Congenital Glioblastoma: A Clinicopathologic and Genetic Analysis

Abstract

Congenital central nervous system (CNS) tumors are uncommon, accounting for 1% of all childhood brain tumors. They present clinically either at birth or within the first 3 months. Glioblastoma (GBM) only rarely occurs congenitally and has not been fully characterized. We examined clinicopathologic features and genetic alterations of six congenital GBMs. Tumors were seen by neuroimaging as large, complex cerebral hemispheric masses. All showed classic GBM histopathology, including diffuse infiltration, dense cellularity, GFAP‐positivity, high mitotic activity, endothelial proliferation and pseudopalisading necrosis. Neurosurgical procedures and adjuvant therapies varied. Survivals ranged from 4 days to 7.5 years; two of the three long‐term survivors received chemotherapy, whereas the three short‐term survivors did not. Paraffin‐embedded tissue sections were used for FISH analysis of EGFR, chromosomes 9p21 (p16/CDKN2A) and 10q ( PTEN/DMBT1); sequencing of PTEN and TP53; and immunohistochemistry for EGFR and p53. We uncovered 10q deletions in two cases. No EGFR amplifications, 9p21 deletions, or mutations of TP53 or PTEN were noted; however, nuclear p53 immunoreactivity was strong in 5/6 cases. Tumors were either minimally immunoreactive (n = 3) or negative (n = 3) for EGFR. We conclude that congenital GBMs show highly variable survivals. They are genetically distinct from their adult counterparts and show a low frequency of known genetic alterations. Nonetheless, the strong nuclear expression of p53 in these and other pediatric GBMs could indicate that p53 dysregulation is important to tumorigenesis.

INTRODUCTION

Congenital central nervous system (CNS) tumors are uncommon, occurring in 1.1–3.6/100 000 newborns and representing 1%–4% of all childhood brain tumors (6, 12, 16, 20, 24, 25). They generally present clinically either at birth or within the first 3 months of life, but there are now numerous examples of prenatal detection (19, 26). The most frequent congenital brain tumors are teratomas, astrocytomas, PNETs and choroid plexus papillomas, with the majority of these presenting in the cerebral hemispheres rather than the posterior fossa (5, 14, 24). Glioblastoma, the highest grade astrocytoma (WHO grade IV), is the most frequent primary brain tumor of adults, but is much less common in childhood and only rarely arises as a congenital neoplasm. Indeed, GBMs account for only 3% of congenital brain tumors and less than 30 cases have been reported (6, 36). In most instances, the diagnosis of GBM carries a dismal prognosis: survivals average 50–60 weeks in adults, with only slightly better outcomes in children. Case reports and small series of congenital GBMs have suggested variable survivals, with the large majority of patients dying within weeks to months, but occasional patients surviving for years (26, 36).

Genetic alterations associated with adult GBMs have been well characterized and include EGFR amplification, p16/CDKN2A deletions, PTEN and TP53 mutations and frequent losses of chromosome 10q (3, 10, 17). In contrast, genetic features of pediatric GBMs are poorly defined and some of the most characteristic genetic alterations present in adult GBMs (ie, EGFR amplification and PTEN mutations) are uncommon in pediatric GBMs (4, 7, 10, 21, 23, 27, 28, 30, 31). Congenital GBMs have not been genetically characterized and it remains unclear if these tumors have their own unique genetic signature or might resemble other pediatric GBMs. We have therefore assembled the largest series of congenital GBMs to date, including six cases, in order to investigate their genetic alterations and clinico‐pathologic features.

MATERIALS AND METHODS

Cases consisted of one autopsy and five surgical specimens obtained from institutional and consultation services of the authors between 1998 and 2005. Data regarding clinical presentation, neuroradiologic appearance, neurosurgical procedures, adjuvant therapies and follow‐up were obtained through review of patient records or through the referring institution’s Tumor Registry. Extent of resection was defined as biopsy, subtotal resection (STR) or gross total resection (GTR), based on the information received from the treating clinician or coded in the Tumor Registry at the treating institution.

Tissues were fixed in 10%‐buffered formalin, routinely processed, paraffin‐embedded, sectioned at 4 to 6 microns and stained with hematoxylin and eosin (H&E). Slides were assessed for histologic differentiation, nuclear features, cellularity, infiltration, mitotic activity, necrosis and vascular proliferation. Only tumors that fulfilled diagnostic criteria of GBM by WHO criteria were included.

For immunohistochemical studies, sections were deparaffinized and subjected to heat‐induced epitope retrieval by steaming for 15 minutes. Slides were then incubated at room temperature with antibodies directed toward glial fibrillary acidic protein (GFAP, monoclonal, 1:100, Dako Co., Carpinteria, CA, USA), synaptophysin (monoclonal, 1:50; Boehringer Mannheim, Mannheim, Germany), MIB‐1 (monoclonal, 1:150, Immunotech, Inc., Westbrook, ME, USA), EGFR (monoclonal, 1:200 clone 528; Oncogene Science, Cambridge, MA) and p53 (mouse monoclonal, 1:20; DO‐7 DAKO Co., Carpinteria, CA). Antibodies were detected using the avidin‐biotin‐peroxidase complex (ABC) method. The antigen‐antibody reaction was visualized using 3,3′‐diaminobenzidine as the chromogen. Standard positive controls were used throughout, normal serum serving as the negative control. For quantitating the mean percent of cells staining with p53, EGFR and MIB‐1, 500 tumor cells were counted manually three times. As known nuclear antigens, only nuclear staining for p53 and MIB‐1 was considered as positive. Antibodies directed at EGFR stain the cytoplasm and cell membrane and only cells with this staining pattern were considered positive. Tumor cell staining for p53 and EGFR was graded as 0 if no cells stained; 1+ if 0%–10% stained; 2+ if 10%–25% stained; 3+ if 25%–50% stained; and 4+ if >50% stained.

FISH analysis of  10q, 9p21 and  EGFR.

Dual‐color FISH was performed and interpreted as previously reported on 5 µm sections derived from tissue blocks (9). A commercial probe cocktail for the p16 gene region on chromosome 9p included a SpectrumGreen‐labeled centromere enumerating probe (CEP9) and SpectrumOrange‐labeled 9p21 ( p16/CDKN2A region) (Vysis, Downers Grove, IL). Paired fluorescein isothiocyanate (FITC)/rhodamine‐labeled DNA probes included CEP7 (Vysis Inc., Downers Grove, IL)/EGFR (RPCI‐11 148P17) and PTEN (10q23)/DMBT1 (10q25‐q26) (both donated by Dr. Robert Jenkins, Mayo Clinic, Rochester, MN). Deparaffinization of the sections was carried out with two 10‐minute immersions in Citrisolv (Fisher Scientific, Pittsburgh, PA), followed by three 3‐minute immersions in isopropanol. Following rinsing, target retrieval was achieved by immersing the slides in citrate buffer, pH 6.0 and steaming for 20 minutes. Slides were exposed to a 0.4% pepsin (P‐7012; Sigma‐Aldrich, St. Louis, MO) digestion for 15 minutes at 37°C, and then a rinse in 2× standard saline citrate (SSC) on a rotator for 5 minutes. Paired probes were diluted from stock with tDenHyb hybridization buffer (Insitus Biotechnologies, Albuquerque, NM) to a concentration of 1:25. Slides were hybridized overnight in a 37°C humidified chamber, washed the following day and counterstained with 4,6‐diamidino‐2‐phenylindole (DAPI) in Fluorgard (Insitus).

TP53 mutations.

DNA was extracted from paraffin sections as reported previously (35). Prescreening for mutations in exons 4 through 8 of the TP53 gene by polymerase chain reaction–single‐strand conformational polymorphism analysis was carried out as described. Primers for exon 4 were 5′‐ACTGCTCTTTTCAC CCATCTAC‐3′ (sense) and 5′‐TCATG GAAGCCAGCCCCTCAG‐3′ (antisense). Samples showing mobility shifts in single‐strand conformational polymorphism analysis can be further analyzed by direct DNA sequencing on an automated sequencing system (ABI PRISM 3100 Genetic Analyzer; Applied Biosystems, Foster City, CA) using an ABI PRISM BigDye Terminator version 1.1 Cycle Sequencing Ready Reaction Kit (Applied Biosystems).

PTEN mutations.

Prescreening for mutations in exons 1 through 9 of the PTEN gene by polymerase chain reaction–single‐strand conformational polymorphism analysis was carried out as described previously (33). Samples showing mobility shifts in single‐strand conformational polymorphism analysis can be further analyzed by direct DNA sequencing on an automated sequencing system as described.

Clinicopathologic correlation.

The extent of surgical resection, the administration of radiation therapy or chemotherapy, the tumor’s histopathologic features, proliferation indices (mitotic rate and MIB‐1 index) and molecular genetic alterations were studied in relation to patient survival in order to establish any correlation between tumor features, treatment and clinical behavior.

RESULTS

Clinical/radiographic features.

Clinical and neuro‐imaging features are listed in Table 1. Tumors occurred in three boys and three girls ranging in age from 4 days to 12 weeks (mean: 25 days). Presenting symptoms included enlarged or enlarging head circumference, tense or bulging fontanelles, irritability, seizures, hypothermia, cranial nerve palsies and vomiting. All tumors affected the cerebral hemispheres of the supratentorial compartment and were centered in the frontal, parietal and occipital lobes (5 cases) or the hypothalamus (1 case). On MRI and CT scans, these congenital GBMs were solitary, large, complex masses that all showed significant enhancement following the administration of contrast agents (Figure 1). They typically occupied expansive regions of the affected cerebral hemisphere and extended across the midline. Outward displacement of the overlying skull on the side of the tumor was uniformly present. Cystic and necrotic regions were present in all tumors, while associated hydrocephalus was variably seen.

Table 1.

Clinical characteristics of congenital glioblastoma. Abbreviations: DOD = dead of disease; AWD = alive with disease.

Case No.	Age/sex	Signs/symptoms	Neuroimaging	Treatment	Outcome
1	7 weeks/F	Irritable; tense fontanelles; increasing head size	Large left cerebral hemispheric mass; ventricular dilatation; CSF dissemination	Subtotal resection	DOD, 4 months
2	6 days/M	Seizures on day of birth	Large left parietal mass; subdural hematoma; posterior cerebral infarct	Gross total resection; chemotherapy (vincristine, Cytoxan, cisplatin, etoposide)	AWD, 67 months
3	7 days/M	Bulging fontanelles; irritability	Large right cerebral hemispheric mass; ventricular dilatation	Subtotal resection	AWD, 90 months
4	6 days/M	Hypothermia at 24 h	Large enhancing hypothalamic mass; hydrocephalus; cerebral atrophy	Biopsy; chemotherapy (vincristine, Cytoxan, cisplatin, etoposide)	DOD, 78 months
5	12 weeks/F	poor feeding, vomiting, increasing head size; right facial drooping, nystagmus.	Large left parieto‐occipital tumor; hydrocephalus	Subtotal resection	DOD, 5 weeks
6	4 days/F	Large head, tense fontanelles	Large left parietal lobe and basal ganglia mass	None	DOD, 4 days

Figure 1.

Neuroimaging and macroscopic features of congenital glioblastoma. A. Axial, post‐contrast MRI demonstrates a large, heterogeneous, contrast‐enhancing mass centered in the parietal lobe of the right cerebral hemisphere and extending across the midline (Case 3). The skull and head are asymmetrically enlarged on the right and there is moderate ventricular enlargement on the left. B. H&E‐stained coronal section of postmortem brain with large, congenital glioblastoma involving the left cerebral hemisphere, centered in the basal ganglia and parietal lobe (Case 6). The tumor showed extensive regions of necrosis and caused significant midline shift and hydrocephalus.

Pathologic features.

All tumors were markedly hypercellular and showed an infiltrative growth pattern and astrocytic morphology (Figure 2). There was a homogeneous, sheet‐like growth of monotonous cells over large areas at low magnification. All tumors were of the fibrillary astrocytoma subtype, although the degree of fibrillarity was low in four cases and only moderate in two. Nuclei were generally small to moderate in size, oval and irregular, with hyperchomasia, but without nucleoli. Mitotic indices ranged from 7 to 42/10 HPF; mean: 18/10HPF. MIB‐1 indices were high, but variable, ranging from 10%–60% and were generally correlated with mitotic rates within individual tumors. All six tumors demonstrated both pseudopalisading necrosis and vascular proliferation. Thus, the tumors fulfilled WHO criteria for glioblastoma. As the differential diagnosis often included primitive neuroectodermal tumor (PNET), immunohistochemistry for synaptophysin and GFAP were performed in all cases. All congenital GBMs were strongly and diffusely positive for GFAP and were negative for synaptophysin.

Figure 2.

Pathologic features of congenital glioblastoma. A. All tumors showed marked hypercellularity and consisted of monotonous sheets of tumor cells with astrocytic morphology that displayed only modest amounts of cytoplasm or fibrillarity. B. “Pseudopalisading” necrosis was present in all tumors. C. All tumors were strongly and diffusely positive for glial fibrillary acidic protein (GFAP) by immunohistochemistry.

Analysis of genetic and immunohistochemical markers.

Results of FISH analysis for EGFR, 9p21 (p16/CDKN2A) and 10q (PTEN/DMBT1) copy numbers are summarized in Table 2. One case gave non‐informative results for all markers, most likely because of suboptimal nucleic acid integrity in this autopsy case (case 6). Deletions of 10q were evident in 2/5 cases (Figure 3). In both of these tumors (cases 4 and 5), we detected only one green (PTEN) signal and one red (DMBT1) signal in each nucleus, consistent with either a large 10q deletion or the loss of an entire chromosome (ie, monosomy 10). No EGFR amplifications or 9p21 deletions were identified in any of the cases. One case (case 4) showed polysomy for both chromosome 7 (EGFR/CEP7) and for chromosome 9 (9p21/CEP9). Single‐strand conformational polymorphism analysis of PTEN and TP53 genes demonstrated no mobility shifts on any of the cases, most likely indicating a lack of mutations in these genes.

Table 2.

Genetic and immunohistologic markers in congenital glioblastoma. Abbreviations: NI = non‐informative; NA = not available; wt = wild type; GBM = glioblastoma.

Case No.	Histology	FISH: EGFR	FISH: 10q ( PTEN )	FISH: 9p21 (CDKN2A)	TP53	PTEN	IHC: EGFR	IHC: p53	IHC: MIB‐1
1	GBM	Normal	Normal	Normal	wt	wt	1+	3+	38%
2	GBM	Normal	Normal	Normal	wt	wt	0	2+	60%
3	GBM	Normal	Normal	Normal	wt	wt	1+	2+	50%
4	GBM	Polysomy	Deletion	Polysomy	wt	wt	1+	0	12%
5	GBM	Normal	Deletion	Normal	wt	wt	0	3+	10%
6	GBM	NI	NI	NI	NA	NA	0	2+	35%

Figure 3.

Representative FISH results from congenital glioblastoma. A. Glioblastoma (GBM) with a normal disomic complement of 9p21. Most cells have two red (9p21) and two green (CEP9) signals. Some nuclei have less than two copies because of the truncation artifact encountered in thin tissue sections. B. Pattern of chromosome 10 deletion in a congenital GBM (case 4). Only one green (PTEN) signal and one red (DMBT1) signal are detected in each nucleus, most likely representing either a large 10q deletion or the loss of an entire chromosome 10.

Representative EGFR and p53 immunohistochemistry is illustrated in Figure 4. We found strong nuclear p53 immunoreactivity in 5/6 cases. In the positive cases, the percentage of neoplastic cells staining ranged from 15%–30% and therefore all were in the 2+ or 3+ category (Table 2). Immunohistochemistry for EGFR showed only focal (1+) staining in three cases, whereas the other three showed no staining (0) (Figure 4).

Figure 4.

p53 and EGFR expression in congenital glioblastoma. A. Immunohistochemistry for p53 showing strong (3+) nuclear immunoreactivity (Case 5). B. Immunostains for EGFR of (case 4) demonstrated mild (1+) reactivity focally.

Treatment, follow‐up and clinicopathologic correlation.

Among the six patients with congenital GBM, two received a gross total resection, two received a subtotal resection, one was biopsied and one received no neurosurgery (a postmortem diagnosis was established) (Table 1). Two patients received adjuvant chemotherapy (both with vincristine, Cytoxan, cisplatin, etoposide); the remaining four did not receive chemotherapy or radiation therapy. Patient survival was highly variable and ranged from 4 days to 7.5 years. There was no relation between extent of neurosurgical resection and survival. Two of the three patients who received chemotherapy had the longest survivals (DOD, 78 months; AWD, 67 months). The three patients with the shortest survivals did not receive adjuvant therapy. Five of six tumors showed strong (2+ or 3+) immunoreactivity for p53 and no association between p53 expression and survival could be established. Similarly, no associations could be established between survival and chromosome 10q loss, EGFR immunoreactivity, or MIB‐1 proliferation.

DISCUSSION

In the current study, we report six congenital GBMs that all presented as large complex, contrast‐enhancing masses in the cerebral hemispheres of children less than 3 months of age. Although the majority of childhood brain tumors show a predilection for the posterior fossa, it has been recognized that congenital brain tumors, including GBMs, occur most frequently in the supratentorial compartment (16). The differential diagnosis of a large malignant cerebral hemispheric tumor in a newborn includes teratoma (usually with immature elements), PNET and glioblastoma. The diagnosis of GBM in these cases was established based on the presence of an infiltrative, densely cellular, GFAP‐positive tumor with high proliferative activity, endothelial hyperplasia and necrosis. The most consistent histological feature among these cases was the monotonous, sheet‐like growth pattern of astrocytic tumor cells with small cell/fibrillary morphology in between necrotic foci. It should be emphasized that all of these tumors arose de novo, as there were no detectable lower grade precursor lesions (eg, ganglioglioma or low grade astrocytoma).

Genetic alterations in adult GBMs have been thoroughly investigated (10, 15, 17). These tumors have been classified as de novo GBMs if they arise without a low grade precursor and secondary if they have progressed from either a WHO grade II or III astrocytoma. De novo GBMs are characterized by EGFR amplifications (30%–40%) and PTEN mutations (20%–30%), whereas secondary GBMs are characterized by a high frequency of TP53 mutations (50%–60%) and a lower frequency of PTEN mutations. Both de novo and secondary GBMs show a high frequency of LOH for chromosome 10 (70%–90%) and loss of 9p21 (40%–70%), the locus of the p16/CDKN2A gene.

The genetic alterations associated with the development of pediatric malignant gliomas have not been as well defined as their adult counterparts (28). Most studies have concluded that many of the genetic alterations in adult GBM can be detected in pediatric GBMs, but are present at much lower frequency. In particular, there is a lower frequency of EGFR amplification, PTEN mutation and p16(CDKN2A) and p14ARF deletions in pediatric malignant gliomas (4, 7, 23). Brain stem gliomas may be an exception to this, in that they demonstrate both EGFR amplification and EGFR protein over‐expression (13). TP53 mutations occur in approximately 30%–40% of pediatric high grade gliomas, which is higher than the frequency in adult de novo GBMs but lower than adult secondary GBMs (21, 22).

The genetic alterations associated with congenital GBMs have not been described. It might be expected that congenital GBMs would occur in the setting of germline mutations of tumor suppressor genes (ie, TP53, PTEN, APC) or be tightly coupled with genetic syndromes associated with brain tumors such as neurofibromatosis, tuberous sclerosis, nevoid basal cell carcinoma syndrome, or von Hippel Lindau disease. However, children with these syndromes tend to develop brain tumors later in childhood and adolescence, while congenital neoplasms only rarely arise (6, 12, 16). One notable exception is the development of congenital rhabdoid tumors of the brain and kidneys in patients with germline INI1 mutations (2, 18). None of the tumors in this series were associated with familial syndromes or known germline mutations. Thus, most congenital brain tumors, including GBMs, appear to arise sporadically and are not associated with well‐defined genetic syndromes.

We uncovered a pattern of molecular and genetic alterations in congenital GBMs roughly similar to other pediatric GBMs and unlike de novo GBMs of adulthood. We demonstrated loss of chromosome 10q in two of five cases (40%), slightly less than other pediatric GBMs. However, we did not identify any EGFR amplifications, TP53 mutations, PTEN mutations or p16/CDKN2A deletions. Although the TP53 gene was not mutated in any of the cases, we found moderate to strong immunoreactivity for p53 protein in five of six cases. Immunoreactivity of p53 is typical in tumors with TP53 mutations; however other mechanisms can be related to its over‐expression, such as extensive DNA damage, p14ARF deletion or MDM2 amplification (22, 30). In a large study of prognostic factors in pediatric high grade gliomas, p53 over‐expression was statistically associated with a shorter survival, whereas TP53 mutation was not (21). We could not demonstrate a significant prognostic association between p53 expression and survival in this study as almost all tumors were p53 positive. However, the high percentage of tumors demonstrating strong nuclear staining for p53 might indicate that this tumor suppressor pathway is dysregulated in congenital GBMs by mechanisms other than TP53 mutation. One possibility is that p53 is upregulated in response to genetic instability associated with the higher rate of DNA mismatch repair (MMR) gene alterations in pediatric high‐grade gliomas (1). Alternatively, disruption of other components of the MDM2/p14/p53 pathway could lead to upregulated p53 (30).

The most striking finding within this small series of congenital GBM was the highly variable survivals. Glioblastoma is almost always a highly aggressive neoplasm. In adults, the mean life expectancy is 50–60 weeks in the setting of optimal surgical resection, radiation therapy and chemotherapy (29). Young age is a favorable prognostic factor for GBM and children with these tumors have slightly longer survivals, yet the 5‐year overall survival rates are still only 5%–15% (32). Clinical outcomes in children depend on many factors including tumor location and extent of resection. For example, malignant gliomas of the brain stem are particularly aggressive and associated with survivals less than a year (34). Previous reports of congenital GBMs indicate that most children succumb to disease within the first few months of life, while only rare patients survive longer than 1 year (26, 36). It was therefore surprising that half of the congenital GBMs in our series survived longer than 5 years (67, 78 and 90 months, with two patients alive at last follow‐up). In contrast, the other three patients died within 4 months of diagnosis. The differences in patient outcome among the short‐ and long‐term survivors in this small series could not be explained based on age, tumor location, extent of resection, histopathologic features, or genetic alterations. Two of three of the long‐term survivors were treated with chemotherapy (both received vincristine, Cytoxan, cisplatin and etoposide), whereas none of the three short‐term survivors received chemotherapy. While chemotherapy has been only modestly effective in the treatment of adults with GBM, there is now good evidence that specific chemotherapeutic regimens prolong survival in children and some infants with GBM (8, 11). The long survivals reported here and in other previously recorded cases should raise the possibility that some congenital GBMs are also highly responsive to chemotherapeutic regimens (36).

In summary, this series of six patients with congenital GBMs showed highly variable survival times, ranging from 4 days to 7.5 years. These tumors were genetically distinct from adult GBMs and showed a low frequency of alterations, similar to other pediatric GBMs. The strong nuclear over‐expression of p53 in these and other pediatric GBMs could indicate that p53 dysregulation is important to tumorigenesis.