Analysis of Oncogene and Tumor Suppressor Gene Alterations in Pediatric Malignant Astrocytomas Reveals Reduced Survival for Patients with PTEN Mutations

INTRODUCTION

During the past 10 years, a fairly extensive body of literature has developed from the analysis of gene alterations in adult malignant astrocytomas (reviewed in Refs. 1 and 2 ). However, there is little information available on the genetics of corresponding tumors from the pediatric population. The emphasis on studying adult malignant astrocytomas for gene alterations is a consequence of the relatively low frequency of cancer in children combined with the fact that PMAs3 constitute a small proportion of brain tumors in children: perhaps 5–10% of childhood intracranial neoplasms are of this type ,(3 , 4) .

The majority of work that has been published on PMAs involves the TP53 tumor suppressor gene (5, 6, 7, 8, 9, 10) , and with reference to these studies, there is considerable variation in the reported TP53 mutation rates. This problem is typical of much of the literature dealing with molecular genetics of PMAs; consequently, it is unclear whether the gene alterations associated with the development of adult astrocytomas occur at comparable frequencies in their pediatric counterparts.

This dearth of information concerning the genetic etiology of PMAs is particularly unfortunate because there is no other group of brain tumors for which patient outcome is so poor. With respect to clinical implications, there are at least two major benefits that would likely result from their genetic analysis. The first involves the identification of markers useful for predicting the clinical course of the disease and length of survival. The other involves the ability of molecular genetic investigations to provide insights concerning the fundamental mechanisms of tumor development and, in doing so, provide information about potential therapeutic targets.

With regard to outcome analyses, hundreds of studies have been published that address survival as a function of tumor genotype. For malignant astrocytomas, such studies have generally involved adult tumor cohorts, and the related investigations have usually failed to demonstrate a consistent relationship between tumor genotype and patient survival when the prognostic significance of specific gene alterations is viewed in the context of other clinical parameters. In particular, tumor malignancy grade and patient age are invariably identified as the most reliable predictors of outcome (11 , 12) . It is because of the contribution of increasing age to the decreased survival of adult astrocytoma patients that genetic outcome analyses of PMA patients should be particularly interesting; a comparable age effect is not anticipated in pediatric patients because their overall health condition should not be adversely affected by increasing age.

In association with the preceding considerations, we have examined a relatively large cohort of PMAs for gene alterations that commonly occur in adult astrocytomas. Our results suggest a significant distinction in the molecular etiology of PMAs with respect to their adult counterparts and further indicate that genetic testing of these tumors may be useful for predicting their clinical behavior.

MATERIALS AND METHODS

Isolation of DNA.

H&E-stained tissue sections were reviewed, and areas of embedded tissues that contained at least 70% tumor cells were identified. These areas were dissected from the corresponding unstained sections, and tissues were deparaffinized in Histoclear (National Diagnostics, Atlanta, GA) for 10 min before DNA extraction with the QIAamp Tissue Extraction Kit (QIAGEN Inc., Arlington Heights, IL).

Assessment of Gene Dosage.

Competitive (multiplex) amplifications were performed in 50-μl volumes containing 200 μm of each dNTP, 1.5 mm MgCl₂, 50 mm KCl, 10 mm Tris-HCl (pH 8.3), 0.001% gelatin, and 1.25 units of Amplitaq Gold (Perkin-Elmer, Foster City, CA) and the following primer (target and reference gene) amounts: (a) 8 and 12 pmol of EGFR and CF primer pairs, respectively; (b) 10 pmol each for CDK4 and DR primer pairs; (c) 13.3 and 6.7 pmol of MDM2 and DR primer pairs, respectively; and (d) 10 pmol each of CDKN2A and APEX primer pairs. Reaction conditions for CDKN2A amplifications were for 43 cycles using a touchdown profile previously described by Burns et al. (13) . For analysis of EGFR, CDK4, and MDM2 amplification, the PCR profiles were 95°C for 9 min; followed by 43 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 1 min; followed by extension at 72°C for 10 min. With the exception of CDK4, all primer sequences have been described previously (13 , 14) . For CDK4, the primers used were AGTTCGTGAGGTGGCTTTACTGAGG (forward) and CTCTCATTATTTCCTCAGGGTCCCC (reverse).

Amplification products were separated in 4% agarose gels that were stained in 0.5 μg/ml ethidium bromide for 30 min at room temperature and then destained in deionized water for 30 min before quantitation of band intensities using a Gel Doc 1000 photodocumentation system (Bio-Rad) and its associated software (Molecular Analyst). The procedure used to assess CDKN2A gene dosage is essentially as we have described previously for the detection of PTEN homozygous deletions (15) . In brief, standard curves were generated from a series of samples containing varying proportions of normal DNA mixed with DNA from a glioblastoma cell line having a CDKN2A homozygous deletion (U87; American Type Culture Collection). Normalized CDKN2A signal intensities (CDKN2A:APEX signal ratios) from tumor DNAs were plotted against standard curves, and cases having less than 20% the CDKN2A signal found in normal DNA were scored as having homozygous deletions. EGFR, CDK4, and MDM2 reaction products from tumor DNAs were normalized against the signal produced by the fragment for the internal reference gene, and these ratios were compared against the corresponding ratios derived from normal DNAs and from DNAs extracted from tissue sections of adult astrocytomas that had been previously identified as having either EGFR, CDK4, or MDM2 amplification (16) . Specimens with target:reference gene signal ratios that were more than three times that determined in normal DNAs were scored as having gene amplification (14) .

Preparatory PCR for DNA Sequencing.

To prepare PCR products for sequencing, 1 μl of stock DNAs was amplified in a 50-μl volume containing 20 pmol of forward and reverse primer, 200 μm of each dNTP, 1.5 mm MgCl₂, 50 mm KCl, 10 mm Tris-HCl (pH 8.3), 0.001% gelatin, 5% DMSO, and 1.25 units of Amplitaq Gold (Perkin-Elmer). Reactions were carried out in a Perkin-Elmer 9600 under the following conditions: 95°C hold for 9 min; followed by 50 cycles of 95°C for 30 s, 55°C or 48°C for 30 s, and 72°C for 1 min; followed by a 10-min hold at 72°C. The primers used for the amplification of PTEN coding sequences (each of the nine exons) and TP53 coding sequences (exons 5–8) have been described previously (17 , 18) .

Sequencing.

Sequencing reactions (6.5 μl) were prepared that contained 100 ng of PCR product from the preparatory PCRs, 2 pmol of primer, 18 mm Tris-HCl (pH 9.5), 4.5 mm MgCl₂, 2.3 μm cold dNTPs, 0.023 μm[α-³³P]dideoxynucleotide triphosphates, and 8 units of ThermoSequenase DNA polymerase (Amersham Life Science, Inc., Arlington Heights, IL). Sequencing reactions were amplified for 30 cycles at 95°C for 20 s, 58°C for 30 s, and 72°C for 1 min using a 1 min ramp time between annealing and elongation phases. After sample denaturation, reaction products were loaded onto a 6% sequencing gel (19:1 acrylamide, 7 m urea, 0.5 × Tris-borate EDTA, and 15% formamide). Electrophoresis was at 75 W at room temperature for 1 h, after which the gels were dried and exposed to Kodak XAR film. Establishment of a tumor mutation was based on the identification of the same sequence alteration in two separate PCR templates generated from the same sample.

Statistical Analysis.

Survival distributions from the date of tissue acquisition were estimated with Kaplan-Meier curves. Comparisons of patient subsets for each clinical or genetic variable were performed using log-rank tests.

RESULTS

A total of 48 PMAs obtained from patients treated at the Mayo Clinic since 1985 were identified through a review of pathology registries and an examination of H&E-stained tissue sections. Blocks containing tissue that yielded DNA of sufficient quantity and quality to support the analysis of all genes of interest were available for 39 of these cases (Table 1)<$REFLINK> . With only one exception, each of the 39 tumors was classified as an astrocytoma; case OA2-1 was determined to be of mixed composition, but it was included in this cohort as a result of its predominant astrocytic character.

Two types of molecular genetic analyses were performed on the DNAs extracted from the archival tissue: (a) competitive PCR reactions to detect tumor-associated alterations in the EGFR, CDK4, MDM2, and CDKN2A gene copy number; and (b) sequence analysis of PCR fragments containing the TP53 and PTEN coding sequence. Because CDKN2A exon 2 sequence was targeted in competitive PCR reactions for dosage analyses of this tumor suppressor gene, specimens determined to have a CDKN2A homozygous deletion would also have homozygous deletion of ARF (19) . Results from the genetic analysis of individual tumors are presented in Table 1<$REFLINK> , and the cumulative data are summarized in Table 2<$REFLINK> . Most notable is the absence of EGFR amplification in each and every specimen (Fig. 1)<$REFLINK> . Based on the incidence of EGFR amplification in adult malignant astrocytomas (20, 21, 22) , one would expect to find eight to nine cases with EGFR amplification in a series of this size and malignancy grade composition.

To create a standard with which to compare the results of gene alteration-survival analyses, patient survival was examined in the context of tumor malignancy grade, a variable that represents one of the most important prognostic factors for adult astrocytomas. Although there was a significant difference in the length of survival when all three malignancy grades were considered (P = 0.002), this significance was not evident when the comparison only involved the grade 3 and grade 4 tumor groups (P = 0.27; Fig. 2A<$REFLINK> ). There was no correlation between patient age and survival for this series of tumors (data not shown).

Because of the lack of a significant association between survival and grade 3 versus grade 4 malignancy classification, we restricted our gene alteration-survival analyses to these two groups of tumors to determine whether the status of any of the genes we had examined might prove useful in predicting outcome for patients with intermediate or high-grade astrocytoma. For patients with TP53 mutations, there was no significant difference in survival compared with patients whose tumors showed only wild-type TP53 sequence (P = 0.49; Fig. 2B<$REFLINK> ). Because MDM2 amplification and ARF homozygous deletion have been suggested to result in the functional inactivation of p53 (23 , 24) , we also compared the survival of patients with either TP53 mutation, MDM2 amplification, or ARF homozygous deletion with the survival of patients whose tumors lacked each of these alterations. This analysis also failed to reveal a significant difference in survival between groups defined in such a manner (P = 0.80).

Patients with amplification of CDK4 or a homozygous deletion of the gene encoding the cdk4 inhibitor p16 did not show a significant difference in survival relative to patients whose tumors did not have either of these alterations (P = 0.60; Fig. 2C<$REFLINK> ). In contrast, the presence of a PTEN mutation (examples are shown in Fig. 3<$REFLINK> ) was strongly associated with decreased survival among grade 3 and grade 4 tumor patients (P = 0.006; Fig. 2D<$REFLINK> ).

DISCUSSION

With the exception of the data involving TP53 mutations, there is little in the published literature with which to compare the results of this study for consistency . For TP53 alterations in PMAs, mutation frequencies between 0% and 66% have been reported in grade 4 astrocytomas, and mutation frequencies between 0% and 50% have been reported anaplastic astrocytomas (5, 6, 7, 8, 9, 10) . Based on data from these studies, the cumulative calculated TP53 mutation frequencies are 31.3% (19 of 51) and 9.7% (3 of 31) in grade 4 and grade 3 PMAs, respectively. These values could lead one to believe that TP53 mutations are associated with increasing astrocytoma malignancy and might therefore suggest that this gene alteration is linked with tumor progression. The opposite is thought to be true for adult astrocytomas, where TP53 mutation frequencies are highest in grade 2 and grade 3 tumors, and, consequently, this gene alteration is generally considered to represent an early event in the development of adult astrocytoma (25 , 26) . In this study, TP53 mutation frequencies of 20% and 24% were observed in grade 4 and grade 3 astrocytomas, respectively; one of seven grade 2 tumors examined showed a TP53 mutation. Importantly, there was no indication that TP53 mutations, or what have been described as functionally equivalent gene alterations (MDM2 amplification or ARF homozygous deletion; Refs. 23 and 24 ), were predictive of outcome for PMA patients. Consequently, these data contrast with results presented in a previous report that indicated that TP53 mutations are correlated with reduced survival for children with malignant astrocytoma (8) .

Gene alterations leading to increased cdk4 activity, either CDKN2A homozygous deletion or CDK4 amplification, have not been examined previously in PMAs. Here these alterations were observed in 17.6% (3 of 17) and 13.3% (2 of 15) of grade 3 and grade 4 tumors, respectively. In previous investigations of adult astrocytomas, CDKN2A homozygous deletion and CDK4 amplification have primarily been identified in grade 3 and grade 4 malignant astrocytomas (27 , 28) , and our data are consistent with these findings because no CDKN2A or CDK4 alterations were identified in grade 2 tumors. As was the case for alterations affecting p53 function, alterations leading to increased cdk4 activity were not predictive of survival for PMA patients.

PTEN mutations were observed in 20% of the grade 4 tumors (3 of 15), 5.9% of the grade 3 tumors (1 of 17), and 0% of the grade 2 tumors. These data are consistent with results associated with the study of adult astrocytomas that indicate that PTEN mutations are highly correlated with increasing malignancy grade and occur primarily in tumors of grade 4 malignancy (29) . Of the genes examined in this study, mutations of PTEN represent the only alteration that shows a significant association with survival (P = 0.006; Fig. 2D<$REFLINK> ). The potential importance of this observation is underscored by the lack of prognostic significance associated with grade 3 versus grade 4 malignancy for these tumors (Fig. 2A)<$REFLINK> , and these are the only malignancy grades of astrocytoma in which PTEN mutations have been observed. Due to the small size of this series of PMAs, it will be necessary to perform a validation study with a larger cohort of tumors to address the reliability of this finding, as well as the findings involving the other gene alterations we have investigated. It is worth noting, however, that our results involving PTEN are consistent with data presented in a recent study of adult grade 4 astrocytomas that indicate that PTEN deletions are a negative prognostic indicator (30) .

In comparing the incidence of specific gene alterations in this cohort of PMAs against data published from the study of adult astrocytomas, the only result that supports an important difference between the genetic etiologies of the two groups of tumors concerns the incidence of EGFR amplification. None of the 15 grade 4 or the 17 grade 3 tumors showed evidence of this gene alteration, and this result is significantly different from the results one would anticipate from the analysis of a series of adult astrocytomas having a similar size and malignancy grade composition. Two other studies have reported data from the analysis of EGFR amplification in PMAs. In one study (5) , amplified EGFR was identified in one of four glioblastomas (25%) and in zero of two anaplastic astrocytomas; analysis of this number of specimens did not allow for a meaningful comparison of the frequency of EGFR amplification in adult versus pediatric tumors. In the second study, no EGFR amplifications were observed among 11 anaplastic astrocytomas and 13 glioblastomas (31) . Our data are consistent with the results presented in the latter report because there were no amplifications of EGFR in a combined 32 cases of grade 3 and grade 4 tumors: this finding points to an important difference between the development of pediatric and adult malignant astrocytomas.

The lack of EGFR amplification in these tumors has important implications for the clinical management of PMAs. Several experimental strategies for the treatment of adult malignant astrocytomas are based on the elevated expression of either mutant or wild-type epidermal growth factor receptor (32, 33, 34, 35) ; the results presented here suggest that such therapeutic approaches may not be effective if used on children with malignant astrocytomas. Conversely, it appears as though TP53 mutations occur at similar frequencies in both pediatric and adult astrocytomas and that therapies designed to exploit aberrant p53 function in tumors cells (36) may be effective for treating some of the astrocytic tumors in both age groups.

This investigation was based entirely on the analysis of DNA extracted from paraffin-embedded material because the utilization of this type of tissue does not require a lengthy prospective collection of specimens that should be anticipated for a relatively rare type of tumor. The lack of reference (normal) DNA for such a study is unfortunate because one might expect constitutional genetic lesions to play a significant role in childhood cancer as compared to cancer occurring in an elderly population. For instance, patients with Cowden’s syndrome are known to be predisposed to tumor development as a result of inheriting a defective PTEN gene (37) . Although our analysis of clinical information for the four patients with PTEN mutations gave no indication of a familial history of cancer, the possibility of individuals in this study having germ-line PTEN or TP53 mutations persists. However, the need to establish a foundation for understanding of the genetic basis of PMAs offsets the liability associated with the lack of reference DNA, and we have demonstrated that extrapolations from the study of adult astrocytomas can lead to erroneous conclusions regarding the molecular etiology of PMAs. Further determination of both the similarities and differences between pediatric and adult astrocytomas will aid in the development of targeted, individualized therapies that will be of benefit to all individuals afflicted with this type of cancer.