It is already well known that hypermethylation of the O6-methylguanine DNA methyltransferase (MGMT) gene promoter is a predictive biomarker of response to temozolomide treatment and of favorable outcomes in terms of overall survival (OS) and progression-free survival (PFS) in glioblastoma (GBM) patients. Nevertheless, MGMT methylation status has not currently been introduced into routine clinical practice, as the choice of the ideal technique and tissue sample specimen is still controversial. The aim of this study was to compare 2 analytical methods, methylation-specific polymerase chain reaction (MSP) and pyrosequencing (PSQ), and their use on 2 different tissue type samples, snap-frozen and formalin-fixed paraffin-embedded (FFPE), obtained from a single-center and uniformly treated cohort of 46 GBM patients. We obtained methylation data from all frozen tissues, while no results were obtained for 5 FFPE samples. The highest concordance for methylation was found on frozen tissues (88.5%, 23/26 samples), using PSQ (76.7%, 23/30 samples). Moreover, we confirmed that OS and PFS for patients carrying methylation of the MGMT promoter were longer than for patients with an unmethylated promoter. In conclusion, we considered MSP a limited technique for FFPE tissues due to the high risk of false-positive results; in contrast, our data indicated PSQ as the most powerful method to stratify methylated/unmethylated patients as it allows reaching quantitative results with high sensitivity and specificity. Furthermore, frozen tumor tissues were shown to be the best specimens for MGMT methylation analysis, due to the low DNA degradation and homogeneity in methylation throughout the tumor.
Int J Biol Markers 2015; 30(2): e208 - e216
Article Type: ORIGINAL RESEARCH ARTICLE
Article Subject: Methods in Biomarkers Research
AuthorsLaura Lattanzio, Marzia Borgognone, Cristina Mocellini, Fabrizio Giordano, Ermanno Favata, Gaetano Fasano, Daniela Vivenza, Martino Monteverde, Federica Tonissi, Annalisa Ghiglia, Claudia Fillini, Claudio Bernucci, Marco Merlano, Cristiana Lo Nigro
- • Submitted on 21/07/2014
- • Accepted on 07/10/2014
- • Available online on 25/12/2014
- • Published online on 26/05/2015
This article is available as full text PDF.
Glioblastoma (GBM) is the most aggressive and highly malignant among the primary brain tumors, and represents the most common subtype in adults, with an annual incidence of 3-4/100,000 (1). In 2005, Stupp and coworkers introduced concomitant and adjuvant temozolomide (TMZ) to postoperative radiotherapy in GBM patients, reporting a 2.5-month increase in median overall survival (2). After this trial, the current standard therapy for GBM patients was defined as surgical resection followed by radiotherapy (RT) plus cytotoxic chemotherapy given concomitantly with and after RT (3).
TMZ is a second-generation DNA-alkylating agent that mediates its cytotoxic effect by forming O6-methylguanines (O6-MeG) which preferentially pair with thymines rather than cytosine during DNA replication (4). These DNA adducts lead to base mispairing and double-strand breaks inducing apoptosis and cell death (5). The major complication of treating cancer patients with alkylating agents is chemotherapy resistance due to cellular intrinsic DNA repair mechanisms. Among these, the human O6-methylguanine DNA methyltransferase (MGMT) protein is well known to remove alkyl groups from the O6-position of guanine by an irreversible transfer of the alkyl group to a cysteine residue at its active site (6). Therefore a transcriptionally active
In human cancer, silencing of the
Many studies have already defined
In this study we systematically compared the qualitative MSP and the quantitative PSQ methods for
Materials and Methods
Patients and samples
Tissue samples were collected between 2006 and 2013 from 46 patients treated at the Department of Neurosurgery at S. Croce University Hospital, Cuneo, after written informed consent by the patients and approval by the ethics committee of S. Croce University Hospital in accordance with the Helsinki Declaration. Patients included in the study were newly diagnosed for GBM and treated with standard TMZ-containing chemoradiotherapy protocols. The group of patients consisted of 35 men and 11 women, with a median age of 64.5 years (range 24-84 years). Fifteen of 46 patients were censored, while 29 of 46 are now dead. Median follow-up was 7.4 months, median OS was 10.5 months and median PFS was 7.2 months. For each patient, 2 samples of the primary tumor were obtained: 1 was collected during surgery, immersed in RNA later (Life Technologies, Carlsbad, CA, USA) and immediately snap-frozen in liquid nitrogen, and the other was assembled from biopsy in FFPE sections by use of standard procedures. FFPE tissues were selected by a single expert pathologist and microdissected to have at least a 80% tumor cell content. FFPE samples were prepared at the same time as frozen samples.
To determine the methylation cutoff value for PSQ analysis, we extracted DNA from a pool of 5 normal brain tissues derived from autopsies; the average percentage of methylation of the 5 samples was 8%; thus we considered methylated any tumor sample carrying ≥9% methylation.
DNA extraction and bisulfite treatment
Genomic DNA was extracted from frozen samples using the tissue protocol of the QIAamp DNA Blood mini Kit (Qiagen, Valencia, CA, USA) following the manufacturer’s instructions. DNA from FFPE tissue sections was obtained performing standard proteinase K digestion and phenol extraction. DNA concentration and quality was assessed with a NanoDrop ND-1000 Spectrophotometer (Celbio-Euroclone, Irvine, CA, USA). Genomic DNA from frozen and FFPE tissues (300 ng and 500 ng, respectively) was subjected to bisulfite conversion using the EZ Methylation Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer’s protocol. A universal methylated human DNA (Zymo Research, Irvine, CA, USA) and a human placenta DNA (Sigma Aldrich, St. Louis, MO, USA) were used, respectively, as positive and negative controls for methylation in both MSP and PSQ procedures.
Methylation-specific polymerase chain reaction
To improve MSP results, we attempted a 2-step PCR approach using a first set of primers amplifying the exon 1 region of the
Frozen and FFPE tissue samples were analyzed in triplicate.
PSQ was performed using the PyroMark ID System (Biotage, Uppsala, Sweden). The PSQ primers used for amplification of bisulfite-treated DNAs were designed to cover a region including 9 CpG sites of the
Survival curves were estimated according to the Kaplan-Meyer method, and the differences of
MGMT methylation status analyzed by MSP and PSQ
In the present study we investigated frozen tissue samples and FFPE tissue samples of 46 GBM patients treated with chemoradiotherapy protocols containing TMZ.
Representative methylation-specific polymerase chain reaction (MSP) (A) and pyrosequencing (PSQ) (B) analyses of MGMT promoter in glioblastoma (GBM) tissues. A) Typical results obtained by MSP: samples #80, #71 and methylated DNA present a band in the M primer set (methylation-positive), whereas samples #65, #84 and placenta DNA are unmethylated (band in the U primer set). Neg ctrl = water (no amplification). B) Typical pyrograms obtained for a methylated sample (#72, upper panel) and for an unmethylated sample (#52, lower panel); shadow squares highlight each CpG site analyzed by pyrosequencing with the corresponding percentage of methylation calculated by the software. The light gray box around the single "C" corresponds to the bisulfite control: there must be no peak in this nucleotide, meaning 0% cytosine incorporation.
Comparison of the 2 techniques and the 2 sample specimens
When comparing in 2-by-2 entry diagrams for all the parameters, we did not include the 5 samples that gave no PCR amplification by MSP, making a total of 41 GBM samples.
Comparison of the 2 techniques and of the 2 sample specimens. Diagrams summarize the numbers of methylated samples and highlight the concordance/discordance between the 2 techniques (A) and the 2 sample specimens (B). The central dotted part of the diagrams shows the concordance between groups; discordant results are highlighted in the corresponding clear part. Black boxes indicate the highest concordance for each comparison. FFPE = formalin-fixed paraffin-embedded; MSP = methylation-specific polymerase chain reaction; PSQ = pyrosequencing.
Comparison of the results in discordant samples
MGMT Methylation status in discordant samples
|Patient ID||Frozen tissues||FFPE tissues|
|MSP results (meth/unmeth)||PSQ results (average % meth)||MSP results (meth/unmeth)||PSQ results (average % meth)|
|Comparison of results obtained in discordant samples by MSP and PSQ. Patient ID = patient identification number; FFPE = formalin-fixed paraffin-embedded; MSP = methylation-specific PCR; PSQ = pyrosequencing; meth = methylated; unmeth = unmethylated.|
|# 16||meth||7||not available||2|
Among FFPE samples, 9/13 were discordant and were methylation-positive only if analyzed by PSQ and not by MSP. Three of 13 frozen tissues were found to be methylated by MSP but not by PSQ. Five of 13 samples were shown to be methylated by all techniques except for MSP on FFPE tissues. Seven of 13 samples analyzed by PSQ showed discordant results if the 2 different specimens were examined, and in particular all of them were methylated only for FFPE tissues. Five samples (ID nos. 50, 56, 59, 67 and 90) were found to be unmethylated on frozen tissues by both techniques but methylated on FFPE tissues by PSQ.
Correlation between methylation status and overall survival
We evaluated the correlation between GBM OS and
Overall survival (OS) of glioblastoma patients. Kaplan-Meyer curves show the correlation between methylation status and OS, based on results from methylation-specific polymerase chain reaction (MSP) on frozen tissues (45 patients analyzed) (A); MSP on formalin-fixed paraffin-embedded (FFPE) tissues (40 patients analyzed) (B); pyrosequencing (PSQ) on frozen tissues (45 patients analyzed) (C); and PSQ on FFPE tissues (45 patients analyzed) (D). In the figures, dark gray lines: methylated; light gray lines: unmethylated; p values calculated as described in “Methods.”
Considering as optimal the analysis on frozen tissues, the median OS of patients with a methylated
Correlation between MGMT methylation status and PFS
On the same cohort of patients, we determined the PFS, for 42 patients analyzed by MSP and PSQ on frozen tissues and by PSQ on FFPE, and for 39 patients by MSP on FFPE tissues (
Progression-free survival data (PFS). Kaplan-Meyer curves show the correlation between methylation status and PFS, based on results from methylation-specific polymerase chain reaction (MSP) on frozen tissues (42 patients analyzed) (A); MSP on formalin-fixed paraffin-embedded (FFPE) tissues (39 patients analyzed) (B); pyrosequencing (PSQ) on frozen tissues (42 patients analyzed) (C); and PSQ on FFPE tissues (42 patients analyzed) (D). In the figures, dark gray lines: methylated; light gray lines: unmethylated; p values calculated as described in “Methods.”
The median PFS for patients with methylated
The main focus of this study was the comparison of analytical methods for
It is already well known that
In this study, we initially determined the percentage of
Afterward, we compared MSP/PSQ and frozen/FFPE tissues in terms of concordance and discordance between techniques and sample specimens. Our data suggested that the most suitable technique seems to be PSQ, as it may be used on both frozen and FFPE tissue types with 100% efficiency (46/46 results obtained). Moreover, in our analysis, the highest concordance in methylation was found on frozen tissues (88.5%, 23/26 samples) and using PSQ (76.7%, 23/30 samples). Discordant results, as expected, were found among FFPE tissues: in particular in 5 FFPE samples, we did not obtain any amplification, and 5/13 discordant samples were unmethylated when analyzed by MSP on FFPE tissue but methylated in all other cases. Considering discordant samples, 9/13 among FFPE tissues were found to be methylation-positive only by PSQ, while among frozen tissues, 3/13 were found to be methylated only by MSP; 7/13 among samples analyzed by PSQ were shown to be methylation-positive only on FFPE tissues and 5 samples were found to be unmethylated on frozen tissues by both techniques but methylated on FFPE tissues by PSQ.
It is already well known that the preservation process can influence DNA quality and integrity. Formalin fixation is the standard storage method for tissues, but induces DNA degradation, leading to a poor DNA quality; this, in addition to damage derived from bisulfite conversion could determine the false-positive results and overestimation of methylation level analysis (26). This could explain the large number of discrepant results obtained in FFPE tissues and in particular the higher number of methylated samples analyzed by PSQ: the inefficient conversion matched with the high sensitivity of PSQ could lead to false positive results. In addition, discordant results obtained by MSP on frozen or FFPE tissues could be explained by subtle bias in PCR efficiency obtained with methylated/unmethylated primers (i.e., the degradated DNA in FFPE tissues could hamper PCR amplification, and unmethylated primers, being longer than methylated ones, could be leading to false-negative results). Obversely, cryopreservation techniques protect DNA from degradation but maintenance of a frozen tissue bank is complex and costly, and such banks are lacking in many laboratories. Both techniques present advantages and pitfalls: MSP requires standard and not expensive laboratory equipment such as a PCR amplifier which is actually nowadays present in almost all diagnostic laboratories, but it is a qualitative method and attempts at quantification based on interpretation of gel band intensities may be operator-dependent. In contrast, PSQ provides quantitative levels of methylation at each individual CpG site, includes a conversion control, and is highly sensitive and specific (27), but it requires expensive equipment and specialized personnel. Recent studies have shown that not all CpG sites in the
Moreover, it must be considered that those discrepancies may also derive from intratumoral heterogeneity of surgical sample collection within the glioblastoma. However, recent studies have demonstrated that methylation status was homogeneous in different regions of the tumor when a frozen tissue was used, even if the comparison between frozen and FFPE tissues presented discordant results (27). If MGMT protein expression is considered, some heterogeneity is present in different regions, decreasing progressively from the inner to the outer layer of tumors (28). Thus, even if frozen tissues seem to be the most suitable specimens for
We are aware that the “take-home” message of our study may be influenced by the low sample number, but we consider the following points to be strengths: (i) all tissues examined were derived from the neurosurgery and pathology departments of a single hospital; (ii) all frozen tissues were collected during surgery by a unique team of surgeons, and this actually considerably decreased the variability among samples and with regard to the quality of biopsies and radical surgery; (iii) FFPE tissues were selected by a single expert pathologist, microdissected when necessary and taken with the same biopsy used for GBM diagnosis (29); (iv) MSP and PSQ are routinely used in the laboratories of S. Croce University Hospital. Altogether, this makes our cohort of patients a valid single-center and uniformly treated GBM patient group in which to compare different techniques and tissue specimens.
Moreover, in the univariate analysis, we confirmed that OS and PFS for patients carrying methylation on the
In conclusion, our clinical and technical data obtained by MSP on FFPE tissues affirmed the limit of this technique, suggesting that, whenever possible, the opportunity should be taken to confirm the level of methylation using PSQ, when the clinician intents to use it for OS and PFS predictions and to tailor TMZ treatment in GBM patients.
We thank Dr. Maura Menghi (Diatech, Jesi, AN) for scientific assistance.
- Lattanzio, Laura [PubMed] [Google Scholar] 1
- Borgognone, Marzia [PubMed] [Google Scholar] 2
- Mocellini, Cristina [PubMed] [Google Scholar] 2
- Giordano, Fabrizio [PubMed] [Google Scholar] 3
- Favata, Ermanno [PubMed] [Google Scholar] 4
- Fasano, Gaetano [PubMed] [Google Scholar] 4
- Vivenza, Daniela [PubMed] [Google Scholar] 1
- Monteverde, Martino [PubMed] [Google Scholar] 1
- Tonissi, Federica [PubMed] [Google Scholar] 1
- Ghiglia, Annalisa [PubMed] [Google Scholar] 1
- Fillini, Claudia [PubMed] [Google Scholar] 5
- Bernucci, Claudio [PubMed] [Google Scholar] 4
- Merlano, Marco [PubMed] [Google Scholar] 6
- Nigro, Cristiana Lo [PubMed] [Google Scholar] 1, * Corresponding Author (email@example.com)
Laboratory of Cancer Genetics and Translational Oncology, S. Croce University Hospital, Cuneo - Italy
Neurology Department, S. Croce University Hospital, Cuneo - Italy
Pathology Department, S. Croce University Hospital, Cuneo - Italy
Neurologic Surgery Department, S. Croce University Hospital, Cuneo - Italy
Radiotherapy Department, S. Croce University Hospital, Cuneo - Italy
Medical Oncology, Oncology Department, S. Croce University Hospital, Cuneo - Italy