Characterization of global 5-hydroxymethylcytosine in pediatric posterior fossa ependymoma

Background 5-Hydroxymethylcytosine (5hmC) is a novel epigenetic mark and may be involved in the mechanisms of tumorigenesis and malignant transformation. However, the role of 5hmC in ependymoma, the third most common brain tumor in children, remains unclear. The aim of this study sought to identify the characterization of 5hmC levels in pediatric posterior fossa ependymoma and to evaluate whether 5hmC levels could be a potential factor to predict clinical outcomes. Results Our results showed that 5hmC levels were globally decreased in posterior fossa ependymoma compared with normal cerebellum tissues (P < 0.001). Group A posterior fossa ependymomas had higher 5hmC levels than group B tumors (P = 0.007). Moreover, 5hmC levels positively correlated with Ki-67 index in posterior fossa ependymoma (r = 0.428, P = 0.003). Multivariate Cox hazards model revealed that patients with high 5hmC levels (> 0.102%) had worse PFS and OS than patients with lower 5hmC levels (< 0.102%) (PFS: HR = 3.014; 95% CI, 1.040–8.738; P = 0.042; OS: HR = 2.788; 95% CI, 0.974–7.982; P = 0.047). Conclusions Our findings suggest that loss of 5hmC is an epigenetic hallmark for pediatric posterior fossa ependymoma. 5hmC levels may represent a potential biomarker to predict prognosis in children with posterior fossa ependymoma.


Background
Ependymoma (EPN) is a relatively rare neuroepithelial tumor that arises throughout the whole neuraxis [1]. Intracranial EPN predominantly occurs in children and adolescents, with two third of those tumors located in posterior fossa [1,2]. Recently, posterior fossa ependymoma (EPN_ PF) has been classified into two molecular subgroups based on DNA CpG island (CpGi) methylation profiles status [3][4][5][6]. Group A ependymoma (EPN_PFA) is characterized by CpGi hypermethylation with 1q gain and occurs predominantly in infancy and young children. These subgroup tumors also exhibit global low H3K27me3 [7][8][9], global DNA hypomethylation [7], and high expression of EZHIP [10]. Conversely, group B ependymoma (EPN_PFB) presents with CpGi hypomethylation and primarily occurs in adolescences and young adults. Moreover, the molecular classification of EPN has provided a superior prognostic prediction and risk stratification [11]. EPN_PFA tumors are often difficult to completely resect and bear a dismal prognosis, while EPN_PFB tumors are less invasive and carry a favorable prognosis [4,5]. It suggests that epigenetic mechanisms play an essential role in EPN_PF pathogenesis and tumor maintenance.
Abnormal DNA methylation at the 5 position of cytosine (5mC) is an epigenetic mark of cancers. Recent studies presented evidence for an active DNA demethylation pathway initiated by the ten-eleven translocation (TET) protein family, resulting in the conversion of 5mC into 5hydroxymethylcytosine (5hmC) [12,13]. As a new epigenetic biomarker, 5hmC is reshaping the view of the tumor epigenome. Several reports have shown that decreased 5hmC level is an indicator of poor survival in the central nervous system (CNS) tumors patients [14][15][16][17]. However, only one report studied the changes of 5hmC as well as its downstream products in two EPN cell lines, which represent a subgroup of supratentorial EPN with RELA fusion [18].
In the present study, we performed the ultra-highperformance liquid chromatography-mass spectrometry (UHPLC-MS/MS) analysis and immunochemistry (IHC) staining analysis to measure global 5hmC and 5mC levels to relate this information to clinical characteristics and survival outcomes in pediatric EPN_PF.

Discussion
Epigenetic modifications are important in normal development and frequently alter during tumorigenesis [23]. Recent studies have demonstrated that the loss of 5hmC in different types of cancers might play an essential role in pathogenesis [24][25][26][27][28]. Kraus et al. [29] reported that loss of 5hmC was found in EPN, but only one case (1/ 23) in this group was EPN_PF. However, it is still unclear whether level of 5hmC is changed in pediatric EPN_PF. To the best of our knowledge, this is the first study to illustrate the alternation of 5hmC level in EPN_ PF. We showed that global levels of 5hmC are strikingly decreased in pediatric EPN_PF tumor tissues compared with normal cerebellum tissues. Our results further support the previous observations of the reduced level of 5hmC that occurred in other types of CNS tumors [14,17,29]. Moreover, we found that 5hmC score based on IHC staining with 5hmC antibody significantly related to 5hmC levels detected by UHPLC-MS/MS, suggesting that the IHC staining approach may be a useful method to identify 5hmC level in pediatric EPN_PF.
The limited number of biomarkers reliably predicting prognosis in EPN highlights the importance of developing more robust prognostic markers [22]. Previous studies indicate that 5hmC levels are associated with clinical outcomes in multiple types of cancers [27,30]. Analysis of multiple cohorts of intracranial ependymoma highlights a wide variance in the utility of the grade II versus grade III distinction as a prognostic marker. However, the utility of histological grading of ependymoma for risk stratification has been controversial and without consistent associations of tumor grade with patient outcome [11]. Recently, DNA  methylation patterns and DNA copy number profiles can be used to predict clinical outcomes for patients with EPN [22]. In this study, our result indicated that high 5hmC level is an independent prognostic factor of poor survival in pediatric EPN_PF on PFS and OS. In most solid tumors, low 5hmC levels relate to high tumor grade and worse outcomes [17,27,29,[31][32][33]. However, our results are in line with the previous report that high levels of 5hmC independently correlate with inferior overall survival in acute myeloid leukemia [34] and suggest that 5hmC may be involved in the distinct mechanisms of tumorigenesis and malignant transformation. These findings remain to be confirmed in future studies. Recent advances in the biological characterization of EPN_PF have demonstrated the existence of two clinically, demographically, and molecularly distinct entities [4]. EPN_ PFA tumors show higher methylation of CpGi. EPN_PFA patients are younger, have laterally located tumors with an increased occurrence of chromosome 1q gain, and behave more aggressively compared with EPN_PFB patients [4]. In addition to CpGi hypermethylation, DNA hypomethylation and global H3K27me3 reduction in the absence of recurrent genetic changes in EPN_PF suggests that epigenetic mechanisms are central to EPN_PF pathogenesis [7,8]. Moreover, several studies [7,8,[35][36][37] showed that reduced H3K27me3 in EPN_PF is not genetically driven but epigenetically deregulated. Our study firstly observed distinct 5hmC levels between two molecular subgroups, suggesting that 5hmC may participate in the abnormal DNA methylation in pediatrics EPN_PF. Future researchers on EPN_PF should focus on the mechanism of epigenetic alternations.
We found a strong positive correlation between 5hmC levels and Ki-67 index in pediatric EPN_PF. Several studies demonstrate that higher Ki-67 index seems to be associated with poor prognosis in pediatrics EPN [20,38,39]. Our data confirm previous results and further suggest that high 5hmC levels associated with inferior outcomes. However, some studies have reported that 5hmC levels are inverse correlation of cellular proliferation in different types of cancers [17,[40][41][42]. These discrepancies may be explained by thedistinct tumorigenesis in this benign tumor compared with other malignant tumors. However, the mechanism of 5hmC influenced tumor cell proliferation in EPN_PF needs further research.
Our study had several limitations. The first limitation of our study was the small sample size and the relatively short follow-up period limited our ability to detect robust survival predictors. Future studies with large sample sizes and long-term follow-up are needed to confirm the results of our findings. Second, the molecular subgroup was classified by IHC. To overcome this limitation, future studies using fresh frozen tumor tissues followed by methylation arrays are needed [4].

Conclusions
In our study, we found that 5hmC is a potential prognostic predictor that may contribute to the improvement of clinical risk stratification for EPN_PF. Our results indicate that the characteristic of 5hmC level is associated with molecular subgrouping and cell proliferation. These findings suggest that the mechanisms responsible for regulating 5hmC may represent a possible future therapeutic target.

Study design and samples
The main objective of this study was to assess the clinical characteristics of DNA hydroxymethylcytosine in pediatric EPN_PF. We used UHPLC-MS/MS to investigate 5hmC abundance in EPN_PF. Further, we conducted molecular classification using IHC. A total of 45 patients (age < 18 years) who were diagnosed with EPN_PF at Beijing Tiantan Hospital between January 2010 and December 2017 were included in this study. Clinical data, including age at diagnosis, gender, tumor size, treatment, recurrence data, and survival, were collected by retrospective chart review. Two neuropathologists reviewed the histopathologic findings according to 2016 WHO classification of CNS tumors [1]. Follow-up evaluations were performed on all patients via either an outpatient consultation or a telephone interview. This study was approved by the ethics committee of Beijing Tiantan Hospital, Capital Medical University. Written informed consent was obtained.
Tumor tissues were obtained during initial surgery before radiation or any other adjuvant treatment. All samples were snap-frozen (− 80°C) or fixed with 4% buffered formalin, paraffin-embedded. All tumor specimens were sterilely stored at Beijing Neurosurgical Institute by the Definition of EPN_PF molecular subgroup by immunohistochemistry IHC analyses were performed as reported [9]. In brief, tissue sections were cut at 5 μm, followed by deparaffinization and rehydration using xylene and ethanol. Then, the slides were incubated in 3% hydrogen peroxide for 10 min in phosphate-buffered saline to block endogenous peroxidase activity. Slides were incubated overnight with rabbit monoclonal anti-H3K27me3 antibody (C36B11, Cell Signaling, Danvers, MA, USA) at a concentration of 1:150 using the standard Leica Bond protocol IHC-F. The Leica Bond Polymer Refine DAB detection kit was used according to the manufacturer's instructions. All IHC slides were evaluated by two independent neuropathologists; the scoring methods were performed as report described [9]. H3K27me3 positive staining was defined as scored positive when more than 80% cells had nuclear positivity and scored negative when they did not.

Evaluation of global 5mC and 5hmC by UHPLC-MS/MS
The absolute amount of 5hmC and 5mC in EPN was measured as previously described [43,44] IHC analysis for 5hmC, 5mC, and Ki-67 The utilized primary antibodies were including 5hmC (1: 800, ab214728, Abcam), 5mC (1:200, ab10805, Abcam), and Ki-67 (1:1500, ab15580, Abcam). Immunohistochemical detection of 5hmC and 5mC was performed as described above exclusive of the step of DNA denatures by 2N HCl [27]. The 5hmC and Ki-67 staining and score methods were performed according to described previously [45]. In brief, positive staining was defined as a dark brown staining pattern, confined to the nuclear region. Scant or fine granular background staining or no staining was considered as negative. The mean value of the five snapshots was calculated to represent the percentage of positive cells in each case.

1q gain by interphase FISH
Dual color interphase fluorescence in FISH analysis was performed on formalin-fixed paraffin-embedded sections as previously described using commercially available 1q25 (spectrum green) and 1p36 (spectrum orange) probe sets (ZytoVision, Germany) [46]. The evaluation criteria and scoring system adopted was based on previously described [46,47].

Statistical analysis
All statistical analyses were performed with SPSS 23 (IBM Corp., New York, NY, USA) and two-sided P values < 0.05 were considered statistically significant. The normality of variables was assessed. Data are expressed as mean ± standard deviation (SD) or median (minimum to maximum). Differences in mean and median values were evaluated by using Student's t test and the Mann-Whitney U test, respectively. Associations between categorical variables were assessed via Fisher's exact test. To interpret the effect of 5hmC level in a more clinically relevant manner, 5hmC levels were dichotomized into two groups using Cutoff Finder [19]. The cutoff values (0.102%) were defined as the points with the most significant split between groups, including PFS and OS. For the survival analysis, overall survival (OS) was calculated from the date of initial surgery that established the pathological diagnosis to the time of death. Progression-free survival (PFS) was derived from the date of initial surgery to the time of progression. Kaplan-Meier curves of OS and PFS were generated and log-rank tests were used to compare OS and PFS between demographic and clinical factors. Multivariate Cox proportional hazards regression models with backward stepwise selection were used to identify significant prognostic factors for OS and PFS. Hazard ratios with corresponding 95% confidence intervals were calculated.