Introduction

Glioblastoma multiforme (WHO Grade IV) is the most frequent and malignant neoplasm of the human nervous system. The majority of glioblastomas develop rapidly in older patients (mean age, approximately 55 years) after a short clinical history and without clinical or histological evidence of a less malignant precursor lesion (primary or de novo glioblastoma). Secondary glioblastomas manifest in younger patients (mean age, approximately 40 years) through progression from low-grade diffuse astrocytoma (WHO Grade II) or anaplastic astrocytoma (WHO Grade III) (Kleihues and Ohgaki, 1999). Recent studies have shown that these glioblastoma subtypes develop through different genetic pathways (Biernat et al, 1997; Kleihues and Ohgaki, 1999; Lang et al, 1994; Tohma et al, 1998; von Deimling et al, 1992; Watanabe et al, 1996). Primary glioblastomas are characterized by EGFR amplification/overexpression, PTEN mutations, p16INK4a homozygous deletion, and loss of heterozygosity (LOH) on chromosomes 10p and 10q (Biernat et al, 1997; Fujisawa et al, 2000; Kleihues and Ohgaki, 1999; Tohma et al, 1998; Watanabe et al, 1996), whereas secondary glioblastomas contain frequent p53 mutations and show LOH preferentially on chromosomes 19q and 10q (Fujisawa et al, 2000; Kleihues and Ohgaki, 1999; Nakamura et al, 2000b; Watanabe et al, 1996, 1997).

The retinoblastoma gene (RB1) at chromosome 13q14 was originally identified as the gene responsible for the development of retinoblastomas (Lee et al, 1987). It functions as a negative regulator of cell proliferation at the G1/S checkpoint of the cell cycle by complexing with cellular proteins, such as the transcriptional factor E2F (Harbour and Dean, 2000; Weinberg, 1995). LOH at the RB1 locus has been found in 25–45% of glioblastomas (Burns et al, 1998; Henson et al, 1994; Ichimura et al, 1996; Nakamura et al, 2000b; Ueki et al, 1996); however, sequencing of all 27 exons of the RB1 gene revealed inactivating mutations in only 5–12% of cases (Henson et al, 1994; Ichimura et al, 1996, 2000; Ueki et al, 1996), suggesting that other tumor suppressor gene(s) are present on chromosome 13q14. Like RB1 mutations in retinoblastomas (Hogg et al, 1993), most RB1 mutations in glioblastomas result in a truncated pRB protein (Henson et al, 1994; Ueki et al, 1996) that does not enter the nucleus. Homozygous deletions in the RB1 locus have been reported in a small fraction of glioblastomas (3 of 120, 3%) (Ichimura et al, 1996). Loss of RB1 expression was detected by immunohistochemistry in 5–27% of glioblastomas (Biernat et al, 1997; Burns et al, 1998; Henson et al, 1994; Nakamura et al, 1996; Ueki et al, 1996); however, there was no clear correlation between loss of RB1 expression and LOH at the RB1 locus (Burns et al, 1998; Henson et al, 1994; Ueki et al, 1996).

The objective of the present study was to assess RB1 promoter hypermethylation in primary and secondary glioblastomas as a potential underlying mechanism of loss of RB1 expression. To assess the timing of these alterations during astrocytoma progression, we also analyzed less malignant precursor lesions, ie, low-grade diffuse astrocytomas and anaplastic astrocytomas, from the same patients.

Results

RB1 Promoter Hypermethylation

RB1 promoter hypermethylation was detected in 14 of 56 (25%) glioblastomas (Tables 1 and 2). Promoter methylation was significantly more frequent in secondary than in primary glioblastomas (14% versus 43%, p = 0.0258, Tables 1 and 2).

Table 1 pRB Expression and RB1 Promoter Hypermethylation in Glioblastomas
Full size table
Table 2 Promoter Hypermethylation of the RB1 Gene in Astrocytic Brain Tumors
Full size table

RB1 methylation was not detected in any of 15 low-grade diffuse astrocytomas or 10 anaplastic astrocytomas (Table 2). Methylated and unmethylated control DNAs showed the expected fragment size of 163 bp (Fig. 1).

Figure 1
figure 1

Methylation-specific PCR of CpG islands of the RB1 promoter in low-grade diffuse astrocytomas (II) and glioblastomas (IV). In glioblastoma 233, only unmethylated DNA (U) was present. In glioblastoma 292, RB1 methylation (M) was restricted (asterisk) to areas lacking pRB immunoreactivity (Figure 2B). In Patient 58, the low-grade diffuse astrocytoma biopsy showed unmethylated status, whereas the glioblastoma derived from it contained a hypermethylated promoter region. Positive control for unmethylated and methylated DNA; normal control, DNA from a normal blood sample.

Full size image

pRB Immunohistochemistry

In 28 of 35 (80%) primary and 15 of 21 (71%) secondary glioblastomas, pRB immunoreactivity was detected, but the fraction of positive tumor cells varied considerably (Table 1). Six (17%) primary and four (19%) secondary glioblastomas showed loss of pRB expression throughout the section (Table 1).

Correlation between RB1 Methylation and pRB Immunoreactivity

Of 13 glioblastomas with loss of RB1 expression, 11 tumors (85%) showed promoter hypermethylation, whereas 40 of 43 (93%) cases with pRB immunoreactivity did not show RB1 methylation (Table 1, p < 0.0001).

In three glioblastomas (Table 1, Cases 292, 59, and 70), most tumor areas showed RB1 expression, but there were also focal areas with neoplastic cells that lacked RB1 expression (Fig. 2). In all three cases, promoter hypermethylation was detected only in areas lacking pRB immunoreactivity but not in the areas with RB1 expression (Fig. 1, Table 1).

Figure 2
figure 2

A, pRB immunohistochemistry showing nuclear immunoreactivity in the majority of glioblastoma cells (Case 70). B, Loss of RB1 expression in clusters of undifferentiated neoplastic cells in a glioblastoma (Case 70), whereas vascular endothelial cells show RB1 expression. Magnification, × 155.

Full size image

Discussion

The essential promoter region of RB1 lies 185–206 bp upstream of the initiation codon and contains putative binding sites for the transcription factors RBF-1, Sp1, ATF, and E2F (Ohtani-Fujita et al, 1993). It contains CpG islands that are frequent sites of methylation (Ohtani-Fujita et al, 1993; Stirzaker et al, 1997). Hypermethylation in the RB1 promoter region has been reported in 13% of unilateral retinoblastomas (Greger et al, 1994; Sakai et al, 1991) and 27% of pituitary adenomas (Simpson et al, 2000), suggesting that promoter hypermethylation is an alternative mechanism for mutational loss of RB1 expression.

In this study, one quarter of the glioblastomas showed promoter hypermethylation in the RB1 gene. RB1 methylation was detected more frequently in secondary than in primary glioblastomas. There was a clear correlation between loss of pRB expression detected by immunohistochemistry and promoter hypermethylation: the majority of glioblastomas with loss of RB1 expression had RB1 promoter hypermethylation (11 of 13, 85%), whereas the majority (93%) of tumors with RB1 expression had normal RB1 gene status. These results indicate that promoter hypermethylation is the major mechanism underlying the loss of RB1 function in glioblastomas. This was supported by the finding that in three glioblastomas, promoter hypermethylation was detected only in areas lacking pRB immunoreactivity but not in the areas where RB1 was expressed.

Loss of RB1 expression has been found to correlate well with LOH at the RB1 locus in hepatocellular carcinomas (Zhang et al, 1994), bladder carcinomas (Xu et al, 1993), and malignant neuroendocrine lung carcinomas (Gouyer et al, 1994). However, it did not correlate with LOH at the RB1 locus in other tumors, including glioblastomas (Burns et al, 1998; Ueki et al, 1996), pituitary tumors (Pei et al, 1995; Simpson et al, 1999), and carcinomas of the breast (Borg et al, 1992), prostate (Cooney et al, 1996), ovary (Dodson et al, 1994; Kim et al, 1994), and head and neck (Yoo et al, 1994). Simpson et al (2000) recently showed that the majority of pituitary adenomas with loss of RB1 expression without LOH on the RB1 locus had promoter hypermethylation. The present study confirms this mechanism and suggests that the lack of correlation between RB1 expression and LOH on the RB1 locus in glioblastomas frequently reflects promoter methylation as the underlying cause of loss of RB1 expression.

Hypermethylation has been considered one of the mechanisms of inactivation of the RB1 gene in the two-hit theory, because the majority of retinoblastomas with promoter hypermethylation showed LOH on the RB1 locus (Greger et al, 1994; Sakai et al, 1991). However, there is recent evidence that hypermethylation without LOH at this locus may be sufficient to cause loss of RB1 expression. A majority of pituitary adenomas with hypermethylation showed loss of RB1 expression without LOH at the RB1 locus (Simpson et al, 2000). We correlated RB1 expression with LOH on RB1 locus in 24 previously published glioblastomas (Nakamura et al, 2000b). Six of nine (67%) glioblastomas with hypermethylation showed loss of RB1 expression, and all of these retained heterozygosity on the RB1 locus (data not shown).

Loss of RB1 expression appears to be a prognostic factor in several human neoplasms, including glioblastomas. The mean survival of patients with glioblastomas showing RB1 expression was 11 months, whereas for those lacking RB1 expression it was 6 months (Nakamura et al, 1996). Similarly, the mean survival of patients with nonsmall cell lung carcinomas with loss or altered RB1 expression was 18 months and therefore significantly shorter than that of patients with RB1 expression (32 months) (Xu et al, 1994). Kornblau et al (1998) reported that the level of expression of pRB determined by immunohistochemistry was a strong prognostic factor in acute myelogenous leukemia, inferior survival rates being associated with no or low RB1 expression. In invasive transitional carcinomas of the bladder, loss of RB1 expression, together with p53 protein accumulation, was associated with significantly shorter survival (Cote et al, 1998). Non-small cell lung cancer with pRB−/p53+ also showed significantly shorter survival (5 year survival 20%) in comparison with pRB+/p53− tumors (5 year survival 73%) (Dosaka-Akita et al, 1997).

It has been reported that loss of RB1 expression is associated with a higher grade of malignancy in several human neoplasms. The majority of parathyroid carcinomas (88%) showed abnormal RB1 expression, whereas none of parathyroid adenomas analyzed had loss of RB1 expression (Cryns et al, 1994). A fraction of endometrial carcinomas lacked RB1 expression, whereas all hyperplastic lesions showed normal RB1 levels (Tsuda et al, 2000). In the present study, we observed RB1 promoter methylation only in glioblastomas and not in the less malignant precursor lesions, ie, low-grade diffuse and anaplastic astrocytoma, suggesting that loss of RB1 expression is a late event during astrocytoma progression.

Materials and Methods

Tumor Samples and DNA Extraction

The surgical specimens of brain tumors were obtained from patients treated between 1977 and 1994 in the Department of Neurosurgery, University Hospital of Zürich, Switzerland. Tumors were fixed in buffered formalin and embedded in paraffin. Pathological diagnosis was made according to the WHO classification (Kleihues and Cavenee, 2000). Thirty-five patients with primary glioblastoma had a preoperative clinical history of less than 3 months (mean, 1.5 months) and histologic diagnosis of a glioblastoma at the first biopsy, without any evidence of a less malignant precursor lesion. Twenty-one patients with secondary glioblastoma had at least 2 biopsies, with clinical and histologic evidence of progression from low-grade diffuse astrocytoma (16 cases, WHO Grade II) or anaplastic astrocytomas (5 cases, WHO Grade III). Low-grade diffuse astrocytomas (15 cases) and anaplastic astrocytomas (5 cases) from the same patients with secondary glioblastomas as well as five additional anaplastic astrocytomas were also examined. DNA was extracted from paraffin sections as described previously (Brüstle et al, 1992). In three glioblastomas (Cases 292, 59, and 70), pRB immuno-positive or negative tumor areas were clearly recognized. These areas were carefully microdissected and analyzed separately.

Methylation-Specific Polymerase Chain Reaction for RB1 Promoter Hypermethylation

DNA methylation patterns in the CpG islands of the RB1 gene were determined by methylation-specific polymerase chain reaction (MSP) (Herman et al, 1996). Sodium bisulfite modification was performed using the CpGenome DNA Modification Kit (Intergen, Oxford, United Kingdom) as described previously (Nakamura et al, 2001). Primer sequences of RB1 for the methylated and unmethylated reaction were as previously reported (Simpson et al, 2000). The PCR was carried out in a 10 μl volume containing PCR buffer (20 mm Tris pH 8.4, 50 mm KCl), 1.5 mm MgCl2, dNTPs (250 μm each), primers (4 pmol each), 0.5 unit of PLATINUM Taq DNA polymerase (GIBCO BRL, Cergy Pontoise, France), and approximately 40 ng bisulfite-modified DNA. Amplification was carried out in a DNA Thermal Cycler 480 (Perkin-Elmer Cetus, Norwalk, Connecticut) with initial denaturing at 95° C for 5 minutes followed by 35 cycles of denaturing at 95° C for 1 minute, annealing for 45 seconds at 65° C (for methylated RB1) or 61° C (for unmethylated RB1), extension for 1 minute at 72° C, and then a final extension for 5 minutes at 72° C. Amplified products were electrophoresed on a 3% agarose gel and were visualized with ethidium bromide.

pRB Immunohistochemistry

pRB immunohistochemistry was performed using the pRB monoclonal antibody (clone G3–245, which recognizes RB1 exons 9–12; PharMingen, San Diego, California) as previously described (Biernat et al, 1997) for Cases 233–344 and 295. The results of pRB immunohistochemistry in other cases were previously reported (Biernat et al, 1997). Fractions of positive cells were recorded as follows: positive in 5–25% tumor cells (+), positive in 25–50% tumor cells (++), or positive in >50% tumor cells (+++). pRB immunoreactivity in less than 5% of tumor cells or presence of clusters of tumor cells showing complete loss of RB1 expression was regarded as negative (Biernat et al, 1997).