Identification of Necrosis-Associated Genes in Glioblastoma by Cdna Microarray Analysis

212 Vol. 10, 212–221, January 1, 2004 Clinical Cancer Research

Identification of Necrosis-Associated Genes in Glioblastoma by cDNA Microarray Analysis

Shaan M. Raza,1 Gregory N. Fuller,2 the 26 genes between grade III necrosis GBM and anaplastic Chang Hun Rhee,4 Suyun Huang,1 astrocytoma (AA) samples; all but 1 of the genes had ele- Kenneth Hess,3 Wei Zhang,2 and vated expression when comparing necrosis grade III with 1 AA samples. Two factors, the ephrin type A receptor 1 and Raymond Sawaya the prostaglandin E receptor EP4 subtype, not previously 1 2 3 2 Departments of Neurosurgery, Pathology, and Biostatistics, Brain considered in this context, were highlighted because of their Tumor Center, The University of Texas M. D. Anderson Cancer Center, Houston, Texas, and 4Department of Neurosurgery, Korean particularly high (positive) correlation coefficients; immu- Cancer Center Hospital, Seoul, Korea nostaining showed the products of these two genes to be localized in perinecrotic and necrotic regions and to be overexpressed in grade III GBMs, but not AAs. These two ABSTRACT molecules also showed significant correlation with survival .in a combined model (0.0034 ؍ Purpose: In the field of cancer research, there has been of GBM patients (P a paucity of interest in necrosis, whereas studies focusing on Conclusions: The application of cDNA expression mi- apoptosis abound. In neuro-oncology, this is particularly croarray analysis has identified specific genes and patterns surprising because of the importance of necrosis as a hall- of gene expression that may help elucidate the molecular mark of glioblastoma (GBM), the most malignant and most basis of necrogenesis in GBM. Additional studies will be common primary brain tumor, and the fact that the degree required to further investigate and confirm these findings. of necrosis has been shown to be inversely related to patient survival. It is therefore of considerable interest and importance to identify genes and gene products related to necrosis INTRODUCTION formation. In cancer research, the subject of cell death has been Experimental Design: We used a nylon cDNA microar- extensively studied with the goal of achieving therapeutically ray to analyze mRNA expression of 588 universal cellular induced tumor death. Among the two major forms of cell death genes in 15 surgically resected human GBM samples with encountered in biology, apoptosis has received a proportionately varying degrees of necrosis. Gene expression was correlated greater degree of attention and emphasis than has necrosis (1, 2). with the degree of necrosis using rank correlation coeffi- A large body of recent literature has provided an understanding cients. The expression of identified genes was compared with of the physiological and molecular events that lead to apoptosis; their expression in tissue samples from 5 anaplastic astro- yet similar characterization of necrosis cannot be found. cytomas (AAs). Immunostaining was used to determine In the field of neuro-oncology, this paucity of interest in whether genes showing the most positive correlation with necrosis is surprising because it is a hallmark of the most necrosis were increasingly expressed in tumor tissues, as common and most malignant primary brain tumor—glioblas- grade of necrosis increased. toma (GBM). In fact, GBM is differentially diagnosed from Results: The hybridization results indicated that 26 lower grade astrocytomas based on the histological presence of genes showed significant correlation with the amount of de novo tumor necrosis and associated microvascular prolifer- necrosis. All 26 genes had functions associated with either ation. Clinical studies indicate that of all clinical, neuroimaging, Ras, Akt, tumor necrosis factor ␣, nuclear factor ␬B, apo- and histopathological characteristics (including age), necrosis ptosis, procoagulation, or hypoxia. Nine genes were posi- that is visible on magnetic resonance imaging scans has the tively correlated with necrosis grade, and 17 genes were greatest prognostic value and is inversely related to patient negatively correlated with necrosis grade. There were sig- survival (3). Based on the clinical implications and potential for nificant differences in the median expression levels of 3 of novel therapeutic interventions, the molecular factors mediating necrogenesis in GBM warrant investigation. An initial step in this process would be the identification of specific gene products associated with necrosis. Received 1/28/03; revised 9/11/03; accepted 9/16/03. Although there have been a few limited studies of one or Grant support: Texas Higher Education Coordinating Board (to W. Z. two necrosis-associated molecules, there have been no efforts to and G. N. F.) and the Anthony Bullock III Brain Tumor Research Fund characterize global gene expression changes associated with (to R. S.). The costs of publication of this article were defrayed in part by the varying degrees of necrosis in GBM (4). In this study, we used payment of page charges. This article must therefore be hereby marked cDNA expression microarrays to identify the molecular finger- advertisement in accordance with 18 U.S.C. Section 1734 solely to prints of 15 GBM samples with differing degrees of necrosis; indicate this fact. this was done with the goal of identifying genes whose mRNA Requests for reprints: Raymond Sawaya, Department of Neurosur- gery–442, The University of Texas M. D. Anderson Cancer Center, expression is correlated with the formation of necrosis. In hopes 1515 Holcombe Boulevard, Houston, Texas 77030. Phone: (713) 792- of establishing a relationship between necrosis and the identified 2400; Fax: (713) 794-4950; E-mail: [email protected]. genes, their expression levels in anaplastic astrocytomas (AAs),

which by definition (St. Anne/Mayo criteria) are tumors that are Table 1 Necrosis grades of human glioblastoma samplesa and not associated with necrosis, were determined and compared survival time of patients with expression levels in grade III necrosis GBM samples, Necrosis Sample Median survival which are tumors characterized by extensive amounts of necrosis. grade no. (weeks) 04 64 MATERIALS AND METHODS I 2 116 II 5 53 Primary Glioma Tissues. All human tumor samples III 4 25 were obtained from the Brain Tumor Center Tissue Bank at The a The following grading system (see Ref. 3) was used: grade 0, no University of Texas M. D. Anderson Cancer Center (Houston, necrosis apparent on the magnetic resonance imaging scan; grade I, TX). After surgical removal, tissue samples were quickly frozen amount of necrosis is less than 25% of the tumor volume; grade II, at Ϫ80°C. All tissue samples used for cDNA analysis were amount of necrosis is between 25% and 50% of the tumor volume; grade evaluated by a neuropathologist; all specimens selected were III, amount of necrosis is greater than 50% of the tumor volume. In all characterized by dense tumor cellularity and the presence of cases, the extent of necrosis was confirmed at surgery. Ͻ10% normal brain tissue. A second neuropathologist confirmed all diagnoses. Grading of astrocytic neoplasms was per- formed according to the St. Anne/Mayo criteria. A total of 5 AAs and 15 GBM samples were analyzed. The 15 GBM sam- volume (Ref. 3; Fig. 1). The extent of necrosis was confirmed in ples obtained were further graded based on the amount of all cases during surgery. Within our sample population, four necrosis observed on magnetic resonance imaging scans. Ne- tumors were grade 0, two were grade I, five were grade II, crosis on magnetic resonance imaging scans appears as a and four were grade III. The sample distribution is shown in hypointense region of T1 signal surrounded by a contrast- Table 1. enhanced region representing viable tumor. Necrosis was graded Isolation of Total RNA and mRNA from Tissue Sam- according to the following previously described system: grade 0, ples. The frozen tissue samples were ground to a powder in no necrosis apparent on the magnetic resonance imaging scan; liquid nitrogen using a mortar and pestle and then treated with grade I, amount of necrosis is Ͻ25% of the tumor volume; grade lysis buffer, TRI Reagent (Molecular Research Center, Cincin- II, amount of necrosis is between 25% and 50% of the tumor nati, OH), to extract nucleic acids and prevent their degradation. volume; grade III, amount of necrosis is Ͼ50% of the tumor Total RNA was isolated as described previously (5). The quality of this RNA was checked by electrophoresis on a denaturing formaldehyde agarose gel. The presence of high-quality undegraded RNA was verified by the observation of a lack of streaking on the lower (anodal) part of the sample lane on the gel. A further indication that high-quality RNA was obtained was the presence on the gel of a 28S rRNA band twice as intense as that of the 18S rRNA band. A poly(dT) mRNA isolation column (Qiagen, Inc., Chatsworth, CA) was used to separate mRNA from high-quality total RNA. With this column, we recovered 1–2 ␮g of mRNA from approximately 200 ␮g of total RNA, but only 0.5–1.0 ␮g of mRNA was required for analysis in the cDNA array. In this study, we obtained high-quality RNA from Ͼ90% of the resected samples. Because undegraded RNA is scarce within regions of necrosis, total RNA and mRNA were predominantly isolated from viable nonnecrotic tumor regions. Hybridization to Human Atlas cDNA Expression Array Blots. cDNA fragments, approximately 200- to 500-bp long, representing 588 human genes with known functions and tight transcriptional controls were immobilized in duplicate on a nylon membrane (Clontech Laboratories, Inc.). To minimize cross-hybridization and nonspecific binding of cDNA probes, each fragment was selected as a unique sequence without a poly(A) tail, highly homologous sequences, or repetitive ele- ments. 32P-labeled cDNA probes were generated through re- verse transcription of 0.5–1.0 ␮g of each analyzed poly(A)ϩ Fig. 1 Magnetic resonance images of four glioblastoma patients with ␣ 32 varying degrees of necrosis. Necrosis appears as the region of decreased RNA sample in the presence of [ - P]dATP; the probes were T1 signal intensity surrounded by a contrast-enhanced region represent- then hybridized to the array. After a high-stringency wash of the ing viable tumor mass. Necrosis grading was done as described in Ref. array, the hybridization pattern obtained was analyzed by auto- 3: grade 0, no necrosis apparent on the magnetic resonance imaging radiography and quantified by phosphorimaging using Image- scan; grade I, amount of necrosis is Ͻ25% of the tumor volume; grade II, amount of necrosis is between 25% and 50% of the tumor volume; Quant software and a Storm 840 PhosphorImager. Because the and grade III, amount of necrosis is Ͼ50% of the tumor volume amount of each cDNA fragment on the membrane was in excess (reproduced with permission from Hammoud et al., Ref. 3). (10 ng), cDNA-probe binding interactions were linear.

For this study, membranes from the same production lot RESULTS were used because our previous work indicates that this makes A statistical analysis confirmed the negative prognostic the results of the technology reproducible. Lastly, tissue samples value of necrosis. Within this study, the median survival times were submitted to analysis in a blind manner with respect to for patients having GBMs with necrosis of grades 0, I, II, and III their degree of necrosis. were 64, 116, 53, and 25 weeks, respectively. This necrosis- Immunohistochemistry. Formalin-fixed, paraffin- survival relationship is shown in Table 1. The hazard ratio for embedded tissues were cut into 5-␮m-thick sections and patient groups 0 and I versus groups II and III was 2.2 (P ϭ mounted on positively charged Superfrost slides. Tissue sections 0.029). Numerous gene expression changes were also found to were deparaffinized in xylene, followed by a graded series of be associated with necrosis grade. Hybridization experiments alcohol washes, and rehydrated in PBS. Sections were then revealed that changes in the expression of 26 genes were sig- incubated with pepsin (Biomedia, Foster City, CA) for 30 min at nificantly correlated with necrosis grade. Of these 26 genes, 9 37°C. To block endogenous peroxidases, slides were placed in a were positively correlated with necrosis grade (Table 2), humidified chamber and incubated for 12 min at room temper- whereas 17 had negative rank correlation coefficient values ature with a solution of 3% H2O2 in methanol. The slides were (Table 3). Based on their potential cellular functions, all 26 washed three times with PBS and blocked for 20 min at room genes were identified as being associated with several specific temperature in PBS supplemented with 1% normal goat serum molecular pathways (and their corresponding biological func- and 5% normal horse serum (protein-blocking solution). The tions). These pathways included those described for Ras, Akt, slides were incubated overnight at 4°C with a rabbit polyclonal tumor necrosis factor (TNF)-␣, nuclear factor (NF)-␬B, apopto- anti-EphA1 (ephrin type A receptor 1) antibody (1:200 dilution; sis, procoagulation, and hypoxia. Based on the functions known Santa Cruz Biotechnology, Santa Cruz, CA) or a rabbit poly- for each gene, Tables 2 and 3 include comments on some of the clonal anti-EP4 (prostaglandin E2 receptor EP4 subtype) anti- potential effects on the necrosis cascade of the observed up- body (1:100 dilution; Cayman Chemical, Ann Arbor, MI); for regulation or down-regulation of these genes. For example, we negative control slides, nonspecific IgG was used in place of believe that the end result of the up-regulation of ephrin type A antibody. The slides were then rinsed three times with PBS, receptor 1, which natively stimulates Akt, is increased Akt incubated for 10 min in protein-blocking solution, and incubated activity. for1hat25°C in peroxidase-conjugated antirabbit IgG solution. As indicated in Table 2, the rank correlation coefficients Next, the slides were washed and incubated in stable diamino- for the nine genes with expression positively correlated with benzidine solution (DAB; Research Genetics, Huntsville, AL). necrosis grade ranged from an extremely strong correlation of The slides were washed three times with distilled water, coun- 0.78 to a weaker but still significant association of 0.52. Paral- terstained with Gill’s hematoxylin solution (Sigma Chemical leling this, the Ps ranged from 0.004 to 0.054. Within this subset Co., St. Louis, MO), and examined with a bright-field micro- of genes, of particular interest were those for the ephrin type A scope. Reddish-brown precipitate in the cell cytoplasm indicated receptor 1 and prostaglandin E2 receptor EP4 subtype, both of a positive reaction. Statistical Analysis. To determine correlation between gene expression and the amount of necrosis present in the tumor samples, gene expression and necrosis were ranked. Necrosis Table 2 Differentially expressed genes positively correlated with increasing necrosis grade in human glioblastoma samples and the was ranked according to the previously described grading sys- implications of this for pathways in necrogenesisa tem; gene expression was ranked after the hybridization results for each array had been corrected for background signal, thresh- Rank Implications for Ͻ correlation proposed necrogenesis olded at 1 (values of 1 were set at 1), and normalized to the Gene coefficient P pathways median of that array. The values were transformed to log base Prostaglandin E2 0.78 0.0037 Increased coagulation; 10 scale for analysis. Ranking provided the advantage of ac- receptor EP4 reduced TNFb activity counting for outlying data points, thereby facilitating robust data subtype analysis. Once the respective variables were ranked, the rank Ephrin type A 0.77 0.004 Increased Akt activity; correlation coefficients were determined, with the spectrum of receptor 1 increased TNF Ϫ activity possible values ranging from 1, a perfect inverse correlation, DNA ligase IV 0.65 0.015 Decreased apoptosis to ϩ1, a perfect correlation between necrosis and gene expres- Thrombomodulin 0.63 0.019 Decreased coagulation sion; scatter plots were also created to visualize gene expres- Muscle-specific 0.61 0.023 Increased apoptosis sion-necrosis correlation. The statistical significance of the cor- DNAse I-like ␬ relation coefficients was then calculated using Ps determined by TNF superfamily 0.57 0.035 Increases in NF- B member 5 (CD40) activity, apoptosis, standard correlation coefficient analysis; statistical significance and coagulation was set at P Ͻ 0.05. Expression values for 26 genes that were Ras-related protein 0.56 0.038 Increased Ras-related significantly correlated with necrosis were compared among RAB3A activity samples of grade III necrosis GBMs, grade 0 GBMs, and AAs Estrogen 0.54 0.043 Increased TNF and NF- sulfotransferase ␬B activity using Wilcoxon’s rank-sum test. Transforming growth 0.52 0.054 Increased Ras activity The hazard ratios and Ps for survival differences between factor-␣ low (grades 0 and I) and high (grades II and III) necrosis GBM a This is based on each individual gene’s correlation with necrosis groups were computed using Cox proportional hazards regres- and known native biological function. sion analysis. b TNF, tumor necrosis factor ␣; NF, nuclear factor.

Table 3 Differentially expressed genes negatively correlated with increasing necrosis grade in human glioblastoma samples and the implications of this for pathways in necrogenesisa Rank correlation Gene coefficient P Implications for proposed necrogenesis pathways Interleukin 12 receptor Ϫ0.65 0.014 Reduced NF-␬Bb activity Interleukin 7 receptor ␣ subunit Ϫ0.64 0.016 Increased apoptosis Placenta growth factor 1 Ϫ0.64 0.016 Decreased hypoxia Cysteine protease ICE-LAP3 Ϫ0.63 0.018 Decreased apoptosis OX40 ligand Ϫ0.62 0.02 Decreased apoptosis Ϫ ␬ G1-S-specific cyclin E 0.59 0.026 Increased NF- B; increased apoptosis; reduced Ras activity Cytoplasmic nuclear factor of activated Ϫ0.59 0.026 Decreased apoptosis; reduced TNF activity T cells 1 Endothelin receptor type B Ϫ0.59 0.027 Reduced TNF activity; increased apoptosis Adenosine A3 receptor Ϫ0.58 0.029 Decreased apoptosis; increased TNF activity RNA polymerase II elongation factor SIII Ϫ0.58 0.03 Increased apoptosis p15 subunit GA-binding protein ␣ subunit Ϫ0.56 0.034 Reduced TNF activity; reduced Ras activity Insulin-like growth factor-binding protein 1 Ϫ0.55 0.039 Increased apoptosis; reduced TNF activity Tyrosine kinase receptor Tie-1 Ϫ0.55 0.039 Increased TNF activity Ski-related oncogene SnoN Ϫ0.53 0.047 Increased transforming growth factor ␤ signaling Nuclear factor I–X Ϫ0.52 0.053 Reduced Ras activity; increased TNF activity Cyclin D2 Ϫ0.52 0.053 Increased NF-␬B and hypoxia; reduced Ras activity TNF superfamily member 6 (FAS) Ϫ0.52 0.053 Decreased apoptosis a This is based on each gene’s correlation with necrosis and known native biological function. b NF, nuclear factor; TNF, tumor necrosis factor ␣.

which showed the strongest positive correlation values. This strong correlation can be appreciated on the scatter plots shown in Fig. 2. Genes for both receptors have correlation coefficients that are Ͼ0.7 and Ps of 0.004. Further statistical analysis determined the extent of correlation of patient survival with the expression of the four genes most positively correlated with necrosis grade (prostaglandin E2 receptor EP4 subtype, ephrin type A receptor 1, DNA ligase IV, and thrombomodulin). Cox proportional regression analysis revealed that combining all four genes in a single model gave an R2ˆ value of 0.53 and an overall

P of 0.012. Furthermore, combining the prostaglandin E2 receptor (EP4 subtype) and ephrin type A receptor 1 in a single model gave an R2ˆ of 0.53 and an overall P ϭ 0.0034. In contrast, Table 3 displays the gradient of rank correlation coefficient values for genes whose expression was negatively correlated with the amount of necrosis. The correlation coefficients ranged from Ϫ0.65 to Ϫ0.52, whereas Ps increased from 0.014 to 0.053. This grouping reveals the importance of the expression of genes for the receptors for interleukin 7 and 12, both of which display the strongest inverse correlation with necrosis grade. Genes for both receptors had the lowest calculated rank correlation coefficient values of approximately Ϫ0.65 and Ps close to 0.015. Similarly, the strong negative correlation coefficients can be seen in the scatter plots for the interleukin 7 and 12 receptors in the bottom panels of Fig. 2. As Table 4 indicates, there were significant differences in Fig. 2 Correlation between necrosis grade (Y axis) and expression the median expression levels of the 26 genes between the grade level (X axis) of four genes in 15 human glioblastoma samples. Top row,

III necrosis GBM and AA samples analyzed. Of the nine genes scatter plots for the prostaglandin E2 receptor EP4 subtype (left panel) positively correlated with necrosis, three genes displayed sig- and ephrin type A receptor 1 (right panel) show a strong positive nificant differences in their expression levels between the his- correlation between necrosis grade and expression level. Bottom row, scatter plots for the interleukin 7 receptor ␣ subunit (left panel) and tological subsets. Two of these, the genes for the ephrin type A interleukin 12 receptor (right panel) genes show a strong negative receptor 1 and thrombomodulin, had the most significant Psof correlation between necrosis grade and expression level. Gene Expr., 0.0159. On the other hand, only 2 of 17 genes (Table 4B) gene expression.

Table 4 Comparison of median expression levelsa of genes correlated with necrosis in human glioblastoma and anaplastic astrocytoma samplesb A. Genes positively correlated with necrosis grade Gene Grade III necrosis P AA P Grade 0 necrosis Ϫ Ϫ Ϫ Prostaglandin E2 receptor EP4 subtype 0.19 0.0317 2.83 0.5556 2.19 Ephrin type A receptor 1 Ϫ0.38 0.0159 Ϫ3.1 0.5556 Ϫ2.73 DNA ligase IV Ϫ0.08 0.0635 Ϫ0.79 0.1905 Ϫ2.73 Thrombomodulin 0.01 0.0159 Ϫ1.31 0.1905 Ϫ0.32 Muscle-specific DNAse I-like 0.08 0.7302 Ϫ0.04 0.2857 Ϫ2.86 Tumor necrosis factor superfamily member 5 (CD40) Ϫ0.08 0.9048 Ϫ0.04 0.7302 Ϫ1.68 Ras-related protein RAB3A 0.16 0.1111 0.75 0.0317 Ϫ0.19 Estrogen sulfotransferase Ϫ0.37 0.9048 Ϫ0.58 0.5556 Ϫ2.86 Transforming growth factor ␣Ϫ0.27 0.1111 Ϫ3.1 0.2857 Ϫ2.73 B. Genes negatively correlated with necrosis grade Gene Grade III necrosis P AA P Grade 0 necrosis Interleukin 12 receptor Ϫ0.55 0.5556 Ϫ0.36 0.1905 0.17 Interleukin 7 receptor ␣ subunit Ϫ3.6 0.0159 Ϫ0.37 0.9048 Ϫ0.48 Placenta growth factor 1 Ϫ3.6 0.2857 Ϫ3.19 0.0159 Ϫ1.72 Cysteine protease ICE-LAP3 Ϫ0.05 0.0635 0.31 1 0.34 OX40 ligand Ϫ3.6 0.2857 Ϫ3.19 0.0635 Ϫ1.39 Ϫ Ϫ Ϫ G1-S-specific cyclin E 3.6 0.0317 0.6 0.7302 2.73 Cytoplasmic nuclear factor of activated T cells 1 Ϫ3.5 0.7302 Ϫ3.19 0.0159 0.15 Endothelin receptor type B 0.4 0.1111 0.73 0.4127 1.01 Adenosine A3 receptor Ϫ3.6 0.2857 Ϫ3.19 0.1111 Ϫ2.73 RNA polymerase II elongation factor SIII p15 subunit 0.88 0.2857 0.97 0.4127 1.15 GA-binding protein ␣ subunit Ϫ3.6 0.1111 Ϫ3.1 0.2857 Ϫ1.65 Insulin-like growth factor binding protein 1 Ϫ3.6 0.2857 Ϫ3.19 0.1111 Ϫ1.68 Tyrosine kinase receptor Tie-1 0.06 1 0.07 0.4127 0.38 Ski-related oncogene SnoN Ϫ3.25 1 Ϫ3.19 0.0159 Ϫ1.34 Nuclear factor I–X 0.62 0.4127 0.87 0.4127 1.01 Cyclin D2 Ϫ3.25 0.4127 Ϫ3.19 0.1905 0.26 Tumor necrosis factor superfamily member 6 (FAS) Ϫ3.6 0.2857 Ϫ3.19 0.1111 Ϫ2.73 a Gene expression levels were log10-transformed from the absolute expression levels determined in the microarray analysis. Thus, an increasingly positive value correlates with a gene expression level higher than that of an extremely negative value. b Grade III necrosis glioblastoma, 4 samples; grade 0 necrosis glioblastoma, 4 samples; anaplastic astrocytoma (AA), 5 samples.

showed significant (P Ͻ 0.05) negative correlation with the molecular events in necrosis and characterized the molecular extent of necrosis. Genes for the interleukin 7 receptor ␣ subunit profiles of malignant gliomas that are associated with varying

and G1-S-specific cyclin E had Ps of 0.0159 and 0.0317, re- amounts of necrosis. spectively. Although we noted some differences in gene expres- As described in our hypothesis, we believe that the Ras and sion between GBMs of necrosis grade 0 and AAs, the signifi- Akt pathways, both of which are integral to GBM formation, cance of this finding is unclear at present. may initiate a series of events leading to necrosis (6). Ras, Immunohistochemical staining (Fig. 3) provided further through activation by various growth factors such as epidermal data corroborating the microarray results. Staining for the pros- growth factor, could lead to increases in TNF-␣ levels and the taglandin E2 receptor EP4 subtype and ephrin type A receptor 1 eventual overexpression of transcription factor NF-␬B. TNF-␣, indicated that these gene products were (a) overexpressed in along with other factors, could initiate procoagulation in the grade III necrosis GBM tumor samples, but not in those with tumor vascular network. Extensive procoagulant activity would grade 0 necrosis or in AA samples, and (b) localized in perine- then lead to hypoxia and nutrient deprivation, resulting in lethal crotic areas within the tumor samples. tumor energy decreases. In coordination with antiapoptotic mechanisms activated by Akt and NF-␬B, this could prevent the DISCUSSION completion of apoptosis induced by ligand-receptor interactions. The different forms of cell death are generally viewed as a As summarized in Fig. 4, one of the many possible mechanisms continuum in which apoptosis and necrosis comprise the two of necrosis formation could involve activation of Ras, Akt, extremes. Whereas apoptosis occurs according to a pathway of TNF-␣, NF-␬B, apoptosis, procoagulation, and hypoxia. Our caspase activation, necrosis can occur when the apoptotic path- results indicate that all 26 of the genes whose expression is way cannot be completed (1, 2). In cancer research, there has significantly correlated with necrosis have functions associated been great interest in inducing apoptosis to fight tumors; yet with the hypothesized biological factors (4, 5, 7–76). necrosis, also a form of cell death, has failed to attract similar The relevance of these genes to the formation of necrosis attention. In light of the clinical relevance of necrosis in neuro- was investigated by comparison of the gene expression profiles oncology, it is surprising to encounter so little data on the among the grade III necrosis GBM, grade 0 necrosis GBM, and subject. Thus, we have formulated a hypothesis regarding the AA subsets. By comparing gene expression between samples

Fig. 3 Immunohistochemical analysis of ephrin A receptor type 1 and prostaglandin E2 receptor EP4 subtype in anaplastic astrocytomas, grade 0 necrosis glioblastomas (GBMs), and grade III necrosis GBMs. Paraffin-embedded sections of anaplastic astrocytomas and GBMs were immunostained with antibodies against either ephrin A receptor type 1 (top row) or prostaglandin E2 receptor EP4 subtype (bottom row). Both were detected at high levels in GBMs with grade III necrosis (far right panels, ϫ200).

containing high degrees of necrosis (grade III) and tumors that One important biological pathway implicated in the necro- do not have necrosis (AA), genes that are uniquely associated sis cascade is apoptosis, in which the largest percentage of the with the formation of necrosis may be identified. Among genes 26 differentially expressed genes is involved. Our model of positively correlated with necrosis (Table 4A), all but one had necrosis predicts potentially increased stimulation of the apo- elevated expression levels in the necrosis grade III samples ptotic program, which could then be switched to a necrotic mode relative to the AA samples. This supports the observation that of death through energy deprivation and antiapoptotic factors these genes are up-regulated in highly necrotic GBM samples; (15). Results of the present study support this concept. Genes however, the difference is significant in only three of these such as adenosine A3 receptor, which inhibits apoptosis at low genes. These results may indicate that although a certain number ligand concentrations, and CD40, a member of the TNF super- of genes are increasingly transcribed in generating necrosis, family that induces apoptosis, are differentially expressed, re- only a fraction of them may be required for the process. Yet sulting in the increased induction of apoptosis (21, 27, 34, 55, necrosis formation probably results from the output of a set of 62). These changes are accompanied by the up-regulation of genes, rather than from the product of a single gene. DNase X and the down-regulation of RNA polymerase II SIII Based on their significant Ps, ephrin type A receptor 1, p15 subunit, which may induce intrinsic apoptosis through DNA prostaglandin E2 receptor EP4 subtype, and thrombomodulin degradation and defective transcription, respectively (37, 51). may be necessary in modulating Akt, TNF-␣, and procoagulant On the other hand, DNase X could potentially be one of the activity. This pattern is also present among genes whose expres- targets for apoptosis-inducing factor, which triggers nuclear sion we found to be negatively associated with necrosis grade. apoptosis after dissipation of the mitochondrial transmembrane Whereas 13 of 17 such genes showed increased expression potential, an event that is crucial to apoptosis (14). levels in the AAs relative to the necrosis grade III GBM subset, The results also show reduced expression of genes for other only 2 genes displayed significant differences in their expres- apoptosis-inducing ligand molecules, such as Fas and OX40 sion between these histological categories. The significant dif- (members of the TNF gene superfamily); interestingly, NF of ␣ ferences in interleukin 7 receptor subunit and G1-S-specific activated T cells (NF-AT), a transcription factor that induces the cyclin E gene expression indicate that these gene products may expression of Fas and OX40, is also down-regulated (52). This be necessary in the control of Ras, NF-␬B, and antiapoptotic apparent paradox could be part of a negative feedback mecha- actions. Thus, although the 26 genes we identified may have nism activated in response to the formation of necrosis. Akt, by roles in pathways resulting in necrosis, only 5 appeared signif- inhibiting glycogen synthase, prevents the phosphorylation of icantly correlated with the process. NF-AT and its translocation to the nucleus, preventing expres-

Aside from indicating a strong association between differentially expressed genes and the biological factors perceived to be crucial to necrogenesis, this study brings additional mole-

cules, such as the prostaglandin E2 receptor EP4 subtype and ephrin type A receptor 1, into consideration. Interestingly, the immunohistochemical staining experiments indicated that these gene products were not only localized to perinecrotic areas within grade III necrosis GBM tumors but were also overexpressed in such samples in comparison with either grade 0 necrosis GBMs or AAs. These data help to validate a potential connection between the above two genes and necrosis formation. The prostaglandin receptor, which had the highest positive correlation coefficient of all of the genes considered, actively modulates coagulation in addition to its capacity to inhibit TNF-␣ synthesis. The EP4 subtype is highly expressed on Fig. 4 Hypothetical pathway leading to necrosis formation in glioblas- platelets (49). A stimulator of adenyl cyclase, the EP4 receptor toma showing possible involvement of the four genes in Fig. 2. These induces coagulant activity by elevating intraplatelet cAMP lev- genes include two showing the most significant positive correlation with els (49, 67). The role of the EP4 receptor in procoagulation and increasing necrosis grade and two showing the most significant negative ␣ correlation with increasing necrosis grade. The roles suggested here for anti-TNF- activity raises an interesting question: what is the these genes are based on the correlation with necrosis observed in this primary function of the prostaglandin E2 receptor EP4 subtype study and on their known native biological functions; however, because in the appearance of necrosis? Is it mainly a mediator of the only two of these genes have been shown to be expressed around areas coagulant cascade, and although it has anti-TNF-␣ activity, are of necrosis within glioblastomas (see Fig. 3), the hypothetical nature of ␣ the gene interactions depicted must be emphasized. In this model, for there more potent pro-TNF- molecules to counterbalance its example, the end result of expression of the interleukin 7 receptor ␣ inhibitory effects? Additional studies will be required to resolve subunit, which is negatively correlated with necrosis and natively acti- these questions. vates the Akt gene, is the decreased stimulation of Akt. Arrowheads Along with the prostaglandin E2 receptor, the ephrin type A indicate activation of the responsive biological factor, whereas cross- receptor 1 and its ligand, ephrin-A1, are molecules not previ- bars (mostly vertical) perpendicular to the ends of solid lines represent inhibition. Solid lines represent increased stimulation or inhibition, ously considered as contributing to the formation of necrosis. whereas dashed lines indicate reduced activity. EPHA1, ephrin type A Identified as an early response growth factor to TNF-␣ stimu- receptor 1; IL7R, interleukin 7 receptor ␣ subunit; IL12R, interleukin 12 lation, ephrin-A1 (which, unlike other ephrins, is not normally receptor; PTGER4, prostaglandin E2 receptor EP4 subtype. found in the nervous system) is a proangiogenic factor and a chemoattractant for migrating endothelial cells (7, 16). Ephrin-A1 is transcribed at increased levels in tumor cells of sion of Fas and OX40 (22). To determine whether these changes epithelial origin, such as advanced melanomas (16). The capa- in gene expression are part of a negative feedback mechanism, bility of ephrin-A1, in contrast to other members of the ephrin future studies will have to ascertain whether these genes are family, to promote tumor cell survival and growth suggests that essential to necrosis formation. The use of pharmacological it has antiapoptotic activity. Ephrin activation of phosphatidyl- inhibitors will aid in this process. inositol 3Ј-kinase, an inducer of Akt and the resulting anti- The results also provide an interesting glimpse into the apoptotic pathways, could mediate such activity (7, 16). There- possible changes in TNF-␣ levels in regions of necrosis. Al- fore, ephrin type A receptor 1, which shows a high positive though our hybridization data indicated that TNF-␣ levels may correlation with necrosis grade, may have angiogenic and anti- not be changed, a previous immunohistochemistry study local- apoptotic roles induced by TNF-␣. ized TNF-␣ to the necrotic periphery (4). In our study, the In addition to the high correlation between necrosis grade up-regulation of genes that are promoted by or initiate TNF-␣ and the expression of the ephrin type A receptor 1 and prosta-

synthesis is observed concurrently with the marked decrease in glandin E2 EP4 receptor subtype, statistical analysis revealed expression of genes that are normally repressed by or inhibitory that when data for these two genes were combined in a single to TNF-␣, such as adenosine A3 receptor (11, 40). However, model, these genes showed a positive and statistically signifi- there are paradoxes, such as that involving the gene for the cant correlation with patient survival. Because the extent of

prostaglandin E2 receptor EP4 subtype. This receptor can inhibit necrosis is negatively correlated with patient survival, any genes TNF-␣ synthesis by interaction with prostaglandin, and expres- correlated with necrosis grade should accordingly be correlated sion of the gene is positively correlated with necrosis grade (60). with survival. Interestingly, a statistically significant positive In this study, no change in TNF-␣ gene expression was detected. correlation with survival was shown when data for the top four This may be caused by a lack of sensitivity of nylon microarray genes positively correlated with necrosis grade (prostaglandin

analysis. Further investigation of this issue will require Western E2 receptor EP4 subtype, ephrin type A receptor 1, DNA ligase blot analysis and more immunohistochemical studies. Neverthe- IV, and thrombomodulin) were combined into a single model, less, the results do suggest that other members of the TNF gene although individually these genes did not show a significant superfamily could play equally important roles in inducing correlation with survival. This suggests an association among procoagulation and apoptosis. these genes in necrosis formation and indicates that it is an

intricate tapestry of gene products, rather than the action of a 4. Roessler, K., Suchanek, G., Breitschopf, H., Kitz, K., Matula, C., single gene, that leads to necrosis in GBMs. Lassmann, H., and Koos, W. T. Detection of tumor necrosis factor ␣ There are several limitations to the present study. First is protein and messenger RNA in human glial brain tumors: comparison of immunohistochemistry with in situ hybridization using molecular the small sample size of 15 GBMs; for more definitive results, probes. J. Neurosurg., 83: 291–297, 1995. future studies will have to examine a larger population of these 5. Fuller, G. N., Rhee, C. H., Hess, K. R., Caskey, L. S., Wang, R., tumors. Moreover, because the genes identified in this study Bruner, J. M., Yung, W. K., and Zhang, W. Reactivation of insulin-like have many functions, it is possible that the differential expres- growth factor binding protein 2 expression in glioblastoma multiforme: sion we monitored in some of them is related to processes other a revelation by parallel gene expression profiling. Cancer Res., 59: than necrosis that also correlate with GBM severity. Neverthe- 4228–4232, 1999. less, it is impressive that such a strong correlation between 6. Raza, S. M., Lang, F. F., Aggarwal, B. B., Fuller, G. N., Wildrick, necrosis grade and the expression of several genes was obtained D. M., and Sawaya, R. Necrosis and glioblastoma: a friend or a foe? A review and a hypothesis. Neurosurgery (Baltimore), 51: 2–13, 2002. in this study. This is evident in the scatter plots of Fig. 2 that 7. Adams, R. H., and Klein, R. Eph receptors and ephrin ligands. show clustering of gene expression values within specific ne- essential mediators of vascular development. Trends Cardiovasc. Med., crosis grades. Additional limitations are the relative inability of 10: 183–188, 2000. nylon microarray analysis to detect changes in genes with low 8. Ahmed, A., Dunk, C., Ahmad, S., and Khaliq, A. Regulation of expression and the possibility of false negatives and positives. placental vascular endothelial growth factor (VEGF) and placenta With regard to the latter, complementary and confirmatory growth factor (PIGF) and soluble Flt-1 by oxygen: a review. Placenta, techniques such as Western blotting and/or immunohistochem- 21 (Suppl. A): S16–S24, 2000. istry are always required to confirm and extend the findings of 9. Airoldi, I., Gri, G., Marshall, J. D., Corcione, A., Facchetti, P., gene expression microarray analysis, as demonstrated in this Guglielmino, R., Trinchieri, G., and Pistoia, V. Expression and function of IL-12 and IL-18 receptors on human tonsillar B cells. J. Immunol., study in the case of the ephrin type A receptor 1 and the 165: 6880–6888, 2000. prostaglandin E2 receptor. 10. Akiyoshi, S., Inoue, H., Hanai, J., Kusanagi, K., Nemoto, N., Our initial cDNA microarray analysis of 15 GBMs, in Miyazono, K., and Kawabata, M. c-Ski acts as a transcriptional co- parallel with 5 AAs, has provided a glimpse of molecular events repressor in transforming growth factor-␤ signaling through interaction in the formation of necrosis in the highest grade of astrocytoma. with smads. J. Biol. Chem., 274: 35269–35277, 1999. The results indicated that Ras, Akt, TNF-␣, NF-␬B, apoptosis, 11. Bowlin, T. L., Borcherding, D. R., Edwards, C. K., III, and Mc- procoagulation, and hypoxia may be important to necrogenesis Whinney, C. D. Adenosine A3 receptor agonists inhibit murine macrophage tumor necrosis factor ␣ production in vitro and in vivo. Cell. Mol. because all of the 26 differentially expressed genes identified Biol. (Noisy-Le-Grand), 43: 345–349, 1997. had functions associated with these factors. Of the 26 genes that 12. Choi, K. C., Kang, S. K., Tai, C. J., Auersperg, N., and Leung, P. C. may play a role in necrogenesis, only the ephrin type A receptor Estradiol up-regulates antiapoptotic Bcl-2 messenger ribonucleic acid ␣ 1, prostaglandin E2 receptor EP4, interleukin 7 receptor sub- and protein in tumorigenic ovarian surface epithelium cells. Endocri- unit, G1-S-specific cyclin E, and thrombomodulin gene products nology, 142: 2351–2360, 2001. showed a significant correlation with the process. The data also 13. Dadi, H., Ke, S., and Roifman, C. M. Activation of phosphatidyli- revealed that necrosis formation may be dependent on a delicate nositol-3 kinase by ligation of the interleukin-7 receptor is dependent on balance between pro- and antiapoptotic activities. Associated protein tyrosine kinase activity. Blood, 84: 1579–1586, 1994. with this revelation was the identification of genes not previ- 14. Daugas, E., Nochy, D., Ravagnan, L., Loeffler, M., Susin, S. A., Zamzami, N., and Kroemer, G. Apoptosis-inducing factor (AIF): a ously implicated in necrosis as possible active participants in the ubiquitous mitochondrial oxidoreductase involved in apoptosis. FEBS necrosis cascade. Together with these conclusions, it is impor- Lett., 476: 118–123, 2000. tant to keep in mind the possibility of negative feedback mech- 15. Denecker, G., Vercammen, D., Declercq, W., and Vandenabeele, P. anisms. This study highlights the concept that end biological Apoptotic and necrotic cell death induced by death domain receptors. events, such as necrogenesis, are determined by the balance of Cell. Mol. Life Sci., 58: 356–370, 2001. pro- and counter-regulatory mechanisms. Future studies will be 16. Easty, D. J., Hill, S. P., Hsu, M. Y., Fallowfield, M. E., Florenes, V. A., Herlyn, M., and Bennett, D. C. Up-regulation of ephrin-A1 aimed at confirming the findings of the present study through during melanoma progression. Int. J. Cancer, 84: 494–501, 1999. complementary methodologies and extending the investigation 17. Eberl, L. P., Egidy, G., Pinet, F., and Juillerat-Jeanneret, L. Endo- of necrogenesis with more comprehensive gene expression thelin receptor blockade potentiates FasL-induced apoptosis in colon analyses. carcinoma cells via the protein kinase C-pathway. J. Cardiovasc. Phar- macol., 36: S354–S356, 2000. 18. Falany, J. L., and Falany, C. N. Regulation of estrogen activity by ACKNOWLEDGMENTS sulfation in human MCF-7 breast cancer cells. Oncol. Res., 9: 589–596, We thank Dr. David M. Wildrick for editorial assistance in pre- 1997. paring the manuscript. 19. Flory, E., Hoffmeyer, A., Smola, U., Rapp, U. R., and Bruder, J. T. Raf-1 kinase targets GA-binding protein in transcriptional regulation of the human immunodeficiency virus type 1 promoter. J. Virol., 70: REFERENCES 2260–2268, 1996. 1. Farber, E. Programmed cell death: necrosis versus apoptosis. Mod. 20. Fowler, D. J., Nicolaides, K. H., and Miell, J. P. Insulin-like growth Pathol., 7: 605–609, 1994. factor binding protein-1 (IGFBP-1): a multifunctional role in the human 2. Kerr, J. F. Shrinkage necrosis: a distinct mode of cellular death. female reproductive tract. Hum. Reprod. Update, 6: 495–504, 2000. J. Pathol., 105: 13–20, 1971. 21. Gao, Z., Li, B. S., Day, Y. J., and Linden, J. A3 adenosine receptor 3. Hammoud, M. A., Sawaya, R., Shi, W., Thall, P. F., and Leeds, N. E. activation triggers phosphorylation of protein kinase B and protects rat Prognostic significance of preoperative MRI scans in glioblastoma basophilic leukemia 2H3 mast cells from apoptosis. Mol. Pharmacol., multiforme. J. Neuro-Oncol., 27: 65–73, 1996. 59: 76–82, 2001.

22. Gold, M. R., Ingham, R. J., McLeod, S. J., Christian, S. L., Scheid, induces the expression of vascular endothelial growth factor by endo- M. P., Duronio, V., Santos, L., and Matsuuchi, L. Targets of B-cell thelial cells and monocytes and promotes angiogenesis in vivo. Blood, antigen receptor signaling: the phosphatidylinositol 3-kinase/Akt/glyco- 96: 3801–3808, 2000. gen synthase kinase-3 signaling pathway and the Rap1 GTPase. Immu- 42. Menouny, M., Binoux, M., and Babajko, S. IGFBP-2 expression in nol. Rev., 176: 47–68, 2000. a human cell line is associated with increased IGFBP-3 proteolysis, 23. Greene, R. A. Estrogen and cerebral blood flow: a mechanism to decreased IGFBP-1 expression and increased tumorigenicity. Int. J. explain the impact of estrogen on the incidence and treatment of Alz- Cancer, 77: 874–879, 1998. heimer’s disease. Int. J. Fertil. Women’s Med., 45: 253–257, 2000. 43. Morawietz, H., Talanow, R., Szibor, M., Rueckschloss, U., Schu- 24. Gruss, H. J. Molecular, structural, and biological characteristics of bert, A., Bartling, B., Darmer, D., and Holtz, J. Regulation of the the tumor necrosis factor ligand superfamily. Int. J. Clin. Lab. Res., 26: endothelin system by shear stress in human endothelial cells. J. Physiol. 143–159, 1996. (Cambridge), 525: 761–770, 2000. 25. Hofmeister, R., Khaled, A. R., Benbernou, N., Rajnavolgyi, E., 44. Morga, E., Faber, C., and Heuschling, P. Stimulation of endothelin Muegge, K., and Durum, S. K. Interleukin-7: physiological roles and B receptor modulates the inflammatory activation of rat astrocytes. mechanisms of action. Cytokine Growth Factor Rev., 10: 41–60, 1999. J. Neurochem., 74: 603–612, 2000. 26. Holland, E. C., Celestino, J., Dai, C., Schaefer, L., Sawaya, R. E., 45. Mosnier, L. O., Meijers, J. C., and Bouma, B. N. Regulation of and Fuller, G. N. Combined activation of Ras and Akt in neural fibrinolysis in plasma by TAFI and protein C is dependent on the progenitors induces glioblastoma formation in mice. Nat. Genet., 25: concentration of thrombomodulin. Thromb. Haemostasis, 85: 5–11, 55–57, 2000. 2001. 27. Jacobson, K. A. Adenosine A3 receptors: novel ligands and para- 46. Naldini, A., and Carraro, F. Hypoxia modulates cyclin and cytokine doxical effects. Trends Pharmacol. Sci., 19: 184–191, 1998. expression and inhibits peripheral mononuclear cell proliferation. 28. Kaltschmidt, B., Kaltschmidt, C., Hehner, S. P., Droge, W., and J. Cell. Physiol., 181: 448–454, 1999. ␬ Schmitz, M. L. Repression of NF- B impairs HeLa cell proliferation by 47. Ogawa, K., Pasqualini, R., Lindberg, R. A., Kain, R., Freeman, functional interference with cell cycle checkpoint regulators. Oncogene, A. L., and Pasquale, E. B. The ephrin-A1 ligand and its receptor, 18: 3213–3225, 1999. EphA2, are expressed during tumor neovascularization. Oncogene, 19: 29. Kast, R. E. Tumor necrosis factor has positive and negative self 6043–6052, 2000. regulatory feed back cycles centered around cAMP. Int. J. Immuno- 48. Ohana, G., Bar-Yehuda, S., Barer, F., and Fishman, P. Differential pharmacol., 22: 1001–1006, 2000. effect of adenosine on tumor and normal cell growth: focus on the A3 30. Kitaura, J., Asai, K., Maeda-Yamamoto, M., Kawakami, Y., adenosine receptor. J. Cell. Physiol., 186: 19–23, 2001. Kikkawa, U., and Kawakami, T. Akt-dependent cytokine production in 49. Paul, B. Z., Ashby, B., and Sheth, S. B. Distribution of prosta- mast cells. J. Exp. Med., 192: 729–740, 2000. glandin IP and EP receptor subtypes and isoforms in platelets and 31. Klaassen, C. D., Liu, L., and Dunn, R. T., II. Regulation of sulfo- human umbilical artery smooth muscle cells. Br. J. Haematol., 102: transferase mRNA expression in male and female rats of various ages. 1204–1211, 1998. Chem. Biol. Interact., 109: 299–313, 1998. 50. Perkins, N. D., Felzien, L. K., Betts, J. C., Leung, K., Beach, D. H., 32. Kondo, N., Matsui, E., Kaneko, H., Fukao, T., Teramoto, T., Inoue, and Nabel, G. J. Regulation of NF-␬B by cyclin-dependent kinases R., Watanabe, M., Kasahara, K., and Morimoto, N. Reduced interfer- associated with the p300 coactivator. Science (Wash. DC), 275: 523– ␥ ␤ on- production and mutations of the interleukin-12 receptor 2 chain 527, 1997. gene in atopic subjects. Int. Arch. Allergy Immunol., 124: 117–120, 2001. 51. Proietti De Santis, L., Garcia, C. L., Balajee, A. S., Brea Calvo, G. T., Bassi, L., and Palitti, F. Transcription coupled repair deficiency 33. Kotov, A., Falany, J. L., Wang, J., and Falany, C. N. Regulation of results in increased chromosomal aberrations and apoptotic death in the estrogen activity by sulfation in human Ishikawa endometrial adenocar- UV61 cell line, the Chinese hamster homologue of Cockayne’s syn- cinoma cells. J. Steroid Biochem. Mol. Biol., 68: 137–144, 1999. drome B. Mutat. Res., 485: 121–132, 2001. 34. Lazaar, A. L., Amrani, Y., Hsu, J., Panettieri, R. A., Jr., Fanslow, 52. Rivera, I., Harhaj, E. W., and Sun, S. C. Involvement of NF-AT in W. C., Albelda, S. M., and Pure, E. CD40-mediated signal transduction in human airway smooth muscle. J. Immunol., 161: 3120–3127, 1998. type I human T-cell leukemia virus Tax-mediated Fas ligand promoter transactivation. J. Biol. Chem., 273: 22382–22388, 1998. 35. Lin, C. G., Lin, Y. C., Liu, H. W., and Kao, L. S. Characterization of Rab3A, Rab3B and Rab3C: different biochemical properties and 53. Rodriguez, A. M., Rodin, D., Nomura, H., Morton, C. C., Wer- intracellular localization in bovine chromaffin cells. J. Biochem. (To- emowicz, S., and Schneider, M. C. Identification, localization, and kyo), 324: 85–90, 1997. expression of two novel human genes similar to deoxyribonuclease I. Genomics, 42: 507–513, 1997. 36. Lin, T. N., Wang, C. K., Cheung, W. M., and Hsu, C. Y. Induction of angiopoietin and Tie receptor mRNA expression after cerebral ische- 54. Roifman, C. M., Zhang, J., Chitayat, D., and Sharfe, N. A partial mia-reperfusion. J. Cereb. Blood Flow Metab., 20: 387–395, 2000. deficiency of interleukin-7R ␣ is sufficient to abrogate T-cell develop- 37. Los, M., Neubuser, D., Coy, J. F., Mozoluk, M., Poustka, A., and ment and cause severe combined immunodeficiency. Blood, 96: 2803– Schulze-Osthoff, K. Functional characterization of DNase X, a novel 2807, 2000. endonuclease expressed in muscle cells. Biochemistry, 39: 7365–7373, 55. Rothstein, T. L. Inducible resistance to Fas-mediated apoptosis in B 2000. cells. Cell Res., 10: 245–266, 2000. 38. Ludviksson, B. R., Seegers, D., Resnick, A. S., and Strober, W. The 56. Salatino, M., Labriola, L., Schillaci, R., Charreau, E. H., and effect of TGF-␤1 on immune responses of naive versus memory CD4ϩ Elizalde, P. V. Mechanisms of cell cycle arrest in response to TGF-␤ in Th1/Th2 T cells. Eur. J. Immunol., 30: 2101–2111, 2000. progestin-dependent and -independent growth of mammary tumors. 39. McKiernan, C. J., Stabila, P. F., and Macara, I. G. Role of the Exp. Cell Res., 265: 152–166, 2001. Rab3A-binding domain in targeting of rabphilin-3A to vesicle mem- 57. Salvatore, C. A., Tilley, S. L., Latour, A. M., Fletcher, D. S., Koller, branes of PC12 cells. Mol. Cell. Biol., 16: 4985–4995, 1996. B. H., and Jacobson, M. A. Disruption of the A(3) adenosine receptor 40. McWhinney, C. D., Dudley, M. W., Bowlin, T. L., Peet, N. P., gene in mice and its effect on stimulated inflammatory cells. J. Biol. Schook, L., Bradshaw, M., De, M., Borcherding, D. R., and Edwards, Chem., 275: 4429–4434, 2000. C. K., III. Activation of adenosine A3 receptors on macrophages inhibits 58. Schonbeck, U., Mach, F., Sukhova, G. K., Herman, M., Graber, P., tumor necrosis factor ␣. Eur. Pharmacol., 310: 209–216, 1996. Kehry, M. R., and Libby, P. CD40 ligation induces tissue factor expres- 41. Melter, M., Reinders, M. E., Sho, M., Pal, S., Geehan, C., Denton, sion in human vascular smooth muscle cells. Am. J. Pathol., 156: 7–14, M. D., Mukhopadhyay, D., and Briscoe, D. M. Ligation of CD40 2000.

59. Sharma, V. K., Bologa, R. M., Li, B., Xu, G. P., Lagman, M., by a cyclic adenosine monophosphate-dependent pathway. Surgery (St. Hiscock, W., Mouradian, J., Wang, J., Serur, D., Rao, V. K., and Louis), 114: 314–323, 1993. Suthanthiran, M. Molecular executors of cell death: differential intrare- 68. Xu, H., Zhang, G. X., Wysocka, M., Wu, C. Y., Trinchieri, G., and nal expression of Fas ligand, Fas, granzyme B, and perforin during acute Rostami, A. The suppressive effect of TGF-␤ on IL-12-mediated im- and/or chronic rejection of human renal allografts. Transplantation (Bal- mune modulation specific to a peptide Ac1–11 of myelin basic protein timore), 62: 1860–1866, 1996. (MBP): a mechanism involved in inhibition of both IL-12 receptor ␤1 60. Shinomiya, S., Naraba, H., Ueno, A., Utsunomiya, I., Maruyama, and ␤2. J. Neuroimmunol., 108: 53–63, 2000. T., Ohuchida, S., Ushikubi, F., Yuki, K., Narumiya, S., Sugimoto, Y., Ichikawa, A., and Oh-ishi, S. Regulation of TNF␣ and interleukin-10 69. Yabkowitz, R., Meyer, S., Black, T., Elliott, G., Merewether, L. A., and Yamane, H. K. Inflammatory cytokines and vascular endothelial production by prostaglandins I2 and E2: studies with prostaglandin receptor-deficient mice and prostaglandin E-receptor subtype-selective growth factor stimulate the release of soluble tie receptor from human synthetic agonists. Biochem. Pharmacol., 61: 1153–1160, 2001. endothelial cells via metalloprotease activation. Blood, 93: 1969–1979, 61. Shirakawa, K., Tsuda, H., Heike, Y., Kato, K., Asada, R., Inomata, 1999. M., Sasaki, H., Kasumi, F., Yoshimoto, M., Iwanaga, T., Konishi, F., 70. Yue, Y., Su, J., Cerretti, D. P., Fox, G. M., Jing, S., and Zhou, R. Terada, M., and Wakasugi, H. Absence of endothelial cells, central Selective inhibition of spinal cord neurite outgrowth and cell survival by necrosis, and fibrosis are associated with aggressive inflammatory the Eph family ligand ephrin-A5. J. Neurosci., 19: 10026–10035, 1999. breast cancer. Cancer Res., 61: 445–451, 2001. 71. Zhou, R. The Eph family receptors and ligands. Pharmacol. Ther., 62. Shneyvays, V., Jacobson, K. A., Li, A. H., Nawrath, H., Zinman, T., 77: 151–181, 1998. Isaac, A., and Shainberg, A. Induction of apoptosis in rat cardiocytes by 72. Ziche, M., Maglione, D., Ribatti, D., Morbidelli, L., Lago, C. T., A3 adenosine receptor activation and its suppression by isoproterenol. Battisti, M., Paoletti, I., Barra, A., Tucci, M., Parise, G., Vincenti, V., Exp. Cell Res., 257: 111–126, 2000. Granger, H. J., Viglietto, G., and Persico, M. G. Placenta growth 63. Simpson, R. B., Chase, C. C., Jr., Spicer, L. J., Carroll, J. A., factor-1 is chemotactic, mitogenic, and angiogenic. Lab. Investig., 76: Hammond, A. C., and Welsh, T. H., Jr. Effect of exogenous estradiol on 517–531, 1997. plasma concentrations of somatotropin, insulin-like growth factor-I, insulin-like growth factor binding protein activity, and metabolites in 73. Chang, F., Steelman, L. S., and McCubrey, J. A. Raf-induced cell ovariectomized Angus and Brahman cows. Domest. Anim. Endocrinol., cycle progression in human TF-1 hematopoietic cells. Cell Cycle, 1: 14: 367–380, 1997. 220–226, 2002. 64. Sokolovsky, M., Ambar, I., and Galron, R. A novel subtype of 74. Green, C. J., Lichtlen, P., Huynh, N. T., Yanovsky, M., Laderoute, endothelin receptors. J. Biol. Chem., 267: 20551–20554, 1992. K. R., Schaffner, W., and Murphy, B. J. Placenta growth factor gene 65. Tilley, S. L., Audoly, L. P., Hicks, E. H., Kim, H. S., Flannery, P. J., expression is induced by hypoxia in fibroblasts: a central role for metal Coffman, T. M., and Koller, B. H. Reproductive failure and reduced transcription factor-1. Cancer Res., 61: 2696–2703, 2001. blood pressure in mice lacking the EP2 prostaglandin E2 receptor. 75. Nomura, M., Yamagishi, S., Harada, S., Yamashima, T., Ya- J. Clin. Investig., 103: 1539–1545, 1999. mashita, J., and Yamamoto, H. Placenta growth factor (PlGF) mRNA 66. Venkitaraman, A. R., and Cowling, R. J. Interleukin-7 induces the expression in brain tumors. J. Neuro-Oncol., 40: 123–130, 1998. association of phosphatidylinositol 3-kinase with the ␣ chain of the 76. Shinagawa, T., Dong, H. D., Xu, M., Maekawa, T., and Ishii, S. The interleukin-7 receptor. Eur. J. Immunol., 24: 2168–2174, 1994. sno gene, which encodes a component of the histone deacetylase com-

67. Williams, J. G., Garcia, I., and Maier, R. V. Prostaglandin E2 plex, acts as a tumor suppressor in mice. EMBO J., 19: 2280–2291, mediates lipopolysaccharide-induced macrophage procoagulant activity 2000.

Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2004 American Association for Cancer Research. Identification of Necrosis-Associated Genes in Glioblastoma by cDNA Microarray Analysis

Shaan M. Raza, Gregory N. Fuller, Chang Hun Rhee, et al.

Clin Cancer Res 2004;10:212-221.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/10/1/212

Cited articles This article cites 76 articles, 20 of which you can access for free at: http://clincancerres.aacrjournals.org/content/10/1/212.full#ref-list-1

Citing articles This article has been cited by 6 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/10/1/212.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/10/1/212. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.