Clinical insights gained by refining the 2016 WHO classification of diffuse gliomas with: EGFR amplification, TERT mutations, PTEN deletion and MGMT methylation

A dataset of adult diffuse glioma samples was obtained from patients diagnosed from 2011 to 2016, in Unidade de Investigação em Patobiologia Molecular of Instituto Português de Oncologia de Lisboa Francisco Gentil (IPOLFG). In this study we included all consecutive glioma patients referred for treatment in our center, previously submitted to surgery, with known histologic diagnosis and biological material available. This study was previously approved by the IPOLFG Ethical Board Committee. Four hundred forty-four glioma samples were reclassified according to the 2016 WHO classification. However, statistical analysis was performed using only 403 samples, due to the exclusion of the NOS (Not Otherwise Specified) glioma group. These samples were previously characterized in the diagnostic routine for: IDH mutations and TERT promoter mutations by Sanger Sequencing. MGMT promoter methylation was determined by Multiplex Ligation-dependent Probe Amplification (MLPA – MRC-Holland according with guidelines defined by van den Bent [30]). PTEN deletion (Vysis PTEN/CEP10), EGFR amplification (Vysis EGFR/CEP7) and 1p/19q codeletion (Vysis,1p36/1q25 and 19q13/19.13 dual color probe) were identified by Fluorescent in situ hybridization (FISH). The definition of numerical alterations was performed according the FISH criteria defined by the International System of Human Cytogenetic Nomenclature (ISCN) 2016 [31].

For samples without the mutational status of IDH1 (n = 92) and TERT promoter mutations (n = 82), we extracted DNA, when not available from the routine diagnosis, which was used at a concentration of 80 ng/μl. Tumor samples were received as fresh tissue or paraffin-preserved tissue for DNA extraction. The tumor sections for DNA extraction were selected by a neuropathologist consider the following criteria: 1) cellular regions without necrosis, 2) representative regions of tumor subtype and 3) areas with a minimum of 2 mm of diameter. The DNA extraction from frozen tissues was performed using the conventional method of phenol-chloroform (MERCK, Germany). From tissues fixed in formaldehyde and preserved in paraffin, the DNA was isolated using the QIAGEN’s Gene Read™ DNA FFPE Kit. Additionally, for some samples included in this project the DNA was extracted using an automatized process by Maxwell® RSC Instrument (Promega, USA), using the RSC DNA FFPE kit (Promega, USA). The extraction was performed according to the manufacture’s protocol. The DNA concentration and quality were assessed using Nanodrop 2000 (Thermo Fisher Scientific, USA).

The mutational analysis directed to exon 4 of IDH1 and TERT promoter was performed using two sets of primers for the detection of hotspot mutations: missense mutations involving a single amino acid change at arginine 132 (R132) of IDH1 and C228T and C250T map − 124 and − 146 bp upstream of TERT ATG site. The target amplification of IDH1 was achieved using the forward primer 5′ CGGTCTTCAGAGAAGCCATT 3′ and the reverse primer 5′ GCAAAATCACATTATTGCCAAC3’ and TERT promoter was amplified using the forward primer 5′ GCACAGACGCCCAGGACCGCGCT 3′ and the reverse primer 5′ TTCCCACGTGCGCAGCAGGACGCA 3′ generating fragments with 129 bp and 196 bp respectively. PCR contained 35 cycles with annealing at 56 °C for IDH1 and 69.5 °C for TERT promoter. Then, an enzymatic method was used to purify each PCR product, using two distinct enzymes: Exonuclease I 20 U/μl (Thermo Fisher Scientific, USA) and FastAP Thermosensitive Alkaline Phosphatase 1 U/μl (Thermo Fisher Scientific, USA). To determine the sequence of interest in IDH1 and TERT promoter gene, an automatic sequencer was used, ABI PrismTM 3130 Genetic Analyser (Applied Biosystems, USA) following the protocol purposed by Big Dye™ Terminator v1.1 Cycle Sequencing Kit (Applied Biosystems, USA).

The primary endpoint was overall survival, defined as the time from the glioma diagnosis to the patient death or last follow up. Survival analysis was done using Kaplan-Meier estimator and the log-rank test for group comparison. Variables with a significant p-value in the univariate analysis were exposed to a multivariate analysis using Cox regression proportional hazard model. The multivariate analysis allowed to study the independent association of the molecular subgroups established with overall survival, while controlling for potential confounders such as age, sex and treatment. In order to eliminate confounder variables, the number of cases of each subtype was reduced because the type of treatment was not accessible for all the cases included in the cohort. Additionally, to evaluate the association between the interest biomarkers and the overall survival was performed a multivariate analysis controlling for: MGMT methylation, PTEN deletion and EGFR amplification. TERT promoter mutations were excluded from this analysis, since the number of samples would reduce the dataset available to determine the impact of the remaining biomarkers.

All tests were two-sided, and we considered a significance level of 5%. The statistical analysis applied here was performed using IBM SPSS Statistics 21.0.

The reorganization of diffuse gliomas according to 2016 WHO classification mainly affected oligodendroglioma and astrocytoma subgroups, reducing the number of oligodendrogliomas (82 to 49) and increasing the astrocytomas (49 to 98), while the number of GBM remained the same (Table 1). Additionally, 41 samples were not included in any glioma subset (Glioma NOS), mainly due to: technical issues, samples with only 1p or 19q deletion or with 1p/19q codeletion and IDH-wildtype. Importantly, in this new classification the subgroup of oligoastrocytomas was reorganized between astrocytomas, oligodendrogliomas or NOS.

This molecular reclassification allowed the division of diffuse gliomas into 6 molecular subgroups according to IDH mutations and 1p/19q codeletion analysis (Table 2). GBM samples corresponded to 57.7% of the entire cohort, of which 55.2% were GBM IDH-wildtype and the remaining 2.5% were GBM IDH-mutant. The astrocytoma subgroup was the second most frequent (22.1%) and the NOS glioma subgroup corresponded to 9.2% of the cohort analyzed (Table 2). IDH-mutant gliomas, whether GBMs or astrocytomas, are predominant in young patients (44 and 38 years respectively) in comparison with IDH-wildtype gliomas (63 and 57 years respectively) (Table 3). In addition, our results also indicated a higher prevalence of GBM IDH-wildtype, astrocytomas IDH-wildtype and IDH-mutant and 1p/19 codeleted gliomas in men (2:1; 1.3:1; 1.4:1 and 2.1 respectively), except for the GBM IDH-mutant subgroup (0.8:1) (Table 3).

Then, we evaluated the prognostic value of histological grade and molecular subgroups. Grade II oligodendroglioma was the histological subgroup associated with longer overall survival (Median Overall Survival (OS):172 months); in contrast, GBM (grade IV) was the subgroup associated with the poorer outcome (OS:11 months) (Fig. 1). Using the molecular classification, gliomas with 1p/19q codeletion were the subgroup associated with better prognosis (OS:198 months). For the remaining molecular subgroups (GBM and astrocytomas), IDH-mutant tumors were associated with better prognosis (OS:25 and 114 months, respectively) when compared to the IDH-wildtype subgroups (OS:10 and 14 months, respectively) (Fig. 1).

The multivariate analysis performed using Cox Regression Hazard model evidenced the prognostic impact of the molecular subgroups and histological grades, after adjustment for age, gender and treatment (Fig. 1). The GBM IDH-mutant subgroup was the only group that did not show a statistically significant p-value (p-value = 0.092), perhaps due to the reduced number of samples (n = 10). These results validated the size and representativeness of each glioma molecular subgroup to perform further studies, as well as, the accuracy introduced by molecular markers in the prognosis and diagnosis of patients with gliomas.

Following the reclassification of gliomas according to the 2016 WHO classification, we investigated the role of TERT promoter mutations, EGFR amplification, PTEN deletion and MGMT promoter methylation in molecular glioma subgroups. Here, we intended to assess whether these molecular alterations are predominantly altered in a specific subgroup, which could help to redefine the established molecular subgroups of gliomas. EGFR amplification was more frequently detected in IDH-wildtype gliomas, both, GBM (38%) and astrocytomas (43%), compared to IDH-mutant gliomas (11 and 4%, respectively). This molecular alteration was absent from the 1p/19q codeleted glioma subgroup (0%). PTEN deletions were identified in 83% of GBM IDH-wildtype, the most aggressive glioma group (Fig. 2), characterized by an OS of 10 months (Fig. 1). However, these alterations were also found in 43 and 50% of GBM IDH-mutant (OS:25 months) and astrocytomas IDH-wildtype (OS:14 months) respectively. In 1p/19q codeleted gliomas and astrocytomas IDH-mutant (OS:198 and 114 months, respectively), the two less aggressive subtypes of gliomas, the incidence of PTEN deletion was reduced (8 and 21% respectively). These results suggested that PTEN deletions are predominantly found in the most aggressive subgroups of gliomas. TERT promoter mutations were mainly found in 1p/19q codeleted (94%) and GBM IDH-wildtype (88%) molecular subgroups (Fig. 2), suggesting that this is not a good biomarker for diagnosis. Regarding MGMT promoter methylation status, we observed that: 100% of 1p/19q codeleted gliomas, 91% of astrocytomas IDH-mutant and 50% of GBM IDH-mutant samples were methylated. Therefore, MGMT promoter methylation, as expected, was inversely associated with aggressiveness, since it appears most frequently in groups with better prognosis (Fig. 2).

Furthermore, we evaluated the prognostic value of these distinct genetic alterations in each molecular subgroup of gliomas. We verified that EGFR amplification did not have significant impact in the overall survival of patients with GBM IDH-wildtype and astrocytoma IDH-wildtype, (p = 0.393 and p = 0.522, respectively) (Fig. 3a and b), as well as, TERT promoter mutations in GBM IDH-wildtype (p = 0.605), although the number of cases was too small for conclusive results (Fig. 3c).

However, as shown in Fig. 3d, MGMT promoter methylation was significantly associated with a prolonged overall survival in GBM IDH-wildtype (p = 0.009), while in astrocytomas IDH-wildtype (Fig. 3e) had no impact in prognosis (p = 0.555). The effect of this biomarker in astrocytomas IDH-wildtype could be related with the low number of methylated samples (n = 9).

Surprisingly, PTEN deletion had a dual effect in the prognosis of GBM and Astrocytomas IDH-wildtype. This molecular alteration was a favorable prognostic for GBM IDH-wildtype (p = 0.042) and a unfavourable prognostic for astrocytoma IDH-wildtype (p = 0.015) (Fig. 3f and g). Moreover, in astrocytomas IDH-mutant, PTEN deletion was not found to have a significant impact on overall survival (p = 0.702) (Fig. 3h). The multivariate analysis, considering EGFR amplification, PTEN deletion and MGMT methylation and controlling for age, gender and treatment, validated the role of PTEN deletion (Hazard Ratio (HR) =0.65; 95% CI 0.43–0.99) and MGMT promoter methylation (HR = 0.61; 95% CI 0.42–0.88) as independent factors of prognosis in GBM IDH-wildtype. In addition, also confirmed the role of PTEN deletion as a prognostic factor of poor outcome (HR = 4.48; 95% CI 1.34–14.94) (Table 4).

To gain further insight into the predictive value of these biomarkers in molecular subgroups of gliomas, we analyzed the effect of EGFR amplification, PTEN deletion and MGMT promoter methylation in the response to therapy using the only group with representative samples - GBM IDH-wildtype (Fig. 4). However, in this group due to the small number of TERT wildtype samples, it was not possible to do this analysis regardless TERT promoter mutations.

Initially, we analyzed the OS of each group of patients treated with radiotherapy (RT) or chemo-radiotherapy (CRT), which was 6 and 16 months, respectively (Fig. 4a). CRT based on temozolomide is the standard treatment for patients with GBM. Patients subjected to RT alone, usually respect the following criteria: age above 70 years, other pathological conditions contra-indicating chemotherapy or a more severe clinical presentation. Patients who were not treated with RT or CRT, were directly to palliative care.

In patients with GBM IDH-wildtype, surprisingly the EGFR amplification was associated with a better response to radiotherapy (p = 0.011) (Fig. 4b left), however it was unable to predict the response to chemo-radiotherapy (p = 0.596) (Fig. 4c right). The multivariate analysis performed in RT subgroup, considering the three genetic alterations and controlling for age and gender revealed that EGFR amplification constitutes an independent predictive factor of response to radiotherapy (HR = 0.56; 95% CI 0.36–0.88) (Table 5). This result suggests a new putative strategy for the management of patients, who may have a better response to radiotherapy, although it should be validated in other cohorts.

The effect of PTEN deletion on response to therapy was inferred only in patients exposed to radiotherapy, since most patients submitted to chemo-radiotherapy had PTEN deleted. According to our results, PTEN deletion had no predictive value in the response to radiotherapy (p = 0.258) (Fig. 4b middle).

Additionally, our results showed that MGMT methylated samples were associated with an improved response to chemo-radiotherapy compared to MGMT unmethylated samples in GBM IDH-wildtype patients (p = 0.003) (Fig. 4c right). MGMT methylation constitutes a well-known predictive biomarker of gliomas used to infer which patients would have a better response to chemotherapy with temozolomide. Despite of its effects in response to chemo-radiotherapy, as described before [25], MGMT was not an important predictor of response to radiotherapy alone (p = 0.733) (Fig. 4b right).