Novel, albeit limited, insights into the molecular basis of human disease have resulted from the use of high-resolution, microarray-based platforms to comprehensively study complex diseases at multiple levels of the genetic program. For example, in central nervous system (CNS) tumors, array-based comparative genomic hybridization (aCGH) and DNA sequencing have led to the identification of tumor-associated chromosomal gains and losses and the corresponding specific gene mutations within these altered loci [[1–3]]. Gene expression-based studies have also profiled thousands of individual transcriptional units to uncover novel genes whose expression is de-regulated, concordant with other clinical features of these tumors, such as histological grade and patient survival [[4–6]]. The combined use of both technologies has identified individual genetic biomarkers associated with CNS tumorigenesis and tumor behavior, although few have yet resulted in specific strategies for targeted drug therapy .
The accumulation of 'genome-wide' data sets from CNS tumors is demonstrating that the processes of tumorigenesis and tumor progression likely involve the coordinated de-regulation of entire molecular networks, rather than single genes. Therefore, mutation, altered copy number, and abnormal expression of single genes observed to vary across individual tumors may be better viewed collectively to identify underlying commonalities at the level of molecular programs, based on known protein-protein interactions, canonical cell signaling pathways, and in silico transcriptional regulatory control predictions. As opposed to identifying single gene/protein targets for anti-cancer drug design, these networks establish broader, multi-target pathways for therapeutic intervention. The combination of higher-order DNA- and RNA-based microarray meta-analyses offers the potential to discover these therapeutically-relevant networks .
Pilocytic Astrocytomas (PAs) are World Health Organization (WHO) grade I glial CNS tumors, accounting for one-fifth of all central nervous system tumors. However, in young children and adolescents, PAs are the most common brain tumor. Compared to other high-grade CNS gliomas, such as Glioblastoma Multiforme (GBM), PAs are characterized by low cell proliferative indices, a biphasic histologic appearances, and microglial infiltration. PAs also lack the genetic alterations observed in high-grade glioma, and in general exhibit few molecular alterations at the DNA level. Previous studies have demonstrated occasional copy number changes involving chromosomes 6, 7, 8, and 11 [9, 10]. In a recent study , we identified genome copy number alterations occurring in cerebellar PA using aCGH on patient-matched tumor and normal samples. However, as in other reported studies, only a limited number of consistent genetic alterations were identified.
Current treatments for these pediatric brain tumors include surgical resection, radiation, and chemotherapy. However, in young children, these treatments are associated with secondary damage to the developing brain, and result in long term neuropsychological and neuroendocrine dysfunction. In addition, since children with PA have good overall survival, it is imperative to identify the key genetic and growth control pathways de-regulated in the tumor and not the normal brain for future therapeutic drug design. Given this clinical imperative and the limited success in identifying therapeutic targets by conventional genomic analyses, we employed a higher-order, network-based analytical approach to discover subtle alterations in genetic pathways operative specifically in PA.
In the present study, we first identified PA-specific gene expression signatures in a meta-analysis of 74 PA microarray datasets. We then utilized the identity of these genes to construct a PA-specific gene network based on transcription factor gene regulatory interactions. Moreover, we identified relationships that are both common and specific to PAs compared to other cell lineage-related glial tumor types by creating similar gene regulatory networks from GBM and oligodendroglioma microarray datasets. Our approach has identified a unique genetic network in PA which may account for the novel phenotype of this brain tumor, and also demonstrates the general utility of using network-based approaches to analyze complex molecular profiling data from other tumor types.