Gene amplification is a common genetic aberration in many different types of human solid tumors.1 On cytogenetic analysis, amplified DNA sequences are typically located in homogenously staining regions or double minute chromosomes, which contain multiple copies of an amplification repeat unit or amplicon.2, 3, 4 Frequently, the amplicons do not only contain the actual amplification target gene(s), whose amplification-associated overexpression provides a selective growth advantage to the tumor cell, but also other genes that are coamplified because of their genomic proximity to the target gene(s).5 The most common amplicons in human malignant gliomas, in particular glioblastomas, involve genes from 7p11.2-p12 and 12q13-q15.6, 7, 8 Detailed molecular analyses have confirmed that the epidermal growth factor receptor gene (EGFR) is the most important amplification target at 7p11.2, whereas the genes for cyclin-dependent kinase 4 (CDK4) and mouse double minute 2 homolog (MDM2) were shown to represent 2 independent target genes from 12q13 and 12q15, respectively.6, 7, 8 However, several other genes from 7p11.2 or 12q13-q15 may be variably coamplified.7, 8
The chromosomal band 1q32 has been repeatedly shown to contain amplified sequences by comparative genomic hybridization (CGH) analysis of gliomas.9, 10, 11 We previously reported on the amplification and overexpression of the mouse double minute 4 homolog gene (MDM4) from 1q32 in a subset of malignant gliomas.12 The absence of TP53 mutations and MDM2 amplification in the tumors with MDM4 amplification indicated that MDM4 amplification and overexpression represents a novel molecular mechanism by which a small subset of malignant gliomas escapes from the p53-dependent growth control. We also found that several other genes from 1q32 were variably coamplified with MDM4, including the glioma amplification on chromosome 1 gene (GAC1), the renin gene (REN) and the retinoblastoma-binding protein 5 gene (RBBP5). More recently, other investigators reported on the amplification of the contactin 2 gene (CNTN2), also known as the transiently expressed axonal glycoprotein 1 gene (TAX1), in 2 of 26 malignant gliomas investigated.13 In 1 of these tumors, CNTN2 was coamplified with REN and GAC1 but not with MDM4. Furthermore, increased CNTN2 transcript levels were detected in several tumors without CNTN2 amplification.13
The physical mapping of chromosome 1 has made considerable progress. According to the GenBank database at the National Center for Biotechnology Information (NCBI), MDM4 and CNTN2 are located within the same contig (GenBank accession number NT_034410) about 0.5 megabases apart from each other. To address the question of whether there are 2 independent amplification targets on 1q32 in malignant gliomas or whether a common amplification target exists apart from MDM4 and CNTN2, we analyzed 17 genes from 1q32 for gene amplification and overexpression in 8 malignant gliomas with amplification or gain of genomic sequences at 1q32. We identified MDM4 as the gene from 1q32 that was commonly amplified or gained in these gliomas. Furthermore, MDM4 amplification was consistently accompanied by overexpression of the respective transcripts. CNTN2 was coamplified with MDM4 in 3 gliomas but overexpressed in only 1 of them. Therefore, our data indicate that MDM4 is the main amplification target on 1q32 in malignant gliomas, with several neighboring genes being variably coamplified.
MATERIAL AND METHODS
The tumors were selected from the brain tumor tissue collection of the Department of Neuropathology, Heinrich-Heine-University, Düsseldorf, Germany. All tumors were diagnosed according to the World Health Organization (WHO) classification of tumors of the central nervous system.14 The 8 malignant gliomas analyzed for amplification and expression of 17 genes from 1q32 comprised 5 glioblastomas of WHO grade IV (GBIV), 2 gliosarcomas of WHO grade IV (GSIV) and 1 anaplastic oligodendroglioma of WHO grade III (AOIII). These tumors were from 5 male and 3 female patients (mean age at operation: 54 years, range: 32–76 years). Five malignant gliomas with amplification of genes from 1q32 (GB31D, GB35D, GB112D, GB216D, AO11D) and 1 glioblastoma (GB96D) with a low-level gene copy number gain were identified in an analysis of 208 gliomas for amplification of MDM4 and GAC1 as reported before.12 The series was supplemented with 2 gliosarcomas (GS4Db, GS9D) that exibited amplification of the PIK3C2B gene at 1q32 in a study analyzing malignant gliomas for aberrations of genes related to the phosphatidylinositol-3′-kinase (PI3K)/protein kinase B (Akt) signal transduction pathway.15 In addition, we selected 102 malignant gliomas without MDM4 and/or PIK3C2B amplification (80 glioblastomas, 5 anaplastic astrocytomas, 3 anaplastic oligoastrocytomas and 14 anaplastic oligodendrogliomas) and screened them for amplification of CNTN2. Parts of each tumor were snap frozen immediately after operation and stored at −80°C. Only tumor pieces with a histologically estimated tumor cell content of 80% or more were used for molecular genetic analyses. As reference tissue for the mRNA expression studies, we used nonneoplastic cerebral tissue (cortex and white matter) from the temporal lobe of a patient operated on for chronic epilepsy.
DNA and RNA extraction
Extraction of high molecular weight DNA and RNA from frozen tumor tissue was carried out by ultracentrifugation as described before.16 Extraction of high molecular weight DNA from peripheral blood leukocytes was performed according to a standard protocol.17
The 8 malignant gliomas were analyzed for amplification of the genes MDM4 (mouse double minute homolog 4), GAC1 (glioma amplified on chromosome 1), REN (renin), RBBP5 (retinoblastoma-binding protein 5),ELF3 (E74-like factor 3), PTPN7 (protein tyrosine phosphatase nonreceptor type 7) and ELK4 (Ets-domain protein 4) using duplex-PCR analyses as reported.12 In addition, all tumors were screened by duplex-PCR for amplification of the genes PIK3C2B (phosphatidylinositol-3′-kinase class 2 β), PEPP3 (phosphatidylinositol-3′-phosphate-binding pleckstrin-homology domain protein 3), KISS1 (KISS1 metastasis suppressor), SNRPE (small nuclear ribonucleoprotein polypeptide E), SOX13 (SRY-box 13),CENPF (centromeric protein F), ATF3 (activating transcription factor 3) and CNTN2 (contactin 2), as well as the anonymous genes KIAA0756 and HDCMD38P. The individual primer sequences used for amplification of fragments from each of these genes, as well as the reference genes GAPDH (12p13), APRT (16q24) and IFNG (12q15), are available on request.
Each PCR reaction was performed in a total volume of 20 μl using 10 ng of template DNA. PCR conditions, including cycle number (25–30 cycles), MgCl2 concentration and annealing temperature, were optimized for each duplex-PCR assay. The PCR products were separated on 3% agarose gels, and ethidium bromide-stained bands were recorded using the Gel-Doc 1000 system (Bio-Rad, Hercules, CA). Quantitative analysis of the signal intensities obtained for each target gene and the corresponding reference gene was performed with the Molecular Analyst software (version 2.1, Bio-Rad). Increases in the target/reference gene ratio between 2- and 5-fold the ratio obtained for constitutional DNA were considered as low-level gene copy number gains, whereas normalized target/reference gene ratios of more than 5 were considered as gene amplification.
Southern blot analysis
For Southern blot analysis of MDM4, GAC1, REN and RBBP5 copy number, we used the PCR-generated probes reported before.12 In addition, we newly generated probes corresponding to a 257 bp fragment of PEPP3 (nucleotides −64 to 193, numbering from the first nucleotide of the start codon, GenBank accession number NT_034410), a 301 bp fragment of SOX13 (nucleotides 6858–7158, GenBank accession number AF149301) and a 251 bp fragment of CNTN2 (nucleotides 29105–29355, GenBank accession number NT_034410). The individual PCR products were purified after separation by agarose gel electrophoresis using the Jetsorb DNA isolation kit (Genomed, Bad Oeynhausen, Germany).
For Southern analysis, 2.5 μg DNA was digested using the restriction enzyme TaqI, separated on 0.8% agarose gels and alkali blotted to Hybond-N+ membranes (Amersham-Pharmacia Biotech, Freiburg, Germany). The membranes were hybridized with DNA probes labeled by random priming with [α-32P]dCTP. Hybridized membranes were exposed to either Kodak Biomax MR films or to phosphoimaging plates (Fuji, Kanagawa, Japan). The films were scanned on a GS-700 imaging densitometer (Bio-Rad) and analyzed with the Molecular Analyst Software (Bio-Rad). The phosphoimaging plates were scanned with the Fuji BioImager 1800 II (Fuji, Kanagawa, Japan), followed by densitometric analysis using the Aida Image Analyzer v.3.11 software (Raytest, Straubenhardt, Germany). As reference for the assessment of gene copy number, the blots were hybridized with probes for the variable number of tandem repeat locus D2S44 on 2q21.3-q22 (pYNH24, obtained from American Type Culture Collection) and a PCR-generated probe18 for the CCNA gene on 4q31-q35. The ratio between target DNA and reference DNA signal intensities was calculated for each tumor and normalized to the target DNA/reference DNA ratio determined for constitutional DNA samples extracted from peripheral blood leukocytes.
Duplex reverse transcription-PCR analysis
The 12 genes from 1q32 exibiting amplification or low-level copy number gains were further investigated for expression of the respective transcripts by duplex reverse transcription-PCR using the β2-microglobulin (B2MG, 15q21-q22) transcript level as reference. Three micrograms of total RNA from each tumor were reverse-transcribed into cDNA in a total volume of 50 μl using random hexanucleotide primers and Superscript reverse transcriptase (Gibco BRL, Eggenstein, Germany). PCR conditions, including cycle number (25–30 cycles), MgCl2 concentration and annealing temperature were optimized for each PCR reaction. The respective primer sequences are available on request. Agarose gel electrophoresis and densitometric analysis of the PCR products were carried out as described above. The ratio between target and reference mRNA signal intensities was calculated for each tumor and normalized to the target/reference mRNA signal ratio determined for a nonneoplastic brain tissue sample.
All 8 investigated malignant gliomas demonstrated increased target/reference gene ratios for MDM4 by duplex-PCR analysis (Figs. 1, 2, 4a). Densitometric determination of the signal intensities obtained by Southern blot hybridization revealed copy number gains of less than 5-fold for the glioblastoma GB96D, of 5–10-fold for 3 glioblastomas (GB35D, GB31D, GB216D) and of more than10-fold for 1 glioblastoma (GB112D), 2 gliosarcomas (GS4Db and GS9D) and 1 anaplastic oligodendroglioma (AO11D). The adjacent GAC1, PIK3C2B and PEPP3 genes were coamplified in 7 tumors with MDM4 amplification. The single glioblastoma with low-level copy number gain for MDM4 (GB96D) demonstrated similar gene copy number gains for GAC1 and PIK3C2B but not for PEPP3 (Fig. 4a). CNTN2, as well as the neighboring genes RBBP5, KIAA0756 and HDCMD38P (which all map distal to MDM4) were coamplified in 3 gliomas (GB35D, GB216D, AO11D) (Fig. 4a).
Amplification of the genes KISS1, SNRPE, SOX13 and REN, which map proximal to MDM4, PIK3C2B and PEPP3, was found in 4 of 8 tumors (Fig. 4a). Thus, the largest amplicons, including 12 of the 17 analyzed genes, were present in tumors GB35D, GB216D and A011D. These amplicons spanned about 1.3 megabases of genomic distance. The glioblastoma GB31D and the 2 gliosarcomas (GS4Db, GS9D) carried the smallest amplicons, which covered about 0.35 megabases and included GAC1,MDM4,PIK3C2B and PEPP3 (Fig. 4a). The low-level copy number gain involving GAC1, MDM4 and PIK3C2B in GB96D spanned an even smaller genomic fragment (approximately 0.2 megabases). The most proximally (ELK4) and most distally located genes (ELF3,PTPN7,ATF3 and CENPF) did not show any gene copy number increases in the 8 investigated gliomas.
To investigate whether CNTN2 is amplified in malignant gliomas independent from MDM4, we screened 102 malignant gliomas without MDM4 amplification for CNTN2 amplification. None of these tumors demonstrated CNTN2 amplification. A total of 95 malignant gliomas were additionally investigated for CNTN2 mRNA expression using duplex reverse transcription-PCR. None of these tumors demonstrated significantly increased CNTN2 mRNA levels relative to the nonneoplastic brain tissue used for reference.
All 7 tumors with MDM4 amplification (the 5 gliomas reported on before12 and the 2 additional gliosarcomas) demonstrated overexpression of MDM4 transcripts by duplex reverse transcription-PCR (Figs. 3, 4b). The MDM4 transcript levels were increased to the highest values in comparison with the transcript levels of the genes coamplified with MDM4 in these tumors (Fig. 4b). The glioblastoma with low copy number gain of MDM4 also exibited elevated MDM4 mRNA expression up to approximately 3-fold relative to nonneoplastic brain tissue.
The GAC1 transcript level was significantly increased relative to the reference brain tissue in only 4 of 7 tumors with GAC1 amplification (Fig. 4b). PIK3C2B amplification was accompanied by mRNA overexpression in 5 of 7 instances (Figs. 3, 4b). The single glioblastoma with a low-level copy number gain of these genes (GB96D) did not demonstrate increased levels of the respective transcripts. Similarly, amplification of PEPP3,KISS1 and HDCMD38P was accompanied by mRNA overrexpression in only 2 of 7, 1 of 4 and 2 of 3 tumors, respectively (Figs. 3, 4b). CNTN2 was amplified in 3 gliomas but demonstrated significant overexpression in only 1 of these cases (GB216D). Among the tumors without CNTN2 amplification, expression levels were approximately equal to nonneoplastic brain tissue except for a single gliosarcoma (GS4Db), which exibited a 4-fold increased expression (Figs. 3, 4b). This particular tumor lacked an increased CNTN2 gene copy number but demonstrated gene amplification of MDM4,GAC1, PIK3C2B and PEPP3, which was accompanied by overexpression of MDM4 and PIK3C2B transcripts. Amplification of RBBP5, KIAA0756, REN and SNRPE was always accompanied by an increased expression of the respective transcripts (Figs. 3, 4b). In contrast, none of the tumors with SOX13 amplification exhibited elevated expression.
Amplified chromosomal regions in human tumors may be of a simple or complex nature. Simple amplicons generally exibit a structural configuration corresponding to the normal chromosomal organization and may involve a single amplication target gene. In contrast, complex amplicons are structurally rearranged and may contain multiple amplification target genes. We previously reported on the amplification and overexpression of the MDM4 gene in malignant gliomas and proposed this gene as the likely target for the amplification at 1q32.12 We now performed a more detailed characterization of the 1q32 amplicon in malignant gliomas using the refined physical mapping information for this region that is available in the public genomic databases. Since a recent study reported on amplification and overexpression of the CNTN2 gene from 1q32 in individual malignant gliomas without MDM4 amplification,13 we wanted to address the question of whether a common amplification target exists apart from MDM4 and CNTN2 or whether 1q32 carries 2 independent amplification targets, as previously shown for the 12q13-q15 (CDK4/MDM2) amplicon.8
Our analysis of 17 genes from 1q32 for amplification and overexpression in malignant gliomas revealed a single center of amplification that included MDM4, GAC1, PIK3C2B and PEPP3. Among these 4 coamplified genes, only MDM4 was invariably overexpressed. The other 13 genes analyzed, including CNTN2, either were coamplified in only a fraction of the cases with MDM4 amplification or were not amplified at all. Thus, our data indicate that MDM4 is the main amplification target on 1q32 in malignant gliomas and argue against 2 independent amplification targets in this region. Our data also do not support the hypothesis of a common amplification target located between MDM4 and CNTN2.
The significance of MDM4 amplification and overexpression for tumor growth is supported by different lines of evidence. The Mdm4 protein has been shown to bind to p53 and inhibits p53-mediated transcriptional transactivation of other genes.19, 20 Mdm4 can also bind to Mdm2 and thereby blocks Mdm2 degradation.21MDM4 knockout mice are not viable and die early in embryonic development.22 However, loss of p53 function completely rescues this phenotype.22 We have reported before that malignant gliomas with MDM4 amplification and overexpression carry neither TP53 mutations nor MDM2 amplification, indicating that MDM4 amplification and overexpression probably represents an alternative mechanism to escape from p53-dependent growth control.12 The 2 additional gliosarcomas with MDM4 amplification included in the present study did not carry TP53 mutations either.23 GS4Db exhibited MDM2 amplification that was, however, restricted to its sarcomatous component.23 Thus, it may be possible that MDM4 amplification is confined to the gliomatous component of this tumor.
The GAC1 gene, which is the closest gene on the telomeric side of MDM4, was found to be coamplified in all cases with MDM4 amplification but was overexpressed in only 4 of 7 tumors with amplification. GAC1 codes for a transmembrane protein belonging to the leucine-reach repeat superfamily that may be involved in cell adhesion or signal transduction.24 However, the functional significance of Gac1 overexpression in malignant gliomas is not known yet.
PIK3C2B is the closest proximal neighbor of MDM4 and encodes a catalytic subunit of the phosphatidylinositol-3′-kinase (PI3K). PI3K plays a crucial role in the transduction of signals from cell membrane-associated growth factor receptors to the nucleus and thereby regulates a number of important cellular processes, such as cell growth and proliferation, apoptosis, migration and invasion, as well as angiogenesis.25 PI3K-dependent signaling is aberrantly activated in the majority of glioblastomas, most frequently due to inactivation of the Pten tumor suppressor protein, which dephosphorylates phosphatidylinositol-3,4,5-triphosphate (PIP3) and thereby counteracts the activity of PI3K.25 Thus, it is possible that PIK3C2B overexpression, as detected in 5 of 7 malignant gliomas with PIK3C2B amplification, provides a growth advantage to glioma cells by virtue of aberrant PIK3/Akt signaling.
The gene product of PEPP3, a recently identified homolog of Pepp1, is expressed in nonneoplastic brain and contains a pleckstrin homology domain with phosphatidylinositol-3′-phosphate-binding activity.26 Thus, Pepp3 may also be involved in phosphatidylinositol-3′-phosphate-dependent signal transduction, but it remains to be shown whether its enhanced expression provides glioma cells with a growth advantage. We found increased PEPP3 mRNA levels in only 2 of 7 malignant gliomas with PEPP3 gene amplification, which indicates that PEPP3 is unlikely to be a major amplification target at 1q32.
CNTN2 was coamplified with MDM4 in 3 malignant gliomas of our series and overexpressed in 1 of these tumors. An additional glioblastoma demonstrated an elevated CNTN2 transcript level in the absence of gene amplification. This latter finding, however, seems to be rare since we did not identify any further tumor with CNTN2 mRNA overexpression among 95 malignant gliomas without CNTN2 amplification. CNTN2 encodes a 135 kDa protein that belongs to the immunoglobulin superfamily and may function as a cell adhesion molecule. The Cntn2 protein is transiently expressed in developing neurons and involved in axonal guidance.27, 28 Rickman et al.13 first reported on CNTN2 amplification and overexpression in individual malignant gliomas and additionally showed that the migration of glioma cells in vitro can be markedly reduced by the application of antibodies or antisense oligonucleotides against the Cntn2 protein or mRNA, respectively. Therefore, these authors concluded that Cntn2 plays a role in glial tumorigenesis and may be a potential target for therapeutic intervention. In contrast, our results indicate that CNTN2 is amplified and overexpressed in only a minor percentage of glioblastomas and seems not to be the major target of 1q32 amplification in these tumors. However, the single malignant glioma that exhibited CNTN2, GAC1 and REN amplification but lacked MDM4 amplification in the study by Rickman et al.13 suggests that individual tumors may carry discontinuous 1q32 amplicons, in which genes other than MDM4 may be the relevant targets.
A number of genes from 1q32, including KISS1, REN, SOX13 and SNRPE on the centromeric side of MDM4, as well as KIAA0756,RBBP5 and HDCMD38P on the telomeric side of MDM4, were variably coamplified in subsets of the malignant gliomas with MDM4 amplification. Our data indicate that any of these genes is unlikely to represent the major amplification target on 1q32. However, the possibility cannot be excluded that amplification and overexpression of 1 or more of these genes provide an additional growth advantage. We previously discussed the possible significance of REN and RBBP5 amplification and overexpression.12 Since the functions of the SNRPE, KIAA0756 and HDCMD38P gene products are not known at present, the biologic significance of amplification and overexpression of these genes remains to be determined. Amplification of SOX13 was never associated with overexpression, a finding that strongly argues against a role of SOX13 in gliomas. Similarly, only a single tumor (AO11D) with KISS1 amplification expressed detectable transcripts of this gene. Thus, KISS1 and SOX13 are probably just coamplified because of their genomic proximity to MDM4.
As reported before,12 we did not detect any amplification of the ELF3 and ELK4 genes from 1q32, both coding for members of the ets family of transcription factors.29, 30 Similarly, the PTPN7 (HePTP) gene, which encodes a hematopoietic tyrosine phosphatase that was reported as a gene amplified and overexpressed in myeloid neoplasms,31 was not amplified in our series of malignant gliomas. The genes for the activating transcription factor-3 (ATF3) and the centromere protein F (CENPF) may be amplified in esophageal squamous cell carcinomas.32CENPF amplification and overexpression were additionally detected in a subset of head and neck squamous carcinomas.33 However, we did not find any amplification of these genes in our series of malignant gliomas.
In summary, our detailed characterization of 1q32 amplicons in malignant gliomas indicates a single region of amplification and confirms that MDM4 is the main amplification target. We did not find any evidence for 2 independent amplification targets on 1q32 or for a common amplification target located between MDM4 and CNTN2. However, our results show that several genes mapping close to MDM4 are frequently coamplified and overexpressed, which may provide an additional growth advantage in some gliomas.