Methylation of O⁶-Methylguanine DNA Methyltransferase and Loss of Heterozygosity on 19q and/or 17p Are Overlapping Features of Secondary Glioblastomas with Prolonged Survival

Patients and Methods

Patients. The 86 patients involved in the study [65 males and 21 females; mean age, 52.8 years (range, 28-67 years)] underwent open surgery for glioblastoma multiforme in our institution between 1995 and 2004. All patients signed an informed consent allowing the use of tumor specimens for biological studies. The majority of glioblastoma multiformes develop rapidly, with a short history and without clinical or histologic evidence of a less malignant precursor lesion (primary or de novo glioblastoma). Secondary glioblastomas develop more slowly by progression from diffuse astrocytoma (WHO grade 2) or anaplastic astrocytoma (WHO grade 3; ref. 12). In our series, 72 patients had a clinical history of ≤12 months (median, 3.0 months), compatible with the diagnosis of primary glioblastoma multiforme; 14 patients were diagnosed as secondary glioblastoma multiforme because of a preoperatory history longer than 12 months (median, 21 months; range, 12-34 months) and radiologic evidence (5 cases) or radiologic and histologic evidence (9 cases) of progression from a lower-grade tumor. Cases who underwent biopsy only were excluded from the study. Pathologic diagnosis was done by two neuropathologists at our institution in agreement with WHO guidelines (13). At recurrence, 28 patients underwent a second surgery: specimens from the last surgery were also considered in the study.

Postoperative irradiation was done according to the following protocol: focal external beam radiation therapy of 60 Gy (split into 180- to 200-cGy daily fractions) to the enhancing portion of the tumor and within a 2- to 3-cm margin. Chemotherapy [cis-diammine-dichloroplatinum 100 mg/m² + 1,3-bis(2-chloroethyl)-1-nitrosourea 160 mg/m² every 6 weeks] was started 3 weeks after surgery and carried on for five cycles or until tumor progression. At progression, patients were considered for reoperation and/or second-line chemotherapy [procarbazine/l-(2-chloroethyl)-3-cyclohexyl-l-nitrosourea/vincristine or temozolamide 150 mg/m²].

Radiologic assessment. All patients had a magnetic resonance imaging (MRI; n = 71) or computed tomography (n = 15) scan with and without contrast done before and 48 to 72 h after surgery to evaluate extent of surgery. T1-weighted, T1-weighted post-gadolinium, fast fluid-attenuated inversion recovery, and T2-weighted axial and coronal sequences were available for MRI analysis. According to previous reports (14–17), the following MRI features were independently reevaluated at the time of this study by two neuroradiologists who were blind to the histologic and molecular signature of the tumor: location; tumor margins (sharp versus undefined); signal intensity on T1 and T2 images (homogeneous versus heterogeneous); pattern of enhancement (ring, nodular, or mixed); presence of necrotic cysts; mass effect, edema. The score was defined by consensus.

MGMT methylation analysis. Methylation patterns in the CpG islands of MGMT were determined by chemical modification of unmethylated cytosines to uracils, followed by methylation-specific PCR, using primers specific for either methylated or modified, unmethylated DNA (18). Treatment of tumor DNA (1 μg) with sodium bisulfite was done with the CpG Genome DNA Modification Kit (Intergen), following the protocol of the manufacturer. Universally methylated DNA (Intergen) was used as a positive control and DNA from normal lymphocytes was used as a negative control.

Methylation-specific PCR was carried out with fluorescently labeled primers; the oligonucleotide primer sequences and the amplification protocol were previously described (18). Methylation-specific PCR products (1.5 μL) were loaded onto 8% polyacrylamide gels, examined on an automated DNA sequencer (Alf Express II, Amersham Biosciences), and quantitated using the Alf Win Fragment Analyser program (version 1.02, Amersham Biosciences). The peak height ratio of PCR products deriving from methylated or unmethylated DNA of the same tumor was calculated. Values >0.1 were scored as evidence of the methylated status of the MGMT gene (Meth+).

Loss of heterozygosity studies. The presence of allelic losses in tumor DNA was investigated by the PCR amplification of the following microsatellite loci: D17S1353 (17p13), D17S520 (17p12), D19S219 (19q13.3), and D19S412 (19q13.3). Primer sequences were downloaded from the GDB Human Genome Database web site.⁷ PCR products were analyzed on an Alf Express II sequencer and a quantitative analysis of signal intensity was carried out with the Fragment Analyser 1.02 program (Amersham Biosciences).

TP53 mutation analysis. The mutation analysis of the TP53 gene was done on 84 patients by denaturing high-performance liquid chromatography. Three pairs of oligonucleotide primers were used to amplify exons 5 to 9 of the TP53 gene. Primer pairs for the amplification were published before (19). PCR reactions were done in a final volume of 50 μL with 30 cycles consisting of a denaturation step at 95°C for 30 s, primer annealing at 58°C for 30 s, and elongation at 72°C for 30 s. Before denaturing high-performance liquid chromatography analysis, heteroduplex formation of the PCR products was carried out by heating for 10 min at 70°C followed by slow cooling to 8°C. Denaturing high-performance liquid chromatography analysis was done on a Wave DNA Fragment Analysis System (Transgenomic) as previously described (19).

The PCR products of any tumor sample showing an aberrant denaturing high-performance liquid chromatography elution profile were sequenced using ABI Prism 377 automated DNA sequencer. Sequence analysis was conducted with the same primers used in the original PCR. The PCR products of tumor samples that showed normal denaturing high-performance liquid chromatography but that at microsatellite analysis (done as previously described) were either not informative or showed loss of heterozygosity (LOH) for the microsatellite D17S1353 were directly sequenced.

Analysis of the amplification status of EGFR. The EGFR gene status was assessed in 84 patients by semiquantitative PCR. A DNA region of 110 bp, encoding the COOH-terminal portion of the oncogene and a fragment of the γ-IFN gene (85 bp, control DNA sequence), was simultaneously amplified by PCR using fluorescently Cy5-labeled primers. Differential PCR was done in a final volume of 20 μL, containing 200 ng of DNA. Initial denaturation at 95°C for 10 min was followed by 24 cycles of denaturation at 95°C for 30 s, annealing at 56°C for 30 s, and extension at 72°C for 30 s. A final extension step of 8 min at 72°C was done.

Fluorescent PCR products were separated on a 8% acrylamide gel and analyzed on an automated DNA sequencer (Alf Express II, Amersham Biosciences). A quantitative analysis of the signal intensity was carried out with the Alf Win Fragment Analyser program (Amersham Biosciences, version 1.02). The amplification status of the EGFR gene was defined by calculating the ratio of PCR products from the EGFR gene and the γ-IFN gene: ratios >2.09 were evidence of more than two copies of the EGFR gene, as previously described (20).

Statistical analysis. Frequency distributions and summary statistics were calculated for all clinical, radiologic, and tumor marker variables. For category variables, cross tabulations were generated and χ² or Fisher test was used to compare their distributions.

Progression-free survival was defined as the time from surgery to the first sign of radiologic progression according to McDonald's criteria (21). Overall survival was defined as the time between surgery and patient death. Survival distributions were estimated by Kaplan-Meier analysis and compared among patient subsets with log-rank tests. The following variables were investigated: age (≤50 or >50 years); glioblastoma multiforme subtypes (primary or secondary); Karnofsky score for performance status (≤80 versus >80); Radiation Therapy Oncology Group class risk (third versus fourth and fifth); extent of surgery (total versus subtotal); tumor location; MGMT methylation; LOHs on chromosomes 19q and 17p; TP53 mutations; and EGFR gene amplification. A multivariate analysis and a Cox proportional hazard regression model analysis were done on these variables to investigate their independent prognostic roles. The patients were classified according to the categories described in univariate analysis, except for age, which was used as a continuous variable.

The statistical analysis was done using software StatView 5.1.

Results

MGMT methylation status, clinical characteristics, and MRI. Methylation of the MGMT promoter was detected in 41 of 86 tumors (47.7%; Meth+). The characteristics of the 86 patients analyzed are shown in Table 1 . MGMT promoter methylation was found more frequently in secondary than in primary glioblastoma multiformes [11 of 14 (79%) versus 30 of 72 (42%); P = 0.01], but was not associated with age, gender, Radiation Therapy Oncology Group class risk, and extent of surgery; a trend to higher Karnofsky performance score was observed in Meth− patients. Overall, the two subgroups of Meth+ and Meth− were well balanced with respect to their baseline clinical characteristics. Patients with secondary glioblastoma multiforme, however, were significantly younger than patients with de novo glioblastoma multiforme (P = 0.005).

In 28 patients that underwent surgery twice because of progressive disease, we analyzed both the first and the second tumor DNA. At first surgery, 14 tumors were Meth+ and 14 were Meth−. At second surgery, 7 of 14 Meth+ glioblastoma multiformes became Meth− whereas all Meth− glioblastoma multiformes maintained their status.

MRI features were evaluated in 71 patients. Correlations between MRI findings, glioblastoma multiforme subtype, and genetics are shown in Table 2 and Fig. 1 . In two cases, presurgical MRI showed no enhancement. Ring enhancement was significantly associated with diagnosis of de novo glioblastoma multiforme (P = 0.006) and with tumors with unmethylated MGMT promoter (P = 0.01; Table 2). This observation was confirmed by further examination of 12 independent tumors (10 primary and 2 secondary glioblastoma multiformes) by three neuroradiologists unaware of the methylation status of the tumors (M.G.B., T.D.S., and L.D.I.). Agreement was present in 11 of 12 cases. Ring enhancement was only present in Meth− tumors (3 of 8) and absent in Meth+ tumors (0 of 4); mixed-nodular enhancement was present in 5 of 8 Meth− and 4 of 4 Meth+ glioblastoma multiformes. No significant correlation was found between the remaining MRI features (T2 signal, mass effect, edema, margins), diagnosis, and methylation status.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

MRI of Meth+ and Meth− glioblastoma. A, Meth− de novo glioblastoma multiforme. MRI axial images (left, T2; right, T1 with contrast). The lesion is located in the left temporal lobe and shows moderate mass effect. The T2 hyperintensity is not homogeneous with a large, more hyperintense round cystic image confirmed by contrast injection (right) that shows also the thin rim of contrast enhancement. B, Meth+ de novo glioblastoma multiforme. MRI axial images (left, T2; right, T1 with contrast). The lesion is located in the left parietal lobe and shows mass effect and some degree of vasogenic edema. The T2 hyperintensity is not homogeneous due to the hyperintense round cystic image. T1 with contrast injection image (right) shows the cystic appearance with a thick surrounding solid portion. C, Meth+ secondary glioblastoma multiforme. MRI axial images (left, T2; right, T1 with contrast). The lesion is located in the left frontal and temporal lobe and shows mass effect and some degree of vasogenic edema. The T2 hyperintensity is slightly inhomogeneous due to a concurrence of a medial portion with homogeneous T2 hyperintensity with no contrast enhancement and a lateral portion showing a mixed, mostly solid portion showing contrast enhancement, clearly recognizable on the right image. D, Meth− secondary glioblastoma multiforme. MRI axial images (left, T2; right, T1 with contrast). The lesion is located in the left frontal lobe with extension mainly to the basal ganglia and insular regions and shows mass effect and vasogenic edema. The T2 hyperintensity is mainly homogeneous and the pattern of contrast enhancement, clearly recognizable in the right picture, is more similar to that in (C) than in (A).

When we considered tumor location, we found that 36 glioblastoma multiformes were located in the frontal lobes (28 in one lobe only and 8 in one frontal lobe plus another lobe), 35 in the temporal lobes (28 in one lobe only and 7 in one temporal lobe plus another lobe), 11 in the parietal lobes (6 in one lobe only and 5 in the parietal and the occipital lobes), and 4 in the occipital lobe. Notably, tumors with methylation of the MGMT promoter were more often located in parietal and occipital lobes, whereas tumors without methylation of the MGMT promoter were frequently in the temporal lobes (P = 0.005).

MGMT methylation status and survival. In all patients, the median progression-free survival was 8 months and the median overall survival was 15 months. In univariate analysis, methylation of the MGMT promoter was positively correlated with prolonged progression-free survival (10 versus 7 months; P = 0.006, log-rank test) and longer overall survival (18 versus 14 months; P = 0.001, log-rank test; Table 3 ; Fig. 2 ). Secondary glioblastoma multiforme also had longer progression-free survival and overall survival than de novo tumors [progression-free survival, 9 versus 8 months (P = 0.057, log-rank test); overall survival, 24 versus 15 months (P = 0.009, log-rank test); Table 3; Fig. 2]. LOH on chromosome 19q (see also below) was positively correlated with longer overall survival (16 versus 14 months; P = 0.02, log-rank test). Other clinical or genetic features were not statistically associated with overall survival or progression-free survival.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Kaplan-Meier analysis of survival of 86 patients with glioblastoma multiforme. A, in univariate analysis, methylation of the MGMT promoter was positively correlated with longer overall survival (18 versus 14 mo; P = 0.001, log-rank test). B, overall survival of primary or secondary glioblastoma multiforme stratified for the MGMT status. C, overall survival of glioblastoma multiforme with or without methylation of the MGMT gene (M+ and M−, respectively) stratified for age (<50 or ≥50 y).

To further analyze the influence of the methylation status of the MGMT promoter, we did a multivariate analysis with the use of the Cox proportional hazard model. The methylation status of MGMT was the only variable independently affecting overall survival (Table 3).

MGMT methylation status and genetic variables. LOHs on chromosomes 17p and 19q were detected in 24 (29%) and 34 (41.%) of 83 tumors, respectively (Table 4 ); overall, 50 (59%) glioblastoma multiforme patients showed LOH on 17p and/or LOH on 19q. MGMT promoter methylation was more frequent in glioblastoma multiforme with LOH on 19q (P = 0.02) and with LOH on 17p and/or 19q (75% versus 45%; P = 0.005). TP53 mutations were detected in 19 of 84 patients (23%; 2 Meth+ cases could not be studied because DNA was insufficient).

As expected, a strong association was found between LOH on 17p and TP53 mutations (P = 0.001). TP53 mutations were more frequent in Meth+ than in Meth− glioblastoma multiforme (28% versus 19%) and the majority of mutations in Meth+ glioblastoma multiforme were G:C > A:T transitions [5 of 11 (45%); Table 4], confirming previous observations by other authors (22–25). In particular, MGMT methylation, which appears as an early event in glioma tumorigenesis leading to secondary glioblastoma multiforme, could imply cytosine mutation in TP53 following deamination or mutation of the paired guanine due to O⁶-alkylguanine mispairing with T (22).

EGFR amplification was observed in 29 of 84 (34%) glioblastoma multiformes, was a rare event in tumors with LOH on 17p (4% in tumors with LOH 17p versus 46% in tumors without LOH 17p; P = 0.0002), and occurred in one patient only with a TP53 mutation. EGFR amplification was more frequent in Meth− than in Meth+ tumors (40% versus 28%, respectively), but the difference did not reach statistical significance. TP53 mutations and EGFR amplification were not correlated with overall survival or progression-free survival even if patients were stratified according to the methylation status of MGMT.

Seven of 13 evaluable secondary glioblastoma multiformes showed LOH on 17p and five showed LOH on 19q; overall, LOH on 17p and/or LOH 19q occurred in 11 of 13 secondary glioblastoma multiformes and was more frequent than in de novo glioblastoma multiforme (85% versus 55%; P = 0.06). TP53 mutations were more frequent in secondary than in de novo glioblastoma multiforme (31% versus 21%), whereas EGFR amplification was less frequent (18% versus 39%).

Discussion

The main findings of this study are that methylation of the DNA repair gene MGMT is highly represented in glioblastoma multiforme resulting from malignant evolution of lower-grade gliomas and tightly associated with LOH on chromosomes 17p and/or 19q: 79% (11 of 14) of secondary glioblastoma multiformes are Meth+ and 85% of Meth+ glioblastoma multiformes show 17p and/or 19q losses. Confirming the majority of previous studies (8, 9), we find that patients with Meth+ glioblastoma multiforme treated by chemotherapy (temozolomide or cis-diammine-dichloroplatinum plus 1,3-bis(2-chloroethyl)-1-nitrosourea) survive significantly longer than others. These findings suggest that MGMT methylation is part of the genetic and epigenetic signature of secondary glioblastoma multiforme.

A more analytic observation of the findings contained in our study as well as in studies from other groups gives further support to this suggestion. First, whereas genetic losses on chromosomes 10q are similarly spread in primary and secondary glioblastoma multiforme (26), losses on 9p are more frequent in primary glioblastoma multiforme (27) and losses on 19q and 17p, albeit not exclusive, are more frequent in secondary glioblastoma multiforme (28, 29). Second, data from other groups show that MGMT methylation is frequent in low-grade gliomas that evolve more rapidly to secondary glioblastoma multiforme (30) and confirm the high frequency of Meth+ tumors in secondary glioblastoma multiforme (22, 24). Third, MRI features, including tumor location and the characteristics of enhancement, contribute to the identification of overlapping features of Meth+ and secondary glioblastoma multiforme. Both de novo and secondary glioblastoma multiformes show undefined margins, T1 and T2 inhomogeneous signals, and important mass effect. The patterns of enhancement could allow a preoperative differentiation between primary, Meth− glioblastoma multiformes, which are large lesions with ring enhancement and large necrotic cysts, most frequently located in the temporal lobe, and secondary glioblastoma multiformes, which are more homogeneously enhancing lesions mainly located in the frontal lobes (Fig. 1). Primary Meth+ glioblastoma multiformes show intermediate characteristics (Fig. 1). Overall, these results support the idea that the Meth+ status is part of the biological signature that characterizes secondary glioblastoma multiforme.

Recent literature is increasingly contributing to the definition of biological features of secondary glioblastoma multiforme: these tumors have specific patterns of gene and protein expression (31, 32), express significantly less survivin that primary tumors (33), and have higher levels of telomerase activity (28) and different angiogenic phenotypes (34). At the clinical level, secondary glioblastoma multiforme is associated with longer survival than primary glioblastoma multiforme (35). Thus, an interesting consequence of our findings could be that part of the survival advantage associated with Meth+ glioblastoma multiforme is not related to chemosensitivity but, rather, is inherent to their tumor biology. Interestingly, results of comparison of the MGMT methylation status of glioblastoma multiforme at first and second surgery show that 7 of 14 Meth+ tumors become Meth− but no Meth− glioblastoma multiforme becomes Meth+, a likely consequence of the selective pressure that chemotherapy is exercising on these cell subpopulations. On the other hand, 7 of 14 Meth+ tumors maintain their status. Further studies will be necessary to investigate the reasons underlying the different evolutions of MGMT methylation status in glioblastoma multiforme.

In our study, we have stretched the criteria for identification of secondary glioblastoma multiforme by including cases with a clinical history longer than 12 months (median, 21 months) and a radiologic evidence of lower-grade glioma. This inclusion criteria was corroborated by the observation that cases with MRI evidence only of secondary glioblastoma multiforme had similar clinical and biological features to cases with MRI and histologic evidence (data not shown).

Our MRI findings indicate that MRI could play a prognostic role in glioblastoma multiforme evaluation and management, suggesting a diagnosis of secondary glioblastoma multiforme early before surgery and genetic analysis. Molecular alterations associated with cancer may confer on the tumor physical or biochemical characteristics that can be imaged; contrast enhancement, reflecting blood-brain-barrier breakdown, is a key variable (36). Many reports have correlated the genetic characterization of gliomas to histology and survival, providing the basis for a reclassification of these tumors (37, 38). Other reports have specifically associated molecular markers of gliomas and MRI features (14–17, 39, 40). Aghi et al., in particular, reported that glioblastoma multiforme overexpressing EGFR has increased T2/T1 ratio and decreased T2 sharpness of the borders. In our series, MRI scores were based on subjective and not computer-based analysis. Ring enhancement was more frequent in EGFR-amplified glioblastoma multiforme, but this did not reach statistical significance. More importantly, we found a significant difference between the MRI appearance of primary and secondary glioblastoma multiformes: the former were more heterogeneous both in T1 and T2 sequences, with rim and mixed contrast enhancement, whereas the latter were more homogeneous both in T2 hypersignal and contrast enhancement (Fig. 1). Borders of the tumor were evaluated also on T2 images and extent of edema was scored, but these were not considered because of the concurrence of steroid therapy.

In our study, 27 glioblastoma multiformes arising de novo are Meth+. It is possible that in part of these 27 cases, a lower-grade glioma went undetected for a few months because of the site of the lesion; in support of this possibility is the observation that Meth+ primary glioblastoma multiforme are located in the right hemisphere more frequently than the others (P < 0.01). Interestingly, 20 of 27 Meth+ primary glioblastoma multiformes and 16 of 41 Meth− primary glioblastoma multiformes had LOH on 17p and/or 19q (χ² P = 0.0097). Thus, even when the clinical presentation implies the presence of a primary glioblastoma multiforme, (epi)genetic analysis and MRI may suggest that the tumor is or will behave as a secondary glioblastoma multiforme. Although therapeutic implications at the moment are rather limited, the improved definition of a subgroup of glioblastoma multiforme may become more relevant in the presence of new molecules, targeting specific pathways of secondary or primary glioblastoma multiforme. A recent study based on DNA microarray, for instance, suggests that genes associated with secondary glioblastoma multiforme include mitotic cell cycle components, whereas those associated with primary glioblastoma multiforme highlight genes typical of a stromal response (31).

Further insights on the potential cell of origin for primary and secondary glioblastoma multiforme may become available in the near future. A series of fascinating observations suggest that glioblastoma multiformes, at least in part, derive from genetic alterations taking place in neural stem cells. Data reviewed by Zhu and Parada (41) were used to propose that this is only for the case primary glioblastoma multiforme. If this will be confirmed, a reasonable speculation would imply that secondary glioblastoma multiforme may rather originate from neural, more committed progenitors or from dedifferentiation of mature glial cells.