The creation of a national, multi-institutional effort to prospectively molecularly characterize over 1000 gliomas: The Glioma Molecular Diagnostic Initiative (GMDI): We have constructed a national, publicly funded effort that we call the GMDI, which coupled with its bioinformatics counterpart (REMBRANDT), is designed to breach the gap between biological and corresponding clinical data related to primary brain tumors in order to better understand the molecular pathogenesis of gliomas and ultimately help patients receive more rationally tailored treatment specific to their tumor's biology. Clinical data on the patients is collected prospectively until the patient's death, through the NCI-sponsored brain tumor clinical trials consortia. DNA, RNA and proteins are assayed using different genomic-scale technologies to determine levels of expression, mutations and chromosomal abnormalities. Beyond the generation of the molecular data and the data population of REMBRANDT, we are also analyzing the data in order to identify new targets for therapy thereby furthering the mission of the NOB.As a first step in this data mining process, we have analyzed the results from the genomic survey in our retrospective sample of 180 gliomas.We have used the data obtained by a) Analysis of Copy Number Alterations (CNA); and b) Loss of heterozygosity (LOH).CNA analysis of these tumors yielded a large number of areas with deletions and amplifications, some of which had been previously reported, as in the case of CDKN2B and EGFR whereas many never previously described areas were detected. We have been able to identify novel regions of LOH while confirming known ones (PTEN on Chr 10, 1p/19q, etc) The high correlation between the results of the LOH and CNA analyses allows us to confirm the regions of loss. An example of how data generated from GMDI may lead to promising new molecular targets is demonstrated by our analyses of chemotherapy sensitive oligodendrogliomas with 1p/19q chromosomal deletions compared to chemotherapy resistant oligodendrogliomas with maintenance at that locus. CGH and SNP analysis allowed us to separate a cohort of 10 oligodendrogliomas with LOH and 24 without. Gene expression profiles between the two groups of tumors revealed 97 genes differentially expressed. One of the most significantly differentially expressed genes was histone deacytelase I (HDAC-1)74, with HDAC1 levels significantly lower in gliomas with 1p/19q LOH compared to those with maintenance of the locus. HDAC1 levels were much lower in chemotherapy sensitive oligodendrogliomas and much higher in other chemotherapy resistant gliomas of all types such as GBMs. This led to a series of in vitro experiments using HDAC1 siRNA and HDAC inhibitors (i.e. valproic acid) demonstrating that down-regulation of HDAC-1 is not only intrinsically toxic to glioma cell lines that normally express high levels of HDAC1, but also sensitize glioma cells to both radiation and a broad array of alkylating agents. We have used these data to convince investigators at CTEP to allow us to conduct a NABTC-sponsored, multi-institutional phase I/II trial of the HDAC inhibitor, Depsipeptide75, in patients with recurrent high-grade gliomas. Recently, through our continuous analysis of the data generated by the GMDI project we have been able to recently generate classifier gene sets using the RNA expression data, that consistently segregate (98% accuracy) the tumors into 6 nested biologically relevant groups. These groups show also segregation when applying relevant clinical parameters, substantiating our hypothesis that better biological knowledge of the tumors can lead to better treatments. These groups are consistent while refining and extending previously published schemas; gaining a greater detail in the sample assignment. These classifier sets have been recently been applied to help the diagnosis of samples on which classical pathology schemas do not yield a definitive answer, thus helping guide patient treatment. Additionally, though correlation of genomic DNA changes with mRNA expression levels in the same samples, we have been able to identify two important sets of genes: a) Genes that are at the genetic origin of Gliomas, and b) putative Tumor Suppressor Genes that, following Knudsons two-hit hypothesis, become deregulated through epigenetic mechanisms in addition to the classical deletion/mutation paradigm. Future work will continue to generate additional data from the hundreds of tumors samples that are now entering our laboratory through the GMDI and to analyze and validate the data in order to identify more potential therapeutic targets. For the next years we are specifically focusing on the following areas: 1) Generation of a classification/diagnostic schema using DNA and RNA data combined, to increase the stability of the classification system thanks to the addition of easily obtainable macromolecule (DNA); 2) Validation of up to 200 genes recognized through our CNA/LOH and DNA/RNA correlation analysis, to this end expression/targeting vectors are being produced to be transfected/transduced into appropriately primed Neural Stem Cells, to asses their in vivo activity; 3) Generation of self-assembling gene networks using the wealth of mRNA data collected to date, in order to obtain a picture of central regulation genes that are ideally suited as possible therapeutic targets.The construction of a data warehouse, with the associated bioinformatics tools, to hold and analyze GMDI and other high-throughput molecular brain tumor data: REMBRANDT is an informatics effort led by NCI's Neuro-Oncology Branch and the Center for Bioinformatics that is aimed at producing a national molecular/clinical database of several thousand primary brain tumors that is fully open and accessible to all investigators. This system will provide informatics support to molecularly characterize a large number of adult and pediatric primary brain tumors and to correlate those data with extensive retrospective and prospective clinical data. We are designing a robust bioinformatics knowledgebase framework (caIntegrator) that leverages data warehousing technology to host and integrate clinical and functional genomic data from various cancer clinical and molecular trials. REMBRANDT is designed to house two sets of valuable data.The first set of data comes from the GMDI, the second type of data is a wide array of molecular and genetic data regarding all types of primary brain tumor studies form the extramural research communities.Easy single-click query options are available wherein users can enter a HUGO gene symbol and visualize either a gene expression profile or copy number alteration status for a chromosomal region across different tumor groups; Kaplan-Meier survival plot for patients with tumors that up-regulate or down-regulate any gene/region of interest; and also survival plots that compare the behavior of any two user-defined sample sets.Advanced search options are available to specify multiple query criteria concurrently (for example, different pathways of interest, Gene Ontogeny terms of interest, etc. and a sophisticated query compound interface allows users to combine these queries from different domains and execute them to visualize data in a user-friendly report format.REMBRANDT provides cross-platform capability by allowing the users [summary truncated at 7800 characters]
Showing the most recent 10 out of 14 publications
