Gliomas are the most common primary tumors of

Neuro-Oncology Advance Access published January 7, 2010 Neuro-Oncology doi:10.1093/neuonc/nop025 N E U RO - O N CO LO GY Differential proteome analy...
8 downloads 0 Views 622KB Size
Neuro-Oncology Advance Access published January 7, 2010 Neuro-Oncology doi:10.1093/neuonc/nop025

N E U RO - O N CO LO GY

Differential proteome analysis of human gliomas stratified for loss of heterozygosity on chromosomal arms 1p and 19q Michael Grzendowski, Marietta Wolter, Markus J. Riemenschneider, Christina B. Knobbe, Uwe Schlegel, Helmut E. Meyer, Guido Reifenberger, and Kai Stu¨hler Medizinisches Proteom-Center, Ruhr-Universita¨t Bochum, Bochum (M.G., H.E.M., K.S.); Institut fu¨r Neuropathologie, Heinrich-Heine-Universita¨t Du¨sseldorf, Du¨sseldorf (M.W., M.J.R., C.B.K., G.R.); Neurologische Universita¨tsklinik, Ruhr-Universita¨t Bochum, Knappschaftskrankenhaus, Bochum, Germany (U.S.)

Combined deletion of chromosomal arms 1p and 19q is an independent prognostic marker in patients with oligodendroglial brain tumors, including oligodendrogliomas and oligoastrocytomas. However, the relevant genes in these chromosome arms and the molecular mechanisms underlying the prognostic significance of 1p/19q deletion are yet unknown. We used two-dimensional difference gel electrophoresis followed by mass spectrometry to perform a proteome-wide profiling of low-grade oligoastrocytomas stratified for the presence or absence of 1p/19q deletions. Thereby, we identified 22 different proteins showing differential expression in tumors with or without combined deletions of 1p and 19q. Four of the differentially expressed proteins, which are vimentin, villin 2 (ezrin), annexin A1, and glial fibrillary acidic protein, were selected for further analysis. Lower relative expression levels of these proteins in 1p/19q-deleted gliomas were confirmed at the protein level by Western blot analysis and immunohistochemistry. Furthermore, sequencing of sodium bisulfite– treated tumor DNA revealed more frequent methylation of 50 -CpG islands associated with the VIM and VIL2 genes in 1p/19q-deleted gliomas when compared with gliomas without these deletions. In summary, we confirm proteome-wide profiling as a powerful means to identify candidate biomarkers in gliomas. In addition, our data support the hypothesis that 1p/19q-deleted gliomas frequently show epigenetic down-regulation of multiple genes due to aberrant methylation of the 50 -CpG islands.

Received November 25, 2008; accepted April 29, 2009. Corresponding Author: Kai Stu¨hler, PhD, Medizinisches ProteomCenter, Ruhr-Universita¨t Bochum, Universita¨tsstraße 150, 44801 Bochum, Germany ([email protected]).

Keywords: 2D-DIGE, glioma, 1p/19q LOH, loss of heterozygosity, proteomics

G

liomas are the most common primary tumors of the central nervous system with an estimated annual incidence of 5– 6 patients per 100 000 population. At present, classification and grading of the diverse types of gliomas are primarily based on morphological and immunohistochemical features as defined in the World Health Organization (WHO) classification of tumors of the central nervous system.1 Thus far, only a few molecular parameters have gained clinical significance as prognostic or predictive markers in the diagnostic assessment of gliomas. In glioblastomas, hypermethylation of the O 6-methylguanine-methyltransferase (MGMT) gene promoter is a predictive molecular marker for response to alkylating chemotherapy and longer survival.2 In oligodendroglial tumors, combined deletion of chromosomal arms 1p and 19q, which is found in up to 80% of oligodendrogliomas and in about 50% of oligoastrocytomas, is associated with favorable response to radio- and chemotherapy as well as prolonged survival.3 – 6 Diagnostic testing for these molecular markers is now frequently performed, because they provide clinically valuable information beyond the conventional histological assessment.7 The relevant genes and proteins conveying the favorable clinical effects associated with losses of 1p and 19q are still unknown. Molecular genetic approaches have mainly focused on the mapping of minimal regions of deletion on 1p or 19q and on the molecular analysis of potential tumor suppressor genes within these regions. Thereby, several candidate tumor suppressor genes have been identified, such as TP73 (1p36.3),8 CAMTA1 (1p36),9 DFFB (1p36),10

# The Author(s) 2010. Published by Oxford University Press on behalf of the Society for Neuro-Oncology. All rights reserved. For permissions, please e-mail: [email protected].

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

SHREW1 (1p36.32),11 CITED4 (1p34.2),12 CDKN2C (1p32),13 and DIRAS3 (1p31),14 on 1p, as well as p190RhoGAP (19q13.3)15 and EMP3 (19q13.3)16 on 19q. In the present study, we used a proteome-based approach involving two-dimensional difference gel electrophoresis (2D-DIGE)17 combined with MALDI-TOF/ TOF mass spectrometry, and focused on the identification of proteins that were able to stratify gliomas according to the 1p/19q allelic status. This approach allowed for the quantitative analysis of approximately 3500 protein spots. Differential expression of selected candidate proteins was validated by Western blotting and immunohistochemistry in an extended series of primary gliomas, and their utility as potential immunohistochemical markers was assessed by receiver operation characteristic (ROC) curves. As epigenetic mechanisms have been shown to frequently cause loss of gene expression in 1p/19q-deleted gliomas,12,14,16 we additionally assessed selected candidate genes for aberrant methylation of 50 -CpG islands.

Materials and Methods Patient Samples Frozen glioma tissue samples were selected from the brain tumor tissue bank at the Department of Neuropathology, Heinrich-Heine-Universita¨t, Du¨sseldorf, Germany, and used in an anonymous manner as approved by the institutional review board. All tumors were classified according to the criteria of the WHO classification of tumors of the central nervous system.1 Tissue samples of each tumor were snap-frozen immediately after operation and stored at 2808C. To ensure that tumor fragments, which were to be used for molecular genetic and proteomic analyses, contained a sufficient proportion (.80%) of tumor cells, we histologically evaluated each specimen. The initial proteomic profiling using the 2D-DIGE approach was carried out on protein extracts from 9 primary oligoastrocytomas of WHO grade II, including 4 tumors with and 5 tumors without combined losses of 1p and 19q. Validation of candidate proteins was performed on an extended series of primary gliomas. In total, we investigated gliomas from 62 different patients (Supplementary Material, Table S1). Extraction of Nucleic Acids High –molecular weight DNA and RNA from unfixed frozen tissue samples were extracted using ultracentrifugation over cesium chloride as described elsewhere.18 DNA extraction from peripheral blood leukocytes was performed according to the standard protocol. Determination of Allelic Losses on Chromosomal Arms 1p and 19q All 62 tumor samples included in this study were investigated for 1p and 19q allelic losses using loss of

2

NEURO-ONCOLOGY

heterozygosity (LOH) analyses at up to 30 different microsatellite markers located along the short arm of chromosome 1, and 5 different microsatellite markers located along the long arm of chromosome 19, as reported in detail elsewhere.19 Tumors were classified only as co-deleted on 1p and 19q in case all investigated informative markers on both chromosome arms exhibited LOH. We compared microsatellite LOH analysis with array-CGH findings in a previous publication and found concordant results for both methods in a series of 20 oligodendroglial tumors.20 Sample Preparation and Protein Labeling for 2D-DIGE For cell lysis, 50 mg of solid tumor tissue was mixed with 70 mL lysis buffer (Tris–HCl 30 mM; thiourea 2 M; urea 7 M; CHAPS 4%, pH 8.5). Cell lysis was completed by subsequent sonification (6  10 s pulses on ice). Cell debris was removed by centrifugation at 12 000g for 15 min. The samples remained on ice during the entire process. Protein concentration was determined using the Bio-Rad Protein AssayTM (Bio-Rad, Munich, Germany). Stock solutions of fluorescent dyes (1 nmol/mL) were diluted with anhydrous DMF p.a. (400 pmol/mL; Sigma, Seelze, Germany) and added to the tissue lysates (400 pmol dye per 50 mg protein). The tumor samples with 1p/19q deletion were labeled with Cy3 and the samples without 1p/19q deletion with Cy5. An internal standard composed of equal amounts of all samples included in the analysis was then labeled with Cy2. Each sample was vortexed, centrifuged briefly, and left on ice for 30 min in the dark. The reaction was stopped by adding 1 mL of 10 mM L-lysine (Sigma) per 400 pmol dye. After further vortexing and centrifugation, the sample again was left on ice for 10 min in the dark. Then, DTT solution (1.08 g/mL; Bio-Rad) and AmpholineTM 2–4 (GE Healthcare, Munich) were added to a final percentage of 10. For protein identification by MALDI-mass spectrometry (MALDI-MS), we ran a preparative gel with 400 mg protein. 2D-gels were subsequently silver stained21 and dried for documentation purposes. Large 2D-gel Electrophoresis Carrier ampholyte-based isoelectric focusing (IEF) was performed using self-casted tube gels of 40 cm in length and 1.5 mm in diameter, as described by Klose22 with minor modifications. Briefly, the tubes were placed into in-house – produced IEF chambers and a voltage gradient was applied as described elsewhere.23 After IEF, the tube gels were ejected and incubated for 10 min in equilibration buffer (125 mM Tris, 40% [w/ v] glycerol, 3% [w/v] SDS, 65 mM DTT, pH 6.8). For the 2D we built an electrophoresis chamber allowing us to run complete 2D gels (40  30 cm) without the need to cut the IEF gel.24 For the 2D, we used self-casted large (40  30 cm) 2D-gels (SDS-PAGE, 15.2% total acrylamide, 1.3% bisacrylamide; Serva, Heidelberg, Germany). Immediately after incubation, the tube gels

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

were placed onto the 2D-gels and fixed with agarose (1%, w/v) containing 0.01% (w/v) bromophenol blue dye (Merck, Darmstadt, Germany).

Scanning and Image Analysis By labeling with fluorescence dye protein, the processing of large gels could be realized.24 SDS-gels remained between the glass plates while image acquisition was done on a Typhoon Trio Fluorescence Imager (GE Healthcare). Excitation and emission wavelengths were chosen specifically for each of the dyes according to recommendations of the manufacturer. Images were preprocessed using ImageQuant TL software (Version 2003.02, GE Healthcare). Intra-gel spot detection and inter-gel matching were performed using the Differential In-gel Analysis (DIA) mode and Biological Variation Analysis (BVA) mode of DeCyder 6.0 software (GE Healthcare), respectively. For standardization and normalization of spot intensities, the internal standard was used. Differentially regulated proteins were determined by Student’s t-test, and all proteins with a P value .05 and fold change 1.8 were chosen for identification by MALDI-MS.

Protein Identification by MALDI-TOF/TOF Mass Spectrometry The proteins of interest were manually cut out from the preparative gel, washed, in-gel digested with trypsin (Promega, Madison, WI), and extracted for mass spectrometry.25 The tryptic peptides of each protein were analyzed by MALDI-TOF MS using UltraFlex II (Bruker Daltonics, Bremen). We used a-cyano-4hydroxycynamic acid as a matrix and applied MALDI AnchorChip targets (Bruker) according to the manufacturer’s instructions. Spectra were acquired in the positive mode with a target voltage of 20 kV and a pulsed ion extraction of 17.25 kV. The reflector voltage was set to 21 kV and detector voltage to 1.7 kV. The calibration of PMF spectra was performed using a peptide mix (monoisotopic masses: bradykinin(1-7), m/z 757.399; angiotensin I, m/z 1296.685; bombesin, m/z 1619.822; ACTH clip(1-17), 2093.086 m/z; and somatostatin, 3147.471 m/z). The PMF spectra were processed using FlexAnalysisTM software (version 2.4, Bruker). Uninterpreted spectra were correlated with the human IPI database (Version v3.15) using the Mascot (Matrix Science, London) algorithms within ProteinScape (Bruker). Database searches were performed using the parameters noted below: fixed cysteine modification with propionamide, variable modification due to methionine oxidation, one missed cleavage site in case of incomplete trypsin hydrolysis, and a mass tolerance of 50 ppm and without definition of details about 2D-PAGE –derived protein mass and isoelectric point (pI). Identified proteins were accepted as correct if they showed a Mascot score higher than 67.

Western Blot Analysis For the validation of selected candidate proteins by Western blot analysis, we extended the number of samples using protein fractions obtained by ultracentrifugation over cesium chloride (ie, the method originally used for extraction of high – molecular weight nucleic acids from frozen tissue samples).18 Excess of guanidine hydrochloride salt load in these protein fractions was removed by replacing solvent with sample buffer using ultrafiltration applying Microcons (Millipore, Schwalbach, 3 kDa cut-off). The proteins were then denatured and separated by gel electrophoresis on precasted NuPAGE Novex 4– 12% Bis– Tris Midi gels (Invitrogen, Karlsruhe). Subsequently, the proteins were transferred to Amersham HybondTM -PVDF membrane (GE Healthcare) using a NovaBlot blot module (GE Healthcare). Membranes were blocked with 3% (w/v) nonfat dried milk, 1% BSA, and 0.1% (v/v) Tween 20 in Tris-buffered saline, pH 8.5, for 1 hour at room temperature. Immunoblots were incubated with primary antibodies at the appropriate dilutions overnight at 48C. The following primary antibodies were used in the Western blot analysis: rabbit polyclonal antihuman villin 2/ezrin antibody (EP886Y, 1:5000; Epitomics, Burlingame), mouse monoclonal anti-human vimentin (VI-RE/1, 1:5000; Acris Antibodies, Hiddenhausen), rabbit monoclonal anti-human annexin A1 (29, 1:1000; BD Transduction Laboratories, Heidelberg), mouse monoclonal antihuman beta-tubulin (SAP.4G5, 1:2000; Sigma-Aldrich, Munich, Germany), mouse monoclonal anti-human glial fibrillary acidic protein (GF-05, 1:50000; Acris Antibodies), mouse monoclonal anti-human peroxiredoxin 1 (4B11-G10, 1:1000; Abcam, Cambridge, MA), mouse monoclonal anti-human peroxiredoxin 5 (4C3, 1:1000, Abcam), mouse monoclonal anti-human peroxiredoxin 6 (1A11, 1:5000, Acris Antibodies), rabbit anti-human enolase-alpha (1:5000, Abcam), and rabbit anti-human annexin A5 (1:1000, Abcam). Membranes were washed in Tris-buffered saline containing 0.1% (v/v) Tween 20 (3  15 min) followed by incubation with the secondary antibody at room temperature for 1 hour. Additional washing was performed with Tris-buffered saline containing 0.1% (v/v) Tween 20 (3  10 min). The immunocomplexes were detected by the Amersham ECL PlusTM Western Blotting Detection System (GE Healthcare). Immunohistochemistry Immunohistochemistry was performed on a panel of 43 formalin-fixed, paraffin-embedded tumor tissue samples (Supplementary Material, Table S1). Tissue sections were mounted on poly-L-lysine– coated slides and dried overnight at 378C. After deparaffinization, rehydration, and quenching of endogenous peroxidases, antigen retrieval was performed in sodium citrate buffer (pH 6, 10 mmol/L) for 15 – 20 min in a steamer. Primary antibodies against annexin A1 (1:100) and villin 2/ ezrin (1:150) were the same as used in the Western

NEURO-ONCOLOGY

3

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

blot analysis. Mouse monoclonal anti-human vimentin antibodies (V9, 1:200) were purchased from BioGenex (San Ramon, CA) and used at a final dilution of 1:200. Polyclonal rabbit anti-GFAP antiserum was obtained from Dako (Hamburg, Germany) and used in a dilution of 1:8000. Immunoreactivity was detected with the Dako EnVisionTM Detection System, HRP using 3,30 -diaminobenzidine tetrahydrochloride as the chromogen. All sections were slightly counterstained with hematoxylin. Appropriate negative and external positive controls were performed for each individual immunohistochemical staining. The immunohistochemical stainings were semiquantitatively evaluated based on the labeling index of the percentage of immunoreactive tumor cells. Labeling indices were grouped into five immunoreactivity scores (IRSs): 0 (no or minimal reactivity, ,1% positive tumor cells), 1 (,10% positive tumor cells), 2 (10 – 50% positive tumor cells), 3 (50 – 90% positive tumor cells), and 4 (.90% positive tumor cells). Real-time Reverse Transcription PCR Analysis RNA for VIM (vimentin; NCBI GenBank accession no.: NM_003380.2), VIL2 (villin 2/ezrin; accession-no.: NM_003379.3), ANXA1 (annexin A1; accessionno.: NM_000700), and GFAP (glial fibrillary acidic protein, accession-no.: NM_002055.2) expression analyses was available from 5 OII, 4 OAII, 7 AII, and 7 AOIII (Supplementary Material, Table S1). The mRNA expression level of each gene was determined by quantitative real-time RT-PCR using the GeneAmp 5700 sequence detection system (Applied Biosystems, Darmstadt, Germany). Continuous quantitative measurement of the PCR product was enabled by intercalation of SYBR green fluorescent dye into the double-stranded DNA. The transcript level of each gene was normalized to the transcript level of ARF1 (ADP-ribosylation factor 1, accession-no: M36340). The primer sequences were as follows: VIM-sense 50 -gggacctctacgaggaggag-30 and VIM-antisense 50 -attccactttgcgttcaagg-30 (237 bp), VIL2sense 50 -gaacagacctttggcttgga-30 and VIL2-antisense 50 -ggatccgcttgttgattct-30 (204 bp), ANXA1-sense 50 -ctgg acctggagttgaaagg-30 and ANXA1-antisense 50 -cttggcaaa gggagatacca-30 (223 bp), and GFAP-sense 50 -acatcg agatcgccacctac-30 and GFAP-antisense 50 -atctccacggtc ttcaccac-30 (166 bp). Primer sequences for ARF1 are published elsewhere.26 As reference tissues, we used commercially available fetal and adult human brain RNA (BD Biosciences, St. Jose; BioChain, Hayward; Stratagene, Amsterdam, The Netherlands). Universal Human Reference RNA (Stratagene) was used as a calibrator to compare different RT-PCR runs. DNA Methylation Analysis by Sodium Bisulfite Sequencing A total of 33 gliomas (Supplementary Material, Table S1) and 3 nonneoplastic brain tissue samples were investigated for VIM and VIL2 50 -CpG island

4

NEURO-ONCOLOGY

hypermethylation by sequencing of sodium bisulfite – modified DNA. Twenty-nine CpG sites were evaluated for methylation within a 336-bp amplicon spanning the VIM transcription start site (chr10: 17311218– 17311533; UCSC genome browser at http://genome .ucsc.edu/). Primers were as follows: VIM-bis-sense 50 -ggtagtgggaggggatt-30 and VIM-bis-antisense 50 -ctac ccaaactataaatac-30 . The primers VIL2-bis-sense 50 -ttaa tttggagttagagtagaatt-30 and VIL2-bis-antisense 50 -accc cctccccatacc-30 were chosen to investigate 16 CpG-sites within a 166-bp fragment covering the VIL2transcription start site (chr6: 159159183 – 159159348). In the case of GFAP, 28 CpG sites of a CpG island covering exon 5, intron 5, and parts of exon 6 were analyzed for methylation in 31 tumors using the primers GFAP-bis-sense 50 -gttttttatttagtttgtagatttgat-30 and GFAP-bis-antisense 50 -atctacctctccaaaaactc-30 (377 bp amplicon; chr17: 40344329– 40344705). In addition, one CpG site (chr17: 40339013) located in the consensus binding sequence of the STAT3 transcription factor in the GFAP promoter region was screened for methylation using the primers GFAP-STAT3-sense 50 -gggtgg gtatagtgtttgt-30 and GFAP-STAT3-antisense 50 -ctactttt atcccaaaataccaaa-30 (126 bp amplicon). DNA was treated with sodium bisulfite as described elsewhere.27 PCR was performed for 45 cycles using HotStarTM Taq DNA polymerase (Qiagen, Hilden, Germany). PCR products were purified and directly sequenced using the BigDyeTM Cycle Sequencing kit v1.1 (Applied Biosystems). The results of sodium bisulfite sequencing were scored according to the ratio of the cytosine (methylated) to thymidine (unmethylated) peak at each CpG site: 0, no methylation; 1, weak methylation (ie, peak height of the methylated signal is lower than 0.5 relative to the unmethylated signal); 2, moderate methylation (ie, peak height of the methylated signal is between 0.5 and 1.5 relative to the unmethylated signal); and 3, strong methylation (ie, peak height of the methylated signal is higher than 1.5 relative to the unmethylated signal). On the basis of these semi-quantitative scores, the tumors were subdivided into two groups: (i) no CpG island hypermethylation (methylation score 1, 2, or 3 in ,50% of the investigated CpG sites) vs (ii) CpG island hypermethylation (methylation score 1, 2 or 3 in 50% of the investigated CpG sites). In addition, a numerical methylation score was calculated for each tumor by summing up the methylation levels (0 –3) at each of the investigated CpG sites. Prior to this analysis, we compared this direct sequencing approach with sequencing after subcloning of bisulfite-modified DNA and obtained similar results for both methods (unpublished data, M.W.).

Statistical Analysis For statistical analysis of the immunohistochemical results, box-and-whisker plots were used to illustrate the distribution of data and Mann– Whitney– Wilcoxson’s U-test was used to depict statistically

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

relevant expression differences stratified according to the 1p/19q allelic status (StatisticaTM software, version 7.1, Statsoft, Tulsa, OK; P values .05 were considered as significant). ROC plots, with the area under the ROC plot curve (AUC) providing a measure of diagnostic accuracy of a test in predicting cases with and without LOH on 1p and 19q, were calculated for each immunohistochemically investigated protein. Two-sided Student’s t-test and Fisher’s exact test were performed with GraphPad Prism 4 software (P , .05 were considered as significant). To assess correlations between DNA methylation and expression independently from arbitrary cut-off levels, we performed Deming (Model II) regression analysis taking into account that both X (methylation score) and Y (relative expression) are subject to error and have different standard deviations.28 The null hypothesis was tested that the computed slope was significantly nonzero with a P value of ,.05, indicating a statistically significant relationship between X and Y.

Results Allelic Status on 1p and 19q The 62 primary gliomas were screened by microsatellite analysis for allelic losses on 1p and 19q (Supplementary Material, Table S1). We detected combined LOH on 1p and 19q in 9 out of 11 (82%) WHO grade II oligodendrogliomas (OII), in 4 out of 11 (36%) WHO grade II oligoastrocytomas (OAII), in 12 out of 17 (71%)

WHO grade III anaplastic oligodendrogliomas (AOIII), and in 7 out of 8 (88%) WHO grade III anaplastic oligoastrocytomas (AOAIII). None of the 15 investigated diffuse astrocytomas (AII) demonstrated combined allelic losses on 1p and 19q. Protein Expression Profiling Using DIGE Minimal Labeling and Large 2D-Gel Electrophoresis To identify proteins differentially expressed in gliomas with or without 1p/19q deletion, we generated protein expression profiles of 9 WHO grade II oligoastrocytomas, including 5 tumors without and 4 tumors with LOH on both chromosomal arms. Oligoastrocytomas of WHO grade II were chosen for the proteomic profiling, because two similarly large groups of tumors with (n ¼ 4) and without (n ¼ 5) 1p/19q losses could be studied from this particular histological entity. All other histologically defined tumor groups from our series were biased either toward frequent 1p/19q deletion (OII, AOIII, and AOAIII) or toward common retention of heterozygosity on these chromosomal arms (astrocytic tumors). Histologically, there was no obvious difference between cases with and without 1p/19q losses. Importantly, there was no obvious bias toward either the astrocytic or the oligodendroglial tumor component in the 1p/ 19q-deleted vs retained tumor groups. The application of high-resolution large 2D-gels in combination with DIGE fluorescence dye protein labeling revealed about 3500 distinct spots per sample (Fig. 1). In our differential analysis, we detected 72 regulated proteins. By means of MALDI-MS, we identified

Fig. 1. Representative large 2D-DIGE gel (40  30 cm) showing approximately 3500 protein spots. The arrows indicate differentially regulated protein spots determined by image analysis and identified by MALDI-MS. The annotated spot numbers correspond to the numbers given in Supplementary Material, Table S2. Case number OA50 is shown in red and OA28 in green. The overlay is displayed.

NEURO-ONCOLOGY

5

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

43 proteins, representing 22 nonredundant proteins (Table 1), by applying the following criteria: P  .05 and fold change 1.8. A total of 19 nonredundant proteins were expressed at higher levels in gliomas without LOH on 1p and 19q, whereas 3 nonredundant proteins were expressed at higher levels in gliomas with LOH on 1p and 19q. The relatively high number of redundant proteins identified was predominantly due to the detection of various isoforms of GFAP and vimentin (Supplementary Material, Table S2). Almost all proteins could be assigned to tumor-relevant processes, in particular, cell proliferation, angiogenesis, and invasion, using the human protein reference database29 (Supplementary Material, Table S2). Fig. 2 shows examples of results obtained by 2D-DIGE for the four differentially expressed proteins that were selected for further validation (vimentin, villin 2/ezrin, annexin A1, and GFAP). Validation of Proteomic Data by Western Blot Analysis A total of 9 differentially expressed candidate proteins (corresponding genes in parentheses), namely vimentin (VIM), annexin A1 (ANXA1), annexin A5 (ANXA5), villin 2/ezrin (VIL2), glial fibrillary acidic protein (GFAP), enolase 1 (ENO1), and peroxiredoxin 1, 5, and 6 (PRDX1, 5, and 6), were selected for validation experiments by Western blot analysis on tumor protein extracts from 19 glioma patients (Supplementary

Material, Table S1). These proteins were selected on the basis of interesting biological functions, high foldchange, and availability of commercial antibodies suitable for Western blot analysis. Western blotting corroborated the DIGE/MALDI-MS results by demonstrating lower expression levels of vimentin, annexin A1, villin 2/ezrin, GFAP, and peroxiredoxin 1 in 1p/ 19q-deleted tumors when compared with tumors without detectable 1p/19q losses (Fig. 3, data of peroxiredoxin 1 not shown). Whereas, in cases of candidate proteins annexin A5 and enolase 1, Western blotting failed to verify DIGE/MALDI-MS results (data not shown). Western blot results obtained with the available antibodies against peroxiredoxin 5 and peroxiredoxin 6 were not interpretable due to poor staining quality (data not shown). Validation of Proteomic Data by Immunohistochemistry Four of the candidate proteins confirmed by Western blot analysis (vimentin, villin 2/ezrin, annexin A1, and GFAP) were additionally validated by immunohistochemical analysis in a panel of 43 primary gliomas (Supplementary Material, Tables S1 and S3; Fig. 4). These four proteins were selected based on the commercial availability of antibodies suitable for immunohistochemical staining of routinely formalin-fixed and paraffin-embedded tissue. In comparison with tumors

Table 1. List of nonredundant proteins identified by proteomic profiling using 2D-DIGE and MALDI-MS as being differentially expressed in 1p/19q-deleted vs 1p/19q-retained oligoastrocytomas of WHO grade II No.

Protein

Gene symbol

Accession

Chromosome

Down-regulated in oligoastrocytomas with 1p/19q deletion 1 Aldose reductase

AKR1B10

IPI00413641.6

7q33

2

Alpha crystallin B chain

CRYAB

IPI00021369.1

11q22.3-q23.1

3 4

Annexin A1 Annexin A5

ANXA1 ANXA5

IPI00218918.4 IPI00329801.11

9q12-q21.2 4q28-q32

5

Chloride intracellular channel protein 1

CLIC1

IPI00010896.2

6p22.1-p21.2

6 7

DnaJ homolog subfamily B member 1 Enolase-a

DNAJB1 ENO1

IPI00015947.4 IPI00465248.4

19p13.2 1p36.3-p36.2

8

Gelsolin precursor

GSN

IPI00026314.1

9q33

9 10

Glial fibrillary acidic protein, splice isoform 1 Glial fibrillary acidic protein, splice isoform 2

GFAP GFAP

IPI00383815.4 IPI00383815.4

17q21 17q21

11

Glial fibrillary acidic protein, splice isoform 3

GFAP

IPI00443478.1

17q21

12 13

Glyceraldehyde-3-phosphate dehydrogenase Heterogeneous nuclear ribonucleoproteins A2/B1

GAPDH HNRPA2B1

IPI00219018.6 IPI00477522.1

12p13 7p15

14

Nucleoside diphosphate –linked moiety X motif 16

NUDT16

IPI00411732.4

3q22.1

15 16

Peroxiredoxin 1 Peroxiredoxin 6

PRDX1 PRDX6

IPI00641244.1 IPI00220301.4

1p34.1 1q25.1

17

Sorbitol dehydrogenase

SORD

IPI00216057.4

15q15.3

18 19

Villin 2/ezrin Vimentin

VIL2 VIM

IPI00479359.5 IPI00418471.5

6q25.2– q26 10p13

Up-regulated in oligoastrocytomas with 1p/19q deletion 20 21

Peroxiredoxin 5, mitochondrial precursor Platelet-activating factor acetylhydrolase IB-a subunit

PRDX5 PAFAH1B1

IPI00024915.2 IPI00218728.2

11q13 17p13.3

22

Syntaxin-binding protein 1

STXBP1

IPI00084828.1

9q34.1

6

NEURO-ONCOLOGY

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Fig. 2. Representative results obtained by 2D-DIGE profiling for four proteins that demonstrated differential expression in 1p/19q-deleted vs 1p/19q-retained oligoastrocytomas, which were selected for further validation by Western blotting and immunohistochemistry. The left and middle columns show 2D-DIGE images from individual tumors with (left column) or without (middle column) 1p/19q deletion. The right column displays box plot diagrams illustrating significantly different expression levels of the respective protein in oligoastrocytomas with 1p/19q deletion when compared with oligoastrocytomas with 1p/19 retention. (A) Annexin A1 (left, case OA50; middle, case OA28). (B) Glial fibrillary acidic protein (GFAP) (left, case OA40; middle, case OA51). (C) Villin 2/ezrin (left, case OA50; middle, case OA28). (D) Vimentin (left, case OA40; middle, case OA51).

with 1p/19q losses, gliomas with retention of heterozygosity on both chromosomal arms displayed significantly higher immunohistochemical labeling scores for vimentin (median immunoreactivity score (IRS) for 1p/ 19q-retained tumors: 4, range 2–4; median IRS 1p/ 19q-deleted tumors: 1, range 0–3; Mann–Whitney– Wilcoxson’s U-test P , .0001). Similarly, annexin A1 immunostaining was significantly stronger in tumors without 1p/19q losses (median IRS for 1p/19q-retained tumors: 3, range 0–4; median IRS 1p/19q-deleted tumors: 1, range 0–2; Mann–Whitney–Wilcoxson U-test P , 0.001). Both vimentin and annexin A1 were expressed in vascular cells in addition to tumor cells. The IRSs for villin 2/ezrin (median IRS for 1p/ 19q-retained tumors: 3, range 2–4; median IRS 1p/ 19q-deleted tumors: 1, range 0–4; Mann–Whitney– Wilcoxson’s U-test P , 0.001) and GFAP (median IRS for 1p/19q-retained tumors: 3, range 1–4; median IRS 1p/19q-deleted tumors: 2, range 0–3; Mann–Whitney– Wilcoxson’s U-test P , 0.01) also were significantly higher in 1p/19q-retained gliomas.

To assess the predictive power of each immunohistochemical assay for subgroup stratification of gliomas according to the allelic status on 1p/19q, we performed ROC curve plots30 (Fig. 5), with the area under the curve as a measure for accuracy of the tests. The areas under the curve were 0.91 for vimentin, 0.84 for annexin A1, 0.82 for vilin 2/ezrin, and 0.83 for GFAP. A decision threshold representing an IRS of 3 (50 –90% positive tumor cells) correlated with the highest accuracy for each tested protein. Although all four proteins revealed high scores in the ROC analysis, vimentin expression demonstrated the best performance for glioma subgroup stratification.

mRNA Expression Analysis of VIM, EZR/VIL2, ANXA1, and GFAP In line with the protein expression data, real-time reverse transcription-PCR analyses revealed significantly lower mean transcript levels of VIM and ANXA1 in

NEURO-ONCOLOGY

7

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Fig. 3. Validation of selected candidate proteins using Western blot analysis. Protein abundances for annexin A1, GFAP, villin 2/ezrin, and vimentin obtained in 10 different gliomas (see Supplementary Material, Table S1) with 1p/19q deletion (lanes 1– 10) and 9 different gliomas without 1p/19q deletion (lanes 11 –19) are shown. b-Tubulin was used as a control for protein loading. The molecular weight of each protein is indicated on the right side of the blots.

1p/19q-deleted gliomas when compared with tumors without these deletions (Supplementary Material, Table 4; Student’s t-test, P , .01 and P , .05, respectively). Similarly, the mean transcript levels of VIL2 and GFAP (relative to the Human Universal reference RNA) were lower in 1p/19q-deleted tumors compared with 1p/19q-retained tumors; however, these differences were statistically not significant (VIL2: mean 3.9 vs 4.7, P ¼ .63; GFAP: mean 69.6 vs 112.4, P ¼ 0.36; Student’s t-test). CpG Island Hypermethylation Analysis of VIM, VIL2, and GFAP The ANXA1 gene lacks an associated 50 -CpG island and therefore was not studied for DNA methylation by sodium bisulfite sequencing. Hypermethylation of the 50 -CpG– rich region of the VIM gene at 10p13 (as defined by a methylation score of 1, 2, or 3 in 50% of the investigated CpG sites) was detected in 4 out of 9 OII (44%), in 2 out of 8 OAII (25%), in 3 out of 9 AOIII (33%), but in none out of 7 diffuse astrocytomas (AII) (Fig. 6A and B). Sodium bisulfite sequencing of the 50 -CpG island associated with VIL2 at 6q25.2 – q26 revealed evidence for hypermethylation in 7 out of 8 OII (88%), in 4 out of 8 OAII (50%), in 5 out of 9 AOIII (56%), and in 5 out of 7 AII (71%) (Fig. 6C). DNA extracted from three nonneoplastic brain tissue samples did not show any hypermethylation of the investigated CpG-rich regions linked to VIM or VIL2. In contrast, we found methylated CpG sites of the CpG island located within the GFAP gene in three nonneoplastic brain tissue samples and in all investigated gliomas (Fig. 6C). The total methylation scores (as defined by the sum of the methylation scores at each investigated CpG site) of each of the three investigated genes were

8

NEURO-ONCOLOGY

significantly higher in gliomas with 1p/19q deletions when compared with gliomas without 1p/19q deletions (P , .001 for VIM and VIL2; P , .01 for GFAP; Student’s t-test). In contrast, methylation of the CpG dinucleotide located within a STAT3-binding element in the GFAP promoter region31 did not significantly differ between 1p/19q-deleted vs nondeleted tumors (P ¼ .07; Student’s t-test). There was a significant inverse correlation between the extent of 50 -CpG island methylation of VIM and the VIM mRNA expression level (P , .05; Deming regression analysis). The extent of CpG island methylation of VIL2 and GFAP was not significantly correlated with reduced transcript levels (VIL2: P ¼ .82; GFAP: P ¼ .26; Deming regression analysis).

Discussion Combined deletion of 1p and 19q has been associated with increased sensitivity to chemo- and radiotherapy as well as prolonged survival in patients with malignant oligodendroglial tumors.3,5,6 Therefore, molecular testing of the 1p/19q status using microsatellite analysis for the detection of LOH is now often performed for prognostic assessment in the routine clinical setting as well as for therapeutic stratification of patients within prospective clinical trials.32 Alternatively, fluorescencein situ-hybridization (FISH) is frequently utilized as a histology-based method that assesses 1p and 19q deletions, especially in cases where only the formalin-fixed, paraffin-embedded tissues and no accompanying blood samples are available.7 Nevertheless, both microsatellite and FISH analysis are labor intensive and thus require an especially equipped laboratory. In contrast, immunohistochemistry combines the advantage of a slide-based method with an easy-to-use assay that is widely accessible within the histopathological community. Thus, it would be very helpful to identify proteins with expression patterns linked to the 1p/19q status, which could then easily be detected by means of immunohistochemistry. Therefore, we performed a screening approach at the proteome level to identify proteins that are differentially expressed between gliomas with and without 1p/19q deletion and could potentially serve as immunohistochemical surrogate markers. We intentionally did not bias our results for proteins encoded on chromosome arms 1p and 19q, but aimed at a proteomewide identification of protein markers that may distinguish gliomas according to the 1p/19q allelic status. 2D-DIGE analysis followed by MALDI-MS was performed on 5 oligoastrocytomas with retention and 4 oligoastrocytomas with LOH on 1p and 19q. This screening step was intentionally restricted to 1 histologically defined tumor entity (ie, oligoastrocytoma of WHO grade II), in which loss or retention of heterozygosity on 1p and 19q each is detectable in approximately half of the cases. Using this proteome-based approach, we identified 22 distinct candidate proteins with the potential to stratify gliomas according to their 1p/19q status. Our list of differentially expressed proteins

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Fig. 4. Selected examples of the immunohistochemical validation of the differential expression of annexin A1 (left, case OA50; middle, case OA51), GFAP (left, case OA29; middle, case O64), villin 2/ezrin (left, case OA29; middle, case OA20), and vimentin (left, case OA29; middle, case OA28) in formalin-fixed and paraffin-embedded primary glioma tissue samples with and without 1p/19q deletion. All sections are counterstained with hematoxylin. The original microscopic magnification of each picture was 400. On the right-hand side of each immunohistochemical panel, box plot diagrams are shown illustrating the results of immunohistochemical evaluation of each protein in 43 different gliomas, including 16 tumors without 1p/19q deletion and 27 tumors with 1p/19q deletion. All four proteins showed significantly lower immunoreactivity scores in the 1p/19q-deleted tumors.

showed no overlap with the 19 proteins identified in a recent proteomic study by another group.33 This discrepancy is likely due to differences in the tumor types used for the proteomic screening (ie, oligoastrocytomas of WHO grade II in our study vs oligodendrogliomas of WHO grades II and III in the study of Okamoto et al.33). In both studies, the vast majority of identified candidate proteins were not encoded by genes located on 1p or 19q. From our candidate list, only the genes for a-enolase (ENO1; 1p36.3 –p36.2) and peroxiredoxin 1 (PRDX1; 1p34.1) map to 1p. Similarly, only

two differentially expressed proteins encoded on 1p, namely glutathione S-transferase M2 (GSTM2, 1p13.3) and F-actin capping protein b-subunit (CAPZB, 1p36.13), were found by Okamoto et al.33 Therefore, we speculate that due to the technical limitations of 2D-DIGE, differential expression of most candidate genes located on 1p and 19q cannot be detected directly. For instance, only proteins with a copy number of approximately 3000–5000 per cell can be detected using fluorescence dye protein labeling, in which 50 mg of protein is analyzed from whole cell lysates.

NEURO-ONCOLOGY

9

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Fig. 5. Receiver operation characteristic curves indicating sensitivity and specificity of the immunohistochemical stainings for vimentin, GFAP, villin 2/ezrin, and annexin A1 with respect to the distinction of 1p/19q-deleted vs 1p/19q-retained gliomas. The results are based on the immunohistochemical evaluation of 43 gliomas. ROC analysis revealed that each analyzed protein has the potential to determine 1p/19q status in glioma with high diagnostic accuracy (AUC ¼ 0.91 for vimentin, 0.84 for annexin A1, 0.83 for GFAP, and 0.82 for villin 2/ezrin staining) using immunohistochemistry.

Expression changes of low abundance proteins, such as transcription factors or growth factors, often become apparent only via their secondary effects on the expression of other proteins. Thus, it is possible that the proteome-wide changes observed in relation to allelic status on 1p/19q are mostly due to secondary effects of low-abundance proteins encoded on 1p/19q. The ENO1 gene has been implicated as a putative tumor suppressor gene in various types of cancer. It maps within a common region of deletion in neuroblastoma and has been reported to induce cell death when transfected into neuroblastoma cells.34 Furthermore, an alternative ENO1 translation product (MBP-1) acts as a negative regulator of c-Myc.35 However, although 2D-DIGE profiling suggested reduced expression of enolase 1 in 1p/19q-deleted gliomas, Western blot analysis did not confirm a significantly different expression between 1p/19q-deleted vs nondeleted gliomas, indicating that enolase 1 does not represent a promising diagnostic marker in gliomas. In a microarray-based screening, PRDX1 has been reported as being differentially expressed at the mRNA level in 1p/19q-deleted vs nondeleted gliomas.36 Here we confirm this finding at the protein level, thus suggesting PRDX1 as a candidate gene for further molecular and functional investigations. The vast majority of the differentially expressed proteins identified by 2D-DIGE profiling showed higher relative expression levels in 1p/19q-retained tumors when compared with 1p/19q-deleted tumors. Among the proteins with higher expression in gliomas lacking 1p/19q deletion were the two intermediate filament

10

NEURO-ONCOLOGY

proteins GFAP and vimentin, both of which are well-established diagnostic markers. GFAP is the major intermediate filament protein in normal astrocytes; germline mutations in the GFAP gene cause Alexander disease, a fatal leukodystrophy with abundant formation of Rosenthal fibers and astrogliosis.37 Immunohistochemical studies on gliomas reported that astrocytic gliomas generally show GFAP immunoreactivity, with the fraction of GFAP-positive cells being inversely proportional to the degree of anaplasia (ie, glioblastomas often demonstrate regionally heterogeneous and sometimes only little GFAP expression).38 In oligodendrogliomas, GFAP positivity is restricted to specific types of tumor cells called gliofibrillary oligodendrocytes and minigemistocytes.39 In contrast, the classic oligodendroglioma cells with round nuclei and artificially swollen cytoplasm are GFAP negative. Since 1p/19q deletion is particularly common in oligodendrogliomas with classic histology40 and oligodendrogliomapredominant oligoastrocytomas,41 our proteomic finding of lower GFAP protein levels in 1p/19q-deleted gliomas, which involved all 3 splicing isoforms, likely reflects the higher fraction of classic oligodendroglial tumor cells in these tumors. It remains to be investigated, however, why GFAP expression is typically low or absent in 1p/19q-deleted glioma cells. One may speculate that this finding reflects a different histogenetic origin and lineage commitment of the tumor cells. Alternatively, 1p/19q deletion may lead to a distinct expression pattern of transcriptional regulators, growth factors, and cytokines known to regulate GFAP expression.42,43 In addition, epigenetic DNA changes may be involved.31,44 Methylation analysis of parts of the CpG island located within the GFAP gene revealed more extensive methylation in 1p/19q-deleted when compared with 1p/19q nondeleted gliomas. However, methylation of this CpG island was also detected in nonneoplastic brain tissue samples and gliomas without 1p/19q deletion (Fig. 6D). In addition, the detected methylation was not associated with significantly different expression of GFAP transcripts. Thus, the functional significance of methylation in this particular CpG island with respect to transcriptional activity of the GFAP gene requires further investigation. A study on embryonic mouse astrocytes suggested that methylation of a single CpG dinucleotide located within a STAT3-binding site in the GFAP promoter may regulate transcriptional activity of the gene.31 However, we did not detect any association between methylation at this particular CpG site and the 1p/19q status or the GFAP expression level in gliomas. Vimentin was the other intermediate filament protein identified as being differentially expressed in 1p/19q deleted vs nondeleted gliomas. Again, previous immunohistochemical studies have reported frequent expression of vimentin in astrocytic gliomas, often together with GFAP.45,46 In oligodendrogliomas, vimentin expression is usually absent in classic oligodendroglioma cells and mostly restricted to cases containing GFAP-positive gliofibrillary oligodendrocytes and minigemistocytes.45,46 One study suggested vimentin expression as an

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Fig. 6. Summary of the results obtained in 33 gliomas concerning DNA hypermethylation of VIM (A), VIL2 (C), and GFAP (D) in relation to the 1p/19q allelic status. The chromosomal location and genomic structure of the investigated genes, including the location of associated CpG islands is indicated on top of each scheme (derived from UCSC genome browser at http://genome.ucsc.edu/). The results of the DNA methylation analyses are represented in a 4-tiered semiquantitative grey-scale pattern: white square, not methylated; light grey square, weakly methylated; grey square, moderately methylated; black square, strongly methylated. Each number represents a single CpG dinucleotide analyzed for methylation by sodium bisulfite sequencing in the respective CpG islands. The total number of investigated CpG sites was 29 (VIM), 16 (VIL2), and 28 (GFAP), respectively. (B) Examples of sequencing of parts of the VIM 50 -CpG– rich region after sodium bisulfite modification showing methylation of CpG sites in the 1p/19q-deleted tumor AO29 (arrowheads in upper lane) but not in tumor OA50, which lacked detectable 1p/19q losses (lower lane) (shown is the reverse sequence). n.d., not determined.

unfavorable prognostic marker in oligodendroglioma patients.47 However, a significant association of vimentin expression with the 1p/19q allelic status had not been reported before. Furthermore, we provide the first evidence indicating that the low or the absence of expression of vimentin in 1p/19q-deleted oligodendroglioma cells is often due to hypermethylation of the

50 -CpG island overlapping with the VIM promoter region.48 Our results show that hypermethylation of VIM is more common in 1p/19q-deleted gliomas and associated with reduced mRNA levels. Aberrant DNA methylation of VIM has also been reported in colorectal carcinomas, in which it has been suggested to be a novel biomarker in fecal DNA.49

NEURO-ONCOLOGY

11

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

Villin 2 (ezrin), the third candidate with significantly lower expression in gliomas with 1p/19q deletion, is a cytoskeletal linker protein of the so-called ERM (ezrin, radixin, and moesin) family of proteins, which is involved in different cellular functions, such as regulation of actin cytoskeleton, control of cell shape, adhesion and motility, and modulation of signaling pathways.50 Immunohistochemical studies on gliomas suggested an association of villin 2 expression with malignant progression of astrocytic tumors.51,52 One of these studies reported that villin 2 was not expressed in normal oligodendrocytes and oligodendrogliomas,51 while the other study detected expression in all glioma types, including oligodendrogliomas.52 Our results indicate that the expression level of villin 2 is lower in oligodendroglial tumors with deletion of 1p and 19q. In addition, we provide the first evidence for a differential methylation of the VIL2 gene in gliomas with and without this genetic marker. However, the precise relationship between DNA methylation and expression of ezrin in glioma cells remains to be further investigated since VIL2 transcript levels were not significantly associated with the methylation status in our glioma cases. The fourth candidate protein, annexin A1 (lipocortin 1), also showed lower expression in gliomas with 1p and 19q losses. Annexin A1 is a calcium-dependent phospholipid-binding protein involved in a variety of cellular processes, such as cell proliferation, differentiation, apoptosis, and inflammation.53 Annexin A1 has phospholipase 2 inhibitory activity and is activated by glucocorticoids in a calcium-dependent manner.54 Loss of annexin A1 expression has been associated with tumor progression and/or advanced disease stage in different types of carcinomas.55 – 57 In brain tumors, a proteomic analysis identified annexin A1 as a potential marker that distinguishes pediatric ependymomas from primitive neuroectodermal tumors.58 Immunohistochemical studies reported on frequent overexpression of lipocortin 1/annexin A1 in astrocytic gliomas, in particular glioblastomas, but not in oligodendroglial tumors.59,60 Our data confirm and extend these findings by linking reduced annexin A1 expression to the hallmark genetic feature of oligodendrogliomas (ie, combined loss of 1p and 19q). However, the molecular mechanisms underlying this association remain to be resolved. In contrast to VIM and VIL2, the ANXA1 gene lacks a 50 -CpG island, which may argue against a role of aberrant DNA methylation, but does not exclude the involvement of epigenetic changes. A number of studies have reported on aberrant promoter methylation affecting a diverse set of genes in oligodendroglial tumors with 1p/19q deletion. Genes showing frequent epigenetic silencing in these tumors include not only a number of candidate tumor suppressor genes on 1p, such as TP73, CITED4, and DIRAS3, or on 19q, such as EMP3, ZNF342, and PEG3, but also a variety of genes from other chromosomes, such as the tumor suppressor genes CDKN2A, p14ARF, CDKN2B, and RB1, as well as DAPK1

12

NEURO-ONCOLOGY

(death-associated protein kinase 1), ESR1 (estrogen receptor 1), THBS (thrombospondin 1), TIMP3 (tissue inhibitor of metalloproteinase 3),61 – 63 and MGMT. 64 In fact, oligodendrogliomas often have hypermethylated several of the genes listed above.61 – 63 Our present data add the VIM and VIL2 genes to the growing list of epigenetically down-regulated genes in 1p/19q-deleted oligodendroglial tumors, thus supporting the hypothesis that these tumors may suffer from a general defect in the regulation of DNA methylation, which results in an aberrant epigenetic phenotype characterized by promoter hypermethylation of multiple genes. In summary, we report that 2D-DIGE combined with MALDI-MS is a suitable approach to identify candidate proteins that are differentially expressed in gliomas stratified for their allelic status on chromosome arms 1p and 19q. Validation of selected candidates by Western blot analysis and immunohistochemistry confirmed four proteins (vimentin, villin 2/ezrin, annexin A1, and GFAP) as being expressed at significantly lower levels in 1p/ 19q-deleted gliomas when compared with 1p/ 19q-retained gliomas. Immunohistochemistry for these proteins thus provides useful information in regard to the differential diagnosis of gliomas. However, whether these or yet other immunohistochemical markers will eventually be able to substitute for 1p/ 19q molecular testing with respect to the prognostic assessment of gliomas requires further investigation. Our finding of frequent VIM and VIL2 promoter methylation in 1p/19q-deleted gliomas supports the observation that these tumors often show epigenetic inactivation of multiple genes.

Supplementary Material Supplementary material is available at Neuro-Oncology online.

Acknowledgments The authors would like to thank Eva Hawranke and Christa Ma¨hler for excellent technical assistance, as well as Dr. Kristin Gierga for critical reading of the manuscript. Conflict of interest statement. None declared.

Funding This study was supported by grants from the Deutsche Krebshilfe (German Glioma Network, grant no. 107940, to G.R., U.S., and K.S.), the Forschungskommission of the Medical Faculty, Heinrich-Heine-University (project no. 9772297 to M.W.; project no. 9772307 to M.J.R.), and the Ministry for Innovation, Science, Research and Technology (MIWFT) of Nordrhein Westfalen.

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

References 1.

Louis DN, Ohgaki H, Wiestler OD, et al., eds. WHO Classification of Tumours of the Central Nervous System. Lyon: IARC Press; 2007.

2. 3.

Hegi ME, Diserens AC, Gorlia T, et al. MGMT gene silencing and benefit Cairncross JG, Ueki K, Zlatescu MC, et al. Specific genetic predictors of

human gliomas using matrix-based comparative genomic hybridization.

Louis DN. Molecular pathology of malignant gliomas. Annu Rev Path

mixed anaplastic oligodendroglioma: intergroup radiation therapy oncology group trial 9402. J Clin Oncol. 2006;24:2707– 2714. van den Bent MJ, Carpentier AF, Brandes AA, et al. Adjuvant procarba-

J.

Large-gel

2-Delectrophoresis.

Methods

Mol

Biol.

1999;112:147 –172. 23. Klose J, Kobalz U. Two-dimensional electrophoresis of proteins: an updated protocol and implications for a functional analysis of the genome. Electrophoresis. 1995;16:1034– 1059. 24. Sitek B, Sipos B, Pfeiffer K, et al. Establishment of “one-piece” large-gel

not overall survival in newly diagnosed anaplastic oligodendrogliomas

2-DE for high-resolution analysis of small amounts of sample using

and oligoastrocytomas: a randomized European Organisation for

difference gel electrophoresis saturation labeling. Anal Bioanal Chem.

2006;24:2715 –2722.

9.

22. Klose

zine, lomustine, and vincristine improves progression-free survival but

Research and Treatment of Cancer phase III trial. J Clin Oncol.

8.

silver stain method allows for 30 minute detection of proteins in polyacrylamide gels. J Biochem Biophys Methods. 1994;28:239 –242.

Cairncross G, Berkey B, Shaw E, et al. Phase III trial of chemotherapy plus radiotherapy compared with radiotherapy alone for pure and

7.

Int J Cancer. 2005;117:95 –103. 21. Nesterenko MV, Tilley M, Upton SJ. A simple modification of Blum’s

Mech Dis. 2006;1:97 –117.

6.

1p/19q status with survival. Brain Pathol. 2004;14:121 –130. 20. Roerig P, Nessling M, Radlwimmer B, et al. Molecular classification of

godendrogliomas. J Natl Cancer Inst.1998;90:1473– 1479.

5.

ment of candidate regions on chromosome arm 1p and correlation of

from temozolomide in glioblastoma. N Engl J Med. 2005;352:997–1003. chemotherapeutic response and survival in patients with anaplastic oli4.

19. Felsberg J, Erkwoh A, Sabel M, et al. Oligodendroglial tumors: refine-

2008;391:361 –365. 25. Marcus K, Immler D, Sternberger J, et al. Identification of platelet pro-

Yip S, Iafrate AJ, Louis DN. Molecular diagnostic testing in malignant

teins separated by two-dimensional gel elctrophoresis and analyzed by

gliomas: a practical update on predictive markers. J Neuropathol Exp

matrix assisted laser desorption/ionization-time of flight-mass spec-

Neurol. 2008;67:1 –15.

trometry

Dong S, Pang JC, Hu J, et al. Transcriptional inactivation of TP73

Electrophoresis, 2000;21:2622 –2636.

and

detection

of

tyrosine-phosphorylated

proteins.

expression in oligodendroglial tumors. Int J Cancer 2002;98:370 –375.

26. Ehrbrecht A, Muller U, Wolter M, et al. Comprehensive genomic analy-

Barbashina V, Salazar P, Holland EC, et al. Allelic losses at 1p36 and 19q13 in

sis of desmoplastic medulloblastomas: identification of novel amplified

gliomas: correlation with histologic classification, definition of a 150-kb

genes and separate evaluation of the different histological components.

minimal deleted region on 1p36, and evaluation of CAMTA1 as a candidate 10. McDonald JM, Dunmire V, Taylor E, et al. Attenuated expression of DFFB is a hallmark of oligodendrogliomas with 1p-allelic loss. Mol Cancer. 2005;4:35.

novel PCR assay for methylation status of CpG islands. Proc Natl Acad Sci USA. 1996;18:9821 –9826. 28. Konings H. Use of Deming regression in method-comparison studies.

11. McDonald JM, Dunlap S, Cogdell D, et al. The SHREW1 gene, frequently deleted in oligodendrogliomas, functions to inhibit cell adhesion and migration. Cancer Biol Ther. 2006;5:300 –304. downregulation of the CITED4 gene at 1p34.2 in oligodendroglial tumours with allelic losses on 1p and 19q. Oncogene. 2007;26:5010–5016. 13. Pohl U, Cairncross JG, Louis DN. Homozygous deletions of the CDKN2C/p18INK4C gene on the short arm of chromosome 1 in anaplastic oligodendrogliomas. Brain Pathol 1999;9:639 –643. inactivation and transcriptional silencing of the DIRAS3 gene at 1p31 in with

1p

loss.

Int

J

biology in humans. Genome Res. 2003;13:2363– 2371. 30. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine. Clin Chem. 1993;39:561– 577. 31. Takizawa T, Nakashima K, Namihira M, et al. DNA methylation is a critical cell-intrinsic determinant of astrocyte differentiation in the fetal

14. Riemenschneider MJ, Reifenberger J, Reifenberger G. Frequent biallelic tumors

Surv Immunol Res. 1982;1:371– 374. 29. Peri S, Navarro JD, Amanchy R, et al. Development of human protein reference database as an initial platform for approaching systems

12. Tews B, Roerig P, Hartmann C, et al. Hypermethylation and transcriptional

oligodendroglial

J Pathol. 2006;208:554 –563. 27. Herman JG, Graff JR, Myo¨ha¨nen S, et al. Methylation-specific PCR: a

tumor suppressor gene. Clin Cancer Res. 2005;11:1119–1128.

Cancer.

2008;122:2503 –2510.

brain. Dev Cell. 2001;1:749 –758. 32. Stupp R, Hegi M. Neuro-oncology: oligodendroglioma and molecular markers. Lancet Neurol. 2007;6:10 –12. 33. Okamoto H, Li J, Gla¨sker S, et al. Proteomic comparison of oligodendro-

15. Wolf MR, Draghi N, Liang X, et al. 190RhoGAP can act to inhibit

gliomas with and without 1p LOH. Cancer Biol Ther. 2007;6:391–396.

PDGF-induced gliomas in mice: a putative tumor suppressor encoded

34. Ejeska¨r K, Krona C, Care´n H, et al. Introduction of in vitro transcribed

on human Chromosome 19q13.3. Genes Dev. 2003;17:476 –487. 16. Kunitz A, Wolter M, van den Boom J, et al. DNA hypermethylation and aberrant expression of the EMP3 gene at 19q13.3 in human gliomas. Brain Pathol. 2007;17:363– 370. 17. U¨nlu¨ M, Morgan ME, Minden JS. Difference gel electrophoresis: a single gel method for detecting changes in protein extracts. Electrophoresis. 1997;18:2071 –2077. 18. van den Boom J, Wolter M, Kuick R, et al. Characterization of gene expression profiles associated with glioma progression using oligonucleotide-based microarray analysis and real-time reverse transcriptionpolymerase chain reaction. Am J Pathol. 2003;163:1033–1043.

ENO1 mRNA in to neuroblastoma cells induce cell death. BMC Cancer. 2005;16:161. 35. Subramanian A, Miller DM. Structural analysis of alpha-enolase. Mapping the functional domains involved in down-regulation of the c-myc protooncogene. J Biol Chem. 2000;275:5958 – 5965. 36. Tews B, Felsberg J, Hartmann C, et al. Identification of novel oligodendroglioma-associated candidate tumor suppressor genes in 1p36 and 19q13 using microarray-based expression profiling. Int J Cancer. 2006;119:792 –800. 37. Quinlan RA, Brenner M, Goldman JE, et al. GFAP and its role in Alexander disease. Exp Cell Res. 2007;313:2077 –2087.

NEURO-ONCOLOGY

13

Grzendowski et al.: Proteome analysis of gliomas according to 1p/19q status

38. Brat DJ, Shehata BM, Castellano-Sanchez AA, et al. Congenital glioblas-

52. Tynninen O, Carpe´n O, Ja¨a¨skela¨inen J, et al. Ezrin expression in tissue

toma: a clinicopathologic and genetic analysis. Brain Pathol.

microarray of primary and recurrent gliomas. Neuropathol Appl Neurobiol. 2004;30:472 –477.

2007;17:276– 281. 39. Herpers MJ, Budka H. Glial fibrillary acidic protein (GFAP) in oligodendroglial tumors: gliofibrillary oligodendroglioma and transitional oligoastrocytoma as subtypes of oligodendroglioma. Acta Neuropathol. 1984;64:265– 272.

53. Lim LH, Pervaiz S. Annexin 1: the new face of an old molecule. FASEB J. 2007;21:968– 975. 54. Castro-Caldas M. Annexin-1: 2nd messanger of the anti-inflammatory actions of glucocorticoids. Acta Reumatol Port. 2006;31:293 –302.

40. Aldape K, Burger PC, Perry A. Clinicopathologic aspects of 1p/19q loss

55. Shen J, Person MD, Zhu J, et al. Protein expression profiles in pancreatic

and the diagnosis of oligodendroglioma. Arch Pathol Lab Med.

adenocarcinoma compared with normal pancreatic tissue and tissue

2007;131:242 –251.

affected by pancreatitis as detected by two-dimensional gel electro-

41. Maintz D, Fiedler K, Koopmann J, et al. Molecular genetic evidence for subtypes

of

oligoastrocytomas.

J

Neuropathol

Exp

Neurol.

1997;56:1098 –1104. 42. Gopalan SM, Wilczynska KM, Konik BS, et al. Nuclear factor-1-X regulates astrocyte-specific expression of the alpha1-antichymotrypsin and glial fibrillary acidic protein genes. J Biol Chem. 2006;281:13126–13133. 43. Laping NJ, Teter B, Nichols NR, et al. Glial fibrillary acidic protein: regulation by hormones, cytokines, and growth factors. Brain Pathol. 1994;4:259 –275.

phoresis and mass spectrometry. Cancer Res. 2004;64:9018 –9026. 56. Garcia Pedrero JM, Fernandez MP, Morgan RO, et al. Annexin A1 down-regulation in head and neck cancer is associated with epithelial differentiation status. Am J Pathol. 2004;164:73 –79. 57. Yu G, Wang J, Chen Y, et al. Tissue microarray analysis reveals strong clinical evidence for a close association between loss of annexin A1 expression and nodal metastasis in gastric cancer. Clin Exp Metastasis. 2008;25:695–702. 58. de Bont JM, den Boer ML, Kros JM, et al. Identification of novel

44. Fukuyama K, Matsuzawa K, Hubbard SL, et al. Analysis of glial fibrillary

biomarkers in pediatric primitive neuroectodermal tumors and ependy-

acidic protein gene methylation in human malignant gliomas.

momas by proteome-wide analysis. J Neuropathol Exp Neurol.

Anticancer Res. 1996;16:1251 –1257.

2007;66:505– 516.

45. Reifenberger G, Szymas J, Wechsler W. Differential expression of glial-

59. Johnson MD, Kamso-Pratt J, Pepinsky RB, et al. Lipocortin-1 immunor-

and neuronal-associated antigens in human tumors of the central and

eactivity in central and peripheral nervous system glial tumors. Hum

peripheral nervous system. Acta Neuropathol. 1987;74:105 –123. 46. Perentes E, Rubinstein LJ. Recent applications of immunoperoxidase his-

Pathol. 1989;20:772 –776. 60. Ruano Y, Mollejo M, Camacho FI, et al. Identification of survival-related

tochemistry in human neuro-oncology. An update. Arch Pathol Lab

genes of the phosphatidylinositol 30 -kinase signaling pathway in glio-

Med. 1987;111:796 –812.

blastoma multiforme. Cancer. 2008;112:1575 –1584.

47. Dehghani F, Schachenmayr W, Laun A, et al. Prognostic implication of his-

61. Wolter M, Reifenberger J, Blaschke B, et al. Oligodendroglial tumors

topathological, immunohistochemical and clinical features of oligodendro-

frequently demonstrate hypermethylation of the CDKN2A (MTS1,

gliomas: a study of 89 cases. Acta Neuropathol. 1998;95:493–504.

p16INK4a), p14ARF, and CDKN2B (MTS2, p15INK4b) tumor suppres-

48. Izmailova ES, Wieczorek E, Perkins EB, et al. A GC-box is required for expression of the human vimentin gene. Gene. 1999;235:69 –75. 49. Chen WD, Han ZJ, Skoletsky J, et al. Detection in fecal DNA of colon cancer-specific methylation of the nonexpressed vimentin gene. J Natl Cancer Inst. 2005;97:1124 –1132. 50. Hughes SC, Fehon RG. Understanding ERM proteins—the awesome power of genetics finally brought to bear. Curr Opin Cell Biol. 2007;19:51 –56.

sor genes. J Neuropathol Exp Neurol. 2001;60:1170– 1180. 62. Dong SM, Pang JC, Poon WS, et al. Concurrent hypermethylation of multiple genes is associated with grade of oligodendroglial tumors. J Neuropathol Exp Neurol. 2001;60:808–816. 63. Alonso ME, Bello MJ, Gonzalez-Gomez P, et al. Aberrant promoter methylation of multiple genes in oligodendrogliomas and ependymomas. Cancer Genet Cytogenet. 2003;144:134 –142. 64. Reifenberger G, Kros JM, Louis DN, et al. Oligodendroglioma. In: Louis,

51. Geiger KD, Stoldt P, Schlote W, et al. Ezrin immunoreactivity is associ-

DN, Ohgaki, H, Wiestler, OD, Cavenee, WK, eds. WHO Classification

ated with increasing malignancy of astrocytic tumors but is absent in oli-

of Tumours of the Central Nervous System. Lyon: IARC; 2007:

godendrogliomas. Am J Pathol. 2000;157:1785 –1793.

54 –59.

14

NEURO-ONCOLOGY

Suggest Documents