Inactivation of FBXW7/hCDC4-β expression by promoter hypermethylation is associated with favorable prognosis in primary breast cancer

A total of 161 primary breast tumor specimens from two breast cancer cohorts were included in this study. Among the 161 samples in which FBXW7/hCDC4-β promoter methylation was analysed, RNA was available from 139 samples, which was further processed for cDNA synthesis and FBXW7/hCDC4-β expression analysis as described below. A total of 68 cases of primary breast cancer were obtained from patients diagnosed at the Department of Obstetrics and Gynecology of the Innsbruck Medical University of Austria (cohort 1) and 93 samples were obtained from patients at the Department of Pathology of Uppsala University, Uppsala, Sweden (cohort 2) . All patients underwent resection of the tumor during surgery and specimens were processed by pathologists at the affiliated hospital. Samples were snap frozen in liquid nitrogen and stored at -80°C until RNA and DNA extraction.

Cohort 1: Clinicopathologic features for this collection of samples have been previously reported [26]. Patients were diagnosed and operated on between 1990 and 2001 and the median age of the patients included in this study was 64 years (range 33 to 88). Patients were treated in compliance with the national recommendations at the time. Forty-one and 27 patients underwent a lumpectomy or a mastectomy, respectively. Thirty-eight patients received loco-regional radiation. Thirty-five patients received adjuvant combination chemotherapy CMF (cyclophosphamide, methotrexate and 5-fluorouracil), and 33 patients received adjuvant anti-hormonal therapy. All samples were collected during surgery in compliance with and approved by the Institutional Review Board and with informed consent from the patients. p53 was sequenced in tumor samples from cohort 1 using either genomic DNA or cDNA as a template with primers derived from intronic or gene specific sequences (primer sequences can be obtained from the authors upon request). PCR amplification and purification was performed as previously described [4, 15] and sequenced at Eurofins MWG Operon (Eurofins MWG Operon, Ebersberg, Germany).

Cohort 2: Breast cancer patients were diagnosed and operated on between 1987 and 1989 at Uppsala University Hospital, Uppsala, Sweden. The median age for the patients used in this study was 63 years (range: 28 to 94). Clinicopathological data and treatment have been previously reported [27, 28]. Briefly, patients were operated on and received postoperative radiotherapy. When adjuvant tamoxifen was given, some patients received this treatment as part of a randomized study comparing two versus five years, and chemotherapy, mostly CMF according to standards in those days [27, 28]. Ethical permission was obtained from the ethical committee at Karolinska Institute and with informed consent from the patients. Data on p53 mutational status in this series of breast tumor specimens have been described [27].

Genomic DNA and RNA from fresh frozen tumor tissue were isolated as previously described [4, 15]. RNA from various normal tissues was purchased from ABI (Applied Biosystems, Foster city, CA, USA). Normal breast tissue was obtained from reduction surgeries as well as from noncancerous tissue DNA extracted from paraffin-embedded breast cancer specimens. cDNA was synthesized from 2 μg of total RNA using Superscript First-strand Synthesis System (Invitrogen, Carlsbad, CA, USA).

A total of 60 human cancer cell lines, originating from breast, brain, prostate, kidney, blood, cervix, lung, skin, bone and thyroid (See Additional file 1 for details), were analyzed for FBXW7/hCDC4 expression and promoter methylation. Cell lines were maintained and cultured according to American Type Culture Collection (ATCC) guidelines or as previously described [15, 29]. All plasmid transfections were performed using LT1 transfection reagent (MIRUS, Madison, WI, USA), as recommended by the manufacturer's protocol. For experiments evaluating the effects of demethylation, cell lines maintained in appropriate media were treated with 5-aza-2'-deoxycytodine (5-aza-dC) (Sigma-Aldrich, St Louis, MO, USA) or DMSO for three to five days.

A 1.6 kb genomic region (-1293 bp to +309 bp from the transcription start site (TSS)) of the FBXW7/hCDC4-β isoform containing 18 CpG sites was amplified by PCR, cloned into the pSC-A vector (Stratagene, Santa Clara, CA, USA) and sequenced. The specific primers used for amplification of the FBXW7/hCDC4-β promoter, were as follows; FP1: 5'-GCC ATT TAC CAC CAT AGC AGA GAG TA-3', RP1: 5'-GCT ATG TGA TTG TGT GTG TAT GCC-3'. Two shorter versions of the promoter, containing 11 and 6 CpGs respectively, were also amplified using the following forward primers, FP2: 5'-AGA CTT ATT TGT GGA AAT GTT CCT TGC TA or FP3: 5'-GCA TTG CTG AAT CCT GGA CTG CAC C and the reverse primer, RP1. Each promoter region was subcloned into pGL3 vector (Promega, Madison, WI, USA) after KpnI and BglII restriction digestion (New England Biolabs, Ipswich, MA, USA). The resulting promoter constructs were termed pGL3hCDC4β-1.6, pGL3hCDC4β-0.8. and pGL3hCDC4β-0.6. The pRL-SV40 Renilla luciferase plasmid was obtained from Promega (Madison). PCR reactions were performed in a BioRad thermocycler (Techne, Burlington, NJ, USA).

Luciferase activities in cell lysates from cells transfected with different pGL3 constructs and pRL-SV40 control plasmid were measured in a luminometer (VICTOR3, PerkinElmer, Waltham, MA, USA) using the Dual-Luciferase Reporter Assay System according to the manufacturer's protocol (Promega). Luciferase activities were quantified and fold change was averaged from at least three separate experiments performed in triplicates.

Methylation of the FBXW7/hCDC4-β promoter was examined in cell lines, normal breast and primary tumors by several methods. Bisulfite-modified genomic DNA was prepared and CpG methylation was analysed by bisulfite-sequence analysis as previously described [30, 31]. The methylation status of the complete FBXW7/hCDC4-β promoter was determined by sequence analysis of at least five individual clones from cell lines if not otherwise stated. For screening of relative methylation levels, the McrBc restriction enzyme (New England Biolabs) was used. McrBc recognize and cuts pairs of purine-methyl cytosines (recognition sequence R^mC(N)_55-103R^mC) and subsequent PCR amplification of methylated DNA segments in comparison with an unmethylated segments thus denotes the methylation status of the region of interest. Briefly, 200 ng of genomic DNA was treated with or without 0.5 unit of McrBc enzyme for 1 hr at 37°C in the reaction buffer provided by the supplier. Each sample was heat inactivated and subsequently amplified by PCR using FP1 and RP1 primers. PCR amplification was performed with 100 ng DNA as template in the following conditions; two minutes denaturation and 30 cycles of amplification (94°C for 30 s, 64°C for 30 s, and 68°C for one minute) using Titanium™taq DNA polymerase (BD) (Becton dickinson, Franklin Lakes, NJ, USA). PCR products were resolved on agarose gels and band intensities were quantified by Image J software (ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA) [32] The mean McrBc ratio (band intensity of McrBc digested DNA divided by undigested DNA) in unmethylated DNA samples obtained from normal breast tissue and tumor-derived cell lines (verified through bisulfite sequence analysis) was 0.808 with a standard deviation (SD) of 0.11. Test samples were judged as methylated if their McrBc ratio had a decreased value greater than 2 SD as compared to the unmethylated control samples. A McrBc methylation ratio of 0.6 was thus used as a cutoff. The McrBc results were also confirmed by restriction digestion of PCR products using Taq I and HpyCHIV enzymes (cuts methylated CG11 and CG18, respectively, data not shown). To verify the effect of promoter methylation on FBXW7/hCDC4-β expression, luciferase reporter constructs were methylated in vitro with Sss I methylase (New England Biolabs) as previously described [33]. Screening for methylation of FBXW7/hCDC4-α in primary tumor samples was carried out using McrBc restriction digestion analysis. The FBXW7/hCDC4-α promoter was amplified using primers; FA1: 5'-AGA CCC AGG AAG AGG AAA AGA GGA-3', RA1: 5'-TGG GTT GGT TCC CTT CCT CCT TC-3' and analysed for methylation as described above.

FBXW7/hCDC4 isoform-specific primers and TaqMan probes for quantitative real-time PCR analysis of different FBXW7/hCDC4 isoforms have been described [12]. The ^ΔΔCt method of relative quantification was performed to determine relative mRNA expression in each sample. The relative expression level of each FBXW7/hCDC4 isoform was obtained by normalizing the expression of FBXW7/hCDC4 mRNA to GAPDH mRNA expression. Primers and conditions for semi-quantitative RT-PCR of FBXW7/hCDC4 isoforms have been described [29]. Amplification of GAPDH mRNA served as an internal control. Oestrogen and progesterone receptor (ER/PR) status was determined by immunohistochemistry as previously described [34].

The association between patient clinicopathological characteristics (such as oestrogen receptor, progesterone receptor, lymph node involvement, grade, stage and age) and methylation status of the FBXW7/hCDC4-β isoform, in addition to P53 mutation, was determined using the Fisher's exact test. Differences in FBXW7/hCDC4-β expression between methylated and unmethylated groups were analysed by means of the (non-parametric) Wilcoxon-Mann-Whitney test. Univariate analyses of survival were conducted by use of the Kaplan-Meier method. Further, the risk of dying in women with methylation of the FBXW7/hCDC4-β isoform compared with women with unmethylated FBXW7/hCDC4-β isoform was modelled by use of multivariable proportional hazards (Cox) models, adjusted for possible confounders of survival such as age at diagnosis, date of primary tumour surgery, oestrogen receptor and progesterone receptor status, and P53 mutation. An arbitrary level of 5% statistical significance was used. Finally, data preparation and analysis was done using the SAS Statistical package, version 9.2 (SAS Institute Inc., Cary, NC, USA).

Mammals express three Fbxw7/hCdc4 splice variants designated α, β, and γ, each encoding a unique N-terminal protein sequence fused to 10 downstream exons, suggesting non-redundant functions as mentioned above. Semi-quantitative RT-PCR analysis of the different FBXW7/hCDC4 isoforms revealed substantial differential expression of the FBXW7/hCDC4-β transcript in specific cell lines from various tumor tissues (Figure 1a). The immortalized breast epithelial cell lines IME and MCF10, expressed high levels of FBXW7/hCDC4-β compared to some breast cancer cell lines with low or absent FBXW7/hCDC4-β expression (Figure 1a, left). This variation in mRNA expression was not generally observed for the FBXW7/hCDC4-α isoform, which was the most abundant and ubiquitously expressed FBXW7/hCDC4 transcript in most cell lines examined (Figure 1a).

As previously reported [35, 36], the beta isoform was expressed at very high levels in tissue from normal brain (Figure S1 in Additional file 1). Significant expression was also observed in tissues from normal breast, ovary and cervix, compared to other tissues with low (spleen, thyroid, liver) or absent (skeletal muscle) FBXW7/hCDC4-β expression (Figure S1 in Additional file 1). To examine whether loss of FBXW7/hCDC4-β expression correlated with hypermethylation of its promoter, we first examined the sequence 1.3 kb upstream and 0.3 kb downstream of the transcription initiation start site in FBXW7/hCDC4-β. Eighteen CpG sites were distributed throughout this region (Figure 1b). To this end, we used bisulfite sequence analysis to determine the methylation status of each of these CpGs in five different cell lines with low/absent FBXW7/hCDC4-β expression (HeLa, U266, PEER, Daudi and DLD1) and five cell lines with high expression (IME, MB-468, T-ALL, 293A and MOLT4). Regarding the methylation status, cell lines were found to fall into two distinct groups; one demonstrating methylation of the majority of CpGs (90 to 100%) correlating with low expression, and the other exhibiting mostly unmethylated CpGs and showing high expression (Figure 1c). Screening for methylation was also carried out using the restriction enzyme McrBc, a methylation specific endonuclease, which cuts DNA containing 5-methylcytosine in the context of a second, arbitrarily spaced methylcytosine, and does not cleave unmethylated DNA. As shown in Figure 1d, screening for methylation of this region with the McrBc enzyme recapitulated the methylation results obtained by bisulphate sequence analysis.

Assessing methylation by McrBc digestion in 50 additional cell lines from various tissues (Table S1 in Additional file 2) demonstrated methylation in 26 of 60 (43%). FBXW7/hCDC4-β expression was next analyzed by real time reverse transcription PCR. A significant difference in expression between methylated and unmethylated cell lines (Figure 2a, P = 0.0001) was found with lower expression in methylated (median expression = 0.02, range; 0 to 0.9) compared to unmethylated cell lines (median expression = 1.6, range; 0 to 8.1). No correlation between methylation of the FBXW7/hCDC4-β promoter and expression of FBXW7/hCDC4-α mRNA was found (data not shown).

To establish a direct link between methylation and silencing of FBXW7/hCDC4-β expression we treated HeLa, BT-20, U2OS and BT-474 cells with the DNA methylation inhibitor 5-aza-dC. As can be seen in Figure 2b, 5-aza-dC treatment increased FBXW7/hCDC4-β mRNA expression levels in methylated cell lines (HeLa and BT-20), with an initial low expression, in contrast to the unchanged high FBXW7/hCDC4-β levels of unmethylated cell lines (U2OS and BT-474) (Figure S2A in Additional file 3). Demethylation of the promoter upon 5-aza-dC treatment was confirmed by McrBc digestion (Figure 2c) and sequencing of bisulfite treated DNA (data not shown). As a control, FBXW7/hCDC4-α mRNA expression was measured after 5-aza-dC treatment. No significant increase of FBXW7/hCDC4-α levels was observed in any of the cell lines examined (Figure S2B in Additional file 4 and data not shown).

To confirm that the FBXW7/hCDC4-β promoter region possesses transcriptional activity, we performed luciferase reporter assays with the 1.6 kb genomic region covering all 18 CpGs (Figure 1b) and two shorter promoter regions (0.8 and 0.6 kb), being deletion constructs of the 1.6 kb region mentioned above. Robust reporter activity was observed in cell lines of different origins with all three constructs, albeit a reduced activity was observed with the shorter promoter constructs (data not shown). To further evaluate the effect of methylation on promoter activity, we next performed reporter assays using transient transfection of constructs where the FBXW7/hCDC4-β promoter was methylated in vitro using the Sss I methylase as previously described [33]. As shown in Figure 2d, in vitro methylation suppressed reporter activity confirming that methylation of the promoter abrogates FBXW7/hCDC4-β expression.

Together, these data demonstrate that CpG-methylation correlates with loss of FBXW7/hCDC4-β expression in tumor cell lines. These data also indicate that methylation could be a significant factor in regulating FBXW7/hCDC4-β expression in some malignancies, including breast cancer.

Based on the results above, we next explored the occurrence and role of FBXW7/hCDC4-β methylation in primary breast cancer specimens. To exclude the possibility that methylation detection is affected by contaminating noncancerous cells, we first analysed FBXW7/hCDC4-β promoter methylation in normal breast tissues (obtained from normal breast reductions) by McrBc digestion (Figure 3a) and bisulfite sequence analysis (data not shown). Importantly, methylation was absent in normal breast tissues as well as in noncancerous tissue DNA extracted from a paraffin-embedded breast cancer specimen (Figure 3a and data not shown). We next analyzed FBXW7/hCDC4-β methylation in two cohorts of breast cancer specimens in which RNA was available and FBXW7/hCDC4 mRNA expression could be analysed. A total of 71 out of 139 (51%) patient samples showed methylation of the promoter as defined by McrBc digestion, compared to its undigested control (Figure 3a). Tumor specimens were classified as low or high methylation (see Materials and methods for details on quantification and classification of promoter methylation), divided into two groups and compared with the mRNA expression levels of FBXW7/hCDC4-β. Similar to the results in tumor cell lines, a statistically significant difference was found between the groups (Figure 3b, P = 0.0002). The methylated group showed significantly lower expression of FBXW7/hCDC4-β (median expression = 0.13, range; 0 to 1.43) as compared to the unmethylated group, which had higher expression levels (median expression = 0.30, range; 0 to 8.83).

Data on methylation status and clinicopathological parameters were available for a total of 161 tumor specimens (68 cases from cohort 1 and 93 cases from cohort 2). The associations between FBXW7/hCDC4-β methylation and clinicopathological features are summarized in Table 1. Methylation was significantly associated with high-grade tumors (P = 0.017) and a trend towards an association with estrogen receptor-negative tumors was also observed in the methylated group (Table 1). No significant associations were observed between methylation of FBXW7/hCDC4-β with age, stage and lymph node status (Table 1) or with p53 mutational status (Table S2 in Additional file 5). As previously reported [27], p53 mutation was associated with high-grade and receptor-negative (ER and PR) tumors (Table S2 in Additional file 5).

To evaluate whether methylation of FBXW7/hCDC4-β was an independent prognostic factor in breast cancer, we examined the significance of methylation using multivariate analysis, including adjustment for other factors known to be associated with clinical outcome. As the cohorts from the two centers were treated during different time periods, and thus partly using different treatment protocols, this outcome analysis was performed separately for the two groups. This also enabled a validation of the findings in the two independent cohorts. In both cohorts, methylation of FBXW7/hCDC4-β was found to be associated with an approximately 50% decreased risk of death (cohort 1 hazard ratio 0.53 (0.23 to 1.23) and cohort 2 (HR) 0.50 (95% CI 0.23 to 1.08), Table 2). We also performed log rank analysis of Kaplan-Meier curves for overall survival in defined prognostic subgroups. Methylation was associated with a significantly improved overall survival in patients with lymph node metastasis, in cohort 1 (Figure 4a, P = 0.017). In this cohort, patients with ER receptor-negative tumors additionally demonstrated a clear trend, although not statistically significant (P = 0.111, data not shown), for improved survival in methylated tumors. In line with the data from cohort 1, the overall survival in women with lymph node-positive tumors from cohort 2 was higher in the group whose tumors demonstrated a methylated FBXW7/hCDC4-β promoter, compared to women with an unmethylated promoter (Figure 4b). However, this difference was not statistically significant, possibly due to the smaller number of patients in this subgroup from cohort 2 (n = 35 out of 93), in comparison with cohort 1 (n = 41 out 68). No clear conclusions could be drawn from cohort 2 regarding the prognostic significance for FBXW7/hCDC4-β methylation in the ER-negative group, due to the limited number of patients (n = 13). Methylation was also associated with a significantly improved overall survival in the p53 mutated subgroup in cohort 2 (Figure 4c, P = 0.048). Thus, in two subgroups with known adverse prognostic features (that is, patients with lymph node metastasis and patients whose tumors harbored p53 mutation) FBXW7/hCDC4-β methylation was found to be associated with improved survival. No association between survival and methylation of FBXW7/hCDC4-β was found in patients with lymph node negative tumors (Figure S3 in Additional file 6). No clear conclusions could be drawn from cohort 1 regarding the prognostic significance for patients with p53 mutation and FBXW7/hCDC4-β methylation due to the limited number of patients and events in the p53 mutated/FBXW7/hCDC4-β methylated subgroup.