Alu hypomethylation and MGMT hypermethylation in serum as biomarkers of glioma

In order to improve prognosis of glioma patients, better tools are required for early diagnosis and treatment. Serum cell-free DNA methylation levels of Alu, MGMT, P16, RASSF1A from 124 glioma patients and 58 healthy controls were detected by the bisulfite sequencing. The median methylation level of Alu was 46.15% (IQR, 36.57%–54.00%) and 60.85% (IQR, 57.23%–65.68%) in glioma patients and healthy controls respectively. The median methylation level of MGMT in glioma samples was 64.65% (IQR, 54.87%–74.37%) compared to 38.30% (IQR, 34.13%–45.45%) in healthy controls, and all revealed significant differences including P16. However, the median methylation level of RASSF1A was not significantly altered in glioma patients. Furthermore, the methylation levels of Alu and MGMT in serum had a good diagnostic value, and was higher than P16. Interestingly, combination of Alu and MGMT identified additional patients, which were missed by either diagnosis alone. In the Alu group, the patients with high levels were associated with an increased survival rate compared to those who with low levels, with similar results observed in the MGMT group. In the present study, we demonstrated that the methylation level of Alu and MGMT in serum had a better diagnostic value than P16. Moreover, combined analysis of Alu and MGMT showed higher sensitivity for glioma diagnosis. Therefore, both serum Alu and MGMT methylation levels may represent a novel prognostic factor for glioma patients.


INTRODUCTION
Glioma is the most common primary brain tumor, and is associated with a poor prognosis in part due to of its invasive nature. Early diagnosis and prompt treatment of glioma can be used to reduce mortality and improve prognosis. However, at present, the diagnosis of glioma mainly depends on imaging examination, which may misrepresent true intracranial tumor burden and phenotype. Although biopsy is used to confirm diagnosis of glioma, it is an invasive approach. Tumor biomarkers such as Alpha Fetoprotein, Carcino Embryonie Antigen, Cancer Antigen 125 have been reported to be useful to screen early malignancy and aid in cancer diagnosis without invasiveness. Furthermore, following a diagnosis of cancer, biomarkers may be used in order to determine prognosis and predict therapeutic response [1,2]. Unfortunately, to date, there is no such biomarker for glioma patients. Therefore, it is desirable to identify biomarkers for use in the early diagnosis of glioma.
Cell-free DNA can be detected in body fluids such as serum/plasma, urine and sputum, and are released into the blood following apoptosis and necrosis of cancer cells in the tumor microenvironment. Therefore, there are greater amounts cell-free DNA in cancer patients compared to healthy controls. As is known, aberrant www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 44), pp: 76797-76806 Research Paper www.impactjournals.com/oncotarget DNA methylation including global hypomethylation and regional hypermethylation, play an important role in the formation and progression of cancer. Over the past decades, many researchers have demonstrated the presence of aberrant methylation in the serum cell-free DNA of patients with various types of tumors [3][4][5][6][7]. However, few studies have investigated serum cell-free DNA in glioma. Furthermore, the published data on tumor suppressor genes MGMT, RASSF1A and P16 are inconsistent due to the different genes or different methods [8][9][10][11][12][13][14][15]. Moreover, almost all previous studies regarding aberrant methylation in glioma are focused on promoter hypermethylation, which is generally associated with gene silencing [16][17][18][19][20], with is limited information on global hypomethylation in serum cell-free DNA of glioma patients to date. Global hypomethylation mainly occurs in repetitive elements and leads to genomic instability and tumorigenesis [21,22]. As such, global hypomethylation is considered a biomarker in many tumors [23]. Alu elements are the most abundant repetitive elements in the human genome, with the estimation of total methylation content of Alu elements reported to be useful in evaluating the global genomic methylation levels. Cadieux et al [24] reported that the degree of genomic hypomethylation was related with disease stage in glioma and thus may serve as a potential prognostic factor.
In the present study, we use the bisulfite sequencing (BSP), to detect the methylation status of Alu, MGMT, RASSF1A, and P16 in serum cell-free DNA of glioma patients compared to healthy controls. Identification of novel serum-based biomarkers may contribute significantly to the early diagnosis and thus improved prognosis of glioma patients.

Diagnostic performance of serum cell-free DNA
ROC curves were constructed to assess the feasibility of serum cell-free DNA for the diagnosis of glioma patients. We analyzed the methylation levels of Alu, MGMT, P16 in serum between patients and healthy controls. The Area Under Curve was 0.904 (95% CI: 0.861-0.946, P<0.01, Figure 2A) in Alu, 0.962 (95% CI: 0.938-0.985, P<0.01, Figure 2B) in MGMT and 0.599 (95% CI: 0.514-0.683, P<0.01, Figure 2C) in P16. These results indicated that the methylation level of Alu and MGMT in serum may have good diagnostic value that is greater than P16. Further analysis revealed a 54.84% sensitivity and 98.28% specificity at the optimal cutoff value of 48.85% (Youden's index) in Alu, compared to 76.61% sensitivity and 98.28% specificity at the optimal cutoff value of 48.85% (Youden's index) in MGMT. Consequently, combination of Alu and MGMT may identify additional patients that may be missed by either diagnosis alone ( Figure 3).
In the present study, we defined > 46.15% as high methylation level and ≤ 46.15% as low methylation level for Alu, and > 64.65% as high methylation level and ≤ 64.65% as low methylation level for MGMT. In Alu group, the patients with high levels were associated with higher survival rates compared to those with low levels, with median OS time was 31 months in the high level group and 23 months in low level group (P<0.01, Figure 5A). Furthermore, similar results were obtained in the MGMT group, with median OS time 32 months in high level group compared to 24 months in the low level group (P<0.01, Figure 5B).

DISCUSSION
Early diagnosis and treatment plays a critical role in improving prognosis for glioma patients. However, currently there are no tumor-specific biomarkers for glioma to assist in early diagnosis. Previous studies [25][26][27][28][29][30] have reported that aberrant methylation of serum cell-free DNA offers the possibility to identify biomarkers for glioma early diagnosis and improve prognosis. Additionally, the detection of serum cell-free DNA as a noninvasive diagnostic method may provide an alternate diagnostic method to tumor tissue biopsy. Moreover, serum cell-free DNA is very stable and can survive harsh conditions for long time. In addition, serum cell-free DNA can be amplified by PCR or other technologies even if only small amounts are present. Moreover, aberrant methylation including global hypomethylation and regional hypermethylation occur in the early stages of tumorigenesis, and thus exist both in early and advanced cancers [31]. Furthermore, this study is the first to detect global hypomethylation and regional hypermethylation in serum at the same time.
In the present study, we used BSP which is the gold-standard technology to detect the methylation status of serum cell-free DNA, and it could sequence more CpG sites to get more accurate results than other technologies. We detected four genes (Alu, MGMT, RASSF1A, P16) in the serum of 124 glioma patients and 58 healthy controls. Results revealed the methylation levels of Alu, MGMT and P16 had  significant differences between patients and controls. Cadieux et al [24] reported that Alu hypomethylation in glioblastoma multiforme, consistent with the results of our study, which demonstrated that Alu was hypomethylation in glioma patients and heavily methylated in healthy people. Similarly, we also found MGMT hypermethylation in patients while hypomethylation in controls consistent with previous reports [32][33][34]. P16 has been reported as hypermethylation in glioma [11], with the present study demonstrating a similar trend that was not significantly different. However, this study obtained different results to the literature in RASSF1A [19,35,36], as we did not identify RASSF1A hypermethylation in patients and hypomethylation in healthy controls. For further analysis of diagnostic value of serum cell-free DNA for glioma, we constructed the ROC curves, the AUC of Alu, MGMT, P16 was 0.904, 0.962, 0.599, respectively and identified that the methylation levels of Alu and MGMT in the serum had better diagnostic value than P16. Furthermore, combined analysis of Alu and MGMT revealed higher sensitivity for glioma diagnosis.
To evaluate the prognostic significance of Alu and MGMT, we analyzed the methylation level of Alu and MGMT in high-grade and low-grade groups. In Alu, the low-grade glioma patients had higher levels than high-grade group. The high-grade group had higher level than low-grade group in MGMT. Moreover, we performed the overall survival curves according to the follow-up data of patients. In the Alu group, results showed that patients in the high level group had higher survival rates compared to the low level group. MGMT analysis, also showed that the high level group had longer survival time compared to the low level group, perhaps due to epigenetically silenced MGMT treated with radiotherapy and alkylating chemotherapy may improve the survival of patients with glioblastoma without MGMT expression [37,38].
In the present study, we detected increased global hypomethylation (Alu) and regional hypermethylation (MGMT, RASSF1A, P16) in serum cell-free DNA of glioma patients compared to healthy controls, and demonstrate that both Alu and MGMT may have good diagnostic value. However, our results suggest that there was much better diagnostic efficiency if a combination of Alu and MGMT was used. Therefore, both serum Alu and MGMT methylation level could be useful tools to predict the prognosis of glioma patients. But could patients that have hypermethylated MGMT with hypomethylated Alu have worst overall survival than patents with only hypermethylated MGMT, hypomethylated Alu only, and hypomethylated MGMT with hypermethylated Alu? We plan to increase the number of patients in the future research for getting more real and effective results.   Table 1 : age at the time of operation, sex, Karnofsky performance score, tumor diameter, tumor location (region, side) and the World Health Organization grade. The study protocol was approved by the Independent Ethics Committee of Affiliated Hospital of Nantong University, with all patients and healthy controls providing informed consent prior to obtaining the serum samples.

Cell-free DNA extraction and sodium bisulfite modification
Peripheral venous blood (5-10 ml) was collected prior to surgery from 124 glioma patients and 58 blood samples from healthy people were also obtained as controls. Serum was isolated by centrifugation at 3500 rpm for 10 min at room temperature and stored at −80°C

Cloning and sequencing
The separated PCR products by the gel were purified using QIAquick Gel Extraction kit (Qiagen) and then cloned using the TOPO TA-cloning kit (Invitrogen) according to the manufacturer's instructions. Colony PCR was undertaken to screen the positive colonies, and the clones of the right size of PCR product were sequenced using the ABI 3730 DNA Analyzer. Additionally, 10 clones of each sample were subjected to standard sequencing analysis.

Statistical analysis
Analysis of all data was performed with GraphPad Prism software (version 5.01). Comparison between two groups was tested by using the t test or Mann-Whitney test. Receiver Operator Characteristic curve curve was constructed to assess the diagnostic value of serum cellfree DNA for glioma. Overall survival (OS) was calculated from the day of first surgery until death or end of followup, and analysis performed by Kaplan-Meier and the log rank test. The level of statistical significance was set at P<0.05.

Author contributions
Mingjie Gong and Rongtao Chu conceived and designed the study. Mingjie Gong, Wei Shi, Jing Qi and Guoping Shao performed the experiments. Mingjie Gong, Zhenghua Shi and Junxiang Wang wrote the paper. Jian Chen and Rongtao Chu reviewed and edited the manuscript. All authors read and approved the manuscript.