|Home | About | Journals | Submit | Contact Us | Français|
Calcitriol (1α, 25(OH)2-Vitamin D3) binds to the vitamin D receptor (VDR) and regulates differentiation of the normal mammary gland, and may therefore be useful in breast cancer treatment or prevention. Many breast cancer cells are, however, resistant to Calcitriol. In this study, we investigated the resistance mechanism and the role of epigenetic silencing of VDR by promoter hypermethylation. Bisulfite sequencing of the VDR promoter region revealed methylated CpG islands at −700 base pairs (bp) upstream and near the transcription start site. VDR CpG islands were demethylated by 5′deoxy-azacytidine treatment, and this was accompanied by a parallel increase in VDR mRNA levels in breast cancer cell lines. Quantitative methylation-specific PCR analyses confirmed hypermethylation of these CpG islands in primary tumors, and its absence in normal breast tissue. VDR transcripts detected in breast cancers were predominantly 5′-truncated, while normal breast tissue expressed full-length transcripts. Consistent with this observation, genes containing the VDR-responsive element (VDRE), such as cytochrome p450 hydroxylases, p21 or C/EBP were underexpressed in breast cancers compared to normal breast samples. Expression of the active longer transcripts of VDR was restored with 5′deoxy-Azacytidine (AZA) treatment, with a concurrent increase in expression of VDRE-containing genes. Thus, promoter methylation-mediated silencing of expression of the functional variants of VDR may contribute to reduced expression of downstream effectors of the VDR pathway and subsequent Calcitriol insensitivity in breast cancer. These data suggest that pharmacological reversal of VDR methylation may re-establish breast cancer cell susceptibility to differentiation therapy using Calcitriol.
Extensive biological and epidemiological data suggest that Calcitriol, the active form of Vitamin D (1α, 25(OH)2-Vitamin D3, VTD), may play an important role in cancer prevention.1 Calcitriol regulates proliferation and induces differentiation of a wide variety of cells, including the normal mammary gland.2 The biological effects of Calcitriol are mediated through the vitamin D receptor (VDR), a nuclear transcription factor which binds to the Vitamin D responsive element (VDRE) present in the promoters of genes responsive to Calcitriol.3 Immunohistochemical data show that VDR expression is higher in differentiated cells than in proliferating cells, and that the Calcitriol signaling pathway participates in negative growth regulation of the mammary gland.4 Furthermore, VDR knockout mice show impaired ductal differentiation and branching in the mammary gland compared to wild type littermates.4
Despite the early promise of Calcitriol as an anti-proliferative and differentiating agent in cancer cells,5 attempts at its therapeutic exploitation have been disappointing, since clinical cancers and cancer cell lines often display vitamin D insensitivity. How directly this insensitivity can be implicated in the pathogenesis and pathology of breast cancer has not been resolved. Since VDR is not commonly mutated in cancer,6 Calcitriol insensitivity through other mechanisms is likely.7,8 While most reports have focused on altered patterns of histone acetylation in the VDR promoter,7 aberrant DNA methylation patterns have been documented in both colon cancer and endometrial cancer.9,10
In order to investigate the role of promoter methylation-induced silencing of VDR expression, we performed an in-silico analysis of the VDR gene,11 focusing on the evolutionarily well-conserved exons 1a and 1d (see Fig. 1).12,13 We found three CpG islands in an area spanning from −790 bp upstream to +380 bp downstream relative to the primary VDR transcription start site in exon 1a. This region contains regulatory elements, including several SP1 and AP-2 sites,14 where methylation may affect the binding of transcription factors.14 We then investigated the possibility that methylation-induced silencing of VDR in breast cancer might account for the Calcitriol insensitivity, and that this could be reversed using demethylating agents, thus mitigating Calcitriol resistance in breast cancer cells.
Calcitriol is a cell differentiating agent and drugs of this class are known to cause growth arrest and induction of differentiation-associated genes. To test the effect on breast cancer cells, four breast cancer cell lines (HS578T, 21PT, MCF7 and T47D) and the immortalized normal breast epithelial cell line HBL100 were treated for 96 h with Calcitriol (1.5 µM), with the demethylating agent AZA (5′deoxy-azacytidine, 5 µM), with Calcitriol + AZA, or with drug vehicle alone. Drug concentrations were selected on the basis of the dose response curves we obtained with our drug formulations in the cell lines studied, and concentrations designed to achieve LD50 concentrations for the individual agents were chosen in order to facilitate detection of additive drug effects. Cell viability was measured by MTT assay. As shown in Figure 2A, Calcitriol alone (light grey bars) had minimal effects on the viability of the five cell lines, compared to vehicle control (white bars). The effects of Calcitriol, however, were clearly amplified by AZA in all five cell lines, with stronger antiproliferative effects than with either agent alone. To determine if this correlated with the induction of VDR expression, Quantitative Real-Time PCR (QRT-PCR) analyses were performed using a commercial “Assay on Demand” VDR primer set, which produces an amplicon spanning exons 7 and 8, as well as our V primer set, producing an amplicon spanning exons 2 and 3, see Figure 1. Both of these assays detect all 5′ VDR transcript variants, and yielded superimposable results. As shown in Figure 2B, the overall expression level of VDR increased (VDR-T), with the strongest increase again seen when both Calcitriol and AZA were used in combination. Together these results suggest that sensitivity to Calcitriol is controlled by the level of expression of VDR, which in turn may be regulated by the status of DNA methylation of the VDR promoter.
Gene promoter hypermethylation is known to silence gene expression in many cancers. To determine whether the VDR gene promoter was hypermethylated in breast cancer cell lines, we first performed an in silico analysis of the VDR promoter to identify potential CpG islands.11 We then performed bisulfite sequencing in the region 790 bp upstream to 380 bp downstream of the exon 1a transcription start site after treatment of cells in the presence or absence of Calcitriol or AZA for 96 hrs (see Fig. 3). CpG dinucleotides are depicted as beads on a string representing the VDR promoter region. As can be seen from the top untreated control line (1) for each of the 5 cell lines examined, methylated CpG dinucleotides were found in several clusters. The most consistent methylation sites were around −760 and −480, with additional scattered methylation found around the transcriptional start site (TS). The pharmacological effects of treating the cell lines with Calcitriol (2) or AZA (3) are shown below the control lines. Calcitriol had no effect on methylation. Three main regions of AZA-responsive hypermethylation were observed. The first two hypermethylated regions flanked the upstream pair of SP1-binding sites around −760 bp as well as the NFγB binding site around −480, and the third flanked the transcriptional start site, but showed lower levels of methylation. The extent of hypermethylation was also less in HBL100 cells, consistent with their derivation from normal breast tissue.
To validate these in vitro data, we extracted DNA from eight freshly frozen human breast cancers, seven adjacent normal breast samples and three normal breast organoid preparations (see bottom section of Fig. 3). Bisulfite sequencing showed that the entire region of the VDR promoter remained largely unmethylated in the three normal organoid samples, whereas the breast tissue adjacent to breast cancer showed low levels of aberrant methylation, i.e., with ~5–15% of CpGs methylated (see lower portion of Fig. 3). In contrast, CpGs were highly methylated in the primary breast tumors at essentially the same residues as were observed in the breast cancer cell lines. More than 40–65% of the CpG dinucleotides were methylated in cancer tissues, albeit with some inter-sample variability.
To validate these results further, we assayed primary breast cancer tissues (n = 15; 10 fresh frozen and 5 formalin-fixed, paraffin-embedded (FFPE) tissues) and adjacent normal breast tissues (n = 7 fresh frozen samples), using methylated DNA (M) and unmethylated DNA (UN)-specific MSP primers, which were designed to interrogate the most consistently methylated CpGs identified in our sequencing experiments (see Fig. 3). As shown in Figure 4, breast cancer samples show significantly greater CpG hypermethylation (average 65%) than normal breast tissue (average 15%) (Wilcoxon rank sum test: p < 0.0002). These results confirm that VDR gene promoter is robustly hypermethylated in breast cancer, compared to normal cells, and support the proposed mechanism of silencing expression of VDR through promoter hypermethylation.
When measuring total VDR transcripts by quantitative RT-PCR, significantly higher levels were observed in primary cancer tissues compared to normal breast tissues. Because of the apparent inconsistency with our in vitro data (Fig. 2B), we compared alternative VDR transcripts in primary breast cancer tissues. It is known that use of secondary VDR promoters and translational start sites as well as alternative splicing can generate a variety of VDR transcripts and proteins.12,13,15–17 Therefore, we examined the electrophoretic patterns generated using PCR primer sets designed to detect several 5′ splice variants of VDR (we generally followed the nomenclature used by Crofts et al.12). Figure 1 illustrates a number of splice variant patterns in relation to the exon structure of VDR. VDR primer set V1 detects three variants that respond to the classic VDR promoter just upstream of exon 1a, and which encode the standard 427 amino acid (aa) VDR peptide, also known as VDRA. Primer set V1d detects three additional variants that use alternative splicing of the 1d exon to encode proteins with N-terminally extended domains, VDRB1 (477 aa) and VDRB2 (450 aa). The V3 and V1d″ splice variants have not been shown to be translated into protein,17 as indicated by the asterisk. We used a downstream primer set (V) spanning exons 3 and 4 to measure total VDR transcript levels. The VT amplicon results corresponded to those obtained with the commercial VDR probe set spanning exons 7 and 8 in quantitative RT-PCR assays.
We performed a series of PCR experiments using the V1 and V1d primer sets specific for Exon 1 and Exon 2 transcript variants. As shown in Figure 5A, the levels and patterns of splice variants of VDR are markedly different in primary cancer compared to normal breast epithelial tissue, with the cancer tissue showing extensive heterogeneity and variability, particularly in the shorter variants, such as V1d″, that are barely detectable in normal tissue. Furthermore, the levels of full-length transcripts (V1, V2, V1d, V1d′) are clearly lower in cancer tissue compared to normal breast tissue.
We sequence-verified the V1, V2, V1d and V1d′ RT-PCR products, and measured them by quantitative imaging of PCR reactions performed at lower amplification cycle numbers (28 cycles), in order to ensure amplification in the linear range of the reactions (shown in Fig. 5B). As shown in the bar graphs in Figure 5C, the results from six primary breast cancer samples and three normal breast organoids indicate that all major full length 5′ splice variants encompassed by these four primers sets are present at lower levels in primary breast cancers when compared to normal breast tissue. Thus these splice variants alone cannot account for the elevation in total VT transcript observed in patient cancer samples compared to normal. A significant fraction of the VDR transcripts found in breast cancer appears to be truncated and generated from more downstream regions.
To investigate if VDR methylation affects expression of the active forms of VDR, we analyzed the VDR variant expression patterns in the same mRNA preparations obtained for the in vitro experiments shown in Figure 1. As shown in Figure 5D, which summarizes the results in the five cell lines, Calcitriol alone had a minimal effect, while levels of V1 and V2 were significantly increased after AZA treatment in all the breast cell lines. V1d and V1d′ variants were less responsive to demethylation. The addition of Calcitriol to AZA showed little additive effect on the variants examined. The results suggest that at least the V1 and V2-specific promoter is partially regulated by DNA methylation, whereas the V1d promoter may be less so.
Since little is currently known about the functional activity of various splice variants of VDR or the functional differences between the active proteins VDRA and VDRB,16 we interrogated the status of VDRE-containing VDR responsive genes in breast cancer tissue by measuring their expression levels. Several cytochrome p450 hydroxylases involved in Vitamin D metabolism are known to contain VDREs, including CYP27B1 (the 25(OH)2-Vitamin D3-activating 1-hydroxylase), CYP24A1 and CYP3A4 (Calcitriol-inactivating 24-OH and 24/25-OH hydroxylases, respectively), which are part of a negative feedback loop, and p21 and C/EBP, which function as tumor suppressors. As summarized in Figure 6, the VDR downstream genes, with the exception of CYP3A4, are expressed at higher levels in normal breast tissue than in cancer tissue (Fig. 6A and B), and can be re-induced in vitro by demethylation (Fig. 6C), although not all cell lines are equally responsive. These data suggest that some alternative transcripts expressed in primary breast cancer tissues may be functionally distinct from normal full-length VDR transcripts, in that they may not fully activate VDR-responsive genes.
Our findings demonstrate that the VDR promoter is hypermethylated in breast cancer, and provide evidence that its demethylation in breast cancer cell lines results in re-expression of the VDR transcripts. Similar epigenetic suppression of VDR mRNA expression has been demonstrated in placental and choriocarcinoma cell lines, which also showed VDR mRNA re-expression after AZA treatment.18 Our initial proliferation assays confirmed the relative Calcitriol insensitivity in breast cancer cells (see Fig. 2A), and an additive antiproliferative response to combined Calcitriol and AZA treatment. Treatment of breast cell lines with AZA resulted in near complete demethylation of the VDR promoter, whereas Calcitriol had almost no effect. Changes in VDR transcript levels mirrored these results (see Fig. 2B). These results were validated in clinical samples, where we show clear hypermethylation of a CpG island in the VDR promoter region in breast cancer tissue by both bisulfite sequencing and a specific MSP assay. Since this CpG island includes Sp1 binding sites, its methylation would be expected to affect Sp1 binding and transcriptional activation,19,20 and although methylation-induced transcriptional silencing most frequently involves CpG sites in the vicinity of the transcriptional start site (TS), there are many examples of critical methylation sites both upstream and downstream of the TS.21 Furthermore, demethylation results in re-expression of the VDR transcripts. Thus, much of the unresponsiveness of the VTD pathway in breast cancer is likely due to epigenetic silencing of VDR transcription, and this may be reversible by pharmacological intervention using demethylating agents.
Numerous conflicting studies have made it difficult to arrive at definitive conclusions regarding the mechanism of Calcitriol insensitivity and the biological impact of this mechanism in cancer, and the data on VDR expression levels in cancer have been particularly inconsistent. Quantitative RT-PCR of VDR has shown that its expression was modestly downregulated in endometrial cancer cells22 and in human colon and lung tumor samples, compared to their normal counterparts, while it was strongly overexpressed in ovarian cancer tissue.8 Reports of VDR expression in breast cancer tissue compared to adjacent normal or independent control tissue have been contradictory,8,23 which may be due to differences in the specific assays used. Our results suggest that a significant proportion of transcripts detected in breast cancer tissue may be N-terminally truncated and may not yield functional peptides, while full-length transcripts are relatively depressed, compared to normal breast epithelial tissue.
Very little is known about the possible roles in breast cancer of the various isoforms of VDR. Initial reports describing several VDR promoters and multiple transcripts with tissue-specific abundance variation12,17 were followed by more detailed biochemical analyses documenting functional differences between the VDRA, VDRB1 and VDRB2 isoforms.15,16,24 There are currently few data, however, on how this regulatory complexity affects breast cancer epidemiology or pathogenesis, let alone what role the potentially untranslated, N-terminally truncated variants we detected in our breast cancer samples might have. A detailed analysis of VDR splice variants, the encoded proteins, and their functional significance is beyond the scope of this report, but our data on VDR expression levels, which strongly depended on the specific amplicons used, suggest that studies failing to take into account the heterogeneity of transcripts in breast cancer tissue could be misleading. In this study, demethylation treatment of breast cancer cell lines with AZA induced the re-expression of active variant transcripts of VDR. These results suggest a correlation between VDR methylation and active transcript variant expression.
Further complicating the role of VDR in breast carcinogenesis, Calcitriol metabolism is controlled by a complex interplay of genetic, nutritional and environmental factors. For example, the dietary intake of Vitamin D only contributes about 10% to Calcitriol synthesis, while the UV-initiated cutaneous conversion of 7-dehydrocholesterol to Vitamin D accounts for about 90%.25 Vitamin D then undergoes a two-step activation process. The initial 25-hydroxylation is performed predominantly in the liver, and the resulting 25OH-VTD is bound to the D-binding transport protein (DBP), constitutes a reservoir, and is the most commonly measured Vitamin D metabolite. 25OH-VTD levels vary considerably depending on access to sunlight and skin pigmentation.26 The production of active Calcitriol by the rate-limiting 1α-hydroxylase CYP27B1 is under tight control in the kidney. It has recently become apparent, however, that CYP27B1 is expressed in a wide range of cell types, including breast epithelial cells, where it is subject to post-transcriptional27 negative feedback inhibition.28–30 Antiproliferative and differentiating effects of CYP27B1 have been detected in many different tissues.31 Similarly, there is a positive feedback loop with CYP24A1, which is induced by ligand-activated VDR. CYP24A1 hydoxylates Calcitriol at the 24-position, the first step in its degradation pathway.32
Our investigation of the expression levels of VDRE-containing, Calcitriol-metabolizing p450 hydroxylases showed that, consistent with a low Calcitriol pathway activity, the rate limiting activating 1α-hydroxylase and the main catabolic 24-hydroxylase mRNAs are underexpressed in breast cancer tissues (Fig. 6). Data from clinical specimens have been contradictory, however, with reports of up or downregulation of the opposing 1- or 24-hydroxylases.8,23,33 Some of these findings could be attributed to expression of splice-variants encoding non-functional peptides, particularly of 1α-hydroxylase (CYP27B1), as reported in several cancers, including breast cancer.34,35 Interestingly, as reported by others,36 we found consistently increased levels of CYP3A4, which has a strong 24-hydroxylase activity and is involved in the metabolism of several steroid hormones and in xenobiotic metabolism.37,38 CYP3A4 polymorphisms have been identified as potential risk factors for predisposition to breast and prostate cancer, and may have pharmacogenetic implications in the tumor response to several chemotherapeutic agents as well.39 In the context of a downregulated Calcitriol pathway in breast cancer, CYP3A4's expression may be less influenced by its VDRE than by triggers related to its role in steroid hormone and xenobiotic metabolism, and its Calcitriol catabolic actions may further depress already low levels of Calcitriol in breast tissue.
Decreased expression of p21 (CDKN1A) has been reported in breast cancer40 and in ovarian cancer.41 p21 contains three VDREs, and is known to be regulated by Calcitriol-induced cyclical chromatin looping.42 In agreement with these reports, we found decreased expression of the p21 tumor suppressor gene in our breast cancer samples, as would be expected in the context of an inactive Calcitriol pathway (see Fig. 6).
Furthermore, we also demonstrated that VDR-responsive genes, including the tumor suppressor C/EBP, could be re-induced in vitro by demethylation treatment, while Calcitriol alone generally had little effect, except in the T47D cell line and on the expression of the CYP3A4 catabolic hydroxylase.
In summary, this report provides correlative evidence for the importance of the Calcitriol/VDR axis in breast cancer, and suggests that potentially reversible epigenetic silencing may be at the center of its inactivation.
The 21PT human breast cancer cell line was derived from a primary tumor and was propagated as described.43 Human breast cell lines were obtained from American Type Culture Collection (Rockville, MD) and propagated as described.44 For pharmacological assays, 1.0 × 106 cells were seeded in 25 cm2 tissue culture flasks. After 24 h, the culture media were changed and cells were treated for 96 h with vehicle (ethanol) alone, or 1.5 µM 1α, 25(OH)2 D3 (Calcitriol), 5 µM 5′deoxy-azacytidine (AZA) (Sigma, St. Louis, MO), or both.
Cell viability was measured using the Cell Titer 96 AQ-One Solution Cell Proliferation Assay kit from Promega Corporation (Madison, WI). Formazan absorbance was read at 490 nm in a 96-well plate reader.
Normal breast tissue organoids were prepared from reduction mammoplasty specimens of women without breast abnormalities as previously described.45 All tissue samples were obtained with the approval of the Johns Hopkins Institutional Review Board.
Primary breast cancer tissues (Invasive ductal carcinomas, pT2-3NxMx) were obtained after surgical resection at the Johns Hopkins Hospital (Baltimore MD), anonymized, and stored frozen at −80°C or fixed in 10% buffered formalin and embedded in paraffin (FFPE). Samples containing >50% tumor cells were processed for molecular studies.
RNA and DNA were purified from cell cultures and tissue samples by organic extraction using the Trizol Reagent (Invitrogen Inc., Carlsbad, CA). Total cellular RNA and DNA were quantified by UV absorption at 260 nm using a Nanodrop spectrophotometer (NanoDrop Technologies Inc., Wilmington, DE).
DNA from cell lines or tissue (1 µg) was treated with sodium bisulfite as previously described.46 Bisulfite-treated DNA was pre-amplified44 using three sets of bisulfite sequencing primers P1, P2 and P3 (see Table 1), encompassing the region from 790 bp upstream to 380 bp downstream of the VDR transcription start site. Primers were 5′-tagged with 21 bp of the M13 universal primer to facilitate subsequent sequencing reactions. PCR products were separated electrophoretically and isolated using a PCR purification kit (Qiagen, Valencia, CA). DNA was sequenced using the M13 Reverse Primer with an Applied Biosystems automated fluorescent sequencer. Percent DNA methylation was determined using the ratio of cytidine to thymidine traces at CpG dinucleotides,44 defined as the average of three independent sequencing runs.
A nested PCR approach was used to assess VDR methylation in tissue samples. Bisulfite-treated DNA was pre-amplified with the P1 sequencing primers and the products isolated as above. QMSP was then performed in triplicates using methylated (M) and unmethylated (UN) DNA-specific primers (see Table 1) with the Qiagen SYBR green PCR Kit (Qiagen, Valencia, CA) in an ABI-PRISM 7900HT instrument. Percent methylation was calculated as the ratio of inverse Ct values of methylated DNA/methylated plus unmethylated DNA (M/[M + U]*100) as previously described.47
Reverse transcription reactions were performed using random hexamer primers as previously described.44 PCR was then performed for 40 cycles using gene-specific primers. To determine total VDR levels (VT), pre-mixed “Assay on Demand” primer-probe sets for VDR (Hs01045843_m1) and GAPDH (Hs99999905_m1) and the GeneAmp® Fast PCR Master Mix were purchased from Applied Biosystems (Foster City, CA) and cDNA was amplified following the manufacturer's instructions in the ABI-PRISM 7900HT instrument. This VDR primer probe set amplifies a region spanning Exons 7 and 8 and does not distinguish between the VDR splice variants that were investigated in this report. The same approach was used for our VF & VR primer set, which spans exons 3 and 4, and similarly detects all 5′ splice variants.
The V1 primers, spanning exon 1a and exon 2, were designed to amplify variant V1 and V2 transcripts (see Fig. 1). The V1d primers, spanning exon 1d and exon 2, were designed to amplify the transcript variants v1d, V1d′ and v1d″. In order to detect all variants simultaneously, the nonspecific V primers, spanning exons 3 and 4, were used.
Predetermined linear ranges of the amplification reactions, i.e., 25 to 28 cycles for VDR and its splice variants, and 30 to 35 cycles for the Calcitriol hydroxylases p21 and C/EBP, were chosen for semiquantitative analyses of transcript levels. The PCR products were separated on 2% agarose gels, scanned in a Gel Doc Imager (BioRad, Philadelphia, PA), and quantified using Image Quant software (BioRad, Philadelphia, PA).
PCR products identified after RT-PCR using the V1 and V1d primers were excised from the agarose gel, isolated using the QIAEX II gel extraction kit (Qiagen, Valencia, CA), and cloned using the TOPO TA cloning Kit (Invitrogen Inc., Carlsbad, CA). Plasmid inserts were verified by EcoR1 digestion and sequenced using the M13 reverse primer.
In vitro experiments were performed in triplicate. For each experiment, data were expressed as means ± standard deviation except where stated otherwise. Significance was determined using the two-tailed Student t-test or the Wilcoxon rank sum test.
Financial Support: Susan G. Komen Foundation and Department of Defense DAMD17-03-1-0547 (C.B.U.), NCI P50 CA088843-06A1 (S.S., C.B.U., V.S.), Breast Cancer Research Foundation (V.S.).
Previously published online: www.landesbioscience.com/journals/cbt/article/11994