DNA methylation analysis with methylation‐sensitive high‐resolution melting (MS‐HRM) reveals gene panel for glioma characteristics

Abstract Introduction Local DNA hypermethylation is a potential source of cancer biomarkers. While the evaluation of single gene methylation has limited value, their selected panel may provide better information. Aims This study aimed to analyze the promoter methylation level in a 7‐gene panel in brain tumors and verifies the usefulness of methylation‐sensitive high‐resolution melting (MS‐HRM) for this purpose. Methods Forty‐six glioma samples and one non‐neoplastic brain sample were analyzed by MS‐HRM in terms of SFRP1, SFRP2, RUNX3, CBLN4, INA, MGMT, and RASSF1A promoter methylation. The results were correlated with patients’ clinicopathological features. Results DNA methylation level of all analyzed genes was significantly higher in brain tumor samples as compared to non‐neoplastic brain and commercial, unmethylated DNA control. RASSF1A was the most frequently methylated gene, with statistically significant differences depending on the tumor WHO grade. Higher MGMT methylation levels were observed in females, whereas the levels of SFRP1 and INA promoter methylation significantly increased with patients’ age. A positive correlation of promoter methylation levels was observed between pairs of genes, for example, CBLN4 and INA or MGMT and RASSF1A. Conclusions Our 7‐gene panel of promoter methylation can be helpful in brain tumor diagnosis or characterization, and MS‐HRM is a suitable method for its analysis.


| INTRODUC TI ON
Moreover, DNA methylation analysis can also serve as a prognostic indicator. In this context, G-CIMP (Glioma CpG island methylator phenotype)-positive phenotype correlates with the presence of IDH mutation and is associated with a good prognosis in GBM (glioblastoma multiforme). 3 Eventually, methylation biomarkers can also predict the response to specific therapy and guide patients' treatment. Multiple large-scale clinical studies proved that GBM patients with O 6 -methylguanine-DNA methyltransferase (MGMT) promoter hypermethylation benefit from alkylating agent therapy, especially temozolomide (TMZ). [4][5][6][7] Thus, MGMT is now the most frequently used methylation-associated biomarker in neuro-oncology. However, to date, there is no consensus on the optimal method for the detection of MGMT as well as other DNA methylation biomarkers in the clinical setting. 8 There is a variety of analytical methods for DNA methylation analysis, allowing both whole-genome methylation profiling and locus-specific DNA methylation assays. The first approach mentioned is often used in the discovery phase allowing the identification of differentially methylated regions and their relevance to the disease.
On the other hand, the second approach-locus-specific DNA methylation analysis, is an optimal technique in the clinics, since most of the established biomarkers rely on DNA methylation differences in the limited number of CpG dinucleotides. 9 Overall, the optimal DNA methylation detection method should be sensitive, specific, quick, cost-effective, and suitable for screening of large sets of clinical samples. 10 Methylation-sensitive high-resolution melting (MS-HRM) has all of these features. [11][12][13] In this method, establishing the DNA methylation level of a particular sequence is based on the differences in the melting profiles of its PCR amplicon. Bisulfite treatment, preceding PCR amplification, creates the difference in the nucleotide sequence corresponding to the presence or absence of methyl groups. In this regard, unmethylated cytosine is oxidatively deaminated to uracyl (read as thymine during PCR), whereas methylated cytosine remains cytosine. Cytosine-guanine pair melts in higher temperatures, as compared to the adenine-thymine pair, resulting in markedly different melting profiles. Due to its simplicity and high reproducibility, MS-HRM is now gaining importance both in screening and in determining new molecular biomarkers. 13,14 The DNA methylation studies in CNS neoplasms usually concentrate on single genes 15 or evaluate methylomes or methylation profiles using microarrays. 16 The first approach carries a disadvantage, as cancer is not a single gene disease. 17 The second methodology provides an enormous amount of data that create a complex picture of a disease, but fail to serve as a simple test. 2,[18][19][20] The golden mean would be defining a set of genes crucial for diagnostic and therapeutic purposes. Thus, the aim of this study was to explore the biomarker potential of the promoter meth- can also exert an angiogenic effect in renal and lung cancer. 22 Runt-related transcription factor functions as a tumor suppressor, and the gene (RUNX3) is frequently deleted or transcriptionally silenced in cancer. 23,24 Its inactivation is frequently caused by promoter methylation. 21 29 It is implicated to contribute to neurodegenerative disorders. 29 INA is overexpressed mostly in oligodendroglial phenotype gliomas and is related to 1p/19q codeletion with >70% specificity. Therefore, it is a favorable prognostic marker. 30 Recently, a set of CpG loci differentially hypermethylated in GBM short-term survivors (overall survival < 1 year) vs. long-term survivors (overall survival > 3 years) was identified. 31 According to this report, methylation of INA was one of the top hypermethylated loci and indicated short-term survival. Another differentially hypermethylated gene was CBLN4 (cerebellin 4 precursor), a trans-synaptic cell adhesion molecule, which is important for the synaptic organization of specific subsets of neurons 32 and it was not previously linked to brain tumors or other cancers. 6 MGMT is already established as a predictive factor in patients with GBM treated with temozolomide (TMZ). 4,33 It encodes O 6methylguanine-DNA methyltransferase, which acts by removing alkyl adducts from the O 6 position of guanine at DNA level, thus antagonizing the cytotoxic effects of alkylating agents, including TMZ, gold standard anti-GBM therapeutic. 7 Recently, also the prognostic value of that marker was established, making it a great descriptor of a link between disease character and therapy response. 16 Loss or altered expression of RASSF1A, Ras association domain family 1 isoform A encoding gene, has been associated with the pathogenesis of a variety of cancers. 34 In our previous study, hypermethylation of RASSF1A analyzed in circulating tumor-derived DNA differentiated primary from metastatic brain cancers. 35 The summary of the proposed panel characteristics is presented in Table S1.

| Sample characteristics
This study was performed on brain glioma samples obtained from 46

| DNA isolation and bisulfite conversion
GeneMATRIX Tissue DNA Purification Kit (EurX, Gdańsk, Poland) was used for DNA isolation. DNA concentration and purity were verified using NanoDrop spectrophotometer. Bisulfite conversion of 500 ng of genomic DNA was performed using EZ DNA Methylation Kit (Zymo Research, USA), following the manufacturer's protocol.
The elution volume after bisulfite conversion was 10 µL, and 1 µL of each converted DNA was taken for the subsequent MS-HRM analysis.

| Standards
CpG Methylated HeLa Genomic DNA (New England Biolabs, USA) and CpGenome Universal Unmethylated DNA Set (Merck, Germany) of equal concentration were mixed in different ratios (0%, 5%, 10%, 25%, 50%, 75%, 100% methylated DNA) to mimic DNA samples with defined levels of DNA methylation. These standards were used for the evaluation of the sensitivity of the assay and the semi-quantitative estimation of gene promoter methylation in the clinical samples. The assays were optimized in terms of primer annealing temperature to obtain the best possible resolution and the highest sensitivity.
Moreover, the whole-genome amplified DNA from pooled peripheral blood lymphocytes was prepared with GenomePlex ® Whole Genome Amplification Kit (Merck, Germany), according to manufacturer's protocol, and was used for each run.

| MS-HRM analysis
MS-HRM analysis was performed using Light Cycler ® 96 (Roche Diagnostics GmbH, Germany). Bisulfite-converted DNA was amplified using 5x Hot FIREPol EvaGreen HRM Mix (Solis BioDyne Co., Estonia). Reactions were carried out in a total volume of 20 µL containing 5× HOT FIREPol EvaGreen HRM Mix, 10 pmol/µL of each primer, depending on the gene assayed and 1µl of the template.

TA B L E 1
The list of patients with brain gliomas evaluated in the present study with their basic characteristics (histopathological diagnosis, WHO grade, gender, age, and overall survival time), as well as promoter DNA methylation of analyzed genes (SFRP1, SFRP2, RUNX3, CBLN4, INA, MGMT, RASSF1A)  Table S2).
The obtained melting curves were normalized automatically by the calculation of the "line of the best fit" in between two normalization regions before and after the significant fluorescence decrease.
The methylation level of each sample was assessed by comparison of the PCR product normalized melting curve/peak with the normalized melting curves/peaks of the controls. Data interpretation was performed according to the guidelines of Smith et al. 42

| Statistical analysis
All variables were measured on an ordinal scale, so they did not require any test to assess normality distribution. Nonparametric tests, Note: The percentage of methylation is expressed with different color intensity (lowest DNA methylation level-brightest color, highest-the darkest one). Indication "nk", not known-indicates samples for which the results were not obtained or were ambiguous.

| MS-HRM analysis of a gene panel of promoter DNA methylation
The normalized melting curves and normalized melting peaks of the analyzed genes obtained from standards and representative samples are presented in Figure S2 (SFRP1), Figure S3 (SFRP2), Figure S4 (RUNX3), Figure S5 (CBLN4), Figure S6 (Table S3).

| Interrelations between promoter methylation levels of genes in the analyzed panel
The strong interrelationship between promoter methylation levels was

| Comparison of MS-HRM and pyrosequencing of MGMT gene
In order to validate the results of MS-HRM, pyrosequencing of MGMT promoter region was applied to all 46 samples analyzed. Ten representative pyrograms and corresponding MS-HRM normalized melting curves are presented in Figure S10. Both methods were highly

| CON CLUS IONS
New epigenetic biomarkers associated with CNS tumors are constantly being sought and tested. Their identification would enable more rational selections of strategies to cure patients and prevent disease relapse. However, it is already known that probably there cannot be a single gene test for glioma detection, as well as for predictive and prognostic purposes, because of the multifactorial process of carcinogenesis. 17 Epigenetics is another area for exploring the disease background and biomarkers, as well as the search for potential therapeutic targets. 45 Additionally, the debate over the best analytical method for methylation markers detection is still ongoing.
We proposed a 7 gene panel for brain glioma characterization. It In this study, the highest median of promoter methylation (>25

F I G U R E 5
The interrelation between the methylation levels of genes analyzed in pairs. * indicates statistical significance at P < 0.05, **P < 0.005, and ***P < 0.0005 [Colour figure can be viewed at wileyonlinelibrary.com] Among many assays for DNA methylation analysis, MS-HRM is one of the most frequently recommended methods. 13,14 It has already been proved useful in the assessment of cancer biomarkers in bladder, colorectal, and breast cancer patients. 11 Our study is the first report presenting the application of MS-HRM for CNS tumor analysis. One of the possible disadvantages of MS-HRM is that the reaction is semi-quantitative since the results of the analysis are presented as a methylation range. Nevertheless, such data presentation is commonly regarded as sufficient for the sample evaluation in the clinics. 51 Many advantages of the method, including its sensitivity as well as specificity, cost-effectiveness, and no sophisticated equipment needed, make this method suitable for the routine clinical application. The fact that MS-HRM analysis requires no manual post-PCR processing and is performed in a closed-tube system, with minimal risk of contamination, is equally important.
In order to validate MS-HRM results, we performed additional pyrosequencing analysis of the most important biomarker from our panel, namely MGMT. For the purpose of this comparison, the pyrosequencing data were categorized into methylation ranges, same as in MS-HRM method. The obtained data showed high agreement among methylation ranges, even though both methods do not analyze exactly the same CpG dinucleotides ( Figure S10). However, when pyrosequencing and MS-HRM results were dichotomized into methylated and unmethylated groups (taking 10% methylation as a cutoff value) the results were fully concordant. Similar results were reported by several groups. [52][53][54] In this study, we also established and optimized protocols for MS-HRM analysis of CBLN4 and INA genes, which were, according to our best knowledge, not reported so far. The reaction was optimized in terms of, for example, primer sequences and primer annealing temperature, so that the fluorescent melting profiles from subsequent methylated DNA standards exhibit significant differences. The sensitivity of both assays was high, allowing the detection of as low as 5% methylation. Indeed, in our tested group, the fluorescent curves of almost all of the samples were found between 0% and 5% methylated standards.
INA encodes a protein that is a member of the intermediate filament family and is involved in the morphogenesis of neurons. It is a novel candidate for a CNS biomarker, indicating GBM with a worse prognosis. 31 In turn, CBLN4 encodes a cerebellin 4 precursor, involved in the regulation of neurexin signaling during synapse development. 55 Their clinically relevant level of promoter methylation and potential role as CNS cancer biomarkers should be determined in further studies.
In the current study, not only INA and CBLN4 but also other genes from the chosen panel did not show any association between DNA methylation level and patients' OS. Thus, their prognostic potential is minimal. However, in our previous study, SFRP1 methylation was associated with shorter OS. 26 Also, MGMT methylation status represents one of the most relevant prognostic factors in GBMs. 4,33 Therefore, the results of our study should be verified on larger patients' cohorts.
CNS tumors have two morbidity peaks. The first one occurs in childhood, and the second one is between 55 and 65 years of age. 56 In this study, patients' age ranged between 16 and 83 years of age, and higher methylation levels of two genes, namely SFRP1 and INA, were observed in older patients, which confirmed our previous observation, 26 indicating a more frequent occurrence of SFRP1 26 and RUNX3 27,28 promoter methylation in older CNS tumor patients. In turn, lack of association between age and SFRP1 methylation level was proved by Chang 57 and Kafka, 58 while no correlation for MGMT in the analysis of 58 anaplastic astrocytomas was reported by Bell et al. 59 The patient's sex is another recognized prognostic factor for brain malignancies. Epidemiological data show that glial tumors are slightly more common (about 1.2×) in men than women. 56

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.