According to a multi-center study design, 20 matched tumor/macroscopically normal samples of breast cancer patients (median patient age: 67 years; range 48–86 years) and 179 unmatched breast carcinomas (median patient age: 57 years; range 28–96 years) were obtained from patients treated by primary surgery for breast cancer at the Departments of Gynecology at the University Hospitals of Aachen, Jena, Regensburg and Düsseldorf, Germany. None of the patients had received neo-adjuvant chemotherapy. Inclusion criterion for ipsilateral normal breast tissue was a distance of > 2 cm to the carcinoma margin. All patients gave informed consent for retention and analysis of their tissue for research purposes and the Institutional Review Boards of the participating centers approved the study. The selection of cases was based on availability of tissue. Cases were not stratified for any known preoperative or pathological prognostic factor. Tumor histology was determined according to the criteria of the WHO (2003), while disease stage was assessed according to UICC 38. Tumors were graded according to Bloom and Richardson, as modified by Elston and Ellis 39. Hormone receptor positivity was defined as an immunoreactivity score (IRS) ≥ 3 40. For 136 patients follow-up data were available with a median time of 64 months (range 1–174 months). Tumor material was immediately snap-frozen in liquid nitrogen after surgery. Hematoxylin/Eosin-stained sections were prepared for assessing the percentage of tumor cells; only samples with greater than 70% tumor cells were selected for analysis. Samples were dissolved in lysis buffer followed by DNA isolation using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's recommendations. For patient characteristics see additional file 1.

A total of 17 archival FFPE normal breast tissues were obtained from the Institute of Pathology, University Hospital of the RWTH Aachen, Germany. These patients had undergone breast reduction surgery without the condition of cancer. The median age in the cancer-unrelated normal breast tissue set was 33 years (range 22–61 years). Per sample, five consecutive sections (each 10 μm) were deparaffinized and rehydrated in a decreasing alcohol series prior to DNA extraction by use of the QIAamp DNA Mini Kit.

SFRP2 protein expression was assessed using a tissue microarray (TMA) consisting of 125 breast carcinomas, four ductal carcinomas in situ (DCIS) and 10 normal breast tissues that have been described previously 41. The TMA contained one tissue core from non-selected, FFPE primary breast carcinoma specimens diagnosed between 1994 and 2002 at the Institute of Pathology, University of Regensburg, Germany. Histological, all tumors were graded according to Bloom and Richardson, as modified by Elston and Ellis 39. Clinical follow-up data were available for 124 breast cancer patients with a median follow-up period of 80 months (range 5–114 months). All patients gave informed consent for retention and analysis of their tissue for research purposes and the Institutional Review Board of the participating centers approved the study.

The TMA was subjected to immunostaining using the K5007 Kit (DAKO, Hamburg, Germany) following the manufacturer's instructions. Antigen retrieval was performed by pretreatment in citrate buffer (pH 7) in a microwave oven (20 min, 200 W). Samples were incubated for 30 min with the primary SFRP2 antibody (rabbit polyclonal IgG; H-140; 1:75; Santa Cruz Biotechnology, Santa Cruz, CA), washed, and incubated for 10 min with the secondary antibody (biotinylated polylink; DAKO). Diaminobenzidin (DAKO) was used for antibody detection. In negative controls the primary antibody was omitted. An experienced breast cancer pathologist (N.B.) scored the immunohistochemical staining according to the scoring system suggested by Remmele and Stegner 40. Feasibility of the antibody for immunohistochemical analysis of breast tissue has been previously demonstrated e.g. by Lee et al. 42.

The benign cell lines HMEC and MCF10A as well as the cancerous breast cell lines BT20, BT474, Hs578T, MCF7, MDA-MB-231, MDA-MB-453, SKBR3, T47D and ZR75-1 were obtained from the American Type Culture Collection (Rockville, MA) and cultured as recommended by the vendor.

RNA isolation, RT-PCR and SYBR Green I realtime PCR (Roche Diagnostics, Mannheim, Germany) were performed as described elsewhere 27. Quality of cDNA was checked after each preparation by standard RT-PCR using glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) primers that yield an amplification product of 510 bp. To ensure experiment accuracy, all quantitative measurements were performed in triplicate. Intron-spanning primer sequences and cycle conditions are given in additional file 2.

Of the genomic DNA, 1 μg was bisulfite-modified using the EZ DNA Methylation Kit (Zymo Research, Orange, CA) according to the manufacturer's recommendations. The final precipitate was eluted in 20 μl TRIS buffer (10 mM). For MSP, one μl of modified DNA was amplified using MSP primers (see additional file 2) that specifically recognized either the unmethylated or methylated gene promoter sequence after bisulfite-conversion. Each primer pair mapped to nine cytosine-phosphate-guanine dinucleotide (CpG) sites in order to specifically discriminate between methylated and non-methylated DNA. Further 11 non-CpG cytosines within the primer pair specific for methylated DNA and 13 non-CpG cytosines within the primer pair specific for non-methylated DNA guaranteed unequivocal amplification of bisulfite-converted DNA. Primers defined an amplicon between +19 and +163 relative to the transcription start site (+1) of the SFRP2 gene. Reaction volumes of 25 μl contained 1 × MSP-buffer 44, 400 nM each of methylation and non-methylation-specific primers, respectively, and 1.25 mM of dNTPs. One drop of mineral oil was added to the reaction tube. The PCR was initiated as "Hot Start" PCR at 94°C and held at 80°C before the addition of 1.25 units Taq DNA polymerase (Promega, Madison, WI). Cycle conditions were: 95°C for 5 min, 35 cycles of 95°C for 30 s, 60°C for 30 s, 72°C for 40 s and a final extension at 72°C for 5 min. Blood lymphocyte DNA from a healthy donor was bisulfite-modified to serve as a control for the unmethylated promoter sequence 45, DNA from the cancerous breast cell line BT20 served as control for methylated alleles. Amplification products were visualized on 3% low range ultra agarose gel (Bio-Rad Laboratories, Hercules, CA) containing ethidium bromide and illuminated under ultraviolet (UV) light.

Quantitative Pyrosequencing of a SFRP2 promoter fragment was performed by use of a Pyromark ID device, PyroGoldSQA Reagent Kit and Pyro Q-CpG software (Biotage, Uppsala, Sweden). Initially, a 291 bp fragment of the SFRP2 promoter (relative position -28 to +263), covering the hybridization sites for MSP primers, was amplified with degenerate primers irrespective of the methylation status, which assures unbiased DNA amplification. To enable single strand preparation the reverse primer was 5'-biotinylated. Reaction volumes of 50 μl contained 1 × GoTaq buffer, 2.5 units GoTaq polymerase (Promega), 2.5 mM of MgCl₂, 400 nM of primers, 500 nM of each dNTP, and 3 μl of bisulfite-converted DNA as template. Reactions were initiated as "Hot Start" PCR at 95°C for 3 min and held at 80°C before addition of Taq polymerase. Cycle conditions were: 94°C for 3 min, 50 cycles of 94°C for 15 sec, 58°C for 30 sec, 72°C for 30 sec, and a final extension step at 72°C for 5 min. PCR was carried out in a PTC-200 cycler (Bio-Rad, formerly MJ Research, Hercules, CA). Prior to sequencing, aliquots of the amplificate were analyzed on a 2% agarose gel containing ethidium bromide under UV light. Single strand separation of the remaining amplificate (40 μl) was performed with a PyroMark Vacuum Prep Workstation (Biotage). Amplificate was immobilized to Streptavidin-Sepharose HP beads (Amersham Biosciences, Uppsala, Sweden), washed, denatured and the biotinylated strands were released into 40 μl of annealing buffer containing 400 nM of forward sequencing primer. Sequencing started with position +3 (relative to the TSS) and was continued to position +167, covering a total of 22 sequential CpG sites. The following sequence represents the SFRP2 promoter sequence that was analyzed by pyrosequencing: AYGGTTTATTTTGTTTTTTYGGGTYGGAGT TTTTYGGAGTTGYGYGYGGGTT TGTAGTGTTTYGTTYGYGTTGTTTTTTYGGTGTTTYGTTTTTTYGYGTT TTAGTYGTYGGTTGTT AGTTTTTYGGGGTTTYGAGTYGTATTTAGYGAAGAGAGYGGGTTYGG.

Universal bisulfite-converted polymethylated and unmethylated DNA (Epi Tect Control DNA Set; Qiagen, Hilden, Germany) served as technical controls for SFRP2 methylation and non-methylation, respectively. Pyrosequencing primers are available on request.

We plated cells at 3 × 10⁴ cells/cm² in a six-well plate on day 0. The demethylating agent DAC (Sigma-Aldrich, Deisenheim, Germany) was added to a final concentration of 1 μM in fresh medium on days 1, 2 and 3. Additionally, 300 nM TSA (Sigma-Aldrich) was added on day 3. Cells were harvested on day 4 for RNA and DNA extraction. Control cells were incubated without the addition of DAC or TSA and fresh medium was also supplied on days 1, 2 and 3.

Cells were seeded at a density of 3 × 10⁴ cells/cm² and transfected 24 hours after incubation with 100 ng/cm² of plasmid DNA in the following manner: 100 ng of empty pCMV-hemagglutinin (HA) vector control (Clontech, Heidelberg, Germany), or 50 ng of pCMV-HA + 50 ng pCMV-HA/SFRP2, or 50 ng of pCMV-HA + 50 ng pcDNA3.1-HisA/WNT1, or 50 ng pCMV-HA/SFRP2 + 50 ng pcDNA3.1-HisA/WNT1 19 applying the FuGENE 6 transfection system (Roche Diagnostics) and a 3:1 transfection ratio according to the manufacturer's instructions.

MCF10A cells were transfected in 96-well plates as described above and an XTT-proliferation assay (Roche Diagnostics) was performed on day 0, 1, 2 and 3 after transfection by determining the optical density of the supernatants at 480 nm minus the optical density of the supernatants at 690 nm. To enhance experimental accuracy, six replicas were seeded. For the colony formation assay cells were transfected accordingly in six-well plates and kept for three weeks under selective force of the antibiotic G418 (700 μg/ml) (Invitrogen, Carlsbad, CA). After incubation, colonies were washed with phosphate-buffered saline, fixed and stained for 30 minutes (0.25% crystal violet in 10% formalin/80% methanol), washed three times with distilled water and photographed.

Statistical analyses were completed using the software package SPSS, version 14.0 (SPSS Inc., Chicago, IL). Differences were considered significant when P-values were below 0.05. A two-sided non-parametric Mann-Whitney U-test and a paired student's t-test were performed to analyze differences in expression levels. Associations between metrical variables were determined by a linear regression analysis. To study statistical associations between clinicopathological factors and SFRP2 expression or SFRP2 promoter methylation status contingency-tables and two-sided Fisher's exact tests were accomplished. Survival curves were calculated using the Kaplan-Meier method, with significance evaluated by two-sided log-rank statistics. Overall survival (OS) (n = 136 for MSP samples) was measured from the day of surgery until tumor-related death (20.6%, n = 28) and was censored for patients alive at last contact (69.1%, n = 94), in case of death unrelated to the tumor (3.7%, n = 5) or when the death cause was unknown (6.6%, n = 9). Disease-free survival (DFS) (n = 136 for MSP samples) was measured from surgery until local or distant relapse (36.8%, n = 50) and censored for patients alive without evidence of relapse at the last follow-up (63.2%, n = 86).

Initiating our analysis, we used the RT-PCR technique to evaluate whether SFRP2 mRNA is differentially expressed in human breast cell lines. SFRP2 mRNA expression was detectable in the benign cell line HMEC, whilst its expression was absent in the cancerous cell lines MDA-MB-231, MCF7, SKBR3, T47D, MDA-MB-453, BT20 and BT474 (Figure 1A). Of the cancerous cell lines only Hs578T exhibited strong SFRP2 expression. However, SFRP2 expression was also absent in the benign cell line MCF10A. Additionally, in a commercially available cDNA of human normal breast tissue (Clontech, Heidelberg, Germany) SFRP2 was strongly expressed while its expression was substantially reduced in Clontech's cDNA of primary breast carcinoma.

Analysis of the SFRP2 gene promoter on chromosome 4q31 46 using the genomic DNA information contained in Ensembl contig ENSG00000145423 47 revealed a CpG island (CGI) between base position -818 to +743 relative to the expected transcription start site (+1), according to the CGI definition of Takai and Jones 48. Further exploration of this CGI using Methprimer software 49 identified three regions of particularly high CpG density (-399 to -151, -6 to +332, and +486 to +685) (Figure 1B). Since sequence integrity of the first SFRP2 exon was demonstrated to be most essential for efficient RNA transcription in a luciferase promoter assay 24 we chose the central CGI (-6 to +332) for subsequent methylation analysis by application of the highly specific MSP primers described by Suzuki et al. 29 and others 2437. First, we assessed SFRP2 promoter methylation in eight cancerous and two non-cancerous cell lines. Six of the analyzed cell lines (MCF10A, MDA-MB-231, MCF7, MDA-MB-453, BT20 and BT474) exhibited a methylated SFRP2 promoter sequence in the analyzed region (Figure 1C). Two cell lines (SKBR3 and T47D) showed partial promoter methylation, since a mixture of unmethylated and methylated DNA sequence could be detected in the same sample. One malignant cell line (Hs578T) and benign HMEC cells revealed solely unmethylated SFRP2 promoter sequence. This result correlates with the above described finding that Hs578T and HMEC cells exhibited strong SFRP2 mRNA expression whereas in all cell lines with aberrant methylation SFRP2 mRNA expression was absent. Interestingly, SFRP2 mRNA was not expressed at detectable levels from unmethyated alleles in SKBR3 and T47D, which may be due to repressing mechanisms that are codominant to the effect of promoter methylation in these cells.

We next asked whether SFRP2 promoter methylation is functionally associated with loss of SFRP2 mRNA expression in breast cell lines. To address this question we treated four representative breast cancer cell lines for four days with 1 μM of the methyltransferase inhibitor DAC and one day with 300 nM of the histone deacetylase inhibitor TSA. As seen in Figure 2A, in three cell lines (BT20, MCF7 and T47D) the DAC/TSA treatment resulted in a clear increase of unmethylated SFRP2 promoter sequence, only in SKBR3 cells there was no visible difference detectable. Interestingly, BT20 and MCF7 showed demethylation after the addition of DAC alone, whereas in T47D cells such effect was only achieved by the combination of DAC and TSA. Before the treatment, none of the cell lines showed detectable SFRP2 mRNA expression (Figure 2B). The treatment with 1 μM DAC alone did not reverse SFRP2 gene repression in these cells. TSA, in contrast, was able to induce some SFRP2 expression in SKBR3 and T47D cells. Importantly, only the combination of both drugs substantially restored SFRP2 mRNA expression in all analyzed cell lines.

To further prove that the demethylating treatment did not result in unspecific upregulation of gene expression, we determined the expression of the growth promoting gene cyclin D1, which is a direct read-out gene of active Wnt signaling 50 and whose expression is commonly elevated in breast cancer 51. Using realtime PCR we observed that cyclin D1 mRNA expression was significantly reduced (P = 0.029, two-tailed Mann-Whitney U-test) after the demethylation treatment (43-fold in BT20, 14-fold in SKBR3, 8-fold in T47D and 6-fold in MCF7, Figure 2C), suggesting that inhibitors of proliferation, such as SFRPs which are downregulated in breast cancer cell lines, have been reactivated and were able to block Wnt signaling in these cells.

In order to answer the question whether SFRP2 promoter methylation occurs in primary breast carcinoma as well we analyzed 199 mammary tumor samples by MSP. For 20 breast tumors corresponding tissues of histological normal breast epithelium were available and analyzed in parallel. Representative results are shown in Figure 3. In total, 165 of 199 tumor samples (83%) showed SFRP2 promoter methylation as a PCR product could be amplified with methylation-specific primers (e.g. #103 in Figure 3), and 34 of 199 tumors (17%) showed no evidence of promoter methylation since exclusive amplification signals were obtained with primers specific to unmethylated DNA (e.g. #108 in Figure 3). None of the 20 matched normal breast tissue samples showed a methylation signal (e.g. #03 in Figure 3). Tumor samples generally exhibited also unmethylated promoter sequences due to possible contamination with small amounts of stromal and endothelial cells, as has also been described by Suzuki et al. 19. To further confirm that SFRP2 promoter methylation in breast cancer is restricted to malignant tissue, we analyzed 17 cancer-unrelated normal breast samples. Again, none of these normal breast tissues (e.g. Figure 3) harbored detectable SFRP2 methylation.

In order to investigate the association of SFRP2 promoter methylation with transcriptional silencing in a quantitative manner we assessed in parallel SFRP2 mRNA expression by realtime PCR and SFRP2 methylation by quantitative Pyrosequencing in breast cell lines. The technical controls for Pyrosequencing revealed median methylation of 4% (unmethylated control) and 90% (methylated control) in the 22 sequential CpG sites, by this defining the detection limits of this assay. HMEC cells exhibited abundant SFRP2 mRNA expression (Δcycle threshold (C_T) GAPDH:SFRP2 = 5.1) (Figure 4A) together with median SFRP2 methylation of 4%, in contrast to e.g. BT20 cells which exhibited mean methylation of 83% together with absence of SFRP2 mRNA expression (ΔC_T GAPDH:SFRP2 = 19.9). Median SFRP2 methylation for Hs578T, MCF10A, MCF7, MDA-MB-231, SKBR3 and T47D was 5%, 73%, 73%, 71%, 61% and 42%, respectively. A direct comparison of SFRP2 mRNA expression and SFRP2 promoter methylation indicates that in breast cell lines SFRP2 methylation is correlated with loss of SFRP2 mRNA expression (P < 0.001; Figure 4B). Thus, when applying an empiric cut-off of > 5% to discriminate between SFRP2 methylation and non-methylation, all semi-quantitative results for SFRP2 methylation in breast cell lines obtained by MSP could be confirmed by the quantitative Pyrosequencing assay.

Immunohistochemical analysis was used to investigate SFRP2 protein expression in normal and malignant breast tissue. A total of 125 informative breast cancer cases, four DCIS and 10 normal breast tissues were analyzed. Intensity of immunohistochemical staining was evaluated using a semi-quantitative IRS 39. SFRP2 protein was clearly detectable in 90% (9/10) of normal breast tissue samples analyzed, as defined by an IRS ≥ 8 (Figure 5C). The mean expression was determined to be IRS = 7.5 (range 3–8; standard deviation (SD) ± 1.6), and median expression to be IRS = 8. Expression was predominantly localized in luminal and basal epithelial cells of the normal breast while weak expression was detectable in adjacent stromal cells. In four DCIS, SFRP2 expression was slightly reduced (mean IRS = 6.5; range 4–8; SD ± 1.9; median IRS = 7; Figure 5D, E) with one sample showing strong reduction (25%, IRS = 4). However, the mean expression in invasive breast carcinomas was determined to be IRS = 4.6 (range 0–8; SD ± 2.1) and median expression to be IRS = 4. Invasive breast carcinomas showed strongly reduced or complete loss (IRS ≤ 4) of SFRP2 expression in 74% (93/125) of cases (Figure 5F, G). The SFRP2 expression difference between tumors and normal breast tissues was statistically significant (P = 0.001).

Clinicopathological characteristics of breast cancer patients were first correlated with SFRP2 protein expression for descriptive data analysis (Table 1). Loss of SFRP2 expression in tumor tissue (IRS ≤ 4) was not associated with age at diagnosis, tumor size, histological grade, histological type, estrogen/progesterone receptor status, Her2 status or expression of p53. A prevalence of more abundant SFRP2 expression was detected in node negative breast tumors (P = 0.033). In univariate survival analysis using log-rank test, loss of SFRP2 protein expression was not associated with DFS (P = 0.237), but a weak trend was detected towards reduced OS (P = 0.071) (Figure 6). SFRP2 promoter methylation in breast carcinomas was not associated with age at diagnosis, tumor size, lymph node status, histological grade, histological type, or estrogen/progesterone receptor status (Table 2). In univariate survival analysis, lymph node status, histological grade and histological type were significantly associated with DFS; lymph node status and histological grade were significantly associated with OS (Table 3). However, SFRP2 methylation was neither associated with DFS (P = 0.192) nor with OS intervals (P = 0.686).

Finally, we asked whether SFRP2 influences proliferation rates in breast cell lines. For gain-of-function experiments we chose mammary MCF10A cells since these cells were found to lack SFRP2 mRNA expression. As shown in Figure 7A, WNT1 overexpressing MCF10A cells notably increased proliferation compared to cells mock-transfected with empty vector. The equimolar co-transfection with SFRP2resulted in a decreased proliferation rate as compared to WNT1 alone. Furthermore, SFRP2-transfectants revealed a slightly reduced proliferation rate compared to cells containing empty vector. In order to support these findings we performed a colony formation assay and selected transfected clones for three weeks by the antibiotic G418. Representative results are shown in Figure 7B. Controls assured the feasibility of this assay, showing that wild type cells without G418 abundantly form colonies, whereas in the presence of G418 wild type cells fail to survive. SFRP2-transfected cells exhibit a reduced number of colonies as compared to mock-transfected cells. The difference in colony numbers from three independent experiments was statistically significant (P = 0.045, two-tailed Mann-Whitney U-test).