Setting the Stage for Personalized Treatment of Glioma

Dr. Li  and Dr. Fine analyzed genomic profiles from patients diagnosed with malignant glial cells  using computer-generated groupings to subclassify the brain tumors into two major groups: O type tumors, which were predominantly oligodendrogliomas and astrocytomas and G type tumors, which  were mostly glioblastomas. They further divided the G tumors into two additional levels of subclassification.

Dr. Li and Dr. Fine analyzed genomic profiles from patients diagnosed with malignant glial cells using computer-generated groupings to subclassify the brain tumors into two major groups: O type tumors, which were predominantly oligodendrogliomas and astrocytomas and G type tumors, which were mostly glioblastomas. They further divided the G tumors into two additional levels of subclassification.

Gliomas, the most common type of primary brain tumors in adults, arise from different types of glial cells, which support and protect the neurons of the central nervous system. How a patient’s glioma is treated depends in part on the type of glial cell from which the tumor developed. Classification of gliomas has traditionally been done by microscopic analysis of tumor sections. This process is subjective and prone to inconsistencies, which may explain in part the wide-ranging and often suboptimal responses of gliomas to treatment.

 

Several efforts have been made to develop a molecular-based scheme for classification of gliomas, but the findings have been limited by several factors. Past studies analyzed expression of only a subset of genes and used a limited number of glioma types and grades. Furthermore, these classification schemes continued to give weight to nonmolecular factors, such as histological classification or clinical features.

Aiguo Li, Ph.D., a researcher in the lab of CCR Neuro-Oncology Branch Chief Howard Fine, M.D., collaborated with other NCI and NIH scientists to develop a more robust classification scheme for glioma. The results were published in a recent issue of Cancer Research.

Li and colleagues decided to create a classification scheme based solely on molecular data. They used a "training set" of 159 gliomas of all types and grades to generate their initial model. Gene chips were used to measure expression of 47,000 gene products from across the entire human genome. The data were then fed into a computer, which mathematically classified tumors into two main types—O and G. Another round of analysis revealed that the tumors could be further organized into two O (OA and OB) and four G (GA1, GA2, GB1, GB2) subtypes. Similar results were obtained when this process was repeated with 189 additional gliomas, illustrating the reliability of the classification scheme.

The computer-generated scheme grouped tumors with similar histological and clinical features. O type tumors were predominantly oligodendrogliomas and astrocytomas while the majority of G type tumors were glioblastomas. O type tumors tended to be of lower grade than G type tumors, and patients with O type tumors were on average younger and survived longer than their G type tumor-bearing counterparts.

The researchers then used another computer program to identify smaller subsets of genes—called classifiers—capable of assigning an unknown glioma patient to one of the six subtypes. These classifiers were used to stratify 341 gliomas that had been analyzed as part of previously published studies using different classification schemes. In many cases, the new molecular-based scheme was able to refine and extend the results of the previous classification efforts, resulting in stronger associations with clinical outcomes such as survival.

The extensive molecular characterization of gliomas achieved by this study will likely have implications for both research on and clinical treatment of glioma. Biological insights gained from the gene expression profiles may help elucidate mechanisms that drive the different types of glioma, possibly leading to the identification of new therapeutic targets. If incorporated into clinical trials, the classifiers may also prove valuable for clinical diagnosis and help link individual patients with therapies most likely to benefit them.

Summary Posted: Sun, 03/01/2009

Reference

Cancer Res. 2009 Mar 1;69(5):2091-9 PubMed Link