Prognostic Value of PLAGL1-Specific CpG Site Methylation in Soft-Tissue Sarcomas

Soft tissue sarcomas (STS) are rare, complex tumors with a poor prognosis. The identification of new prognostic biomarkers is needed to improve patient management. Our aim was to determine the methylation status of the 118 CpG sites in the PLAGL1 tumor-suppressor gene P1 CpG island promoter and study the potential prognostic impact of PLAGL1 promoter methylation CpG sites in STS. Training cohorts constituted of 28 undifferentiated sarcomas (US) and 35 leiomyosarcomas (LMS) were studied. PLAGL1 mRNA expression was investigated by microarray analysis and validated by RT-qPCR. Pyrosequencing was used to analyze quantitative methylation of the PLAGL1 promoter. Associations between global promoter or specific CpG site methylation and mRNA expression were evaluated using Pearson’s product moment correlation coefficient. Cox univariate and multivariate proportional hazard models were used to assess the predictive power of CpG site methylation status. Sixteen CpG sites associated with PLAGL1 mRNA expression were identified in US and 6 in LMS. Statistical analyses revealed an association between CpG107 methylation status and both overall and metastasis-free survival in US, which was confirmed in a validation cohort of 37 US. The exhaustive study of P1 PLAGL1 promoter methylation identified a specific CpG site methylation correlated with mRNA expression, which was predictive for both metastasis-free and overall survival and may constitute the first US-specific biomarker. Such a biomarker may be relevant for identifying patients likely to derive greater benefit from treatment.


Introduction
Sarcomas are uncommon tumors of mesenchymal origin, responsible for approximately 1-2% of adult cancers. They are currently classified according to their differentiation line. However, three main groups of STS have been described on a genetic level: a simple genetic profile, exclusively composed of well/dedifferentiated liposarcomas [1][2][3], a complex genomic profile observed in leiomyosarcomas (LMS) and undifferentiated sarcomas (US) [2,4,5], and a profile characterized by translocations [6]. These genetic alterations induce tumor development by targeting tumor suppressors (p53, pRb) [7,8] and genes related to ploidy status and proliferation.
It has now been clearly established that changes in DNA methylation occur early in tumorigenesis. Cancer cells are characterized by global DNA hypomethylation, which promotes chromosomal breaks and rearrangements, and the silencing of tumor-suppressor genes, based on hypermethylation of their promoters [9] [10].
PLAGL1 is a maternally-imprinted gene, which maps to the human chromosome 6q24.2, a region often rearranged in many cancers [11][12][13], including sarcomas [14]. PLAGL1 was first described as a tumor-suppressor gene as it showed antiproliferative properties and shared the ability to concomitantly regulate apoptosis and cell-cycle arrest with p53 [15] [16,17]. However, recent studies have described an oncogenic role of PLAGL1 in glioblastomas [18] and rhabdomyosarcomas [19], suggesting that PLAGL1 functions depend on the cell context. Except in peripheral blood cells [20], PLAGL1 expression is conditioned by methylation of a promoter (P1) containing a CpG island consisting of 118 CpG sites [21]. Loss of PLAGL1 expression, due to either loss or hypermethylation of the active paternal allele, has been reported in various tumors [22].
The main treatment of STS combines primary radical surgery, chemotherapy, and radiation. Patient therapy depends mostly on the stage of the disease, however, the 5-year survival rate for STS patients is approximately 50%. Among STS subtypes, LMS and US exhibit the highest metastatic relapse rate and, consequently, the poorest prognosis. Therefore, it is necessary to develop new prognostic molecular biomarkers for future patient management. DNA methylation has been proposed as an alternate pathway to gene deletion and mutation-induced tumors. Surprisingly, in STS, while extensive studies have explored genetic alterations, the methylation status of specific genes involved in oncogenesis have yet to be elucidated [23][24][25]. Even if PLAGL1 is frequently altered in various cancer types, to date, no study has extensively investigated the possible prognostic role of PLAGL1 promoter methylation status in STS and its possible association with cancer patient outcome.
This work investigated the putative clinical prognostic role of specific CpG-site methylation levels of the PLAGL1 promoter in leiomyosarcomas and undifferentiated sarcomas. A strategy was developed for performing an exhaustive study of the methylation levels of the 118 CpG sites on the PLAGL1 P1 CpG island using pyrosequencing. This approach determined the methylation profile of each CpG site of the P1 CpG island and identified specific CpG sites particularly correlated with PLAGL1 mRNA expression and clinical outcomes.

Patients and ethics statement
The initial study cohort of 63 sarcoma samples included 35 LMS and 28 US, and a second cohort of 37 US was used to validate the prognostic value of CpG sites. All samples were primary tumors from patients who had not been previously treated with radiotherapy or chemotherapy and exhibited no other diagnosed cancers. All tumors were provided by the French Sarcoma Group (GSF) database, part of the Conticabase (www.conticabase.org), containing data from adult soft tissue sarcomas [26]. All samples studied were freshfrozen tumors which exhibited a high tumor-cell content (>80%). All cases were histologically reviewed by the pathologist subgroup and classified according to the 2002 WHO classification.
The samples used in this study formed part of the Biological Resources Center of the Institut Bergonié Comprehensive Cancer Center (CRB-IB). In accordance with the French Public Health Code (articles L. 1243-4 and R. 1243-61), the CRB-IB has been approved by the French authorities to deliver samples for scientific research (number AC-2008-812, on February 2011). These patient samples were requalified for research. The patients signed an informed consent form approved by the Committee for the Protection of Individuals (CPP).

PLAGL1 mRNA expression analysis
The expression of PLAGL1 mRNA in sarcoma samples was studied using data obtained with the Human Genome U133 plus 2.0 array (Affimetrix) by Chibon et al. [26] and deposited at Gene Expression Omnibus: accession number GSE21050. PLAGL1 mRNA expression was assessed by three probe sets which were found to be highly correlated (> 0.97 Pearson's Correlation Coefficient, Figure S1). The mean of the GCRMA normalized expression values of the three probes was calculated and used to identify associations with the methylation of CpG sites and clinical factors.
PLAGL1 mRNA expression data were then confirmed by RT-qPCR in tumors where total RNA was available. For RT-qPCR, 400 ng total RNA were reverse transcribed in a 20 µl mixture to generate first-strand cDNA using the high-capacity cDNA reverse-transcription site kit (Applied Biosystems) according to the manufacturer's instructions. Reverse transcriptions were performed in triplicate. All qPCR were performed in a 20 µl final volume containing 20 ng cDNA preparation using power Sybrgreen PCR master mix (Applied Biosystems) according to the manufacturer's instructions and the Step-One plus real-time PCR system (Applied Biosystems). Primers for PLAGL1, BACTIN, and PRPL0 were designed using Primer 3 software (Table S1). BACTIN and PRPL0 were used for data normalization.
Correlations between mRNA expression values from microarray and RT-qPCR data were evaluated with GraphPad Prism v4 software, using Spearman's product-moment correlation coefficient.

CGH array
Genomic profiling was performed using a DNA microarray developed in our laboratory. Three thousand eight hundred seventy-four BAC/PAC DNAs (BACPAC Resources Center, Children's Hospital, Oakland Research Institute) were spotted in triplicate on ultraGAPS slides (Corning). The probes were prepared and hybridized as previously described [27]. The data were analyzed with software developed at Institut Curie (CAPweb, http://bioinfo-out.curie.fr/CAPweb/). Cyanine-5/ cyanine-3 ratios between 0.8 and 1.2 were considered as normal status and ratios >1.2 and <0.8 were considered as gains and losses, respectively. Analysis of array-CGH (computation of genomic alterations) was provided by the VAMP interface (http://bioinfo.curie.fr/vamp) [28].
Genomic DNA was isolated using a standard phenolchloroform extraction protocol. DNA samples (100-500 ng), including positive controls for methylated and unmethylated status (EpiTect PCR control DNA; Qiagen, Hilden, Germany), were modified by sodium bisulfite treatment, using the Epitect bisulfite Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. Bisulfite-modified DNA was then used as a template for PCR amplification.
PCR was carried out as follows: 1X PCR buffer, 200 μM of each dNTP, 1.5 to 2.5 mM MgCl 2 , 10-50 ng DNA, 200 nM of each primer, and 2 units Platinium Taq Polymerase (Invitrogen, CA). Initial denaturing for 3 min at 94°C was followed by 50 cycles, starting with a step at 94°C for 20 s. Different annealing temperatures were used according to the primers for 20 s (Table S1), followed by a 30-45 s step at 72° C, with a final extension at 72° C for 7 min. Primers used for PCR and sequencing are listed in Table S1.
Amplification and sequencing primers were designed for the bisulfite-converted DNA using PyroMark Assay Design Software 2.0 (Qiagen) and synthesized by Eurogentec (France). To account for DNA fragmentation introduced by bisulfite treatment, 4 different amplification reactions (A0 to A3) were designed to cover all 118 CpG sites. Since the optimum pyrosequencing reading length is 60-120 nucleotides per primer, more than one sequencing primer was needed to cover the multiple CpG sites contained in the various amplicons ( Figure S2, Table S1). Pyrosequencing was performed using the PyroGold kit and PSQ 96 ID instrument (Qiagen), as directed by the manufacturer. CpG sites were quantified using Pyro Q-CpG 2.0 methylation software (Qiagen).
The coefficient of variance for methylation levels of CpG sites was calculated using demethylated and hypermethylated control DNA supplied by the manufacturer. As expected, it was < 2%.

Statistical analysis
Associations between global promoter or specific CpG site methylation and mRNA expression were evaluated with R software, version 2.13.0, using Pearson's product-moment correlation coefficient.
A screening analysis was performed using moderated t-tests with the p-value adjusted according to the Benjamini and Hochberg method for False Discovery Rate (FDR) and the R/ Bioconductor package limma [29] to identify CpG sites associated with clinical factors in the 28 US and 35 LMS from the tumor training sets. The association of CpG-site methylation with clinical prognosis, such as overall (OS) and metastasis-free survival (MFS) was confirmed using Cox univariate analysis. Duration of OS was calculated from the date of diagnosis until the date of last follow-up or death from any cause. MFS was calculated from the date of first diagnosis until the date of first metastasis. Patients who were alive or dead without metastasis at the last follow-up were censored. All variables significant at p<0.05 in the univariate analyses were included in stepwise, ascending, Cox-regression models. To validate the prognostic value of previously-identified CpG sites in terms of OS or MFS, Cox univariate analyses were performed on a second cohort (validation set) of 37 US with a pre-fixed cut-off, defined as the median methylation percentage of the first cohort. SPSS software version 16.0 was used for statistical analyses. Survival curves were computed by the Kaplan-Meier method. P ≤ 0.05 was considered significant.

Clinical characteristics of patients
Primary STS tumors were obtained before any treatment. Thirty-five LMS and 28 US, constituting the training sets, were randomly selected from 52 LMS and 65 US included in the Chibon et al., study [26]. The remaining 37 US were studied as a validation set. The clinical characteristics of the samples are described in Table 1. In LMS, median OS was 37 months (range: 4-296 months) and median MFS was 15 months (0-92 months). Tumor grade, size, and differentiation stage were not associated with MFS or OS of LMS. In US, median OS was 40 months (range: 2-203 months) and median MFS was 13 months (from 0 to 52 months). Median OS in the US validation set was 28 months (4-166 months) and median MFS was 17 months (range: 0-166 months). In both cohorts of US tumors, tumor grade and size were not associated with MFS or OS (Tables 2 and 3). In the three cohorts studied, patient age was significantly associated with overall survival (in LMS: P=0.025, HR=2.75; in US-training set: P=0.021, HR=4.3; in US-validation set: P=0.034, HR=3.9) (Tables 2 and 3). The differentiation stage could not be tested due to small number of patients with low differentiation scores.
Tumors were also classified according to the CINSARC (Complexity INdex in SARComas) scoring system [26] where C1, the lower CINSARC score, indicates a good prognosis and C2, the higher CINSARC score, corresponds to a poor prognosis. In LMS, CINSARC was associated with MFS (P=0.003, HR=9.8). In the US training set, CINSARC was associated with MFS (P=0.025, HR=5.8) and OS (P=0.047, HR=3.8). In the US validation set, CINSARC appeared to be associated with OS, although the values were not statistically significant (P=0.12, HR=5), but it was impossible to confirm the association with MFS due to the absence of metastasis in the C1 group (Tables 2 and 3).

PLAGL1 expression in STS
PLAGL1 mRNA expression was analyzed in 35 leiomyosarcomas (LMS) and 28 undifferentiated sarcomas (US), using the data obtained by Chibon et al [26] with the Human Genome U133 plus 2.0 array (Affimetrix). Prior analysis, we confirmed array-based PLAGL1 mRNA expression measures by RT-qPCR. Good correlation was observed between microarray expression data and RT-qPCR (Spearman r=0.93, p<0.0001 normalized by BACTIN and r=0.86, p<0.0001 normalized by RPLP0) (Figure S3), thus validating the microarray PLAGL1 mRNA expression values.
PLAGL1 mRNA expression was not significantly different in LMS and US in terms of the median (9.07 in LMS and 8.96 in US) or heterogeneity of the expression values, which ranged from 2.6 to 12.3 in LMS and from 3 to 9.8 in US. The wellknown high level of chromosome alterations in sarcomas like LMS and US may explain this heterogeneity. To verify this hypothesis, we studied the correlation between PLAGL1 mRNA expression and its genomic status, obtained by CGHarray. No significant alteration in the genomic profile of the locus containing the PLAGL1 gene (6q24.2) was detected in US and, consequently, no correlation between the genomic profile and mRNA expression of PLAGL1 was observed in these sarcomas (Pearson=0.131 P=0.533). In contrast, 6 out of 35 LMS exhibited chromosome alterations: PLAGL1 was deleted in 2 tumors (ratio<0.8) and gained in 4 tumors (ratio>1.2) ( Figure  1). These alterations correlated with PLAGL1 mRNA expression, since PLAGL1 expression increased in tumors exhibiting gene gain and decreased in those with gene deletion (Pearson= 0.354 P=0.043). Nevertheless, this phenomenon alone did not account for the heterogeneity of PLAGL1 mRNA expression observed in all LMS, since a majority of the tumors had no gene alteration but exhibited a large range of PLAGL1 mRNA expression (range 3.5 to 12.3).
Thus, we investigated whether PLAGL1 expression in STS was regulated by methylation of its promoter, which would account for its heterogeneity.

Promoter methylation and PLAGL1 mRNA expression
Pyrosequencing was used to study the methylation status of all 118 CpG sites on the PLAGL1 P1 CpG island promoter. We found that the mean methylation percentage of the PLAGL1 P1 CpG island (mean methylation percentage of all CpG sites) was not significantly different in LMS and US in terms of the median methylation value (27.5±7% in LMS and 30.4±4.6% in US). However, a pronounced variation was observed in mean methylation values in the tumors, ranging from 13.4% to 41.1% in LMS and 20.7% to 40.7% in US. Thus, the next stage was to assess the correlation between the mean methylation percentage of the P1 CpG island and PLAGL1 mRNA expression in US and LMS. No significant correlations were found in either tumor type (Pearson=-0.12, P=0.53 for US; Pearson=-0.019, P=0.91 for LMS) (Figure 2 Aa, Ba). Analysis of the percentage methylation of each CpG site on the P1 CpG island revealed highly variable values between two consecutive CpG sites in the same tumor and between the same CpG site in different tumors ( Figure S4). This suggested that methylation of specific CpG sites may be particularly correlated with PLAGL1 mRNA expression, rather than global PLAGL1 promoter methylation.
Statistical analysis revealed that the methylation status of 6 and 16 out of the 118 CpG sites correlated with PLAGL1 mRNA expression in LMS and US, respectively ( Figure 2) ( Table 4). As expected, in LMS, these CpG sites were negatively correlated with mRNA expression (Pearson=-0.50, P=0.0022) (Figure 2Bb (Figure 2Ac). These results suggest that the methylation status of only a few of the 118 CpGs in the P1 CpG island is relevant for mRNA expression of PLAGL1 regulation in STS and that these CpG sites are tumortype-specific.

Prognostic value of PLAGL1 promoter methylation and mRNA expression
Investigation of a putative association of promoter methylation with STS clinical prognosis revealed that the mean methylation value of the PLAGL1 promoter CpG island had no prognostic value in US (P=0.88, HR=1.  (Table 2) and between OS and CpG 107 and 110 (Table 3). In LMS, the association between CpG19 and OS found during screening analysis was not confirmed with Cox univariate analysis (P=0.067), suggesting that PLAGL1 P1 CpG site methylation is not prognostic in LMS. Methylation levels of significant CpG sites in a validation set of 35 US were studied to validate the association between CpG site methylation and clinical prognosis. The cut-off determined in the training set, corresponding to the median methylation level of each significant CpG site, was applied to the validation set. Associations between CpG107 and both MFS (P=0.001) ( Table  2) and OS (P= 0.001) ( Table 3) were confirmed in the validation set by univariate Cox analysis. The Kaplan-Meier method was used to produce survival curves stratified by DNA methylation (Figure 3). Low CpG107 methylation levels (<35%) were associated with poor prognosis for both MFS and OS in US. The methylation status of CpG107 showed significant concordance with mRNA expression in both cohorts (Pearson=-0.46, P=0.013 in training set, Pearson=-0.58, P=0.0002 in validation set) ( Figure 4A). When analyzed by group, the low-methylation group consisted of a subset of samples with high mRNA expression, compared to the highmethylation group (unpaired t-test P=0.027 and P=0.0045 in training and validation set, respectively) ( Figure 4B). Investigation of the prognostic value of PLAGL1 mRNA expression in US revealed significant associations between PLAGL1 mRNA expression and both overall and metastasisfree survival in the training set, but these results were not confirmed in the validation set (Tables 2 and 3).
Given that more variables were significantly associated with OS and MFS in the training set than in the validation set, multivariate analyses were performed on the US training set to determine whether CpG107 was a significant independent prognosticator for MFS or OS. For MFS, CpG49 and CpG107 were significant markers in the multivariate analysis including CpG8, CpG46 PLAGL1 mRNA expression and CINSARC ( Table 2). For OS, CpG110 and age were significant markers in the multivariate analysis including CINSARC, PLAGL1 mRNA expression, and CpG107 (Table 3).

Discussion
LMS and US are rare, complex diseases and their pathogenesis is still poorly understood. They exhibit a high metastatic relapse rate leading to poor prognosis. The identification of new metastatic and survival prognosis markers is important for their management. The aims of this study were to evaluate the importance of individual PLAGL1 promoter P1 CpG sites for PLAGL1 mRNA expression and their potential association with sarcoma patient outcomes.
In this study, we identified CpG107 methylation status of the PLAGL1 P1 promoter as the first prognostic biomarker for both  In US and LMS, transcriptomic data analysis revealed a high variability of PLAGL1 expression, not entirely explained by amplification or deletion of the gene. This indicated that the methylation status of the P1 PLAGL1 promoter in US and LMS may be responsible for the heterogeneity of PLAGL1 expression in STS. The exhaustive study of the P1 PLAGL1 promoter CpG island using the pyrosequencing method revealed the considerable heterogeneity of the PLAGL1 methylation profiles of CpG sites along the CpG island and for a given CpG site in different tumors. Accordingly, we showed that the mean methylation percentage of all PLAGL1 P1 CpGs was not correlated with mRNA expression levels, while methylation of only a few specific CpG sites was correlated, their number and localization varying according to the STS subtype studied. These data showed that most CpGs exhibit a large range of methylation compatible with mRNA expression values (Table 4). However, in US, some CpGs (CpG27, 29, 30, 32), located just before the transcription start site (TSS) (CpG32), exhibited a narrow range of methylation (0 to 20%). The specific promoter regions, called core regions, often situated around the TSS within a CpG island, have been shown to be crucial for regulating gene expression [30,31] and are probably more sensitive to small variations in methylation. Surprisingly, this study also revealed that 3 CpG sites (CpG45, 51, 58) were positively correlated with mRNA expression in US. This may be due to the fixation of transcriptional repressors since in silico studies using the TFsearch and PROMO databases identified CpG 45 and 51 as putative binding sites for P53, PAX5, and SP1, which may act as transcription repressor factors [32][33][34]. These results confirmed, as described in previous studies [35,36], that specific CpG sites were particularly relevant for directed mRNA expression. Importantly, these findings revealed that the CpG sites differed from one STS subtype to another, suggesting that the methylation modification of certain CpG sites may be a critical event for tissue-specific expression of PLAGL1. This may constitute a specific methylation signature, possibly resulting from a difference in tumor microenvironment [37] or differentiation lineage. These results demonstrate the relevance of performing methylation studies according to histological subtypes rather than in STS in general, even if large patient cohorts are difficult to constitute.
Epigenetic alterations are now clearly defined as early events during oncogenesis. Tumor-specific epigenetic alterations act as early, relatively stable molecular markers of malignancy, leading to better diagnosis, prognosis and therapy [10]. Investigation of the prognostic value of the methylation status of the PLAGL1 promoter CpG island, showed that, as expected, the mean methylation value of the PLAGL1 promoter had no prognostic value in either of these STS sub-types, possibly due to a lack of correlation between mean methylation value of the PLAGL1 promoter and PLAGL1 mRNA expression. In contrast, the screening analysis of the association between each CpG site's methylation value and clinical outcomes, followed by univariate Cox analyses, revealed that PLAGL1 promoter methylation was a tumor typespecific biomarker. Indeed, while no CpG site was found to have any prognostic value in LMS, methylation of CpG107 was identified and validated as a prognostic factor for MFS and OS in US. Moreover, multivariate analysis showed that CpG107 methylation status was an independent prognosticator for MFS, suggesting that it is probably a reliable prognostic biomarker for US. Interestingly, CpG107 is one of the most strongly correlated with mRNA expression, suggesting that it is likely to be an important site for the transcriptional regulation of PLAGL1 in US. The relevance of a specific, single, CpG site in mRNA expression and/or predictive of prognosis has recently been reported in various cancers. In liver cancer, the hypermethylation of one CpG silenced transcriptional expression of TTP [36] and, in chronic lymphatic leukemia, a single CpG dinucleotide has been identified to be important for ZAP-70 expression and prognosis [38]. Thus, these data highlight the interest of studying the detailed mapping of methylation patterns within a CpG island, as they revealed the complexity of gene regulation by the DNA methylation process and the relative importance of some CpG sites as clinical biomarkers. Low methylation (<35%) of CpG107 was correlated with significantly higher PLAGL1 expression and poor prognosis. This is consistent with high PLAGL1 mRNA expression having an adverse effect on survival in the training set. Together, these data suggest that the role of PLAGL1 in US is more consistent with the phenotype of an oncogenic than a tumorsuppressor gene. Although PLAGL1 was initially described as a tumor-suppressor gene, recent studies have shown its oncogenic role in glioblastomas [18] and rhabdomyosarcomas [19]. Moreover, the embryonic growth-retardation phenotype observed in ZAC1-deficient mice is unexpected for a putative tumor-suppressor gene [39] and more consistent with the phenotype of an oncogenic gene. Thus, PLAGL1 functions apparently depend on the cell context and the biological role of PLAGL1 in STS has yet to be established.
This study confirmed that the CpG-site methylation had a better prognostic value than mRNA expression since, in contrast to CpG site methylation, the prognostic value of mRNA expression was not confirmed in the validation set. Indeed, the clinical impact of biomarkers is dependent on the stability of the information provided. While DNA methylation is a covalent modification of the cytosine base and remains relatively stable in vitro and in vivo over time [40], RNA and protein expression results from a balance of distinct factors, including RNA polymerases, transcriptional activators and repressors, miRNA, and other processes leading to a significant variability in gene expression. Thus, recent studies have reported that DNA methylation analysis is a more accurate predictor of outcome than expression [38,41,42]. Finally, the possibility cannot be excluded that the methylation pattern of PLAGL1 has an impact on the expression of other genes, for example by sequestering transcription factors. In conclusion, this research identified a specific CpG site where determination of the methylation status was associated with both metastasis-free and overall survival in undifferentiated sarcoma. Current guidelines recommend doxorubicin and iphosphamide-based chemotherapies for patients with STS, while their clinical outcomes remain quite diverse. In this context, the identification the methylation status of a single CpG may provide a useful epigenetic biomarker of primary US at increased risk of relapse and metastasis, thus improving patient management by selecting patients who would benefit from this treatment. Moreover, the methylation study of a single CpG site dramatically simplifies analysis, thus facilitating its application in pathology laboratories.  figure). The asterisk identifies the transcription start site. B: Illustration of pyrosequencing assays. CpGs contained in the CpG island P1 of PLAGL1 are shown in bold underlined type. Each CpG is identified by a superscript number. PCR and sequencing primers are symbolized by full and dotted arrows, respectively. Four PCR were necessary to sequence the entire CpG island. The first assay required two sequencing primers (in blue, for dinucleotides CpG1 to 15). The second assay required three sequencing primers (in red, for dinucleotides CpG13 to 48). The third assay required four sequencing primers (in green, for dinucleotides CpG44 to 92). The fourth assay required four sequencing primers (in purple, for dinucleotides CpG77 to 118). b: biotinylated primer. (TIF)