Hypermethylation of RAD9A intron 2 in childhood cancer patients and tumor cell lines

Background Most childhood cancers occur sporadically and cannot be explained by an inherited mutation or unhealthy lifestyle. The prenatal origins hypothesis postulates a role for (epi)genetic mutations which occur stochastically in rapidly dividing cells. This study aims to investigate the impact of adverse methylation in tumor relevant genes in former childhood cancer patients which may be associated with an enhanced risk to develop primary secondary cancers Methods We performed an epimutation screen of several candidate genes (APC, CDKN2A, EFNA5, RAD9A, and TP53), in skin fibroblasts of 20 patients with a primary cancer in childhood and subsequent second primary cancer (2N), 20 matched patients with only one primary cancer in childhood (1N), 20 cancer free controls (0N) and unrelated leukemia cancer samples, using bisulfite pyrosequencing and deep bisulfite sequencing. radiation, colony formation assays, cell proliferation, PCR and molecular karyotype SNP-array, experiments were performed to further characterize RAD9A hypermethylation in fibroblasts and FaDu sub clones. Data were analyzed using the Kruskal-Wallis rank sum test and Benjamini-Hochberg procedure, the linear mixed-effects model was fit using REML and R-scripts. Results Four 1N patients and one 2N patient displayed elevated (10%) mean methylation levels of RAD9A intron 2. Deep bisulfite sequencing of RAD9A in these patients revealed 2% allele methylation errors (defined as alleles with >60% methylated CpGs). We found RAD9A hypermethylation in bone marrow of patients with pre-ALL (pre-acute lymphoblastic leukaemia), AML (acute myloid leukaemia), and plasmablastic lymphoma (PBL), and in EBV-(Epstein Barr virus) transformed lymphoblastoid cells. RAD9A methylation in fibroblasts or tumor cells (FaDu) was not affected by in vitro aging, or DNA damage induced by radiation or the chemotherapeutical daunorubicin. Molecular karyotyping of FaDu sub clones revealed a homozygous these patients displayed elevated allelic methylation errors, which must be considered as functionally relevant epimutations. We propose that the constitutive RAD9A epimutations arose early in development, predisposing the compromised cells to tumorigenesis and patients to increased childhood cancer risk. Analyses of tumor cell clones with high methylation levels of RAD9A suggest a connection between methylation levels of the RAD9A intron 2 locus and a homozygous inactivation of important genes with followed effects on carcinogenesis and progression. Larger prospective studies are needed to correlate constitutive epimutations in RAD9A and other susceptibility genes with life-long cancer risk and progression of carcinogenesis.

in RAD9A. Conclusion We propose that constitutive RAD9A epimutations may have arisen early in development, predisposing the compromised cells to tumorigenesis and increased childhood cancer risk. Analyses of tumor cell clones with high methylation levels of RAD9A suggest a connection between methylation levels of the RAD9A intron 2 locus and a homozygous inactivation of important genes Background Tumorigenesis is a multistep process, involving an accumulation of genetic and epigenetic changes in multiple genes resulting in both, the, inactivation of tumor suppressor (TS) genes and/or activation of oncogenes [1,2]. Tumor epigenomes are characterized by a global loss of DNA methylation, leading to reactivation of retrotransposons and genome instability, as well as regional hypermethylation and silencing of TS genes, compromising DNA repair and cell cycle control [3,4]. In sporadic cancer (epi)genetic changes, which may arise by stochastic DNA replication errors or adverse environmental exposures, are usually restricted to the tumor and its precursor cells. In contrast, most hereditary forms of cancer are caused by germline mutations in tumor suppressor genes, which predispose patients to tumor development, which itself is triggered by inactivation of the second TS allele. Accumulating evidence suggests that, similar to germline mutations, constitutive epimutations involving soma-wide hypermethylation of tumor suppressor genes in normal body cells, can cause phenocopies of cancer syndromes such as hereditary non-polyposis colon cancer (HNPCC) [5,6] as well as breast-and ovarian cancer [7,8]. Since constitutive epimutations usually occur in a mosaic state with variable proportions of affected cells in different tissues, they are most likely not transmitted through the germline but may arise during early development. For some cancer-predisposing genes, i.e. MLH1 [9], MSH2 [10], and DAPK1 [11], the probability for de novo epimutations depends on cis-regulatory genetic sequence variants.
Compared to the aging population, cancer is rare among children and young adults, representing <1% of all cancers. Children are usually not exposed to an unhealthy lifestyle or an adverse environment, and only 5-10% of children with cancers carry germline mutations increasing their cancer risk [12]. Therefore, most childhood cancers should occur sporadically. One explanation for sporadic childhood cancers is somatic mosaicism. A high proportion of human preimplantation embryos are chromosomal mosaics of normal and aneuploid cells [13]. One embryonal cell carrying a de novo chromosome or genetic mutation can be propagated into different tissues and organs during somatic development, laying dormant for many years. Approximately 30% of all human tumors are endowed with RAS mutations, consistent with an essential role of the RAS signaling pathway for tumorigenesis. Several RASopathies including neurofibromatosis type 1 (NF1), Noonan syndrome, Costello syndrome (CS) and cardiofaciocutaneous syndrome (CFCS) are associated with increased cancer risks [14,15]. Although most congenital RASopathies are caused by germline RAS mutations, NF1 and other RASopathies can also be present in a mosaic state, due to postzygotic or somatic mutations.
The prenatal origins hypothesis postulates that childhood cancers arise from postnatally persisting embryonal or more differentiated prenatal cell with predisposing mutations [16,17]. Several studies have demonstrated prenatal oncogenic events underlying acute leukemia in childhood [18][19][20][21]. During each cell division not only the DNA sequence but also its epigenetic modifications are copied to the daughter cells. Considering that the error rate during this copying process is at least one magnitude higher for epigenetic information than for genetic information [22], constitutive epimutations may occur much more frequently than prenatal DNA sequence mutations. We have previously described monozygotic twin sisters discordant for childhood cancer and a constitutive epimutation in the BRCA1 TS gene [23].
To analyze the role of constitutive epimutations in childhood cancer, we performed an epimutation screen for several cancer-relevant genes in a unique cohort, consisting of fibroblasts derived from individuals who survived childhood cancer and subsequently developed a second primary cancer (2N) and matched individuals with childhood cancer but without a second cancer (1N) hypermethylation, consistent with an epimutation [23]. Cell lines 0N24 and 2N24 are a pair of discordant monozygotic twins. One twin suffered from childhood leukemia and later on from a thyroid carcinoma whereas her sister was completely healthy until adulthood.
Control fibroblast strains for cell proliferation and senescence analyses were established from excess skin materials from surgical interventions in non-cancer patients [70].
Epstein Barr virus (EBV) transformation of resting B cells (in peripheral blood lymphocytes) to proliferating lymphoblastoid cells occurs in early stages of infection and is widely used to obtain immortalized lymphoblast cell lines [49]. Lymphoblastoid cells were harvested in an early passage after stable infection. CVS-Samples (excess material) were obtained during the course of routine diagnostics. Bone marrow samples (excess material from routine chromosome diagnostics) were obtained from patients with pre-B-ALL (46,XY,t(9;22)(q34;q11.2)[2]/46,XY [25], AML (46,XY,der(7)(q-).ish del(16)(q22) [12]/ 46,XY [10], and plasmablastic lymphoma (PBL) with complex aberrations in 60% of cells, were used. Cells were exposed to X-rays with a D3150 X-Ray Therapy System (Gulmay Ltd, Surrey, UK) at 140 kV and a dose rate of 3.62 Gy/min at room temperature. Sham irradiated control cells were kept at the same conditions in the radiation device control room. The FaDu tumor cell line was irradiated at 80% of confluency. Cells were exposed to single doses ranging from 2-8 Gy and harvested at 15 min, 2 h, and 24 h after irradiation. For fractionalized irradiation, fibroblasts were irradiated 4x, 8x, or 10x within a period of 20 days with doses of 2 Gy and 4 Gy, respectively, within a period of 20 days. Cells were given one day of recovery time between exposures and the medium was changed twice a week. Cells were harvested one day after the final irradiation.
FaDu derived single cell clones were generated by dilution of the primary cell line and propagated in Minimal Essential Medium with Earle's salts (Invitrogen,), supplemented with 15% FBS, 1% vitamins and 1% antibiotics (Pen/Strep) using 96 plates (Cellstar). After 48h wells with only one cell were selected via microscopic examination. Subsequent propagation was performed in conditioned medium (one day old medium of primary culture at 50% confluency, sterile filtered using 0,2µm filter) and cells were transferred in in to a 24 well plate, later a 6 well and finally 10 cm petri dishes when they reached 80% confluency.

Daunorubicin treatment
Fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 15% fetal bovine serum (FBS), 1% vitamins and 1% antibiotics (Pen/Ptrep) in a 90% humidified incubator with 5% CO 2 at 37°C. The cells were cultured in T25 flasks to 80-90% confluence, than a non-lethal dose [48] of 3µM daunorubicin (Pfizer Pharma PFE GmbH) was added. After 2 hours the medium was replaced and at time points of 2 h, 4 h and 24 h post treatment the cells were harvested by trypsinization. Quantification of γH2AX was performed as previously described [71].

Growth kinetics of FaDu and sub clones
Tumor cell lines were cultured as described above. Cells were collected and seeded at 5*10 4 density in a T75 cell culture flask in triplicates. The cells were harvested by trypsinization and total cell numbers were determined by counting using Moxi z automatic cell counter Fa. Orflo, at every time point. The cellular proliferation rate was calculated as cumulative population doublings (CPD). The statistical analysis was done using the linear mixed-effects model fit by REML setting the biological replicate as random variable.

Colony formation assay
Clonogenic survival was determined in colony formation assays adapted after Menegakis et al [72]. At passage 8, cells were seeded (1x10 5 ) in 10 cm diameter petri dish in triplicates. After 5 days and one change of medium, the cells were irradiated with 2 Gy, 4 Gy, 6 Gy and 8 Gy. Sham irradiated cells were kept at the same conditions in the radiation device control room (0 Gy). After 24 h cell suspensions were obtained for each dose and different seeding densities were plated in triplicates. Remaining cells were pelleted, washed with PBS and stored at -80°C for further experiments. 14 days after irradiation colonies were fixed and stained with crystal violet. Colonies defined as >50 cells were counted and surviving fractions were expressed in terms of plating efficiency. Survival data after radiation dose were fitted to linear quadratic regression models employing the maximum likelihood approach (r package CFAssay). Differences between curves were evaluated using the F-test. Adjustment of p-values was done using the method of
Bisulfite conversion of 0.2 µg DNA was performed with the EpiTect Bisulfite Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. PCR and sequencing primers for the genes above-mentioned were designed with PyroMark Assay Design 2.0 software (Qiagen) (Supplementary Table 1

Deep bisulfite sequencing (DBS)
Next-generation sequencing (NGS) libraries for DBS were generated as described previously [8]. PCR amplification of APC, CDKN2A, TP53, and RAD9A was performed using primers containing a target-specific part and partial adapter overhangs (Supplementary Table 2

DBS Analysis
The sequences in FASTQ format were processed using the Amplikyzer2 [73] pipeline, which provides a detailed nucleotide-level analysis including the calculation of CpG methylation rates. All sequences were aligned to the genomic sequence of each amplicon using default settings. For the subsequent extraction of read and CpG-wise methylation status only reads with an overall bisulfite conversion rate of over 95% have been considered. Further downstream processing of Amplikyzer2 output files and subsequent analyses of methylation rates was performed using R-Scripts. Statistical analyses were performed with the statistical software package R (Version 3.2.2) [74].

Selection of genes for an epimutation screen
We aimed to analyze gene methylation of the most likely candidate genes which are involved in cancer. No pathological germline TP53, BRCA1, BRCA2 and RAD9A mutations were detected using Sanger sequencing in our patients. According to Magnusson et al., [24] there is a connection between familial mutations in BRCA1/2 and CDKN2A genes and childhood cancer. Although we do not have any information about the mutational status of family members, the disease spectrum seen in our patients matches the collective screened in the study conducted by Magnusson and colleagues. Germline mutations in the TS gene BRCA1 account for 25-30% of hereditary breast-and ovarian cancer. It plays a major role in homologous recombination and other DNA repair mechanisms [25]. BRCA1 haploinsufficiency may induce genomic instability and premature senescence in breast epithelial cells [26]. The CDKN2A TS gene is deleted, mutated or epigenetically silenced in a wide variety of cancers [27,28]. It inhibits the catalytic activity of the CDK-cyclin D complex, a regulator of the G1 checkpoint of the cell cycle [29]. Mutations in the APC TS gene cause familial adenomatous polyposis and colon cancer [30]. Inactivation of APC promotes tumorigenesis through loss of cell adhesion [31] and chromosome instability [32]. Aberrant signaling of the Eph family of receptor tyrosine kinases and their ephrin ligands have also been implicated in human cancers. Aberrant methylation of EFNA5 have been found in acute lymphocytic leukemia [33] and non-Hodgkin lymphoma [34]. Germline mutations in the TP53 TS gene cause Li-Fraumeni syndrome which is characterized by familiar occurrence of childhood soft tissue sarcomas and a wide spectrum of early-onset cancers. Somatic mutations in TP53 are among the most frequent genetic alterations in cancer [35]. Mutations in TP53 exon 6 can lead to truncated protein with a putative oncogenic gain of function in cancer cells [36].
The RAD9A gene is involved in different DNA repair pathways, including base excision repair, homologous recombination and mismatch repair, alongside multiple cell cycle phase checkpoints and apoptosis [37]. At least some of these functions are mediated by the RAD9A-HUS1-RAD1 complex. Overexpression due to gene amplification or intron 2 hypermethylation is found in a variety of cancers, consistent with an oncogenic function [38,39]. On the other hand, reduced expression levels in normal body cells of childhood cancer patients with a second primary tumor also suggest a function as genomic caretaker [40].
Epimutation screen of fibroblasts derived from childhood cancer patients Using bisulfite pyrosequencing, we have determined mean methylation of the promoter regions in APC, CDKN2A, and EFNA5, a cis-regulatory region in RAD9A intron 2, and a mutation hotspot in TP53 exon 6 in 20 primary fibroblast cell lines of cancer free controls and 20 1N and 20 2N matched patients (Figure 1). There were significant variation in methylation among the groups. APC ( Figure 1A) proved to be hypermethylated in 1N in comparison to the 0N control (p value = 0.03). CDKN2A (Figure 1B), which is often mutated in a variety of cancers, exhibited hypermethylation in the 1N group (p value = 0.006) while it was hypomethylated in 2N group in comparison to 0N group (p value = 0.008). The 2N group showed a hypomethylation in TP53 ( Figure 1E) in comparison to the control group 0N (adj. p value = 0.01). No significant differences between the groups, were detected for EFNA5 ( Figure 1C) and RAD9A genes ( Figure 1D). In a previous work we showed that outliers could be considered as likely candidates for an abnormal methylation pattern, indicative for a mosaic epimutation [23]. We identified one patient (1N08) who exhibited a conspicuous hypermethylation (9%) of the APC promoter, while patient 1N15 showed a hypermethylation of CDKN2A (7%), and patient 2N12 displayed hypomethylation in TP53 (95%). Five patients (1N04, 1N07, 1N14, 1N20, and 2N21) showed increased RAD9A intron 2 mean methylation, ranging from 10% to 31%. The results for the analyzed outliers are compiled in Table 1.
Average methylation of millions of DNA molecules in a genomic DNA sample is a surrogate marker which can sometimes be difficult to interpret. Methylation changes could be due to either single CpG methylation errors at different positions in a large number of alleles or to a few allele methylation errors, where all or most CpGs in individual DNA molecules are aberrantly methylated. Because it is usually the density of CpG methylation in a cisregulatory region rather than individual CpGs that turns a gene "on" or "off" [41,42], allele methylation errors must be considered as putative functionally relevant epimutations. Deep bisulfite sequencing (DBS) can determine the methylation profiles of many thousands of individual DNA alleles for multiple genes and samples in a single experiment and, thus, directly measure epimutation rates (EMRs). In the present study we performed DBS on the patients with suspected Epimutations (Table 1)

RAD9A methylation
As most of our patients (except 1N20) did receive chemo-and radiotherapy during treatment and the donation of the fibroblast was done in adulthood, several years after the first malignancy, we designed experiments which may indicate treatment related changes in the methylation of our studied genes.

DNA methylation during in vitro aging
Cellular aging in vitro has been frequently used to study cancer pathologies that are caused by accelerated/premature molecular aging [43]. A recent methylation array analysis revealed >500 genes with significant methylation changes during extended fibroblast culture [44]. To study age or tissue culture effects on methylation, fibroblast strains of controls (0N01, 0N02, 0N03, and 0N24) and childhood cancer patient 2N024 were propagated in culture from passage 7 until senescence. Strain 0N03 showed the lowest (n = 20) and 0N01 (n = 41) the highest number of population doublings ( Fig. 2A).
Methylation of APC, BRCA1, CDKN2A, and RAD9A was measured at passages 7, 17 and 25 by bisulfite pyrosequencing. None of the analyzed genes showed a correlation between mean methylation (measured by bisulfite pyrosequencing) and in vitro aging (Fig. 2B-F).

Effects of radiation
We have shown previously that DNA methylation remains rather stable in primary fibroblasts within the first cell cycle after irradiation [45]. In contrast, significant methylation changes in >250 genes and the MAP kinase signaling pathway were associated with delayed radiation effects in irradiated fibroblast single cell clones [44]. To study radiation effects on the methylation of RAD9A intron 2 site, the control cell line 0N18 was analyzed at 15 min, 2 h and 24 h after irradiation with 0 Gy, 2 Gy, 5 Gy, and 8 Gy, at each time point. RAD9A methylation values remained virtually unchanged between 7% and 9% ( Table 2). Furthermore three fibroblast strains (0N12, 1N08, and 2N12) were irradiated in fractions of 8x 2 Gy, 4x 4 Gy, 10x 2 Gy, 8x 4 Gy, and/or 10x 4 Gy within a 20 day period. Again RAD9A methylation remained rather constant at 5% in 1N08 and 2N12, and at 8-9% in 0N12 (Table 2). In addition, exponentially growing FaDu tumor cells were analyzed at 2 h, 4 h, and 24 h after irradiation with a single dose of 0 Gy, 2 Gy, 5 Gy, and 8 Gy, respectively. RAD9A methylation varied within a narrow range between 54% and 57% ( Table 2) and there was no difference between irradiated and non-irradiated cells.

Effects of daunorubicin treatment
Although tumor therapy varied between patients, daunorubicin and doxorubicin were frequently used in the treatment regimens. As both drugs have similar properties [46] and the cellular uptake of daunorubicin is superior to that of doxorubicin [47], we analyzed the influence of daunorubicin, on the RAD9A methylation. Treatment of normal fibroblasts with 3µM danorubicin, as stated in the study of Przybylska and colleagues [48], yields 60% of  [49]. Global changes in DNA methylation may contribute to the pathogenesis of EBV [50]. We therefore tested EBV transformed lymphoblasts for changes in methylation of RAD9A in intron 2. The mean methylation varied in six different EBV transformed cell lines from 6% to 41% (data shown in supplementary Figure 1 and FaDu cell line (mean methylation 54%) with up to three copies [51][52][53]. Triggered by the idea that changes in chromosomal integrity, as seen in many tumorigenic cells including EBV transformed lymphoblasts and the bone marrow samples analyzed in this study, we presumed that this might cause hypermethylation of intron 2 site in RAD9A.
Therefore we generated single cell line clones from the parental FaDu cell line. Sub clonal events with gene mutations were reported by Nisar et al., [54] and recently by Ben-David and colleagues [55] with the consequence of copy number gains and losses and different drug responses. The FaDu cell line exhibits homozygous loss of function mutations in TP53 and CDKN2A genes [56], suggesting that chromosomal changes caused by the lack of proper DNA repair may be frequent. During the cultivation of the parental FaDu cell line, a certain amount of divergent cells may develop ( Figure 3C). We were able to generate thirteen sub clonal cell lines with divergent RAD9A methylation patterns (Figure 3 B). Two of the sub clones, 4 and 6 exhibit high methylation values (mean methylation 75% and 73% repectively) in comparison to the parental cell line (mean methylation 54%) while others showed a lower methylation level (e.g. clone 2 and 9). Despite different methylation levels of RAD9A, the methylation levels remained stable during cultivation for all clones. Clone 2, 4 and 9 and the parental cell line were chosen for further characterization.
Monolayer culture growth kinetics, for the sub cell clones 2, 4, 9 and the parental FaDu cell line performed using triplicates revealed significant delayed growth for the sub clone In order to elucidate the connection between loss of function of those important genes and hypermethylation of RAD9A we examined two fibroblast cell lines known to have homozygous mutations either in BRCA2 ( FANCD1) or in SLX4 ( FANCP1) in contrast to normal primary fibroblast cell lines (N = 20). As expected, control cell lines exhibited mean methylation values ranging from 3-11%, in contrast to FANCD1 (mean 28%) and FANCP1 (mean 47%).

Discussion
The vast majority or childhood cancers occurs sporadically and cannot be explained by inherited mutations in known tumor susceptibility genes [12]. Therefore, it seems plausible to assume that stochastic or adverse exposure events during early (intrauterine and postnatal) development increase cancer susceptibility through epigenetic reprogramming [16,17]. Consistent with the developmental programming of cancer hypothesis, we previously identified a monozygotic twin pair discordant for childhood cancer with a constitutive BRCA1 epimutation in the twin with cancer [23]. Here, we It has been previously shown that hypermethylation of RAD9A intron 2 is associated with mRNA and protein overexpression, which may be a critical step in the development of prostate, breast and other cancers [39,62]. According to the Cancer Genome Atlas/TGCA The methylation-sensitive region in intron 2 is endowed with three regulatory elements, annotated in ORegAnno (http://www.oreganno.org/). Element OREG1137234 is a binding side for the transcription factor ZNF263, which functions as transcriptional repressor. It is plausible to assume that intron 2 hypermethylation interferes with ZNF263 binding, which then activates RAD9A expression. The RAD9A-HUS1-RAD1 complex is recruited to sites of DNA damage and required for cell cycle arrest and DNA damage repair [63]. When translocated to mitochondria, RAD9A binds and neutralizes the anti-apoptotic activity of BCL-2 and BCL-xL proteins, thus promoting cell death [64,65]. However, permanent RAD9A overexpression may also have harmful, tumor-promoting effects. We propose that RAD9A methylation is an early (either stochastic or environmentally induced) event, which may lead to malignant transformation in normal body cells. This in turn could affect downstream genes that are involved in further deregulation during tumor development. Constitutive epimutations (allele methylation errors) which arise in early development are likely to be present in a mosaic state in different normal tissues of an individual. Although here we only analyzed primary skin fibroblasts, we assume that the identified RAD9A epimutations were also present in other tissues, particularly in those, which developed a tumor. Previously, we have shown that epimutations in BRCA1 and RAD51 can originate in single precursor cells [7,8]. It is difficult to define a threshold for constitutive epimutations in normal tissues that can be associated with tumor formation.
In our experience with TS epimutation screening >800 individuals [7,8], mean methylation values of 10% and allele methylation errors of 2-3% (depending on the gene and assay) are outside the normal methylation variation range (depending on the tissue origin analyzed). The error rate for copying DNA methylation patterns during DNA replication is estimated to be 10-100 times higher than for non-replicating DNA [22,66].
Therefore, rapidly dividing cells, i.e. stem cells during embryonal development and organogenesis, may be particularly vulnerable to acquiring methylation defects [67,68].
Although mutations and/or epimutations in a single cell may be potentially harmful, the body must be able to cope with a (very) small proportion of cells, which need to be corrected or eliminated. Nevertheless, the risk for malignant transformations may increase with the percentage of cells with compromised genomic caretakers. When the analyzed childhood cancer patients were recruited for this study, they all had survived an initial cancer treatment (in most cases including radiation therapy) and had been tumorfree for several years. Although we did not find evidence for irradiation or chemotherapy associated RAD9A epimutations, we cannot completely exclude the possibility that the observed RAD9A hypermethylation in some patients is a consequence of tumor treatment which leads to mutation of important genes. We propose that the constitutive RAD9A