KLK10 exon 3 unmethylated PCR product concentration: a new potential early diagnostic marker in ovarian cancer? - A pilot study

DNA methylation is one of the well-studied epigenetic modifications in DNA/chromatin metabolism. It is a dynamic process and involves the reversible and heritable methylation of the 5′ carbon of cytosine residues to yield 5-methylcytosine (5-MC) [1]. The reaction belongs to one carbon metabolism where DNA methyltransferases (DNMT1, DNMT3a and DNMT3b) are the biocatalysts and S-adenosylmethionine (SAM) is the methyl-donor [2]. Besides having important physiological roles in cell differentiation, development and gene regulation [3], DNA methylation can provide clues to other physiological processes, e.g. cell and tissue aging [4] and establishment of memory [5].

In DNA the cytosine residues occur either in frequencies that are far less than expected or in CpG-rich short stretches (0.5–4 kbp) in gene promoters and other regulatory regions known collectively as CpG islands [6]. In the CpG context the two cytosines on the opposing DNA strands are usually symmetrical as for their methylation status, i.e. both are either unmethylated or methylated [7]. The effect of DNA methylation on gene regulation may differ according to the context in which it occurs; however, in CpG-rich gene promoters it is well known to share in gene silencing [8]. Deregulated gene methylation was implicated in several diseases including cancer [9]. Nonetheless, aberrant gene methylation in cancer can be a promising diagnostic and prognostic target in tumor and naked DNA samples; e.g. in lung cancer [10].

The kallikrein-related peptidase 10 gene (KLK10) is one of 15 members in a serine protease gene subfamily located in tandem on chromosome 19q13.3–13.4 [11]. The gene product (KLK10) is a secreted protein found in normal human mammary epithelial cells (MECs) but downregulated or absent in the majority of human breast cancer cell lines [12]. KLK10 has a wide tissue expression [13] and is regulated by mechanisms that include steroid hormones [14] and micRNA [15]. The gene was reported as candidate tumor suppressor in some cancer types, e.g. in prostate cancer cells [16]; however, it showed contrasting expression profiles in different cancers, e.g. underexpressed in breast [17], testis [18] and prostate [19], and overexpressed in ovary [20], colon [21], and pancreas [22].

Liu and colleagues did not find mutation in KLK10 gene in different cancer types [12]. Indeed, tumor-specific lack of KLK10 expression correlated to KLK10 exon 3 hypermethylation in a majority of cell lines and in primary breast cancer [23]. The rationale behind such correlation may be understood as KLK10 exon 3 methylation was shown to perfectly follow that of KLK10 gene promoter [24]. Results of previous works were consistent about the confinement of KLK10 exon 3 hypermethylation in malignant samples; but not normal ones, in cell lines [23‐25] and in primary tumors [23‐26]. However, in some sample sets KLK10 expression did not show simple correlation with KLK10 exon 3 methylation [23‐26]. In the present study, KLK10 exon 3 methylation is assessed for its possible role in the biology, diagnosis and/or prognosis of ovarian tumors.

In the literature we could find only one study on KLK10 exon 3 methyaltion in primary ovarian tumors that was reported by Sidiropoulos and colleagues [25]. Our results agree with those in the previous work with some logic approximation due to differences in the used technique for gene methylation assessment (MSP vs. direct DNA sequencing, respectively) and the examined KLK10 expression parameter (serum KLK10 vs. cytosolic KLK10, respectively). The two studies showed KLK10 exon 3 methylation only in malignant samples (methylated-specific PCR positive samples: 1/16 and samples with partially methylated KLK10 exon 3: 6/7, respectively) but not in normal samples (methylated-specific PCR negative samples: 13/13 and samples with fully unmethylated KLK10 exon 3: 2/2, respectively). Additionally, we studied KLK10 exon 3 methylation in benign ovarian tumors – not included in the previous work – and they all were unmethylated (Table 3). Differences in the internal working of MSP and direct DNA sequencing would make comparison of frequencies of methylated samples in the two studies practically unfeasible (see later).

Moreover, as for the correlation between KLK10 gene expression and KLK10 exon 3 methylation the two studies showed similar results. In the previous work there was fair KLK10 expression in normal samples (fully unmethylated) and high expression in the partially methylated malignant samples (3 out of 6). In concordance to those results our samples showed negative correlation between serum KLK10 level (ng/ml) and KLK10 exon 3 unmethylated PCR product concentration (ng/μl) by Spearman’s correlation and by comparing the unmethylated PCR product concentration mean ± SD in overall samples according to KLK10 level (Table 5).

While our qualitative MSP results did not show statistically significant differences in patient group samples (Table 3), KLK10 exon 3 unmethylated PCR product concentration (ng/μl) mean ± SD showed statistically significant differences in normal and tumor samples in the two patient groups (Fig. 1). Because our MSP experiment conditions were almost constant for all PCR tests (including the operator, instruments and machines, kits and reagents, DNA template concentration, PCR program, simultaneous unmethylated- and methylated-specific PCRs, and random sample sets in 4 PCR runs done in 4 different days) those statistically significant differences of KLK10 exon 3 unmethylated PCR product concentration should be due to differences in the methylation patterns of samples rather than due to unfounded chance (see later).

Individual cells in the same tissue may show heterogeneity as regard their DNA methylation patterns [28]. Indeed, understanding that methylation pattern heterogeneity is fundamental in interpreting qualitative and quantitative DNA methyaltion data and also for appreciating and understanding several aspects of gene functions. Each cell contains two copies of each CpG site; one on each of the two homologous chromosomes, and, therefore, can have on its own 3 different methylation patterns for that particular CpG site: fully unmethylated, fully methylated or partially methylated. The possible methylation patterns of a number (n) of CpG sites are given mathematically as 2ⁿ. For example, when studying an average number of 12 CpGs by MSP, the possible methylation patterns in the DNA sample are 2¹² = 4096, which reflects possible immense methylation pattern heterogeneity when the cell can have any two of these patterns. However, in the context of KLK10 exon 3 in ovarian cancer the property of methylation pattern heterogeneity would be present almost exclusively in malignant samples but not in normal ones because in the previous work the malignant samples were partially methylated, whereas the normal samples were fully unmethylated [25].

In Fig. 3 we present a theoretical rationale for qualitative and quantitative MSP results based on possible methylation pattern heterogeneity in individual cells of a sample [28], the independency of the primer pairs in MSP from one another [23, 29] and previous DNA methylation results by MSP and direct DNA sequencing in cell lines and primary breast cancer [23]. We hypothesize that the existence of different partially methylated patterns in the sample would result in the stratification of the starting DNA template into template classes with varying MSP-expressivities according to the number of methylation-wise matching CpG sites with their corresponding nucleotides in the respective PCR primer. While some patterns would be more or less favorable for primer annealing (Fig. 3: elements C, and E-J), other patterns may not be at all (Fig. 3: element D). Differences of the proportions of those differentially MSP-expressible partially methylated patterns in the samples may be effected practically as differences in the unmethylated PCR product concentration. Given that those partially methylated patterns would occur only in malignant samples but not in normal – and probably also benign – samples (see above), such proportionate decreases in the unmethylated PCR product concentration would happen in malignant samples only.

Our results showed statistically significant positive correlation between KLK10 exon 3 unmethylated PCR product concentration (ng/μl) and patient age by Spearman’s correlation and on comparing its mean ± SD in overall samples and in malignant samples according to age groups (Table 4). However, because there was no statistically significant difference of patient age in the two patient groups (Table 1), comparison of KLK10 exon 3 unmethylated PCR product concentration in patient group samples should be valid and justifiable.

KLK10 exon 3 methylation correlated to high disease grade (breast cancer [30]), late stage (lung cancer [26]) and bad patient prognosis (acute lymphoblastic leukemia [24]). Similarly, the only methylated-specific PCR positive sample in our results was serous carcinoma of high grade (3) and late stage (III/IV) whose patient died within one year after starting the treatment. However, in our results comparison of KLK10 exon 3 unmethylated PCR product concentration (ng/μl) according to disease grade, stage or patient survival did not show statistically significant differences that may be attributable to early implication of deregulated KLK10 exon 3 methylation in the disease or lack of significant information about site-specific CpG methylation in MSP data (see later).

The notion that information about the exact location of DNA methylation maybe crucial for correct clues about disease diagnosis, prognosis and gene function [31] would greatly restrain the translational impact of DNA methylation data by MSP. For example, lack of site-specific CpG methylation data (in MSP) and/or data about their pattern distribution in the sample (in MSP and direct DNA sequencing, respectively) may explain – at least in part – the inconsistent results, in sample subsets or sample types, about the downregulation of KLK10 through KLK10 exon 3 hypermethylation in previous works [23‐26]. At the functional aspect, the reported antitumor role of KLK10 was demonstrated in vector-mediated transfection experiments in vitro and/or in tumor xenograft in mice using different cancer cell lines; e.g. breast [32], prostate [16] and gastric cell [33], including ovarian cell line [34]. Noteworthy is that overexpression of KLK10 was found in malignant sample subsets (tissues or sera) in cancer types in which KLK10 was shown experimentally as tumor suppressor; breast [35], prostate [19], gastric [36] and ovary [20, 27]. Taken together, the antitumor role of KLK10 may be understood only in relation to the context of the given study as regard type of tissue and the underlying pathophysiologic processes [37].

Indeed, the translational impact of DNA methylation data can vary greatly with the pros and cons of the used tools. While MSP studies only a few of CpGs without regard to their exact location, direct DNA sequencing can give site-specific CpG methylation data for all CpGs of interest as percent methylation in the sample. Unfortunately, both tools fall short of providing information about site-specific CpG methylation pattern distribution in sample. The more recently discovered next generation sequencing have the advantage of presenting site-specific DNA methylation data in absolute numbers and may tackle the problem of methylation pattern distribution in hand of sophisticated computational methods [38]. Nonetheless, considering possible dynamicity of DNA methylation as a mechanism for gene regulation [1] another dimension for DNA methylation pattern resolution may be necessary in order to spot and trace such spatiotemporal DNA methylation events rather than making statements based merely on single static snapshot image.

In summary, our results agreed with those in the previous work as for the exclusive occurrence of KLK10 exon 3 methylation in malignant but not in normal ovarian samples. Additionally, we assessed KLK10 exon 3 methylation in benign ovarian tumors (not included in the previous work) and found that they were unmethylated. Although our qualitative MSP results did not show statistically significant differences in patient group samples, semi-quantitative KLK10 exon 3 unmethylated PCR product concentration (ng/μl) showed statistically significant differences not only in tumor samples but also in otherwise histologically normal samples in benign and malignant patients. These findings would suggest, for the first time, KLK10 exon 3 unmethylated PCR product concentration as potential early epigenetic diagnostic marker in ovarian cancer. The diagnostic potential of KLK10 exon 3 unmethylated PCR product concentration in overall patient group samples was also evident in ROC curve analysis, and it could enhance the sensitivity of preoperative serum markers CA125, KLK10 and KLK6. Taking into account the limitations in the present study (small sample size and semi-quantitative assessment of PCR product concentration) further studies are needed before making a statement about the validity and reproducibility of these results. Moreover, we highlighted the translational limitations of DNA methylation data in MSP and direct DNA sequencing and addressed future challenges of the more recent next generation sequencing (NGS) to further enhance the translational impact of DNA methylation data.