the subtypes has not revealed subtype-specific treatment strategies. In this study, Dr. Patrick Paddison (Human Biology Division) and collaborators at Mount Sinai Medical School propose that the sensitivity of GBM tumor cells to inhibition of BUB1B, a molecule involved in the checkpoint before cell division, may be an effective predictive marker of tumor aggressiveness and responsiveness to specific treatments.
BUB1B encodes Bub1-like pseudo-kinase, BubR1 that is implicated in mitotic checkpoint function and chromosome segregation fidelity. The mitotic checkpoint delays anaphase until chromosomes are properly attached to the mitotic spindle. A previous study from the Paddison lab found BUB1B to be the top hit from an shRNA genetic screen knocking down each kinase in the genome and testing subsequent viability in tumor-initiating GBM stem cells (GSCs) for genes required for GSC expansion. Cells that are sensitive to BUB1B inhibition (BUB1BS), have shorter average distances between sister kinetochores during mitosis when microtubule attachments have formed than cells that are resistant to BUB1B-inhibition (BUB1BR). Abnormalities in chromosome segregation machinery are often a sign of genetic instability associated with cancer. Thus, this measurement can serve as a reliable predictor of BUB1B sensitivity. However, this method is too laborious and time consuming to be useful for characterizing tumor samples.
In this study, the authors found patterns of gene expression that are associated with BUB1BS or BUB1BR and create a method of predicting BUB1BR/S status. First, they classified 18 GSCs from 18 patients for BUB1BR/S status by sensitivity to BUB1B silencing and/or measurement of inter-kinetochore distance at metaphase. Then, they used a linear model to fit the variation in mRNA expression level for each gene in terms of BUB1BR/S. They identified 838 genes whose expression is associated with BUB1BR/S. Pathway analysis showed that these genes are enriched for cell cycle related pathways particularly the mitotic cell cycle. The resulting framework examines expression levels of 838 genes to classify a tumor sample as sensitive or resistant.
When the authors applied the classifying method to GBM patient tumor data, they found that the molecular signal was different between the normal cells included in the tumor and the malignant cells. They collected 9 pairs of GSCs and tumor tissue from which they were derived and measured gene expression levels. When taken together, the predicted BUB1B status was not consistent. However, when the authors “decomposed” the tumor tissue expression into expression of the major cell types in the brain, they were able to detect the fraction of GBM tumor cells present in the tumor sample. The predicted proportion of GBM tumor cells in the sample was highly associated with the predicted BUB1BS/R status. This was also consistent with histological results. Based on the predicted fraction and molecular profile of each cell type, the authors could determine the molecular signature of the GBM tumor cells.
The authors applied the “deconvolution” procedure, as the authors termed the partitioning of GBM molecular signatures from other cell types, to three GBM data sets including the Cancer Genome Atlas (TCGA). They found that BUB1BS/R status was significantly associated with survival rate; GBM patients with BUB1BS tumors have worse prognoses. In 58 GBM xenograft mouse models of GBM patients, BUB1BS/R status corresponded to the invasiveness of the xenograft tumor.
Currently there are no therapeutic strategies for inhibiting BUB1B activity. Therefore, the authors aimed to identify drugs that would impair the viability of BUB1BS cells. The authors used the Connectivity Map (CMAP) database to do this. CMAP encompasses 1.5 million gene expression profiles in multiple cell types before and after treatment with around 5,000 small-molecule compounds and 3,000 genetic reagents. They analyzed this data for drugs that inhibit genes within the network specific to BUB1BS cells. They identified several potential drugs including MS-275, Etoposide, Camptothecin, Irinotecan, Thioridazine, and Azacitidine.
To validate these candidate drugs, the authors analyzed data sets with detailed characterization of cell lines including the Cancer Cell Line Encyclopedia (CCLE) and the Compositae Genome Project (CGP), with gene expression and found 5 glioma cell lines that classified into BUB1BR subtype and 4 that classified into BUB1Bs subtype. Based on drug sensitivity measurements included in the data sets, the authors tested whether the two groups showed distinct drug sensitivity. Out of 13 potential drugs, data for four of the drugs were available in the data sets. Three out of the four drugs were significantly associated with BUB1BS status: Etoposide (targeting topoisomerase II), MS-275 (targeting histone deacetylase), and Irinotecan (targeting topoisomerase I). Etoposide and Irinotecan were tested in vitro in 9 GSC patient samples, 4 BUB1BS 4 BUB1BR and one ranked in between. The BUB1BS cells were significantly more sensitive to drug treatment with both drugs.
Sensitivity to BUB1B-inhibition is a new way to classify GBM tumors. BUB1BS tumors have worse prognoses but respond better to Etoposide and Irinotecan. The tumor heterogeneity of GBM tumors poses challenges for treatment. For instance, patients with a mixture of tumor cells from BUB1BS and BUB1BR subtypes would be a particular challenge to treat. Also, some samples were not clearly classified into one of the two subtypes. Effective precision medicine for GBM will require deeper understanding of these classifications and identification of more subtypes.
This work was supported by the National Institutes of Health.
Lee E, Pain M, Wang H, Herman JA, Toledo CM, DeLuca JG, Yong RL, Paddison P, Zhu J. 2017. Sensitivity to BUB1B inhibition defines an alternative classification of glioblastoma. Cancer Res, 77(20).