Introduction

Ependymomas occur predominantly in children and adolescents, comprise approximately 9% of childhood brain tumors, and represent the third most frequent kind of brain tumors in children. Similar to other malignant gliomas, anaplastic ependymomas (WHO grade III) differ from ependymomas (WHO grade II) by the presence of high mitotic activity, cellular pleomorphism, and nuclear anaplasia. However, even when present, anaplasia does not appear to be consistently associated with poor prognosis [29]. Age at diagnosis, extent of resection, tumor location, and proliferation indices seem to be significant prognostic factors in this disease. Moreover, spinal ependymomas are associated with a more favorable prognosis than intracranial ependymomas of the respective WHO grade [21]. Adult ependymoma patients usually have a better prognosis than children with histologically comparable tumors [19].

Little is known regarding molecular genetic alterations underlying the formation and progression of ependymomas. Cytogenetic and molecular genetic studies have revealed structural and numerical abnormalities of chromosomes 6, 7, 9, 10, 11, 13, 17, and 22 [2, 7]. In comparative genomic hybridization (CGH) studies, these tumors showed mostly complex aberration patterns with frequent deletions or gains of chromosomes. The main losses occurred on chromosomes 1p, 4q, 6q, 9, 10, 13q, 16, 17, 19q, 20q and 22q [10, 24, 28, 36, 38]. This method, however, allows the detection of genetic aberration over the genome only at a resolution of 10–20 Mbp. Previous LOH studies in ependymomas have been restricted by the small number of cases studied; only a limited number of microsatellite markers and genes were investigated at one time [6, 16, 17]. Recently, an LOH study in ependymomas with 384 microsatellite markers was published [35]. However, the investigators focused on ependymomas WHO grade II and did not examine either the relationships between various genetic abnormalities detected or the relationships between genetic abnormalities and clinical features. Our group has recently identified two common regions showing deletions on the long arm of chromosome 22 in 33 ependymomas of brain and spinal cord using microsatellite markers [11]. In the present study, we have conducted allelotyping of 33 ependymomas using 34 microsatellite markers distributed over 15 autosomes to detect possible aberrations characterizing distinct clinicopathological subgroups.

Materials and methods

Tumor samples and DNA extraction

Tumor specimens (n=30) were obtained from patients treated at the Department of Neurosurgery, University Hospital of Würzburg (Würzburg, Germany) between 1992 and 2001 and 3 cases from the University Hospital of Münster (Münster, Germany). Histological diagnoses were established according to the WHO classification criteria [37]. Twenty-five cases were primary preoperatively untreated tumors; 8 tumors were recurrent. The series consisted of 5 myxopapillary ependymomas WHO grade I, 11 ependymomas WHO grade II, and 17 anaplastic ependymomas WHO grade III. 20 of the ependymomas were intracranial tumors; 13 were spinal cord lesions; 15 female and 18 male patients were included (Table 1). The mean patient age at diagnosis was 19 years (range 8 months to 69 years). For DNA extraction, only areas of the solid tumor specimens were analyzed; some cases required microdissection. DNA was prepared as described previously [11].

Table 1. Clinical and LOH data on 33 patients with ependymomas of various grades. LOH events out of the total number of informative genotypes are shown; results for ependymoma subgroups displaying frequent aberrations are in bold (LOH loss of heterozygosity, Chr. chromosome IT infratentorial, ST supratentorial)
Full size table

PCR and microsatellite analysis

We used a panel of 34 microsatellite repeats spread over 15 chromosomes for the analysis (Table 2). The repeats were chosen to cover loci of known and putative oncogenes or tumor suppressor genes and regions showing frequent aberrations detected in the previous CGH studies [10, 24, 28, 38]. Primer sequences for PCR amplification were retrieved from the Genome Database (http://gdbwww.gdb.org/). PCR primers were synthesized at MWG Biotech (Munich, Germany) and one oligonucleotide of each primer pair labeled with fluorescent dye phosphoramidites FAM, NED or HEX. Paired constitutional DNAs from peripheral blood leukocytes and tumor DNA samples of each patient were amplified with AmpliTaq Gold enzyme (ABI, Foster City, CA) in multiplex PCR reactions using 50 ng genomic DNA as template. Thirty cycles were carried out in a Mastercycler thermal cycler (Eppendorf, Hamburg, Germany). Aliquots of the PCR reactions were then mixed with ROX-labeled size standard and formamide, denatured, and subjected to electrophoresis on a 377 DNA Sequencer (ABI). The automatically collected data were analyzed by GENESCAN and GENOTYPER software as described in the manufacturer's manual. LOH and amplification were scored as described previously [11]. In summary, the unchanged allele was identified, and the ratios of the allele showing decreased or increased signal relative to the unchanged allele were calculated, first for control DNA, and then for the tumor. Decrease of the ratio by 40% in the tumor compared with the control was called LOH, increase by 40% amplification.

Table 2. Thirty-four microsatellites used in the analysis of ependymomas, their chromosomal location, and genes of interest
Full size table

Statistical methods

Correlation between LOH results and clinicopathological variables (age, gender, tumor location, grade of differentiation) was evaluated by chi-square test. Associations between 6q LOH and LOH on chromosome 9 were also analyzed.

Results

LOH analysis

Representative examples of electropherograms are shown in Fig. 1. LOH with any of the 34 markers was observed in 26 of 33 tumors (78.8%). Out of 878 informative loci, 85 genotypes (9.7%) showed LOH. LOH frequency for individual chromosomal arms varied from 3.7% (20q13.1) to 26.7% (6q24-25.3). Amplifications were found in only 3 tumors (case 3 in 13q14, case 5 in 9q32-33 and case 24 in 20q13.1); microsatellite instability was found in one case. The frequency of genetic aberrations in specific regions is depicted in Fig. 2: 6q24-25.3 (26.7%), 6q15-16 (23.8%), 9p21 (22.6%), 6q24 (21.4%), 6q21-22.1 (19.2%), 11q23-24 (14.3%), 3p14.2 (13.3%), 11q13.1 (12.5%), 10q23 (10.3%), 13q14 (9.7%), 9q34 (9.4%), and 17p13.1 (9.3%). The long arm of chromosome 6 displayed the most frequent aberrations, 10 of 33 cases (30.3%) showing LOHs in two hot spots: 6q24-25.3 and 6q15-16 (Table 2 and Figs. 2, 3) as detected by markers D6S441 and D6S1709 in 8 (26.7% of informative cases) and 5 (23.8% of informative cases) tumors, respectively. The second most frequently altered chromosome was chromosome 9 displaying aberrations in 9 tumors (27.3% of informative cases) involving both its long and short arms (Fig. 3).

Fig. 1.
figure 1

Representative LOH in the 11q23-24 region as detected with the D11S2179 microsatellite marker in patient 11. Control and tumor samples were amplified in a multiplex reaction with markers D11S2179 located in the 11q23-24 region and D15S231 used as internal control. Arrows indicate LOH (LOH loss of heterozygosity)

Full size image
Fig. 2.
figure 2

Chromosomal regions showing frequent LOH. Aberration frequency for regions suffering more than three LOHs is plotted in the diagram

Full size image
Fig. 3.
figure 3

Summary of LOH results from 20 ependymoma patients with deletion of main markers on chromosome 3, 6, 9, 10,11, 13, and 17. Only cases with more than three aberrations are plotted in the diagram. Red, LOH; black, retention of both alleles; gray, noninformative; L/A: allelic imbalance

Full size image

Clinicopathological correlations

Intracranial tumors were found more frequently in children (mean age 10.6 years) than in adult patients (mean age 31.8 years) who had more intraspinal tumors (P<0.01, χ2 test). While the vast majority (84%) of intraspinal ependymomas were histologically benign, corresponding to WHO grade II or I, most of the intracranial tumors (75%) were classified as anaplastic, WHO grade III (P<0.01, χ2 test, Table 1). The differences in tumor grade and location were also reflected in the differences in patient survival. Three patients had died by the time of this study, all of these cases were children with intracranial anaplastic ependymomas.

LOH on chromosome 9 was frequently associated with male patients (8 male versus 1 female patients, P=0.02, χ2 test). Five anaplastic ependymomas with LOH on chromosome 9 were intracranial: four of them with loss of 9p were infratentorial and one with loss of 9q was supratentorial (Table 1). While the numbers in each group were small, there was no significant difference in the incidence of LOH between the intracranial and spinal ependymomas or between infratentorial and the supratentorial ependymomas. There was no correlation between allelic loss and a specific tumor grade. However, a trend (P=0.09, χ2 test) was found between LOH and age distribution: adult ependymomas seemed to display more frequent LOHs than pediatric ependymomas (93.3% versus 72.2%). In particular, genotypes displaying LOH at D9S2136 were found more frequently in adults (23%) than in children (9%) (P=0.047, χ2 test). Comparing cases with LOH on chromosomal arm 6q and/or chromosome 9, we found 9 tumors showing only LOH on the long arm of chromosome 6 and 7 tumors only on chromosome 9. Only 2 out of 18 tumors showed concurrent LOH on 6q and on chromosome 9 (Fig. 3). However, a statistical inverse correlation between LOH 6q and LOH 9 failed to reach significance (P=0.5). We found no association between losses of the two chromosomes and loss of chromosome 22. There was no relationship between LOHs and whether the tumor was primary or recurrent.

Discussion

A previous study from our group suggested that LOH on the long arm of chromosome 22 plays an important role in pathogenesis of ependymoma. In this work, we have identified other chromosomal regions that may contain candidate genes involved in the development of ependymomas. The most frequent LOH occurred on 6q, as 30.3% of tumors tested showed a deletion of a part or the whole long arm of chromosome 6. Deletions and rearrangements involving 6q have been reported in a number of human malignancies, including gliomas [18], primitive neuroectodermal tumor [34], breast carcinoma [33], acute lymphoblastic leukemia [8], and thymoma [39]. Previous cytogenetic studies on ependymomas showed deletions or rearrangements of 6q with variable frequencies [15, 23]. Loss of 6q has been reported as the most common genetic aberration using the CGH approach [24]. Tong et al. [35] performed allelotyping studies of ependymomas using 384 microsatellite markers. A high frequency of LOH was detected on 6q (68.75%) and a hot spot deletion region was defined at chromosome 6q25.2-27. However, only 16 ependymomas (grade II) were included in that study and no relationship between genetic alterations and clinicopathological features was detected. To our knowledge, our study represents the largest series to date of ependymomas investigated by microsatellite analysis for genetic aberrations on the long arm of chromosome 6q. The 6q24-25.3 was the most frequent deletion region in this work, frequent aberrations were detected also in 6q15-16 and 6q21-22.1. In a previous lymphoma study, one of the hot spots was detected in the region 6q23.3-25.3 between the markers D6S310 to D6S441 overlapping our above-mentioned region [31]. This region contains several already cloned genes, e.g., IGF2R, MAP3K5, and ZAC. Interestingly, the ZAC tumor suppressor gene has been shown to inhibit tumor cell proliferation in vitro and in vivo in nude mice [30] and its expression was found to be lost or reduced in breast cancer cell lines and primary breast tumors [3]. Putative tumor suppressor loci at 6q22 and 6q23-q24 are involved in the malignant progression of sporadic endocrine pancreatic tumors [1]. Whatever gene is the target of deletions here, it seems to define a particularly important event in ependymoma pathogenesis. Regarding region 6q15-16, several studies reported the deletion of this region in breast tumors [26], acute lymphoblastic leukemia and non-Hodgkin's lymphomas [20]. A transcription factor, BACH2, which possesses an inhibitory effect on cell proliferation, was localized to this region [27]. Additional studies are in progress to narrow the 6q15-16 and 6q24-25.3 regions and identify potential tumor suppressor genes that are involved in ependymoma tumorigenesis.

Chromosome 9 was the second most frequently altered chromosome in this study. Region 9p21 contains CDKN2A (encoding p16INK4a) and CDKN2B (encoding p15INK4b), well-known tumor suppressor genes that are frequently altered in various types of human neoplasms, including brain tumors. The inactivation of CDKN2A may play a role in progression of thymoma and thymic carcinoma [9]. In brain tumor, CDKN2A deletions were also associated with evolution of neuroblastoma [32], glioblastomas [22] and worse prognosis in oligodendrogliomas [4]. CDKN2A inactivation may not play an important role in the malignant transformation of ependymomas [5]; however, inactivation of the p14ARF seems to be involved in ependymoma progression [14].

Comparing cases with LOH on chromosome 6q and/or chromosome 9, there appears to be a trend to inverse correlation between LOH on chromosome 6q and 9. This negative correlation was also found by CGH methods [28].

Regarding genetic aberrations on other chromosomes, relatively frequent LOHs were detected in regions 3p14.2, 10q23, 11q13.1, 11q23, 13q14, and 17p13. Of these chromosomal regions, 3p has rarely been reported in the previous studies of ependymomas. With conventional cytogenetics, losses of chromosome 10 were reported in about 9% of ependymomas [12] as well as in gliomas [13]. Losses of chromosome 13 were described in approximately 5% ependymomas [23]. The tumor suppressor genes Fhit, PTEN, MEN1, ATM, Rb and P53 were localized to the above-mentioned loci, respectively.

Interestingly, LOH-positive genotypes were found preferentially in adults (93.3%) as opposed to children (72.2%). Nevertheless, children ependymoma patients have a worse prognosis than adult patients with histopathologically comparable tumors [19]. This suggests that the clinical behavior of ependymomas may not be simply correlated to the number of LOHs per case. LOHs on chromosome 9 were preferentially associated with male patients (8 male versus 1 female patients, P=0.02). Our results support the notion that molecular biological analysis might show a different aberration pattern in various subtypes of ependymomas, as has already been shown for astrocytomas [25]. As most of the cases that we collected are fairly recent, a longer waiting period will be needed to correlate LOH results with patient survival.

In conclusion, our data demonstrate that ependymal tumors of all WHO grades suffer allelic losses during carcinogenesis. The most frequent LOHs were found on the long arm of chromosome 6 with hot spots in 6q15-16 or 6q24-25.3 and chromosome 9. We conclude that important tumor suppressor genes are situated on these two chromosomes.