Global DNA methylation analysis reveals miR-214-3p contributes to cisplatin resistance in pediatric intracranial nongerminomatous malignant germ cell tumors

Abstract

Background

Pediatric central nervous system germ cell tumors (CNSGCTs) are rare and heterogeneous neoplasms, which can be divided into germinomas and nongerminomatous germ cell tumors (NGGCTs). NGGCTs are further subdivided into mature teratomas and nongerminomatous malignant GCTs (NGMGCTs). Clinical outcomes suggest that NGMGCTs have poor prognosis and survival and that they require more extensive radiotherapy and adjuvant chemotherapy. However, the mechanisms underlying this difference are still unclear. DNA methylation alteration is generally acknowledged to cause therapeutic resistance in cancers. We hypothesized that the pediatric NGMGCTs exhibit a different genome-wide DNA methylation pattern, which is involved in the mechanism of its therapeutic resistance.

Methods

We performed methylation and hydroxymethylation DNA immunoprecipitation sequencing, mRNA expression microarray, and small RNA sequencing (smRNA-seq) to determine methylation-regulated genes, including microRNAs (miRNAs).

Results

The expression levels of 97 genes and 8 miRNAs were correlated with promoter DNA methylation and hydroxymethylation status, such as the miR-199/-214 cluster, and treatment with DNA demethylating agent 5-aza-2ʹ-deoxycytidine elevated its expression level. Furthermore, smRNA-seq analysis showed 27 novel miRNA candidates with differential expression between germinomas and NGMGCTs. Overexpresssion of miR-214-3p in NCCIT cells leads to reduced expression of the pro-apoptotic protein BCL2-like 11 and induces cisplatin resistance.

Conclusions

We interrogated the differential DNA methylation patterns between germinomas and NGMGCTs and proposed a mechanism for chemoresistance in NGMGCTs. In addition, our sequencing data provide a roadmap for further pediatric CNSGCT research and potential targets for the development of new therapeutic strategies.

Importance of the study

NGMGCTs are malignant pediatric brain tumors, which have a poorer prognosis than germinomas. However, few studies have discussed this phenomenon. We employed methylation and hydroxymethylation DNA immunoprecipitation sequencing and identified a mechanism that contributed to the chemotherapy resistance in NGMGCTs. Furthermore, our findings not only provide insights into the possible molecular mechanisms underlying the therapeutic resistance of NGMGCTs but also provide clues that may help in the development of new target therapies and therapeutic strategies in the future.

Central nervous system germ cell tumors (CNSGCTs) are rare and heterogeneous. Their incidence varies significantly according to geography. CNSGCTs constitute 3% of tumors in Western countries compared with 11%–15% of all pediatric CNS tumors observed in Japan, Korea, and Taiwan. The histological types of CNSGCTs are germinomas and nongerminomatous GCTs (NGGCTs).¹ NGGCTs are further subdivided into mature teratomas and nongerminomatous malignant GCTs (NGMGCT), including immature teratomas, teratomas with malignant transformation, yolk sac tumors, embryonal carcinomas, choriocarcinomas, and mixed GCTs. Common locations for the development of CNSGCTs are the pineal gland, sellar region, basal ganglia, and ventricles.² The diagnosis is made based on the clinical features, neuroimaging findings, serum tumor marker levels (alpha-fetaprotein and beta human chorionic gonadotropin), histological findings, and response to radiation therapy and/or chemotherapy in selective patients. A histopathological comparison of the degrees of differentiation and malignancy of NGMGCTs indicates that germinomas are the most undifferentiated GCT. Germinomas and NGMGCTs carry different prognoses and sensitivities to chemotherapy and radiotherapy.³ Germinomas also show a good response to dose reduction in extended focal radiation therapy and a good prognosis with or without less intensive chemotherapy.⁴ However, NGMGCTs require not only standard dose radiotherapy but also intensive chemotherapy⁵; furthermore the outcomes are markedly worse than those of germinomas. The mechanisms underlying this chemotherapy-related difference between NGMGCTs and germinomas are still unclear.

DNA methylation at the fifth position of cytosine (5mC) is an important epigenetic modification of the mammalian genome that is involved in the regulation of many biological processes, including embryonic development, gene and microRNA transcription, cellular differentiation, X chromosome inactivation, genomic imprinting, and chromosome instability.⁶ Accordingly, 5mC has long been considered a stable epigenetic marker of DNA. In 2009, ten-eleven-translocation (TET) family dioxygenases were identified, which can oxidize 5mC to intermediate products, including 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC), and 5-carboxylcytosine (5caC).⁷ Aberrant 5mC and 5hmC expression patterns are associated with tumor formation.^8–10 Hypermethylation at specific promoter regions of tumor suppressor genes, such as DNA repair pathway genes (hMLH1, MGMT, and BRCA1), cell cycle control genes (p16^ink4a, p15^ink4b, and RB), apoptotic or pro-apoptotic proteins (APAF-1 and DAPK1), and p53 network genes (p14^ARF, p73, and SFRP1), has been associated with drug resistance.^10,11 However, the mechanisms of methylation-regulated coding and noncoding genes that cause chemotherapy resistance in NGMGCTs are still unclear.

MicroRNAs (miRNAs) are 21- to 23-nucleotide (nt), single-strand, noncoding RNAs that play essential roles in many cellular processes, including development, stem cell division, differentiation, apoptosis, disease, and cancer formation.¹² Many studies have distinguished between the global miRNA profiles of germinomas and NGMGCTs.^13,14 Wang et al reported that some miRNAs, such as miR-142-5p, were downregulated in NGMGCTs and some, such as miR-335 and miR-654-3p, were upregulated.¹³ Palmer et al indicated that the miR-371–373 and miR-302 clusters were upregulated in malignant GCTs.¹⁴

In the present study, we performed methylation and hydroxymethylation DNA immunoprecipitation sequencing (MeDIP-seq and hMeDIP-seq, respectively) to analyze the differential DNA methylation and hydroxymethylation signatures between germinomas and NGMGCTs. We identified several genes and miRNAs, such as FZD7 and the miR-199a/-214 cluster, that were hypomethylated and upregulated in NGMGCTs. Furthermore, we demonstrated that in NGMGCTs, miR-214-3p overexpression enhanced cisplatin resistance by downregulating the expression of its target, the apoptotic protein BCL2-like 11 (BCL2L11/BIM).

Materials and Methods

Biological Samples

The parent/legal guardian of the patients in this study provided informed consent, and all procedures were approved by the institutional review board of VGH-TPE (VGHIRBNU.:20l004018IA). Fresh-frozen tumor tissues were collected during surgery from patients with intracranial GCTs. Data of the study cases were retrieved from the surgical pathology files of the Department of Pathology and Laboratory Medicine at Taipei Veterans General Hospital.

Cell Culture

Human embryonic kidney (HEK)293T cells were obtained from the American Type Culture Collection. Human embryonal carcinoma NCCIT cells were obtained from Dr Huang’s lab (Taipei Medical University, Taiwan). HEK293T cells and NCCIT cells were maintained in Dulbecco’s modified Eagle’s medium and Roswell Park Memorial Institute 1640 medium, each supplemented with 10% fetal bovine serum (all Gibco/Life Technologies). These cells were incubated at 37°C in a humidified atmosphere of 5% CO₂.

Methylated and Hydroxymethylated DNA Immunoprecipitation Analysis

Total DNA was extracted from biological samples and sequenced using the Illumina Hiseq2000 and Nextseq system according to the manufacturer’s instructions. Briefly, double-strand DNA was sonicated and then ligated to adaptor for further amplification (Qiagen GeneRead Library Prep). Next, the double-strand DNA was denatured and immunoprecipitated with anti–5ʹ-methylcytosine and anti–5ʹ-hydroxymethylcytosine antibodies (1 μg, Active Motif), respectively, overnight. The enriched methylated DNA fragment was amplified by polymerase chain reaction (PCR) (GeneRead DNA I Amp Kit), and high-throughput sequencing was performed.

Filtered reads were first mapped using Hg19 as the reference. To identify differential methylation and hydroxymethylation regions (DMRs and DhMRs), bam format files were used as input for the MEDIPS algorithm.¹⁵ The window of each DMR was set to 250 bps, and the minimum number of reads for each group was set to 10. The identified regions of DMRs and DhMRs in NGMGCTs were compared with germinomas. DMRs and DhMRs were considered significant when P < 0.05 and < 0.01, respectively. DMRs and DhMRs were annotated according to their genomic location by using ChIPseeker, an R package.¹⁶

Small RNA Sequencing Analysis

Total RNA was collected, and small RNA fractions were sequenced using an Illumina HiSeq 2000 and Nextseq, according to the manufacturer’s instructions. Briefly, total RNA was subjected to directional RNA adapter ligation, first strand cDNA synthesis, and size selection. The cDNA corresponding to small RNA (16 to 30 nt) was collected and sequenced. Raw sequencing reads in the FASTQ format were generated and then subjected to our in-house bioinformatics pipelines for miRNA profiling and discovery.¹⁷

Lentivirus Production and Transduction

Knockdown or miRNA expression plasmids were co-transfected with 2 packaging plasmids, pCMV-dR8.91 and pMD2.G, into HEK293T cells by using the TurboFect transfection reagent (Thermo Scientific). The virus suspension was harvested by filtering the culture media through 0.22-μm filters (Nalgene) at 48 and 72 h after transfection. For lentivirus transduction, the virus suspension was added to NCCIT cells, and transduced cells were selected by the antibiotic resistance marker blasticidin (10 μg/mL) or puromycin (2 μg/mL) for 1 week. All procedures involving the manipulation of infectious materials were performed in a Biosafety Level 2 laboratory.

RNA and Reverse Transcription Quantitative PCR

Total RNAs of tumor tissues and cultured cells were isolated using TRIzol reagent (Invitrogen/Life Technologies). Total RNA was reverse-transcribed (RT) into complementary DNA through miRNA-specific or random hexamer priming using a RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific). MiRNA-specific and U6-specific RT primers are shown in Supplementary Table S1. Quantitative PCR was performed in duplicate with miRNA- or gene-specific primers by using a Maxima SYBR FAST quantitative (q)PCR kit (Thermo Scientific), and the specific product was detected and analyzed using the StepOnePlus Real-Time PCR System. MiRNA-specific and U6-specific qPCR primers are shown in Supplementary Table S1. Quantitative PCR was performed in a 20-µL reaction volume consisting of 10 µL 2× SYBR green mix, 0.5 µL of 10 μM forward primer, 0.5 µL of 10 μM reverse primer, 8 µL nuclease-free water, and 1 µL template. For gene detection, thermal cycle was programmed for 10 minutes at 95°C as initial denaturation, followed by 40 cycles of 15 sec at 95°C for denaturation, 15 sec at 58°C for annealing, and 30 sec at 72°C for extension. For miRNA detection, thermal cycle was programmed for 5 minutes at 95°C as initial denaturation, followed by 40 cycles of 10 sec at 95°C for denaturation, and 60 sec at 60°C for annealing and extension. Relative gene expression changes were evaluated with the 2^−ΔΔC_T method.¹⁸ The cycle threshold (C_T) values of the target genes or miRNAs were subtracted to internal control glyceraldehyde 3-phosphate dehydrogenase or U6 expression levels, respectively, within the same sample to determine ∆C_T, and then average ∆C_T values of technical replicates were calculated. For each gene or miRNA of interest, the ∆C_T values of the testing group were then subtracted to the average ∆C_T of the control group to obtain ∆∆C_T. The expression levels were presented as fold change using 2^−ΔΔC_T.

5-Aza-2ʹ-Deoxycytidine Treatment

In the 5-aza-2ʹ-deoxycytidine (Sigma) treatment, NCCIT cells were incubated with 2.5, 5, and 10 μM of 5-aza-2ʹ-deoxycytidine or with a corresponding amount of dimethyl sulfoxide for 96 h. The medium was changed every 24 h. The cells were then harvested for RNA extraction.

MTT Assay and Cisplatin Treatment

To evaluate cell viability, cells were seeded at 1 × 10⁴/well and incubated at 37°C. After incubation for 24 h, the cells were treated with different concentrations of cisplatin for 48 h. After 48 h, the cells were treated with 1% thiazolyl blue tetrazolium for 30 min at 37°C, followed by 0.1% sodium dodecyl sulfate in 2-propanol, and thorough mixing was performed. The results were obtained by measuring the absorbance at wavelengths of 570 and 650 nm using a multiwell scanning spectrophotometer.

Immunoblotting

Knockdown or miRNA expression plasmids were introduced into NCCIT cell lines using lentiviruses. After 48 h, the cells were harvested and lysed using NET lysis buffer containing protease and phosphatase inhibitors. Immunoblotting was performed using anti-BCL2L11 (1:1000; Cell Signaling) and anti–caspase-3 (1:1000; Cell Signaling) antibodies, followed by visualization using horseradish peroxidase–conjugated secondary antibodies and an enhanced chemiluminescence detection system (Merck Millipore).

In Vitro Methylation and Reporter Assay

In brief, 1 µg of the purified plasmid DNA was incubated with the cytosine-phosphate-guanine (CpG) methyltransferase M.SssI (New England Biolabs) and S-adenosylmethionine at 37°C for 1 h. The methylated plasmids were extracted by using PCR cleanup kits (Bioman). The protein expression plasmid (600 ng), methylated plasmid (300 ng), and Renilla luciferase plasmid (15 ng) were co-transfected into 1.5 × 10⁵ HEK293T cells by using the TurboFect transfection reagent. Luciferase activity was measured using a luminometer (model LB593, Berthod). Firefly luciferase activity was normalized to Renilla luciferase activity.

Immunohistochemistry

Immunohistochemistry (IHC) sample preparation and staining were performed as described previously.¹⁹ The antibodies used were anti-BCL2L11 (1:100) and anti–BCL-XL (1:1000; Cell Signaling).

Statistics Analysis

Independent sample t-tests were performed to compare the continuous variation between the 2 groups. P < 0.05 was considered significant. All data are reported as mean ± SD.

Results

Distribution of DMRs and DhMRs and Identifying Methylation-Regulated mRNAs

MeDIP- and hMeDIP-seq were performed to examine the differential distribution of 5mC and 5hmC between germinomas and NGMGCTs. Our study included 6 germinomas and 6 NGMGCTs; the supporting clinical information is listed in Table 1. Regions that showed significant changes in the 5mC and 5hmC methylation in NGMGCTs compared with germinomas are defined as DMRs and DhMRs, respectively. Both 5mC and 5hmC modifications were widely distributed across the genome, and the DMRs and DhMRs occurred more frequently at introns and distal intergenic regions (Fig. 1A). Occurrence of DMRs at promoter region was about 11% (NGMGCT hypomethylation) and 10% (NGMGCT hypermethylation); DhMRs at promoter region were about 4.5% (NGMGCT hypohydroxymethylation) and 14% (NGMGCT hyperhydroxymethylation). Gene expression was associated with the presence of higher 5hmC and lower 5mC at the gene promoter regions. Furthermore, non-CpG island promoter methylation silences gene expression.²⁰ To determine genes regulated through methylation, genes with DMRs and DhMRs occurring at the promoter region were selected and their gene expression values were obtained from our previously published mRNA microarray dataset.¹³ We identified 26 downregulated genes that had higher promoter 5mC and lower 5hmC and 71 upregulated genes that had lower promoter 5mC and higher 5hmC in NGMGCTs compared with germinomas (Fig. 1B and C). The top 10 upregulated and downregulated genes are listed in Supplementary Table S2.

Table 1.

Clinical details of sequencing patients enrolled in MeDIP-seq, hMeDIP-seq, small RNA-seq, and mRNA array

Sample No.	Age, y	Sex	Pathological Diagnosis	Surgical Excision	Treatment	AFP	HCG	MeDIP- seq	hMeDIP- seq	smRNA- seq	mRNA Array
G1	14.1	M	G	STR	RT	N	N	+	+	+	+
G2	11.6	M	G	PR	RT	N	N	+	+	–	+
G3	13.9	M	G	NTR	RT	N	N	+	+	+	–
G4	14.0	M	G	PR	RT+CMT	H	N	+	+	+	+
G5	11.4	F	G	PR	RT	N	N	+	+	+	–
G6	12.4	M	G	PR	RT	N	N	+	+	+	–
G7	16.0	M	G	STR	RT	N	N	–	–	+	+
G8	12.2	M	G	PR	RT	N	N	–	–	–	+
G9	10.4	F	G	PR	RT	N	N	–	–	–	+
NG1	11.5	M	MGCT	STR	RT+CMT	H	N	+	+	+	–
			(G+IT+YSC)
NG2	4.3	M	MGCT	NTR		N	N	+	–	–	+
			(G+IT)
NG3	0.7	M	IT	NTR	RT+CMT	H	N	+	+	+	–
NG4	14.5	M	MGCT	PR	RT+CMT	H	N	+	+	+	+
			(G+YSC)
NG5	7.6	M	MGCT	NTR	RT+CMT	H	H	+	+	+	+
			(EC+IT+YSC)
NG6	10.3	M	MGCT	NTR	RT+CMT	N	N	+	+	+	+
			(IT+G)
NG7	0.0	F	IT	NTR		H	N	–	+	+	+
NG8	0.1	F	IT	NTR	RT+CMT	H	N	–	–	+	–
NG9	10.6	M	MGCT	PR	RT+CMT	H	H	–	–	–	+
			(IT+YSC)

Abbreviations: G, germinoma; NTR, near total removal, >95%; MGCT, mixed GCT; STR, subtotal removal, >75%; IT, immature teratoma; PR, partial removal, >25% <75%; YSC, yolk sac tumor; RT, radiotherapy; CMT. chemotherapy; normal range of AFP <10 ng/dL; EC, embryonal carcinoma; normal range of betaHCG <10 mIU/dL; “+”: sequenced; “‒”: not sequenced; NG, NGMGCTs or nongerminomatous malignant germ cell tumors; N, normal level; H, high level.

Fig. 1.

Identification of methylation-regulated mRNAs between germinomas and NGMGCTs. (A) Distribution of DMRs (left) and DhMRs (right) within genomic regions. (B–C) Venn diagram demonstrating the principle of putative methylation- and hydroxymethylation-regulated mRNAs. The number of intersections indicates the overlapped mRNAs across each group. (B) Comparison of downregulated, hypermethylated, and hypohydroxymethylated genes in NGMGCTs. (C) Comparison of upregulated, hypomethylated, and hyperhydroxymethylated genes in NGMGCTs.

Small RNA Sequencing Reveals Differential Expression of Known and Novel MiRNAs

To identify the miRNome differences between germinomas and NGMGCTs, the global miRNA expression patterns of 6 germinomas and 7 NGMGCTs were investigated using small RNA sequencing (smRNA-seq) and analyzed with our in-house bioinformatics pipelines.²¹ In addition to known miRNAs, novel miRNA candidates were predicted using 3 independent bioinformatics algorithms (miRDeep2, mireap, and miRanalyzer) after removal of mRNA contamination and known miRNAs. In total, we identified 159 known and 27 novel miRNAs differentially expressed between germinomas and NGMGCTs (q < 0.01; Fig. 2A and Supplementary Figure S1A). The top 20 upregulated and downregulated miRNAs are listed in Table 2, and the other significant known miRNAs are listed in Supplementary Table S3. Through RNA fold analysis (http://rna.tbi. univie.ac.at/cgi-bin/RNAfold.cgi), we verified that these novel miRNA candidates can fold into a hairpin secondary structure (Supplementary Figure S1B); all the differentially expressed novel miRNA candidates are listed in Supplementary Table S4.

Fig. 2.

Identification of methylation-regulated miRNAs and miR-199a/214 cluster expression regulated through DNA methylation. (A) Small RNA-Seq revealed differentially expressed known miRNAs between germinomas and NGGCTs (q < 0.01). (B) Schematic representation for identifying methylation-regulated miRNAs. The putative targets were obtained by overlapping the differentially expressed miRNAs and differentially methylated miRNA promoter region identified from MeDIP and hMeDIP-sequencing. A total of 8 miRNA expressions might be regulated by methylation (6 upregulated and 2 downregulated in NGMGCTs). (C) MiR-214-3p (left) and miR-199a-5p (right) were upregulated in NGMGCTs (n = 11) compared with germinomas (n = 12) (*P < 0.05, **P < 0.01, by t-test). (D) Scatter plots illustrating negative correlations between MeDIP values and the reads per million of miR-214-3p (left) and miR-199a-5p (right). Every dot denotes one sample in which MeDIP-seq and smRNA-seq were both performed. (E) Expression of miR-214-3p (left) and miR-199a-5p (right) after treatment with different concentrations of 5-aza-2′-deoxycytidine (2.5, 5, and 10 μM). Expression values expressed as mean ± SD in duplicates and representative of 3 independent experiments. (F) Constructs of the miR-199/214 promoter (length: 1000 bps) that was methylated with M.SssI in vitro and co-transfected with Twist1 (left) and Twist2 (right) expression vectors into 293T cells. Luciferase activities of the methylated expression vectors were normalized to the expression of Renilla firefly luciferase and are representative of 3 independent experiments (**P < 0.01, by t-test).

Table 2.

Small RNA-sequencing showed 81 upregulated and 78 downregulated miRNAs (top 20 differential miRNAs [left: upregulated; right: downregulated])

Name	Average RPM (G)	Average RPM (NG)	LogFC (NG/G)	Name	Average RPM (G)	Average RPM (NG)	LogFC (NG/G)
hsa-miR-675-5p	0.00	18.63	∞	hsa-miR-520a-5p	20.04	0.14	−7.15
hsa-miR-122-5p	1.61	9179.46	12.48	hsa-miR-520f-3p	164.30	1.23	−7.06
hsa-miR-206	5.07	5796.09	10.16	hsa-miR-520b	95.75	0.76	−6.97
hsa-miR-194-3p	0.06	32.15	9.15	hsa-miR-520c-3p	82.19	0.67	−6.94
hsa-miR-585-3p	0.11	49.72	8.78	hsa-miR-515-3p	5.40	0.05	−6.87
hsa-miR- 203b-3p	0.04	13.51	8.48	hsa-miR-526b-3p	17.91	0.25	−6.16
hsa-miR-490-5p	0.08	21.22	8.13	hsa-miR-516b-3p	38.98	0.68	−5.85
hsa-miR- 203a-3p	9.80	2275.29	7.86	hsa-miR-516a-3p	38.98	0.68	−5.85
hsa-miR-196a-5p	6.47	1348.96	7.70	hsa-miR-520e	13.61	0.24	−5.81
hsa-miR-135a-3p	0.72	134.85	7.55	hsa-miR-150-5p	1671.90	31.56	−5.73
hsa-miR- 196b-5p	5.35	949.85	7.47	hsa-miR-3614-5p	4.11	0.10	−5.36
hsa-miR-675-3p	0.78	135.87	7.45	hsa-miR-5701	52.75	1.54	−5.10
hsa-miR- 200b-5p	0.32	39.08	6.93	hsa-miR-96-3p	6.91	0.20	−5.10
hsa-miR-1262	0.28	30.73	6.76	hsa-miR-548x-3p	7.50	0.24	−4.97
hsa-miR- 323b-3p	0.38	40.32	6.72	hsa-miR-202-5p	2.64	0.09	−4.89
hsa-miR-375	7.48	657.19	6.46	hsa-miR-3180-5p	3.89	0.14	−4.84
hsa-miR- 216b-5p	2.02	167.70	6.38	hsa-miR-548aj-3p	10.27	0.39	−4.70
hsa-miR-1-3p	70.71	5608.09	6.31	hsa-miR-6502-5p	1.75	0.07	−4.62
hsa-miR- 487a-5p	0.19	12.97	6.10	hsa-miR-422a	42.78	1.75	−4.62
hsa-miR- 449b-5p	1.38	94.52	6.09	hsa-miR-548ae- 3p	10.24	0.45	−4.52
G: Germinomas	NG: NGMGCTs

Abbreviations: RPM, reads per million; G, germinoma; NG, NGMGCTs or nongerminomatous malignant germ cell tumors.

Identifying Epigenetically Regulated miRNAs in Germinomas and NGMGCTs

To decipher methylation-regulated miRNAs, we first determined the miRNA transcription start site by using miRStart (http://mirstart.mbc.nctu.edu.tw/browse.php) and selected miRNAs with DMRs and DhMRs occurring at the promoter region. By comparing the expression of the selected miRNAs using the smRNA-seq dataset, we identified 2 downregulated (miR-142-5p and miR-142-3p) and 6 upregulated miRNAs (miR-199a-3p, miR-199a-5p, miR-214-3p, miR-214-5p, miR-218-5p, and miR-585-3p) in NGMGCTs that might be under methylation-dependent transcriptional regulation (Fig. 2B). The miR-214/199a cluster contains 4 genes (miR-214-3p, miR-214-5p, miR-199a-3p, and miR-199a-5p) and its promoter DMRs and DhMRs are located at chr1:172,114,251–172,114,750 and chr1:172,116,251–172,117,000, respectively (Supplementary Figure S2A). MiRNAs within this cluster were found to be upregulated in NGMGCTs. We validated the expression levels of miR-214-3p and miR-199a-5p using reverse transcription quantitative PCR (RT-qPCR), and the PCR data correlated well with the sequencing results (Fig. 2C). Furthermore, the reads per million of miR-214-3p and miR-199a-5p were negatively correlated with their methylation status (correlation = −0.8458 and −0.7454, respectively; Fig. 2D). NCCIT is an embryonal carcinoma cell line and its expression pattern was positively correlated with intracranial germ cell tumors (Pearson correlation coefficient r = 0.8‒0.9; Supplementary Figure S2B). The miR-214/199a promoter showed higher 5mC methylation and lower 5hmC methylation, which was similar to the pattern of germinomas (Supplementary Figure S2A). Therefore, we treated NCCIT cells with different concentrations of 5ʹ-aza for 96 h and both miR-214-3p and miR-199a-5p were upregulated after 5ʹ-aza treatment (Fig. 2E). Bisulfite sequencing confirmed demethylation of the miR-214/199a promoter after 5’Aza treatment (Supplementary Figure S2C). Gene promoter methylation inhibits transcription through 2 basic mechanisms: (i) modification of CpG sites within transcription factor binding sites, and (ii) specific binding of methyl-CpG-binding domain proteins to methylated CpG sites to indirectly affect transcription factor binding.²² To determine whether CpG methylation status in the DMRs of the miR-214/199a promoter affects miR-214 and miR-199a expression, we conducted in vitro methylation assays using luciferase reporters containing the full-length miR-214/199a minimal promoters. The cloned sequence and plasmid map are shown in Supplementary Figure S3. CpGs in the promoter region were methylated in vitro using CpG methyltransferase M.SssI before transfection into 293T cells for the luciferase reporter assay. The promoter activity of the miR-214/199a cluster was reduced upon CpG methylation by M.SssI (Fig. 2F). Next, we wish to determine if miR-214/199a promoter methylation affects transcription factor–dependent expression. The Twist1 and Twist2 transcription factors were upregulated in NGMGCTs (Supplementary Figure S4), therefore we co-transfected 293T cells with methylated miR-214/199a promoter reporter plasmids and Twist expression plasmids (Twist1 and Twist2) to detect luciferase activity. Twist proteins have been demonstrated to regulate miR-214/199a expression by binding to E-box elements in the promoter region.^23,24 We found that the Twist-binding site does not contain CpG sites; the expression of Twist1 and Twist2 should enhance the luciferase activity of the miR-214/199a promoter reporter plasmid. However, the luciferase activity was reduced when the miR-214/199a promoter reporter plasmid was methylated (Fig. 2F). These data suggested that miR-214/199a promoter methylation affects transcription factor–dependent miRNA expression.

MiR-214-3p Overexpression Leads to Cisplatin Resistance Through Downregulation of BCL2L11

Previous studies showed that aberrant miRNA expression was associated with chemotherapy resistance by targeting apoptosis-related genes. For example, overexpression of MET protein resulted in tyrosine kinase inhibitor resistance in lung cancers. Garofalo et al demonstrated that MET could induce miR-221/222 and miR-30b/c expression, which in turn targeted the pro-apoptotic proteins apoptotic protease activating factor 1 (APAF-1) and BIM. They showed that overexpression of miR-221/222 and miR-30b/c in HCC827 and PC9 cells significantly reduced gefitinib-induced cell death. They co-transfected wild-type or mutated miRNA binding sites of BIM and APAF-1–containing expression vectors with miR-221/222 and miR-30b/c expression vectors and performed caspase-3/7 and viability assays. They observed increased caspase-3/7 activity when cells were co-transfected with APAF-1 and BIM containing mutated miRNA binding sites and miRNAs. These data suggested that the effects of apoptotic proteins on chemotherapy resistance were directly related to miRNA expression.²⁵ We showed methylation-dependent upregulation of miR-214/199a expression in NGMGCTs. Furthermore, miR-214-3p has been associated with cisplatin resistance in ovarian cancer²⁶ and tongue squamous cell carcinomas.²⁷ To identify whether miR-214-3p is involved in cisplatin resistance in NGMGCTs, we first determined the half-maximal inhibitory concentration values of cisplatin in NCCIT cells to be 6.52 μM (Supplementary Figure S5). Next, we stably expressed miR-214-3p in NCCIT cells (Fig. 3A) and monitored the corresponding cell viability after treatment with different concentrations of cisplatin and etoposide. NCCIT-miR-214-3p cells showed significant resistance to cisplatin cytotoxicity compared with NCCIT-vector cells (Fig. 3B). However, no difference was observed in the survival rates between the NCCIT-miR-214-3p and NCCIT-vector cells treated with etoposide (data not shown). To determine whether miR-214-3p enhances chemotherapy resistance through the regulation of cell proliferation, we seeded cells and assessed the proliferation rate through MTT assay at days 0, 1, 3, 5, and 7. Both NCCIT-miR-214-3p and NCCIT-vector cells had similar proliferation rates (Supplementary Figure S6), which suggests that cisplatin resistance mediated by miR-214-3p does not involve cell proliferation. Next, we used Targetscan (http://www.targetscan.org/) to identify which expressed genes were direct targets of miR-214-3p, and we compared these results with the downregulated genes in NGMGCTs. We determined that miR-214-3p potentially targeted 314 genes, among which 4 were apoptosis-related genes localized in the mitochondria (Fig. 3C). BCL2L11, also known as BIM, is a pro-apoptotic protein.²⁸ BCL2L11 was shown to be the direct target of miR-214-3p.²⁹ In our study, miR-214-3p overexpression reduced BCL2L11 mRNA and protein expression (Fig. 3D). Knockdown of BCL2L11 resulted in increased cisplatin resistance compared with control cells (NCCIT-shvec; Fig. 3E). BCL2L11 knocked-down cells produced an additional cleaved form of caspase-3 after cisplatin treatment (Fig. 3F). These results demonstrate that BCL2L11 is the direct target of miR-214-3p, and the loss of BCL2L11 expression inactivates the apoptosis pathway and confers cisplatin resistance.

Fig. 3.

MiR-214-3p enhanced cisplatin resistance by targeting the pro-apoptotic protein BCL2L11. (A) Overexpression of miR-214-3p in NCCIT cells confirmed with RT-qPCR. (B) NCCIT-vector cells and NCCIT-miR-214-3p cells were treated with cisplatin at different concentrations (4, 6, 8, 10, and 12 μM). Cell viability was detected through MTT assay. Viabilities are expressed as mean ± SD from duplicate wells and are representative of 3 independent experiments (*P < 0.05, by t-test). (C) Schematic representation for identifying miR-214-3p targets. The putative targets were obtained by overlapping the downregulated genes in NGMGCTs and the software-predicted targets. The target genes that we were interested in, that is, the apoptotic genes, were localized in the mitochondria. Four genes were listed. (D) Immunoblotting (right) and RT-qPCR (left) confirmed the levels of BCL2L11 in NCCIT-miR-214-3p cells. RT-qPCR data are shown as mean ± SD in duplicates, and immunoblotting and RT-qPCR are representative of 3 independent experiments. (E) BCL2L11 knocked-down (KD) cells were treated with cisplatin at different concentrations (6, 8, 10, and 12 μM). Cell viability was detected through MTT assay. Viabilities are expressed as mean ± SD from duplicate wells and are representative of 3 independent experiments (**P < 0.01, *P < 0.05, by t-test). (F) BCL2L11-KD cells were treated with 7.5 μM cisplatin, and the levels of caspase-3 were detected through immunoblotting.

Expression of BCL2L11 Was Downregulated in NGMGCTs

We validated BCL2L11 expression in our clinical samples and determined that out of 12 cases of germinomas, 7 (60%) showed upregulated BCL2L11 expression and 5 showed similar or downregulated BCL2L11 expression, compared with NGMGCTs (Fig. 4A). IHC staining indicated that germinomas had strong and intermediate BCL2L11 immunoreactivity (Fig. 4B and C), whereas NGMGCTs had weaker positivity for BCL2L11 (Fig. 4D–F). IHC assessment of another important apoptotic factor, BCL-XL, showed that both types of GCTs had similar staining intensities.

Fig. 4.

BCL2L11 expression was downregulated in NGMGCTs. (A) RT-qPCR confirmed lower mRNA levels of BCL2L11 in NGMGCTs (n = 11) than in germinomas (n = 12) (**P < 0.01, by t-test). (B–F) IHC results for hematoxylin and eosin staining of BCL2L11 and BCL-XL in germinoma and NGMGCT specimens. (B–C) Germinomas. (D) Immature teratoma. (E) Yolk sac tumor. (F) Mixed germ cell tumors. Scale bar: 100 μm.

Discussion

Clinical treatments of NGMGCTs include surgery and a combination of standard dose radiotherapy and intensive chemotherapy. However, the overall survival of NGMGCTs is still markedly lower than that in germinomas.¹³ Previous studies demonstrated that treatment with DNA demethylating agent 5ʹ-aza changed the methylation patterns of radiation-resistant and chemoresistant cells, and restored the chemo- and radiotherapy sensitivity of these cells.^28,29 The results suggest that one of the mechanisms for tumor cell resistance to chemo- and radiotherapy is change in DNA methylation patterns to affect gene expression.^30,31 DNA mutation or aberrant expression of components and cofactors involved in the epigenetic modification and its regulation has been observed in tumors.^32,33 Rare germline variants have been found in histone demethylase, Jumonji domain containing 1C (JMJD1C) in Japanese intracranial GCT patients.^34–36 However, genes encoding for the major methylation-regulating enzymes, including DNMT1, DNMT3A, DNMT3B, TET1, TET2, and TET3, were not mutated in CNSGCTs. Also, our data showed that the TET1 gene was slightly upregulated in germinomas,¹³ whereas other DNA methylation–associated genes had no significant difference in expression compared with NGMGCTs. Therefore, the mechanism for the differential 5mC and 5hmC patterns between germinomas and NGMGCTs still remains to be elucidated.

Cisplatin has been used to treat several types of cancers. However, many solid tumors can easily acquire cisplatin resistance, thus limiting its therapeutic efficacy. Several mechanisms have been proposed that contribute to cisplatin resistance in various cancers, including (i) reduced cisplatin accumulation by regulating drug transport and uptake, (ii) increased levels of glutathione, glutathione-S-transferase, or metallothioneins to detoxify cisplatin, (iii) altered DNA repair involving loss of mismatch repair or increased nucleotide excision repair, (iv) altered DNA damage tolerance, and (v) reduced apoptosis induction.³⁷ The pro-apoptotic protein BCL2L11 (BIM) is associated with cisplatin resistance in several cancers, including ovarian cancer³⁸ and hepatocarcinoma.³⁹ Furthermore, better prognosis correlates with higher BCL2L11 expression in nasopharyngeal carcinomas.²⁹

Dysregulated miRNA expression is well known in cancers and can contribute to chemotherapy resistance through the binding and transcription regulation of downstream targets.⁴⁰ In the present study, 159 miRNAs were differentially expressed in NGMGCTs compared with germinomas, and several miRNAs have been reported to regulate cisplatin sensitivity in other cancers. Previously, miR-449a was reported to regulate cisplatin sensitivity by targeting BCL2 expression in the gastric cancer cell line SGC7901.⁴¹ In addition, miR-214 overexpression has been shown to target phosphatase and tensin homolog gene expression to affect cell survival and cause cisplatin resistance in ovarian cancers.²⁶ In our study, we showed that miR-214-3p overexpression contributed to cisplatin resistance in NGMGCTs by inhibiting BCL2L11 expression, whereas there is no significant difference in expression of phosphatase and tensin homolog between germinomas and NGMGCTs. In nasopharyngeal carcinoma cells, BCL2L11 expression is regulated by miR-214 and miR-214 silencing, resulting in cell apoptosis and suppression of cell proliferation and tumor growth. Since germinomas and NCCIT cells showed low miR-214-3p expression, we treated cisplatin in miR-214-3p overexpressing NCCIT cells and observed significantly reduced cell death. In addition to its biological significance, miR-214-3p can serve as a plasma biomarker. Li D et al reported that the secretion of miR-214-3p–containing exosomes from osteoclast inhibits osteoblast activity and osteoblastic bone formation.⁴² Hao M et al showed that a high level of serum miR-214 expression was associated with bone disease and poor prognosis in patients with multiple myelomas.⁴³ The level of miR-214-3p in the CSF and serum of NGMGCT patients remains to be determined. Moreover, Kuninty PR et al demonstrated that miR-214-3p and miR-199a-3p inhibition can downregulate cell proliferation, differentiation, and migration in cancer-associated fibroblasts in pancreatic tumor stroma.⁴⁴ Therefore, these findings suggest that miR-214 serves as a potential therapeutic candidate and biomarker for NGMGCT treatment.

In summary, integrative analysis of DNA methylome, transcriptome, and miRNome uncovered important differences in the epigenetic and miRNA-mediated transcriptional regulation between germinomas and NGMGCTs. Several methylation-regulated genes and miRNAs, specifically miR-214-3p, are hypomethylated and upregulated in NGMGCTs compared with germinomas. Our study determined that miR-214-3p expression contributed to cisplatin resistance by targeting the pro-apoptotic protein BCL2L11. The CNSGCT omics data from this study are a valuable resource that provides insight into the mechanisms of treatment resistance and could aid in the development of new antitumor therapeutic strategies.

Supplementary Material

Supplementary material is available at Neuro-Oncology online.

Funding

This work was supported by the Ministry of Science and Technology (MOST; 102-2314-B-038-060-MY3, 105-2314-B-038-009), Taipei Medical University (TMU-102-AE1-B44), Taipei Medical University Hospital (104 TMU-TMUH-11, 105TMU-TMUH-01-01), and Comprehensive Cancer Center of Taipei Medical University/Health and Welfare Surcharge of Tobacco Products (MOHW105-TDU-B-212-134-001).

Conflict of interest statement. None declared.