Comparative assessment of 5 methods (methylation-specific polymerase chain reaction, methylight, pyrosequencing, methylation-sensitive high-resolution melting, and immunohistochemistry) to analyze O6-methylguanine-DNA-methyltranferase in a series of 100 glioblastoma patients
There is a strong need to determine the best technique for O6-methylguanine-DNA-methyltranferase (MGMT) analysis, because MGMT status is currently used in clinical trials and occasionally in routine clinical practice for glioblastoma patients.
The authors compared analytical performances and predictive values of 5 techniques in a series of 100 glioblastoma patients who received standard of care treatment (Stupp protocol).
MGMT promoter was considered methylated in 33%, 33%, 42%, and 60% of patients by methylation-sensitive high-resolution melting, MethyLight, pyrosequencing (with an optimal risk cutoff at 8% for the average percentage of the 5 CpGs tested), and methylation-specific polymerase chain reaction (MS-PCR), respectively. Fifty-nine percent of the samples had <23% (the optimal risk cutoff) of MGMT-positive tumor cells. The best predictive values for overall survival (OS), after adjustment for age and performance status, were obtained by pyrosequencing (hazard ratio [HR], 0.32; P < .0001), MS-PCR (HR, 0.37; P < .0001), and immunohistochemistry (HR, 0.43; P = .0005) as compared with methylation-sensitive high-resolution melting (HR, 0.52 P = .02) and MethyLight (HR, 0.6; P = .05). For progression-free survival (PFS), the best predictive values were obtained with pyrosequencing (HR, 0.35; P < .0001), methylation-sensitive high-resolution melting (HR, 0.46; P = .002), and MS-PCR (HR, 0.49; P = .002). Combining pyrosequencing and immunohistochemistry slightly improved predictive power for OS, but not for PFS. Poor reproducibility and interobserver variability were, however, observed for immunohistochemistry.
Glioblastoma multiforme (GBM) is the most common primary brain tumor in adults. The standard of care for GBM patients currently involves surgical resection, then temozolomide chemotherapy with concomitant radiotherapy, followed by cycles of adjuvant temozolomide.1 O6-Methylguanine-DNA-methyltranferase (MGMT) status is recognized as a very powerful predictor of response to temozolomide for newly diagnosed GBM patients.2
MGMT status can be analyzed at the DNA, RNA, or protein level.3MGMT expression is mainly regulated at the epigenetic level through CpG island promoter methylation.4, 5 Promoter methylation tests are thus used for MGMT assessment at the DNA level. Most studies reporting a link between MGMT promoter methylation and survival in GBM patients have used methylation-specific polymerase chain reaction (MS-PCR).6, 7-10 One of the major problems of this type of technique is the subjectivity linked to eye reading of the gel and the lack of automation. This led to the development of alternative techniques including real-time quantitative MS-PCR, MethyLight, pyrosequencing, methylation-sensitive high-resolution melting, COBRA (COmbined Bisulfite Restriction Analysis), and multiplex ligation-dependent probe amplification (MLPA). All of these methods except MLPA require a first step of bisulfite conversion. Pyrosequencing is a sequence-by-synthesis method that is the only method analyzing methylation levels at each CpG separately (for review and comparison of techniques, see Weller et al2 and Preusser3). Currently, among these alternative techniques, only pyrosequencing11, 12 and MethyLight13 have been shown to have predictive value for GBM patients. Furthermore, a real-time quantitative MS-PCR approach, proposed by MDxHealth (formerly Oncomethylome Sciences), is being used in several international clinical trials. This assay determines the number of copies of methylated MGMT, which is then normalized to the number of copies of the ACTB gene.14
MGMT assessment at the protein level is usually done by immunohistochemistry (IHC), a low-cost method commonly used in diagnostic histopathology. Some studies of high-grade glioma patients treated with alkylating agents, including GBM patients, have reported significant association of MGMT expression by IHC with patient outcome.15-18 However, in other recent studies of cohorts of GBM patients treated by radiotherapy plus temozolomide, no correlation between MGMT expression and patient survival was found.7, 11, 19
In this study, we compare the analytical performances and the predictive values of 5 techniques (MS-PCR, methylation-sensitive high-resolution melting, pyrosequencing, MethyLight, and IHC) for MGMT analysis in a series of 100 newly diagnosed GBM patients treated with temozolomide.
MATERIALS AND METHODS
Patients with newly diagnosed primary GBM given standard care treatment and followed-up for at least 18 months were selected in 4 French centers (Marseille, Paris, Poitiers, and Rennes). For each patient, a frozen tumor sample and paraffin-embedded tissue specimens had to be available. The protocol was approved by the ethics committee of Rennes, and informed consent was obtained from patients.
DNA Extraction and Bisulfite Treatment
Tumor samples obtained during surgery were stored at −80°C. For each sample, at least 1 slice was stained with hematoxylin and eosin to control the percentage of tumor cells. Only samples containing at least 60% tumor cells were retained, and DNA extraction was performed at each center according to local procedures. DNAs were then centralized in Rennes and anonymized, and 2.5 μg (5 × 500 ng) was treated with sodium bisulfite using the EZ DNA Methylation Gold kit according to the specified protocol (Zymo Research, Orange, Calif). For each sample, modified DNA was eluted in 5 × 10 μL, pooled, and stored at −80°C in aliquots before being sent and/or analyzed. The 100 DNA samples were tested in 10 different series of 10 samples (a series corresponds to samples that were bisulfite modified and then tested at the same time). One sample was excluded because of an insufficient quantity of DNA. In each series, anonymized DNA extracted from 3 primary cell lines obtained in the laboratory in Rennes (GB2, GB3, and GBM2) were used as quality controls to assess reproducibility. DNA was also extracted from peripheral blood mononuclear cells (nonmethylated control), and a universal methylated DNA (Millipore, Billerica, Mass) was used as fully methylated control.
MGMT Promoter Methylation Analysis
MethyLight was performed in Marseille, methylation-sensitive high-resolution melting in Paris, MS-PCR in Poitiers, and pyrosequencing in Rennes. As described above, the same batch of bisulfite-modified DNA samples were used for each technique, to avoid any bias that could be induced by the sodium bisulfite treatment. Each series was tested separately. Figure 1 describes the CpG island region analyzed and the different CpGs interrogated by the assays. Numbers are given to the CpGs from 5′ to 3′ on the coding strand of the gene, starting from CpG1 located −452 bp before the transcription initiation site.20
MethyLight was performed as described previously.13 All primers and probes have been published elsewhere.21 The differences in amounts of input genomic DNA were normalized by the collagen type II, alpha 1 gene (COL2A1). The percentage of methylated reference was calculated as follows: the methylated MGMT/COL2A1 ratio for each sample was divided by the same ratio obtained for a SssI-treated genomic DNA used as standard, and values were multiplied by 100.
Methylation-Sensitive High-Resolution Melting
PCR amplification and high-resolution melting analysis were performed using a Mx3000P apparatus (Stratagene, La Jolla, Calif). PCR was carried out in a 25-μL total volume containing: 2 × Brilliant EvaGreen QPCR Master Mix (Biotium, Hayword, Calif), 300 nM of each primer (forward: 5′ GCGTTTCGGATATGTTGGGATAGT 3′ and reverse: 5′ AACGACCCAAACACTCACCAAA 3′),22 and 1 μL of bisulfite-modified template. The amplification consisted of 10 minutes at 95°C, followed by 50 cycles of 15 seconds at 95°C, 30 seconds at 60°C, and 30 seconds at 72°C. After amplification, a postamplification melting curve program was initiated by heating to 95°C for 1 minute, cooling to 70°C for 30 seconds, and increasing the temperature to 95°C (heating rate 0.01°C/s) while continuously measuring fluorescence. Control DNAs were extracted from blood. The methylated control was obtained by treatment with CpG Methylase M.sssI M0226S (New England Biolabs, Ipswich, Mass). Sample melting curves were compared with control melting curves obtained with unmethylated and methylated controls. When the peak corresponding to methylated DNA was >50% of the peak corresponding to unmethylated DNA, the sample was considered methylated. All reactions were performed in duplicate.
Methylation status of the CpG islands of MGMT promoter was obtained using a 2-stage PCR as described before.11 Primers for the second PCR were those commonly used to detect either methylated or unmethylated modified DNA.6, 23
Pyrosequencing was performed with the PyroMark Q96 MGMT kit (Qiagen, Courtaboeuf, France) on a PSQTM96 MA system (Biotage, Uppsala, Sweden), as described previously.11 Templates for pyrosequencing were obtained by amplifying bisulfite-treated DNA with primers that are biotinylated for template strands. The biotinylated PCR products were then immobilized on streptavidin-coated Sepharose beads (GE Healthcare, Orsay, France), and the single-stranded DNA templates were analyzed in the pyrosequencing instrument. For each CpG island tested, a mean result was calculated. For data analysis, the average percentage of the 5 CpGs was determined (PYRmean) as well as the results of each tested CpG (PYRCpGx).
Immunohistochemistry for MGMT Protein Analysis
Unstained slides from the 4 centers were centralized and anonymized in Rennes before being stained in Marseille in 6 different series. For 1 patient, paraffin-embedded tissue was not available. In each series, anonymized slides issued from 2 selected blocks previously found with low and high percentages of MGMT-expressing cells were included as controls. The staining was performed as previously described.18 The percentage of positive tumor cells was determined by Dominique Figarella-Branger as follows. The percentage of the tumor immunoreactive nuclei was counted at high magnification. On average, 400 nuclei were counted in each tumor. Slides were examined as carefully as possible to take into account and exclude the immunoreactive nuclei of inflammatory cells included in the tumor. A second pathologist (Stephan Saikali) from a different neuropathology laboratory analyzed the same slides a second time using the same criteria. One sample was excluded because of a technical problem during the staining process.
To analyze the intra-assay precision of molecular assays, GB2, GB3, and GBM2 cell lines were tested 5× in the same series. To analyze interassay precision, the same cell lines were tested in 10 different series. Limits of quantification (defined as the lowest amount in a sample that can be measured with a specified precision) and linearity for quantitative techniques were determined by serial dilutions of the fully methylated control with the unmethylated control. Thirteen samples with theoretical methylation percentages of 100%, 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, 10%, 5%, 2.5%, and 0% were tested 5× in a single run.
For IHC, slides from the 2 selected control blocks were tested in 6 different series.
The statistical analysis was performed with R statistical software (version 2.10.1, http://www.Rproject.org). For continuous and, respectively, categorical variables, the median (interquartile range) and number of patients in each category are reported. The Bland-Altman method was used to compare the results from the 2 pathologists who analyzed the same IHC slides. Range agreement was defined as mean bias ± 2 standard deviations (SD). Approaches based on the structure of the data to define a cutoff (such as the median-based method) have been shown to have a poor performance and do not allow an easy comparison of different studies.24 Optimal risk cutoffs were therefore determined as the threshold values of the continuous distribution that best separated low- and high-risk patients according to their outcomes (outcome-based method). More precisely, they were defined as the thresholds that optimized the area under the receiver operating characteristic (ROC) curve obtained with a Cox model25 using overall survival (OS) adjusted for age and Karnofsky score (the proportional hazard assumption was checked). The function riskset AUC (package risksetROC) in R statistical software was used to obtain the area under the ROC curve.
To study OS and progression-free survival (PFS), cumulative event curves (censored endpoints) were established using the Kaplan-Meier method. Univariate analyses were performed using Cox regression models adjusted for age and Karnofsky score. Multivariate Cox models including no, 1, or 2 biological tests to assess MGMT status, adjusted for age and Karnovsky score, were built. Two criteria were used to choose the best models: the area under the ROC curve and the Harrell C index26 obtained with the function validate (package Design).
The study population included 100 adult patients with newly diagnosed primary GBM, excluding giant cell GBM. The median follow-up was 17.9 months. A total of 75 patients had died when the database was closed. Clinical patient characteristics are summarized in Table 1. The median PFS was 10.5 months (95% confidence interval [CI], 9.0-12.6), and median OS was 17.9 months (95% CI, 15.8-20.4).
Three cell lines, GB2, GB3, and GBM2, were used to analyze the intra-assay and interassay precision of the molecular techniques (Fig. 2A, B). In most cases, GB2 was detected unmethylated by MS-PCR, methylation-sensitive high-resolution melting, and MethyLight. PYRmeans were 7% (coefficient of variation [CV], 35%) and 5% (CV, 11%) for intra-assay and interassay precision, respectively. In most cases, GB3 was detected unmethylated by methylation-sensitive high-resolution melting and MethyLight. It was detected methylated in almost every case with MS-PCR. PYRmeans were 10% (CV, 6%) and 9% (CV, 6%) for intra-assay and interassay precision, respectively. GBM2 was detected methylated in all cases with MS-PCR and methylation-sensitive high-resolution melting. Mean percentages of methylated reference were 28 (CV, 36%) and 42 (CV, 72%) for MethyLight, and PYRmeans were 82% (CV, 1%) and 77% (CV, 7%) for intra-assay and interassay precision, respectively.
For IHC, the series of slides labeled IHC01 were analyzed in 6 different series. Mean percentage of tumor cells positive for MGMT was 31% (CV, 63%). Slides in the series IHC02 were analyzed in 5 different series, because a technical problem occurred during 1 series. Mean percentage of tumor cells positive for MGMT was 8% (CV, 87%; Fig. 2C).
A fully methylated control was diluted to obtain 13 samples with theoretical percentages of methylation from 0 to 100% (Fig. 3A). The sample with the lowest proportion of methylated DNA tested (2.5%) was found methylated by MS-PCR and methylation-sensitive high-resolution melting. This 2.5% methylation mixture was measured as 4% (CV, 14%) and 5% (CV, 13%) with pyrosequencing and MethyLight, respectively. Thus, all 4 molecular tests could detect relatively low levels of methylation. As the limit of quantification is usually defined as the lowest value with a CV <20%, it was set at 4% for pyrosequencing. A good linearity was also observed for pyrosequencing and MethyLight techniques (Fig. 3B).
MGMT Analysis in the Cohort of Patients
MGMT promoter was found methylated in the studied population in 60% and 33% of cases by MS-PCR and methylation-sensitive high-resolution melting, respectively (Table 2). For MethyLight, the percentage of methylated reference values ranged from 0 to 231.36. When searching for the optimal clinical cutoff, the value of 0 was obtained; therefore, patients were dichotomized as methylated or unmethylated, with 33% of cases methylated. The cutoffs determined for pyrosequencing and IHC by our outcome-based method were as follows: 4% (PYRCpG1), 11% (PYRCpG2), 4% (PYRCpG3), 6% (PYRCpG4), 5% (PYRCpG5), 8% (PYRmean) and 23% (IHC). The percentages of samples with results above the cutoffs were 46%, 38%, 48%, 44%, 41%, and 42% for PYRCpG1, PYRCpG2, PYRCpG3, PYRCpG4, PYRCpG5, and PYRmean, respectively. Fifty-nine percent of the samples had ≤23% MGMT-positive tumor cells.
Table 2. MGMT Status According to the Different Techniques and Their Significance Concerning OS and PFS in Univariate Analysis
The level of significance represented by P values is derived from 2-sided tests (P < .05 was considered to indicate statistical significance).
8.1 × 10−5
2.4 × 10−3
1.7 × 10−2
1.7 × 10−3
4.8 × 10−2
2.2 × 10−2
2.9 × 10−4
1.5 × 10−5
2.4 × 10−5
1.2 × 10−4
1.2 × 10−4
5.5 × 10−5
1.4 × 10−5
4.6 × 10−5
3.4 × 10−4
2.0 × 10−4
1.7 × 10−5
1.2 × 10−5
5.1 × 10−4
3.6 × 10−2
Positive cells ≤23%
Positive cells >23%
Relation Between MGMT Analysis by the Different Techniques and Clinical Outcome
For the univariate analysis, the variables were dichotomized as methylated and unmethylated or according to their optimized cutoffs. MGMT promoter methylation assessed by either MS-PCR, methylation-sensitive high-resolution melting, MethyLight, or pyrosequencing was significantly associated with a better OS (Table 2).
Patients carrying tumors found methylated by MS-PCR had a median OS of 22.4 months versus 14.9 months for patients carrying tumors found unmethylated (P = 8.1 × 10−5). For methylation-sensitive high-resolution melting and MethyLight, the OS for the 2 groups was 20.4 and 16.7 months (P = 1.7 × 10−2 and 4.8 × 10−2, respectively). Patients with a pyrosequencing-assessed mean percentage of methylation >8 had a median OS of 26.2 months, whereas patients with a mean percentage of methylation of ≤8 had a median OS of 15.7 months (P = 1.7 × 10−5). When each CpG was considered separately, longer OSs were observed for patients with a level of methylation above the selected cutoff, whatever the CpG position (P = 2.9 × 10−4 for PYRCpG1, P = 2.4 × 10−5 for PYRCpG2, P = 1.2 × 10−4 for PYRCpG3, P = 1.4 × 10−5 for PYRCpG4, and P = 3.4 × 10−4 for PYRCpG5). We also found that patients with ≤23% MGMT-positive tumor cells as assessed by IHC had significantly longer OS (22.4 months vs 14.6 months, P = 5.1 × 10−4).
Rather similar correlations were observed between MGMT status and PFS. Patients with methylated tumors assessed by MS-PCR (P = 2.4 × 10−3), methylation-sensitive high-resolution melting (P = 1.7 × 10−3), or MethyLight (P = 2.2 × 10−2), or yielding a level of methylation above the selected cutoff (P = 1.5 × 10−5 PYRCpG1; P = 1.2 × 10−4 PYRCpG2; P = 5.5 × 10−5 PYRCpG3; P = 4.6 × 10−5 PYRCpG4; P = 2.0 × 10−4 PYRCpG5; P = 1.2 × 10−5 PYRmean) had statistically longer PFS. This was also true for patients with MGMT protein expression below the selected cutoff (P = 3.6 × 10−2).
Influence of the Observer on IHC
As observer variability had been described as a major drawback of anti-MGMT IHC,19 stained slides were evaluated by a second pathologist from another neuropathology laboratory. Median percentage of positive cells was 12% for the series of slides interpreted by the second pathologist, and 74% of patients had tumors with ≤23% positive cells. If using a cutoff of 23%, 26 patients were classified differently by the 2 pathologists. Despite this interobserver variability, shown by the Bland-Altman plot (Fig. 4A), a significant association between MGMT expression and OS (P = 2.0 × 10−4) or PFS (P = 5.7 × 10−4) was still observed when the slides were evaluated by the second pathologist (Fig. 4B).
Multivariate Cox models were applied to determine whether the predictive value of MGMT status could be improved by the use of >1 method of evaluation (data not shown). Prediction errors were globally evaluated and reported as the area under the ROC curve and the Harrell C index. The best area under the ROC curve value of 0.68 for OS was obtained with a model including PYRCpG4, PYRCpG2, or PYRmean and IHC. The improvement when adding IHC was nonetheless weak. PYRCpG1, PYRCpG4, or PYRmean gave the best area under the ROC curve values for PFS, with no improvement when adding IHC in these 3 models.
MGMT status is currently recognized as a strong predictive factor for newly diagnosed GBM patients given standard care treatment, despite the lack of consensus concerning the best technique for its assessment. In a previous study we had compared MS-PCR, a semiquantitative SYBR Green-based MS-PCR, pyrosequencing, IHC, and MGMT expression at the RNA level in a cohort of GBM patients. The best predictive value was obtained by pyrosequencing.11 The aim of this study was to compare pyrosequencing to other promising techniques (MethyLight and methylation-sensitive high-resolution melting), to refine the cutoff in an independent cohort of patients (only 14 patients are common to both cohorts) and to carefully study the analytical performance of these techniques, which has not been done before.
Our study confirms the prognostic value of MGMT promoter methylation assessed by MS-PCR. Sixty percent of patients were classified as methylated in our series of 99 GBM patients, which is in the range reported in the literature with this technique (34%-68%, with a mean of 46%).2 We also confirm that the percentages of methylation estimated by pyrosequencing are highly correlated with outcome. One challenge with quantitative techniques is the definition of cutoff points. In a previous study, we chose the medians as cutoffs: 5%, 9%, 8%, 8%, 8%, and 8% for CpG1, CpG2, CpG3, CpG4, CpG5, and CpGmean, respectively.11 In this series, an outcome-based technique gave similar results: 4%, 11%, 4%, 6%, 5%, and 8% for PYRCpG1, PYRCpG2, PYRCpG3, PYRCpG4, PYRCpG5, and PYRmean, respectively. For CpG1 and CpG3, the cutoff was set at 4%, which is also the limit of quantification of the method. There is therefore a risk of misclassification for patients displaying a low percentage of methylation using these single CpGs. Dunn et al performed a pyrosequencing assay for 12 CpG sites, including the 5 CpGs of our assay, and chose as cutoff the mean average methylation + 2 SD obtained for 6 non-neoplastic brain samples (9%). Fifty-three percent of patients were considered methylated and had a median OS of 16.8 months.12 Studying each CpG separately is of interest, as in some tumors methylation is very heterogeneous (Fig 5). We found that methylation at CpG4 was the most significantly correlated with OS, and PYRmean was the most significantly correlated with PFS. In multivariate Cox models with 1 technique, PYRCpG4 was retained for both OS and PFS. Interestingly, methylation at this CpG site has recently been shown to be correlated (with others) to MGMT mRNA expression and protein expression.27 PYRCpG4 could therefore be of particular interest, all the more so because similar results were found in a different cohort of patients.11
In our hands, methylation-sensitive high-resolution melting and MethyLight had a weaker predictive value. Among the series, 13 and 17 patients were differently classified by pyrosequencing and MethyLight and by pyrosequencing and methylation-sensitive high-resolution melting, respectively (Fig. 5). The majority of conflicting results were observed for the 12 patients with PYRmean results between 9% and 19%. This is consistent with the finding that cell line GB3, used to test reproducibility, had a PYRmean of 9%, but was found unmethylated by methylation-sensitive high-resolution melting or MethyLight. These 2 techniques appear to have lower sensitivity than pyrosequencing and MS-PCR, which could explain in part why they had less predictive value in our series of GBM patients. Interestingly, a great majority of patients in whom heterogeneous methylation was observed by pyrosequencing were classified as unmethylated by MethyLight and methylation-sensitive high-resolution melting. As MethyLight is designed to detect sequences fully methylated at CpGs covered by the primers and the probe, this method could classify as unmethylated some patients displaying heterogeneous methylation. Theoretically methylation-sensitive high-resolution melting should be less affected by inhomogeneous methylation. Methylation-sensitive high-resolution melting relies on a first step of PCR amplification with a unique set of primers designed to bind both methylated and unmethylated DNA. The primers used in this study contain a limited number of CpG sites. They have been shown to be highly sensitive in detecting methylated template depending on the annealing temperature. These experiments were done using mixes of fully methylated template in unmethylated controls, as we did.22 It is possible that a heterogeneous pattern of methylation at the site of annealing can induce a PCR bias directed toward the unmethylated allele, which could explain a lower sensitivity of methylation-sensitive high-resolution melting for some samples. Finally, it can also be hypothesized that testing CpGs separately is a more clinically relevant option. By using a luciferase reporter assay, it has recently been shown that alteration at a single CpG of MGMT could dramatically reduce transcriptional activity.4 That may explain why examination at individual well-selected CpGs, as is done by pyrosequencing, better predicts therapeutic response.
Although the clinical value of immunohistochemical detection of MGMT protein is controversial, in particular because of a high interobserver variability, we decided to include this technique because 1) it is an inexpensive and widespread technique, which theoretically allows the visualization of tumor cells; and 2) several studies have reported significant associations between MGMT protein expression and outcome for glioma patients.17, 18, 28, 29 In our study, a percentage of MGMT-positive cells >23% was significantly correlated with worse OS and PFS. We chose an optimized cutoff, in contrast with other studies where median, proportion of MGMT-positive cells in normal brain, proportion of MGMT-positive non-neoplastic cells, or arbitrary values were chosen. This cutoff was therefore optimized for this series of samples and readings. When a second pathologist analyzed the percentage of MGMT-positive cells in our series of slides, divergent results were noted for a substantial number of cases, similar to previous reports.19 This can explain in part why in a previous study without centralization of IHC, we did not find any association between MGMT immunohistochemical expression and survival.11 This highlights the need for harmonization of not only the staining process (slide pretreatment, antibody type, and dilution), but also the way the percentage of positive cells is established, to render IHC more reliable.
In conclusion, we show in this study that many patients can be classified differently, either as potential responders or as nonresponders to temozolomide, depending on the technique used to define their MGMT status. We found that IHC assessment of MGMT expression was significantly associated with outcome. However, this method also exhibited poor reproducibility and observer variability, impeding current recommendation for use. All the techniques analyzing promoter methylation were of predictive value, but pyrosequencing was found to be the most powerful predictor of prognosis in GBM patients. It also presented very good analytical performances. As frozen tumor tissue is not always available in daily practice, we also validated the feasibility of this technique with paraffin-embedded samples on a subset of patients. All of these points make pyrosequencing an attractive method for determining MGMT status. We are currently implementing an evaluation of pyrosequencing in different laboratories within the framework of a prospective study.
We thank M. Marty and Pierre Rivet for their invaluable technical contribution. Samples in Rennes were collected and stored by the Centre de Ressources Biologiques. The specimens provided by the Marseille team were stored in the AP-HM tumor bank (authorization number 2008/70). Pyrosequencing was performed on the Biogenouest Transcriptome Platform of the University of Rennes 1.
Funding was provided by the French ministry of health (Support for Costly Cancer Technical Evaluation–STIC–Gov-0478).