DNA methylation patterns in bladder cancer and washing cell sediments: a perspective for tumor recurrence detection

Urinary bladder cancer is the fourth most common malignancy in the Western world, with a male:female ratio of nearly four to one and a median age at diagnosis between 65 and 70 years [1]. Histologically, 90% to 95% of malignant bladder tumors are urothelial carcinoma (UC), formerly designated transitional cell carcinoma (TCC) [2]. Although more than 70% of the lesions are detected as non-invasive papillary carcinomas, which commonly recur, a poor prognosis is related to tumors that are already invasive at diagnosis (~20%) [3]. After transurethral resection of superficial bladder cancer, periodic cystoscopic monitoring is performed for early recurrence detection, with some cases requiring intravesical prophylactic instillation chemotherapy. Muscle invasive disease calls for more aggressive treatment, often consisting of radical cystectomy and bladder substitution [4].

At present, conventional diagnosis for urinary bladder cancer is based on morphological, histological and pathological features. These criteria provide essential prognostic information, but show insufficient power to precisely predict patient outcome. The need for accurate predictive markers has led to the search for molecular markers in bladder cancer patients [5]. The use of genetic and epigenetic alterations for the early detection of bladder cancer is promising because it is believed that some molecular events occur at the beginning of the carcinogenesis process. Thus, molecular diagnosis may allow detection before clinical or radiographic manifestations. In this context, a sensitive and specific noninvasive test could prescreen patients with clinical symptoms as well as those at high risk, and would also be useful in monitoring patients post-surgically for early detection of recurrence.

DNA-, RNA-based or/and immunohistochemical methods have been applied to identify new tumor markers or to estimate risk of tumor progression in UC. Several DNA alterations have been described in bladder cancer, such as allele losses or deletions [6], gene amplifications [7], DNA mutations [8] and microsatellite instabilities [9]. Furthermore, aberrant DNA methylation patterns have been recognized as common epigenetic changes in human cancer and are already detected in early cancer stages [10]. DNA methylation occurs on cytosine residues located at the 5' position of guanines in CpG dinucleotides [11]. Its distribution on the mammalian genome is not random and is especially important in CpG-rich areas, also called CpG islands. The promoter region of actively transcribing genes is frequently rich in this dinucleotide sequence, almost always unmethylated [12].

Dense DNA methylation in CpG islands of growth-regulating gene promoter regions is now recognized as a common alternative mechanism for gene inactivation in human cancer, an event as important as somatic mutations in coding regions of tumor suppressor genes (TSG) [13]. Usually both genetic and epigenetic events represent complementary hits involved in TSG inactivation [14]. A large number of studies have shown that loci of epigenetically inactivated TSG generally coincide with overlapping regions of allelic losses in human cancer, including UC [6, 15‐19]. In fact, loss of heterozygosity (LOH) assays have been widely used as indirect approaches in the search for a new TSG [20, 21]. In the last few years, genetic studies have indicated that allelic loss in many distinct chromosomal regions, including 1p, 3p, and 16q, are associated with UC tumorigenesis [17‐19, 21]. It is important to notice that RASSF1A (Ras association (RalGDS/AF-6) domain family 1) and RARB (retinoic acid receptor, beta) mapped at 3p (3p21.3 and 3p24, respectively), SFN (stratifin, also known as 14-3-3σ) located at 1p35.3, and CDH1 (cadherin 1, type 1, E-cadherin [epithelial]) at 16q22.1, are epigenetically silenced TSGs located at loci that overlap with LOH minimal regions in human cancer.

RASSF1A protein probably modulates some of the growth inhibitory responses mediated by Ras, although its interaction with activated Ras remains unclear. This gene is considered a bona fide tumor suppressor epigenetically inactivated during human carcinogenesis, whose hypermethylation has also been reported in UC [22‐30]. The RARB gene is a member of the thyroid-steroid hormone receptor superfamily of nuclear transcriptional regulators that binds retinoic acid (the biologically active form of vitamin A), and also mediates cellular signaling during embryonic morphogenesis, cell growth, and differentiation [31]. Retinoic acids exhibit tumor-suppressor activity due to their antiproliferative and apoptosis-inducing effects [32]. RARB has also presented high methylation frequencies in urinary bladder tumors (varying from 15% to 93%) [15, 22, 24, 25, 30].

Initially, it was suggested that loss of stratifin expression could contribute to malignant transformation by disabling the cell cycle arrest at the G2 checkpoint, allowing the accumulation of genetic defects [33]. Subsequently, the down-regulation of SFN gene in various human cancers was generally attributed to the hypermethylation of its CpG island. To the best of our knowledge, there is only one previous study addressing SFN hypermethylation in UC where the highest frequency was found for squamous cell carcinomas irrespective of their grade of cellular malignancy (80%). Furthermore, the authors found hypermethylation of 57.1% for high grade, high stage nonpapillary TCC; 28.6% for low grade, low stage papillary TCC; and 28% for undifferentiated small cell carcinomas, the lowest rate [34].

The CDH1 gene encodes for a calcium-dependent cell-cell adhesion glycoprotein, whose loss of function may contribute to cancer progression by increasing proliferation, invasion, or metastasis [35]. These findings suggest that CDH1 is a tumor- and invasion-suppressor gene [36]. Its relation to urinary bladder carcinogenesis was demonstrated through the observation of altered expression due to epigenetic changes in many studies, ranging from 9.5% hypermethylation [29] up to 87% in TCCs [37]. However, some studies have also evaluated TCCs, both in squamous cells and in situ carcinoma, and have shown a variable spectrum of hypermethylation for the same gene [15, 22, 24, 25, 27, 30, 38‐40].

In an effort to identify a possible association among epigenetic changes, urinary bladder cancer prognosis and early-recurrence, we analyzed the methylation pattern of CDH1, RARB, SFN and RASSF1A genes in 54 fresh samples of urinary bladder cancer, 14 of which were paired with tumor-adjacent normal urothelium; 39 paraffin-embedded UC primary tumor and/or recurrence matched with 23 bladder washing sediments obtained from 20 patients under post-surgical monitoring. In addition, we analyzed a hospital-based control group consisting of 24 bladder washings from patients that reported urological complaints, but without any bladder tumor history and showing negative cytology for tumor cell presence.

Epigenetic alterations are a hallmark of human cancer. In particular, DNA hypermethylation is a common mechanism for inactivating tumor-suppressor and other cancer genes in tumor cells [48]. The aberrant methylation patterns have been used as targets for the detection of tumor cells in clinical specimens such as tissue biopsies or body fluids [49].

In our MSP analysis performed on DNA obtained from fresh tumor samples, a hypermethylated pattern predominated for CDH1 and SFN genes. Commonly, a large spectrum of hypermethylation frequencies has been reported for several genes in bladder cancer. For example, CDH1 gene methylation frequencies range from 9.5% to 84% [15, 22, 25, 27, 30, 37‐40], independently of histological classification. We detected CDH1 hypermethylation frequency of 100% in bladder UCs, squamous cell carcinomas and adenocarcinomas samples, as well as in normal adjacent urinary bladder tissue and in exfoliated urothelial cells from cancer-free controls. Similarly, the SFN gene was also hypermethylated in these samples. Costa et al. [50] also detected high frequencies of CDH1 methylation in clear cell renal carcinomas and normal renal tissues (82.7% and 87.1%, respectively); in addition, SFN was hypermethylated in 100% of normal and tumoral renal tissues analyzed.

Interpretation of differential DNA-methylation patterns in cancer has proven difficult, in part because the functional consequences depend on the genomic region involved, the specific CpG dinucleotides, and the inter- and intratumoral heterogeneity. Apart from this, methodological issues such as the different primer sets interrogating methylation at distinct CpG dinucleotides of a specific promoter region could explain the range of frequencies reported in the literature. In our study, the protocol used (which included the addition of urea in the DNA modification step in order to improve the efficiency of unmethylated cytosine conversion [42]) and the CDH1 gene analysis based on the nested-PCR approach may have contributed to the high methylation prevalence observed.

The dynamic nature of epigenetic alterations is partially due to polymorphisms in some methyl group metabolism genes [51, 52] and in genes coding for proteins that mediate these changes (DNA methyltransferases, methyl-CpG-binding domain proteins) [53]. In addition, genomic profiles of DNA methylation are also influenced by aging [38, 54], dietary intake [55, 56] and environmental exposure [53, 57, 58]. In this context, DNA methylation heterogeneous patterns should be expected and already detected for some genomic regions as reported by Eckkhardt et al. [59]. These authors have found that 30.2% of investigated loci on chromosomes 6, 20, and 22 exhibit heterogeneity of methylation status, mainly due to the mosaic patterns in the studied tissue. Furthermore, 10% of the analyzed regions showed tissue-specific differences in DNA methylation.

We observed hypermethylated CDH1, SFN, and RARB genes in the normal-adjacent tissue of urinary bladder tumor. Aberrant methylation patterns have been associated with chronic inflammation [60, 61], viral infection [62] and aging [63]. Smith and Pereira-Smith [64] have previously reported that epigenetic alterations are involved in both the etiology and consequences of aging. Thus, hypermethylation in normal tissue as detected in the present study agrees with the results previously found by Bornman et al. [38], who observed a similar pattern for CDH1 in normal bladder tissue from patients older than 70 years. Furthermore, aberrant methylation patterns of the CDH1 promoter region were also described in pre-adenoma stages of colorectal cancer, in hyperplasic polyps [65, 66] and in ulcerative colitis (a chronic inflammatory condition of the large intestine that predisposes to cancer) [60, 67]. During breast cancer progression, CDH1 gene methylation occurs in about 30% of the ductal carcinomas in situ, with a significant increase in invasive and metastatic lesions [68]. Moreover, this gene has been also found methylated in pre-malignant and invasive bladder cancers. In mammary tissue, SFN is usually unmethylated in normal epithelium, but methylated in atypical hyperplasias, intraductal papillomas, ductal in situ carcinomas, infiltrating carcinomas and in stromal cells [69, 70]. SFN and CDH1 methylation have been reported in peripheral blood cells [70] as well as in infiltrating leukocytes in breast cancer [71]. Overall, these observations suggest that both genes, CDH1 and SFN, are not effective biomarkers for MSP analysis in bladder cancer.

The MSP analysis of RARB and RASSF1A genes showed respective hypermethylation of 82.9% and 24.4% in 49 UCs analyzed. Investigating the methylation at the same CpG dinucleotides, Maruyama et al. [22], and Catto et al. [25], found 15% and 24% hypermethylation for RARB, and 35% and 54% for RASSF1A, respectively. In order to verify the specificity of RARB and RASSF1A hypermethylation in relation to malignant phenotype in bladder tissue, we evaluated these genes in exfoliated urothelial cells from patients without cancer (control group): RARB and RASSF1A hypermethylation were detected in 8.3% and 33.3%, respectively. The comparison of these data revealed that RARB hypermethylation provides better diagnostic coverage and specificity than RASSF1A hypermethylation. However, the hypermethylated pattern of these genes in normal adjacent tissue in matched samples, especially for the RARB gene, was an unexpected finding. In this context, we could hypothesize that molecular alterations precede morphological changes in the exposed urinary bladder epithelium, since patients with bladder neoplasia frequently show genetic instability on apparently normal mucosa besides alterations of surrounding tissue [72]. Aberrant methylation patterns appear to reflect a pre-malignant characteristic of the urinary bladder mainly because UC is a neoplasia with multifocal lesions and elevated recurrence indices [73], thus corroborating the hypothesis that epigenetic alterations in cancer may preexist in morphologically normal cells [74].

Thus, genetic and epigenetic alterations may be present before cancer detection by imaging or traditional pathology investigations. Therefore, molecular tests that target these alterations have conceptual advantages for the successful early detection of neoplasias [48]. DNA represents an ideal substrate for molecular detection because it is robust, survives adverse conditions that many clinical specimens undergo and, can be readily amplified by powerful PCR-based approaches [75]. Tiny amounts of DNA from early pre-neoplastic lesions or small cancers can be used to permit the sensitive detection of one cancer cell in a background of hundreds of normal cells.

Hence, the predictive value of MSP in identifying tumor cells in washing sediments was evaluated in bladder cancer patients under post-surgical monitoring to detect tumor recurrence. Positive cytology was found in 33.3% of patients with urinary bladder tumor history. The hypermethylation patterns of RARB and RASSF1A genes observed in cells obtained from urinary bladder washing sediments were not concordant: some hypermethylated cases in tumor tissue and recurrence by cytological analysis did not show this marker in the same exfoliated cells. Contrarily, in four cases, the cells taken from urinary bladder washings exhibiting hypermethylation for these genes did not match the hypermethylation of the correspondent TCC. The heterogeneity of the intra- and intertumoral methylation patterns could partially explain these discrepancies. Thus, hypermethylation could already be present in the urinary bladder epithelium of cancer patients but not necessarily in cells exfoliated from the urinary bladder of cancer-free patients. RARB hypermethylation confirmed the presence of tumor cells in only 1 out of 5 recurrent cases and was absent in all cases showing negative cytology. Importantly, for the eight patients whose tumors did not present RARB methylation, the paired cell washing sediments DNA were also negative for methylation. This finding corroborates the idea that the methylation pattern of this gene is specific for tumor cells. RASSF1A gene MSP analysis in washout cells showed discordant results since its hypermethylation was not detected in 4 recurrent cases, although 2 negative cases for tumor cells using cytology showed this tumor tag, which suggest that these patients are under high risk for tumor recurrence. Some studies using promoter hypermethylation identified in tumor DNA as a target for cancer detection in the correspondent urine sample have shown sensitivities ranging from 49% to 91% [15, 26, 30, 75]. Recently, Yu et al. [76] included the RASSF1A in an 11-gene set to assessment of DNA methylation in urine sediments for sensitive/specific detection of bladder cancer. Although two studies have reported that the overall methylation sensitivity was significantly higher than cytology [15, 77], several factors may contribute to the lower sensitivities of MSP analysis in cells from urinary bladder fluids including the incomplete diagnostic coverage of selected gene sets, limited quantity of cells sampled, and the intrinsic heterogeneity of methylation patterns in the exposed epithelium, among others.

In the literature, no single gene was found to be consistently methylated in most bladder tumors. Thus, panels of genes that are methylated in urinary bladder cancer have been investigated to define methylation profiles associated with urinary bladder cancer diagnosis, prognosis and early recurrence detection. DNA hypermethylation of CDH1 and SFN genes was detected indistinctly among urinary bladder tumoral and normal tissues as well as in urinary bladder exfoliated cells, suggesting that these epigenetic features do not satisfy enough specificity criteria for use as prognostic or early detection markers. The methylation of RARB and RASSF1A genes appears to be an initial event in urinary bladder carcinogenesis maintained during tumor progression and should be included in the panels of differentially methylated genes in urinary bladder cancer in order to maximize the diagnostic coverage of epigenetic markers.