This site uses cookies, tags, and tracking settings to store information that help give you the very best browsing experience. Dismiss this warning

Genomic and transcriptome analysis revealing an oncogenic functional module in meningiomas

Xiao Chang Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Xiao Chang in
jns
Google Scholar
PubMed Close
 Ph.D.
,
Lingling Shi GHM Institute of CNS Regeneration, Jinan University, Guangzhou, China;

Search for other papers by Lingling Shi in
jns
Google Scholar
PubMed Close
 Ph.D.
,
Fan Gao Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Fan Gao in
jns
Google Scholar
PubMed Close
 Ph.D.
,
Jonathan Russin Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Jonathan Russin in
jns
Google Scholar
PubMed Close
 M.D.
,
Liyun Zeng Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Liyun Zeng in
jns
Google Scholar
PubMed Close
 Ph.D.
,
Shuhan He Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Shuhan He in
jns
Google Scholar
PubMed Close
 B.S.
,
Thomas C. Chen Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Thomas C. Chen in
jns
Google Scholar
PubMed Close
 M.D.
,
Steven L. Giannotta Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Steven L. Giannotta in
jns
Google Scholar
PubMed Close
 M.D.
,
Daniel J. Weisenberger USC Epigenome Center, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Daniel J. Weisenberger in
jns
Google Scholar
PubMed Close
 Ph.D.
,
Gabriel Zada Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by Gabriel Zada in
jns
Google Scholar
PubMed Close
 M.D.
,
Kai Wang Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, California;
Department of Psychiatry, Keck School of Medicine, University of Southern California, Los Angeles, California; and
Division of Bioinformatics, Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California

Search for other papers by Kai Wang in
jns
Google Scholar
PubMed Close
 Ph.D.
, and
William J. Mack Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, California;
Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, California;

Search for other papers by William J. Mack in
jns
Google Scholar
PubMed Close
 M.D.
Page Range:
E3
Volume/Issue:
Volume 35: Issue 6
DOI link:
https://doi.org/10.3171/2013.10.FOCUS13326
Free access

Object

Meningiomas are among the most common primary adult brain tumors. Although typically benign, roughly 2%–5% display malignant pathological features. The key molecular pathways involved in malignant transformation remain to be determined.

Methods

Illumina expression microarrays were used to assess gene expression levels, and Illumina single-nucleotide polymorphism arrays were used to identify copy number variants in benign, atypical, and malignant meningiomas (19 tumors, including 4 malignant ones). The authors also reanalyzed 2 expression data sets generated on Affymetrix microarrays (n = 68, including 6 malignant ones; n = 56, including 3 malignant ones). A weighted gene coexpression network approach was used to identify coexpression modules associated with malignancy.

Results

At the genomic level, malignant meningiomas had more chromosomal losses than atypical and benign meningiomas, with average length of 528, 203, and 34 megabases, respectively. Monosomic loss of chromosome 22 was confirmed to be one of the primary chromosomal level abnormalities in all subtypes of meningiomas. At the transcriptome level, the authors identified 23 coexpression modules from the weighted gene coexpression network. Gene functional enrichment analysis highlighted a module with 356 genes that was highly related to tumorigenesis. Four intramodular hubs within the module (GAB2, KLF2, ID1, and CTF1) were oncogenic in other cancers such as leukemia. A putative meningioma tumor suppressor MN1 was also identified in this module with differential expression between malignant and benign meningiomas.

Conclusions

The authors' genomic and transcriptome analysis of meningiomas provides novel insights into the molecular pathways involved in malignant transformation of meningiomas, with implications for molecular heterogeneity of the disease.

Abbreviations used in this paper:

SNP = single-nucleotide polymorphism; UCLA = University of California, Los Angeles; WGCNA = weighted gene correlation network analysis.

Object

Meningiomas are among the most common primary adult brain tumors. Although typically benign, roughly 2%–5% display malignant pathological features. The key molecular pathways involved in malignant transformation remain to be determined.

Methods

Illumina expression microarrays were used to assess gene expression levels, and Illumina single-nucleotide polymorphism arrays were used to identify copy number variants in benign, atypical, and malignant meningiomas (19 tumors, including 4 malignant ones). The authors also reanalyzed 2 expression data sets generated on Affymetrix microarrays (n = 68, including 6 malignant ones; n = 56, including 3 malignant ones). A weighted gene coexpression network approach was used to identify coexpression modules associated with malignancy.

Results

At the genomic level, malignant meningiomas had more chromosomal losses than atypical and benign meningiomas, with average length of 528, 203, and 34 megabases, respectively. Monosomic loss of chromosome 22 was confirmed to be one of the primary chromosomal level abnormalities in all subtypes of meningiomas. At the transcriptome level, the authors identified 23 coexpression modules from the weighted gene coexpression network. Gene functional enrichment analysis highlighted a module with 356 genes that was highly related to tumorigenesis. Four intramodular hubs within the module (GAB2, KLF2, ID1, and CTF1) were oncogenic in other cancers such as leukemia. A putative meningioma tumor suppressor MN1 was also identified in this module with differential expression between malignant and benign meningiomas.

Conclusions

The authors' genomic and transcriptome analysis of meningiomas provides novel insights into the molecular pathways involved in malignant transformation of meningiomas, with implications for molecular heterogeneity of the disease.

Keywords: meningioma; copy number alteration; WGCNA; hub genes

Meningiomas are common intracranial tumors that arise from the meninges, the membranous covering of the brain and spinal cord. They constitute 25%–30% of all primary brain tumors, with an annual incidence of 4–6 per 100,000 persons.43,53 Most meningiomas are considered benign and grow slowly.51 The WHO classifies meningiomas into 3 subtypes: Grade I (benign meningioma), Grade II (atypical meningioma), and Grade III (malignant/anaplastic meningioma). The prevalence rates of these subtypes are roughly 80%–90%, 6%–15%, and 2%–5%, respectively, based on several reports.28,35 Malignant meningiomas are typically characterized by aggressive behavior, with infiltration into the adjacent brain, focal necrosis, and increased vascularity.

Atypical and malignant meningiomas are often associated with high recurrence rates, even following complete resection, and are associated with shorter progression-free and overall survival compared with their benign counterparts.43 Currently, the mechanisms underlying malignant transformation in meningiomas are poorly understood. Although several genetic studies have been performed on meningiomas to identify disease-susceptibility genes, few have focused on genetic and chromosomal alterations that may specifically contribute to the malignant transformation of these tumors.

Prior investigations have demonstrated an association between monosomic loss of chromosome 22 and meningioma formation.40 Other chromosome losses have been reported on 1p, 3q, 6q, 9p, 10q, 14q, 17p, 18p, 18q, and 22q.22 Relative levels of chromosomal abnormalities have been noted to correlate with higher meningioma grade.23 A recent study indicated that benign meningiomas with alterations in chromosome 14 may have a predilection for aggressive behavior and recurrence.36 A previous genome-wide association study revealed that a common variant at 10p12.31 near MLLT10 influences the risk of meningioma. This was the first genome-wide association study signal reporting on this disease and expanded our knowledge of genetic events to that initiate tumorigenesis.8 Studies on the NF2 gene in sporadic meningiomas suggest that approximately one-third to one-half of these tumors have an inactivating mutation, often accompanied by loss of the other allele.40 Other candidate tumor suppressor genes that have been suggested to promote meningioma formation include DAL-1, BAM22, MN1, and LARGE.43 However, current knowledge is insufficient to fully understand the genetic underpinnings of this disease.

Genome-wide expression profiling analysis has also identified differentially expressed genes and pathways in meningiomas.10,18,23 Several tumor suppressor genes are downregulated in malignant meningiomas, compared with benign and atypical ones.10 Furthermore, a specific deregulated pathway with enrichment of underexpressed genes regulated by the transcription factors Sp1 and AGP/EBP was identified in fibroblastic meningiomas compared with normal dura.18 A recent study suggested that meningiomas can be classified into 5 groups based on gene expression profiles.23

In the current study, we collected a sample set of 19 resected meningiomas comprising all 3 WHO grades, and we performed DNA copy number analysis using single-nucleotide polymorphism (SNP) genotyping arrays. In parallel, we also performed transcriptome analysis using gene expression microarrays to complement the copy number analysis. Considering the relatively small sample size, we also reanalyzed 2 publicly available meningioma expression data sets from the GEO database, totaling 9 additional malignant samples. Finally, we take advantage of weighted gene coexpression network analysis to identify dense coexpression subnetworks in meningiomas and detect oncogenic modules that are associated with malignant meningiomas. Our integrative genomics analysis yielded novel insights into the malignant transformation of meningiomas.

Methods

Gene Expression Microarray

The study was approved by the University of Southern California institutional review board. Written informed consent was obtained from all patients prior to collection of surgical tumor specimens. Following selection of benign, atypical, and malignant meningioma case samples with confirmed histopathology, total RNA samples were extracted from fresh-frozen meningioma tissue following a Qiagen RNA extraction protocol. A liquid nitrogen grinding method was used to prepare tissue samples for extraction. First, tissue samples were removed from the −80°C refrigerator, cut into fragments, and weighed immediately to ensure that no more than 100 mg tissue was present in every 1.5-ml tube. Next, liquid nitrogen was added into the tube as soon as possible, followed by tissue grinding. Next, 900 μl of QIAzol Lysis Reagent is added to each tube, vortexed for at least 60 seconds, and stored at room temperature (15°C–25°C) for 5 minutes. The standard protocol of the RNeasy Kit is then applied to each tissue sample, with the entire process being performed on ice (Rneasy Plus Universal Mini Kit, Cat. no. 73404). The quality of total RNA was assessed using an Experion RNA StdSens Chip on a Bio-Rad bioanalyzer. Genome-wide expression profiles of approximately 47,000 transcripts were quantified using an Illumina HumanHT-12 v4 Expression BeadChip. Raw data were processed and normalized without background subtraction using the Illumina GenomeStudio software suite. Among the 19 samples, mRNA from 11 samples (5 benign, 2 atypical, and 4 malignant meningiomas) were successfully extracted with an RNA integrity number value over 7. Considering the small sample size, we added 1 more malignant sample to perform the DNA SNP microarray assay.

Single-Nucleotide Polymorphism Array

DNA extraction was performed from 10 to 20 mg of tumor fragments using the DNeasy Blood & Tissue kit from (Qiagen), according to the supplied protocol. Genomic DNA (250 ng) was assayed by the HumanOmni-Express BeadChip SNP array (Illumina) with 730,525 markers, including 392,197 SNP markers. All microarray analysis was performed at the USC Epigenome Center. Raw microarray data were processed by the GenomeStudio Software (Illumina) to generate log R ratio and B allele frequency values for each marker on the array.

Data Analysis

Copy number alternations in 19 meningioma genomes were detected with a computational tool called OncoSNP from the SNP genotyping data.56 Both stromal contamination and intratumor heterogeneity models were activated in OncoSNP to perform a joint analysis and estimate the baseline level of stromal contamination and intratumor heterogeneity. This is the most accurate analysis at the expense of computing resources. The SNP genotyping arrays show “genomic wave patterns” in which signal intensity is correlated to local guanine-cytosine content, so the signal intensity value was adjusted by the genomic_wave.pl program in the PennCNV package.7,50 We used an R package LIMMA in Bioconductor to performed pairwise class distinction to identify genes differentially expressed among the benign, atypical, and malignant meningiomas (false discovery rate adjusted p value < 0.01).47 Since no differentially expressed genes were found in atypical and benign samples, we reanalyzed the microarray data by combining atypical and benign samples and comparing these to malignant samples. Two hundred eighty-eight differentially expressed genes were finally identified between malignant and benign-atypical samples. Copy number alternations and gene expression data were clustered using MultiExperiment Viewer software suite v.4.7.29 We used DAVID web server (http://david.abcc.ncifcrf.gov/) to test enrichment in gene sets with GO, SwissProt, and InterPro terms compared with the background list of all genes.15

We also downloaded raw data (as CEL files) for 2 meningioma whole-genome gene expression studies from the GEO database (accession nos. GSE16581 and GSE4780, respectively). These samples were assayed on the Affymetrix Human Genome U133 Plus 2.0 platform, so we used the Affymetrix Expression Console software to preprocess the CEL files with Robust MultiChip Analysis protocol and then exported the probeset-level expression levels as a tab-delimited text file. The LIMMA package in R47 was used to identify differentially expressed genes.

Weighted Gene Coexpression Network Analysis

To construct a network from gene expression data, we selected 3600 of the most varying genes and calculated the Pearson correlation for all pairs of selected genes. The correlation matrix was converted into an adjacency matrix with a power function, so that the connection strength between 2 genes xi and xj was defined as aij = |cor(xi, xj)|β. The parameter β was determined by the criterion that the resulting adjacency matrix approximately fit a scale-free topological feature according to a previously proposed model-fitting index.57 The model-fitting index of a perfect scale-free network is 1. Here, we chose the smallest value of β (β = 9), which is the minimum value required to make the model-fitting index 0.85 (Fig. 1).9,57

Fig. 1.
Fig. 1.

Analysis of network topology for various soft-thresholding powers. Left: The scale-free fit index (y axis) as a function of the soft-thresholding power (x axis). Right: The mean connectivity (in degree, y axis) as a function of the soft-thresholding power (x axis).

The adjacency matrix was further transformed into a topological overlap matrix, which captures not only the direct interaction between 2 genes but also their indirect interactions through all the other genes in the network. A similarity measure was defined:
article image
where
article image
is the node connectivity.42,57 Subsequently, 1 − TOMij was used as a distance matrix in the hierarchical clustering of the transcript units for module detection.42

Oncogenic Module Analysis

In the oncogenic module, member genes pairs with a Mutual Rank (http://coxpresdb.jp/help/mr.shtml) < 200 or Pearson correlation coefficient > 0.4 are considered connected.33 The core region was identified with MCODE, a plugin planted in Cytoscape.1,44 Protein-protein interaction information was queried from Michigan Molecular Interactions (MiMI).17 The network topology was visualized with Cytoscape.44

Results

Chromosomal Profiles of Genetic Alternations

We assayed copy number alteration profiles in 19 histopathologically verified human meningioma tumor samples: 10 benign (WHO Grade I), 5 atypical (WHO Grade II), and 4 malignant (WHO Grade III) meningiomas (Fig. 2). The average length of copy number gains was 423, 701, and 270 megabases in benign, atypical, and malignant meningiomas, respectively. In comparison, the average length of copy number loss was 34, 203, and 528 megabases in benign, atypical, and malignant meningiomas, respectively. In general, hierarchical clustering analysis classified the samples into 2 groups. Three malignant samples and 1 atypical sample with significant gain or loss of chromosomes were clustered into a small group. One malignant, 2 atypical, and all 10 benign samples with fewer chromosomal abnormalities were clustered into a separate larger group. The larger group could further be divided into 2 additional subgroups based on the relative degree of copy number amplification. The chromosomal copy number patterns suggested that higher grades of meningiomas demonstrate more variability in their genomic structures, especially with regard to genomic losses.

Fig. 2.
Fig. 2.

Chromosomal profiles of genomic copy number alternations and hierarchical clustering results of 19 meningioma samples. The 2 major clusters are colored by red and blue. Copy numbers of the genomic regions are marked from green (loss), black (normal), and red (gains). The regions of 13p, 14p, 15p, 21p, and 22p are ignored here, since they are not well covered by the SNP array data. The malignant samples are separated from benign/atypical samples, except for one atypical sample (Sample ID 351). A = atypical; B = benign; M = malignant.

High Chromosomal Arm–Level Abnormality Occurrences in High-Grade Meningiomas

Previous studies have suggested that a strong association between increased frequency of chromosomal losses with increased histological grades exists in meningiomas.34 Our copy number variation analysis corroborated the findings that malignant and atypical meningiomas present more chromosomal arm–level abnormalities than benign meningiomas (Fig. 2). Obvious deletion or amplification events in chromosomal arms were found in 2 of 4 malignant samples (Sample ID 249: amplification on 3 chromosomal arms, deletion on 9 chromosomal arms; Sample ID 255: deletion on 3 chromosomal arms) and 2 of 4 atypical samples (Sample ID 351: amplification on 17 chromosomal arms; Sample ID 678: amplification on 4 chromosomal arms, deletion on 6 chromosomal arms), whereas only 1 arm-level deletion occurred in 1 of 15 benign samples. Three of 4 malignant samples showed a monosomic loss of chromosome 4, but this event was not observed in atypical and benign samples. However, we also found a malignant sample (Sample ID 254) and an atypical sample (Sample ID 489) that showed a significant repression of chromosomal alternations compared with other samples from the same WHO grade (Fig. 2). Thus, other mechanisms may be involved in the development and malignant transformation of meningiomas.

Loss of Chromosome 22

Our results also showed that nearly half of meningioma samples have significant deletions in chromosome 22q, including 2 of 4 malignant samples, 2 of 4 atypical samples, and 5 of 15 benign samples (Fig. 2). To confirm this observation, we applied a statistical framework to distinguish genuine chromosomal aberrations from random background events.4 The results showed that loss of chromosome 22q was clearly statistically significant (Q value = 0; Table 1). In addition, loss of chromosome 1p was also detected with high statistical significance (Q value = 0.0484; Table 1). Loss of chromosome 22q may be a primary event with respect to chromosomal abnormalities in meningiomas, as demonstrated by the few benign samples with monosomy of chromosome 22 and loss of chromosome 22 in most high-grade meningiomas.23

TABLE 1:

Significance of chromosomal arm abbreviations predicted by GISTIC software*
ArmNo. of GenesAmplificationDeletion
Frequencyz-ScoreQ ValueFrequencyz-ScoreQ Value
1p14230−1.170.8870.262.680.0484†
1q12740−1.210.8870.161.160.433
2p5900−1.030.8870.161.740.202
2q10170−1.180.8870.110.4880.623
3p6610.06−0.09070.8870.110.8120.508
3q7320.110.760.8870.06−0.1270.77
4p2910−0.940.8870.162.070.152
4q6670−1.050.8870.161.660.21
5p1690.050.08510.8870−0.9570.887
5q10090.05−0.3760.8870−1.210.887
6p7680.06−0.2040.8870.06−0.2040.77
6q5500.06−0.03170.8870.110.8970.481
7p3900−1.030.8870.05−0.05510.77
7q8150−1.160.8870.05−0.2840.77
8p3700.110.9640.8870−0.9940.887
8q5540.110.8110.8870−1.050.887
9p2750−1.020.8870−1.020.887
9q7310−1.130.8870.05−0.2420.77
10p2480−0.9560.8870.111.080.458
10q8170−1.120.8870.110.6190.615
11p5780−1.020.8870.161.750.202
11q10480−1.190.8870.110.4690.623
12p3740.05−0.04550.8870−1.020.887
12q9430.05−0.3450.8870−1.190.887
13q4000−1.060.8870−1.060.887
14q8740−1.030.8870.263.260.0109
15q7760.06−0.2080.8870.06−0.2080.77
16p5850.060.0120.8870.171.850.202
16q4560.060.02090.8870.110.9730.46
17p4200.060.04170.8870.1110.46
17q10860.06−0.2930.8870.110.5270.623
18p930−0.9320.8870.050.1380.755
18q2650−0.9890.8870.050.0220.77
19p7210−1.030.8870.212.530.0552
19q10550−1.150.8870.161.330.36
20p2490.111.160.8870.060.1470.755
20q5280.110.9140.8870.06−0.01960.77
21q3240.172.140.6360.060.1620.755
22q5980−0.7730.8870.538.380†

GISTIC = Genomic Identification of Significant Targets in Cancer.

Values are below 0.05, indicating that the copy number variations in this chromosome are significantly associated with disease.

Gene Expression Analysis

Hierarchical clustering analysis indicated that 5 benign and atypical samples were represented by similar whole-genome expression patterns and were grouped closely together. The malignant samples, however, showed varying patterns and were not clustered together. These observations imply that gene expression profiles of malignant meningiomas are likely to be disordered compared with benign and atypical tumors (Fig. 3). We identified 260 differentially expressed genes between malignant and benign samples and 59 differentially expressed genes between malignant and atypical samples. Interestingly, we did not identify any differentially expressed genes between benign and atypical samples (Fig. 4). In a previous study, we also demonstrated that genome-wide methylation patterns cannot differentiate benign and atypical meningiomas but can easily differentiate malignant from benign/atypical samples.11 Considering the small sample size for the atypical group and the low power to detect differential expression, we combined the benign and atypical samples and subsequently identified 288 differentially expressed genes between the malignant and benign/atypical sample groups. Hierarchical clustering analysis confirmed that the differentially expressed genes were clearly divided into these malignant and nonmalignant groups (Fig. 5).

Fig. 3.
Fig. 3.

Hierarchical clustering results of transcriptome data using the 1000 most variably expressed genes. Genes with high, medium, and low expression values are colored by red, black, and green, respectively. The malignant samples are clearly separated from benign/atypical samples.

Fig. 4.
Fig. 4.

Venn diagram of differentially expressed genes among 3 classifications of meningiomas. Red circle shows the number of differentially expressed genes between malignant and benign meningiomas, purple circle shows the number of differentially expressed genes between malignant and atypical meningiomas, and green circle shows the number of differentially expressed genes between benign and atypical meningiomas.

Fig. 5.
Fig. 5.

Hierarchical clustering of 288 genes with differential expression between 5 malignant groups and 5 benign and 2 atypical groups. The cluster is color coded using red for upregulation, green for downregulation, and black for median expression.

Notably, a probable tumor suppressor gene, MN1, showed significant repression in all the malignant samples analyzed in our study (Table S1, http://bionas.usc.edu/meningioma/Table_S1.xlsx). Previous research also suggests that inactivation of the MN1 gene in meningiomas may contribute to their pathogenesis.24 We also found that the IGFBP5 gene (whose promoter transcription can be activated by the MN1 gene product) was significantly suppressed (Table S1).31 Interestingly, MN1 can both stimulate and inhibit transcription of this gene. Overexpression of MN1 was also reported in the development of acute myeloid leukemia.30

Considering the relatively small sample size, to further validate our gene expression result, we downloaded the raw data from 2 meningioma gene expression studies from GEO, hereafter referred to as the University of California, Los Angeles (UCLA) study (GSE16581, 63 samples)23 and the Scheck study (GSE4780, 56 samples). We identified 410 and 465 genes differentially expressed between benign and malignant samples from the UCLA and Scheck study, respectively. However, the number of overlapping genes among the 3 data sets is low: only 3 common genes were identified: LEPR, FXYD5, and KCNMA1. LEPR was associated with pituitary adenoma,12,20,46 FXYD5 was associated with a broad spectrum of cancers, and KCNMA1 was also associated with giloma.3,19,27,41 The number of overlapping genes between each pair of the 3 data sets was higher: 26 between the UCLA and USC studies, 34 between the Scheck and USC studies, and 35 between the UCLA and Scheck studies. Interestingly, we found that the MN1 gene was also differentially expressed in the UCLA data set. We next mapped the identified differentially expressed genes from the 3 data sets on the Cancer Genome Anatomy Project by DAVID and extracted the top 10 significantly enriched pathways (Table S2, http://bionas.usc.edu/meningioma/Table_S2.xlsx). For our data set, the top 4 pathways were all relevant to meningiomas, indicating a strong correlation of our data set and the Cancer Genome Anatomy Project. The first and third most significantly enriched pathways in the Scheck study were also related to meningiomas. However, no pathway associated with meningiomas was detected in the UCLA study.

Association Between DNA Copy Number and Gene Expression Levels

We calculated the Pearson correlation coefficient between DNA copy number and gene expression level. Our results showed that a certain proportion of genes were closely correlated (Pearson correlation coefficient > 0.5; Fig. 6). Gene ontology enrichment analysis suggested that genes related to mitochondrial function were significantly enriched in DNA copy number and correlation to high gene expression (Table S3, http://bionas.usc.edu/meningioma/Table_S3.xlsx). Analysis using Fisher's exact test also lent support to the hypothesis that these genes are enriched in chromosome regions with clear copy number abnormalities (chromosome 22q13.1 [p = 2.6e-5], chromosome 4p16.3 [p = 1.15e-8]; Table S3).

Fig. 6.
Fig. 6.

Distribution of DNA number and expression correlation. The x axis represents the Pearson correlation coefficient, and the y axis represents the frequency. The average is significantly higher than zero (p = 0).

Previous studies on meningiomas revealed several aberrant signaling pathways that may be involved in tumorigenesis.39 These pathways include membrane-associated protein 4.1 family–relevant pathways, vascular endothelial growth factor–involved angiogenic pathways, Hedgehog signaling pathway, MAPK and PI3K signaling pathways, Notch signaling pathway, and growth factor–and cytokine-induced pathways. We next examined our copy number variation and gene expression data with respect to the candidate meningioma pathways. We found that copy number alterations disrupt genes in some of these pathways and may result in gene expression changes. Besides MN1, a few genes in the PI3K/Akt pathway and some bone morphogenetic proteins were also differentially expressed between the benign/atypical and malignant samples. These results suggested that these pathways may play a role in tumor progression toward malignancy (Table S4, http://bionas.usc.edu/meningioma/Table_S4.xlsx).

An Oncogenic Module Identified by Weighted Gene Coexpression Networks

We further analyzed the gene expression data with the weighted gene correlation network analysis (WGCNA) algorithm, which has been widely applied in gene coexpression network construction and module detections.5,16,21,32,57 Twenty-three coexpression modules were identified (Fig. 7). Functional annotations of these modules were performed by analyzing the gene compositions within each module. Three modules were significantly enriched in genes with specific functional categories—Module I: glycoprotein/membrane/blood vessel development/cell migration, 356 genes; Module II: cell cycle, 97 genes; and Module III: synapse/membrane, 1289 genes. Of these, Module I is considered an oncogenic module since it is enriched in cancer-related functional processes. Not surprisingly, the differentially expressed gene MN1, as well as its target gene IGFBP5, was also contained within this oncogenic module.

Fig. 7.
Fig. 7.

Weighted gene coexpression network identified multiple functional modules. The upper portion of the figure is the cluster dendrogram of genes, and the lower section shows the modules of coexpressed genes. Genes within the same module are colored in the same color.

In a biological network, a few highly connected nodes that hold the whole network together are referred to as hubs.48 Similarly, genes with a high number of interactions are believed to play an important role in organizing the biological process of functional modules.2,13,48 We therefore zoomed in to identify hub genes in the oncogenic module. Unlike classic gene coexpression networks or protein-protein interaction networks, WGCNA defined a whole network connectivity measure (kTotal) for each gene, based on its Pearson correlation coefficient with all of the other genes, and an intramodular connectivity measure (kWithin) when only considering the connection strength of each gene with all the other genes within the specific module.57 In this study, the intramodular connectivity is far more meaningful than whole network connectivity in the oncogenic module. We hypothesized that intramodular hub genes may be associated with meningioma tumorigenesis. As expected, 2 intramodular hub genes (GAB2 and KLF2) with the highest values of connectivity were related to tumorigenesis, and 2 additional intramodular hub genes (ID1 and CTF1) are known oncogenes (Table 2). These 4 hub genes were also significantly differentially expressed between malignant and benign meningiomas (Table S1). Interestingly, the 4 hub genes are also related to leukemia.25,37,52,54 As described above, the meningioma tumor suppressor MN1 in this module was also associated with leukemia.26,30

TABLE 2:

Connectivity strength of genes in the oncogenic module*
GenekTotalkWithinkOutkDiff
GAB2†91.93722661.377602430.55962430.8179788
KLF2†88.97509761.06527827.90981933.15545911
SNORA59A93.75832560.520712133.23761327.28309894
PTGDR87.80106359.787816728.01324631.77457035
SNORA59B89.13884957.417304631.72154425.69576019
SLC47A193.17295157.340897935.83205321.50884502
SCARA592.7363957.275731235.46065921.81507259
CLIC284.4949357.051045827.44388529.60716128
PDZRN379.20021755.771497123.4287232.34277751
ENPP681.05655353.419829627.63672325.78310628
FGL280.55355953.192776127.36078325.83199294
ID1†93.68675552.194640941.49211410.70252651
THSD484.48684851.317381633.16946618.14791539
PDE9A77.03495950.906806926.12815224.77865462
CTF1†79.88053150.803479929.07705121.72642927
SULF281.41658950.29174831.12484119.16690698
SLC26A279.1633550.025187129.13816320.88702448
KCNMA181.47027950.009355931.46092418.54843236
LRRN4CL75.79444949.297935526.49651422.8014219
LOC10013127770.84504148.605818622.23922226.36659624
HSPB279.86516648.324683731.54048216.78420127
EPHX179.47603847.93751931.53851916.399
ST6GALNAC267.86502647.902207219.96281927.93938834
CTXN368.09697247.431434120.66553826.76589649
NHEDC271.41354247.23459124.17895123.05563971
WNK472.82252246.76134626.06117620.70016994
ST6GAL172.5784246.619731325.95868920.66104219
TMEM10072.37063246.070259126.30037319.76988595
PGR69.89052145.949960723.94056122.00940005
SEMA6C72.65128945.799664826.85162418.94804063
SLC7A288.09327845.568168342.525113.043058586
PENK60.90428845.154673915.74961429.40506027
DHRS371.61559545.021278226.59431718.426961
PGM566.51124944.417981522.09326722.32471413
SLPI83.39038943.447017239.9433713.503645876
TMOD172.80169242.881554729.92013812.96141702
SMAD961.96153542.689424519.27211123.41731393
MSRB373.59828642.330870531.26741611.06345497
PHGDH74.77743242.295363932.4820699.81329536
MAL258.9044442.05172516.85271525.19901013
IGFBP564.65026641.884094822.76617119.11792343
PLEKHA675.3531741.667463533.6857067.981757175
C9ORF12575.37608740.838380534.5377076.300673862
PODN63.91877140.611151923.30761917.30353258
PLCD470.39059340.60387829.78671510.81716307
AFAP1L168.03330340.44174727.59155612.8501905
LOC64634761.7055540.109815521.59573418.514081
SLC7A159.58177439.905496819.67627720.22921977
HS.56916252.54903639.829308112.71972727.10958066
SLC38A279.24787739.554677839.693199−0.138521551
ALPL69.24647339.497581229.7488929.748689668
LPAR561.82962239.272558722.55706316.71549586
ADRB267.23823639.255603827.98263211.27297214
A4GALT78.23973339.077628539.162105−0.084476175
C10ORF11661.50440538.560431922.94397315.61645855
SOD355.2449238.549978416.69494121.85503705
LOC64434853.88864538.288218515.60042722.68779164
LOC72871565.67253338.2312727.44126310.7900071
LOC64259058.95901838.116047720.8429717.27307792
HS.52592254.11184638.084814316.02703222.05778262
HCG452.0250538.073666113.95138424.12228233
MN169.2024838.047843231.1546376.893206166
CLEC1A71.98529237.763278334.2220143.541264242
FAM20A77.82080737.72971840.091089−2.361371218
ALDH266.54058237.33450229.206088.128422392
EHD152.96511137.111097615.85401421.257084
NMNAT263.61948736.261554827.3579328.903622913
KIAA119966.22330335.818954530.4043495.414605574
KLHL1360.5982635.689905924.90835410.78155167
LEPR55.7741835.685456220.08872415.59673201
SMS54.92561135.640565819.28504516.35552082
JAM255.79243435.540936720.25149715.28943996
TOB174.26579835.530148538.735649−3.205500765
ACSL560.5013735.370965625.13040410.24056162
SFRP275.1354135.11672640.018684−4.9019577
UGT3A249.86158835.06008414.80150420.25858018
HNMT54.07531634.977091819.09822415.87886739
PHLDB268.83188834.468246934.3636410.104606126
HS.28666656.06840634.358338321.71006812.64827032
SELENBP166.80843134.256029932.5524011.703628871
TBC1D863.33181534.051495729.280324.771176193
GPR454.44204934.000821920.44122713.55959473
ARL4D46.53918233.352027213.18715520.164872
SERPINB159.5975833.221875626.3757046.846171609
TCN257.20740933.042852224.1645578.878295542
FGFRL153.02940832.95839520.07101312.88738164
MAN1C155.22611232.51412622.7119869.802139895
OXTR62.35078532.045356830.3054281.739928946
C18ORF153.01127531.906712521.10456210.80215021
FXYD547.947531.598176116.34932415.24885214
TSC22D156.53599131.569399524.9665916.602808354
MBOAT146.9516831.28749415.66418615.62330761
PDE6A45.61697731.234202114.38277516.85142689
HMHA157.92842130.865078827.0633423.801736731
SULF147.6193830.769466116.84991413.91955223
CLCNKB43.96268530.75440213.20828317.54611929
FAT446.92860930.74017116.18843814.55173324
CALML446.7152630.718349615.9969114.72143933
TMEM2051.91054330.52106521.3894789.131586859
CYP4F1247.64448330.080701717.56378112.51692053
SLC16A955.30962129.960073925.3495474.610526902
TP53I1148.9197129.924110918.99559910.92851171
CHMP1B50.51734729.649819820.8675278.782292372
EYA150.15631129.039855921.1164557.923400575
RIPK441.63827128.828844512.80942716.01941788
SPTLC368.87706328.720610940.156452−11.43584107
GADD45B56.38784528.530376427.8574680.672908301
LZTS174.16472328.014245546.150477−18.13623175
PIK3C2B50.78973527.991446822.7982895.193158155
MET55.9537227.815318528.138402−0.32308342
UNC13B50.00587727.692869122.3130085.379861043
ACSF242.34623327.610597414.73563512.87496221
C15ORF5247.79962927.548883520.2507457.298138423
PLCE153.35136627.375552425.9758141.39973842
SIX249.80905427.204226522.6048284.599398799
TMEM19541.56852926.831196414.73733312.09386376
NR4A242.08857326.806342115.28223111.52411149
REEP152.47320126.723318825.7498820.973436371
SHB44.15913526.660098517.4990369.161062391
ADCY360.79641826.658108634.13831−7.480200999
ASTN247.36149226.541992320.81955.722492236
PGF38.77838626.319546612.4588413.86070701
TNFAIP8L346.17473226.260613319.9141196.346494281
CXCR752.39470325.896056226.498647−0.602590624
SLC20A253.49957425.651870827.847703−2.195832492
HS6ST241.85044225.512338116.3381049.174234209
EYA257.1816225.288230531.893389−6.605158749
AFAP145.7344525.051467320.6829834.368484348
PDE1A45.41118425.032928820.3782554.654673372
MYO5B40.07882524.81402215.2648039.549219448
HMGCLL138.51017924.639477813.87070110.7687769
SEMA3C47.94602824.391021923.5550060.836016004
NBL165.59420624.275457941.318748−17.04329014
BCAT253.4367323.901126529.535604−5.63447704
TEKT352.6864323.689679428.996751−5.307071562
CCDC3448.15128323.647223924.504059−0.856835355
APLP247.08097123.3654723.715501−0.350030569
C8ORF3436.9119523.216262113.6956889.52057455
SLC26A737.30534423.16626814.1390769.027191859
C10ORF2644.12955823.13156720.9979912.133576376
C2ORF4047.34822722.717150824.631076−1.913925259
ALX340.12367522.544628917.5790464.965582441
ADI137.44683922.478630214.9682097.51042108
RFTN142.38941922.440187819.9492312.490957076
LOC5410342.85835822.317418420.540941.776478254
RRAD37.15965721.963559115.1960986.76746107
CYP4X134.62360321.829630212.7939739.035657623
KLK136.74733221.718772915.0285596.69021392
MARCH332.66755121.455035811.21251510.24252109
RSPO335.18194421.422984413.7589597.664025135
SLC24A343.29628121.364106821.932174−0.568067224
MICAL140.46294821.287935319.1750122.112922943
TRPC332.61381421.140653311.4731619.667492816
FLJ2092033.06382820.967714512.0961138.871601067
ZNF68045.76102320.807062324.95396−4.146897924
CLCNKA32.41141520.669687511.7417278.927960491
ANO230.64206820.66837479.97369310.6946812
C10ORF3335.26124620.663220314.5980256.065194948
PDE5A34.44358120.656981313.7865996.870382082
SQRDL38.79087720.578129618.2127472.365382227
CTSZ36.01316820.557057615.4561115.100946672
PROS145.2220820.365007624.857072−4.492064393
LOC34297938.7876720.311344118.4763261.835018433
NCAPD349.29621320.217758929.078454−8.860695375
AFAP1L242.8035320.196261922.607268−2.411006256
NPTX239.7795920.027587119.7520030.27558387
BAIAP2L237.36507819.881250617.4838282.397422803
C10ORF7335.71563519.747092215.9685423.778549776
PCOLCE234.43599119.662686914.7733044.889382806
SLITRK436.20240619.412263516.7901432.622120752
TGFBR347.36205919.306790628.055268−8.748477346
BMP533.00115819.233965313.7671935.466772139
CHST439.85282619.164138120.688687−1.524549272
ITM2C48.76402119.020414729.743606−10.72319115
PSAT145.13918518.910632426.228553−7.31792061
ELOVL636.39240618.832291417.5601151.272176728
IGFBP332.75302918.752529714.0004994.75203061
CTSK58.1148718.596069939.5188−20.92273007
BOC36.2837218.278578518.0051410.273437118
C10ORF1141.19881518.182779723.016035−4.833255754
CDH133.94236518.107181415.8351832.271998064
ST3GAL533.77905118.102242515.6768082.425434052
IL3430.4903918.093920212.396475.697450237
NELL245.24785517.984239827.263615−9.279375304
GPR137C43.82508217.961536725.863546−7.902008965
SFRP127.27118417.88380749.3873778.496430893
LOC72821127.57957117.78382339.7957477.988075957
DEFB138.23668617.400509320.836176−3.435667055
ASAM46.90068517.311055529.58963−12.27857443
SHF26.36808216.97775559.3903267.587428967
COPZ248.2574616.80144831.456012−14.65456387
PLD132.687716.705592715.9821080.723485181
BTG231.20068916.468599514.732091.736509694
STARD828.94585316.288443112.657413.631033206
CHSY327.14349916.232967410.9105325.322435723
FOXO131.70005516.111757115.5882980.523458878
MPZL228.15970316.021574212.1381293.883445349
LOC38849428.6475215.909506512.7380133.171493282
TMEM16B22.76817815.6536037.1145758.539027983
FAM63B32.98917115.648405417.340766−1.692360696
CD5526.19304815.627542210.5655065.062036543
AHNAK233.93432815.601284518.333044−2.731759213
POL3S33.01049415.557466217.453027−1.895561148
PDGFRL38.2842715.389518722.894751−7.505232417
C14ORF7831.37441715.141577616.232839−1.091261493
DACT127.43704714.92469312.5123542.412339487
COL18A126.59864214.650965711.9476772.703289168
PTPN1328.14495314.394034213.7509190.643115561
FOXP131.03543214.292497916.742934−2.450436567
SETBP128.75464614.29070314.463943−0.173240412
MIR18530.75193914.242923916.509015−2.266090945
PPFIBP138.21645714.165179924.051277−9.886097078
PVRL345.4393114.084466131.354843−17.27037735
ADAMTSL126.51767114.023105612.4945651.528540236
EMILIN126.86886113.903404712.9654560.937948195
TMEM30B28.07724913.863188514.214061−0.35087206
LOC64797924.72542413.827080410.8983442.928736384
F1022.9762113.65890049.3173094.341591077
CYP3A526.12245813.549280412.5731780.976102708
RERG37.78642813.351686324.434741−11.08305495
KLHL1419.1986513.28029355.9183577.361936841
GMPR25.50925213.202838912.3064130.896425719
KLF437.57433913.170061524.404277−11.23421574
TMEM3724.22587713.040630911.1852461.855385222
ZNF50328.28862512.911599815.377025−2.465425451
FAM101A18.58435712.81869325.7656647.053029009
C11ORF8030.3789912.786641817.592348−4.805706326
ATOH819.69778712.75774266.9400445.81769829
C10ORF6522.30059212.69567169.6049213.090751068
TAGLN28.57118612.676485415.8947−3.218215036
EDG724.21711612.673604811.5435111.130093344
CTHRC135.1669812.542804222.624176−10.08137184
RASEF17.30033512.4778714.8224647.655406849
SALL421.86256812.39215319.4704152.921738591
C11ORF7026.18460712.237031213.947576−1.710544996
AGPAT922.62974712.072425510.5573221.515103936
SERTAD421.01561412.03765228.9779613.059690825
NPNT29.06140111.726025417.335375−5.609350123
INMT20.77393911.7230389.0509012.672137019
B4GALT128.0334511.583186616.450263−4.867076274
SMAD624.77593111.582568113.193362−1.610794394
ANLN38.97035711.534727827.43563−15.90090182
NTN320.82366211.41675129.4069112.009840221
PBX321.97424911.334140110.6401090.694030737
C1ORF13321.20822411.3213149.886911.434403989
SRGAP123.16362611.258624511.905002−0.646377281
BOP133.91715711.190537822.726619−11.53608163
FREM220.9720810.861607310.1104730.751134643
RHOB20.90382410.832152410.0716720.760480515
ANO641.97806410.644536631.333528−20.68899102
TJP323.07241910.60180212.470617−1.868814754
FGF1123.88424710.593212213.291034−2.697822252
ARHGAP1023.39738410.56365312.833731−2.270077847
HES144.48682510.54147433.945351−23.40387648
NQO122.69069210.538868712.151824−1.612955021
LOC38788228.9674310.536280118.43115−7.89486971
RPS6KA534.27425110.528758723.745492−13.21673371
KLK719.82666210.50580159.3208611.184940615
MEOX224.26207410.285971513.976103−3.690131443
MND132.63870110.153367122.485333−12.33196635
EDN128.69933310.088397218.610936−8.52253882
SIX126.21979310.004929116.214864−6.209935026
HSPB634.075869.928973324.146886−14.21791312
FGR19.3396029.91956769.4200340.499533463
SMOX26.8196259.904195116.91543−7.01123496
DSP19.1832979.89447499.2888220.605652459
NFKBIZ17.5763629.86519667.7111652.154031451
CDC719.6956479.86139189.8342560.027136139
CDH515.3420029.76815845.5738444.194314517
S100P16.6382059.67953626.9586692.720867288
MR123.2811259.645705313.63542−3.989714456
KCNK514.6996849.62022995.0794544.540775581
LPAR317.4386789.58337157.8553071.728064961
FZD719.9267069.536757410.389948−0.853190853
FABP322.2694819.531026812.738454−3.207427529
COL23A120.1852729.417149210.768122−1.350973071
ZBTB1620.4858599.297516411.188343−1.890826658
PDZD235.4358579.249279126.186577−16.93729831
STOX135.1033249.096999526.006324−16.90932479
C10ORF10521.2919218.999853612.292067−3.292213634
RNF2418.4540958.95910579.494989−0.535883418
C1ORF2148.7344788.872050739.862427−30.99037668
CFB27.5602488.865768718.69448−9.828710926
CHODL16.8201168.55121578.2689010.282315188
GCGR13.3887288.39065714.9980713.392586019
LOC10013355113.7715898.23389365.5376962.696197965
RCSD114.0709338.17942945.8915042.287925604
GDE120.054498.131233411.923257−3.792023541
HRCT117.2051178.00244749.202669−1.200221661
GALNTL418.8312257.849918210.981307−3.131388378
PRDM620.2456297.83621912.40941−4.573191005
SLC25A2915.6927947.44812528.244669−0.796544006
TIMP313.910597.44298616.4676040.975382132
KLK818.0090857.362790910.646294−3.283503569
SLC22A1814.0916177.20461286.8870050.317608148
CHCHD617.0187616.943616710.075144−3.131527701
FGF218.0146746.905337111.109337−4.203999999
LOC38858815.0422246.71829828.323926−1.60562803
TCF319.1228776.547554812.575322−6.027767481
GALNT1211.0822856.49592774.5863571.909570346
SLC13A413.1497746.30122366.848551−0.547327043
XYLT111.2254076.24359024.9818171.26177294
ENPP114.3395336.24057628.098957−1.858380392
MSLN12.8637676.23343416.630333−0.396899349
HS.57244413.6587886.21075387.448034−1.237280446
AOX117.0796736.189781710.889891−4.700109256
SLC6A209.6775736.1643313.5132422.651088869
IL18R113.320976.10672967.214241−1.107511012
PRG411.9068756.07721855.8296570.24756167
PLXNA212.02175.82866376.193036−0.364372752
MST413.7706835.75003758.020646−2.270608299
CYB5R216.5258445.745134310.78071−5.035575398
SPATA1814.3221055.73842828.583677−2.845248513
NXPH38.3734185.69066022.6827583.007902253
FMO19.6578895.470144.1877491.282391373
C6ORF2610.180495.44538864.7351010.710287393
GADL112.2565585.31581416.940744−1.624930024
CLDN79.5337335.15599344.3777390.778254026
AXUD110.3040965.11101765.193079−0.082061128
GINS426.0818544.908152421.173701−16.26554896
LHCGR8.317114.8842743.4328361.451438378
FRZB11.0677154.83881846.228896−1.390077976
FLJ303759.3238564.77186444.5519920.219872435
TNFRSF218.7219264.75172133.9702040.78151713
C12ORF4820.3866334.728598315.658035−10.92943636
CCDC6812.1829184.67674197.506176−2.829434386
FANCC22.8454534.425678918.419774−13.99409526
HS.4447859.2808584.35345484.927403−0.57394811
LPHN211.4267244.31286217.113862−2.8010001
NAPSB8.75744.27689684.480503−0.203605997
CPXM210.0823314.27527545.807055−1.531780019
HS.11993310.1182764.24386225.874414−1.630551804
MYO1B8.2138934.10764964.1062430.001406504
MLF1IP20.2991154.023520816.275595−12.25207393
LYPLAL18.0380863.99621144.041875−0.045663261
MFAP216.5841843.777984212.8062−9.028215801
CLIC39.9250673.71553526.209532−2.493996661
ABCG211.4505553.65942627.791129−4.131703004
ENG10.6492393.58113157.068107−3.486975962
COL8A29.1396343.57788185.561753−1.983870823
LYPD6B6.6943173.49474093.1995760.295165126
ANGPTL79.9852163.36297586.62224−3.259264177
GPSM216.0004063.347130312.653276−9.306145687
SIX421.3971283.292304518.104824−14.81251918
ZBTB7C9.0632493.21553785.847711−2.632172957
PRRX16.6122493.13302233.479227−0.346204832
TMEM15410.3025652.93475147.367814−4.433062307
COL12A110.963932.92581598.038115−5.11229867
HS.5750386.7187152.82897543.88974−1.060764291
MSX14.5993362.44433862.1549970.289341444
PLEKHG37.8578781.95327375.904604−3.951330301
ST3GAL68.2766741.75605596.520618−4.764561802
CD1773.7330921.54032442.192767−0.652442683
PPP1R3C3.5246281.48467652.039952−0.555275244
FAM107B4.8113331.36963133.441702−2.072070786
ADAMTSL52.4970670.66256211.834505−1.17194317

kTotal = total connectivity; kWithin = connectivity within the oncogenic module; kOut = connectivity outside the oncogenic module; kDiff = difference between kWithin and kOut.

Important hubs that are addressed in the text.

Validation of the Oncogenic Module With Other Biological Networks

We further validated the oncogenic module with a coexpressed gene database, COXPRESdb, which collected large mammalian coexpression information from publicly available GeneChip data.33 We created a subnetwork by mapping the genes of the oncogenic module onto the COXPRESdb database (Fig. 8). To further explore the oncogenic module in detail, we identified a highly connected core region of 27 genes with MCODE from the oncogenic module (Fig. 9; see Methods). The 27 core genes containing GAB2 and KLF2 are mostly related to cancer. ID1 and CTF1 are also directly connected with members of the core region (Fig. 7). Although MN1 is not directly connected to the core region, the protein-protein interaction networks show that MN1 can connect to KLF2 through EP300. These results show a layered structure in which GAB2 and KLF2 function as central hubs. Genes or gene products interacting with them can form a core region in the oncogenic module. Most of the genes in the module can associate with GAB2 and KLF2 through those in the core region.

Fig. 8.
Fig. 8.

Network topology of the oncogenic module. The core region is marked by yellow, hub genes within the core region are marked in green, and hub genes outside the core region are marked in white.

Fig. 9.
Fig. 9.

The network topology of the core region; hub genes within the core region are marked in green.

Discussion

In this study, we compared different subtypes of meningiomas at both the genomic and transcriptome levels. Our results identified chromosomal level alternations and differentially expressed genes between benign and malignant meningiomas. Gene coexpression network analysis revealed a module enriched with oncogenes that may represent a potential developmental pathway in meningioma pathogenesis and malignant transformation.

At the genomic level, copy number alteration analysis revealed that the degree of chromosomal abnormality correlates with the WHO meningioma grade. Monosomic loss of chromosome 22 may be considered a significant and early step in meningioma pathogenesis. Nearly half of our samples harbored significant deletions in chromosome 22q, a finding that substantiates previous reports.23,40 Loss of chromosome 22, however, was not apparent in all of the malignant samples; indeed only 2 of 4 exhibited deletion in this region. Loss of chromosome 4 was evident in 3 of 4 malignant samples, suggesting a potential role in malignant transformation. One malignant sample (Sample ID 254) demonstrated only minor changes in copy number across the entire genome. The sole chromosomal abnormality noted in this tumor was the gain of chromosome 21 (Fig. 1). Although these observations support a role for chromosomal-scale abnormalities in the development and progression of meningiomas, they reaffirm previously known challenges in extrapolating complicated mechanisms from a relatively small sample size.

We further investigated gene expression differences between nonmalignant and malignant meningiomas at the transcriptome level. The whole-genome expression profile was consistent among the nonmalignant (benign and atypical) samples. These patterns, however, were altered in the malignant samples, suggesting disruption of normal gene regulatory mechanisms. Furthermore, each malignant sample presented a unique expression pattern, implying heterogeneity in the factors involved with alterations in the gene regulatory system (Fig. 2). Additionally, based on analysis of previously identified abnormal signaling pathways in meningiomas, we found that genes such as MN1 and bone morphogenetic proteins are also differentially expressed between benign and malignant meningiomas.

We also identified some genes that displayed a high association between DNA copy number and gene expression. These genes were enriched in chromosome 1, 2, 3, 4, 16, and 22, which is consistent with regions harboring high levels of copy number alterations. As previously stated, loss of chromosome 22 was identified in the majority of samples, whereas loss of chromosome 4 was noted in 3 of 4 malignant meningiomas. Copy number alterations may therefore exert a dose-dependent effect on gene expression level.

Since biological network studies provided systematical understanding of the biological process, we also applied WGCNA to identify gene coexpression modules and potential oncogenes.2,6,45 The WGCNA approach has been widely used to identify potential oncogenic coexpression modules and predict the molecular targets and prognosis markers.14,16,38,55,58 For instance, Horvath et al. have successfully identified the gene ASPM as a molecular target for glioblastoma, another primary brain tumor.14 Ivliev et al. also identified transcriptional modules related to proastrocytic differentiation and sprout signaling in gliomas.16 In our study, among the 23 coexpression modules, we identified one that was strongly associated with tumorigenesis and could be considered an oncogenic module. This module contains 356 genes and is enriched in glycoprotein/membrane/blood vessel development and cell migration. We also identified the intramodular hub genes that are thought to contribute significantly to the topological architecture and biological process of functional modules. Four oncogene hubs (GAB2, KLF2, ID1, and CTF1) were detected among the top 15 most connected genes within the oncogenic module. Prior studies have suggested that these 4 genes are associated with leukemia.25,37,52,54 Interestingly, the meningioma 1 gene, often abbreviated as MN1, was also identified in this oncogenic module. MN1 was first discovered in 1995 from a meningioma patient with a balanced translocation that disrupts MN1 in its first exon. In this case, no expression of MN1 mRNA was observed, so MN1 was considered as a candidate tumor suppressor gene, affecting meningioma formation via inactivation.24 However, further research has suggested that overexpression of the MN1 gene can induce acute myeloid leukemia. Our study demonstrated differential MN1 gene expression between benign and malignant meningiomas. Other differentially expressed oncogenes found in this module were IGFBP5, MET, and TGFBR3. IGFBP5 is also upregulated by MN1 and linked to other cancer types, including glioblastoma multiforme.31,49

We further identified a core region within the oncogenic coexpression module, consisting of the hub genes GAB2 and KLF2. The additional hub genes ID1 and CTF1 were directly connected to these members of the core region. Although MN1 was not directly connected in the coexpression network, protein-protein interaction networks show that MN1 can connect to KLF2 via EP300. These results suggest that GAB2 and KLF2 may play an important role in the meningioma developmental pathway. Oncogenes interacting with GAB2 and KLF2 can function together as a small group. In turn, oncogenes indirectly connected to GAB2 and KLF2 via neighbors can form a large oncogenic module for meningiomas.

Considering the small sample size, we analyzed 2 relatively large gene expression data sets including 68 and 56 meningioma samples, respectively, although the number of malignant samples is still quite small (6 and 3, respectively). Functional enrichment analysis indicated that our data (the USC study) and those from the Scheck study were highly consistent with the Cancer Genome Anatomy Project database. We also applied WGCNA to these 2 data sets. However, WGCNA failed to find a significant meningioma-related module. A possible reason may be the imbalance of sample proportion from different subtypes because only 6 malignant samples of the total 68 samples are involved in the UCLA study and only 3 malignant samples of the total 56 samples are involved in the Scheck data set. This issue will lead to a bias in calculating the expression correlation between gene pairs. Therefore, the relative proportion of subtypes may be more important than sample size when applying WGCNA to construct gene coexpression networks.

Although limited by sample size, this systematic analysis of genomic and transcriptome differences between benign and malignant meningiomas revealed novel insight into the machinery of meningioma pathogenesis and transformation. If confirmed in larger studies, these alterations may serve as targets for novel diagnostic modalities and therapeutic interventions.

Conclusions

In summary, copy number alteration analysis confirmed the high frequency of chromosome 22 monosomy and the association between WHO grade and chromosomal abnormalities. Transcriptome analysis identified several genes that were highly differentially expressed between the benign and malignant groups, including a meningioma candidate gene, MN1. In addition, we used a weighted gene coexpression network analysis and identified an oncogenic module associated with meningiomas. Hub oncogenes in this module are also associated with the development of leukemia, implying potential common pathways between the 2 types of neoplastic diseases.

Acknowledgment

We thank members of the Wang lab for helpful comments and suggestions on the analytical strategies.

Disclosure

This project was supported in part by grant IRG-58-007-51 from the American Cancer Society to K.W., and in part by Southern California Clinical and Translational Science Institute (NIH/NCRR/NCATS) through grant KL2RR031991. The authors declare that they have no competing interests.

Author contributions to the study and manuscript preparation include the following. Conception and design: Mack, Zada, Wang. Acquisition of data: Gao, Russin, Zeng, He, Giannotta, Zada. Analysis and interpretation of data: Chang, Shi, Gao, Weisenberger, Zada, Wang. Drafting the article: Chang, He, Wang. Critically revising the article: Mack, Russin, Zada, Wang. Reviewed submitted version of manuscript: Mack, Zada, Wang. Approved the final version of the manuscript on behalf of all authors: Mack. Statistical analysis: Chang, Shi, Weisenberger, Wang. Administrative/technical/material support: Chen, Wang. Study supervision: Mack, Chen, Giannotta, Wang.

Disclaimer

This article provides links to supplemental material on an external website hosted on the file server in Dr. Kai Wang's lab at the University of Southern California. These links are provided for informational purposes and as a convenience to the reader. Provision of these links does not imply any endorsement or approval of the website or its contents by the JNS Publishing Group or the American Association of Neurosurgeons.

Supplemental online information: Supplementary tables are hosted on the file server in Dr. Kai Wang's lab at the University of Southern California, Los Angeles.

Supplementary Table 1 (“Expression signal value and copy number of differentially expressed genes”): http://bionas.usc.edu/meningioma/Table_S1.xlsx.

Supplementary Table 2 (“Comparison between 3 gene expression data sets”): http://bionas.usc.edu/meningioma/Table_S2.xlsx.

Supplementary Table 3 (“Genes with high correlation between copy number and expression level”): http://bionas.usc.edu/meningioma/Table_S3.xlsx.

Supplementary Table 4 (“Aberrant signaling pathways in meningiomas”): http://bionas.usc.edu/meningioma/Table_S4.xlsx.

References

  • 1↑

    Bader GD, & Hogue CW: An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 4:2, 2003

  • 2↑

    Barabási AL, & Oltvai ZN: Network biology: understanding the cell's functional organization. Nat Rev Genet 5:101–113, 2004

  • 3↑

    Batistatou A, , Charalabopoulos AK, , Scopa CD, , Nakanishi Y, , Kappas A, & Hirohashi S, et al.: Expression patterns of dysadherin and E-cadherin in lymph node metastases of colorectal carcinoma. Virchows Arch 448:763–767, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4↑

    Beroukhim R, , Getz G, , Nghiemphu L, , Barretina J, , Hsueh T, & Linhart D, et al.: Assessing the significance of chromosomal aberrations in cancer: methodology and application to glioma. Proc Natl Acad Sci U S A 104:20007–20012, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5↑

    Chang X, , Liu S, , Yu YT, , Li YX, & Li YY: Identifying modules of coexpressed transcript units and their organization of Saccharopolyspora erythraea from time series gene expression profiles. PLoS ONE 5:e12126, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6↑

    Chang X, , Wang Z, , Hao P, , Li YY, & Li YX: Exploring mitochondrial evolution and metabolism organization principles by comparative analysis of metabolic networks. Genomics 95:339–344, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7↑

    Diskin SJ, , Li M, , Hou C, , Yang S, , Glessner J, & Hakonarson H, et al.: Adjustment of genomic waves in signal intensities from whole-genome SNP genotyping platforms. Nucleic Acids Res 36:e126, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8↑

    Dobbins SE, , Broderick P, , Melin B, , Feychting M, , Johansen C, & Andersson U, et al.: Common variation at 10p12.31 near MLLT10 influences meningioma risk. Nat Genet 43:825–827, 2011

  • 9↑

    Emilsson V, , Thorleifsson G, , Zhang B, , Leonardson AS, , Zink F, & Zhu J, et al.: Genetics of gene expression and its effect on disease. Nature 452:423–428, 2008

  • 10↑

    Fèvre-Montange M, , Champier J, , Durand A, , Wierinckx A, , Honnorat J, & Guyotat J, et al.: Microarray gene expression profiling in meningiomas: differential expression according to grade or histopathological subtype. Int J Oncol 35:1395–1407, 2009

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11↑

    Gao F, , Shi L, , Russin J, , Zeng L, , Chang X, & He S, et al.: DNA methylation in the malignant transformation of meningiomas. PLoS ONE 8:e54114, 2013

  • 12↑

    Giusti M, , Bocca L, , Florio T, , Corsaro A, , Spaziante R, & Schettini G, et al.: In vitro effect of human recombinant leptin and expression of leptin receptors on growth hormone-secreting human pituitary adenomas. Clin Endocrinol (Oxf) 57:449–455, 2002

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13↑

    Han JD, , Bertin N, , Hao T, , Goldberg DS, , Berriz GF, & Zhang LV, et al.: Evidence for dynamically organized modularity in the yeast protein-protein interaction network. Nature 430:88–93, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14↑

    Horvath S, , Zhang B, , Carlson M, , Lu KV, , Zhu S, & Felciano RM, et al.: Analysis of oncogenic signaling networks in glioblastoma identifies ASPM as a molecular target. Proc Natl Acad Sci U S A 103:17402–17407, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15↑

    Huang W, , Sherman BT, & Lempicki RA: Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4:44–57, 2009

  • 16↑

    Ivliev AE, , 't Hoen PA, & Sergeeva MG: Coexpression network analysis identifies transcriptional modules related to proastrocytic differentiation and sprouty signaling in glioma. Cancer Res 70:10060–10070, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17↑

    Jayapandian M, , Chapman A, , Tarcea VG, , Yu C, , Elkiss A, & Ianni A, et al.: Michigan Molecular Interactions (MiMI): putting the jigsaw puzzle together. Nucleic Acids Res 35:Database issue D566–D571, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18↑

    Keller A, , Ludwig N, , Backes C, , Romeike BF, , Comtesse N, & Henn W, et al.: Genome wide expression profiling identifies specific deregulated pathways in meningioma. Int J Cancer 124:346–351, 2009

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19↑

    Kraft R, , Krause P, , Jung S, , Basrai D, , Liebmann L, & Bolz J, et al.: BK channel openers inhibit migration of human glioma cells. Pflugers Arch 446:248–255, 2003

  • 20↑

    Kristiansen MT, , Clausen LR, , Nielsen S, , Blaabjerg O, , Ledet T, & Rasmussen LM, et al.: Expression of leptin receptor isoforms and effects of leptin on the proliferation and hormonal secretion in human pituitary adenomas. Horm Res 62:129–136, 2004

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21↑

    Langfelder P, & Horvath S: WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9:559, 2008

  • 22↑

    Lee JY, , Finkelstein S, , Hamilton RL, , Rekha R, , King JT Jr, & Omalu B: Loss of heterozygosity analysis of benign, atypical, and anaplastic meningiomas. Neurosurgery 55:1163–1173, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23↑

    Lee Y, , Liu J, , Patel S, , Cloughesy T, , Lai A, & Farooqi H, et al.: Genomic landscape of meningiomas. Brain Pathol 20:751–762, 2010

  • 24↑

    Lekanne Deprez RH, , Riegman PH, , Groen NA, , Warringa UL, , van Biezen NA, & Molijn AC, et al.: Cloning and characterization of MN1, a gene from chromosome 22q11, which is disrupted by a balanced translocation in a meningioma. Oncogene 10:1521–1528, 1995

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25↑

    Lister J, , Forrester WC, & Baron MH: Inhibition of an erythroid differentiation switch by the helix-loop-helix protein Id1. J Biol Chem 270:17939–17946, 1995

  • 26↑

    Liu T, , Jankovic D, , Brault L, , Ehret S, , Baty F, & Stavropoulou V, et al.: Functional characterization of high levels of meningioma 1 as collaborating oncogene in acute leukemia. Leukemia 24:601–612, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27↑

    Liu X, , Chang Y, , Reinhart PH, , Sontheimer H, & Chang Y: Cloning and characterization of glioma BK, a novel BK channel isoform highly expressed in human glioma cells. J Neurosci 22:1840–1849, 2002

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28↑

    Louis DN, , Scheithauer BW, , Budka H, , von Deimling A, & Kepes JJ, Meningiomas. Kleihues P, & Cavenee WK: World Health Organization Classification of Tumours. Pathology and Genetics: Tumours of the Nervous System Lyon, IARC Press, 2000. 176–189

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29↑

    Mar JC, , Wells CA, & Quackenbush J: Defining an informativeness metric for clustering gene expression data. Bioinformatics 27:1094–1100, 2011

  • 30↑

    Meester-Smoor MA, , Janssen MJ, , Grosveld GC, , de Klein A, , van IJcken WF, & Douben H, et al.: MN1 affects expression of genes involved in hematopoiesis and can enhance as well as inhibit RAR/RXR-induced gene expression. Carcinogenesis 29:2025–2034, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31↑

    Meester-Smoor MA, , Molijn AC, , Zhao Y, , Groen NA, , Groffen CA, & Boogaard M, et al.: The MN1 oncoprotein activates transcription of the IGFBP5 promoter through a CACCC-rich consensus sequence. J Mol Endocrinol 38:113–125, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32↑

    Mumford JA, , Horvath S, , Oldham MC, , Langfelder P, , Geschwind DH, & Poldrack RA: Detecting network modules in fMRI time series: a weighted network analysis approach. Neuroimage 52:1465–1476, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33↑

    Obayashi T, , Hayashi S, , Shibaoka M, , Saeki M, , Ohta H, & Kinoshita K: COXPRESdb: a database of coexpressed gene networks in mammals. Nucleic Acids Res 36:Database issue D77–D82, 2008

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 34↑

    Ozaki S, , Nishizaki T, , Ito H, & Sasaki K: Comparative genomic hybridization analysis of genetic alterations associated with malignant progression of meningioma. J Neurooncol 41:167–174, 1999

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 35↑

    Perry A, , Gutmann DH, & Reifenberger G: Molecular pathogenesis of meningiomas. J Neurooncol 70:183–202, 2004

  • 36↑

    Pfisterer WK, , Coons SW, , Aboul-Enein F, , Hendricks WP, , Scheck AC, & Preul MC: Implicating chromosomal aberrations with meningioma growth and recurrence: results from FISH and MIB-I analysis of grades I and II meningioma tissue. J Neurooncol 87:43–50, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 37↑

    Plun-Favreau H, , Perret D, , Diveu C, , Froger J, , Chevalier S, & Lelièvre E, et al.: Leukemia inhibitory factor (LIF), cardiotrophin-1, and oncostatin M share structural binding determinants in the immunoglobulin-like domain of LIF receptor. J Biol Chem 278:27169–27179, 2003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 38↑

    Presson AP, , Yoon NK, , Bagryanova L, , Mah V, , Alavi M, & Maresh EL, et al.: Protein expression based multimarker analysis of breast cancer samples. BMC Cancer 11:230, 2011

  • 39↑

    Ragel BT, & Jensen RL: Aberrant signaling pathways in meningiomas. J Neurooncol 99:315–324, 2010

  • 40↑

    Ragel BT, & Jensen RL: Molecular genetics of meningiomas. Neurosurg Focus 19:5 E9, 2005

  • 41↑

    Ransom CB, , Liu X, & Sontheimer H: BK channels in human glioma cells have enhanced calcium sensitivity. Glia 38:281–291, 2002

  • 42↑

    Ravasz E, , Somera AL, , Mongru DA, , Oltvai ZN, & Barabási AL: Hierarchical organization of modularity in metabolic networks. Science 297:1551–1555, 2002

  • 43↑

    Riemenschneider MJ, , Perry A, & Reifenberger G: Histological classification and molecular genetics of meningiomas. Lancet Neurol 5:1045–1054, 2006

  • 44↑

    Shannon P, , Markiel A, , Ozier O, , Baliga NS, , Wang JT, & Ramage D, et al.: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13:2498–2504, 2003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 45↑

    Sharan R, & Ideker T: Modeling cellular machinery through biological network comparison. Nat Biotechnol 24:427–433, 2006

  • 46↑

    Shimon I, , Yan X, , Magoffin DA, , Friedman TC, & Melmed S: Intact leptin receptor is selectively expressed in human fetal pituitary and pituitary adenomas and signals human fetal pituitary growth hormone secretion. J Clin Endocrinol Metab 83:4059–4064, 1998

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 47↑

    Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3:Article3 2004

  • 48↑

    Vidal M, , Cusick ME, & Barabási AL: Interactome networks and human disease. Cell 144:986–998, 2011

  • 49↑

    Wang H, , Wang H, , Zhang W, & Fuller GN: Overexpression of IGFBP5, but not IGFBP3, correlates with the histologic grade of human diffuse glioma: a tissue microarray and immunohistochemical study. Technol Cancer Res Treat 5:195–199, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 50↑

    Wang K, , Li M, , Hadley D, , Liu R, , Glessner J, & Grant SF, et al.: PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data. Genome Res 17:1665–1674, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 51↑

    Wilson CB: Meningiomas: genetics, malignancy, and the role of radiation in induction and treatment. The Richard C. Schneider Lecture. J Neurosurg 81:666–675, 1994

  • 52↑

    Wöhrle FU, , Daly RJ, & Brummer T: Function, regulation and pathological roles of the Gab/DOS docking proteins. Cell Commun Signal 7:22, 2009

  • 53↑

    Wrobel G, , Roerig P, , Kokocinski F, , Neben K, , Hahn M, & Reifenberger G, et al.: Microarray-based gene expression profiling of benign, atypical and anaplastic meningiomas identifies novel genes associated with meningioma progression. Int J Cancer 114:249–256, 2005

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 54↑

    Wu J, & Lingrel JB: KLF2 inhibits Jurkat T leukemia cell growth via upregulation of cyclin-dependent kinase inhibitor p21WAF1/CIP1. Oncogene 23:8088–8096, 2004

  • 55↑

    Xiang Y, , Zhang CQ, & Huang K: Predicting glioblastoma prognosis networks using weighted gene co-expression network analysis on TCGA data. BMC Bioinformatics 13 Suppl 2 S12, 2012

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 56↑

    Yau C, , Mouradov D, , Jorissen RN, , Colella S, , Mirza G, & Steers G, et al.: A statistical approach for detecting genomic aberrations in heterogeneous tumor samples from single nucleotide polymorphism genotyping data. Genome Biol 11:R92, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 57↑

    Zhang B, & Horvath S: A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4:Article17 2005

  • 58↑

    Zhu XL, , Ai ZH, , Wang J, , Xu YL, & Teng YC: Weighted gene co-expression network analysis in identification of endometrial cancer prognosis markers. Asian Pac J Cancer Prev 13:4607–4611, 2012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • Collapse
  • Expand
Volume 35: Issue 6

Contributor Notes

Address correspondence to: William J. Mack, M.D., Department of Neurosurgery, Keck School of Medicine, University of Southern California, Los Angeles, 1200 N. State St., Ste. 3300, Los Angeles, CA 90033. email: mackw@usc.edu.

Please include this information when citing this paper: DOI: 10.3171/2013.10.FOCUS13326.

  • View in gallery

    Analysis of network topology for various soft-thresholding powers. Left: The scale-free fit index (y axis) as a function of the soft-thresholding power (x axis). Right: The mean connectivity (in degree, y axis) as a function of the soft-thresholding power (x axis).

  • View in gallery

    Chromosomal profiles of genomic copy number alternations and hierarchical clustering results of 19 meningioma samples. The 2 major clusters are colored by red and blue. Copy numbers of the genomic regions are marked from green (loss), black (normal), and red (gains). The regions of 13p, 14p, 15p, 21p, and 22p are ignored here, since they are not well covered by the SNP array data. The malignant samples are separated from benign/atypical samples, except for one atypical sample (Sample ID 351). A = atypical; B = benign; M = malignant.

  • View in gallery

    Hierarchical clustering results of transcriptome data using the 1000 most variably expressed genes. Genes with high, medium, and low expression values are colored by red, black, and green, respectively. The malignant samples are clearly separated from benign/atypical samples.

  • View in gallery

    Venn diagram of differentially expressed genes among 3 classifications of meningiomas. Red circle shows the number of differentially expressed genes between malignant and benign meningiomas, purple circle shows the number of differentially expressed genes between malignant and atypical meningiomas, and green circle shows the number of differentially expressed genes between benign and atypical meningiomas.

  • View in gallery

    Hierarchical clustering of 288 genes with differential expression between 5 malignant groups and 5 benign and 2 atypical groups. The cluster is color coded using red for upregulation, green for downregulation, and black for median expression.

  • View in gallery

    Distribution of DNA number and expression correlation. The x axis represents the Pearson correlation coefficient, and the y axis represents the frequency. The average is significantly higher than zero (p = 0).

  • View in gallery

    Weighted gene coexpression network identified multiple functional modules. The upper portion of the figure is the cluster dendrogram of genes, and the lower section shows the modules of coexpressed genes. Genes within the same module are colored in the same color.

  • View in gallery

    Network topology of the oncogenic module. The core region is marked by yellow, hub genes within the core region are marked in green, and hub genes outside the core region are marked in white.

  • View in gallery

    The network topology of the core region; hub genes within the core region are marked in green.

Analysis of network topology for various soft-thresholding powers. Left: The scale-free fit index (y axis) as a function of the soft-thresholding power (x axis). Right: The mean connectivity (in degree, y axis) as a function of the soft-thresholding power (x axis).

Chromosomal profiles of genomic copy number alternations and hierarchical clustering results of 19 meningioma samples. The 2 major clusters are colored by red and blue. Copy numbers of the genomic regions are marked from green (loss), black (normal), and red (gains). The regions of 13p, 14p, 15p, 21p, and 22p are ignored here, since they are not well covered by the SNP array data. The malignant samples are separated from benign/atypical samples, except for one atypical sample (Sample ID 351). A = atypical; B = benign; M = malignant.

Hierarchical clustering results of transcriptome data using the 1000 most variably expressed genes. Genes with high, medium, and low expression values are colored by red, black, and green, respectively. The malignant samples are clearly separated from benign/atypical samples.

Venn diagram of differentially expressed genes among 3 classifications of meningiomas. Red circle shows the number of differentially expressed genes between malignant and benign meningiomas, purple circle shows the number of differentially expressed genes between malignant and atypical meningiomas, and green circle shows the number of differentially expressed genes between benign and atypical meningiomas.

Hierarchical clustering of 288 genes with differential expression between 5 malignant groups and 5 benign and 2 atypical groups. The cluster is color coded using red for upregulation, green for downregulation, and black for median expression.

Distribution of DNA number and expression correlation. The x axis represents the Pearson correlation coefficient, and the y axis represents the frequency. The average is significantly higher than zero (p = 0).

Weighted gene coexpression network identified multiple functional modules. The upper portion of the figure is the cluster dendrogram of genes, and the lower section shows the modules of coexpressed genes. Genes within the same module are colored in the same color.

Network topology of the oncogenic module. The core region is marked by yellow, hub genes within the core region are marked in green, and hub genes outside the core region are marked in white.

The network topology of the core region; hub genes within the core region are marked in green.
  • 1

    Bader GD, & Hogue CW: An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 4:2, 2003

  • 2

    Barabási AL, & Oltvai ZN: Network biology: understanding the cell's functional organization. Nat Rev Genet 5:101–113, 2004

  • 3

    Batistatou A, , Charalabopoulos AK, , Scopa CD, , Nakanishi Y, , Kappas A, & Hirohashi S, et al.: Expression patterns of dysadherin and E-cadherin in lymph node metastases of colorectal carcinoma. Virchows Arch 448:763–767, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Beroukhim R, , Getz G, , Nghiemphu L, , Barretina J, , Hsueh T, & Linhart D, et al.: Assessing the significance of chromosomal aberrations in cancer: methodology and application to glioma. Proc Natl Acad Sci U S A 104:20007–20012, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5

    Chang X, , Liu S, , Yu YT, , Li YX, & Li YY: Identifying modules of coexpressed transcript units and their organization of Saccharopolyspora erythraea from time series gene expression profiles. PLoS ONE 5:e12126, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Chang X, , Wang Z, , Hao P, , Li YY, & Li YX: Exploring mitochondrial evolution and metabolism organization principles by comparative analysis of metabolic networks. Genomics 95:339–344, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7

    Diskin SJ, , Li M, , Hou C, , Yang S, , Glessner J, & Hakonarson H, et al.: Adjustment of genomic waves in signal intensities from whole-genome SNP genotyping platforms. Nucleic Acids Res 36:e126, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Dobbins SE, , Broderick P, , Melin B, , Feychting M, , Johansen C, & Andersson U, et al.: Common variation at 10p12.31 near MLLT10 influences meningioma risk. Nat Genet 43:825–827, 2011

  • 9

    Emilsson V, , Thorleifsson G, , Zhang B, , Leonardson AS, , Zink F, & Zhu J, et al.: Genetics of gene expression and its effect on disease. Nature 452:423–428, 2008

  • 10

    Fèvre-Montange M, , Champier J, , Durand A, , Wierinckx A, , Honnorat J, & Guyotat J, et al.: Microarray gene expression profiling in meningiomas: differential expression according to grade or histopathological subtype. Int J Oncol 35:1395–1407, 2009

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Gao F, , Shi L, , Russin J, , Zeng L, , Chang X, & He S, et al.: DNA methylation in the malignant transformation of meningiomas. PLoS ONE 8:e54114, 2013

  • 12

    Giusti M, , Bocca L, , Florio T, , Corsaro A, , Spaziante R, & Schettini G, et al.: In vitro effect of human recombinant leptin and expression of leptin receptors on growth hormone-secreting human pituitary adenomas. Clin Endocrinol (Oxf) 57:449–455, 2002

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Han JD, , Bertin N, , Hao T, , Goldberg DS, , Berriz GF, & Zhang LV, et al.: Evidence for dynamically organized modularity in the yeast protein-protein interaction network. Nature 430:88–93, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14

    Horvath S, , Zhang B, , Carlson M, , Lu KV, , Zhu S, & Felciano RM, et al.: Analysis of oncogenic signaling networks in glioblastoma identifies ASPM as a molecular target. Proc Natl Acad Sci U S A 103:17402–17407, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15

    Huang W, , Sherman BT, & Lempicki RA: Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4:44–57, 2009

  • 16

    Ivliev AE, , 't Hoen PA, & Sergeeva MG: Coexpression network analysis identifies transcriptional modules related to proastrocytic differentiation and sprouty signaling in glioma. Cancer Res 70:10060–10070, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17

    Jayapandian M, , Chapman A, , Tarcea VG, , Yu C, , Elkiss A, & Ianni A, et al.: Michigan Molecular Interactions (MiMI): putting the jigsaw puzzle together. Nucleic Acids Res 35:Database issue D566–D571, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 18

    Keller A, , Ludwig N, , Backes C, , Romeike BF, , Comtesse N, & Henn W, et al.: Genome wide expression profiling identifies specific deregulated pathways in meningioma. Int J Cancer 124:346–351, 2009

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Kraft R, , Krause P, , Jung S, , Basrai D, , Liebmann L, & Bolz J, et al.: BK channel openers inhibit migration of human glioma cells. Pflugers Arch 446:248–255, 2003

  • 20

    Kristiansen MT, , Clausen LR, , Nielsen S, , Blaabjerg O, , Ledet T, & Rasmussen LM, et al.: Expression of leptin receptor isoforms and effects of leptin on the proliferation and hormonal secretion in human pituitary adenomas. Horm Res 62:129–136, 2004

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Langfelder P, & Horvath S: WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics 9:559, 2008

  • 22

    Lee JY, , Finkelstein S, , Hamilton RL, , Rekha R, , King JT Jr, & Omalu B: Loss of heterozygosity analysis of benign, atypical, and anaplastic meningiomas. Neurosurgery 55:1163–1173, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    Lee Y, , Liu J, , Patel S, , Cloughesy T, , Lai A, & Farooqi H, et al.: Genomic landscape of meningiomas. Brain Pathol 20:751–762, 2010

  • 24

    Lekanne Deprez RH, , Riegman PH, , Groen NA, , Warringa UL, , van Biezen NA, & Molijn AC, et al.: Cloning and characterization of MN1, a gene from chromosome 22q11, which is disrupted by a balanced translocation in a meningioma. Oncogene 10:1521–1528, 1995

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Lister J, , Forrester WC, & Baron MH: Inhibition of an erythroid differentiation switch by the helix-loop-helix protein Id1. J Biol Chem 270:17939–17946, 1995

  • 26

    Liu T, , Jankovic D, , Brault L, , Ehret S, , Baty F, & Stavropoulou V, et al.: Functional characterization of high levels of meningioma 1 as collaborating oncogene in acute leukemia. Leukemia 24:601–612, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27

    Liu X, , Chang Y, , Reinhart PH, , Sontheimer H, & Chang Y: Cloning and characterization of glioma BK, a novel BK channel isoform highly expressed in human glioma cells. J Neurosci 22:1840–1849, 2002

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Louis DN, , Scheithauer BW, , Budka H, , von Deimling A, & Kepes JJ, Meningiomas. Kleihues P, & Cavenee WK: World Health Organization Classification of Tumours. Pathology and Genetics: Tumours of the Nervous System Lyon, IARC Press, 2000. 176–189

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 29

    Mar JC, , Wells CA, & Quackenbush J: Defining an informativeness metric for clustering gene expression data. Bioinformatics 27:1094–1100, 2011

  • 30

    Meester-Smoor MA, , Janssen MJ, , Grosveld GC, , de Klein A, , van IJcken WF, & Douben H, et al.: MN1 affects expression of genes involved in hematopoiesis and can enhance as well as inhibit RAR/RXR-induced gene expression. Carcinogenesis 29:2025–2034, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 31

    Meester-Smoor MA, , Molijn AC, , Zhao Y, , Groen NA, , Groffen CA, & Boogaard M, et al.: The MN1 oncoprotein activates transcription of the IGFBP5 promoter through a CACCC-rich consensus sequence. J Mol Endocrinol 38:113–125, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 32

    Mumford JA, , Horvath S, , Oldham MC, , Langfelder P, , Geschwind DH, & Poldrack RA: Detecting network modules in fMRI time series: a weighted network analysis approach. Neuroimage 52:1465–1476, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 33

    Obayashi T, , Hayashi S, , Shibaoka M, , Saeki M, , Ohta H, & Kinoshita K: COXPRESdb: a database of coexpressed gene networks in mammals. Nucleic Acids Res 36:Database issue D77–D82, 2008

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 34

    Ozaki S, , Nishizaki T, , Ito H, & Sasaki K: Comparative genomic hybridization analysis of genetic alterations associated with malignant progression of meningioma. J Neurooncol 41:167–174, 1999

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 35

    Perry A, , Gutmann DH, & Reifenberger G: Molecular pathogenesis of meningiomas. J Neurooncol 70:183–202, 2004

  • 36

    Pfisterer WK, , Coons SW, , Aboul-Enein F, , Hendricks WP, , Scheck AC, & Preul MC: Implicating chromosomal aberrations with meningioma growth and recurrence: results from FISH and MIB-I analysis of grades I and II meningioma tissue. J Neurooncol 87:43–50, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 37

    Plun-Favreau H, , Perret D, , Diveu C, , Froger J, , Chevalier S, & Lelièvre E, et al.: Leukemia inhibitory factor (LIF), cardiotrophin-1, and oncostatin M share structural binding determinants in the immunoglobulin-like domain of LIF receptor. J Biol Chem 278:27169–27179, 2003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 38

    Presson AP, , Yoon NK, , Bagryanova L, , Mah V, , Alavi M, & Maresh EL, et al.: Protein expression based multimarker analysis of breast cancer samples. BMC Cancer 11:230, 2011

  • 39

    Ragel BT, & Jensen RL: Aberrant signaling pathways in meningiomas. J Neurooncol 99:315–324, 2010

  • 40

    Ragel BT, & Jensen RL: Molecular genetics of meningiomas. Neurosurg Focus 19:5 E9, 2005

  • 41

    Ransom CB, , Liu X, & Sontheimer H: BK channels in human glioma cells have enhanced calcium sensitivity. Glia 38:281–291, 2002

  • 42

    Ravasz E, , Somera AL, , Mongru DA, , Oltvai ZN, & Barabási AL: Hierarchical organization of modularity in metabolic networks. Science 297:1551–1555, 2002

  • 43

    Riemenschneider MJ, , Perry A, & Reifenberger G: Histological classification and molecular genetics of meningiomas. Lancet Neurol 5:1045–1054, 2006

  • 44

    Shannon P, , Markiel A, , Ozier O, , Baliga NS, , Wang JT, & Ramage D, et al.: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13:2498–2504, 2003

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 45

    Sharan R, & Ideker T: Modeling cellular machinery through biological network comparison. Nat Biotechnol 24:427–433, 2006

  • 46

    Shimon I, , Yan X, , Magoffin DA, , Friedman TC, & Melmed S: Intact leptin receptor is selectively expressed in human fetal pituitary and pituitary adenomas and signals human fetal pituitary growth hormone secretion. J Clin Endocrinol Metab 83:4059–4064, 1998

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 47

    Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3:Article3 2004

  • 48

    Vidal M, , Cusick ME, & Barabási AL: Interactome networks and human disease. Cell 144:986–998, 2011

  • 49

    Wang H, , Wang H, , Zhang W, & Fuller GN: Overexpression of IGFBP5, but not IGFBP3, correlates with the histologic grade of human diffuse glioma: a tissue microarray and immunohistochemical study. Technol Cancer Res Treat 5:195–199, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 50

    Wang K, , Li M, , Hadley D, , Liu R, , Glessner J, & Grant SF, et al.: PennCNV: an integrated hidden Markov model designed for high-resolution copy number variation detection in whole-genome SNP genotyping data. Genome Res 17:1665–1674, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 51

    Wilson CB: Meningiomas: genetics, malignancy, and the role of radiation in induction and treatment. The Richard C. Schneider Lecture. J Neurosurg 81:666–675, 1994

  • 52

    Wöhrle FU, , Daly RJ, & Brummer T: Function, regulation and pathological roles of the Gab/DOS docking proteins. Cell Commun Signal 7:22, 2009

  • 53

    Wrobel G, , Roerig P, , Kokocinski F, , Neben K, , Hahn M, & Reifenberger G, et al.: Microarray-based gene expression profiling of benign, atypical and anaplastic meningiomas identifies novel genes associated with meningioma progression. Int J Cancer 114:249–256, 2005

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 54

    Wu J, & Lingrel JB: KLF2 inhibits Jurkat T leukemia cell growth via upregulation of cyclin-dependent kinase inhibitor p21WAF1/CIP1. Oncogene 23:8088–8096, 2004

  • 55

    Xiang Y, , Zhang CQ, & Huang K: Predicting glioblastoma prognosis networks using weighted gene co-expression network analysis on TCGA data. BMC Bioinformatics 13 Suppl 2 S12, 2012

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 56

    Yau C, , Mouradov D, , Jorissen RN, , Colella S, , Mirza G, & Steers G, et al.: A statistical approach for detecting genomic aberrations in heterogeneous tumor samples from single nucleotide polymorphism genotyping data. Genome Biol 11:R92, 2010

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 57

    Zhang B, & Horvath S: A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4:Article17 2005

  • 58

    Zhu XL, , Ai ZH, , Wang J, , Xu YL, & Teng YC: Weighted gene co-expression network analysis in identification of endometrial cancer prognosis markers. Asian Pac J Cancer Prev 13:4607–4611, 2012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1180 181 29
PDF Downloads 939 131 10
EPUB Downloads 0 0 0