Promoter methylation of Wnt-antagonists in polypoid and nonpolypoid colorectal adenomas

A panel of 9 CRC cell lines (Caco2, Colo205, Colo320, HCT116, HT29, SW480, SW620, LS174T and LS513) was used in this study. Colo205, Colo320, HCT116, HT29, SW480, SW620, LS174T and LS513 were cultured in DMEM (Lonza Biowhittaker, Verviers, Belgium) supplemented with 10% fetal calf serum (FCS) (Hyclone, Perbio, Etten-Leur, The Netherlands). Caco2 was cultured in RPMI1640 (Lonza Biowhittaker) supplemented with 20% FCS. Both cell culture media were supplemented with 2 mM L-Glutamine, 100 IU/ml sodium-penicillin (Astellas Pharma B.V., Leiderdorp, The Netherlands) and 100 μg/ml streptomycin (Fisiopharma, Palomonta (SA), Italy). The cervical cancer cell line CaSki was used as positive control and cultured as described before [26]. All cell lines were cultured using coated flasks and dishes (Greiner Bio-One, Frickenhausen, Germany).

Collection, storage and use of archival tissue and patient data were performed in compliance with the “Code for Proper Secondary Use of Human Tissue in the Netherlands” (http://​www.​fmwv.​nl and http://​www.​federa.​org). This study was approved by the VU University medical center (2011-03), the Leeds University (CA02/014 Leeds (West)) and the Hospital Vitkovice (EK/140/10).

This study followed the ethical guidelines of the Institutional Review Board (IRB). The IRB waived the need for consent for use of the archive samples, and the samples were analyzed anonymously.

Formaldehyde-fixed, paraffin-embedded (FFPE) colorectal tissue samples were collected at three different institutes; Leeds General Infirmary in Leeds, UK, Hospital Vitkovice in Ostrava, Czech Republic and VU University medical center in Amsterdam, The Netherlands [5, 13, 27]. Patients with a hereditary form of CRC, inflammatory bowel disease were excluded. The final series contained 44 nonpolypoid adenomas, 44 polypoid adenomas and 18 carcinomas. Normal colorectal mucosa was collected from age matched non-cancerous patients. Classification of the adenomas was performed using the Paris classification [28]. A summary of all clinical characteristics is listed in Table 1.

DNA and RNA from cell lines was isolated using TRIzol Reagent (Life Technologies, Breda, The Netherlands) according to the manufacturers’ instructions [29]. DNA from FFPE material was isolated after macro-dissection as described before [14].

DNA methylation analysis of SFRP2, WIF-1, DKK3 and SOX17 was performed using quantitative methylation specific PCR (qMSP) as described before [25, 30]. All samples were tested in duplicate and average Ct values were used for further analysis. Samples with delta Ct values between duplicates more than 1.5 were excluded.

In addition, the modified, unmethylated sequence of the housekeeping gene B-actin (ACTB) was amplified as a reference to verify sufficient DNA quality and successful DNA modification [31]. Samples with Ct-values > 32 for ACTB were excluded from further analysis. In all qMSP runs, a negative (non-bilsulfite-treated cell line DNA (CaSki)) and a positive (bisulfite-treated CaSki DNA) control were included. For all samples the delta Ct ratio between the gene of interest and ACTB was calculated using the 2^-ΔCt method [32]. The upper limit of the 99% confidence interval of normal controls was used as cut-off value to determine methylation positivity. The reproducibility of these assays has been demonstrated previously [25].

CaSki cells were incubated with 0.2 and 5 μM 5-aza-2′-deoxycytidine (DAC; Sigma Chemical Co, St Louis, MO, USA) diluted in PBS for five days. All incubations were performed in duplicate, and cells were directly harvested for DNA and RNA isolation. Gene expression was evaluated by RT-PCR as previously described [30].

Statistical analysis was performed using SPSS 20. We used a significance level of p < 0.013 (0.05/4 genes), to adjust for multiple testing according to the correction suggested by Bonferroni [33]. Comparisons between the methylation levels in CRC cell lines and normal colon were done using a non-parametric Mann-Whitney U test. After that, a positivity score was performed using a cut off level based on the upper limit of the 99% confidence interval of the normal controls. All group comparisons were performed using two-sided chi-square or Fisher exact test when expected values in the cross table are below 10. In the overview tables it is indicated which test was applied.

Logistic regression was used to examine the relationship between WIF-1 methylation and the independent variables phenotype, location, APC methylation, 5q loss and APC mutation. First the univariate relationships between gene methylation and the dependent variables were examined. Multivariate analyses were performed including all variables with a univariate p-value of less than 0.1. Next, we used a stepwise procedure and removed the variable with the largest p-value in each step, until only variables with a p-value < 0.05 remained in the multivariate model.

Comparison of the SFRP2, WIF-1, DKK3 and SOX17 promoter methylation levels between nine CRC cell lines and eighteen normal colon mucosa samples revealed significantly elevated methylation levels in CRC cells for all four genes (p = 0.00001, p = 3.1*10^-5, p = 3.1*10^-5 and p = 0.001 for SFRP2, WIF-1, DKK3 and SOX17, respectively). For SFRP2, WIF-1 and DKK3 increased methylation levels were observed in all nine CRC cell lines, whereas SOX17 showed higher methylation levels in all but two cell lines (LS513 and LS174T) (Figure 1).

Methylation lead to decreased expression, as upon treatment with demethylating agents, an increase in expression was observed for SFRP2, DKK3 and SOX17 but not for WIF-1 (Additional file 1: Figure S1).

Since the findings in cell lines are supportive of a role of promotor methylation of these genes in colorectal carcinogenesis, we next investigated a series of tissue specimens consisting of 18 carcinomas, 44 nonpolypoid and 44 polypoid adenomas.

Increased methylation levels for all four genes were detectable in all carcinomas and in both polypoid and nonpolypoid adenomas (Figure 2).

Interestingly, methylation levels in nonpolypoid adenomas were more similar to those observed in carcinomas than those in polypoid adenomas. No relation in methylation levels of any of the four genes and the different carcinoma stages was observed (data not shown).

To dichotomize the qMSP results into positive or negative for methylation, a cut off was calculated for each gene based on the 99% confidence interval of the normal controls (Figure 3). Significantly increased positivity rates were observed for all four genes in both types of adenomas and in carcinomas compared to normal colorectal mucosa (p < 0.003, Additional file 2: Table S1). Interestingly, DKK3 and WIF-1 methylation frequencies were significantly higher in polypoid adenomas (DKK3; 97% (38/39), WIF-1; 87% (34/39)) compared to nonpolypoid adenomas (DKK3; 76% (32/42), WIF-1; 57% (24/42); p = 0.005 and p = 0.003, respectively). WIF-1 methylation was also significantly higher in polypoid adenomas compared to carcinomas (WIF-1; 47% (8/17), p = 0.003).

When the methylation results of the four Wnt-antagonists were combined into one value that was positive if all four markers were methylated 79% (31/39) of the polypoid adenomas were positive in comparison to only 40% (16/40) of the nonpolypoid adenomas (p = 0.0004), indicating a lower level of Wnt-antagonist methylation in nonpolypoid adenomas in general (Figure 3, Additional file 2: Table S1).

To investigate the relation between methylation of SFRP2, WIF-1, DKK3 and SOX17 and the anatomical location of the adenoma, we divided all the adenomas into left- and right-sided adenomas. This showed no statistical difference for the investigated genes, except for WIF-1 methylation, which showed more methylation in the left colon 82% (40/49) compared to the right colon 56% (18/32), p = 0.01 (Table 2).

Methylation of the Wnt-antagonists may provide an alternative mechanism of Wnt-pathway activation next to APC mutations, methylation and loss of the APC locus on chromosome 5q. Therefore, we combined the promoter methylation results for SFRP2, WIF-1, DKK3 and SOX17, in polypoid and nonpolypoid adenomas, with previously obtained molecular data on APC mutation [13], APC methylation (Voorham et al., submitted) and chromosome 5q loss [14] in the same samples. For APC methylation as well as for chromosome 5q loss, no relation was found with the promoter methylation results for SFRP2, WIF-1, DKK3 and SOX17 when combining both adenomas types or in nonpolypoid adenomas or polypoid adenomas, separately (Tables 3 and 4). For APC mutation, a positive trend with WIF-1 as well as with DKK3 methylation was observed (Table 5). Of the APC mutated adenomas 83% (33/40) showed WIF-1 methylation and of the APC wild type adenomas 61% (25/41) showed WIF-1 methylation (p = 0.048). For DKK3, 95% (38/40) of the APC mutated samples showed DKK3 methylation whereas only 78% (32/41) showed DKK3 methylation in the APC wild type adenomas (p = 0.026) (Table 5). When we combined APC methylation, APC mutation and chromosome 5q loss together into one value called APC disrupting event, no significant difference was found (Additional file 2: Table S2).

To investigate the possible interaction between the different variables (phenotype, location, APC mutation, APC methylation and chromosome 5q loss), a multivatiate analysis was performed for WIF-1 methylation. For the genes SFRP2, DKK3 and SOX17, we were unable to perform a valid multivariate analysis, due to the limited number of unmethylated samples[34].

For the WIF-1 gene, we first performed univariate analyses showing that phenotype (p-value = 0.004), location (p-value = 0.016) and APC mutation (p = 0.035) were related to WIF-1 methylation. In the multivariate analysis, location and APC mutation were removed from the model (due to high p-values), leaving only phenotype in the model. This suggests that phenotype is the major contributor to the observed difference in WIF-1 methylation in our samples.