Influence of MMR, MGMT Promotor Methylation and Protein Expression on Overall and Progression-Free Survival in Primary Glioblastoma Patients Treated with Temozolomide

Glioblastoma is the most common malignant brain tumor in adults. Standard treatment includes tumor resection, radio-chemotherapy and adjuvant chemotherapy with temozolomide (TMZ). TMZ methylates DNA, whereas O6-methylguanine DNA methyltransferase (MGMT) counteracts TMZ effects by removing the intended proteasomal degradation signal. Non-functional MGMT mediates the mismatch repair (MMR) system, leading to apoptosis after futile repair attempts. This study investigated the associations between MGMT promoter methylation, MGMT and MMR protein expression, and their effect on overall survival (OS) and progression-free survival (PFS) in patients with glioblastoma. MGMT promoter methylation was assessed in 42 treatment-naïve patients with glioblastoma WHO grade IV by pyrosequencing. MGMT and MMR protein expression was analyzed using immunohistochemistry. MGMT promoter methylation was present in 52%, whereas patients <70 years of age revealed a significantly longer OS using a log-rank test and a significance threshold of p ≤ 0.05. MGMT protein expression and methylation status showed no correlation. MMR protein expression was present in all patients independent of MGMT status and did not influence OS and PFS. Overall, MGMT promoter methylation implicates an improved OS in patients with glioblastoma aged <70 years. In the elderly, the extent of surgery has an impact on OS rather than the MGMT promoter methylation or protein expression.


Introduction
Despite advances in surgery, radiation and chemotherapy, the diagnosis of glioblastoma multiforme (GBM) is devastating and the outcome poor. Next to neurological examination, diagnostic tools, i.e., magnetic resonance imaging (MRI), computed tomography [1] and classical histopathology, as well as other invasive and advanced methods [2,3], provide the basis for histopathological diagnosis of GBM. Tumor resection might be a curative intervention in the case of a gross total resection (GTR) compared to a subtotal resection (STR). GTR is defined as the complete removal of all tumor tissue, supported by imaging approaches. In addition to the surgical intervention, treatment including temozolomide (TMZ)-based chemoradiotherapy is used as a standard of care for GBM [4]. Many studies showed that O6-methylguanin DNA methyltransferase (MGMT) promoter methylation serves as a prognostic factor for progression-free survival (PFS) and overall survival (OS) undergoing STR and 68.3% GTR. The majority of the patients in this study were treated with the Stupp protocol (32; 76.2%), whereas two patients were treated with radiotherapy alone (4.8%) and eight patients did not receive any further treatment (19.0%). The IDH1 status was assessed in 88.1% of the included patients, demonstrating a 7.1% IDH1 mutation rate, which is within the published range [36,37]. Table 1. Characteristics of patients with glioblastoma WHO grade IV. The characteristics are listed for all patients (All), patients younger than 70 years of age (<70 years) and patients with an age of 70 years or more (≥70 years). Overall, the patients' characteristics reflect a balanced study cohort, with a high percentage of GTR and a representative KPS.

Overall Survival, Sex, Age and Resection Association
The median OS for the entire GBM patient cohort was 11.5 months (95% CI: 5.0-17.0). Stratifying by age, patients <70 years revealed a longer median OS, of 15 months, compared to 4 months for patients ≥70 years (p = 0.032) ( Figure 1A).
A sex-related difference in OS was not detected (Figure S1), with a median OS for females of 14.5 months (95% CI: 3.0-34.0) compared to 8 months for males (95% CI: 4.0-17.0) (p = 0.37) in the total cohort. Within the subgroup of patients with GTR an OS advantage compared to patients with STR (median OS of 13 and 4 months, respectively; p = 0.3) was observed ( Figure S2), although not reaching statistical significance. For PFS, neither age-related nor a sex-related effects were determined (p = 0.460 and p = 0.180, respectively). However, GTR compared to STR demonstrated a benefit for PFS (median PFS of 214 and 102 days, respectively, p = 0.003). The results are summarized in Table 2. ing MGMT promoter methylation. The hazard ratio (HR) and the 95% confidence interval in are stated in brackets. Statistical significance between the groups was evaluated using a log-rank test.
A sex-related difference in OS was not detected (Figure S1), with a median OS for females of 14.5 months (95% CI: 3.0-34.0) compared to 8 months for males (95% CI: 4.0-17.0) (p = 0.37) in the total cohort. Within the subgroup of patients with GTR an OS advantage compared to patients with STR (median OS of 13 and 4 months, respectively; p = 0.3) was observed ( Figure S2), although not reaching statistical significance. For PFS, neither age-related nor a sex-related effects were determined (p = 0.460 and p = 0.180, respectively). However, GTR compared to STR demonstrated a benefit for PFS (median PFS of 214 and 102 days, respectively, p = 0.003). The results are summarized in Table 2. MGMT promoter methylation. The hazard ratio (HR) and the 95% confidence interval in are stated in brackets. Statistical significance between the groups was evaluated using a log-rank test. In summary, the data underline the beneficial influence on OS in younger patients with GBM and, in general, patients with GTR.

MGMT Promoter Methylation Causes a Survival Benefit in Younger GBM Patients
In total, MGMT promoter methylation (MGMT met+ ) was present in 50% of the patients (representative pyrograms are depicted in Figure S3). Patients with unmethylated MGMT promoter (MGMT met− ) did show a slight advantage during the first 6 months after surgery ( Figure 1B). This slight advantage could likely be explained by a resection effect, since 76.2% of patients with MGMT met− underwent GTR compared to 60.0% GTR in the MGMT met+ group. However, the median OS was observed to be longer in patients with MGMT met+ compared to patients with MGMT met− , at 14 and 11 months, respectively, (p = 0.028) ( Figure 1B). Stratification of MGMT methylation status and age revealed a statistically significant longer OS for younger patients compared to elderly patients. Patients <70 years with MGMT met+ (n = 11) showed a median OS of 31 months compared to the MGMT met− subgroup (n = 12), with a median OS of 13 months (p = 0.003) ( Figure 1C). In the ≥70 years subgroup of patients, patients with MGMT met− (n = 9) and patients with MGMT met+ (n = 10) revealed a similar OS ( Figure 1D). The univariate Cox proportional hazard model revealed an increased risk for patients <70 years and an unmethylated MGMT promoter (hazard ratio (HR): 4.85 (95% CI: 1.59-14.74), p = 0.005) compared to the patients with MGMT met+ (Table 3). STR represented a risk for patients ≥70 years (HR: 4.05 (95% CI: 1.11-14.79), p = 0.035) compared to GTR. Analyses of PFS demonstrated no survival benefit for any subgroup of GBM patients, independent of age and MGMT promoter methylation status. Table 3. Univariate Cox proportional hazard model of OS in patients with glioblastoma WHO grade IV. The data are shown for all patients (All), patients younger than 70 years of age (<70 years) and patients with an age of 70 years or more (≥70 years). References in the model were female sex, methylated MGMT promoter, low MGMT protein expression, gross total resection and a Karnofsky performance score below 70. The hazard ratio (HR), the 95% confidence interval (CI) and the corresponding p-value are listed. Overall, the results showed an OS advantage for younger patients with MGMT met+ , while the MGMT met− showed a concomitant risk for tumor-related death.

MGMT Promoter Methylation Does Not Correlate with MGMT Protein Expression
Classification by IHC revealed high MGMT protein expression for 26% of patients (MGMT high ), while in 74% of patients, low MGMT protein expression (MGMT low ) could be detected (exemplary images are shown in Figure 2). For both groups, the MGMT high and MGMT low group, comparable percentages of MGMT promoter methylation, 45.5% (n = 5/11) and 52% (n = 16/31), respectively, were determined ( Table 2). No correlation could be monitored between MGMT promoter methylation and MGMT protein expression (phi = 0.054).
Association analyses of the different subgroups and the PFS revealed no statistical differences. However, a trend could be monitored, whereby the MGMT met+ /MGMT low patients (n = 12) showed the longest PFS with a median of 281 days compared to all other subgroups. The MGMT met− /MGMT low patients (n = 14), demonstrated a median PFS of 149 days. In both subgroups expressing MGMT protein, a decreased median PFS was observed (MGMT met− /MGMT high , n = 4, median PFS 211 days and MGMT met+ /MGMT high , n = 4, median PFS 118 days), albeit the sub-cohorts encompassed low numbers of patients.
To summarize, epigenetic silencing of MGMT by promoter methylation does not correlate with protein expression of MGMT. Although methylated, a sub-cohort of patients showed a clear protein expression pattern (MGMT met+ /MGMT high ), which might indicate that the epigenetic silencing must not result in a complete downregulation of protein expression. Furthermore, our data suggest a tendency for a decreased median PFS in the MGMT high sub-cohort.

MGMT Promoter Methylation Status as a Prognostic Factor Dependent on Patient Age
A multivariate Cox proportional hazard model, including sex, MGMT promoter methylation status, MGMT protein expression status and type of resection, was performed and is summarized in Table 4. In the entire cohort, patients with MGMT met− showed an increased associated risk for a reduced OS (HR: 2.69 (95% CI: 1.19-6.07), p = 0.017). Patients <70 years indicated a more prognosticating effect of the MGMT promoter methylation status (HR: 6.14 (95% CI: 1.78-21.21), p = 0.004). For STR, the entire cohort revealed an increased risk of death, although not statistically significant. In patients ≥70 years, a statistically significant increased risk was observed (HR: 5.74 (95% CI: 1.43-23.05), p = 0.012).
The results of our analyses underline the importance of a GTR, with a focus on the ≥70 years subgroup of patients compared to STR and evaluation of MGMT status in patients <70 years as prognostic factors. Table 4. Multivariate Cox proportional hazard model of OS in patients with GBM. The data are shown for all patients (All), patients younger than 70 years of age (<70 years) and patients with an age of 70 years or more (≥70 years). References in the model were female sex, methylated MGMT promoter, low MGMT protein expression and gross total resection. The hazard ratio (HR), the 95% confidence interval (CI) and the corresponding p-value are listed.

MGMT Promoter Methylation Status as a Prognostic Factor Dependent on Patient Age
A multivariate Cox proportional hazard model, including sex, MGMT promoter methylation status, MGMT protein expression status and type of resection, was performed and is summarized in Table 4. In the entire cohort, patients with MGMT met− showed an increased associated risk for a reduced OS (HR: 2.69 (95% CI: 1.19-6.07), p = 0.017). Patients <70 years indicated a more prognosticating effect of the MGMT promoter methylation status (HR: 6.14 (95% CI: 1.78-21.21), p = 0.004). For STR, the entire cohort revealed an increased risk of death, although not statistically significant. In patients ≥70 years, a statistically significant increased risk was observed (HR: 5.74 (95% CI: 1.43-23.05), p = 0.012). Table 4. Multivariate Cox proportional hazard model of OS in patients with GBM. The data are shown for all patients (All), patients younger than 70 years of age (<70 years) and patients with an age of 70 years or more (≥70 years). References in the model were female sex, methylated MGMT promoter, low MGMT protein expression and gross total resection. The hazard ratio (HR), the 95% confidence interval (CI) and the corresponding p-value are listed. The results of our analyses underline the importance of a GTR, with a focus on the ≥70 years subgroup of patients compared to STR and evaluation of MGMT status in patients <70 years as prognostic factors.

Stable Expression of MMR Proteins Independent of MGMT Status
Evaluating the expression results of proteins related to the MMR system, including MLH1, MSH2, MSH6 and PMS2 by IHC, revealed a stable expression in all GBM patients of the tested cohort. MMR protein expression in tumor cells was present with a score of 2 to 3, independent of MGMT status (promoter methylation and protein expression), sex, age, IDH1 mutation status or treatment (Figure 3).

Stable Expression of MMR Proteins Independent of MGMT Status
Evaluating the expression results of proteins related to the MMR system, including MLH1, MSH2, MSH6 and PMS2 by IHC, revealed a stable expression in all GBM patients of the tested cohort. MMR protein expression in tumor cells was present with a score of 2 to 3, independent of MGMT status (promoter methylation and protein expression), sex, age, IDH1 mutation status or treatment (Figure 3).

Discussion
Prognostic factors are of particular interest in brain tumors. The prognostic value of MGMT promoter methylation in newly diagnosed, as well as recurrent GBM is confirmed by a plethora of studies [38]. In our study, we could demonstrate that the MGMT promoter methylation status and the combination with MGMT protein expression are of prognostic value in newly diagnosed GBM treated with standard therapy. In particular, the prognostic value of MGMT on epigenetic and protein level is associated with age in GBM patients. Further, the observed effects on OS and PFS seem to be independent from the MMR system. As demonstrated in this study, all tested tumor samples expressed MLH1, MSH2, MSH6 and PMS2 in normal and tumor cells to a high level (both, score and intensity), and there was no correlation with any patient characteristics. Therefore, the evaluation of our data suggests that analyzing protein expression of the MMR system has no impact as a prognostic factor for GBM. Epigenetic silencing of MGMT or MGMT protein expression does not influence the expression of MMR.
Generally, the median OS of the entire cohort of this study is comparable to other studies investigating patients with GBM [39,40]. However, other studies show a longer median OS [4,41,42]. This difference is likely due to a higher number of included elderly patients in our study. This reflects published data including elderly individuals [43].
Fifty-two percent of the GBM tested showed MGMT promoter methylation resulting in an age-related survival benefit in favor of the <70 year sub-cohort. The relationship between MGMT protein expression and MGMT silencing by epigenetic regulation, like hypermethylation of the promoter region, is still controversially discussed. Several studies showed that MGMT protein expression has no clear effect on the clinical outcome of patients [44][45][46]. However, in our study, a subgroup of patients with a low MGMT protein expression and MGMT promoter methylation demonstrated a benefit in median OS compared to patients expressing the MGMT protein in tumor cells. Interestingly, even the subgroup with an unmethylated MGMT promoter, yet with low MGMT protein expression, revealed a better survival compared to those patients expressing the MGMT protein. The results of both of these defined subgroups are in line with findings of Lalezari et al. [47]. It was indicated that low protein expression alone was inadequate for a better outcome. Interestingly, there is a sub-cohort of MGMT-promoter-methylated patients with MGMT protein expression showing a worse median OS. This outcome might be due to a partial escape of the epigenetic silencing of the MGMT protein and an efficient DNA repair via the functional MMR or other DDR systems, such as BER, in the TMZ-treated tumor cells [47][48][49].
Overall, it can be concluded, that a partial correlation of an effective MGMT silencing via epigenetic alterations is present in a portion of patients with GBM, while in others a "mosaic-like" pattern, such as unmethylated MGMT promoter and low protein expression or methylated MGMT promoter and present MGMT protein expression, exist.
Evaluation of our data suggests that analyzing protein expression of the MMR system has no impact as a prognostic factor for GBM. Epigenetic silencing of MGMT or MGMT protein expression does not influence the expression of MMR. However, mutations or hypermethylation of the MMR proteins, such as MLH1, a dynamic expression pattern or the involvement of other DDR pathways [48] upon/after TMZ treatment, remain to be clarified in future studies.
Besides molecular factors, the extent of neurosurgical intervention has prognostic value on the survival in GBM [50][51][52]. The results of this study show that GTR of the tumor results in an OS and PFS advantage, whereas STR is associated with decreased survival probabilities dependent on age. This is in line with published data [40,53,54]. Although only elderly patients, ≥65-year-old patients were included in the study by Bruno et al., and a clear advantage of GTR compared to STR was observed for OS and PFS [53]. Further study performed by Almenawer et al., showed an OS of 8.68 and 14.04 months for STR and GTR, respectively, and a PFS of 4.31 months for STR and 7.03 months for GTR [54]. These findings underline the importance of surgery in the neuro-oncological management.
The strengths of the study are (i) exclusively treatment-naïve GBM grade IV patients, (ii) with inclusion of MGMT protein expression with the evaluation of the nucleus and cytoplasmic distribution, the cytoplasmic localization could indicate the trafficking of MGMT loaded with methyl towards proteasome for degradation, and (iii) balanced groups for age, MGMT promoter methylation status and surgical resection type. The major limitation of our study is that stratifying into the groups of interest resulted in small sample sizes in the respective subgroups, thus, solely large effects could be observed.

Patient Selection
The implementation of this study is subject to the ethical votes (AN5220 329/4.4, AN2014-0068 334/424) of the Medical University of Innsbruck. Only patients with written informed consent were included in the study. Tissue samples of diagnosed GBM patients were collected from 2015 to 2018. The following inclusion criteria were applied: (i) patients diagnosed with GBM WHO grade IV, (ii) absence of other concurrent tumors, (iii) treatment-naïve patients, and (iv) availability of tumor tissue for analyses. Medical history, including age, sex, KPS, surgical resection and therapeutic intervention, were documented and evaluated. OS and PFS were calculated from the date of surgery. Surgical re-section was performed via craniotomy, depending on tumor location and preoperative MRI. The extent of resection was measured on postoperative contrast-enhanced T1 MRI (within 48 h after resection). The extent of the resection was determined as GTR (≥98% tumor removal) and STR (90 to 98% tumor removal).

Tissue Sampling
Surgical specimens were formalin-fixed and, after macroscopic evaluation by a trained pathologist, paraffin-embedded (FFPE). After standard hematoxylin/eosin staining (Diapath, Sanova, Vienna, Austria), diagnosis was performed according to the WHO classification (2016 version) [55] by experienced neuropathologists, at the Medical University of Innsbruck and at the Institute of Neuropathology and Neurochemistry, Medical University of Vienna.

Methylation Analysis
Following the histopathological evaluation by an experienced pathologist, three to five 5 µm thick slices of the FFPE material were used for further DNA isolation. Genomic DNA was isolated using the EZ1 DNA investigator kit or QIAamp DNA FFPE Tissue Kit, and bisulfite conversion was performed using the EpiTect Fast FFPE bisulfite Kit (all Kits from Qiagen, Hilden, Germany), according to manufacturer's recommendation. For pyrosequencing, the Pyro Mark Therascreen MGMT Kit (Qiagen, Hilden, Germany) was used to assess four CpG sites. In total, 50 ng of bisulfite-converted DNA was used for amplification PCR. All analyses were performed as duplicates. As sequencing controls, an unmethylated and a methylated control DNA were processed in parallel. The CpG sites were located in exon 1 of the human MGMT gene (CpG 76-79). A mean methylation score of <8% was classified as unmethylated, as previously published [56,57].

Immunohistochemical Analysis
For the 42 cases, immunohistochemistry was performed on 2 µm thick slices of FFPE material. Following deparaffination of the slices, MMR protein staining was performed using the OptiView DAB IHC Detection Kit or the Ventana system, as recommended by the manufacturer (Roche, Basel, Switzerland). Immunohistochemical staining of the MGMT protein was performed manually on FFPE sections. For antigen retrieval, a heat-induced epitope retrieval was performed in a water bath at 98.5 • C for 40 min, in a citrate buffer (pH 6.0). For detection, the Dako REAL EnVision Detection System (Agilent, Santa Clara, CA, USA) and peroxidase/DAB+/rabbit/mouse was used. The following antibodies and dilutions were used: anti-MGMT antibody, clone MT3.1 (Merck Millipore, Burlington, MA, USA), 1:100 in blocking solution; MLH1 (M1) (Roche Cell Marque, Basel, Switzerland), 1.4 µg/mL pre-diluted; MSH2 (G219-1129) (Roche Cell Marque, Basel, Switzerland), 3.62 mg/mL pre-diluted; MSH6 (BC/44) (Biocare Medical, Pacheco, CA, USA), 1:50 in sCC1; PMS2 (EPR3947) (Roche Cell Marque, Basel, Switzerland), 14.37 µg/mL pre-diluted; IDH1 R132H (clone H09, Dianova, Hamburg, Germany) 1:30. The intensity of the immunohistochemical staining was scored from 0 to 3, representing very weak, weak, moderate and strong staining, respectively. The density of MMR and MGMT protein expression was scored in percent of the tumor area. The MGMT expression scores were assigned to low (scores: 0 and 1) and high (scores: 2 and 3) expression. The expression patterns of MGMT, MMR and IDH1 were evaluated by at least two independent researchers (J.H., S.S., R.H. and A.M.B.-T.), blinded to MGMT promoter methylation status and further clinical information.

Statistics
Correlation analyses for dichotomous data were assessed using the phi coefficient. Survival probabilities were assessed for OS and PFS. Kaplan-Meier curves were compared using the log-rank test. Univariate and multivariate analyses were carried out using Cox proportional hazards models. A p-value of <0.05 was considered as statistically significant. All statistical analyses were performed using R version 4.0.3.

Conclusions
Summarizing the results of this study, determination of MGMT promoter methylation status and concomitant MGMT protein expression status have an additional prognostic value, whereby the former is particularly of prognostic value in patients younger than 70 years. The analysis of the MMR system by IHC could not demonstrate any diagnostic or prognostic relevance in order to improve the treatment of GBM patients. Further analyses are needed in the future to define the role of the epigenetic MGMT silencing regulation and its effectiveness and to strengthen its application as a prognostic and predictive molecular marker in GBM.