Correlation of epigenetic markers with glioblastoma patients’ survival and molecular subtypes: insights from methylation-specific MLPA analysis

Samples and survival data collection

Following approval from the Ethical Committee of Farhat Hached University-Hospital, Sousse, Tunisia (IORG 0007439 ERC 02 12 2024), tissue samples of 50 Tunisian adult GB cases were selected. They had a mean age of 49 years old and sex ratio 1.4. Samples and clinical data were retrospectively collected from the patients who consulted Neurosurgery Department and referred to Histopathology Department since 2012. All pediatric cases as well as other types of gliomas were excluded. Two independent pathologists confirmed the histopathological diagnosis separately (Histopathology Department in Farhat Hached University Hospital, Sousse, Tunisia and Institute of Cancer Research “ICR’’, London, UK). All cases were diagnosed as GB and classified as grade IV gliomas according to the 2007 WHO classification criteria [1], 3]. Survival duration was collected from all patients and calculated from the time of diagnosis to death or censor.

DNA extraction

The DNA was extracted from brain tumor Formalin-Fixed Paraffin-Embedded (FFPE) tissue specimens according to the manufacturer’s protocol (QIAGEN QIAMP FFPE kit).

Methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA)

The methylation profile of all samples was determined using methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA) previously described as a sensitive technique with FFPE samples and showed reproducible and accurate results [16].

MS-MLPA was chosen because it presented the advantage of targeting a number of CpG sites in 65 tumor suppressor genes and MMR genes simultaneously using ME001, ME002, ME003, ME004 probe mix kits (MRC Holland, Amsterdam, Netherlands). Data analysis was performed with GeneMarker^® software (SoftGenetics, USA). For each gene, the average of all promoter probe ratios was calculated. The threshold value of 0.25 was chosen on the basis of previous reports to detect methylation [13], 18], 19]. Thus, we interpreted average ratios <0.25 as absence of methylation and >0.25 as methylated.

Statistical analysis

Statistical analysis was performed using SPSS software, version 21.0 (SPSS Inc., Chicago, IL). Associations between categorical variables were evaluated with the chi-square test (Chi²) and two-tailed Fisher’s exact test. Survival curves were estimated by the Kaplan-Meier method. A p value less than 0.05 was considered significant for both tests.

Methylation analysis and survival correlation

As a first step, we analyzed the methylation status of targeted genes. Results were summarized in supplementary material presenting the methylated status in each gene. Some genes were present twice. Each value corresponded to a specific probe for different CpG sites in the promoter region. LMNA, MSH6, RASF1, TWIST1, SFRP1, MGMT, CCND2 and PMXMP4 were the most frequently methylated genes (methylated in more than the 50 % of GB samples) (Supplementary Material 1).

Only seven genes revealed significant overall survival differences when the promoter was methylated or not. Genes that showed longer OS when the promoter was unmethylated are ATM (p=0.001), TP53 (p=0.001), TP73 (p=0.002), BRCA1 (p=0.012), MSH6 (p=0.008) and BRCA2 (p=0.008). However, the MGMT gene was the unique gene with methylated promoter that was significantly associated with longer OS (p=0.046) (Table 3, Figure 1). As expected, OS was significantly lower for patients with methylated ATM, BRCA1 and BRCA2 (from 6 to 11 months).

Figure 1:

Significant survival correlation with promoters’ methylation status. (Blue) unmethylated, (green) methylated. OS, overall survival (Month), p value of Kaplan-Meier survival test.

In addition, the methylation of tumor suppressor genes TP53 and TP73 showed a significant lower OS (10–9 months). Regarding the MSH6 as MMR genes, our results showed that patients with MSH6 presented significant lower survival (14 months, p=0.008). In parallel, our analysis revealed the significant OS difference of patients with MGMT methylations, taking into consideration that the studied samples are analyzed retrospectively and that the patients included, have already undergone treatment. This explains the longer OS with patients with MGMT methylation, as they were more sensitive to the Temozolomide chemotherapy. This finding further confirms the MGMT methylation as valuable theranostic marker to TMZ treatment (Table 3).

Correlations between genes methylation and molecular subtypes

GB with amplified EGFR gene presented a specific methylated genes panel. We found that 100 % of tumors with amplified EGFR (16 cases), were unmethylated in ATM, CD44, BRCA2 and VHL genes (Panel 1) with significant Chi² p values: 0.014, 0.009, 0.01 and 0.027 respectively. TP73, THBS1, GSTP1, and ESR1 genes (Panel 2) were highly unmethylated in amplified EGFR group (16 cases) (Table 1). Remarkably, more than 57 % of GB with amplified EGFR showed simultaneously unmethylated BCL2, BMIP3 and ID4 genes (Panel 3) that is significantly different from GB with normal EGFR (Chi² p values: 0.033, 0.001 and 0.028 respectively) but without influencing the OS (Table 1).

Survival correlation showed a significant difference in ATM unmethylation between GB with amplified EGFR and GB with normal EGFR (p=0.012). Significant results were obtained when analyzing the simultaneous unmethylation effect of the four genes ATM, BRCA2, CD44 and VHL on OS (p=0.005) (Table1).

Interestingly, when TP73, GSTP1, THBS1 and ESR1 (Panel 2) are all together unmethylated with the absence of EGFR amplification, the patients showed significant good overall survival (OS=15 months, p=0.001) (Table 1). Inversely, simultaneous methylation of the genes of Panel 2 in GB with amplified EGFR had the worst prognosis among other groups. (OS=3 months, p=0.041) (Table 1).

When analyzing the data using IDH mutation marker, we noticed that GB with IDH1 R132H mutation have the lowest percentage of methylation in tumor suppressor and MMR genes promoters (9 cases) (Table 2). Only four genes (HLTF, SFRP5, BRCA1 and CASP8) showed a significant distribution when methylated between wild type IDH1 and IDH1 R132H groups (Chi² p values: 0.002, 0.019, 0.048, 0.009, respectively) (Table 2).

96 % of wild type IDH1 GB subtype (25 out of 26 cases) were methylated in HLTF and SFRP5 genes. Meanwhile, 89 % of IDH1 R132H GB subtype were unmethylated for both genes. Significant Chi² p values were also reported: p=0.002 and p=0.019, respectively (Table 2).

Good overall survival was detected (OS=23 months, p=0.026) when the tumors were IDH1 R132H and unmethylated simultaneously in HLTF and SFRP5 genes (Table 2). Similarly, significant good OS was detected within IDH1 R132H and unmethylated BRCA1 group (OS=19 months, K-M p=0.002).

Prolonged survival GB+

The comparison of genes methylation distribution between GB+ and GB with low OS is presented in Table 4. We noticed that 80 % (4 out of 5) of GB+ samples were simultaneously methylated in promoters of RBM14 and PCCA genes. However, only one tumor presented this methylation from the GB group with low OS (Table 4).

Interestingly, all GB+ did not have the IDH1 mutation and were simultaneously unmethylated in BRCA1, SFRP5 and HTLF promoters (Supplementary material 2). 80 % of GB+ (4 cases out of 5) had a normal EGFR and were simultaneously unmethylated for ATM, BRCA1, BRCA2, CD44 and VHL genes (Supplementary material 2). The unique GB+ case with methylated promoter of ATM, BRCA1, BRCA2, CD44 and VHL genes had a lower OS (34 months) than the rest of GB+ group (OS from 36 to 48 months). The molecular and epigenetic subtyping according to patients’ survival is summarized in (Figure 2).

Figure 2:

Methylation status of genes panels according to GB subtypes. Glioblastomas are subtyped into 4 groups according to IDH and EGFR markers: Normal EGFR (Glioblastom with normal EGFR status), amplified EFGR (Glioblsatoma with amplified EGFR), WT IDH (Glioblastoma with wild type IDH), IDH1 R312H (Glioblastoma with IDH1 mutation), only simultaneous promoter methylation (yellow), or unmethylation (blue) for the gene panel are presented. Only genes with significant correlation K.M p values are presented with good prognosis group marker (green) and bad prognosis group marker (red).

Methylation genes and prognosis impact

Few reports demonstrated that ovarian tumors that are wild-type for BRCA1/2 and BRCA1 promoter methylation were associated with good patient outcome following therapy with platinum-taxane [26]. Methylation frequencies of BRCA1/2 varied considerably between CpG sites across the BRCA1 and BRCA2 promoters. CpG sites were more frequently methylated in sporadic carcinomas and were located outside the promoter region [27]. By contrast, in GB samples, BRCA methylation was weakly described. In a previous study of GB epigenetic description, BRCA1/2 were not cited [28].

According to the results of the present study, BRCA1 or BRCA2 were methylated in 20 and 14 % of GB respectively. Survival correlation indicated that the promoter unmethylation of these genes had a significant better prognostic in patients’ outcomes (mean OS=26 months, p=0.012 and p=0.008) (Table 3). The promoter CpG sites hypermethylation may lead to downregulation of tumor suppressor genes [11]. Our results confirm that promoter methylation of BRCA1/2 is involved in tumor aggressivity.

TP53 and TP73 genes were previously observed methylated in brain tumors especially in GB samples, whereas normal brains have been reported to be unmethylated for these two genes. Furthermore, within low grade glioma, the TP73 was found methylated in 39 % of oligodendroglioma [29]. Similarly, the promoter region of TP53 was found methylated in low grade astrocytoma, oligoastrocytoma and oligodendrogliomas [30].

In the current study, 25 and 37 % of analyzed GB tumors presented a methylated status in promoters of TP73 and TP53 respectively. Survival correlations indicated that when both genes are methylated, they give a negative impact on patient outcome (mean OS=9 and OS=10 months, p=0.002 and p=0.001, respectively).

These results could be explained by the protective role of the tumor suppressor protein TP53 during genotoxic stress [30]. Once methylated, the TP53 could not display its role of tumor suppressor and the cell misses its protection.

The gene MSH6 is an MMR gene that was detected methylated in 65 % of GB cases. The OS mean of patients with methylated MSH6 is 14 months, whereas it is 20 months within patients without MSH6 methylation. The survival correlation showed that, the MSH6 methylation is associated with poor patient prognosis (p=0.008). Our finding is similar to a previous study that investigated the MSH6 gene mutation and immunohistochemical MSH6 protein expression prior to and following treatment. It indicated that in recurrent gliomas, reduced MSH6 protein expression was detected in 50 % of tumors that received treatment [8]. Thus, methylation of MSH6 in GB is not a favorable prognosis.

The assessment of MGMT methylation is widely carried out in GB, as an important theranostic marker, to predict response to alkylating chemotherapy. Indeed, MGMT methylation is a good prognosis molecular marker [31].

In our study, MGMT methylation was assessed in all GB samples. Only 57 % were methylated. The survival correlation showed a better prognosis of patients with MGMT methylated (mean OS=29 months, p=0.046).

Molecular subtypes and methylation distinctive profiles

EGFR amplification subtype was investigated regarding tumors’ methylation profile. GB tumors group with normal EGFR status showed BRCA2, ATM, CD44 and VHL promoter methylation respectively in 57, 36, 21 and 29 % of cases. Meanwhile, no GB cases with amplified EGFR were found methylated in the cited genes.

Previous data suggested that EGFR amplification indicated a poor prognosis in glioma that regulate early DNA methylation [9], 17]. Moreover, EGFR hypermethylation is a typical event of secondary GB and tumors are more resistant to tyrosine kinase inhibitor [32]. Recently, ATM methylation was described in a subset of head and neck tumors [33].

In our results, 100 % of GB with amplified EGFR subtype showed no methylation in ATM, BRCA2, CD44 and VHL genes (OS=7 months, p=0.005). Our findings demonstrated that these genes were methylated only in subtype presenting normal EGFR status. Even in absence of EGFR amplification, the GB remain high grade and aggressive glioma (OS=13 months) when ATM, BRCA2 and CD44 are methylated. Thus, the reason for aggressivity in the tumor group ''normal-EGFR and methylated genes panel 1” is not only due to EGFR, but probably also due to an epigenetic event which is the methylation of ATM, CD44 and BRCA2 promoters. This event could occur in early stage of gliomagenesis. This hypothesis is supported by the relatively moderate OS registered in the group of “amplified EGFR and unmethylated genes panel 1” (OS=7 months, p=0.005).

Moreover, tumors with EGFR amplification and simultaneous methylation of genes panel 2 (TP73, THBS1, GSTP1 and ESR1) showed the lowest survival rate (OS=3 months, p=0.001).

In a previous study, the decreased expression of THBS1 has been observed in some tumors, including GB-multiform (GBM). Some cell lines were methylated at several CpG sites within the THBS1 5’CpG island, and had no detectable expression by RT- PCR [34]. THBS1 appears to be a potent anti-angiogenic factor by inhibiting endothelial cell adhesion, growth and motility. In vitro experiments showed that THBS1 has the ability to induce apoptosis in endothelial cells, and its down-regulation enhances angiogenesis [34]. This previous finding concurred with our results that showed the association of THBS1 methylation with poor survival GB.

Another study, using the multi-dimensional GBM-associated data sets analysis containing DNA mutation, copy number, DNA methylation, mRNA and miRNA expression, showed that ESR1 was connected by the largest number of member genes [35]. The high expression of ESR1 was also associated with favorable survival in TCGA GBM data set [35]. Other previous studies have demonstrated that the encoded protein of ESR1 was involved in pathological processes of multiple cancers [35], 36]. Uhlmannet et al. discovered a putative association between ESR1 and gliomagenesis [37]. These findings further support our results of poor survival GB group with ESR1 methylation (OS=3 months, p=0.001).

However, methylation of genes panel 3 in amplified EGFR group resulted in a poor prognosis, and we were not able to find a significant statistical correlation with the OS. This insignificant correlation could be due to the relatively small number of studied tumors.

It was previously confirmed that EGFR status is a prognostic marker [5]. However, according to our findings, the combination of EGFR status and tumor suppresser genes (ATM, CD44, BRCA2, VHL, TP73, THBS1, GSTP1 and ESR1) methylation profile will add more precision to the prognosis evaluation.

In sum, combined unmethylation of BRCA2, ATM, CD44 and VHL genes in normal EGFR GB group will lead to a better prognosis (OS=15, p=0.005). Meanwhile, combined methylation of TP73, THBS1, GSTP1 and ESR1 genes in amplified EGFR GB group will result in a poor prognosis (OS=3 months, p=0.001).

IDH1-R132H subtype was also investigated and correlated with methylation profile. IDH1-R132H mutant GB samples have the lowest methylation ratio in all analyzed genes (65 genes promoter). Only 10–25 % of GB were methylated in the genes BRCA1, CASP8, HLTF and SFRP5. This result indicates that the good prognosis is not only conferred by IDH mutation, but also by MMR gene activation. Whereas, another study showed that IDH-mutant promotes DNA hypermethylation, and changes the methylome [38]. In our study, we focused only on MMR genes specifically in GB and not all glioma methylome.

Similarly, CASP8 expression is retained in GB, suggesting that it may support cancer growth [39]. A previous report has shown that among the 66 glioma samples, 52 samples exhibited CASP8 gene methylation. CASP8 silencing was significantly associated with the high-grade glioma [40]. In our results, 44 % of wild-type-IDH GB were methylated for CASP8 gene. This result explains the low mean survival of wild-type-IDH in comparison with IDH1-R132H. The CASP8 methylation or inactivation will inhibit the apoptotic mechanisms and therefore therapeutic resistance. Meanwhile, in the opposite GB group with IDH1-R132H mutation that showed good patient survival, CASP8 was unmethylated in 87 % of analyzed tumors. This result emphasizes more the good prognosis of IDH1-R132H GB with active CASP8 gene. However, due to the small number of tumors with IDH1-R132H and CASP8 methylated promoter (7 cases) no significant correlation was found.

33 % of IDH1-R132H mutant tumors presented a methylated promoter in BRCA1 gene. BRCA1 unmethylation in IDH1 groups was significantly associated with good prognosis (p=0.002). From another angle, BRCA1 methylation was significantly associated with poor prognosis (OS=6 months, p=0.002) in wild-type IDH1 subtype.

Moreover, the combined unmethylation of HLTF and SFRP5 genes showed a better survival within IDH groups (p=0.002). By contrast, simultaneous methylation of HLTF and SFRP5 genes was correlated with GB wild type IDH1 group defining therefore a poor prognosis group (OS=1 month, p=0.026). This result could be explained by a specific pathway driving the tumors to GB stage.

Previous reports proved the implication of Wnt pathway proteins such as SFRP1 and SFRP2 commonly altered in GB. However, distinct patterns of hypermethylated Wnt pathway inhibitor genes (SFRP1 and SFRP2) were observed between primary GB and low-grade glioma. Nevertheless, SFRP5 methylation was rarely described [41]. The HLTF gene methylation is reported as a poor prognosis marker in colorectal cancer and associated with tumor metastasis [42]. HLTF is a tumor suppressor gene. In vitro experiments showed that the downregulation of HLTF is associated with tumor cell invasion. Furthermore, HLTF expression is associated with better patient survival in colorectal cancer [42].

Methylation of the PCCA and RBM14 promoters has been found correlated with GB+ who did not present any specific molecular profile with copy number variation in a previous study [17]. In the current study, the RBM14 and PCCA were detected co-methylated in 80 % of GB+ samples (Chi²-test, p=0.005).

RBM14 acts as DNA repairing gene that belongs to non-homologous end joining pathway. In a previous study, RBM14 has been reported to promote radio-resistance in GBM by regulating DNA repair and cell differentiation [43]. When RBM14 is methylated, the tumors could lose the radio-resistance and become more sensitive to the treatment. Moreover, it was noticed that the knock-down of RBM14 blocks GB regrowth after irradiation in vivo [43]. Targeting the RBM14 dependent pathway may prevent recurrence of tumors and eradicate the disease.

Moreover, PCCA gene has been found in some studies as a contributor to tumors pathogenicity. Indeed, high microsatellites instability was associated with frame shift mutation in PCCA gene leading to further tumor pathogenicity in colorectal cancer. Once PCCA is expressed, it contributes to amino acid nucleotide metabolic pathway that is involved in cancer pathogenicity [44]. However, it was reported in another paper that PCCA gene was methylated in high grade non muscle invasive bladder cancer without being able to demonstrate its effect to immunotherapy sensitivity due to lack of samples [45]. Meanwhile, our result indicated that silencing of PCCA by methylation could provide a better prognosis. The correlation becomes more interesting when we associate both markers. Never described before, this simultaneous methylation of PCCA and RBM14 was found in 80 % of GB+ (p=0.005).

However, all GB+ tumors were wild-type IDH1, which is considered as bad prognostic marker. They presented also, unmethylation of HLTF, SFRP5 and BRCA1 genes. In addition, 4 tumors out of 5 showed normal EGFR and simultaneous unmethylation of BRCA1, BRCA2, ATM, VHL and CD44 genes.

The genes methylation profile of GB+ explains well their prolonged survival. Indeed, in the absence of EGFR amplification and with the activation of tumor suppresser genes (BRCA1, BRCA2, ATM, VHL, CD44, HLTF and SFRP5), GB+ tumors seem to respond better to treatment, avoid the relapse and confer prolonged survival (more than 34 months).

Meanwhile, the first limitation of the current study is the small sample size of tumor specimens. To address it, future studies should include a larger samples cohort. To overcome the second limitation of absence of normal brain tissue future studies should incorporate it in comparative molecular profile analysis or AI-driven molecular profiling of normal brain tissue [24].

To sum up our findings, MGMT methylation is associated with a good prognosis. Interestingly, most of these genes’ methylation status was found in correlation with prognostic markers (EGFR amplification and IDH1-R132H).

Our results showed that methylation of BRCA2, ATM, CD44 and VHL genes on the one hand and TP73, THBS1, GSTP1 and ESR1 genes on the other, are associated with EGFR molecular GB subtype. They seem to play an important role in GB progression and aggressivity.

Our findings highlighted the association between activation of HLTF and SFRP5 and the good prognosis in GB subtype with IDH1 R132H mutation. The same finding was noticed for BRCA1 gene.