Dissecting the Methylomes of EGFR-Amplified Glioblastoma Reveals Altered DNA Replication and Packaging, and Chromatin and Gene Silencing Pathways

Simple Summary Glioblastoma is the most malignant brain tumor. To date, there is no curative therapy available. Since EGFR is an interesting candidate in precision medicine, we performed an integrated molecular analysis in glioblastoma with and without EGFR amplification. We found that there are significant molecular differences in glioblastoma, depending on the EGFR amplification state. Analysis of top differences revealed DNA replication and packaging, and chromatin and gene silencing pathways. Targeting these altered pathways may open novel targets in precision medicine. Abstract Glioblastoma IDH wildtype is the most frequent brain tumor in adults. It shows a highly malignant behavior and devastating outcomes. To date, there is still no targeted therapy available; thus, patients’ median survival is limited to 12–15 months. Epithelial growth factor receptor (EGFR) is an interesting targetable candidate in advanced precision medicine for brain tumor patients. In this study, we performed integrated epigenome-wide DNA-methylation profiling of 866,895 methylation specific sites in 50 glioblastoma IDH wildtype samples, comparing EGFR amplified and non-amplified glioblastomas. We found 9849 significantly differentially methylated CpGs (DMCGs) with Δβ ≥ 0.1 and p-value < 0.05 in EGFR amplified, compared to EGFR non-amplified glioblastomas. Of these DMCGs, 2380 were annotated with tiling (2090), promoter (117), gene (69) and CpG islands (104); 7460 are located at other loci. Interestingly, the list of differentially methylated genes allocated eleven functionally relevant RNAs: five miRNAs (miR1180, miR1255B1, miR126, miR128-2, miR3125), two long non-coding RNAs (LINC00474, LINC01091), and four antisense RNAs (EPN2-AS1, MNX1-AS2, NKX2-2-AS1, WWTR1-AS1). Gene ontology (GO) analysis showed enrichment of “DNA replication-dependent nucleosome assembly”, “chromatin silencing at rDNA”, “regulation of gene silencing by miRNA”, “DNA packaging”, “posttranscriptional gene silencing”, “gene silencing by RNA”, “negative regulation of gene expression, epigenetic”, “regulation of gene silencing”, “protein-DNA complex subunit organization”, and “DNA replication-independent nucleosome organization” pathways being hypomethylated in EGFR amplified glioblastomas. In summary, dissecting the methylomes of EGFR amplified and non-amplified glioblastomas revealed altered DNA replication, DNA packaging, chromatin silencing and gene silencing pathways, opening potential novel targets for future precision medicine.

Focusing on gliomas of adults being allocated to astrocytomas, oligodendrogliomas, and glioblastomas, the main molecular hallmarks are as follows [1]: astrocytomas harbor isocitrate dehydrogenase (IDH) 1/2 mutations; and oligodendrogliomas harbor loss of 1p and 19q (1p/19q) combined with IDH1/2 mutations. Diffuse gliomas with IDH1/2, H3F3A (Histone H3 Family 3A) and HIST1H3B/C wildtype status are classified as glioblastoma [1]. Thereby, the degree of anaplasia is represented by the addition of a CNS WHO grade, depending on the degree of anaplasia, i.e., the presence of mitoses, microvascular proliferation, and necrosis: astrocytomas IDH mutated are graded as CNS WHO grade 2 to 4, oligodendrogliomas IDH mutated 1p/19q co-deleted are graded as CNS WHO grade 2 to 3, and glioblastomas IDH wildtype are graded as CNS WHO grade 4 [1].
Of all tumors of the CNS, glioblastomas IDH wildtype are the most malignant brain tumors in adults [1]. With 3-4 reported cases per 100,000 population in Western world countries, glioblastomas are also the most frequently diagnosed primary brain tumor [1]. The discovery of the methylation of the O6-methylguanin-DNA-methyltransferase (MGMT) promoter in 2008 opened up new possibilities in glioblastoma therapy [2][3][4]: hypermethylation of the MGMT promoter is associated with significantly longer survival in patients that receive an adjuvant radio-chemotherapy with temozolomide, according to the EORTC/NCIC protocol [5]. However, besides the addition of adjuvant tumor treating field therapy, to date, no further significant progresses in glioblastoma therapy has been made [1].
A promising novel molecular target for individualized patient care is the amplification of epidermal growth factor receptor (EGFR) [6]. EGFR is a receptor tyrosine kinase [7] contributing to cell proliferation and differentiation [7,8], bound to the epidermal growth factor (EGF) and other growth factors [9,10]. Genomic alterations of EGFR result in signaling independent of physiological ligands propagating pro-oncogenic processes [10]. In case of EGFR amplifications, there is a direct impact on protein levels with overexpressed EGFR protein [11]. Thus, detection of EGFR amplification, and subsequently altered pathways offers novel targets in molecular glioblastoma therapy.
Here, we performed an integrated analysis of 50 glioblastomas IDH wildtype with or without EGFR gene amplification, and performed epigenome-wide methylation analysis to reveal distinct differences in the glioblastoma methylome and subsequently affected pathways.

Tissue Collection
We analyzed 50 anonymized glioblastoma IDH wildtype CNS WHO grade 4 (Supplementary Table S1). Integrated diagnosis and re-classification were performed according to the 2021 CNS WHO classification including histology, immunohistochemistry, and molecular genetics [1]. We used formalin-fixed and paraffin-embedded (FFPE) tissue samples that underwent routine histological analysis and, subsequently, were stored in the archive of the University Institute of Pathology, University Hospital Salzburg, Paracelsus Medical University. Prior to study inclusion, samples were anonymized (non-identifiable samples).

Molecular Genetic Analysis
Molecular genetic analyses were performed as previously described [12][13][14][15]. Molecular genetic testing was performed according to the 2021 CNS WHO classification [1]. Thereby, essential diagnostic criteria for glioblastoma IDH wildtype are an IDH wildtype H3 wildtype diffuse astrocytic glioma, with one or more of the following criteria: microvascular proliferation, necrosis, TERT promoter mutation, EGFR gene amplification, and/or +7/−10 chromosome copy-number alterations. As further desirable criteria, there is a DNA methylation profile of glioblastoma IDH wildtype. Thus, in the cases with TERT promoter wild type status, or without information on TERT promoter mutations status (NA), diagnosis of glioblastoma IDH wildtype was assessed by applying the essential criteria of IDH wildtype H3 wildtype diffuse astrocytic glioma with the morphologic features of microvascular proliferation and necrosis, and the desirable criteria of a DNA methylation profile of glioblastoma IDH wildtype were applied. Due to this, we did not perform further molecular analyses on TERT promoter mutations in cases with unknown TERT status (4 cases), and we did not perform +7/−10 chromosome copy-number alteration analyses in cases with TERT wild type status (3 cases) (Supplementary Table S1).
Raw data files (idat-files) of Infinium Methylation Bead Chips were analyzed using the current brain tumor classifier of the molecularneuropathology.org bioinformatics pipeline [16]. Amplifications of copy-number variation (CNV) plots were interpreted as intensities of more than 0.6 on the log2-scale after baseline correction [14,17].

Computational Data Analysis
We used the Genome Studio (Illumina, San Diego, CA, USA) and RnBeads [18] for methylation analysis, as previously described with identical analysis parameters [19]. Analogous to other epigenome-wide association studies (EWAS), we used nominal pvalues [20,21]. Further data visualization was performed applying GraphPad Prism 9 software suite and Morpheus.

A Subfraction of Glioblastomas IDH Wildtype CNS WHO Grade 4 Shows EGFR Gene Amplification
We performed an integrated analysis on glioblastomas IDH wildtype CNS WHO grade 4 that were processed in routine diagnostic workup. Analyses of copy number variation (CNV) plots that were checked routinely during sample workup (e.g., 1p and 19q losses) found that a subfraction of glioblastomas showed EGFR amplification. For this study, we subsequently selected 50 glioblastomas with known EGFR amplification status: 25 with amplified EGFR gene region (Figure 1a,b), and 25 without amplified gene region (Figure 1c,d). Thereby, intensities of more than 0.6 on the log2-scale after baseline correction were interpreted as EGFR amplifications, the correction of the baseline of chromosome 7 enabled distinguishing between chromosome 7 polysomy and EGFR amplification, as described by Stichel et al. [17].  Of all 50 glioblastomas (Figure 1e), the mean EGFR probe intensity levels (reflecting EGFR gene amplification) of amplified glioblastomas was 0.99, while the mean EGFR probe intensity levels of non-amplified glioblastomas was 0.04 (Figure 1f). These samples were subsequently processed on epigenome-wide DNA-methylome analysis. Analysis on molecular glioblastoma subtypes (mesenchymal, RTK I, RTK II) showed that all three molecular subtypes were evenly distributed in the analyzed cohort (Supplementary Figure S1a); however, there was a predominance of RTK II subtype in EGFR amplified glioblastomas (60%, Supplementary Figure S1b) and a predominance of RTK I subtype in EGFR nonamplified glioblastomas (48%, Supplementary Figure S1c). Further CNV alterations were not analyzed in the current study.

Differential Methylation Analysis of EGFR Amplified Glioblastomas IDH Wildtype CNS WHO Grade 4 Shows Distinct Epigenomic Alterations
By using the Illumina Infinium EPIC bead chip, we interrogated 866,895 probes per sample. During data preprocessing, we removed 17,371 probes overlapping with SNPs, 7508 probes during greedy cut filtering, 2937 context specific probes, and 18,695 probes located on sex-chromosomes; of the remaining 820,384 probes, 245,013 were annotated with tiling regions, 43,306 were annotated with promoters, 33,801 were annotated with genes, and 25,763 were annotated with CpG islands (Figure 2a). Beta value distributions showed high densities of unmethylated methylation values in CpG islands (Figure 2b). Generation of a volcano plot showed the distribution of differential methylation values (∆β values) and significance level (−log10); probes with ∆β values of ≥0.1 and p-value < 0.05 are indicated in red color (Figure 2c). Subsequent hierarchical clustering of the top 100 differentially methylated CpGs (DMCG) showed distinct clusters in the generated heat map using Manhattan distance (Figure 2d). Each pairwise comparison of probes resulted in 9849 significantly differentially methylated probes, with ∆β values of ≥0.1 and p-value < 0.05 ( Figure 2e); thereby, 5178 probes were hypermethylated and 4671 probes were hypomethylated (Figure 2f).

Discussion
Targeting EGFR overexpression may be a promising novel approach in glioblastoma therapy, since small molecule tyrosine kinase inhibitors (TKIs) and monoclonal antibodies have already been approved in other tumor entities [6].
Interestingly, the list of top significant differentially methylated genes reveals that the top hypo-and hypermethylated genes are both functionally important RNAs (Supplementary Table S2). The top position of hypermethylated genes in EGFR non-amplified glioblastomas (i.e., hypomethylated in EGFR amplified glioblastomas) is micro-RNA miR128-2, and the top position of hypomethylated genes in EGFR non-amplified glioblastomas (i.e., hypermethylated in EGFR amplified glioblastomas) is antisense RNA NKX2-2-AS1 (Supplementary  Table S2).
According to Budi et al., miR128 is involved in cancer development, progression and therapy response [22]. Thereby, miR128 is associated with many different cellular processes, such as cell proliferation and self-renewal, chemotherapeutic resistance, cell growth and differentiation, angiogenesis and senescence, and apoptosis [22]. Decreased expression of miR128 has been described in a broad range of cancers, such as breast cancer, lung cancer, thyroid cancer, laryngeal cancer, head and neck cancer, melanoma, osteosarcoma, leukemia, and glioma [22]. Shi et al. found that overexpression of miR128 inhibits tumor growth and neovascularization through the miR128/p70S6K1 pathway [23]. Ye et al. found that miR128 promotes apoptosis, and that downregulation of miR128 expression reduced glioma cell death [24].
NKX2.2 is also well known in the context of other tumor types, such as colorectal cancer, lung cancer, neuroendocrine tumors, osteosarcomas, Ewing sarcomas and brain tumors [25].
Cao et al. found that NKX2-2-AS1 was highly expressed in Ewing sarcomas, functioning as a core co-regulator of the driver mutation fusion gene, and is recognized as a specific biomarker for Ewing sarcomas [26]. Muraguchi et al. described that NKX2-2 suppresses the self-renewal of glioma-initiating cells [27]. In glioma models, downregulation of NKX2.2 was correlated with increased tumor malignancy and accelerated tumor growth and progression [27].
As a limitation of this study, the results were not validated in an independent cohort or using published data, such as TCGA data. This should be added in additional analyses, to increase the validity of results. Furthermore, correlation of methylation data with gene expression should be added in further studies to evaluate the biological relevance of methylation differences. Furthermore, in the current study, we used nominal p-values for analysis and did used p-values with correction for multiple testing; this is in accordance with other epigenome-wide association studies (EWAS) [20,21].
In summary, dissecting the methylomes of EGFR amplified and non-amplified glioblastomas revealed altered DNA replication, DNA packaging, chromatin silencing and gene silencing pathways, opening potential novel targets and targetable pathways for future precision medicine.

Conclusions
Despite intensive research, there still is no curative therapy for glioblastoma patients available. Since EGFR amplification is a promising novel candidate in precision medicine, we performed an epigenome-wide DNA-methylation analysis of EGFR amplified and nonamplified glioblastomas. We identified distinct DMCGs, depending on EGFR amplification status. Interestingly, the top DMCGs showed an overrepresentation of functionally relevant RNAs, and GO analysis showed enrichment of DNA replication, DNA packaging, chromatin silencing, and gene silencing pathways being hypomethylated in EGFR amplified glioblastomas.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/cancers15133525/s1, Figure S1: Analysis on molecular glioblastoma subtypes; Table S1: Patient details; Table S2: Top significant differentially methylated genes; Table S3: GO enrichment analysis of top hypomethylated and hypermethylated pathways in EGFR amplified glioblastomas.  Institutional Review Board Statement: All procedures performed in this study were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards, and approved by the Institutional Review Board and Ethics Committee of the district of Salzburg (415-E/2509/2-2019, 24 April 2019), all samples were anonymized prior to study inclusion (non-identifiable samples).

Informed Consent Statement: Not applicable.
Data Availability Statement: Data is available from the corresponding author on request.

Conflicts of Interest:
The authors declare no conflict of interest.