Methylation pattern of tumor-suppressor gene promoters as putative noninvasive diagnostic markers for prostate cancer

Aim. To assess the rate of promoter methylation of putative TSGs for PCa in tumor tissue and in urine of PCa patients for better understanding of regulation of gene expression upon the PCa development and to evaluate the possibility to use the data on TSGs’ methylation for the development of noninvasive PCa markers. Methods. A quantitative methyl-specific PCR (qMSP) was used for the analysis of a methylation rate in prostate tissues and cell lines, and an ordinary MSP was performed for the study of urine samples. Results. We found that the RASSF1A promoter demonstrated a higher methylation rate in the TMPRSS2:ERG fusion positive PCa. The methylation of NKX3.1, PTEN and RASSF1A in DNA from urine was more common for cancer patients than for healthy donors. The promoters of CDH1 and GDF15 were methylated more frequently in PCa patients, than in patients with inflammatory disease. Conclusions. The abovementioned five genes can form a panel for early non-invasive detection of PCa. This set can be combined with the detection of the TMPRSS2:ERG fusion transcript. More work should be done to understand the molecular mechanisms explaining the functional role of promoter methylation of the selected genes.


Introduction
Prostate cancer (PCa) is one of the most abundant malignancies among men worldwide [1]. In Ukraine, the number of PCa new cases in 2018 was estimated as 7 936 (Cancer today https://gco.iarc.fr/). For now, the blood-based PSA (Prostate Specific Antigen) test is the most widely used approach for prostate cancer screening but it shows rather low sensitivity and specificity (75 %-85 % and 25-35 %, respectively) [2]. Of note, tumors of a prostate gland are highly heterogeneous, demonstrating histological, cellular and molecular differences even within one tumor from the same patient [3,4]. Therefore, it is important to seek additional PCa diagnostic markers of high sensitivity and specificity as well as the markers for estimation of disease severity and prediction of its progression.
The aberrant methylation of promoters of tumor-suppressor genes (TSGs) is one of the earliest events in cancer development and, therefore, the its detection can be used for early cancer diagnosis. However, many various cell populations with different methylome can be found in PCa lesions. Due to this fact, the quantitative analysis of gene methylation patterns and comparison with clinical tumor characteristics are needed to shed the light on a role of methylation events in PCa development [5,6].
The methylation patterns can be easily assessed in liquid biopsies [7,8]. Combining the investigations on PCa patients, patients with inflammation in the prostate gland and the healthy individuals, it is possible to study the impact of promoter methylation of defined TSGs on PCa development, and the optimal set of potential markers can be created.
In the present paper, we analyze the patterns of promoter methylation of a set of TSGs that was investigated previously in our laboratory [9]. Among them there are the genes, which are involved in cytoskeleton integrity (KRT18, VIM, CDH1, RASSF1A), cell division and signaling (RASSF1A, PTEN, NKX3.1), and other cancerrelated cellular processes [9][10][11][12][13][14]. KRT18 encodes keratin 18 [10], participating in maintenance of epithelial cell integrity, together with E-cadherin, encoded by CDH1 [11]. Vimentin, encoded by VIM gene, oppositely, represents the type III intermediate filament, which is more common for the mesenchymal cells type [12]. The PTEN gene encodes the protein PTEN, the well-known inhibitor of phosphatidylinositol 3-kinase pathway, involved in carcinogenesis [13]. The proteins, encoded by RASSF1A and NKX3.1 have multiple functions and were reported to participate in malignant transformation [14,15]. In our previous work we described the differential expression of those genes in PCa compared to the paired conventionally normal tissues (CNT), which pointed to their potential role in development of PCa, and, especially, in epithelial-mesenchymal transition [9]. GDF15 was included in our analysis as the cancer-related gene, according to the previously reported data [16].
Hence, the aim of the present study was to assess the rate of promoter methylation of the set of genes in tumor tissue and in urine of PCa patients for better understanding of regulation of gene expression upon the PCa development and to evaluate the possibility to use the data on TSGs' methylation for the development of noninvasive PCa markers.

Materials and Methods
Cell culture. Three human PCa cell lines -PC3, DU145 and LNCaP were grown in a DMEM medium, supplemented with 10 % FBS at 37 °C in a humidified atmosphere of 5 % CO2.
Sample collection and primary processing. All samples were collected according to the Declaration of Helsinki and approved by the guidelines issued by the Ethic Committee of the Institute of Urology and National Cancer Institute, Kyiv, Ukraine.
Tissues. The PCa samples and the paired CNTs were collected at the Institute of Urology of National Academy of Medical Sciences of Ukraine and National Cancer Institute (NCI) (Kyiv, Ukraine) as described earlier [9]. The most important clinical and pathological characteristics (CPC), such as Gleason score (GS), PSA levels in blood, tumor stage, patient age and the presence of the TMPRSS2:ERG fusion transcripts in cancer samples are presented in S1.
Urine. Urine samples of PCa patients were collected at National Cancer Institute (NCI) (Kyiv, Ukraine). The urine samples of patients with inflammatory disease of genitourinary system as well as the samples of healthy donors were collected at the Institute of Urology of National Academy of Medical Sciences of Ukraine. The urine from cancer patients was taken before surgical treatment into sterile containers and immediately transported to Institute of Molecular Biology and Genetics.
Primary processing of urine. All samples were centrifuged at 1500 rpm for 20 minutes (at +4 °C) to collect all cells. The supernatant was replaced to other tubes and the CTAB reagent was added for precipitation of DNA. After incubation at +4 °C (until visually detected cloudy precipitate was formed in urine) the samples were centrifuged at 2000 rpm for 15 minutes. The supernatant was discarded and a pellet of CTAB and DNA was stored at -20 °C until the DNA isolation [17].

DNA isolation
DNA from cells, tissues and urine were extracted according to the protocol [18].

The bisulfite conversion and PCR
Bisulfite conversion of DNA was performed, using the EZ DNA Methylation lightning kit (Zymo Research, USA), according to the manufacturer's instructions.
Quantitative methylation-specific PCR (qMSP). The primers for methylated and unmethylated promoter sequences were designed, using MethylPrimer software, the sequences and product sizes are listed in the S2. qMSP was performed, using the HOT FIREPol® EvaGreen® qPCR Mix (Solis BioDyne), according to the manufacturer's instructions using the CFX96 Touch Real-time PCR Detection System (Bio-Rad, USA). Reaction conditions were 95 °C for 12 min, 40 cycles of dissociation at 95 °C for 20 sec, annealing at 60 °C for 20 sec and elongation at 72 °C for 20 sec, followed by melting of PCR products from 65 °C to 95 °C increment 0.5 °C, 5 msec, 60 repeats. The Col2A1 gene was used as a reference gene. The quantification of relative amount of methylated and unmethylated forms of gene promoters was performed, using 2 ΔCt method, where ΔCt=Ct(Col2A1) -Ct(gene of interest), meth-ylated or unmethylated. Then the methylation rate for each gene was calculated by the equation [relative amount of methylated gene]/ [relative amount of methylated+relative amount of unmethylated gene].
Methylation-specific PCR (MSP). PCR of bisulfite treated DNA from urine was performed, using the 5xFIREPol® Master Mix Ready to Load (Solis BioDyne). Reaction conditions were: 95 °C for 12 min, 40 cycles of dissociation at 95 °C for 20 sec, annealing at 60 °C for 20 sec and elongation at 72 °C for 20 sec, elongation 72 °C for 7 minutes.
Agarose gel electrophoresis. All samples after MSP were analyzed by the electrophoresis, using 2.5 % agarose gel in 0.5xTBE buffer, 100 mA, for the presence or absence of the methylated and unmethylated fragments.

Statistical Analysis
According to the fact, that obtained data cannot demonstrate the normal Gaussian distribution, a nonparametric statistical approach was used. The Mann-Whitney test was performed for comparison of independent samples. For comparison of the tumor-normal (TN) pairs we used the Wilcoxon matched pair test. The Kruskal-Wallis test with the Dunn-Bonferroni post-hoc test for multiple comparisons was performed to find differences between multiple experimental groups. The non-parametric Spearman correlation analysis was used to find correlations. Finally, the MDR, simple logistic regression and the receiver operator characteristic (ROC) analysis were used for identification of the combinations of the methylated studied genes with the best accuracy, sensitivity and specificity for PCa. The software STATISTICA10, Microsoft Excel, Sigma Plot, Displayr (https://app.displayr.com/), OpenEpi and GraphPadPrism were used to perform all listed tests and data visualization.

Results and Discussion
Methylation of genes' promoters in cell lines. Among all studied genes, NKX3.1 was the most highly and constitutively methylated gene (Fig. 1A). In contrast, the CDH1 gene demonstrated a low methylation rate in all cell lines, with the highest methylation in LNCaP and the lowest -in DU145. Promoter methylation of another epithelial cell marker, KRT18, was the highest in DU145 and lowest in PC3 cells. All data are shown in Fig. 1A.
Gene promoter methylation patterns in prostate tissues. Promoters of the genes described above (except VIM), showed differential methylation patterns in adenocarcinomas and paired CNTs (Fig. 1B). The VIM promoter was totally methylated in all samples. NKX3.1 and KRT18 were highly methylated whereas the methylation of PTEN and CDH1 was low. RASSF1A demonstrated the most differential methylation, ranging from 0.03 to 1 in various samples (Fig. 1B).
There was no difference either in the methylation rate of PCa adenocarcinomas compared to CNTs, or in the groups, sorted by the early (1-2) and late (3-4) stages of the cancer disease according to results of the Wilcoxon matched pairs test.

A qualitative analysis of promoter methylation patterns in cell-free DNA samples from urine
Three groups of patients were included in this study: one group with confirmed PCa (n=41), one group with the inflammatory disease of a The NKX3.1 promoter was not methylated in any healthy donor but it was methylated more frequently in the samples of patients with inflammation than in PCa patients (75 % vs 63 %) ( Fig. 2A). Oppositely, CDH1 and GDF15 were frequently methylated in the samples of PCa patients, but rarely in patients with inflammation. The RASSF1A promoter was methylated in the majority of PCa and inflammatory disease patients and only in 6 % of healthy donors. All data are presented in Fig. 2A). A representative image of MSP products is shown in Fig. 2B.
All five genes showed differences in methylation between healthy donors and PCa patients ( Table 1). Three genes -NKX3.1, PTEN and RASSF1A differ between healthy donors and patients with inflammation. Only one gene, CDH1, demonstrated the differential methylation pattern in patients with inflammation and PCa.
Relation of the methylation pattern with a tissue type, gene expression levels and CPCs. We found that for the same patient a higher methylation rate of CDH1, PTEN and NKX3.1, observed in tumors, corresponds to a higher methylation rate of these genes in surrounding CNTs. The correlation coefficient for CDH1 was r s = 0.771, p = 0.000231, for PTEN -r s = 0.668, p = 0.00614, and for NKX3.1 -r s = 0.851 with p<0.0001.
Using the presented and observed earlier data on the relative gene expression [11], the putative correlations between methylation and expression of studied genes were analyzed. We found that the methylation rates of NKX3.1 and CDH1 negatively correlated with each other (s = -0.68633, p <0.01) in tumor tissues, Table 1. The statistically significant differences between promoter methylation of DNA, isolated from urine of patients with PCa and inflammation, and from healthy donors. Of note, the level of KRT18 methylation in PCa tissues negatively correlated with Gleason score (GS) (s = -0.593, p<0.05). There were no correlations between the methylation rate of the studied genes and other CPCs (S 3).
The differential methylation pattern in adenocarcinomas, bearing and negative for the TMPRSS2:ERG fusion transcript. Earlier, we detected TMPRSS2:ERG fusion transcript in PCa samples [9].
We analyzed the promoter methylation rates of all studied genes in PCa samples, grouped depending on the presence or absence of the TMPRSS2:ERG fusion (F+ and F-, respectively), using the Mann-Whitney rank sum test (Fig. 3A-E). The significantly higher levels of methylation of the RASSF1A promoter were shown for the F+, than for the F-groups (t=4.98, p<0.001) (Fig 3A). There were no other differences in methylation rates (Fig. 3B-E). Taking into consideration that our study included PCa patients and patients with inflammation of genitourinary system, the Odds ratio (OR) was calculated for each methylated gene, associated with PCa. The results of such analy sis are presented in S4.

A regression analysis of the qualitative methylation of genes' promoters in DNA isolated from urine and a putative combination of markers for non-invasive PCa diagnostics
The calculations demonstrate, that the study of all genes in combination can help to distinguish PCa patients from non-PCa individuals. The area under ROC curve (AUC) is 0.7905, p<0.0001 (Fig. 4A).
Using the MDR analysis, we selected PTEN, RASSF1A and NKX3.1 that are differen tially methylated in PCa patients and healthy individuals. The AUC for those genes is estimated as 0.9512, p<0.0001 (Fig. 4B). The best combination for differentiation of PCa patients and patients with inflammation was CDH1 with GDF15. The AUC for those genes was estimated as 0.8049 (Fig. 4C). We generated ROC curve, combining promoter methylation of PTEN, RASSF1A and NKX3.1 with CDH1 and GDF15, to evaluate their potential for diagnostics. The AUC was 0.8175, p<0.0001 (Fig. 4D).

Conclusions
Summarizing the obtained data on methylation in prostate tissues and urine, we found that the RASSF1A promoter demonstrates a higher methylation rate in the TMPRSS2:ERG fusion positive PCa. The frequencies of methylated NKX3.1, PTEN and RASSF1A in DNA, isolated from urine, were higher in cancer patients compared to healthy donors. The promoters of the CDH1 and GDF15 genes were methylated more frequently in PCa patients, than in patients with inflammatory disease. PTEN, RASSF1A,