The chronic lymphocytic leukaemia cell lines MEC-1 20 and EHEB 21 were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany) and were grown in a medium consisting of 90% Iscove's Modified Dulbecco's Medium (Invitrogen, San Diego, USA) or Roswell Park Memorial Institute Medium (Invitrogen), respectively, supplemented with 10% foetal bovine serum (Invitrogen). Treatment of cells with 5-aza-2´-deoxycytidine (5-aza-dC; Merck Chemicals, Nottingham, UK) was performed for 72 h and 96 h, starting cultures from 5 × 10⁵ cells/ml. For the 96 h-treatment, the medium was changed after 48 h in order to supply fresh drug. Since a treatment with 0.5 μM 5-aza-dC resulted in only a small decrease of methylation of hypermethylated genes (on average by 9.5%), further experiments were performed at higher drug concentrations (1.0 and 2.0 μM), which did not affect the per cent of viable cells in cultures (>90%) as shown with Vi-Cell Automated Cell Viability Analyzer (Beckman Coulter, Brea, USA). These 5-aza-dC concentrations, but not lower doses, proved to be effective in DNMT1 depletion and transcriptional reactivation of different genes in CLL cell line WaC3CD5 in an earlier study 22 as well as in other cell lines of B cell lineage 23 24 .

Peripheral blood was obtained from 12 patients of Volgograd Regional Clinical Oncological Dispensary No.1 (Volgograd, Russia) and the National Haematology Research Centre of the Russian Academy of Medical Sciences (Moscow, Russia). Written informed consent was given and the experiments were approved by the institutional Ethical Review Boards. Diagnosis of CLL was established according to standard morphologic and immunophenotypic criteria 25 . All patients were untreated at the time of blood collection. Peripheral blood mononuclear cells were isolated by a standard procedure using Ficoll-Hypaque gradient centrifugation as described elsewhere 26 . A total of five buffy-coats from the peripheral blood of healthy individuals were provided by the Institute for Clinical Transfusion Medicine and Cell Therapy (Heidelberg, Germany). CD19⁺ cells were isolated from control samples using Dynabeads CD19 pan B (Invitrogen, Carlsbad, USA) according to the manufacturer’s protocol and employed as a reference. The cells were collected and snap-frozen for the subsequent extraction of nucleic acids. The patient data are summarised in Additional file 1: Table S 1.

DNA was extracted from the samples using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) as suggested by the manufacturer. DNA concentration was measured in a ND-1000 spectrophotometer (Thermo Scientific, Wilmington, USA). A total of 1.9 μg DNA was treated with sodium bisulphite using the EpiTect Bisulfite kit (Qiagen). The efficiency of bisulphite conversion averaged 98.8% and was computed from the sequences of 231 cloned PCR-products of CDH1, DACT1, DKK1, DKK2, DKK3, DKK4, SFRP1, SFRP2, SFRP3, SFRP4, SFRP5 and WIF1 using the BISMA software, which considers the non-CpG cytosines within the sequences 27 .

PCR-amplification of the loci interrogated was carried out in 25 μl reactions that contained 2.0 μl bisulphite-converted DNA, 1.5 mM MgCl₂, 125 mM dNTP, 200 nM primers, 0.65 units HotStar Taq DNA polymerase and 1x Q-solution (Qiagen). A previously reported amplification protocol was employed 28 . Briefly, amplification was started by an initial activation of the HotStar Taq DNA polymerase at 95°C for 15 min. The first amplification cycle was denaturation at 95°C for 1 min, annealing at 62°C for 2 min and elongation at 72°C for 3 min. This procedure was continued for 20 cycles, reducing the annealing temperature by 0.5°C each cycle, followed by 25 cycles of 1 min denaturation at 95°C, 2 min annealing at 52°C and 2 min elongation at 72°C. The sequences of the PCR primers are listed in Table 1. About 5 μl of each reaction was examined on 2% agarose gels.

* Primers of the same sequence but without biotin modification were employed for amplification of the PCR-products used for subcloning and bisulphite sequencing.

In order to control for possible amplification bias, appropriate calibration was performed as described in detail 28 . Fully methylated and unmethylated human control DNA that had been bisulphite-treated was purchased (EpiTect PCR control DNA; Qiagen) and mixed in different ratios to obtain calibration samples that represent distinct methylation percentages of 0, 12.5, 25, 37.5, 50, 62.5, 75, 87.5 and 100%, respectively. A total of 15 ng calibration DNA was used for the amplification of each locus.

A volume of 20 μl of each PCR product was mixed with 2 μl Streptavidin Sepharose High Performance (GE Healthcare, Uppsala, Sweden), 38 μl PyroMark binding buffer (Qiagen) and 20 μl water. The Vacuum Prep Workstation (Biotage, Uppsala, Sweden) was used to prepare single-stranded DNA according to the manufacturer’s instructions. The Sepharose beads with the single-stranded templates attached were released into a PSQ 96 Plate Low (Biotage) containing 15 μl of 0.6 μM corresponding sequencing primer in annealing buffer. Pyrosequencing reactions were carried using the Pyro Gold Reagent Kit (Biotage) in a PSQ HS 96 Pyrosequencing System (Biotage) according to the manufacturer’s protocol. The sequences of the pyrosequencing primers are listed in Table 1. Quantification of CpG methylation was performed using the Software PyroQ-CpG v.1.0.9 (Biotage). The initial amplification result containing a bias towards unmethylated alleles was corrected using the calibration data derived from the control samples as previously described 28 . The PCR-product of SFRP1 was sequenced directly using Sanger chemistry. Despite optimisation efforts, accurate quantification was not possible by alternative pyrosequencing assays, which resulted either in enormous bias towards the methylated allele or readouts with lack of specificity.

The PCR products were purified with the QIAquick PCR Purification kit (Qiagen) and cloned using the TOPO TA Cloning kit (Invitrogen). Clones were picked at random and sequenced with Sanger chemistry at GATC Biotech (Constance, Germany). The sequencing data were visualised using the CpGviewer software 35 .

For each locus, the average methylation percentage across the interrogated CpG sites was calculated. Differences observed between the patient and control groups were tested using generalised least squares (GLS) models 36 . As the country of origin (Russia or Germany), sex, and age of the probands might have been influential covariates, they were taken into account. Thus, we fitted the following model for each gene:

Methylation = β 0 + β 1 Country + β 2 Status + β 3 Sex + β 4 Age + ε ,

where Methylation are the methylation measurements that were square root-transformed to achieve normality; β ₀ is the model intercept; β ₁ is the effect of the country of origin, which is used as a binary variable that takes value 0 for Germany and 1 for Russia; β ₂ is the effect of the proband’s status (a binary variable 0 for healthy and 1 for a diseased individual); β ₃ is the effect of sex (0 for a female and 1 for a male individual); β ₄ is the effect of age, and ε represents the model residuals. Ideally, model residuals should be normally distributed with mean 0 and a certain constant variance. However, an exploratory analysis revealed high variation of methylation between the countries as well as between the healthy and diseased individuals. The homogeneity of variance assumption was thus relaxed by allowing the variance to be different in each stratum, i.e. in each of the combinations of country and disease status 36 .

The full model was then reduced by stepwise backward elimination of insignificant terms. For each gene, only the final optimal model found by this approach is reported. Validation of the optimal models was performed by an examination of the quantile-quantile plots of their residuals 36 . Estimation of the model parameters was conducted based on the restricted maximum log-likelihood algorithm using the nlme v3.1-102 package 37 for the R computing environment 38 . P-values of less than 0.05 were regarded as statistically significant.

Total RNA was isolated using the miRNeasy Mini Kit (Qiagen) as suggested by the manufacturer. The RNA concentration was measured in a ND-1000 spectrophotometer (Thermo Scientific). The quality of the RNA samples was confirmed in a 2100 Bioanalyzer (Agilent, Santa Clara, CA). The RNA integrity number 39 of the samples averaged 9.7. Reverse transcription reactions were carried out with 1 μg of total RNA using SuperScript III Reverse Transcriptase (Invitrogen) and 0.5 μg oligo(dT) 12–18 primer (Invitrogen). Quantitative RT-PCR was performed in triplicate with the ABI PRISM 7900 Sequence Detection System (ABI, Foster City, USA) using the Absolute QPCR SYBR Green mix (Thermo Scientific). The identity codes of the commercially available PCR-primers are as follows: Hs_CDH1_1_SG, Hs_DACT1_2_SG, Hs_DKK1_1_SG, Hs_DKK2_1_SG, Hs_DKK4_1_SG, Hs_SFRP1_1_SG, Hs_SFRP2_1_SG, Hs_FRZB_2_SG, Hs_SFRP4_3_SG, Hs_WIF-_1_SG, Hs_GAPDH_1_SG (Qiagen), DKK3-qRTF1 and DKK3-qRTR1 (Thermo Scientific). Primer sequences for SFRP5 have been reported earlier: 5´-CTGACGGCCTCATGGAGCAGATGT-3´ (forward) and 5´-TCCAATCAGCTTCCGG TCCCCATT-3´ (reverse) 16 . Universal Human Reference (UHR) total RNA was used as a calibration control (Stratagene, La Jolla, USA). The calibration graph (c_t vs. log unit of the standard template) was obtained as described 40 . Two housekeeping genes – glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin (ACTB) – were tested. GAPDH was superior over ACTB exhibiting only minor variation of expression in all studied conditions (CV 10.4% vs. 34.9%, respectively) and therefore served as an internal control. The thermal cycler conditions were as follows: 50°C for 2 min and 95°C for 15 min, followed by a two-step PCR of 45 cycles at 95°C for 15 sec and 60°C for 60 sec.

For protein extraction, cells were lysed shaking at 4°C for 30 min in a buffer containing 50 mM Tris/HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1x Complete Protease Inhibitor (Roche Diagnostics, Mannheim, Germany), 1x phosphatase inhibitor I and III (Sigma-Aldrich, St. Louis, USA) and 2.5 U/ml benzonase (Qiagen). Cells were passed through a 25 gauge needle, and cell debris was removed by centrifugation at 15,000 g for 30 min at 4°C.

Protein quantification was performed by an immunoassay as previously described 41 . An antibody that is specific for the C-terminus of β-catenin (BD Biosciences, Franklin Lakes, USA) was covalently immobilised on magnetic xMAP microspheres (Luminex, Austin, Texas) 42 . Per well of a microtiter plate (non-binding surface; Corning, New York, USA), 2,000 beads in 20 μl blocking reagent for ELISA (Roche Diagnostics) were incubated with 20 μg protein from cell lysates in 40 μl with lysis buffer. In calibration experiments, defined amounts of recombinantly expressed GST-β-catenin were used. Incubation was overnight at 4°C. The samples were transferred to a blocked filter plate (Millipore, Billerica, USA) and washed twice with 100 μl PBS using a vacuum manifold (Millipore). For detection, 30 μl of anti total β-catenin antibody (Invitrogen; 1:200 diluted in the assay buffer) were added to the beads. Incubation was at room temperature for 120 min. Unbound antibody was removed by washing with PBS as described above. A volume of 30 μl of 2.5 μg/ml donkey anti-rabbit phycoerythrin-conjugated antibody (Jackson ImmunoResearch, West Grove, USA) was added next and incubated for 45 min at room temperature. After another washing with PBS, the beads were resuspended in 100 μl assay buffer and analysed with a Luminex 100 IS system (Luminex Corp, Austin, USA). Calculation of the absolute β-catenin expression was based on a seven-point dilution series of the recombinant standard protein.

For the quantification of E-cadherin, an E-cadherin-specific antibody (R&D Systems, Minneapolis, USA) was employed as the bead-bound capture reagent. All dilution, incubation, washing and measurement steps were performed as described above. However, the detection system consisted of biotinylated detection antibodies specific for E-cadherin (R&D Systems) and phycoerythrin-conjugated Leandro, USA). Recombinant human E-cadherin Fc Chimera (R&D Systems) was used as standard.

To find out if the expression of the Wnt/β-catenin pathway inhibitors may be regulated by DNA methylation, the two CLL cell lines EHEB and MEC-1 were studied. First, bisulphite pyrosequencing was performed to examine the methylation status of twelve inhibitor genes. After correction of the raw data for PCR-bias and artificial variations introduced by the pyrosequencing process as described 28 , accurate quantification of the methylation degree was performed for CpG dinucleotides that are located in proximity to the transcriptional start sites or those associated with expression down-regulation of the respective genes in solid tumours 30 34 (Figure 1A). A high degree of methylation was recorded (average methylation percentage in EHEB/average methylation percentage in MEC-1) for CDH1 (79/79), DKK1 (68/75), DKK2 (88/83), SFRP3 (94/71) and WIF1 (71 in EHEB). Partial methylation was observed for DKK3 (23/37), DKK4 (31 in EHEB), SFRP2 (32/17) and SFRP4 (24/29). The genes DACT1 (10/0), SFRP5 (14/6), DKK4 (9 in MEC-1) and WIF1 (11 in MEC-1) were essentially unmethylated. Pyrosequencing of SFRP1 failed, although various primers were tested. However, extensive hypermethylation of SFRP1 in both cell lines was confirmed by semiquantitative bisulphite Sanger sequencing. Overall, ten out of twelve antagonists of Wnt/β-catenin signalling exhibited substantial methylation of the gene in at least one of the two CLL cell line models.

Because DNA methylation in established cancer cell lines may not always adequately reflect the reality in primary tumours, also patient and control samples were studied for confirmation (Figure 1B). Consistent with the observations in the cell lines, eleven genes were found to be at least partially methylated in the patient material (average methylation percentage across twelve patient samples; lowest and highest value): CDH1 (56; 22 to 88), DKK1 (34; 17 to 84), DKK2 (35; 14 to 78), DKK3 (18; 5 to 61), DKK4 (32; 0 to 71), SFRP2 (21; 2 to 70), SFRP3 (26; 6 to 50), SFRP4 (17; 3 to 68), SFRP5 (15; 6 to 27), WIF-1 (30; 8 to 73). Again, extensive hypermethylation of SFRP1 was shown by semiquantitative bisulphite Sanger sequencing. In contrast, and in accordance with the cell line data, DACT1 was basically unmethylated (2; 1 to 5) in all samples. All the loci interrogated were essentially unmethylated in control CD19⁺ B cells of five healthy individuals with the exception of DKK4, which was found to be methylated (36; 33 to 39) to a level very similar to that in the cancer samples.

The methylation patterns of most genes were heterogeneous within the patient group. However, the levels of methylation for CDH1 DKK1 DKK2 DKK3 SFRP3 and WIF1 were significantly higher than in the control group (Figure 2). Although there was no statistically significant difference between the patient and control groups for the other genes, abnormally high methylation – defined as an increase of the average methylation level beyond that of the average observed in the group of normal tissues plus twice the standard deviation 43 – was a frequent event in patient CLL samples but for DACT1 and DKK4.

To exclude a possible contribution of additional covariates such as country of origin, sex or age of the individuals to the observed alterations of DNA methylation, we took them into account by using generalised least squares models. A disease-specific nature of methylation differences between compared groups was confirmed for all the genes, which exhibited differential methylation between the groups [see Additional file 2: Table S 2]. The results obtained from bisulphite pyrosequencing were validated by genomic bisulphite Sanger sequencing of randomly cloned PCR-products (Figure 3). The data obtained were in full agreement with the results of the pyrosequencing assay. In combination, the sequencing results indicate that the genes of most Wnt/β-catenin pathway inhibitors are prone to aberrantly high methylation in CLL.

In order to ascertain the role of methylation in the regulation of expression, mRNA levels of the twelve genes were analysed by quantitative PCR in the cell lines EHEB and MEC-1 after treatment with 5-aza-2´deoxycytidine (5-aza-dC). Incorporation of this nucleoside analogue into DNA is known to capture covalently DNA methyltransferases, thereby inducing DNA hypomethylation 44 . At a concentration of 1.0 and 2.0 μM, the drug induced on average a 25% decrease in the methylation level of the strongly methylated genes (Figure 4A). This decrease in DNA methylation was associated with a significant transcriptional re-activation of CDH1 DKK3 DKK4 SFRP2 SFRP3 SFRP4 and WIF1 in at least one of the two cell lines (Figure 4B ).

It is interesting to note that the expression of SFRP3 was induced 13.5- and 9.1-fold in EHEB and MEC-1, although no CpG island has been annotated from the genomic sequence 16 . This fact may highlight the rather artificial character of CpG island definition. The effect of the methylation decrease on expression was striking, which may actually be particularly due to the small number of CpG sites in the region (although some indirect activation mechanism cannot be excluded either). DKK1 DKK2 and SFRP1 were also highly methylated in both cell lines. Still, treatment with 5-aza-dC did not result in a significant induction of expression, although methylation was reduced to a degree that was similar to that of SFRP3. The genes DKK1 DKK2 and SFRP1 do have CpG islands in their sequences. Apparently, the demethylation effect was too little or insufficient on its own to change the overall blockage of expression.

Expression of DACT1 exhibited yet another, very contrasting result. It was substantially up-regulated after incubation with 5-aza-dC (10.5-fold and 3.5-fold in EHEB and MEC-1, respectively), even though the gene was essentially unmethylated even prior to the treatment. This clearly indicates an indirect control of this gene’s activity by the variation of the degree of DNA methylation; its own sequence is not affected, however.

In order to demonstrate the functional relevance of epigenetic control of antagonists to the overall regulation of the Wnt/β-catenin pathway, we looked at the variation in the amounts of transcriptionally active (non-phosphorylated) β-catenin and E-cadherin in EHEB and MEC-1 cell lines. The amount of non-phosphorylated β-catenin correlates directly with the activity of the pathway. For this analysis, we took advantage of suspension bead arrays, which had already been used for the detection of different cytoplasmic fractions of β-catenin 41 . No constitutive Wnt/β-catenin signalling could be detected in the cell lines. The non-phosphorylated fraction of β-catenin was hardly detectable (data not shown) at either condition, suggesting dormancy of this pathway route or a very efficient degradation process. However, the β-catenin level strongly increased in cell lysates after a treatment of the cells with 2.0 μM 5-aza-dC for 72 h and 96 h (Figure 5A), indicating a methylation controlled silencing in cell lines. Concomitantly to the increase in β-catenin levels, also the amount of E-cadherin increased substantially upon CDH1 hypomethylation, while the protein was basically undetectable under control conditions (Figure 5A). As expected, β-catenin and E-cadherin formed a complex, which as a result of the higher expression levels of CDH1 was also present at much higher concentrations after treatment with 5-aza-dC (Figure 5C).