Analysis of TET Expression/Activity and 5mC Oxidation during Normal and Malignant Germ Cell Development

During mammalian development the fertilized zygote and primordial germ cells lose their DNA methylation within one cell cycle leading to the concept of active DNA demethylation. Recent studies identified the TET hydroxylases as key enzymes responsible for active DNA demethylation, catalyzing the oxidation of 5-methylcytosine to 5-hydroxymethylcytosine. Further oxidation and activation of the base excision repair mechanism leads to replacement of a modified cytosine by an unmodified one. In this study, we analyzed the expression/activity of TET1-3 and screened for the presence of 5mC oxidation products in adult human testis and in germ cell cancers. By analyzing human testis sections, we show that levels of 5-hydroxymethylcytosine, 5-formylcytosine and 5-carboxylcytosine are decreasing as spermatogenesis proceeds, while 5-methylcytosine levels remain constant. These data indicate that during spermatogenesis active DNA demethylation becomes downregulated leading to a conservation of the methylation marks in mature sperm. We demonstrate that all carcinoma in situ and the majority of seminomas are hypomethylated and hypohydroxymethylated compared to non-seminomas. Interestingly, 5-formylcytosine and 5-carboxylcytosine were detectable in all germ cell cancer entities analyzed, but levels did not correlate to the 5-methylcytosine or 5-hydroxymethylcytosine status. A meta-analysis of gene expression data of germ cell cancer tissues and corresponding cell lines demonstrates high expression of TET1 and the DNA glycosylase TDG, suggesting that germ cell cancers utilize the oxidation pathway for active DNA demethylation. During xenograft experiments, where seminoma-like TCam-2 cells transit to an embryonal carcinoma-like state DNMT3B and DNMT3L where strongly upregulated, which correlated to increasing 5-methylcytosine levels. Additionally, 5-hydroxymethylcytosine levels were elevated, demonstrating that de novo methylation and active demethylation accompanies this transition process. Finally, mutations of IDH1 (IDH1 R132) and IDH2 (IDH2 R172) leading to production of the TET inhibiting oncometabolite 2-hydroxyglutarate in germ cell cancer cell lines were not detected.


Introduction
In the past decade, epigenetic profiles of various cell types and tissues were analyzed with great ambition. In general, methylation of cytosines within CpG dinucleotides of promoter regions leads to epigenetic gene silencing. The protein family of DNA methyltransferases (DNMTs) catalyzes the methylation of cytosines. DNMT1 is a maintenance methylase that remethylates the newly synthesized DNA strand during replication, requiring the hemimethylated parental strand as a blueprint. The DNMTs 3A and 3B are able to methylate DNA de novo in cooperation with the cofactor DNMT3L. In contrast to this, the question how DNA can become demethylated is only poorly understood. The observation that the fertilized zygote and primordial germ cells (PGCs) lose methylation within one cell cycle led to the hypothesis that these cells are able to actively remove methylation marks in their genome [1] [2] [3] [4] [5]. Several publications of the last years could demonstrate that oxidation of 5-methylcytosine (5mC) to 5hydroxymethylcytosine (5hmC) by the ten-eleven-translocation (TET) proteins is a key process of active DNA demethylation (ADD) [6] [7]. It is hypothesized that 5hmC is converted into a cytosine by two mechanisms -the oxidation and the deamination pathway [8]. The oxidation pathway involves further oxidation steps of 5hmC to 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) by the TET proteins. 5fC and 5caC can be replaced by an unmodified cytosine by the base excision repair (BER) pathway, involving the enzyme thymine DNA glycosylase (TDG) [9] [10]. Nevertheless, it is also suggested that 5fC and 5caC might become deformylated or decarboxylated by so far unidentified deformylases or decarboxylases, eventually resulting in replacement of modified cytosines by unmodified cytosines [8]. The deamination pathway involves deaminases of the AID or APOBEC family to deaminate 5mC or 5hmC to thymidine or 5-hydroxymethyluracil, respectively. Both deamination products are thought to be replaced by cytosines by the BER pathway, involving TDG, MBD4 or the single strand selective monofunctional uracil DNA glycosylase (SMUG1) [8].
In mice, it has been shown that oxidation of 5mC to 5hmC is mainly catalyzed by Tet1 [11]. Tet1 knock-down resulted in an increase in global and Nanog promotor specific DNA methylation, while 5hmC levels become reduced [12]. Furthermore, TET2 and TET3 are able to catalyze the same reactions as TET1, but TET3 has only been reported to be active in ADD of the male pronucleus at zygote stage [12] [13].
Testicular type II germ cell cancers (GCC) arise from a common precursor lesion, the carcinoma in situ (CIS) [14] [15]. CIS cells are thought to arise from a PGC, which is arrested in development. GCCs can be subdivided into two groups -the seminomas and the non-seminomas [15]. Seminomas are considered to be the default developmental pathway of CIS cells [16]. The non-seminomas comprise of the pluripotent undifferentiated embryonal carcinomas (EC) and the more differentiated teratomas, yolk-sac tumors and choriocarcinomas.
Recently, it has been demonstrated that mutations of the isocitrate dehydrogenases (IDH) 1 and 2 genes (IDH1 R132 , IDH2 R172 ) in glioblastomas, diffuse astrocytic and oligodendroglial tumors are associated with the production of the oncometabolite 2-hydroxyglutarate and decreased levels of alpha-ketoglutarate, leading to inhibition of alpha-ketoglutarate-dependent oxygenases, which are involved in DNA and histone tail demethylation [17] [18] [19] [20]. Additionally, mutations of the IDH genes are associated with increased DNA methylation at CpG islands [21]. Since TET enzymes require alpha-ketoglutarate and oxygen [22], mutations of IDH1 and IDH2 might have an impact on ADD processes in GCCs.
Here we present a comprehensive study of human adult germ cells, GCCs and their respective cell lines with regard to ADD pathway keyplayers, like TET1-3 and global cytosine modifications. Further, we asked if changes in ADD keyplayer expression and cytosine modification levels (5mC/5hmC) accompany the transition from the seminoma-like cell line TCam-2 into an EClike state in vivo.
Our results suggest that in spermatogonia of the adult testis oxidation of 5mC to 5hmC and further to 5fC as well as 5caC occurs. While 5mC levels are maintained until post-meiotic stages of spermatogenesis, 5hmC, 5fC and 5caC levels gradually decrease during mitosis/meiosis. By extending our analysis to human type II GCCs, we found the majority of CIS and seminoma cells to be hypomethylated and hypohydroxymethylated, while non-seminomas displayed high 5mC and 5hmC levels, demonstrating correlation between 5mC and 5hmC. Interestingly, 5fC and 5caC was detectable in all analysed GCCs, including hypomethylated and hypohydroxymethylated specimen. Quantification of 5mC, 5hmC and 5fC levels in GCC cell lines as well as qRT-PCR/cDNA microarray analysis of GCC cell lines and tissues suggest that GCCs preferentially utilize the oxidation pathway for ADD, involving TET1. Furthermore, we could find no mutation of the IDH1 (IDH1 R132 ) and IDH2 (IDH2 R172 ) genes in GCC cell lines, which are associated with aberrant production of the TET inhibiting oncometabolite 2-hydroxyglutarate. Finally, xenograft experiments demonstrated that the transition of seminomatous TCam-2 cells to an EC-like state is accompanied by strong upregulation of DNMT3B, DNMT3L and an increase in 5mC suggesting de novo methylation. Furthermore, increasing 5hmC levels suggest an ADD process during the in vivo transition.

Ethics statement
The ethics commitee of the Rheinische Friedrich-Wilhelms-Universitä t Bonn approved the analyses of formalin fixed, paraffin-embedded type-II germ cell cancer tissues in context of this study. No personal patient data will be collected or stored. Written permission to use the tissue for scientific purposes was obtained from the patients and was approved by the 'Ethik-Kommission für klinische Versuche am Menschen und epidemiologische Forschung mit personenbezogenen Daten der Medizinischen Fakultä t der Rheinischen Friedrich-Wilhelms-Universität Bonn' (The ethics committee for clinical trials on humans and epidemiological research with patient-related data of the medical faculty of the Rheinischen-Friedrich-Wilhelms-University Bonn). All animal experiments were conducted according to the German law of animal protection and in agreement with the approval of the local institutional animal care committees (

Testicular and cancer tissues for mircoarray analysis
Description of testicular tissue sample collection and preparation for cDNA microarray expression analysis was described in Biermann et al. [33]. Briefly, all testicular tissue samples were obtained immediately after orchiectomy. Normal testicular specimens were obtained from three patients with prostate carcinomas. All tumors were classified according to the WHO classification based on their histology as pT1 stage. Samples were examined by frozen section to assure a significant tumor cellularity. Immunohistochemical staining of CIS samples with an antibody to TFAP2C ensured that samples were composed of CIS only (100% of the tubules composed of CIS) without any normal testicular tubules or invasive tumors. DNA, RNA and protein isolation DNA, RNA and proteins were isolated as described previously [34] [35]. Briefly, DNA was isolated by phenol-chloroformisoamyl-alcohol extraction, RNA by TRIzol (Invitrogen, Karlsruhe, Germany) and proteins by RIPA buffer (50 mm TRIS-HCl pH 8, 150 mm NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS +1 tablet of protease inhibitor cocktail (Roche, Mannheim, Germany)). DNA/RNA quality was assessed by photometric measurement of ratios 260/280 nm and 260/ 230 nm using a NanoDrop photometer (Peqlab, Erlangen, Germany).

Western blot
Western blot analysis was performed as described previously [35]. See table 1 (Table S1) for antibody details and dilution ratios.

Immunohistochemistry
Tumor tissues were dissected, fixed in 4% phosphate-buffered formalin for 1-2 days at 4uC and processed in paraffin wax. 4 mm thick tissue sections on glass slides were pretreated in the Lab Vision PT Modul (Thermo Scientific, Munich, Germany) and in PT Modul Buffer (pH 6) (TA-250-PM, Medac, Hamburg, Germany) for 20 minutes and at 99uC, followed by a cool-down phase of 20 minutes at RT. Endogenous peroxidases were blocked by incubation in peroxidase blocking buffer (TA-125-HP, Medac, Hamburg, Germany) for 10 minutes. Primary antibodies were incubated for 30 minutes at RT. Signal detection was performed semiautomatically in the Autostainer 480 S (Medac, Hamburg, Germany) using the Bright Vision + polymer detection system (Medac, Hamburg, Germany) and the following settings: Enhancer for 10 minutes, polymer (Poly-HRP-Goat anti-mouse/-rabbit IgG) for 20 minutes, 3, 39-diaminobenzidine (DAB) (415192F, Medac, Hamburg, Germany) for 8 minutes. Afterwards, nuclei were stained by hematoxylin for 3 minutes. See table 1 (Table S1) for antibody details and dilution ratios.

Quantification of 5mC, 5hmC
The ELISA-based 'MethylFlash Methylated DNA Quantification Kit (Colorimetric)' and the 'MethylFlash Hydroxymethylated DNA Quantification Kit (Colorimetric)' (all from Epigentek, distributed by BioCat GmbH, Heidelberg, Germany) were used to quantify the amount of 5mC and 5hmC in the DNA of GCC cell lines. Quantification was performed in triplicates according to the manual. 100 ng of total DNA was used for 5mC quantification and 200 ng for 5hmC quantification. The initial incubation time was 90 min and the final developing time was 10 min. For absolute quantifications standard curves were generated by plotting the concentration of the positive control supplied with the assay against the optical density (OD) at 450 nm after performing the assay. For relative quantification the OD at 450 nm of 5 ng/ml of assay positive control was used as reference.

DNA dot blot
Genomic DNA was spotted onto a positively charged nylon membrane (Roth, Karlsruhe, Germany). The DNA spots were air dried for 15 min and UV-crosslinked (20 sec, 1200J/cm 2 ). Membranes were blocked in 5% dry milk powder in PBS/0,.1% Tween20 (PBSTM) for 1h at room temperature (RT). Afterwards, the membranes were washed in PBS for 365 min at RT. The membranes were incubated with the first antibody, diluted in 5% PBSTM for 3 h, followed by another wash step. The secondary antibody, also diluted in 5% PBSTM was applied for 1 h. After a final wash step, the membranes were incubated in 2 ml PierceSuper Signal West Pico chemiluminescent substrate (Perbio, Bonn, Germany) and the signal was detected using Kodak X-Ray films (Sigma-Aldrich, Taufkirchen, Germany). Staining of membranes by methylene blue (0,.04% methylene blue in 5M sodium actetate) confirmed equal loading. Swirling in distilled water destained membranes. See table 1 (Table S1) for antibody details and dilution ratios.

Whole genome gene expression array analysis
The whole procedure has already been published [36]. The array was reanalyzed in context of this study.

Xenotransplantation of TCam-2 and 2102EP cells
Xeontransplantation was performed as described in a previously [37]. Briefly, TCam-2 and 2102EP cells were cultivated for eight days in standard growth medium. Cells were trypsinized, counted and 1610 7 cells were re-suspended in 500 ml of 4uC cold Matrigel (BD, Heidelberg, Germany) and injected into the flank of CD1 nude mice using a cold syringe.

Statistical analysis
qRT-PCR analysis, quantification assays and activity measurements were performed in technical triplicates. Standard deviations were calculated and two-paired t-tests were performed to determine the significance of the data. Significance level a = 5%.

Results
Since murine PGCs are able to actively demethylate their genome during their migration to the genital ridges [38], we asked if adult human gonocytes and germ cells of different developmental stages also display ADD.
ELISA-based absolute quantification of 5mC and 5hmC levels in DNA of GCC cell lines (1 seminoma, 8 EC and 1 choriocarcinoma cell line(s)) revealed that all cell lines display high levels of 5mC (0.53-1.60%), with the highest level in the EC cell line 2102EP ( Fig. 2A), confirming the findings by Wermann et al. [42]. Li et al. analysed levels of 5hmC in different human tissues and cancer cell lines and demonstrated high levels of 5hmC in brain (0.67%), liver (0.46%), kidney (0.38%), colon (0.45%) and rectum (0.57%) tissue and little 5hmC content was detected in four cancer cell lines (Hela, HCT116, SW620, AN3Ca; ,0.02%) [43]. Compared to these results, GCC cell lines also display low levels of 5hmC (0.014-0.027%). Of note, low levels of 5hmC had been demonstrated for several various cancer cell lines compared to normal, healthy tissues [44] [45]. Standard curves used for absolute quantification of 5mC and 5hmC levels are given in supplemental figures S2 (Fig. S2). A dot blot analysis demonstrated the presence of 5fC and 5caC in all GCC cell lines analyzed (Fig. 2 B, 5fC, 5caC). A methylene blue staining served as loading control (Fig. 2B, MB). IHC staining of global DNA methylation by a 5mC specific antibody further validated the 5mC status in three different GCC cell lines (Fig. S2), i.e. moderate 5mC levels in TCam-2, moderate to strong levels in NCCIT and high levels of 5mC in 2102EP. These results suggest that GCC cell lines are able to oxidize 5mC to 5hmC, 5fC and 5caC, which are part of the oxidation pathway of ADD.
To analyze the expression of genes involved in ADD we performed quantitative real-time RT-PCR (qRT-PCR) analysis of the GCC cell lines and compared the results to data from a cDNA microarray of GCC tissues [46] (Fig. 3A, B). Seminoma and EC cell lines and tumors display most prominent expression of TET1 while TET2 and TET3 are expressed at very low levels, suggesting that here TET1 is the main molecule involved in 5mC oxidation. In the choriocarcinoma derived JAR cells, high expression of all TETs could be found, with the highest expression for TET3 (Fig. 3A), suggesting that either TET3 is preferentially used for ADD or all three TET proteins can be utilized for ADD. We used an ELISA-based method to measure the demethylating activity of nuclear extracts and detected similar levels in the nuclear extracts of TCam-2, 2102EP and NCCIT cells, while demethylating activities of the cytoplasmatic extracts were comparably low (Fig. S3). Since TET1 expression is much higher in TCam-2 than in 2102EP and NCCIT cells, but demethylation activities are comparable ( Fig. 2A), we speculate that EC cells are also able to utilize different mechanisms to demethylate their genome. DNTM3B is highly expressed in ECs (Fig. 3A, B) and is assumed to be also a 5hmC dehydroxymethylase [47], so we assume that ECs are additionally able to utilize DNMT3B for ADD.
It has been demonstrated that mutations of the IDH1 and IDH2 gene (IDH1 R132 , IDH2 R172 ) in glioblastomas are associated with the production of the TET inhibiting oncometabolite 2-hydroxyglutarate [17] [18] [19]. In the GCC cell lines analyzed here IDH1, but not IDH2 is strongly expressed (Fig. 3A). We screened all analyzed GCC cell lines for mutations in the IDH1 (IDH1 R132 ) and IDH2 (IDH2 R172 ) by pyrosequencing, but did not detect any mutations (Data S1, example given in Fig. S3). Glioblastoma samples with confirmed mutations of the IDH1 or IDH2 gene was used as positive controls. Therefore, it is unlikely that GCC cell lines accumulate the oncometabolite 2-hydroxyglutarate and reduce alpha-ketoglutarate levels that might influence TET enzyme mediated DNA demethylation.
The DNA glycosylases TDG, MBD4 and SMUG1 are thought to activate the BER at the end of the ADD cascade [8]. It is hypothesized that TDG is part of both, the deamination and oxidation pathway, while MBD4 and SMUG1 operate in the deamination pathway [8]. Since we hypothesized that the oxidation pathway of ADD is used in GCC cell lines, we believe that TDG is the main DNA glycosylase involved in this process. In line with this, we found a strong signal for TDG by western blotting (Fig. S3). Furthermore, we measured TDG activity by an ELISA-based principle and found comparably high TDG activity in the nuclear extract of TCam-2, 2102EP, NCCIT and JAR cells (Fig. S3).
The role of GADD45A/B, which is thought to recruit AID/ AICDA and APOBECs during ADD is still subject to discussion [48] [49]. In GCC tissues and cell lines, GADD45A, GADD45B, AID/AICDA and APOBEC1 are expressed at very low levels suggesting that these factors might not be essential for the ADD mechanism in GCC (Fig. 3A, B). These results indicate further that GCCs preferentially utilize the oxidation pathway of ADD. Finally, we found comparable expression of the maintenance DNA methylase DNMT1 on mRNA level in all cell lines and tissues, confirming the results by Netto et al. [41] who detected no differences in DNMT1 expression between CIS/seminomas and non-seminomas. We verified this finding by measuring a similar DNMT1 protein amount in nuclear and cytoplasmatic extracts of seminomatous TCam-2 as well as EC cell lines 2102EP and NCCIT (Fig. S3). The de novo methylase DNMT3A is nearly absent in all GCC cell lines and tissues, while DNMT3B is highly expressed in most EC cell lines and tissues confirming previous results [50] [33]. Furthermore, the choriocarcinoma cell lines JEG-3 and JAR are strongly positive for DNMT3B, while the seminoma cell line TCam-2 and seminoma tissues displayed only very low expression levels, suggesting that seminomas posses only weak de novo DNA methylation activity, but are able to maintain DNA methylation patterns during replication by the maintenance DNA methyltransferase DNMT1. Strong expression of DNMT3L could be detected in both choriocarcinoma (JEG-3, JAR) and two EC (GCT27, 1777N) cell lines, while all other cell lines display absent or low expression of DNMT3L. These expression profile analysis suggests that all GCC tissues and cell lines utilize the oxidation pathway of ADD as well as replication-dependent maintenance DNA methylation (DNMT1), while de novo DNA methylation seems to be restricted to ECs and choriocarcinomas.
During our experiments we recognized varying TET1 mRNA levels in TCam-2 cells. To analyze this phenomenon in detail, we performed qRT-PCR analysis of TET1, TET2 and TET3 expression in TCam-2, 2102EP and JAR cells during 2-8 days of in vitro cultivation (Fig. 4A). We found that TET1 expression levels increased over time in all cell lines analyzed (TCam-2: 2fold; 2102EP: 2-fold; JAR, 2.7-fold), indicating that TET1 expression correlates to cell density. In contrast, TET2 expression decreased slightly over time in TCam-2, while in JAR cells TET2 expression increased (1.6-fold). In TCam-2 and 2102EP cells, expression of TET3 expression remained relatively constant. In JAR cells, TET3 expression decreased after eight days. We measured the TET activity after 2 and 8 days of in vitro cultivation to correlate these data to the expression profile of the TETs (Fig. 4B). Here, the increased TET activity in TCam-2 and 2102EP cells is most likely due to the upregulation of TET1 levels. In JAR cells, induction of TET activity was most pronounced and correlated to upregulation of TET1 and TET2. Furthermore, in 2102EP and JAR cells the increase in TET/TET levels and activity clearly correlated with rising 5hmC levels (relative quantification to 5 ng/ml positive control provided in the kit; oligonucleotide containing either 50% 5mC or 20% 5hmC) (Fig. 4C). Of note, TCam-2 cells, despite displaying the highest levels of TET1 when compared to 2102EP and JAR show only a  very moderate TET activity. In addition, the increase in 5hmC levels was not significant (Fig. 4C).
Finally, we asked how expression of ADD keyplayers and cytosine modification levels (5mC/5hmC) are affected when seminomatous TCam-2 cells transit into an EC in the somatic microenvironment of the murine flank [37]. Therefore, TCam-2 cells were xenografted into the flank of nude mice; 2102EP cells served as an EC control. The marker genes SOX17 and SOX2 are differentially expressed between seminoma and ECs and serve as marker for transition [16]. Furthermore, TCam-2 cells downregulate SOX17 and upregulate SOX2 during in vivo transition to an EC [37]. Thus, we screened for changes in SOX17 and SOX2 expression as an indicator of a successful transition of TCam-2 cells to an EC-like tumor. qRT-PCR analysis demonstrates that SOX17 expression gradually decreases, while SOX2 is strongly upregulated during the time course of the xenograft experiment (Fig. S4). During the in vivo transition, we could not detect a significant change in TET1 expression. TET2 and TET3 were slightly downregulated although overall expression levels were very low (Fig. 5A). Since expression of TET1 is highly variable in TCam-2 cells, we determined the TET1 expression levels at eighteen randomly picked time-points, ranging from 1*10 ' 4 cells to a confluent T75-flask and from one to eight days during cultivation of TCam-2. The average expression of TET1 was calculated based on these values (Fig. S4). The high standard deviation reflects the variability of TET1 expression in TCam-2 cells. Expression of TDG and DNMT3A (also very low) remained unchanged, while DNMT1 expression slightly decreased (Fig. 5A). DNMT3B and DNMT3L were dramatically upregulated in tumors from xenografted TCam-2 cells, suggesting increased de novo DNA methylation (Fig. 5A). This increase was validated on protein level by IHC. The first cells that displayed a strong signal for DNMT3B were detected after one week (Fig. 5B, red arrows) and their number steadily and strongly increased. These results clearly demonstrate that upregulation of DNMT3B and its regulatory cofactor DNMT3L accompanies the transition of TCam-2 from seminoma-like to an EC-like tumor. Relative (to 5 ng/ml positive control provided in the kit; oligonucleotide containing either 50% 5mC or 20% 5hmC) quantification of 5mC and 5hmC levels demonstrates that both, 5mC and 5hmC levels increase (Fig. 5C), indicating that DNA de novo methylation and ADD accompanies the transition to an EC-like status.

Discussion
In this study, we analyzed expression/activity of active and passive DNA (de)methylation keyplayers as well as presence of cytosine modifications (5mC/5hmC/5fC/5caC) during normal and malignant germ cell development. First, we could demonstrate that 5mC levels remain constant during spermatogenesis, while global 5hmC, 5caC and 5fC levels of the DNA declined. These data suggest that the oxidation pathway of ADD is active in spermatogonia but is turned off at later stages of spermatogenesis. Hence, levels and sites of 5mC modification become fixed reflecting a certain epigenetic ground state of sperm, with regard to methylation of imprinted and other epigenetically regulated loci [51] [52] [53]. In consequence this leads to conservation of this epigenetic ground state in sperm until fertilization of an oocyte.
We also analyzed expression of ADD keyplayers in GCC tissues and corresponding cell lines. We found the majority of CIS and seminoma tissues to be hypomethylated and hypohydroxymethylated compared to EC tissues. Therefore, we claim that induction of a process leading to DNA methylation and/or hydroxymethylation is a hallmark of non-seminomatous development from CIS.
Interestingly, we detected 5fC and 5caC in all GCC tissues. With regard to CIS and seminomas, which present as weakly methylated and hydroxymethylated, the question arises when these 5fC and 5caC modifications were established.
Analysis of GCC cell lines showed that 5mC, 5hmC, 5fC, and 5caC are detectable in all samples, implicating that GCC cell lines utilize the oxidation pathway of ADD to demethylate their genome. Nevertheless, compared to regular somatic tissue the amounts of 5hmC in GCC cell lines measured are quite low, a fact already known for various other cancer cells and cancer derived cell lines [44] [43] [45].
Our study demonstrated comparable levels of 5mC, 5hmC and 5fC in the seminoma-like TCam-2 and in the non-seminomatous cell lines. This seminoma-atypical high methylation of TCam-2 cells has been described before [42]. Since our study reveals that 29% of seminomas are hypermethylated compared to CIS and the other seminoma samples analyzed (Fig. S1), it seems plausible that TCam-2 was derived from such a strongly methylated seminoma. Of note, it should be considered for future experiments that TCam-2 cells might not represent seminomas with low 5mC levels. Therefore, TCam-2 cells might behave different with respect to DNA methylation and treatment resistance compared to hypomethylated seminomas. Interestingly, Werman et al. showed that treatment of TCam-2 cells with the demethylating agent 5aza-29-deoxycytidine leads to an enhanced sensitivity towards cisplatin, demonstrating that the DNA methylation level of TCam-2 cells modulates drug resistance [42].
GCC cell lines and tissues display high levels of TET1 expression, compared to TET2 and TET3 as well as moderate TDG expression and activity. In addition they show extremely low GADD45A/B, AID/AICDA and APOBEC expression, which were assumed essential for the deamination pathway [8]. It is unlikely that other APOBEC enzymes compensate for APOBEC1, since we detected only weak expression of other APOBEC family members in GCC tissues by cDNA microarray analysis (Fig. S4). The DNA glycosylase MBD4 is expressed in GCC cell lines and tissues, but has been demonstrated to display a very low excision activity towards 5hmC, 5fC or 5caC [9]. So, we believe that MBD4 plays only a minor and subordinate role in ADD. In summary, the data suggest that GCC tissues and cell lines can utilize the oxidation pathway involving TET1 and TDG for ADD.
Seminomas and the seminoma-like cell line TCam-2 seem to utilize the replication dependent maintenance DNA methylation pathway since they display high DNMT1 expression and low/ absent DNMT3A/B/L levels. In non-seminoma tissues and cell lines, maintenance-and de novo-methylation seem to occur in parallel, indicated by expression of DNMT1, DNMT3B and DNMT3L. In line with this, comparable DNMT1 expression levels between seminomas and non-seminomas and strong DNMT3B expression in ECs, but not seminomas have already been demonstrated [41] [33] [54]. In murine embryonic stem cells (ESC) de novo methylases Dnmt3a/b are highly expressed [55] [56] [57]. De novo methylase expression and 5mC levels are very low in PGCs [58] [2]. Seminomas are assumed to be the default developmental pathway of CIS cells, which are similar in gene expression and morphology to PGCs, while ECs are considered to be a malignant counterpart of ESCs, sharing a similar gene expression profile and differentiation potential. These similarities indicate that in general reduced 5mC levels and low de novo DNMT expression is a hallmark of PGC-like cells (PGC, CIS, seminoma), while ESC-like cells (ESC, EC) are hypermetyhlated compared to PGCs and possess de novo methyltransferase activity.
We could show that TET1, but not TET2 or TET3 expression increased with cell density in TCam-2 and 2102EP cells, while in JAR cells expression of both, TET1 and TET2 increased. Elevated TET1 expression correlated with an increase in total TET activity and increase in 5hmC levels, demonstrating further that TET1 is the main oxygenase involved in oxidation of 5mC to 5hmC, 5fC and 5caC in these cell lines. JAR cells seem to be able to additionally utilize TET2 for 5mC oxidation.
A mutation in IDH1 (IDH1 R132 ) or IDH2 (IDH2 R172 ) causes the production of the TET inhibiting oncometabolite 2-hydroxyglutarate [44] [20]. Thus, despite TET proteins being present in GCCs, a block of 5mC oxidation might be a result of 2hydroxyglutarate production. Sequencing revealed, that IDH1, which is highly expressed in GCC tissues and cell lines was not mutated in GCC cell lines, making an inhibition of the TETs by 2-hydroxyglutarate unlikely. IDH2 expression is nearly absent in GCC cell lines and tissues and IDH2 is not mutated (IDH2 R172 ) in GCC cell lines. Low IDH2 expression was also found in melanoma cells [59]. Additionally, Lian et al. could show that IDH2 plays an important role in maintaining proper 5hmC levels in melanocytes [59]. Thus, we hypothesize, that low levels of IDH2 might contribute to overall low 5hmC levels in GCCs.
The transition of the seminomatous TCam-2 cells into an EC in the murine flank is accompanied by significantly increased DNMT3B, DNMT3L and 5mC levels, suggesting that de novo methylation is induced. In parallel, we measured an increase in 5hmC levels during the in vivo growth of TCam-2 cells. These findings indicate that both, de novo methylation and oxidation of 5mC to 5hmC are necessary to establish TCam-2s EC-like state in vivo. We hypothesize that genes important for a PGC-/CIS-/ seminoma-like cell fate are silenced by de novo methylation, while EC associated genes become actively demethylated. Figure S1 Analyses of cytosine modifications in GCC tissues and cell lines. A) Demonstration of specificity of 5mC and 5hmC antibody by DNA dot blotting. Indicated amounts (20-5 ng) of oligonucleotides containing either 5mC (50%) or 5hmC (20 %) were spotted on positive charged nylon membrane and detected either with 5mC or 5hmC antibody. B) Pie charts of 5mC/5hmC staining intensities in CIS, seminoma and EC tissues (in %). Staining intensities for 5mC were grouped in hypo-or hypermethylated, while 5hmC staining intensities were classified as hypo-or hyperhydroxymethylated. C) IHC of 5mC and 5hmC in teratoma, choriocarcinoma and yolk-sac tumor tissue. Secondary antibody controls are given in the lower right corner. Scale bar: 200 mm. D) IHC staining of 5mC, 5hmC, 5fC and 5caC in hypermethylated seminoma tissue. Scale bar: 100 mm. Standard deviations are given above each bar. E) Measurement of total DNMT1 protein amount in the nuclear (NE) and cytoplasmatic (CE) extracts of TCam-2, 2102EP and NCCIT cells. Standard deviations are given above each bar. Significant changes are labelled by an asterisk (p-value ,0.05) (calculated by twopaired t-test). (TIF) Figure S4 Supporting information to xenograft experiments. A) qRT-PCR analysis of SOX2 and SOX17 expression at indicated time points in xenografted TCam-2 and 2102EP cells. GAPDH was used as housekeeping gene and for data normalization. Standard deviations are given at the top of each bar. B) qRT-PCR analysis of TET1 expression in eighteen different TCam-2 RNA isolates. Sample 19 represents the average expression. GAPDH was used as housekeeping gene and for data normalization. Standard deviation is given at the top of the last bar. C) cDNA microarray analysis of different APOBEC family members in normal testis tissue (NTT), CIS, seminoma, embryonal carcinoma (EC), teratoma and mixed germ cell cancers (mGCC). Standard deviation is given at the top of each bar. (TIF) Data S1 Results of the pyrosequencing-based mutation analysis of IDH1 (R132) and IDH2 (R172) in GCC cell lines. (PDF)