Glioblastoma (GBM; WHO grade IV astrocytoma) is the most common malignant and heterogeneity primary brain tumor. Recently, advanced MR neuroimaging has shown its promising values in quantifying tumor heterogeneity from the perspectives of blood-brain barrier disruption, tumor angiogenesis, cellularity, and metabolism. Radiogenomic studies reveal that the imaging phenotypes can be correlated to the underlying genomic and molecular pathways in heterogeneous GBM. Accordingly, combined microarray data with MR neuroimaging of GBM may develop a new method that leverages and integrates large datasets to provide an approach for predictive and prognostic biomarkers, and therefore embodies the precision medicine in personalized therapies. In this study, we aim to establish a locus specific radiogenomic map based on advance MR image and RNA microarray. We focus on the correlation between ADC images and several molecular mechanisms, such as cell proliferation (epidermal growth factor receptor, EGFR), mitosis (Cyclin-dependent kinases family, CDK) and immune response (C-C chemokine receptors family, CCR)¹.

Patients and Samples: To study the tumor heterogeneity, 9 locus-specific primary GBM tissue samples from four different patients for enhancing (A), infiltrative (B), and necrotic tumor (C) portions were extracted by MR-guided stereotaxic surgery at Taipei Medical University Hospital. Patients were all conducted in compliance with IRB approvals and informed consents were obtained.

RNA extract and Microarray assay: RNA from stereotactic surgical specimens were isolated by mirVana miRNA kit, and the quality of the extracted RNA were evaluated with the RNA 6000 Nano LabChip on the Agilent 2100 Bioanalyzer. Five hundred nanograms of total RNA were amplified and labeled by Agilent Quick Amp Labeling Kit and analysis by One-color Agilent 60-mer Whole Human Genome Array Kit. The signal intensities were corrected for the background intensities with the normexp method, followed by the quantile normalization across arrays with the limma software². The reported microarray data were deposited in the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo).

Conventional MR image: Pre- and post-contrast T1-weighted, T2-weighted and fast low-angle shot MR imaging were obtained using a 1.5T MR unit (HDx, GE).

Diffusion-weighted image: Diffusion-weighted imaging (DWI) was acquired for each patient with the following parameters: FOV=240×240 mm², matrix=256×256 pixels, 23 contiguous slices, slice thickness/gap=5.0/1.0 mm, TR/TE=7000/82.2 ms, number of excitation=2, and a maximal b of 1,000 s/mm² . The apparent diffusion coefficient (ADC) map was then calculated from DWI. ROI were selected manually and image ratio was analyzed by OLEA SPHERE™.

Gene Ontology (GO): GO analyses were executed using The Database for Annotation, Visualization and Integrated Discovery (DAVID)³. For each GO term⁴, the p-value of function clustering and the p-value following multiple detecting correction, such as Benjamini correction or false discovery rate (FDR) correction, were calculated in DAVID.

Experimental flow chart of this study was showed as figure 1. In figure 2, locus specific MR guide biopsies were showed. Using correlation coefficient analysis, overexpression of T cell activation genes such as CD4 and CD8a were found in the low-ADC site, namely, the enhancing tumor regions (Figure 3). Moreover, the regional ADC also exhibited negative correlations with cell proliferation and mitosis-related genes (Figure 3). We found that EGFR and CCR5, which plays role in cell proliferation and immune response, were be down-regulated in high ADC region. Mapping of MR ADC image with GO_biological process was defined as figure 4.A previous study by Diehn et al. showed discrepant gene expression within the tumor of different spatial loci by the integrated functional genomic datasets and MR imaging⁵. However, whether physiological MR imaging can predict genetic expression patterns within a heterogeneous tumor remains unclear. Here, our preliminary results show that ADC cellularity map may provide a new insight into tumor progression-related mechanism, such as down-regulation of T cell immune response and over-expression of mitosis and cancer cell proliferation. A further study on the immune-mediated mechanism may improve our understanding of the physiological MR imaging and its relevance to genetic-modulated tumor behavior. Our results imply that ADC ratio difference within a heterogeneous GBM can be controlled by various molecular mechanisms during tumor progression. This may suggest that MR radiogenomic map may improve our understanding of the treatment strategy design.