Microarray Analysis of MRI-defined Tissue Samples in Glioblastoma Reveals Differences in Regional Expression of Therapeutic Targets : Diagnostic Molecular Pathology

Journal Logo

00019606-200612000-00002ArticleDiagnostic Molecular PathologyDiagnostic Molecular Pathology© 2006 Lippincott Williams & Wilkins, Inc.15December 2006 p 195-205Microarray Analysis of MRI-defined Tissue Samples in Glioblastoma Reveals Differences in Regional Expression of Therapeutic TargetsOriginal ArticlesVan Meter, Timothy PhD*; Dumur, Catherine PhD†; Hafez, Naiel MD†; Garrett, Carleton MD†; Fillmore, Helen PhD*; Broaddus, William C. MD, PhD**Department of Neurosurgery and Harold F. Young Neurosurgical Center†Molecular Diagnostics Division, Department of Pathology, Virginia Commonwealth University, Medical College of Virginia Campus, Richmond, VA, 23298Reprints: Dr Timothy Van Meter, PhD, Box 980631, Department of Neurosurgery, West Hospital, 8th Floor, 1200 E Broad St, MCV Campus, Virginia Commonwealth University, Richmond, VA, 23298 (e-mail: [email protected]).Timothy Van Meter and Catherine Dumur contributed equally to this work.AbstractMicroarray technologies have come into prominence for the assessment of molecular diagnostic profiles in cancer tissue biopsies. To better understand the effect of sampling bias, we paired image-guided stereotactic biopsy and microarray technology to study regional intratumoral differences in tumor periphery and core regions of untreated glioblastoma. RNA was extracted from serial frozen sections using an integral histopathologic scoring approach. Gene expression analysis was performed using high-density oligonucleotide microarrays (22,283 probe sets). A consensus list of 643 genes (784 probe sets) with greater than 2-fold difference between intratumoral periphery and core samples was obtained using Microarray Suite 5.0, model-based expression indexes, and robust multiarray analysis algorithms. Results were validated using quantitative polymerase chain reaction and Western blotting analyses. Reproducible profiles emerged, in which multiple therapeutic targets significant to glioblastoma [matrix metalloproteinases, AKT1 (v-akt murine thymoma viral oncogene homolog 1), epidermal growth factor receptor, vascular endothelial growth factor] showed significant differences in regional expression that may affect treatment response. This study suggests important intratumoral regional differences in the molecular phenotype of glioblastoma.Malignant primary brain tumors are known to be heterogeneous cancers, both in terms of histologic features such as growth rate and angiogenesis, and in terms of their molecular genetic composition.1–3 Glioblastoma multiforme, the most heterogeneous and most malignant intrinsic brain tumor, frequently recurs within 1 to 3 cm from the primary tumor resection site, and at distant sites throughout the brain. Molecular mediators of the invasive spread of glioblastoma, and the relationship of gene expression patterns to specific tumor regions, are still poorly understood.Several recent reports have attempted to assess differences in gene expression between glioma cells within the tumor mass versus invasive tumor cells from the margin of brain surrounding the tumor mass.4–6 Mariani and coworkers4 examined differential gene expression in a single case of glioblastoma using RNA extracts prepared from invasive tumor cells isolated from the adjacent brain versus tumor core using laser capture microdissection. In a subsequent paper, the same group expanded this study to validate a larger number of genes using quantitative reverse transcription-polymerase chain reaction (QRT-PCR) and immunohistochemistry using a tissue microarray approach.5 Both studies used a 5700 gene cDNA array to obtain differential gene expression profiles between individual invading tumor cells and the cells of the tumor mass itself. This approach has made a laudable and lasting contribution to elucidating regional gene expression differences in subsets of glioblastoma multiforme (GBM) cells, but may underestimate the role of important regulators of invasiveness concentrated in the contiguous actively growing rim. Similar studies have been performed by Suzuki and coworkers6 with genomic microarray systems, in which DNA extracted from 30 GBM cases was studied for chromosomal changes, confirming well-documented molecular genetic changes in GBM and describing some novel changes such as deletion of the PI3 kinase subunit gene PI3CA. Global genomic studies, however, have not yet been applied specifically to the study of regional gene expression in GBM.To further elucidate the heterogeneous nature of glioblastoma, we have prospectively collected tumor tissue samples from multiple distinct regions of enhancing periphery and tumor core from untreated glioblastoma using magnetic resonance (MR) image-guided stereotactic neurosurgic techniques, a novel approach to precise tissue acquisition which holds the prospect of combining neuroimaging techniques such as MR spectroscopy with molecular biologic and biochemical profiles (Fig. 1). This study provides proof of principle experiments performed on multiple samples of 6 cases of untreated GBM, with the objective of demonstrating that equivalent high-quality data about regional differences in gene expression can be studied by relying on image-guided stereotactic biopsy and an integral histologic scoring method, without the need for laborious laser capture microdissection methods. To examine differences in genome-wide gene expression profiles between GBM periphery and tumor core, tumor samples were examined by high-density oligonucleotide microarray holding 22,283 probe sets. Further validation of gene candidates in QRT-PCR and Western blot assays suggests a role for invasion-related genes in distinct regions of the hypoxic tumor core, including several novel candidate genes which warrant further study.JOURNAL/dimp/04.03/00019606-200612000-00002/figure1-2/v/2021-02-17T195932Z/r/image-jpeg GBM sample acquisition. Representative preoperative MR images taken as intraoperative “snap-shots” of the precise location of excision using Stealth stereotactic biopsy to select specific tumor areas during prospective tissue collection from untreated glioblastoma. A, Sampling from tumor core, and (B) sampling from the enhancing tumor periphery, both taken after gadolinium contrast.MATERIALS AND METHODSTissue Acquisition, Characterization, and ProcessingTumor tissue was prospectively collected in the operating room in accordance with VCU IRB-approved protocols (VCU IRB no. 3031). Briefly, samples of the tumor were first obtained to allow full neuropathologic evaluation and diagnosis, as required for the clinical management of the patient's disease. After this, tumor study samples were obtained that would otherwise have been discarded. The site of origin of tumor samples was planned preoperatively by selecting areas of tumor on the periphery of the enhancing mass, in which a lower percentage of necrotic cells was expected, and selecting areas of tumor within the poorly enhancing central core of the tumor mass, in which a high percentage of necrosis was expected. The sites of origin were monitored and recorded during stereotactic tumor resection, using the intraoperative neuronavigation system (Medtronic StealthStation) that is used routinely for these tumor resections (Fig. 1). Overall, 6 GBM cases were included in this study. In particular, for the microarray analysis, 6 samples were taken from the first GBM patient, including samples from 3 separate enhancing tumor periphery regions and 3 separate necrotic or perinecrotic regions. Paired core and periphery samples were subsequently collected from 5 additional GBM resections, and analyzed for gene and protein expression levels. Samples obtained were routinely snap-frozen in liquid nitrogen within 5 minutes of removal from the brain and stored at −86°C until ready for sectioning and extraction.Tissue Characterization and RNA ProcessingDuring processing, tumor samples were removed from storage and placed on dry ice. Samples were embedded in Tissue-Tek OCT compound and cooled before sectioning on a cryostat (Leica, CM1850). Integral histopathologic scoring of standard features (% tumor vs. normal cells, % necrosis, and extent of endothelial proliferation) was performed by a neuropathologist on serial frozen sections at 800 μM increments during Trizol RNA/protein extraction from tissue samples. Adjacent frozen sections were placed directly in Trizol for extraction. Frozen sections (80×10 μM) were prepared such that representative serial sections were available for staining at a minimum of 3 points throughout the specimen, with the remainder of the sections placed directly in Trizol reagent (Invitrogen, Gaithersburg, MD) for extraction. Total RNA isolation was performed using Trizol reagent (Invitrogen Life Technologies, Carlsbad, CA), as reported previously,7 followed by a cleanup process with RNeasy kits (Qiagen, Inc, Valencia, CA) according to the manufacturer's protocols. The quality of total RNA sample as well as cDNA and cRNA synthesis products was assessed by running 1 μL of every sample in RNA 6000 Nano or DNA 7500 LabChips on the 2100 Bioanalyzer (Agilent, Palo Alto, CA), following the manufacturer's protocol. Furthermore, RNA integrity from high necrotic content samples was confirmed by monitoring cDNA and cRNA synthesis products following previously established quality control criteria.7Affymetrix GeneChip Standard ProtocolThe Affymetrix standard protocol has been extensively described elsewhere.34 Briefly, starting with 5 μg of total RNA from every sample we generated double-stranded cDNA using a 24-mer oligodeoxythymidylic acid primer with a T7 RNA polymerase promoter site added to the 3′ end (Superscript cDNA Synthesis System; Life Technologies, Inc, Rockville, MD). After second-strand synthesis, in vitro transcription was performed using the Enzo BioArray High Yield RNA Transcript Labeling Kit (Enzo Diagnostics, Farmingdale, NY) to produce biotin-labeled cRNA. Twenty microgram of the cRNA product was fragmented and hybridized for 18 to 20 hours into HG-133A microarrays, containing 22,283 probe sets. Each microarray was washed and stained with streptavidin-phycoerythrin and scanned at a 6 μm resolution by the Agilent G2500A Technologies Gene Array scanner (Agilent Technologies, Palo Alto, CA) according to the GeneChip Expression Analysis Technical Manual procedures (Affymetrix, Santa Clara, CA). After scanning the chips, the raw intensities for every probe were stored in electronic files (in .DAT and .CEL formats) by the Microarray Suite 5.0 software (Affymetrix, Santa Clara, CA). We have previously assessed the robustness of the Affymetrix platform by running 16 technical replicates during a precision and reproducibility study.7 In the current study, it was decided to run 1 array per biologic sample, due to the limited amount of RNA available per sample for additional validation assays, and because we have already validated that, in our laboratory, there is very little variance between technical replicates on this platform.7Microarray Data AnalysisFor every probe set, normalization, background subtraction, and expression summaries were calculated using 3 commonly used methods.35 First the MAS5 method (Affymetrix, Santa Clara, CA) was used to obtain probe set summaries. A detailed description of this algorithm has been published elsewhere.36 Numerical expression summaries were stored in electronic files (in CHP format). Second, MBEI37 were calculated, which uses a multiplicative model to account for probe affinity effects in calculating probe set expression summaries. Third, the RMA average method was used38; this method uses quantile normalization followed by a median polish to remove probe affinity effects when calculating probe set summaries. Fold changes between core and periphery samples were obtained using BRB-ArrayTools v3.1.0,9 an Excel Add-in that performs analyses of microarray data. Probe sets that systematically showed greater than 1.3-fold-change in expression levels of their corresponding gene, between the core samples and those from the periphery of the tumor, regardless of the algorithm used to calculate their expression summaries, were further considered for analysis.Statistical AnalysisThe “significance-score” algorithm (S-score) developed by Dr Li Zhang39 was used to produce a score for the comparisons of the expression summaries between core and periphery samples. The S-score produces a robust measure of expression changes by weighting oligonucleotide pairs according to their signal strength above empirically determined noise levels. The procedure produces scores centered on “0” (no change) with a standard deviation of 1. Thus, scores >2 or <−2 from a single comparison have, on average, a 95% chance of being significant hybridization changes, corresponding to a P value of P<0.05. A P value derived from S-score analysis, does not necessarily reflect that the observed changes in gene expression are biologically significant. To overcome this limitation, we performed QRT-PCR for candidate genes in a larger set of GBM tumors, corresponding to pairs of core and periphery samples.QRT-PCRProbes and primer sets for detection of MMP-1, VEGF-A, and AKT1, transcripts were designed to span intron-exon junctions using GenBank sequences (MMP-1, GB NM_002421; VEGF-A, GB NM_003376; AKT1, GB NM_005163; www.ncbi.nlm.nih.gov/entrez/nucleotide) using Primer Express, version 2.0. The probes were labeled in the 5′ end with FAM (6-carboxyfluorescein) and in the 3′ end with TAMRA (6-carboxytetramethyl-rhodamine). Also, inventoried assays (Applied Biosystems, Foster City, CA) were used to measure the gene expression of MMP-19, VEGF, and PIK3R1 in a larger set of GBM tumors, corresponding to pairs of core and periphery samples. For all the samples, cyclophilin A (PPIA) from the Predeveloped TaqMan Assay Reagents (Applied Biosystems, Foster City, CA) was used as endogenous control. The experiments were performed in the ABI Prism 7900 Sequence Detection System (Applied Biosystems, Foster City, CA) using the TaqMan One Step PCR Master Mix Reagents Kit. All the samples were tested in triplicate. The cycling conditions were 48°C for 30 minutes; 95°C for 10 minutes; and 40 cycles of 95°C for 15 seconds and 60°C for 1 minutes. The 2−ΔΔCt method was used to calculate fold changes in the expression levels of the genes of interest.15 The reactions and the synthesis of the probes and primers were performed in the VCU Nucleic Acid Research Facilities.Western Blotting ProcedureBriefly, samples were prepared by grinding 40 to 70 mg frozen tissue specimens with a mortar and pestle on dry ice. Proteins were extracted from pulverized frozen tissue by adding 500 μL TPER (Tissue Protein Extraction Reagent, Pierce Biochemical Company, Rockville, IL), vortexing briefly, passing 6 times through a 21 gauge needle, centrifuging at 14,000×g in an Eppendorf 1810C microcentrifuge. Supernatants were stored at −86°C in aliquots before use. Total protein was analyzed in 10 μg samples via Novex gel electrophoresis system (Invitrogen, Gaithersburg, MD). NuPAGE 4% to 12% Bis-Tris gels were used to separate proteins by electrophoresis, performed at 200 V for 50 minutes. Proteins were transferred to 0.45 μM nitrocellulose, at 35 V for 2.5 hours at room temperature (RT) in 2-(N-Morpholino) ethanesulfonic acid transfer buffer (20% methanol; Invitrogen). After transfer, blots were briefly transferred to Tris-buffered saline-Tween-20 buffer (TBST: 10 mM Tris, 150 mM NaCl, and 0.5% Tween-20, pH 8.0), blocked for 1 hour at RT in TBST, supplemented with 5% nonfat dry milk, and the primary antibody applied for a further 1 hour at RT in fresh blocking buffer. After four 10-minute washes with TBST, species-specific horse-radish peroxidase-conjugated secondary antiserum (Rockland, Gilbertsville, PA) was applied for 1.5 hours at RT with agitation, washed for 1 hour with frequent buffer changes, treated with enhanced chemi-luminescence (ECL) reagents (Amersham Lifesciences, Piscataway, NJ) for 1 minute at RT, and subjected to autoradiographic film (Marsh Bio Products, Inc, Rochester, NY). Film is developed on an automatic film processor and densitometry performed, analyzed with Imagequant software (Amersham Biosciences, Piscataway, NJ).RESULTSHistologic Staining and Analysis of Frozen Sections Prepared From GBM SpecimensCryostat sectioning was used to analyze each frozen specimen before RNA extraction. Hematoxylin and eosin staining was performed (Fig. 2) during the extraction process using a coordinated “integral” histologic scoring technique. Evaluation of tissues from tumor core regions revealed highly necrotic tissue, varying between 30% and 80% (Table 1). Consistent with glioblastoma, prominent mitotic activity among tumor cells and vascular proliferation were evident in samples from both the tumor core regions and from the enhancing periphery. Tumor cells comprised the majority of the cellular component of each section chosen for the study, with a threshold of 90% tumor set for each specimen.JOURNAL/dimp/04.03/00019606-200612000-00002/figure2-2/v/2021-02-17T195932Z/r/image-jpeg Hematoxylin and eosin staining. Representative 5 μm frozen sections prepared for histopathologic analysis from each specimen during RNA extraction of (A) a poorly enhancing glioblastoma tumor core sample with 80% necrosis, and (B) a periphery sample with 0% necrosis from the same patient (GBM no. 1). magnification ×400.JOURNAL/dimp/04.03/00019606-200612000-00002/table1-2/v/2021-02-17T195932Z/r/image-tiff Clinical Description and Histopathologic Features of Glioblastoma Specimens Used in This StudySample and Microarray Quality ControlThe quality of the total RNA extracted correlated with the levels of necrosis ascertained by histopathologic evaluation of the different regions of the tumor. Thus, tumor sections with 80% necrosis yielded a mixture of degraded and undegraded total RNA. The latter, arising from the 20% viable cells within the necrotic area, was successfully transcribed into full-length cDNA and cRNA molecules, with similar profiles to RNA preparations from tumor tissue sections containing 0% necrosis, as assessed by capillary electrophoresis (Fig. 3). All the RNA samples met the quality control criteria previously established in our laboratories,7 which are summarized in Table 2.JOURNAL/dimp/04.03/00019606-200612000-00002/figure3-2/v/2021-02-17T195932Z/r/image-jpeg Total RNA, cDNA, and cRNA profiles. Quality control parameters were obtained from electropherograms using capillary electrophoresis on the Agilent Bioanalyzer 2100, from (A) a core sample containing 80% necrosis, and (B) a periphery sample containing less the 10% necrosis.JOURNAL/dimp/04.03/00019606-200612000-00002/table2-2/v/2021-02-17T195932Z/r/image-tiff Quality Control Criteria for RNA Sample Purity, Integrity, cDNA and cRNA Preparation, and Hybridization in the HG-U133A Microarrays for all the SamplesGenome-wide Gene Expression Analysis of Glioblastoma Core and Periphery SpecimensUntreated glioblastoma tissue samples were compared by microarray analysis. The average expression summary for every probe set from the core samples was compared against the average expression summary for every probe set from the periphery samples. Probe sets that showed 1.3-fold or greater change between the 2 groups were selected for each algorithm, Microarray Suite 5.0 (MAS5), model-based expression indexes (MBEI), and robust multiarray analysis (RMA). Only the probe sets exhibiting differential gene expression according to the 3 algorithms were chosen for further analysis (Fig. 4A). From the concordant 3069 probe sets, 784 showed greater than a 2-fold change in expression levels between the 2 phenotypes, using the MAS5 algorithm. These 784 probe sets were successful in stratifying the different phenotypes when used in a supervised cluster analysis of all-6 samples (Fig. 4B).JOURNAL/dimp/04.03/00019606-200612000-00002/figure4-2/v/2021-02-17T195932Z/r/image-jpeg Graphical representation of genome-wide gene expression analysis. A, Venn diagram representing the comparison of differentially expressed probe sets according to MAS5, the MBEI, and the RMA algorithms. The total number of affected probe sets per algorithm is given in parentheses. The different lists of probe sets were established by comparing the average signal intensity between the 2 groups of GBM samples: core and periphery. B, Dendrograms resulting from hierarchical clustering of genes and samples using a Pearson (centered) correlation between the core and periphery GBM samples. Hierarchical cluster analysis was performed with 784/3096 probe sets that showed greater than 2-fold change between 2 groups of GBM samples.Biologic Function of Differentially Expressed Genes Between Core and PeripheryAmong the genes that were found to be differentially expressed between the core and the periphery samples, 171 were involved in cell adhesion and motility, 79 in inflammation mediated by chemokine and cytokine signaling pathways, 67 in the integrin signaling pathway, 46 in angiogenesis, 23 in the epidermal growth factor receptor (EGFR) signaling pathway, 21 in the fibroblast growth factor (FGF) signaling pathway, 20 in the transforming growth factor-β signaling pathway, 15 in the vascular endothelial growth factor (VEGF) signaling pathway, and 7 in the hypoxia response via HIF1 activation, among other biologic processes. Of interest are distinct biologic pathways that appear to be enhanced in either the core or the periphery region of glioblastoma that have been previously linked to tumor aggressiveness, but not previously linked to specific regional expression patterns. A representative list of genes involved in such processes is summarized in Table 3. A complete list of differentially expressed genes found in this study is available at: http://www.ctrf-cagenomics.vcu.edu/publiclyavaildata.htm (This website can be viewed only with Microsoft Internet Explorer, version 6.0 or greater, or comparable browser).JOURNAL/dimp/04.03/00019606-200612000-00002/table3-2/v/2021-02-17T195932Z/r/image-tiff Summary of Some of the Most Substantial Gene Expression Changes Occurring Between Core and Periphery of GBM, Grouped by Biologic Process and Cellular FunctionTranscript Levels of Select Genes are Shown to be Over-expressed in Core Versus Periphery by QRT-PCRQRT-PCR assays were used to examine expression levels of representative differentially expressed genes involved in biologic aggressiveness of glioblastoma such as matrix metalloproteinase (MMP)-1, and VEGF-A (both enhanced in tumor core regions) and AKT1 (v-akt murine thymoma viral oncogene homolog 1) (enhanced in periphery) to validate differences in expression seen in the Affymetrix microarray studies. The 2−ΔΔCt method8 was used to calculate fold changes in the expression levels of the genes of interest compared with one of the periphery samples. The 2−ΔΔCt method assumes that the efficiencies for the endogenous control amplicon (PPIA) and the gene of interest amplicon are the same. Efficiencies were determined for the amplicons MMP-1, VEGF-A, AKT1, and PPIA on 1:5 dilution series. The slopes of Ct/log dilution plots for the reactions were −2.99, −3.12, −2.89, and −3.18, respectively; thus, all amplicons amplify with similar efficiencies. In each case, a significant difference in transcript expression level was detected. Figure 5A shows the result of comparing cycle threshold values for each gene to normalized fluorescence intensity values obtained by microarray analysis. This analysis also substantiates the microarray data obtained by demonstrating the directionality of the fold change in differential expression by QRT-PCR assays, despite differences in the magnitude of fold changes detected. These differences may presumably be due to the greater linear dynamic range of QRT-PCR (up to 11-log in some reactions), as compared with microarray assays (up to 2-log). Spearman correlation was also used to assess the correlation between Affymetrix probe set intensity and quantitative PCR for the 3 genes analyzed, for which a value of 0.5 was calculated. Furthermore, to investigate if the changes in gene expression summarized in Table 2, found in the microarray study on 1 GBM case, were replicated in a larger cohort of GBM cases, we measured the gene expression levels of MMP-19, VEGF, and PIK3R1 in 5 new GBM samples. Figure 5B shows the results of QRT-PCR assays for these genes on 5 pairs of core and periphery GBM samples. Thus, MMP-19 and VEGF consistently show over-expression in the core region of the tumors; whereas, PIK3R1 gene expression levels were systematically overexpressed in the peripheral regions of GBM.JOURNAL/dimp/04.03/00019606-200612000-00002/figure5-2/v/2021-02-17T195932Z/r/image-jpeg Differential gene expression in core and peripheral GBM samples. A, Validation of microarray data by QRT-PCR. Bar chart representation of cyclophilin A mRNA (PPIA)-normalized gene expression fold changes for MMP-1, VEGF-A, and AKT1 in core samples relative to periphery samples of 1 GBM case, determined by QRT-PCR and by oligonucleotide microarray analysis. B, Bar chart representation of log2-transformed, PPIA-normalized, gene expression ratios between core and periphery samples from 5 additional GBM cases, for MMP-19, VEGF-A, and PIK3R1 transcripts, determined by QRT-PCR. Negative fold-change values correspond to genes that are overexpressed in periphery samples compared with core samples.Protein Levels of Gene Candidates Detected by Western Blot Correlate With Regional Transcript Expression ProfilesWestern blot analysis was used to detect differences in protein levels between tumor core and periphery regions. Matrix metalloproteinase-1 was detected strongly in 3/3 core tissue samples, compared with much weaker detection in the 3 samples from tumor periphery and normal nontumor brain (Fig. 6A). The 54 kd proform of this MMP was predominantly detected, with an additional band detected at 44 kd, reported to correspond to active MMP-1, in which the propeptide has been removed by endopeptidases.9 Similarly, VEGF protein expression was found to be highly expressed in tumor core regions. Several molecular weight species were detected, with the predominant band detected at 55 kd. This is the expected molecular weight for VEGF-A under reducing conditions.10 In contrast, the oncogenic serine-threonine kinase AKT1 was detected in both periphery and core, but periphery samples were found to have more abundant protein expression, detected at 57 kd (Fig. 6B).11 Normalized (% cyclophilin A) densitometric quantification of MMP-1, VEGF, and AKT1 protein levels are shown in Figure 6C, values were derived from 3 independent Western blots per antigen.JOURNAL/dimp/04.03/00019606-200612000-00002/figure6-2/v/2021-02-17T195932Z/r/image-jpeg Differential protein expression in core and periphery GBM samples: validation of microarray data by Western blotting analysis. Western blot analysis of MMP-1, VEGF-A, and AKT1 expression in GBM core and periphery. Cyclophilin A protein level was used as an endogenous control for loading.DISCUSSIONThe present study was designed as a proof of principle experiment to evaluate the use of combining the precision of neuroimaging-assisted stereotactic neurosurgery with the descriptive power of high-density oligonucleotide microarray technology, to investigate intratumoral heterogeneity in GBM. The study used a novel combination of techniques to examine the heterogeneity of gene expression in 6 separate intratumoral regions derived concurrently from the enhancing periphery or from the poorly enhancing core of a single case of untreated de novo glioblastoma. The data obtained was validated for selected candidate genes using regional sampling of 5 additional untreated de novo glioblastomas. Our hypothesis was that hypoxia-induced genes would have greater expression in the poorly perfused, nonenhancing core samples. Conversely, we expected to find increased expression of genes involved with proliferation and/or invasion in the periphery of the tumor. Gene expression profiles were in fact found to differ considerably between the regional samples, revealed by analyzing the microarray intensity data with 3 separate algorithms, identifying a concordance of 3069 candidate probe sets with greater than 1.3-fold difference in expression between collective core and periphery.Selected candidate genes with biologic significance in glioblastoma were chosen from the list of 784 probe sets (623 genes) that differed greater than 2-fold by at least 2 of the 3 algorithms used in the analysis. To validate these findings, we performed quantitative RT-PCR to examine transcript levels of 3 candidate genes, AKT1, MMP-1, and VEGF-A. Each of these candidates was detected at higher transcript expression levels by replicate quantitative RT-PCR assays within in the same tumor region and the direction of differential expression detected in microarray data sets was confirmed. In fact, greater mean regional differences between periphery and core were demonstrated than found with microarray data alone, presumably due to the greater linear dynamic range of QRT-PCR (up to 11-log in some reactions), as compared with microarray assays (up to 2-log). Such differences in these and other candidate genes could have important implications for the biology of these tumors and for treatment response. Antiangiogenic VEGF-targeting drugs, for example, are already in use in preclinical models and in human clinical trials.12–14To analyze these differences at the protein level, Western blotting of tumor tissue lysates from the same regions was performed. This analysis was important to assess whether differences in translation of expressed transcripts exist in different intratumoral regions that could lead to differences in levels of functional protein. This analysis revealed that relative levels of expression between mRNA and protein were maintained for the candidate genes examined.Several different cellular functions were highlighted by the nature of the genes that showed differential gene expression between core samples and peripheral samples in GBM. Among the biologic processes that appear to be enhanced in the perinecrotic cells found in the tumor core, we found those involved in tumor aggressiveness, such as cell migration and survival, as well as angiogenesis (Table 3 and supplementary data).Over-expression of thrombospondin-1 and metalloproteinases, accompanied with a down-regulation of tissue inhibitors of metalloproteinases, suggests an enhancement of cell migration mechanisms or matrix turnover in the perinecrotic cells of the tumor core (supplemental table).15 In addition, other cell survival mechanisms were represented, including detection of over-expression of superoxide dismutase 2 (SOD2) and inhibitor of NF-κB kinase (IKKα).16–17 Evidence of enhanced angiogenesis, probably due to the hypoxic conditions of the core region of GBM, was associated with the over-expression of HIF1A and VEGFs (PGF and VEGF) and their receptors (FLT1 and KDR).18–19In addition, several genes involved in the integrin signaling pathways were found to be differentially expressed in perinecrotic cells compared with cells from the GBM periphery. We found several genes involved in inside-out signaling, such as the interaction of the integrin receptors (ITGB1, 2 and 4) with extracellular matrix proteins (collagens, laminins, and FN1), which might be expected to change the affinity of the receptor and affects cell adhesion, migration, and cell-cell communication by interacting with other transmembrane proteins, such as TM4SF6.20–22 Also, genes were identified that are involved in outside-in signaling derived from the interaction of the integrin receptors (ITGA 5, 6, and 7) with extracellular matrix. This interaction is thought to stimulate cascades of signal transduction proteins, such as TLN1, VCL, ARGEF7, PAK6 and 3, MYLK, and MRLC3, that affect proliferation, and cell motility.23–26The gene profiles seen in the tumor periphery, in turn, were consistent with a separate program of survival and cell growth, including heightened expression of EGFR, the glial mitogen FGF-9, and up-regulated AKT1 expression. We have reported previously that EGFR and AKT1 mRNA expression, protein expression and activity are up-regulated in the tumor periphery relative to tumor core and normal brain.27 Both EGFR and multiple FGFs are known to play a role in central nervous system development and are active in precursor cell expansion during development in vivo, and are able to maintain multipotent central nervous system progenitor phenotype in vitro.28,29 This finding and previous reports of enhanced EGFR signaling in the tumor periphery support a role for EGFR in maintaining the proliferative zone and supporting an invasion program at the area of greatest mitotic activity.1,3,30 Many studies in vitro and in vivo support this notion, including one recent study by Lal et al,31 in which enhanced volumetric growth and invasiveness was demonstrated in U87 xenografts transfected with the constitutively active vIII mutant cDNA. Transgenic mouse models in which EGFR is over-expressed can cause tumorigenesis, exacerbated by additional losses such as CDKN2A deletion.The heightened expression of FGF-9 within tumor periphery represents a new finding that may lead to further understanding of the role this glial mitogen plays in tumor development or progression. Importantly, this dual growth factor signaling could synergistically affect tumor growth in such a way that direct targeting of the EGFR (eg, by one of the numerous EGFR-selective therapeutics in development) is rendered less effective. This possibility should be studied further in model systems and in a large sample of GBMs with well-characterized regional stereotactic biopsy.We believe the methods outlined in the proof of principle experiments presented here are well suited to this work. Our strict quality control parameters included: (1) integral histopathologic evaluation of each snap-frozen specimen studied, (2) requiring a high percentage of tumor tissue for each set of sections used, (3) presence of well-documented tumor features such as necrosis, and (4) the highest quality and integrity of the resultant RNA. Our results demonstrate that this approach can yield high-quality data regarding gene expression from the viable cells within the specimen, and confirm the heterogeneous nature of glioblastoma. The algorithms used in the analysis of microarray data are also of paramount importance, as they will yield widely varying data sets of differentially expressed genes. We have used the consensus gene list derived from analyses by 3 separate algorithms to provide more robust data to inform follow-on studies.Previous studies in glioblastoma have examined patterns of gene expression using classic molecular genetic techniques such as loss of heterozygosity studies and Southern blot analysis and, more recently, have employed microarray-based genomic methods.1,5,6 These studies have suggested a redefinition of molecular subclasses of glioblastoma.32 This still leaves an open question of the relationship of tumor geography to specific patterns of gene expression, particularly because most of the sample material acquired for these studies arises from tumor core regions during surgical debulking. Recently, one group has examined the relationship of extent of necrosis in 17 clinical samples of GBM using an MR image grading system to categorize the tumors. Using microarray-profiling techniques, a short list of necrosis-related gene candidates was reported.33 Intertumoral heterogeneity limited the information gathered by this technique. Our method has the potential to expand the definition of regional or feature-associated gene expression programs, which may depend on the specific repertoire of molecular genetic changes in each tumor case, but may reveal general rules and signaling systems that pertain to molecular subtypes of GBM. Further work in this direction and using the methods outlined here, along with corollary immuno-histologic and in vitro studies, should be pursued toward this end.Our results with multiple samples from a single tumor surgery further demonstrate the significant heterogeneity in gene expression analysis results that may be obtained simply on the basis of the sampling variability (Fig. 4B), though it is reassuring that unsupervised cluster analysis successfully grouped the samples together with respect to the portion of the tumor from which they were taken (core vs. periphery). Further studies of intratumoral heterogeneity in a larger number of tumors are therefore justified, to shed light on techniques best suited to comprehensive characterization of the gene expression pattern for individual tumors.ACKNOWLEDGMENTSThe authors thank Hord, Cullather, and Crone families for generous support of the VCU Brain Tumor Research Program and the MCV Foundation.REFERENCES1. Frederick L, Wang XY, Eley G, et al. Diversity frequency of epidermal growth factor receptor mutations in human glioblastomas. Cancer Res. 2000;60:1383–1387.[Context Link][Medline Link]2. Cheng Y, Ng HK, Ding M, et al. Molecular analysis of microdissected de novo glioblastomas paired astrocytic tumors. J Neuropathol Exp Neurol. 1999;58:120–128.[Context Link][Full Text][CrossRef][Medline Link]3. Okada Y, Hurwitz EE, Esposito JM, et al. Selection pressures of TP53 mutation microenvironmental location influence epidermal growth factor receptor gene amplification in human glioblastomas. Cancer Res. 2003;63:413–416.[Context Link][Medline Link]4. Mariani L, Beaudry C, McDonough WS, et al. Death-associated protein 3 (Dap-3) is overexpressed in invasive glioblastoma cells in vivo and in glioma cell lines with induced motility phenotype in vitro. Clin Cancer Res. 2001;7:2480–2489.[Context Link][Medline Link]5. Hoelzinger DB, Mariani L, Weis J, et al. Gene expression profile of glioblastoma multiforme invasive phenotype points to new therapeutic targets. Neoplasia. 2005;7:7–16.[Context Link][Medline Link]6. Suzuki T, Maruno M, Wada K, et al. Genetic analysis of human glioblastomas using a genomic microarray system. Brain Tumor Pathol. 2004;21:27–34.[Context Link][Full Text][CrossRef][Medline Link]7. Dumur CI, Nasim S, Best AM, et al. Evaluation of quality-control criteria for microarray gene expression analysis. Clin Chem. 2004;50:1994–2002.[Context Link][Full Text][CrossRef][Medline Link]8. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR the 2(-Delta Delta C(T)) Method. Methods. 2001;25:402–408.[Context Link][CrossRef][Medline Link]9. Huntington JT, Shields JM, Der CJ, et al. Over-expression of collagenase 1 (MMP-1) is mediated by the ERK pathway in invasive melanoma cells: role of BRAF mutation and fibroblast growth factor signaling. J Biol Chem. 2004;279:33168–33176.[Context Link][CrossRef][Medline Link]10. Connolly JM, Rose DP. Angiogenesis in two human prostate cancer cell lines with differing metastatic potential when growing as solid tumors in nude mice. J Urol. 1998;160:932–936.[Context Link]Buy Now[Medline Link]11. Gagnon V, Mathieu I, Sexton E, et al. AKT involvement in cisplatin chemoresistance of human uterine cancer cells. Gynecol Oncol. 2004;94:785–795.[Context Link][CrossRef][Medline Link]12. Gordon MS, Margolin K, Talpaz M, et al. Phase I safety pharmacokinetic study of recombinant human anti-vascular endothelial growth factor in patients with advanced cancer. J Clin Oncol. 2001;19:843–850.[Context Link][Full Text][CrossRef][Medline Link]13. Winkler F, Kozin SV, Tong RT, et al. Kinetics of vascular normalization by VEGFR2 blockade governs brain tumor response to radiation: role of oxygenation angiopoietin-1, matrix metalloproteinases. Cancer Cell. 2004;6:553–563.[Context Link][CrossRef][Medline Link]14. Jayson GC, Mulatero C, Ranson M, et al. European Organisation for Research and Treatment of Cancer (EORTC). Phase I investigation of recombinant anti-human vascular endothelial growth factor antibody in patients with advanced cancer. Eur J Cancer. 2005;41:555–563.[Context Link][CrossRef][Medline Link]15. Van Meter TE, Rooprai HK, Kibble MM, et al. The role of matrix metalloproteinase genes in glioma invasion: co-dependent interactive proteolysis. J Neurooncol. 2001;53:213–235.[Context Link][Full Text][Medline Link]16. Tatton WG, Chalmers-Redman RM, Ju WJ, et al. Propargylamines induce antiapoptotic new protein synthesis in serum- and nerve growth factor (NGF)-withdrawn, NGF-differentiated PC-12 cells. J Pharmacol Exp Ther. 2002;301:753–764.[Context Link][CrossRef][Medline Link]17. Gasparian AV, Yao YJ, Kowalczyk D, et al. The role of IKK in constitutive activation of NF-kappaB transcription factor in prostate carcinoma cells. J Cell Sci. 2002;115:141–151.[Context Link][Medline Link]18. Nomura M, Yamagishi S, Harada S, et al. Placenta growth factor (PlGF) mRNA expression in brain tumors. J Neuro-oncol. 1998;40:123–130.[Context Link]19. Hatva E, Kaipainen A, Mentula P, et al. Expression of endothelial cell-specific receptor tyrosine kinases growth factors in human brain tumors. Am J Pathol. 1995;146:368–378.[Context Link][Medline Link]20. Rooprai HK, Vanmeter T, Panou C, et al. The role of integrin receptors in aspects of glioma invasion in vitro. Int J Dev Neurosci. 1999;17:613–623.[Context Link][CrossRef][Medline Link]21. Calderwood DA, Fujioka Y, de Pereda JM, et al. Integrin beta cytoplasmic domain interactions with phosphotyrosine-binding domains: a structural prototype for diversity in integrin signaling. Proc Natl Acad Sci USA. 2003;100:2272–2277.[Context Link][CrossRef][Medline Link]22. Berditchevski F, Zutter MM, Hemler ME. Characterization of novel complexes on the cell surface between integrins proteins with 4 transmembrane domains (TM4 proteins). Mol Biol Cell. 1996;7:193–207.[Context Link][CrossRef][Medline Link]23. Arnaout MA. Integrin structure: new twists turns in dynamic cell adhesion. Immunol Rev. 2002;186:125–140.[Context Link][Full Text][CrossRef][Medline Link]24. Takagi J, Petre BM, Walz T, et al. Global conformational rearrangements in integrin extracellular domains in outside-in and inside-out signaling. Cell. 2002;110:599–611.[Context Link][CrossRef][Medline Link]25. Hashimoto S, Tsubouchi A, Mazaki Y, et al. Interaction of paxillin with p21-activated Kinase (PAK). Association of paxillin alpha with the kinase-inactive and the Cdc42-activated forms of PAK3. J Biol Chem. 2001;276:6037–6045.[Context Link][CrossRef][Medline Link]26. Vijayan KV, Liu Y, Dong JF, et al. Enhanced activation of mitogen-activated protein kinase and myosin light chain kinase by the Pro33 polymorphism of integrin beta 3. J Biol Chem. 2003;278:3860–3867.[Context Link][CrossRef][Medline Link]27. VanMeter T, Dumur C, Hafez H, et al. Intratumoral heterogeneity in glioblastoma: up-regulation of invasion-related genes in peri-necrotic tissue. Neuro-Oncology. 2004;6:340; Abstract.[Context Link]28. Tropepe V, Sibilia M, Ciruna BG, et al. Distinct neural stem cells proliferate in response to EGF FGF in the developing mouse telencephalon. Dev Biol. 1999;208:166–188.[Context Link][CrossRef][Medline Link]29. Kuhn HG, Winkler J, Kempermann G, et al. Epidermal growth factor fibroblast growth factor-2 have different effects on neural progenitors in the adult rat brain. J Neurosci. 1997;17:5820–5829.[Context Link][CrossRef][Medline Link]30. Van Meter TE, Broaddus WC, Rooprai HK, et al. Induction of membrane-type-1 matrix metalloproteinase by epidermal growth factor-mediated signaling in gliomas. Neuro-Oncology. 2004;6:188–199.[Context Link][CrossRef][Medline Link]31. Lal A, Glazer CA, Martinson HM, et al. Mutant epidermal growth factor receptor up-regulates molecular effectors of tumor invasion. Cancer Res. 2002;62:3335–3339.[Context Link][Medline Link]32. Kim S, Dougherty ER, Shmulevich I, et al. Identification of combination gene sets for glioma classification. Mol Cancer Ther. 2002;1:1229–1236.[Context Link][Medline Link]33. Raza SM, Fuller GN, Rhee CH, et al. Identification of necrosis-associated genes in glioblastoma by cDNA microarray analysis. Clin Cancer Res. 2004;10:212–221.[Context Link][Medline Link]34. Affymetrix GeneChip®. Expression Analysis Technical Manual. Santa Clava, CA: Affymetrix; 2001.[Context Link]35. Sugita M, Geraci M, Gao B, et al. Combined use of oligonucleotide tissue microarrays identifies cancer/testis antigens as biomarkers in lung carcinoma. Cancer Res. 2002;62:3971–3979.[Context Link][Medline Link]36. Irizarry RA, Bolstad BM, Collin F, et al. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res. 2003;31:e15.[Context Link][Full Text][Medline Link]37. Li C, Wong WH. Model-based analysis of oligonucleotide arrays: model validation design issues standard error application. Genome Biol. 2001;2:RESEARCH0032.[Context Link][CrossRef][Medline Link]38. Hubbell E, Liu W-M, Mei R. Robust estimators for expression analysis. Bioinformatics. 2002;18:1585–1592.[Context Link][Full Text][CrossRef][Medline Link]39. Zhang L, Wang L, Ravindranathan A, et al. A new algorithm for analysis of oligonucleotide arrays: application to expression profiling in mouse brain regions. J Mol Biol. 2002;317:225–235.[Context Link][CrossRef][Medline Link]glioblastoma; microarray; heterogeneity; therapeutic target; MRI00019606-200612000-0000200004892_1998_160_932_connolly_angiogenesis_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e952_citationRF_FLOATING))|25||ovftdb|SL00004892199816093225citation_FROM_JRF_ID_d1008e952_citationRF_FLOATINGBuy Now>&interface=OvidSP00019606-200612000-0000200004892_1998_160_932_connolly_angiogenesis_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e952_citationRF_FLOATING))|11065405||ovftdb|SL00004892199816093211065405citation_FROM_JRF_ID_d1008e952_citationRF_FLOATING[Medline Link]972059000019606-200612000-0000200003979_2004_94_785_gagnon_chemoresistance_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e980_citationRF_FLOATING))|11065213||ovftdb|SL0000397920049478511065213citation_FROM_JRF_ID_d1008e980_citationRF_FLOATING[CrossRef]10.1016%2Fj.ygyno.2004.06.02300019606-200612000-0000200003979_2004_94_785_gagnon_chemoresistance_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e980_citationRF_FLOATING))|11065405||ovftdb|SL0000397920049478511065405citation_FROM_JRF_ID_d1008e980_citationRF_FLOATING[Medline Link]1535037400019606-200612000-0000200005083_2001_19_843_gordon_pharmacokinetic_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING))|11065404||ovftdb|SL0000508320011984311065404citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING[Full Text]00005083-200102010-0003300019606-200612000-0000200005083_2001_19_843_gordon_pharmacokinetic_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING))|11065213||ovftdb|SL0000508320011984311065213citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING[CrossRef]10.1200%2FJCO.2001.19.3.84300019606-200612000-0000200005083_2001_19_843_gordon_pharmacokinetic_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING))|11065405||ovftdb|SL0000508320011984311065405citation_FROM_JRF_ID_d1008e1014_citationRF_FLOATING[Medline Link]1115703800019606-200612000-0000200136248_2004_6_553_winkler_metalloproteinases_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1048_citationRF_FLOATING))|11065213||ovftdb|SL001362482004655311065213citation_FROM_JRF_ID_d1008e1048_citationRF_FLOATING[CrossRef]10.1016%2FS1535-6108%2804%2900305-800019606-200612000-0000200136248_2004_6_553_winkler_metalloproteinases_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1048_citationRF_FLOATING))|11065405||ovftdb|SL001362482004655311065405citation_FROM_JRF_ID_d1008e1048_citationRF_FLOATING[Medline Link]1560796000019606-200612000-0000200002212_2005_41_555_jayson_investigation_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1082_citationRF_FLOATING))|11065213||ovftdb|SL0000221220054155511065213citation_FROM_JRF_ID_d1008e1082_citationRF_FLOATING[CrossRef]10.1016%2Fj.ejca.2004.11.02100019606-200612000-0000200002212_2005_41_555_jayson_investigation_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1082_citationRF_FLOATING))|11065405||ovftdb|SL0000221220054155511065405citation_FROM_JRF_ID_d1008e1082_citationRF_FLOATING[Medline Link]1573756000019606-200612000-0000200005084_2001_53_213_vanmeter_metalloproteinase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1116_citationRF_FLOATING))|11065404||ovftdb|SL0000508420015321311065404citation_FROM_JRF_ID_d1008e1116_citationRF_FLOATING[Full Text]00005084-200105320-0001200019606-200612000-0000200005084_2001_53_213_vanmeter_metalloproteinase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1116_citationRF_FLOATING))|11065405||ovftdb|SL0000508420015321311065405citation_FROM_JRF_ID_d1008e1116_citationRF_FLOATING[Medline Link]1171607200019606-200612000-0000200005231_2002_301_753_tatton_propargylamines_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1150_citationRF_FLOATING))|11065213||ovftdb|SL00005231200230175311065213citation_FROM_JRF_ID_d1008e1150_citationRF_FLOATING[CrossRef]10.1124%2Fjpet.301.2.75300019606-200612000-0000200005231_2002_301_753_tatton_propargylamines_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1150_citationRF_FLOATING))|11065405||ovftdb|SL00005231200230175311065405citation_FROM_JRF_ID_d1008e1150_citationRF_FLOATING[Medline Link]1196108200019606-200612000-0000200004646_2002_115_141_gasparian_transcription_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1185_citationRF_FLOATING))|11065405||ovftdb|SL00004646200211514111065405citation_FROM_JRF_ID_d1008e1185_citationRF_FLOATING[Medline Link]1180173200019606-200612000-0000200000457_1995_146_368_hatva_endothelial_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1253_citationRF_FLOATING))|11065405||ovftdb|SL00000457199514636811065405citation_FROM_JRF_ID_d1008e1253_citationRF_FLOATING[Medline Link]785674900019606-200612000-0000200002823_2000_60_1383_frederick_glioblastomas_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e652_citationRF_FLOATING))|11065405||ovftdb|SL00002823200060138311065405citation_FROM_JRF_ID_d1008e652_citationRF_FLOATING[Medline Link]1072870300019606-200612000-0000200000100_1999_17_613_rooprai_receptors_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1287_citationRF_FLOATING))|11065213||ovftdb|SL0000010019991761311065213citation_FROM_JRF_ID_d1008e1287_citationRF_FLOATING[CrossRef]10.1016%2FS0736-5748%2899%2900051-900019606-200612000-0000200000100_1999_17_613_rooprai_receptors_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1287_citationRF_FLOATING))|11065405||ovftdb|SL0000010019991761311065405citation_FROM_JRF_ID_d1008e1287_citationRF_FLOATING[Medline Link]1057142200019606-200612000-0000200006702_2003_100_2272_calderwood_phosphotyrosinebinding_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1321_citationRF_FLOATING))|11065213||ovftdb|SL000067022003100227211065213citation_FROM_JRF_ID_d1008e1321_citationRF_FLOATING[CrossRef]10.1073%2Fpnas.26279199900019606-200612000-0000200006702_2003_100_2272_calderwood_phosphotyrosinebinding_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1321_citationRF_FLOATING))|11065405||ovftdb|SL000067022003100227211065405citation_FROM_JRF_ID_d1008e1321_citationRF_FLOATING[Medline Link]1260671100019606-200612000-0000200002470_1996_7_193_berditchevski_characterization_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1355_citationRF_FLOATING))|11065213||ovftdb|SL000024701996719311065213citation_FROM_JRF_ID_d1008e1355_citationRF_FLOATING[CrossRef]10.1091%2Fmbc.7.2.19300019606-200612000-0000200002470_1996_7_193_berditchevski_characterization_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1355_citationRF_FLOATING))|11065405||ovftdb|SL000024701996719311065405citation_FROM_JRF_ID_d1008e1355_citationRF_FLOATING[Medline Link]868855200019606-200612000-0000200004210_2002_186_125_arnout_structure_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING))|11065404||ovftdb|SL00004210200218612511065404citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING[Full Text]00004210-200208000-0001200019606-200612000-0000200004210_2002_186_125_arnout_structure_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING))|11065213||ovftdb|SL00004210200218612511065213citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING[CrossRef]10.1034%2Fj.1600-065X.2002.18612.x00019606-200612000-0000200004210_2002_186_125_arnout_structure_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING))|11065405||ovftdb|SL00004210200218612511065405citation_FROM_JRF_ID_d1008e1388_citationRF_FLOATING[Medline Link]1223436800019606-200612000-0000200002867_2002_110_599_takagi_conformational_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1410_citationRF_FLOATING))|11065213||ovftdb|SL00002867200211059911065213citation_FROM_JRF_ID_d1008e1410_citationRF_FLOATING[CrossRef]10.1016%2FS0092-8674%2802%2900935-200019606-200612000-0000200002867_2002_110_599_takagi_conformational_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1410_citationRF_FLOATING))|11065405||ovftdb|SL00002867200211059911065405citation_FROM_JRF_ID_d1008e1410_citationRF_FLOATING[Medline Link]1223097700019606-200612000-0000200004613_2000_276_6037_hashimoto_interaction_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1444_citationRF_FLOATING))|11065213||ovftdb|SL000046132000276603711065213citation_FROM_JRF_ID_d1008e1444_citationRF_FLOATING[CrossRef]10.1074%2Fjbc.M00585420000019606-200612000-0000200004613_2000_276_6037_hashimoto_interaction_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1444_citationRF_FLOATING))|11065405||ovftdb|SL000046132000276603711065405citation_FROM_JRF_ID_d1008e1444_citationRF_FLOATING[Medline Link]1109607300019606-200612000-0000200004613_2003_278_3860_vijayan_mitogenactivated_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1478_citationRF_FLOATING))|11065213||ovftdb|SL000046132003278386011065213citation_FROM_JRF_ID_d1008e1478_citationRF_FLOATING[CrossRef]10.1074%2Fjbc.M20868020000019606-200612000-0000200004613_2003_278_3860_vijayan_mitogenactivated_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1478_citationRF_FLOATING))|11065405||ovftdb|SL000046132003278386011065405citation_FROM_JRF_ID_d1008e1478_citationRF_FLOATING[Medline Link]1246099100019606-200612000-0000200002006_1999_208_166_tropepe_telencephalon_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1517_citationRF_FLOATING))|11065213||ovftdb|SL00002006199920816611065213citation_FROM_JRF_ID_d1008e1517_citationRF_FLOATING[CrossRef]10.1006%2Fdbio.1998.919200019606-200612000-0000200002006_1999_208_166_tropepe_telencephalon_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1517_citationRF_FLOATING))|11065405||ovftdb|SL00002006199920816611065405citation_FROM_JRF_ID_d1008e1517_citationRF_FLOATING[Medline Link]1007585000019606-200612000-0000200005095_1997_17_5820_kuhn_progenitors_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1552_citationRF_FLOATING))|11065213||ovftdb|SL00005095199717582011065213citation_FROM_JRF_ID_d1008e1552_citationRF_FLOATING[CrossRef]10.1523%2FJNEUROSCI.17-15-05820.199700019606-200612000-0000200005095_1997_17_5820_kuhn_progenitors_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1552_citationRF_FLOATING))|11065405||ovftdb|SL00005095199717582011065405citation_FROM_JRF_ID_d1008e1552_citationRF_FLOATING[Medline Link]922178000019606-200612000-0000200005072_1999_58_120_cheng_microdissected_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING))|11065404||ovftdb|SL0000507219995812011065404citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING[Full Text]00005072-199902000-0000200019606-200612000-0000200005072_1999_58_120_cheng_microdissected_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING))|11065213||ovftdb|SL0000507219995812011065213citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING[CrossRef]10.1097%2F00005072-199902000-0000200019606-200612000-0000200005072_1999_58_120_cheng_microdissected_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING))|11065405||ovftdb|SL0000507219995812011065405citation_FROM_JRF_ID_d1008e686_citationRF_FLOATING[Medline Link]1002909500019606-200612000-0000200129338_2004_6_188_meter_metalloproteinase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1586_citationRF_FLOATING))|11065213||ovftdb|SL001293382004618811065213citation_FROM_JRF_ID_d1008e1586_citationRF_FLOATING[CrossRef]10.1215%2FS115285170300048600019606-200612000-0000200129338_2004_6_188_meter_metalloproteinase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1586_citationRF_FLOATING))|11065405||ovftdb|SL001293382004618811065405citation_FROM_JRF_ID_d1008e1586_citationRF_FLOATING[Medline Link]1527971100019606-200612000-0000200002823_2002_62_3335_lal_epidermal_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1620_citationRF_FLOATING))|11065405||ovftdb|SL00002823200262333511065405citation_FROM_JRF_ID_d1008e1620_citationRF_FLOATING[Medline Link]1206796900019606-200612000-0000200134960_2002_1_1229_kim_identification_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1654_citationRF_FLOATING))|11065405||ovftdb|SL0013496020021122911065405citation_FROM_JRF_ID_d1008e1654_citationRF_FLOATING[Medline Link]1247970400019606-200612000-0000200043746_2004_10_212_raza_identification_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1688_citationRF_FLOATING))|11065405||ovftdb|SL0004374620041021211065405citation_FROM_JRF_ID_d1008e1688_citationRF_FLOATING[Medline Link]1473447200019606-200612000-0000200002823_2002_62_3971_sugita__|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1734_citationRF_FLOATING))|11065405||ovftdb|SL00002823200262397111065405citation_FROM_JRF_ID_d1008e1734_citationRF_FLOATING[Medline Link]1212432900019606-200612000-0000200006178_2003_31_e15_irizarry_affymetrix_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1768_citationRF_FLOATING))|11065404||ovftdb|SL00006178200331e1511065404citation_FROM_JRF_ID_d1008e1768_citationRF_FLOATING[Full Text]00006178-200302150-0003100019606-200612000-0000200006178_2003_31_e15_irizarry_affymetrix_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1768_citationRF_FLOATING))|11065405||ovftdb|SL00006178200331e1511065405citation_FROM_JRF_ID_d1008e1768_citationRF_FLOATING[Medline Link]1258226000019606-200612000-0000200129669_2001_2__li_oligonucleotide_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1802_citationRF_FLOATING))|11065213||ovftdb|SL001296692001211065213citation_FROM_JRF_ID_d1008e1802_citationRF_FLOATING[CrossRef]10.1186%2Fgb-2001-2-8-research003100019606-200612000-0000200129669_2001_2__li_oligonucleotide_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1802_citationRF_FLOATING))|11065405||ovftdb|SL001296692001211065405citation_FROM_JRF_ID_d1008e1802_citationRF_FLOATING[Medline Link]1153221600019606-200612000-0000200124336_2002_18_1585_hubbell_estimators_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING))|11065404||ovftdb|SL00124336200218158511065404citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING[Full Text]00124336-200212000-0000600019606-200612000-0000200124336_2002_18_1585_hubbell_estimators_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING))|11065213||ovftdb|SL00124336200218158511065213citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING[CrossRef]10.1093%2Fbioinformatics%2F18.12.158500019606-200612000-0000200124336_2002_18_1585_hubbell_estimators_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING))|11065405||ovftdb|SL00124336200218158511065405citation_FROM_JRF_ID_d1008e1829_citationRF_FLOATING[Medline Link]1249044200019606-200612000-0000200005027_2002_317_225_zhang_oligonucleotide_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1861_citationRF_FLOATING))|11065213||ovftdb|SL00005027200231722511065213citation_FROM_JRF_ID_d1008e1861_citationRF_FLOATING[CrossRef]10.1006%2Fjmbi.2001.535000019606-200612000-0000200005027_2002_317_225_zhang_oligonucleotide_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e1861_citationRF_FLOATING))|11065405||ovftdb|SL00005027200231722511065405citation_FROM_JRF_ID_d1008e1861_citationRF_FLOATING[Medline Link]1190283900019606-200612000-0000200002823_2003_63_413_okada_microenvironmental_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e720_citationRF_FLOATING))|11065405||ovftdb|SL0000282320036341311065405citation_FROM_JRF_ID_d1008e720_citationRF_FLOATING[Medline Link]1254379600019606-200612000-0000200043736_2001_7_2480_mariani_overexpressed_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e754_citationRF_FLOATING))|11065405||ovftdb|SL0004373620017248011065405citation_FROM_JRF_ID_d1008e754_citationRF_FLOATING[Medline Link]1148983000019606-200612000-0000200042394_2005_7_7_hoelzinger_glioblastoma_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e789_citationRF_FLOATING))|11065405||ovftdb|SL0004239420057711065405citation_FROM_JRF_ID_d1008e789_citationRF_FLOATING[Medline Link]1572081300019606-200612000-0000200075348_2004_21_27_suzuki_glioblastomas_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING))|11065404||ovftdb|SL000753482004212711065404citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING[Full Text]00075348-200402000-0000600019606-200612000-0000200075348_2004_21_27_suzuki_glioblastomas_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING))|11065213||ovftdb|SL000753482004212711065213citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING[CrossRef]10.1007%2FBF0248217400019606-200612000-0000200075348_2004_21_27_suzuki_glioblastomas_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING))|11065405||ovftdb|SL000753482004212711065405citation_FROM_JRF_ID_d1008e823_citationRF_FLOATING[Medline Link]1569696600019606-200612000-0000200003030_2004_50_1994_dumur_evaluation_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING))|11065404||ovftdb|SL00003030200450199411065404citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING[Full Text]00003030-200411000-0001000019606-200612000-0000200003030_2004_50_1994_dumur_evaluation_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING))|11065213||ovftdb|SL00003030200450199411065213citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING[CrossRef]10.1373%2Fclinchem.2004.03322500019606-200612000-0000200003030_2004_50_1994_dumur_evaluation_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING))|11065405||ovftdb|SL00003030200450199411065405citation_FROM_JRF_ID_d1008e857_citationRF_FLOATING[Medline Link]1536488500019606-200612000-0000200034071_2001_25_402_livak_quantitative_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e891_citationRF_FLOATING))|11065213||ovftdb|SL0003407120012540211065213citation_FROM_JRF_ID_d1008e891_citationRF_FLOATING[CrossRef]10.1006%2Fmeth.2001.126200019606-200612000-0000200034071_2001_25_402_livak_quantitative_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e891_citationRF_FLOATING))|11065405||ovftdb|SL0003407120012540211065405citation_FROM_JRF_ID_d1008e891_citationRF_FLOATING[Medline Link]1184660900019606-200612000-0000200004613_2004_279_33168_huntington_collagenase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e918_citationRF_FLOATING))|11065213||ovftdb|SL0000461320042793316811065213citation_FROM_JRF_ID_d1008e918_citationRF_FLOATING[CrossRef]10.1074%2Fjbc.M40510220000019606-200612000-0000200004613_2004_279_33168_huntington_collagenase_|00019606-200612000-00002#xpointer(id(citation_FROM_JRF_ID_d1008e918_citationRF_FLOATING))|11065405||ovftdb|SL0000461320042793316811065405citation_FROM_JRF_ID_d1008e918_citationRF_FLOATING[Medline Link]15184373 GBM sample acquisition. Representative preoperative MR images taken as intraoperative “snap-shots” of the precise location of excision using Stealth stereotactic biopsy to select specific tumor areas during prospective tissue collection from untreated glioblastoma. A, Sampling from tumor core, and (B) sampling from the enhancing tumor periphery, both taken after gadolinium contrast. Hematoxylin and eosin staining. Representative 5 μm frozen sections prepared for histopathologic analysis from each specimen during RNA extraction of (A) a poorly enhancing glioblastoma tumor core sample with 80% necrosis, and (B) a periphery sample with 0% necrosis from the same patient (GBM no. 1). magnification ×400. Clinical Description and Histopathologic Features of Glioblastoma Specimens Used in This Study Total RNA, cDNA, and cRNA profiles. Quality control parameters were obtained from electropherograms using capillary electrophoresis on the Agilent Bioanalyzer 2100, from (A) a core sample containing 80% necrosis, and (B) a periphery sample containing less the 10% necrosis. Quality Control Criteria for RNA Sample Purity, Integrity, cDNA and cRNA Preparation, and Hybridization in the HG-U133A Microarrays for all the Samples Graphical representation of genome-wide gene expression analysis. A, Venn diagram representing the comparison of differentially expressed probe sets according to MAS5, the MBEI, and the RMA algorithms. The total number of affected probe sets per algorithm is given in parentheses. The different lists of probe sets were established by comparing the average signal intensity between the 2 groups of GBM samples: core and periphery. B, Dendrograms resulting from hierarchical clustering of genes and samples using a Pearson (centered) correlation between the core and periphery GBM samples. Hierarchical cluster analysis was performed with 784/3096 probe sets that showed greater than 2-fold change between 2 groups of GBM samples. Summary of Some of the Most Substantial Gene Expression Changes Occurring Between Core and Periphery of GBM, Grouped by Biologic Process and Cellular Function Differential gene expression in core and peripheral GBM samples. A, Validation of microarray data by QRT-PCR. Bar chart representation of cyclophilin A mRNA (PPIA)-normalized gene expression fold changes for MMP-1, VEGF-A, and AKT1 in core samples relative to periphery samples of 1 GBM case, determined by QRT-PCR and by oligonucleotide microarray analysis. B, Bar chart representation of log2-transformed, PPIA-normalized, gene expression ratios between core and periphery samples from 5 additional GBM cases, for MMP-19, VEGF-A, and PIK3R1 transcripts, determined by QRT-PCR. Negative fold-change values correspond to genes that are overexpressed in periphery samples compared with core samples. Differential protein expression in core and periphery GBM samples: validation of microarray data by Western blotting analysis. Western blot analysis of MMP-1, VEGF-A, and AKT1 expression in GBM core and periphery. Cyclophilin A protein level was used as an endogenous control for loading.Microarray Analysis of MRI-defined Tissue Samples in Glioblastoma Reveals Differences in Regional Expression of Therapeutic TargetsVan Meter Timothy PhD; Dumur, Catherine PhD; Hafez, Naiel MD; Garrett, Carleton MD; Fillmore, Helen PhD; Broaddus, William C. MD, PhDOriginal ArticlesOriginal Articles415p 195-205

You can read the full text of this article if you:

Access through Ovid