The Cancer Genome Atlas (TCGA) project has generated gene expression data that divides glioblastoma (GBM) into four transcriptional classes: proneural, neural, classical, and mesenchymal. Because transcriptional class is only partially explained by underlying genomic alterations, we hypothesize that the tumor microenvironment may also have an impact. In this study, we focused on necrosis and angiogenesis because their presence is both prognostically and biologically significant. These features were quantified in digitized histological images of TCGA GBM frozen section slides that were immediately adjacent to samples used for molecular analysis. Correlating these features with transcriptional data, we found that the mesenchymal transcriptional class was significantly enriched with GBM samples that contained a high degree of necrosis. Furthermore, among 2422 genes that correlated with the degree of necrosis in GBMs, transcription factors known to drive the mesenchymal expression class were most closely related, including C/EBP-β, C/EBP-δ, STAT3, FOSL2, bHLHE40, and RUNX1. Non-mesenchymal GBMs in the TCGA data set were found to become more transcriptionally similar to the mesenchymal class with increasing levels of necrosis. In addition, high expression levels of the master mesenchymal factors C/EBP-β, C/EBP-δ, and STAT3 were associated with a poor prognosis. Strong, specific expression of C/EBP-β and C/EBP-δ by hypoxic, perinecrotic cells in GBM likely account for their tight association with necrosis and may be related to their poor prognosis.

Glioblastoma (GBM) is an aggressive primary brain tumor with an average survival of approximately 1 year. A recently recognized subtype, glioblastoma with oligodendroglioma component (GBM-O), was designated by the World Health Organization (WHO) in 2007. We investigated GBM-Os for their clinical and molecular characteristics as compared to other forms of GBM. Tissue samples were used to determine EGFR, PTEN, and 1p and 19q status by fluorescence in situ hybridization (FISH); p53 and mutant IDH1 protein expression by immunohistochemistry (IHC); and MGMT promoter status by methylation-specific polymerase chain reaction (PCR). GBM-Os accounted for 11.9% of all GBMs. GBM-Os arose in younger patients compared to other forms of GBMs (50.7 years vs. 58.7 years, respectively), were more frequently secondary neoplasms, had a higher frequency of IDH1 mutations and had a lower frequency of PTEN deletions. Survival was longer in patients with GBM-Os compared to those with other GBMs, with median survivals of 16.2 and 8.1 months, respectively. Most of the survival advantage for GBM-O appeared to be associated with a younger age at presentation. Among patients with GBM-O, younger age at presentation and 1p deletion were most significant in conferring prolonged survival. Thus, GBM-O represents a subset of GBMs with distinctive morphologic, clinical and molecular characteristics.

Multi-modal, multi-scale data synthesis is becoming increasingly critical for successful translational biomedical research. In this paper, we present a large-scale investigative initiative on glioblastoma, a high-grade brain tumor, with complementary data types using in silico approaches. We integrate and analyze data from The Cancer Genome Atlas Project on glioblastoma that includes novel nuclear phenotypic data derived from microscopic slides, genotypic signatures described by transcriptional class and genetic alterations, and clinical outcomes defined by response to therapy and patient survival. Our preliminary results demonstrate numerous clinically and biologically significant correlations across multiple data types, revealing the power of in silico multi-modal data integration for cancer research.

In this paper, we present a comprehensive framework to support classification of nuclei in digital microscopy images of diffuse gliomas. This system integrates multiple modules designed for convenient human annotations, standard-based data management, efficient data query and analysis. In our study, 2770 nuclei of six types are annotated by neuropathologists from 29 whole-slide images of glioma biopsies. After machine-based nuclei segmentation for whole-slide images, a set of features describing nuclear shape, texture and cytoplasmic staining is calculated to describe each nucleus. These features along with nuclear boundaries are represented by a standardized data model and saved in the spatial relational database in our framework. Features derived from nuclei classified by neuropathologists are retrieved from the database through efficient spatial queries and used to train distinct classifiers. The best average classification accuracy is 87.43% for 100 independent five-fold cross validations. This suggests that the derived nuclear and cytoplasmic features can achieve promising classification results for six nuclear classes commonly presented in gliomas. Our framework is generic, and can be easily adapted for other related applications.

No abstract