DNA Hypermethylation of a panel of genes as an urinary biomarker for bladder cancer diagnosis

Purpose The aim of this study was to explore the diagnostic potential of a panel of ve hypermethylated gene promoters in bladder cancer. Individuals with primary BCa and control individuals matching the gender, age and smoking status of the cancer patients were recruited. DNA methylation was assessed for the gene promoters of RASSF1, RARβ, DAPK, hTERT and APC in urine samples collected by spontaneous urination. Fifty patients and 35 healthy controls were recruited, with average age of 70.26 years and average smoking status of 44.78 pack-years. In the BCa group, DNA methylation was detected in 27(61.4%) samples. RASSF1 was methylated in 52.2% of samples. Only 3(13.6%) samples from the control group were methylated, all in the RASSF1 gene promoter. The specicity and sensitivity of this panel of genes to diagnose BCa was 86% and 61% respectively. The RASSF1 gene could diagnose BCa with specicity 86.4% and sensitivity 52.3%. Promoter DNA methylation of this panel of ve genes could be further investigated as urine biomarker for the diagnosis of BCa. The RASSF1 could be a single candidate biomarker for predicting BCa patients versus controls. Studies are required in order to develop a geographically adjusted diagnostic biomarker for BCa.

The application of epigenetics may allow for a risk-adapted approach and more cost-effective method of diagnosis of BCa. Numerous epigenetic changes such as DNA methylation, histone modi cations, microRNA expression and nucleosome positioning are characteristic of the epigenome of BCa cells [5,6].
Similarly, hypermethylation of DAPC promoter is almost six times more frequent in BCa patients than in Page 3/12 healthy individuals (OR:5.81; 95%CI: 3.83-8.82, P < 0.00001) [11]. Finally, the TERT leads to upregulated activity of the enzyme resulting in cancer cells' immortalization [9]. To date, no single diagnostic biomarker could replace cystoscopy as the primary diagnostic tool for BCa [12]. As, additionally, geographical and ethnic differences in methylation patterns exist [13] a more potent panel of diagnostic biomarkers may be in demand.
The aim of this study was to explore the diagnostic potential of a panel of ve hypermethylated gene promoters, whose sensitivity and speci city has been proven when studied in a separate fashion. Ideally, this would lead to the development of a DNA methylation-based diagnostic protocol in urine samples and optimization of its sensitivity and speci city.

Study design -participants
This was a prospective, case-control study conducted in the Urology Department of a public teaching Hospital. The Hospital's Review Board and the Bioethics Committee of the School of Medicine of the Aristotle University of Thessaloniki approved the study protocol (decision No. 3/4.6.2014). The study is registered in the Australian New Zealand Clinical Trial registry (Registration number ACTRN12620000258954). All methods adhered to the tenets of the declaration of Helsinki as revised in 1989. All study participants were recruited following written informed consent.
Urine samples were collected from patients who were attending the cystoscopy clinics. Recruits were subjects with non-muscle invasive or muscle invasive BCa who were able to provide a urine sample and undergo a cystoscopy before any treatment for BCa. Control subjects matching the gender, age and smoking status of the BCa patients were recruited among healthy volunteers with no known urological disease or malignancy (Table 1). Individuals with metastasis in urinary bladder or other malignancies were excluded.   DNA extraction and treatment DNA was extracted from urine sediments using the Cells and Tissue DNA Isolation Kit (NORGEN BIOTEK CORP., Thorold, Canada). Both integrity and purity were con rmed via spectrophotometry and agarose gel electrophoresis. Extracted DNA was stored at -40°C until the modi cation with sodium bisul te using the EZ DNA Methylation-GoldKit (Zymo Research, Orange, CA). Modi ed DNA was then stored at -20°C until further analysis.

Gene promoter methylation assay
Quanti cation of the percentage of methylation of DNA in the gene promoter of DAPK [14], APC [15], RAR-β2 [15], RASSF1 [14] and TERT [14] was  Table 3. All primers were synthesized by IDT (Integrated DNA Technologies, Iowa, United States). In order to quantify and compare the ampli cation products, Cq data corresponding to the target genes were normalized relative to those of the internal housekeeping gene, B-ACTIN (14). Furthermore, a standard 100% methylated control human DNA and a 100% non-methylated control human DNA were used (EpiTect PCR Control DNA set, Qiagen, Germany. Methylation speci c quantitative PCR (MSP-qPCR) was run in duplicate. Table 3. Table 3 The sequences of primers and probes for the quantitative methylation speci c real-time PCR.

Gene
Primer

Statistical Analysis
Sample size calculation. Assuming that the percentage of DNA methylation among controls is 20% [10], and the desired OR will be 4, then the total sample size needed would be 78 subjects (39 patients and 39 controls) in order to achieve 80% power with alpha set at 5% (for each of the 5 biomarkers). Since this was a pilot study, no correction for multiple testing was made. The gene promoters were hypermethylated in 57% of individuals with non-muscle invasive BCa; in particular, 24% of them had one, 16% two, 14% three, 3% four and none ve hypermethylated gene promoters. By comparison, 80% of subjects with muscle invasive tumor had hypermethylated gene promoters: 20% one, 40% two and 20% three genes respectively (Fig. 1). According to the grade classi cation of tumors, the gene promoters were hypermethylated only in 50% of the patients with lowgrade urothelial cancer (one in 29%, two in 17% and three in 4%) while up to 72% of patients with high grade or CIS had hypermethylated gene promoters (one in 17%, two in 22%, three in 28% and four in 5%) (Fig. 2). The hypermethylation was not signi cantly different between patients with muscle and nonmuscle invasive BCa as well as between patients with high-grade and low-grade (p = 0.369 and p = 0.148, respectively). The gene promoters of DAPK and TERT were hypermethylated in one patient each.
According to our statistical analysis the speci city and sensitivity of this diagnostic panel of biomarkers were 86.4% and 61.4%, respectively, while the positive and negative predictive values were estimated at 90% and 53%, respectively. The area under the curve (AUC = 0.7634) derived from a multivariate logistic regression model (Fig. 3). The diagnostic panel was considered positive when it had at least one methylated promoter.
Due to the high correlation observed, only RASSF1, APC and RAR-β2 promoters remained in the model; however, APC and RAR-β2 had very unstable estimates (Quasi-complete separation of data points was detected). Thus, further analysis using exact logistic regression was performed in order to explore the multicollinearity effects. As a result, RASSF1 gene promoter could be a single candidate for predicting patients versus controls with speci city 86.4% and sensitivity 52.3%. The odds ratio (OR) estimate for the diagnostic biomarker RASSF1 only is 6.9 (95%CI: 1.792-26.857) (P = 0.005) and the AUC was 0.6932.
Further analysis was performed for the diagnostic accuracy of the current panel for MIBC or high-grade tumors. Based on the 5-year recorded prevalence of BCa in the Greek population [16] and the known prevalence of 20-30% for MIBC or high-grade tumors among rstly diagnosed BCa patients, we calculated that the current panel had a positive predictive value of 0.21 and a negative predictive value of 0.86, with 0.86 speci city and 0.72 sensitivity.

Discussion
In this pilot, controlled study the gene promoters of individuals with BCa were more commonly hypermethylated compared to healthy controls. The panel of genes tested was found to have 86.4% speci city and 61.4% sensitivity in the diagnosis of BCa, quite similar to the speci city and sensitivity of the RASSF1 promoter gene alone (86.4% and 52.3% respectively with OR 6.9).
Despite the relatively small sample size, the study sample re ects the disease's demographics. In general, non-muscle invasive cancer (NMIBC) can be found in 70-80% of all BCa and only 10-30% constitute muscle invasive BCa (MIBC) [17]. In our sample 84% of the patients had NMIBC and 11% MIBC.
Regarding the diagnostic potential of the hypermethylated gene promoters' panel of our study, the sensitivity (61%) is lower compared to the cystoscopy's sensitivity for all kinds of BCa (68-83%) but higher compared to cytology, particularly for low-grade tumors (50%) [18]. Furthermore, the speci city is considerably higher than the cytology's speci city for patients with low-grade cancer and comparable to the speci city of the invasive cystoscopy [4]. However, the diagnostic accuracy of the methylation of the panel of Twist Family BHLH Transcription Factor 1 (TWIST1) and NID2 genes as urine biomarker was higher than the diagnostic accuracy of our panel of genes, with 90% sensitivity and 93% speci city [19]. Similarly, a meta-analysis and systematic review of 24 articles revealed that the overall sensitivity and speci city of DNA methylation urine biomarkers was 84% and 92% respectively, higher than our results [20]. But since DNA methylation varies among different human groups regarding macro-and microgeographical scales, numerous studies from different areas are required in order to investigate the methylation pro le of the patients with BCa across human populations [13,23]. This may lead to a common panel of gene promoters that could be used worldwide to differentiate the BCa from healthy subjects but, in addition, a more individual approach may be necessary depending on subjects' residency.
Our results also indicate that the hypermethylation of suppressor gene promoter of RASSF1 might be a potential single urine biomarker in BCa with speci city 86.4%, sensitivity 52.3% and OR 6.9. By contrast, a previous study which investigated the diagnostic accuracy of RASSF1 in BCa, showed lower speci city and sensitivity of RASSF1 than in our study (17% and 58% respectively), whereas a recent meta-analysis revealed that the risk for  [25]. Functional analysis of RASSF1 shows a potential involvement in inhibition of cell proliferation, promoting cell apoptosis and aging and the maintenance of microtubule stability. It is also known that the expression of RASSF1 is absent in many tumor cells as a consequence of the methylation of the gene promoter [25]. However, Chen et al. found that the use of a panel of genes had higher diagnostic accuracy compared to the use of a single gene promoter [19].
In our study population, the gene promoter of DAPK was hypermethylated only in one patient contrary to a meta-analysis concluding that DAPK promoter methylation was associated with BCa risk (OR:5.81; 95% CI = 3.83-8.82, P < 0.00001) [11]. Similarly, the gene promoter of TERT was methylated in only one subject with BCa; a recent study by a research group suggests that THOR (TERT Hypermethylated Oncological Region) hypermethylation is associated with disease progression and increased TERT expression, which leads to carcinogenesis [9]. To date, there are no published studies to con rm whether our results re ect a low prevalence of TERT and DAPK promoter in the Greek population.
An attempt was made to explore the diagnostic accuracy of our panel of genes for MIBC or high-grade tumors. We found a high negative predictive value of 0.86, with 0.86 speci city and 0.72 sensitivity Previous literature has sparsely investigated the value of methylation biomarkers in the diagnosis of MIBC or high-grade tumors. In a recent study, molecular analysis of the methylation pro le of the promoters of p14ARF, p16INK4A, RASSF1A, DAPK and APC from urine sediments demonstrated correlations with BCa grade and stage [7], while other researchers found that the progression to MIBC in patients with primary pTaG1/2 BCa could be predicted with the methylation analysis of the gene promoters TBX2 and TBX3 [26].
Finally, 13.6% of our study controls had methylated the RASSF1 gene promoter, in accordance with a previous survey, which showed that 12% of the loci in apparently normal urothelium from cancerous bladder tissues were hypermethylated, indicating an epigenetic eld defect [27].
In the control group, we detected DNA in 63% of the urine samples, which can be explained by the decreased cell exfoliation of normal urothelium [28]. Furthermore a recent study shows that the procedure for collection of urine sediments can be in uenced by the co-sedimentation of normal cells and the presence of crystals and substances that may inhibit downstream PCR analyses [29].
Finally, our study was adequately powered and achieved an OR (6.9) higher than the initially desired OR = 4. The attained sample size of the control group was almost half of the initially planned. This might have had an effect on the genes with nonsigni cant results especially for APC where methylation was 34% as opposed to 13.6% of the control samples.
Cost-effectiveness is always an issue with novel technologies. The detection of hypermethylation of speci c genes from urine samples has been previously shown to be cost-effective in the diagnosis of BCa [20] [30]. When using our panel of genes, the real costs were signi cantly lower than cystoscopy costs in the Greek National Healthcare System (89 Euro vs. 230 Euro). However, the cost-effectiveness needs to be examined in light of the diagnostic accuracy of our panel of genes which remains to be proven in larger longitudinal case-control studies.