mor biology. Advancement may depend upon new model systems in which the multi-faceted microenvironment can be deconvoluted in order to assess the contributions that specific subsets of cells, matrix, and three-dimensional structures make to tumor tissue biology. Prior efforts to identify the mediators and targets of tumor promoting cell-cell interactions have frequently relied on conditioned media experiments, however one drawback to this approach is that changes induced by physical cell-cell interactions or reciprocally induced interactions cannot be detected. Ideally, model systems like these should support the discovery of all of the genes and pathways that mediate the important reciprocal interactions that occur between tumor cells and their microenvironment. We were interested in understanding the complete set of interactions between GBM and endothelial cells as these are purported to be essential for of PDE7B expression on clonogenic PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19674226 activity in an extreme limiting dilution assay. Briefly, replicate cultures were established with cell numbers that ranged from 130 cells/well in sphere formation media and allowed to grow for 3 weeks. Under these conditions, PDE7B in the GBM Perivascular Niche tumorigenesis, tumor progression, and recurrence after standard treatment. While the deconvolution of get Peretinoin transcriptional data has been previously described, here we report a new algorithm PubMed ID:http://www.ncbi.nlm.nih.gov/pubmed/19673983 tailored for mixed expression signals from 
two cell types and its application to the novel setting of GBM tumor cells interacting with ECs. Direct global expression profiling of co-cultured GBM and endothelial cells, enabled by our computational deconvolution algorithm, revealed a number of pathways that were induced or suppressed by the interaction of these cells. Among them was upregulation of the cAMP-specific phosphodiesterase, PDE7B, in GBM cells. This computational approach has its own limitations. First, the detected gene expression changes are a mixture of the signal from both of the constituent cell types. Therefore, for each gene, further studies are required to determine in which of the cell types the change occurred, and whether this change was a function of PDE7B in the GBM Perivascular Niche secreted factors or cell-cell contact. Furthermore, reciprocal gene induction and suppression could theoretically result in no apparent change in signal and a diminution of the model’s sensitivity. We suspect that processes like these contribute to the marginally weaker significance we detected for changes in gene expression as compared to many other global expression studies. However, we were able to identify 45 genes whose expression was significantly altered. After multiple hypothesis correction, validation by qRTPCR, and in vivo experiments we did not uncover any false positives, reassuring us that the analysis provided a high quality list of changes in gene expression as a consequence of functional interactions between GBM and endothelial cells. Currently, the system serves as a proof of principle and evaluates a single set of interactions between two cell types from the tumor microenvironment. Further work will be necessary to systematically add other tumor microenvironmental cells types and determine the additional effects on gene expression. Of the genes discovered by our assay, PDE7B was chosen for further studies. This focus was supported by our previous work demonstrating that PDE4A1 and cAMP suppression can contribute to tumorigenesis and drive intra