Quantitative analysis of MGMT promoter methylation status changes by pyrosequencing in recurrent glioblastoma

MGMT promoter methylation status can change in response to several factors, treatment with alkylating therapy being the mechanism more commonly cited in the literature. Some authors have attempted to quantify these alterations, with inconsistent results. This study aims to determine changes in MGMT promoter methylation status by pyrosequencing, which quantitatively yields results, in a cohort of patients reoperated for recurrent glioblastoma and having previously completed the Stupp protocol. Methylation status of the MGMT promoter gene of a total of 24 pairs of glioblastoma preselected tumor samples was retrospectively analyzed using pyrosequencing and depicted as percentages or categories (hypermethylated, intermediate methylation, unmethylated). Matched samples were compared using Wilcoxon signed‐rank test, and log‐rank test was used to establish a correlation with survival data. The median value of MGMT promoter methylation status declined after adjuvant treatment from 20.35% to 14.25% (p = 0.346). A significant correlation between methylation in primary samples and overall survival (p = 0.05) and progression‐free survival (p = 0.024) was found. Intermediate methylation status at recurrence was linked to greater survival after progression, without reaching statistical significance (post‐progression survival [PPS]) (p = 0.217). Although treatment with alkylating chemotherapy was a common feature in all patients of our cohort, switching in both directions was observed when MGMT promoter methylation status was analyzed as a continuous variable. These data suggest that the dynamics of epigenetics may be very complex and not entirely explained by clonal selection influenced by temozolomide.


INTRODUCTION
Gene expression can be controlled by promoter regions located in close proximity to sites where the transcription of DNA into RNA begins. Some of these genomic areas are characterized by clusters of cytosine and guanine dinucleotides, called CpG islands. It is well known that hypermethylation of these CpG islands, catalyzed by enzymes called DNA methyltransferases, can hinder the codification of DNA into proteins without altering the nucleotide sequence, leading to epigenetic silencing. 1 Methylation of the promoter region of the DNA repair enzyme O 6 -methylguanine-DNA-methyltransferase gene (MGMT) has been independently implicated in high-grade glioma prognosis and sensitivity to alkylating agents, such as temozolomide (TMZ), for more than a decade. 2 The MGMT enzyme prevents the process of programmed cell death by removing alkyl groups from guanine residues, 3 which, if left unrestored, inevitably leads to futile attempts of repair and eventually the apoptotic cascade.
Most alkylating agents, including TMZ, utilize MGMT pathways and induce cell death by targeting O 6 -methylguanine adducts. Consequently, lowered MGMT levels (as an effect of epigenetic silencing) result in a greater chemotherapeutic response.
Different direct and indirect methods can address MGMT status in histological samples. MGMT enzymatic activity can be measured from frozen tumor specimens; however, this intricate processing technique and the potential contamination by nonglioma cells render this method valueless for daily clinical practice. 4 The detection of MGMT protein levels by immunohistochemistry has been linked to survival benefits in glioma patients with conflicting results; low interobserver correlation and contamination by nonneoplastic tissue are some possible explanations for such determinations. MGMT mRNA levels can be measured from fresh samples, but this implies a challenging technique that produces results that are difficult to interpret. 5 MGMT promoter hypermethylation correlates with prolonged overall survival (OS) and progression-free survival (PFS) in glioblastoma patients receiving alkylating agents 2,6 and constitutes a routine molecular analysis in many institutions. Several techniques can be used to study MGMT promoter methylation status; most such methods are based on an initial DNA treatment by bisulfite. This reaction induces different conversions depending on whether or not the nucleotides are methylated. Gel-based methylation-specific polymerase (MSP) is the most commonly used test, which is qualitative in nature. By contrast, pyrosequencing (PSQ) provides quantitative information on the percentage of CpG methylation, and it was recently validated in a meta-analysis as a prognostic tool that correlates with OS and PFS. 7 Moreover, a novel meta-analysis based on a Cochrane systematic review suggested that PSQ may be a slightly better predictor of survival than MSP. 8 Additional methylation-specific assays can be reviewed elsewhere. 5 Although DNA methylation patterns remain relatively constant, they can change in response to microenvironmental factors. 9 Several groups have attempted to quantify these genetic alterations focusing on recurrent glioblastoma to detect a possible mechanism that explains universal treatment failure and tumor recurrence, albeit with inconsistent results. 10,11 In this study, 24 matched pre-and post-treatment glioblastoma samples were assessed to identify modifications of the proportion of methylated CpG sites in the CpG island. We postulate that the percentage of methylated residues may decrease after several months, influenced by chemoradiotherapy. Thus, this study aimed to explore the data quantitatively using pyrosequencing, which could reveal undetectable changes when considering methylation as a categorical variable.

MATERIALS AND METHODS
Tissue samples were retrospectively obtained from patients diagnosed and treated with two or more interventions at our institution between 2009 and 2021. Only patients with a confirmed histology of glioblastoma, according to the 2021 WHO classification, 12 in both specimens were selected for genetic analysis, which meant excluding radionecrosis, grade 4 IDH1 mutated lesions, and malignant transformation from the study. Clinical data included sex, age at diagnosis, type of treatment after the first surgery, OS, and PFS in months. OS was defined as the period from initial diagnosis until death or last followup and PFS as the interval from diagnosis to objective tumor recurrence in neuroimaging based on RANO criteria. 13 Finally, post-progression survival (PPS) was defined as the time elapsed from tumor recurrence to death.
Pathologic specimens were preselected to avoid contamination from necrotic and inflammatory tissue. Areas with the highest amount of neoplastic cells were marked by our pathologist and manually dissected under microscopic guidance for DNA analysis. DNA extraction from formalin-fixed paraffin-embedded samples was performed using the Qiagen QIAmp DNA FFPE Tissue Kit (Qiagen, Valencia, CA), and the DNA amount was measured via the NanoDrop Lite technique. Bisulfite conversion was accomplished according to the Qiagen EpiTec Fast protocol. Afterward, PCR amplification of the MGMT promoter genomic sequence in exon 1 on chromosome 10 (from 131 265 519 to 131 265 537 pb: CGACGCCCGCAGGTCCTCG; 76-79 CpG sites) was completed using the therascreen MGMT Pyro Kit. Before the final pyrosequencing analysis on the PyroMark Q24 System, samples were prepared as stated in the manufacturer's handbook. Quantitative pyrosequencing results of the 4 CpG analyzed dinucleotides were controlled by samples with pre-established results (i.e., methylated and unmethylated) to detect possible incomplete bisulfite conversion, which can result in biased methylation quantification.
MGMT promoter methylation was classified according to recently designed cut-off levels, 14 in unmethylated (<9%), intermediate (9-29%), and highly methylated (>29%) tumors, and measured as a continuous variable, indicated by the average percentage value of the examined CpGs. Statistical analysis was performed using SPSS 22.0 (IBM, Armonk, NY). We initially investigated our data distribution pattern by a Shapiro-Wilk test. Based on our results, nonparametric tests were used. Wilcoxon signedrank test was used to compare the methylation status of matched samples from the same patient after treatment influence. The log-rank test was used to establish a correlation between survival data and methylation status, and pvalues less than or equal to 0.05 were considered significant.

RESULTS
In our institution, 36 patients with glioblastoma underwent surgical treatment for recurrence between 2009 and 2021. Four cases of malignant transformation from anaplastic gliomas, two cases of grade 4 IDH1 mutant astrocytoma, and a case of radionecrosis were excluded from the analysis. In five patients, there was no tissue availability, and thus they were not included. Therefore, MGMT promoter methylation status was examined in 24 paired samples.
The mean age of the cohort was 55 years (range 36-70). All patients received microsurgery assisted by different surgical tools, including neuronavigation, 5-ALA fluorescence, and intraoperative magnetic resonance imaging (MRI) depending on the period. Based on an early postoperative MRI, gross total resection was achieved in 22 out of 24 patients (91.6%). Twenty-one (87.5%) patients completed the Stupp protocol, consisting of concomitant chemoradiotherapy followed by six cycles of adjuvant TMZ. The remainder (12.5%) did not accomplish all the cycles because of early radiological progression or chemotherapy toxicity. Median PFS and OS were 12.5 and 25 months, respectively.
Eight (33.3%) primary malignancies were considered to be highly methylated; 10 (41.7%) were intermediate, and six (25%) were unmethylated. Subsequent histological analysis of paired cases revealed that methylation status remained in the same category in 15 (62.5%) and changed in nine (37.5%). Modification in methylation status was registered in both directions, although was slightly more common from higher grade to lower. Patients' main clinical characteristics are illustrated in Table 1.
The median value of MGMT promoter methylation status decreased after adjuvant treatment from 20.35% to 14.25%. Wilcoxon signed-rank test was used to compare the methylation status of matched samples without obtaining statistical significance in the difference between paired values (p = 0.346).
A log-rank test was used to determine the relationship between the degree of methylation and survival data. A statistically significant correlation was found between the proportion of methylation in primary samples and OS. Median OS was 32, 17, and 19 months in the highly methylated, intermediate, and unmethylated groups, respectively (p = 0.05). Likewise, a correlation was found when studying primary methylation and PFS (20, 10, and 10 months, respectively; p = 0.024). Kaplan-Meier survival curves for OS and PFS are presented in Figure 1. Tumors whose methylation status at recurrence was intermediate were associated with greater PPS, without reaching statistical significance (9, 16, and 8 months, respectively; p = 0.217). Figure 2 represents percentage changes in MGMT methylation status of the whole cohort.

DISCUSSION
Following recent and validated cut-off values, 14 a change in the category of MGMT promoter methylation status between primary and recurrent glioblastoma was noted in nine (37.5%) cases. Of these, switching from higher to lower methylation levels was slightly more frequent than contrariwise. These results partially agree with a recent meta-analysis 10 in which a change from methylated to unmethylated was found to be almost twice as frequent. It must be taken into account that a high proportion of studies included in this meta-analysis used the MSP technique and dichotomized tumors into methylated or unmethylated without considering the intermediate methylation category. When analyzing the methylation level as a continuum, we observed increments or decrements nearly equally distributed in our cohort. However, the median value for the whole series dropped by approximately 6% between the first and second pathologic analyses.
Survival data of patients and the proportion of tumors exhibiting higher degrees of methylation of MGMT promoter gene were higher in our series than in previous reports. [14][15][16] This may reflect our highly selective indications for redo procedures since we do not usually consider a reoperation for incomplete resections at first surgery or PFS intervals lower than 6-9 months. Thus, we might be excluding patients with more aggressive lesions (i.e., unmethylated). Survival benefit among patients whose tumors contain a methylated MGMT promoter is well established. 2 Still, to the best of our knowledge, this was the first study to consider methylation status as a continuous variable and investigate its correlation with survival. As in previous reports, 10,17,18 we observed that methylation was linked to OS and PFS, which may reflect responsiveness to TMZ. The association between methylation status at recurrence and PPS was unexpected. Tumors whose methylation status was intermediate were more linked to post-progression survival than hypermethylated ones. One possible explanation is that the presence of hypermethylated promoters and a probable lower expression of MGMT favors the accumulation of other DNA lesions that could confer a worse prognosis compared to the group of lesions classified as "intermediate methylation." As an example, it has been shown that epigenetic inactivation of the MGMT gene can facilitate the appearance of mutations in TP53 and other genes implicated in the process of tumorigenesis or disease progression, 19,20 even though this may be again influenced by other factors such as the limited size of our study group.
Methylation status results may be dependent on the laboratory test used for its determination. MSP has been, to date, the most used technique to validate MGMT methylation status change detection. 17,[21][22][23][24][25] However, Park et al. 26 concluded that methylation changes might remain undetectable when analyzing tumors by MSP, whereas other techniques, such as methylation-specific multiplex ligation probe amplification (MS-MLPA), can readily uncover the downward shift of methylation in recurrent tumors. In our study, pyrosequencing was used to detect these changes, but results similar to those previously published were obtained.
On the other hand, tumor samples for methylation analysis can provide different results if not previously selected. To avoid this kind of contamination in our study, areas with a higher amount of neoplastic cells were marked and manually dissected to exclude nonneoplastic elements as previously described. 21 Intratumor heterogeneity concerning epigenetic silencing of the MGMT gene has been reported by some groups; 21,27 others found that MGMT promoter methylation status was relatively homogeneous within tumors 22,28 and therefore we only  examined one preselected area from each tumor. By contrast, MGMT protein expression evaluated by immunochemistry appears to have a decreasing gradient from the inner to the outer portion of tumors. 27 These data support the idea that there is low concordance between MGMT promoter methylation and protein levels and that other mechanisms may be involved in its expression apart from epigenetic silencing. 29 Another issue is the definition and validation of the technically and clinically relevant cut-off value for MGMT promoter methylation and the fact that most samples will be classified similarly by most tests, although different CpGs are interrogated. 20 Due to its mechanism of action, clonal selection by TMZ has been elucidated as a primary factor that could explain treatment resistance and methylation pattern reshaping within the clinical course. TMZ may deplete cell lines with low MGMT expression (i.e., hypermethylated promoter), inducing chemoresistance by selecting clones with low levels of methylation that are more resistant to alkylating drugs. 18 Although exposed in a mathematical model, 30 this feature remains controversial since changes from unmethylated to highly methylated status in paired samples after chemotherapy can be detected in several studies, including ours. 17,18,21,26,31,32 In contrast, other authors found differences in the methylation pattern of recurrent glioblastoma depending on whether the patient was treated with RT followed by TMZ versus concurrent RT/TMZ. 18,24 Furthermore, it has been reported that switching from methylated to unmethylated status can happen without chemotherapy in lower-grade glioma cases. 23 We consider our cohort to be very homogeneous in terms of post-surgical treatment since all patients received simultaneous chemoradiotherapy, and 87.5% were treated with at least six cycles of adjuvant TMZ. These data suggest that the epigenetic dynamics of MGMT promoter methylation status over the clinical course might be very complex and influenced by several processes that are still poorly understood. Additional mechanisms have been theorized in which chemotherapy with alkylating agents may promote dedifferentiation from noncancer stem cells to glioma stem cells, responsible for tumor growth and chemoresistance acquisition. 33 Several drawbacks need to be considered in the present study. Its small sample size limits the statistical power of our results. A future multicenter, more extensive study would be of interest that considered multiarea biopsies and other methylation measuring techniques that might lead to different conclusions regarding modifications in epigenetic changes.
Although MGMT promoter methylation status was examined quantitatively, we did observe both increments and decrements in paired samples after TMZ treatment. These data suggest that clonal selection caused by alkylating agents does not entirely explain the dynamics of MGMT epigenetic changes. Factors related to MGMT tumor heterogeneity, laboratory tests used, and sample quality may also influence results. Despite our small cohort size, MGMT promoter methylation status seems to correlate with OS and a greater chemotherapeutic response, which impacts longer PFS.

ACKNOWLEDGMENTS
This study received financial support from IdIsBa (Institut d'Investigaci o Sanitària Illes Balears), a regional institution that receives public funds to support medical research. We were allowed to acquire a pyrosequencing kit to investigate MGMT methylation status of paired samples.

DISCLOSURE
The authors declare that they have no conflicts of interest in connection with this article.

APPROVAL OF THE RESEARCH PROTOCOL
The study protocol was approved by the institutional ethics committee (Comité de Etica de la Investigaci on de las Islas Baleares (CEI-IB)). We assert that all methods were performed in accordance with the relevant guidelines and regulations.Informed consent: Because our research was retrospective in nature, the CEI-IB approved the waiver for informed consent.

REGISTRY AND REGISTRATION NO. OF STUDY/TRIAL
IB 4893/22PI (see attached file).

ANIMAL STUDIES
NA.

RESEARCH INVOLVING RECOMBINANT DNA
NA.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.