Molecular pathogenesis of pediatric astrocytic tumors

Abstract

Astrocytomas are the most common pediatric brain tumors, accounting for 7%–8% of all childhood cancers. Relatively few studies have been performed on their molecular properties; therefore, classification of pediatric astrocytic tumors into genetic subtypes similar to that of adult tumors remains to be defined. Here, we report an extensive characterization of 44 pediatric astrocytomas—16 diffuse astrocytomas (WHO grade II), 10 anaplastic astrocytomas (WHO grade III), and 18 glioblastomas (WHO grade IV)—in terms of genetic alterations frequently observed in adult astrocytomas. Some form of p53 mutation was found in three diffuse astrocytomas, in three anaplastic astrocytomas, and in six glioblastomas examined; PTEN mutations were detected only in two glioblastomas. EGFR amplification was detected in only one anaplastic astrocytoma and two glioblastomas, but no amplification was observed for the PDGFR-α gene. Loss of heterozygosity (LOH) on 1p/19q and 10p/10q was less common in pediatric astrocytic tumors than in those seen in adults, but the frequency of LOH on 22q was comparable, occurring in 44% of diffuse astrocytomas, 40% of anaplastic astrocytomas, and 61% of glioblastomas. Interestingly, a higher frequency of p53 mutations and LOH on 19q and 22q in tumors from children six or more years of age at diagnosis was found, compared with those from younger children. Our results suggest some differences in children compared to adults in the genetic pathways leading to the formation of de novo astrocytic tumors. In addition, this study suggests potentially distinct developmental pathways in younger versus older children.

Brain tumors amount to less than 2% of all malignant neoplasms. In children, however, they are the most common solid tumors, causing nearly one-quarter of all childhood cancer deaths; the incidence of pediatric brain tumors also seems to be increasing more rapidly than any other tumor type (Bleyer, 1993; Rickert and Paulus, 2001). While high-grade astrocytomas are characteristically the most common primary CNS malignancy in adults, low-grade astrocytomas are the most common lesions in children, accounting for approximately 40% of all pediatric cerebral tumors (Kaatsch et al., 2001; Kleihues et al., 2000). Adult brain tumors are typically supratentorial, whereas pediatric tumors are localized predominantly in the posterior fossa and brainstem (Cohen et al., 2001). Many studies have indicated that, as a group, malignant gliomas in children and young adults carry a more favorable prognosis than do comparable lesions in older patients (Campbell et al., 1996; Salcman et al., 1994). When taken together, these results imply age-related differences in tumor biology and/or host–tumor interactions.

The majority of adult glioblastomas (WHO grade IV) occur in older patients after a short clinical history with no clinical or histopathologic evidence of less malignant precursor lesions (primary or de novo glioblastoma). In younger patients, another type of glioblastoma, so-called secondary glioblastoma, shows a slow progressive development from low-grade diffuse (WHO grade II) or anaplastic astrocytoma (WHO grade III) (Kleihues et al., 2000). Adult glioblastomas display distinct pathways of genetic alterations during progression characterized by EGFR amplification/overexpression and PTEN mutation (Lang et al., 1994a; von Deimling et al., 1993; Watanabe et al., 1996), while, in contrast, secondary glioblastomas developing in younger patients demonstrate frequent p53 mutations (Tohma et al., 1998; Watanabe et al., 1996). Amplification and overexpression of PDGFR-α are typical in the pathways leading to secondary glioblastomas (Kleihues et al., 2000). In recent studies, we have observed that primary glioblastomas seem to be characterized by loss of hetero-zygosity (LOH)³ throughout chromosome 10 (Fujisawa et al., 2000), while secondary glioblastomas preferentially show LOH on chromosomes 10q, 19q, and 22q (Fujisawa et al., 2000; Nakamura et al., 2000, 2005b) and promoter methylation of RB1, TIMP-3, and HRK (Nakamura et al., 2001c, 2005a, 2005b). Pediatric glioblastomas are mostly primary tumors and resemble secondary glioblastomas only in that the patients are young (Sure et al., 1997).

Compared with the extensive efforts that have been directed at characterizing the molecular features of adult astrocytoma progression, relatively few data have been collected on pediatric tumors, reflecting, in part, the fact that these lesions are less common. Although a few studies have suggested that de novo pediatric malignant gliomas rarely show EGFR gene amplification and more commonly exhibit the p53 mutations seen in secondary gliomas (Bredel et al., 1999; Pollack et al., 1997; Raffel et al., 1999; Sung et al., 2000), the question remains as to whether childhood astrocytomas can be classed into genetic subtypes as has been done for adult astrocytomas. In addition, few studies have been conducted on low-grade diffuse astrocytomas of children specifically to assess them for the more common genetic alterations associated with adult astrocytoma progression.

The aim of this study, then, was to investigate whether the genetic alterations documented for adult astrocytomas are, in fact, also operative in children. To address this issue, we collected 44 primary pediatric astrocytic tumors and screened each lesion for genetic alterations that may be involved in tumorigenesis. Our results show different genetic aberrations both between astrocytic tumors in younger and older pediatric patients and between pediatric and adult tumors, supporting the notion of different genetic mechanisms in patient-age-related tumor development.

Materials and Methods

Tumor Samples and DNA Extraction

Sixteen diffuse astrocytomas, 10 anaplastic astrocytomas, and 18 glioblastomas were retrieved from the archives of the Nara Medical University School of Medicine, Nara, Japan, with a full review of all available clinical data. Between 1985 and 2005, all 44 patients (22 males and 22 females) had undergone surgical resection, autopsy, or surgical pathology consultation for sporadic pediatric astrocytic tumors. Patients within this pediatric group were younger than 18 years of age at the time of operation, ranging in age from one month to 17 years, with a median age of nine years. Fourteen of our patients were younger than six years, whereas 30 patients were treated surgically between the age of 6 and 17 years. Of the lesions selected, 70% (31 tumors: 8 diffuse astrocytomas, 8 anaplastic astrocytomas, 15 glioblastomas) were located supratentorially, 25% (11 tumors: 7 diffuse astrocytomas, 2 anaplastic astrocytomas, 2 glioblastomas) developed in the brainstem, and only 4.5% (two tumors: one diffuse astrocytoma, one glioblastoma) were found in the cerebellum. The histologic diagnosis was reevaluated according to the WHO classification of CNS tumors (Kleihues et al., 2002). The distribution of the tumors was as follows: 11 in brainstem; 10 in frontal lobe; 8 in parietal lobe; 4 in occipital lobe; 3 each in temporal lobe, hypothalamic region, and thalamus; 2 in cerebellum. All of these tumors were sporadic and primary in which patients had no prior history of low-grade astrocytomas. Table 1 summarizes patient gender and age, in addition to the WHO tumor classifications and sites of removal. The archived tumor samples had been fixed in 10% buffered formalin and embedded in paraffin. We extracted DNA from mounted paraffin sections, as described previously, to be used for mutation, amplification, and microsatellite analyses (Nakamura et al., 2001a; Watanabe et al., 1996). Control genomic DNA was extracted from related but separate blocks of tissue not involved with the tumor.

Table 1.

Genetic alterations in pediatric astrocytic tumors

			Exon		Amplification/Overexpression		LOH
Case Number	Age (years)/Sex	Location	p53 Mutation	PTENMutation	EGFR	PDGFR	1p/19q	10p/10q	22q
Diffuse astrocytoma, WHO grade II
1	3/F	Frontal	–	–	−/−	−/−	−/−	−/−	–
2	4/M	Brainstem	–	–	−/−	−/−	−/−	−/−	–
3	5/M	Cerebellum	–	–	−/−	−/−	−/−	−/−	LOH
4	5/M	Brainstem	–	–	−/−	−/+	−/−	−/−	–
5	6/F	Brainstem	–	–	−/−	−/−	−/−	−/−	–
6	7/M	Parietal	–	–	−/−	−/−	−/−	−/−	LOH
7	8/F	Brainstem	–	–	−/−	−/−	−/−	−/−	–
8	9/M	Frontal	–	–	−/−	−/−	−/−	−/−	LOH
9	11/F	Brainstem	–	–	−/−	−/−	−/−	−/−	–
10	12/F	Thalamus	–	–	−/−	−/−	−/−	−/−	LOH
11	13/M	Brainstem	–	–	−/−	−/+	−/−	−/−	LOH
12	13/F	Parietal	E 8		−/−	−/−	−/−	−/−	–
13	15/M	Temporal	–	–	−/−	−/−	−/−	−/−	LOH
14	15/F	Brainstem	E 5	–	−/−	−/+	−/LOH	−/−	–
15	15/F	Hypothalamus	–	–	−/−	−/−	–	−/−	–
16	16/M	Temporal	E 7	–	−/−	−/−	–	−/−	LOH
Anaplastic astrocytoma, WHO grade III
1	4/F	Frontal	–	–	+/+	−/−	–	−/−	LOH
2	4/M	Brainstem	–	–	−/−	−/−	–	−/−	–
3	6/F	Brainstem	–	–	−/−	−/−	–	−/−	–
4	7/F	Thalamus	–	–	−/−	−/+	–	−/−	LOH
5	7/F	Parietal	–	–	−/+	−/−	–	−/−	–
6	9/M	Parietal	E 6	–	−/−	−/−	–	−/LOH	–
7	11/F	Frontal	–	–	−/−	−/−	−/LOH	−/−	LOH
8	13/M	Hypothalamus	E 5	–	−/−	−/+	–	−/−	–
9	15/M	Occipital	E 7	–	−/−	−/−	–	−/−	–
10	17/M	Frontal	–	–	−/−	−/−	–	−/LOH	LOH
Glioblastoma, WHO grade IV
1	1 mo/M	Parietal	–	E 6	−/+	−/−	–	LOH/LOH	LOH
2	1/F	Cerebellum	–	–	−/−	−/−	–	−/−	–
3	2/F	Occipital	–	–	+/+	−/−	–	−/−	–
4	3/M	Frontal	–	–	+/+	−/−	–	LOH/LOH	–
5	3/M	Brainstem	–	–	−/−	−/−	–	−/−	LOH
6	3/F	Hypothalamus	E 7	–	−/−	–/+	–	LOH/LOH	–
7	4/F	Temporal	–	E 5	−/−	−/−	–	−/−	LOH
8	4/M	Frontal	–	–	–/+	−/−	–	−/−	–
9	8/M	Occipital	–	–	−/−	−/−	–	−/−	LOH
10	9/F	Parietal	–	–	−/−	−/−	–	−/−	LOH
11	12/M	Thalamus	E 6	–	−/−	−/−	LOH/LOH	−/−	LOH
12	13/F	Frontal	–	–	−/+	−/−	–	−/−	–
13	15/M	Occipital	E 8	–	−/−	−/+	–	−/−	LOH
14	16/F	Brainstem	–	–	−/−	−/−	−/LOH	−/LOH	LOH
15	16/F	Frontal	E 5	–	−/−	−/+	–	−/−	–
16	16/M	Parietal	–	–	−/−	−/+	LOH/LOH	−/−	LOH
17	17/M	Frontal	E 7	–	−/−	−/+	–	−/LOH	LOH
18	17/F	Parietal	E 5	–	−/+	−/−	−/LOH	−/−	LOH

Mutation Analyses of p53 and PTEN

We prescreened for mutations in exons 5–8 of the p53 gene and exons 1–9 of PTEN gene by PCR–single-strand conformational polymorphism (SSCP) analysis. The sequencing primers we used have been described previously (Nakamura et al., 2001b; Watanabe et al., 1998). Samples that showed mobility shifts in SSCP analysis were further analyzed by DNA sequencing; after PCR amplification with the same set of primers used in SSCP, PCR products were sequenced on the Genetic Analyzer 310 (Applied Biosystems, Foster City, Calif.) using ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kits (Applied Biosystems).

Evaluation of EGFR and PDGFR-α Amplification

Real-time quantitative PCR was carried out according to previously published methods (Biernat et al., 2004; Nigro et al., 2001). Two sets of primers and fluorescent internal probes for the EGFR genome and the internal control β-actin were used. Both PCR products have similar sizes, to ensure equal amplification efficiency. Two sets of primers and probes for the PDGFR-α and the internal control IFN-82 were also used. To calculate the relative gene dosages and the level of EGFR and PDGFR-α amplification, we used the C_t values (the PCR cycle number in which the first fluorescence signal measured above threshold), dividing the difference of C_t values of the target gene and the reference gene of the tumor sample (ΔC_tT) by the difference in C_t values of the target and reference gene of the normal brain sample (ΔC_tN). Multiplication by 2 of the ratio of ΔC_tT:ΔC_tN then yielded the level of amplification.

LOH Assay

Forty-eight highly polymorphic markers were selected from the Genome Database (www.gdb.org/) and the National Center for Biotechnology Information (www.ncbi.nlm.nih.gov/genemap) based on the frequency of heterozygosity, as well as by coverage and flanking of the region of interest. For each marker, we labeled the sense primer with a fluorescent dye and then paired normal and tumor DNA samples from each patient. The paired samples were amplified for 30 cycles with an annealing temperature of 56–58°C. We mixed aliquots of the PCR reactions with a size standard and formamide, denatured them, and subjected them to capillary electrophoresis on a Genetic Analyzer 310 (Applied Biosystems); collected data were analyzed using GeneScan software (Applied Biosystems). We independently repeated the analysis of each marker at least twice and found a variation of no more than 3% in the allelic ratios. Only those samples heterozygous for a given locus were regarded as informative; locus homozygosity and/or microsatellite instability rendered any particular sample noninformative. Samples were considered to show LOH when a peak allele signal from tumor DNA was reduced by 50% compared with its normal-tissue counterpart (Nakamura et al., 2003).

Immunohistochemical Analysis of EGFR and PDGFR-α Expression

We assessed the expression of EGFR and PDGFR-α immunohistochemically, using monoclonal antihuman EGFR antibody (EGFR113; Novocastra Laboratories, Newcastle upon Tyne, U.K.) and polyclonal antihuman PDGFR-α antibody (RB-1691; NeoMarkers, Lab Vision, Fremont, Calif.). After deparaffinization, unstained tissue slides were heated to boiling in a pressure cooker for 5 min in 10 mM sodium citrate (pH 6.0) buffer. They were then incubated overnight at 4°C with anti-EGFR and anti-PDGFR-α antibody at a dilution of 1:100 and 1:150, respectively. We visualized binding reactions using the chromagen diaminobenzidine with the Histofine SAB-PO kit (Nichirei, Tokyo, Japan) and hematoxylin counterstaining.

Statistical Methods

Fisher’s exact test was used to examine possible associations between genetic alterations and tumor grade. The association between patient age and genetic status was calculated by the Mann-Whitney U-test. Statistical significance was established as P < 0.05.

Results

p53 and PTEN Mutations

In the samples evaluated, we found p53 mutations in 19% (3 of 16) of diffuse astrocytomas, in 30% (3 of 10) of anaplastic astrocytomas, and in 33% (6 of 18) of glioblastomas examined (Table 1). There was a significant association between patient age and p53 mutations in each grade. Of the six glioblastomas we collected containing detectable mutations in p53, only one occurred in a child younger than 6 years (Table 2). However, fully half of the glioblastomas harvested from patients 6 or more years of age contained p53 mutations.

Table 2.

Genetic alterations in pediatric astrocytic tumors according to patient age

		Mutation		EGFR		PDGFR		LOH
Tumor Type	Number	p53	PTEN	Amp.	Over.	Amp.	Over.	1p	19q	10p	10q	22q
Diffuse astrocytoma	16	3 (19%)	0	0	0	0	3 (19%)	0	1 (6%)	0	0	7 (44%)
(6 years)	4	0	0	0	0	0	1 (25%)	0	0	0	0	1 (25%)
(≥6 years)	12	3 (25%)	0	0	0	0	2 (17%)	0	1 (8%)	0	0	5 (42%)
Anaplastic astrocytoma	10	3 (20%)	0	1 (10%)	2 (20%)	0	2 (20%)	0	1 (10%)	0	2 (20%)	4 (40%)
(6 years)	2	0	0	1 (50%)	1 (50%)	0	0	0	0	0	0	1 (50%)
(≥6 years)	8	3 (38%)	0	0	1 (13%)	0	2 (25%)	0	1 (13%)	0	2 (20%)	3 (38%)
Glioblastoma	18	6 (33%)	2 (11%)	2 (11%)	6 (33%)	0	5 (28%)	2 (11%)	4 (22%)	3 (17%)	5 (28%)	11 (61%)
(6 years)	8	1 (13%)	2 (25%)	2 (25%)	4 (50%)	0	1 (13%)	0	0	3 (38%)	3 (38%)	3 (38%)
(≥6 years)	10	5 (50%)	0	0	2 (20%)	0	4 (40%)	2 (20%)	4 (40%)	0	2 (20%)	8 (80%)

Abbreviations: Amp., amplification; Over., overexpression.

PTEN mutations were detected in only 2 (11%) of 18 glioblastomas examined (Table 1). Both tumors were in children younger than 6 years who did not have p53 mutations.

EGFR and PDGFR-α Gene Amplification/Protein Overexpression

As shown in Table 1, EGFR gene amplification was detected only in one anaplastic astrocytoma (10%) and in two glioblastomas (11%); we saw no amplification in any of the low-grade diffuse astrocytomas. In addition, those tumors positive for EGFR amplification were negative for p53 mutations. In contrast, no amplification was observed for the PDGFR-α.

We found that overexpression of EGFR protein was somewhat more frequently encountered. Of the 28 malignant astrocytomas analyzed (anaplastic astrocytomas and glioblastomas combined), 28.6% showed overexpression, although immunopositivity to EGFR appeared to be confined to small foci of tumor cells. Given that the sample sizes were not equal, glioblastomas appeared to be more likely to contain high levels of EGFR than were anaplastic astrocytomas (33% vs. 20%, respectively). Overexpression of PDGFR-α protein was less frequently found in three diffuse astrocytomas, two anaplastic astrocytomas, and five glioblastomas, respectively. Two anaplastic astrocytomas and five glioblastomas showed overexpression of PDGFR-α, which was mutually exclusive with EGFR overexpression.

Analysis of LOH

Tables 1 and 2 and Fig. 1 also illustrate the frequency of LOH in our collection of tumors. LOH on chromosome 1p was detected in 2 of 18 glioblastomas only (11%). Each case, which had few oligodendroglial components, revealed allelic loss at either D1S548 or D1S508. LOH on chromosome 19q was found in 1 of 16 diffuse and 1 of 10 anaplastic astrocytomas (6.2% and 10%, respectively) and in 4 of 18 glioblastomas (22%); two of the affected glioblastomas, cases 11 and 16, showed a common region of deletion (CRD) on 19q13.1–13.3 between D19S425 and D19S596. It should also be noted that, even though LOH on 1p and 19q is an overall infrequent event in pediatric compared to adult astrocytomas, we did not detect this particular alteration in tumors from children younger than six years of age.

Fig. 1.

Allelic patterns of chromosomes 1p, 19q, 10p, and 10q in 10 anaplastic astrocytomas and 18 glioblastomas from pediatric patients. A G-banded ideogram of chromosome is shown on the left, and the corresponding polymorphic loci are shown to the right of the ideogram. Case numbers are listed at the top of each column. The CRDs on chromosomes 19q and 10q are indicated by the brackets on the right.

All of the diffuse astrocytomas in our study were negative for LOH on chromosome 10p and/or 10q. We did detect LOH on both 10p and 10q in 3 of 18 glioblastomas (16.7%), all of which were from children younger than six years; this was interpreted as the loss of an entire copy of chromosome 10. 10q LOH occurred in 2 of 10 anaplastic astrocytomas (20%) and in 5 of 18 glioblastomas (27.8%), frequencies that are somewhat lower than is reported in adult tumors. The deletions on 10q in one of the anaplastic astrocytomas and two of the glioblastomas were partial; these tumors were removed from older children. However, in each of the seven tumors demonstrating 10q LOH, the CRD occurred on 10q24 distal to D10S1264 and encompassing DMBT1 (Mollenhauer et al., 1997) and FGFR2 (Moschonas et al., 1996).

LOH on chromosome 22 was detected among all classifications of tumors across all age groups but seemed to be more frequent in tumors in children older than six years (Table 2). Seven of 16 grade II diffuse astrocytomas (~44%), 4 of 10 anaplastic astrocytomas (40%), and 11 of 18 glioblastomas (61%) exhibited LOH in at least one locus on 22q. We identified two CRD at 22q12.3 (proximal CRD [PCRD]) and at 22q12.3–13.32 (distal CRD [DCRD]) in the 11 primary glioblastomas studied (Fig. 2). The PCRD was defined with proximal and distal boundaries marked by D22S1176 and D22S1172, respectively, and spanned a distance of 7.5 Mb, while the DCRD was defined with nine markers within a span of 30 Mb. In the five informative grade II diffuse astrocytomas occurring in older patients, the CRD appeared to be defined proximally by marker D22S280 and distally by D22S1172 (Fig. 2). Representative cases are depicted in Fig. 3.

Fig. 2.

Analysis of 22q LOH in 16 diffuse astrocytomas, 10 anaplastic astrocytomas, and 18 glioblastomas. Ideogram of chromosome 22 and the location and order of microsatellite markers are on the left; case numbers are listed at the top. Rectangular areas indicate CRDs in the cases of diffuse astrocytoma, in glioblastomas of younger patients, and in glioblastomas of older children.

Fig. 3.

GeneScan electrophorogram of several 19q and 22q markers in normal tissue and in glioblastoma. The numbers in the right corner of each panel refer to the ratio of larger to shorter allele. Arrows indicate LOH. Markers are indicated at the bottom of the panels. (A) Glioblastoma case 1 retains heterozygosity for both D19S219 and D19S596, while glioblastoma case 16 shows allelic loss for both D19S219 and D19S596. (B) Glioblastoma case 1 shows LOH for D22S1147 and D22S928 while retaining heterozygosity for D22S1176. For glioblastoma case 17, D22S1176 exhibits allelic loss, whereas D22S1147 and D22S928 exhibit retention of both alleles.

Discussion

With the exception of data involving p53 mutations, EGFR amplification/overexpression, and microsatellite instability in malignant pediatric astrocytomas (Alonso et al., 2001; Bredel et al., 1999; Pollack et al., 1997; Raffel et al., 1999; Sung et al., 2000), there is little in the published literature with which to compare the results of this study. Our results, which are at variance with those observed in analyses of adult astrocytic tumors, indicate that, although childhood astrocytic tumors seem to be similar histologically to their adult counterparts, they also have some distinctive molecular features. All of our glioblastoma cases were categorized as primary glioblastomas because none of them had a history of prior, precursory lower-grade astrocytomas. In contrast to primary glioblastomas occurring in older adults (Louis, 1997; Ohgaki et al., 2004), this series of pediatric tumors frequently exhibited p53 mutations, but rarely showed EGFR amplification/overexpression, suggesting that the biological behavior and molecular pathogenesis of most childhood glioblastomas may be more similar to secondary glioblastomas than to primary glioblastomas. This study further suggests the existence of distinct molecular tumorigenic pathways in younger versus older children, particularly based on the higher frequency of p53 mutations and LOH on 19q and 22q detected in tumors from children six or more years of age at diagnosis compared to those from children diagnosed at younger than six years of age.

The reported frequencies of p53 alterations in pediatric astrocytic tumors vary considerably. Although initial studies indicated that childhood malignant gliomas rarely exhibited p53 mutations (Lang et al., 1994b; Litofsky et al., 1994), recent reports have suggested that a subgroup of anaplastic astrocytomas and glioblastomas of childhood frequently exhibit such mutations (Raffel et al., 1999; Sung et al., 2000; Sure et al., 1997). The overall incidence of p53 mutation in 27% of our lesions is broadly similar to the results obtained from pediatric astrocytic tumors by other authors (Pollack et al., 1997; Sure et al., 1997). It is worth noting that the p53 mutation rate of pediatric glioblastomas is apparently much less than the 60%–80% detected in adult diffuse astrocytomas and secondary glioblastomas.

EGFR amplification occurs in fully one-third of adult glioblastomas, while the majority of high-grade pediatric astrocytomas apparently lack EGFR amplification. These results both refute previously published negative data (Cheng et al., 1999; Raffel et al., 1999; Sung et al., 2000) and corroborate other, positive studies (Sure et al., 1997; Wasson et al., 1990). EGFR gene amplification was not found in the low-grade astrocytomas of our series, which is consistent with what is generally accepted (Kleihues et al., 2000). Therefore, we think it safe to postulate that overexpression/amplification of EGFR does not seem to be a major genetic factor in the development of these tumors, regardless of their clinical history.

The number of specimens in this study did not allow for a meaningful comparison of the frequency of EGFR amplification correlated with different locations or specific histologic subtypes. In one study, EGFR amplification was a feature of supratentorial malignant gliomas of childhood, although found at a lower frequency than in adult tumors (Bredel et al., 1999). Bredel et al. (1999) further reported that, among three tumors showing EGFR amplification, one glioblastoma was of a small-cell type, indicating a possible relationship between EGFR amplification and small-cell histology, at least in adult patients (Burger et al., 2001). All three of the malignant gliomas showing EGFR amplification in our series had a hemispheric location, but they showed the classical glioblastoma histology. In addition, several studies report that, in adult tumors, EGFR amplification and p53 mutations seem to be mutually exclusive. None of the tumors with amplified EGFR exhibited p53 mutations in our study. Based on this, we can hypothesize that astrocytic tumors of both adults and children share at least some molecular mechanisms of tumor progression and, further, that pediatric glioblastomas are more similar to secondary glioblastomas, even if they differ in some clinicopathologic aspects.

PTEN mutations were detected in only two (11%) glioblastoma cases. This lower frequency is similar to that found in the study of adult secondary glioblastomas. In addition, we found no consistent correlation between PTEN mutations and EGFR amplification as in primary glioblastomas, even though our figures are too low for a definite conclusion. Raffel et al. (1999) pointed out that the PTEN mutation was the only alteration of the genes examined that was significantly associated with survival. Due to the small size of our series and low frequency of PTEN mutations, it will be necessary to perform a validation study with a larger cohort of tumors to address the reliability of this finding.

We did not observe amplification of the PDGFR-α gene, but this amplification in pediatric gliomas has been previously reported in one study (Di Sapio et al., 2002). PDGFR-α gene amplification was detected in only 1 of 38 gliomas, and Di Sapio et al. (2002) suggested a possible role of PDGFR-α in the progression of pediatric gliomas. In our series, as in adult cases, PDGFR-α overexpression and EGFR overexpression did not occur in the same tumors. In addition, PDGFR-α is expressed approximately equally in all grades of tumors. We cannot exclude the possibility that such overexpression could be related to the initial stages of pediatric astrocytoma formation.

LOH analyses have reported losses on 1p/19q in pediatric astrocytomas, and there is evidence in the adult for the presence of tumor suppressor gene(s) on 19q involved in the malignant progression of astrocytomas (Nakamura et al., 2000; Smith et al., 1999). In the present study, LOH at 1p/19q was significantly less common in pediatric tumors compared to adult lesions of similar size and malignancy (Nakamura et al., 2000; Watanabe et al., 2002). These data have several important implications. One possible explanation for lower LOH frequency in pediatric patients is that the histologic distribution of tumors in the pediatric age group differs from that in adults, since LOH is observed most commonly in gliomas with oligodendroglial features. Clearly, pediatric oligodendroglial tumors are uncommon, and the small number of oligodendroglial lesions represented in our series might preclude definitively identifying the association between 1p/19q loss and pathogenesis in this subset. However, our previous LOH study in secondary glioblastomas, which had few oligodendroglial components, showed that 19q loss was more common in secondary glioblastomas than in primary glioblastomas (54% vs. 6%, respectively; Nakamura et al., 2000). Thus, differences in histologic distributions between pediatric and adult tumors do not provide an obvious explanation for the observations noted in the present study.

Genomic loss on 10q appears to be a common chromosomal change overall and showed a similar frequency in pediatric anaplastic astrocytomas and glioblastomas (30% vs. 28%, respectively). The frequency of 10q LOH, however, is somewhat lower than that in adult primary glioblastomas but is closer to the LOH rate seen in secondary glioblastomas in adults (Fujisawa et al., 2000). Multiple tumor suppressor genes on the long arm of chromosome 10 have been implicated in the development of astrocytic tumors (Ichimura et al., 1998), and LOH on 10q has been found more frequently in primary than in secondary glioblastomas; 10q LOH in primary glioblastomas is also associated predominantly with PTEN mutations (Fujisawa et al., 2000; Tohma et al., 1998). Although losses on 10p seem to play little or no role in our pediatric cases, we do note an intriguing concomitant LOH on 10q in the only three tumors that showed LOH in 10p. In each case, this suggests a complete chromosomal deletion, an event detected more commonly in primary glioblastomas (Fujisawa et al., 2000).

We found that 44% of low-grade astrocytomas, 40% of anaplastic astrocytomas, and 61% of glioblastomas exhibited LOH at 22q in our panel. This is comparable to 22q LOH detected in 33% of adult diffuse astrocytomas, in 40% of adult anaplastic astrocytomas, and in 53% of adult glioblastomas; these frequencies are further consistent with a role of perturbation in 22q as an early genetic event in astrocytoma tumorigenesis and progression (Nakamura et al., 2005b). To our knowledge, our study is the first report of a high frequency of LOH at 22q in each grade of pediatric astrocytic tumors. We specifically identified a CRD at 22q12.3 between D22S1176 and D22S1172 in a majority of cases. This chromosomal region includes the tumor suppressor gene TIMP-3 and is a segment frequently deleted in secondary glioblastomas as well as in some diffuse astrocytomas. However, the deleted region is large enough that we cannot exclude involvement of other tumor suppressor gene(s).

Another interesting observation is that astrocytic tumors from adolescents and children older than six years had significantly higher frequencies of 19q and/or 22q deletions than did those from younger children. In addition, glioblastomas from the older children showed 22q LOH at distal region 22q12.3–13.31, as is found in adult primary glioblastomas (Nakamura et al., 2005b). This indicates that glioblastomas in older children follow a molecular pathway distinct from that in very young children, and that the molecular phenotypes of tumors from older children tend to have a greater overlap with that of adult secondary glioblastomas. Since the pattern of progression to glioblastoma is not characteristically observed in pediatric glioblastoma, however, it is not certain whether such tumors are truly analogous to each other. And whereas an age correlation has been noted for p53 mutations, the small number of cases also precludes determination of whether similar chromosomal losses in secondary glioblastomas begin to occur more commonly in older children. Pollack et al. (1997) showed that p53 mutations were rare in tumors from children younger than four years, but in tumors from older children, the percentages were similar to those found in adults. These data are in line with those obtained from studies on adult glioblastomas, suggesting the existence of at least two different subsets of childhood glioblastomas. Although we cannot clearly define genetic pathways characteristic of pediatric astrocytoma progression, specific molecular pathogenic age-related differences might exist in the development of pediatric tumors, especially in younger children. Further efforts to identify minimal common deleted regions or some genetic alterations relevant to pediatric astrocytomas should be supported.

In summation, our data suggest that pediatric astrocytic tumors follow genetic pathways different from those operating in adults. Further, the association between age and genetic alterations among pediatric glioblastomas indicates the probable existence of distinct pathways of molecular tumorigenesis in younger versus older children. Taken together, although the changes detected in pediatric glioblastomas are comparable to those observed in the development of adult secondary glioblastomas, there are sufficient differences in frequency and occurrence to say that the presumed tumorigenic pathways do not always fit.