Background
Epidemiological data indicates higher incidence and mortality rates from breast cancer in low latitude regions. Among the mechanisms suggested for a relationship between sunlight and cancer is the genesis of vitamin D in the skin, resulting from the UV light action. In accordance with this hypothesis, there is evidence that lower 25(OH)D
3[
1‐
5] and 1,25(OH)
2D
3[
6,
7] serum concentrations are encountered in patients with breast cancer, as compared with women without cancer, as well as in patients with advanced or metastatic disease in comparison with those with early-stage disease [
8,
9]. In addition, 25(OH)D
3 deficiency at diagnosis was related with poor prognosis, evaluated as metastasis-free and overall survival [
10].
In human breast xenografts established in immunossupressed mice 1,25(OH)
2D
3 exerts growth inhibitory effects, and in mouse mammary organ culture exposed to chemical carcinogens, both 25(OH)D
3 and 1,25(OH)
2D
3 mediate preventive effects [
11‐
13]. However, the chemopreventive effect of vitamin D is still controversial, as supplementation trials on vitamin D
3 and colon or breast cancer incidence have been inconsistent [
14,
15]. One critical issue is that the appropriate supplementation dose for cancer prevention trials was not well established [
16]. On the other hand, clinical studies point to a clinical benefit for 1,25(OH)
2D
3 (or analogues) alone or in combination with chemotherapy in the treatment of hormone refractory prostate cancer and breast cancer skin lesions [
17,
18]. However, concerns about hypercalcemic side effects limit the dose of 1,25(OH)
2D
3 (or analogues) that can be safely administered
in vivo.
Phase I clinical studies indicate that subcutaneous doses of calcitriol given every other day result in peak 1,25(OH)
2D
3 serum concentration of 0.25-0.75 nM [
19] while weekly pulses of oral calcitriol allow higher dose administration and peak serum concentrations of 1–15 nM [
20]. Although these vitamin D concentrations represent about 1.3-83 times the upper limit of physiologic serum levels, they are well below the concentrations (10-100nM) typically used to investigate hormone actions in cell culture studies. At these concentrations, 1,25(OH)
2D
3 exerts antiproliferative and pro apoptotic effects [
21] and modulates angiogenesis [
22,
23], invasion and metastasis [
24,
25]. Among the downstream targets of the hormone are cyclin dependent kinase inhibitors as p21
WAF1/CIP1 and p27
KIP1; growth factors, receptors and associated proteins as TGFβ, TGFβ receptors and insulin-like growth factor binding protein-3 (IGFBP-3) [
26‐
31]. In addition, gene expression profiling of breast cancer cell lines MCF7 and MDA-MB-231 have identified many potential 1,25(OH)
2D
3 target genes, [
24] but again, these studies were conducted with supra physiological concentrations of calcitriol (50-100nM). Furthermore, experiments in cell lines do not reflect the complex array of interactions among malignant and stromal cells, secreted factors and extracellular matrix proteins taking place in the tumor microenvironment, which also modulate the hormone actions.
Although the majority of human breast cancers express vitamin D receptors (VDR) [
7,
32,
33], there have been no demonstrations that 1,25(OH)
2D
3 modulates gene expression in human breast cancer samples. To address this research gap, a physiologically relevant
in vitro model to study 1,25(OH)
2D
3 actions, represented by short term culture of fresh breast cancer tissue slices, which maintain the epithelial mesenchymal relationship and preserve tissue morphology and proliferation rate, was established [
25,
34,
35]. With this organotypic culture system the transcriptional effects of 1,25(OH)
2D
3 at 0.5nM, a concentration that can be safely attained
in vivo, and 100nM, the concentration typically used in cell culture studies, was compared. In addition, mammary cell lines and fibroblasts obtained from breast cancer samples were used to validate transcriptional targets of 1,25(OH)
2D
3 in epithelial and stromal cell types. Cancer associated fibroblasts (CAF) are interactive cells that infiltrate tumor specimens, influencing their behavior [
36‐
38], which are also potential targets of the hormone. Although VDRs have been detected in fibroblasts obtained from prostate and breast tumors, few studies have compared 1,25(OH)
2D
3 mediated genomic effects in epithelial and stromal cells [
39,
40]. The present study indicates that physiologically relevant concentrations of 1,25(OH)
2D
3 may influence gene expression in breast tumor slices cultured
ex vivo, and that regulation of target genes likely occurs in both epithelial and stromal compartments of the tumor.
Methods
Patients
Post-menopausal breast cancer patients clinical stages I-III were invited to take part in the study. This protocol was carried out in compliance with the Helsinki Declaration and was approved by the Institutional Ethics Committee (Comitê de Ética do Instituto Brasileiro de Controle do Câncer, protocol number 108/2006/7; Comitê de Ética em Pesquisa do Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo, protocol number 626/06; Comitê de Ética do Hospital do Câncer A. C. Camargo, protocol number 1131/08). A written informed consent was signed by all participants. Twenty one patients were prospectively accrued at Instituto Brasileiro de Controle do Câncer and Hospital do Câncer A. C. Camargo, São Paulo, from August 2007 to September 2009. Characteristics of these patients are described on Table
1.
Table 1
Characteristics of patients
Median age | 70 (56–76) | 56.5 (49–72) | 0.068 |
CS III | 3 (60%) | 5 (33%) | 0.347 |
N(+) | 3 (60%) | 9 (60%) | 1.000 |
IDC | 5 (100%) | 10 (62%) | 0.262 |
ER(+) | 2 (40%) | 11 (69%) | 0.325 |
PR(+) | 3 (60%) | 9 (56%) | 1.000 |
HER2(+) | 2 (40%) | 5 (31%) | 1.000 |
Tissue slice preparation and treatment
Tumor fragments were obtained immediately after tumor resection by the pathologist, who selected an involved area for this study. Fragments were placed into culture medium (RPMI 1640 with antibiotics and fungicide) and tissue slices were prepared using the Krumdieck tissue slicing system (Alabama Research and Development Corporation, Birmingham, AL, USA). Fragment thickness varied between 400–800 μm. Slices were cultured for 24 hours in 6-well plates (1 slice/well; 1–3 slices per treatment) containing 2 mL of culture media, RPMI supplemented with 10% v/v FBS, antibiotics and 0.001% ethanol (vehicle) or 1,25(OH)2D3 (Calbiochem, Darmstadt, Germany) 0.5nM or 100nM (from now on called physiological and supra-physiological concentrations, respectively). One slice of each sample was processed by FFPE and hematoxilin-eosin stained slides revealed that tumor samples contained > 50% malignant cells.
Fibroblasts primary culture
Primary fibroblast culture was established from tumor samples obtained from another five post-menopausal patients, diagnosed with invasive ductal carcinoma (histological grades II or III, three of them hormone receptor positive). Tumor samples were cut into small pieces and fibroblast primary culture was established through the explant methodology. After three cell passages, mesenchymal origin of the cells was confirmed by their spindle cell morphology and positive expression of vimentin [mouse anti human vimentin monoclonal antibody, clone Vim 3b4 (1:200); DAKO Corporation, Carpinteria, CA, USA] and alpha smooth muscle actin [(mouse monoclonal antibody anti human alpha smooth muscle actin, clone 1A4 (1:50); R&D Systems] and negative expression of cytokeratin [mouse monoclonal antibody anti human cytokeratin clone AE1/AE3 (1:100); DAKO] by immunocytochemistry (data not shown). Fibroblasts were then exposed to 1,25(OH)2D3 (Calbiochem, Darmstadt, Germany) 0.5nM or vehicle for 24 hours and after RNA extraction, RT-qPCR was performed to evaluate expression of candidate genes.
Culture of mammary epithelial cell lines
HB4A (normal mammary epithelial cell line) and C5.2a (HB4A transfected with HER2), both donated by Drs. Mike O’Hare and Alan Mackay, Ludwig Institute for Cancer Research, London, UK; SKBR3: breast cancer cell line overexpressing HER2; MDA MB-231: breast cancer cell line triple negative; and MCF-7: breast cancer cell line ER(+), acquired from American Type Culture Colection (Manassas, Virginia, USA), were cultured in RPMI-1640 supplemented with 10% fetal calf serum (FCS). After 24 hours, medium was replaced and 1,25(OH)2D3 0.5 nM (treated cells) or ethanol (control cells) was added. After 24 hs of treatment, total RNA was isolated using Trizol reagent and used in RT-qPCR.
RNA extraction and microarray hybridization
Tumor specimens were pulverized (Bio-Pulverizer™ BioSpec Products Inc., Oklahoma, USA) under liquid nitrogen and total RNA was isolated using RNeasy kit (Qiagen, Valencia, CA, USA), according to the manufacturer’s protocol. RNA integrity was verified in a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA) and samples with RNA integrity number ≥ 6.6 were analyzed. Beginning with 100 ng total RNA, a two-round linear amplification was carried out, according to Affymetrix protocol (Two Cycle Target Labeling Kit, Affymetrix, Santa Clara, CA, USA). Afterwards, biotin-labeled cRNA was synthesized from double strand cDNA, using IVT labeling kit (Affymetrix) and 20 μg of biotinylated fragmented aRNA was hybridized onto Human Genome U133 Plus 2.0 GeneChip (Affymetrix
Hybridized arrays were scanned using Affymetrix GeneChip Scanner 3000 and after visual inspection, images were subjected to Affymetrix GeneChip Operating Software (GCOS) analysis to generate report files for quality control. Data normalization was performed using the Robust Multi-Array Average (RMA). Samples were categorized according to treatment in three groups: 1,25(OH)
2D
3 0.5nM, 1,25(OH)
2D
3 100nM and control. To establish a differential gene expression profile between vitamin D treated and untreated samples, SAM two class paired, provided on MEV (MultiExperiment Viewer – Boston, MA, USA) was used, after selecting 50% of the genes with the highest standard deviation. False discovery ratio (FDR) ≤0.10 was considered significant. In addition, results obtained with FDR≤0.01 are presented. Unsupervised hierarchical clustering based on Euclidean distance and average linkage was used to verify association patterns. The reliability of the clustering was assessed by the Bootstrap technique. Raw data complying with MIAME format was deposited at the Gene Expression Omnibus (GEO) data repository (
http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE27220) accession number GSE27220.
To explore functional enrichment associated with calcitriol treatment based on Ontologies (GO, Pathway), Regulome (TFBS, transcription factor binding site) Pharmacome (Drug-gene associations) among other features, differentially expressed genes were subject to subsequent analysis using ToppFun, available on ToppGene Suite (
http://toppgene.cchmc.org/enrichment.jsp) and were considered significant if
P < 0.05 [
41].
Gene set enrichment analysis (GSEA) method was used to identify whether predefined gene sets might associate with gene expression differences between phenotypes. In this pairwise comparison, all genes are ranked based on signal-to-noise ratio and the alternative hypothesis that rank ordering of distinct pathway members is associated with a specific phenotype is tested [
42]. This methodology makes it possible to detect situations where all genes in a predefined set change in a small but coordinated way. FDR<0.10 was considered significant.
Real time RT-PCR
Reverse transcription was performed with random primers and Superscript III (Invitrogen Corporation, Carlsbad, CA, USA). Quantitative PCR (qPCR) was carried out using specific primers (Additional file
1: Table S1) and SYBR-green I (Sigma, St. Louis, MO, USA) in a Rotor-gene system (Corbett Research, Mortlake, Australia). Relative expression of target genes was calculated as 2
-ΔΔCT, using
GAPDH or
ACTB as internal control (as indicated) and the average value of the target gene in control samples, as reference level.
Western blot
Protein lysates from cell lines were made using RIPA buffer (1% NP-40, 0.1% SDS, 0.5% Sodium Deoxycholate in 1 × PBS) supplemented with complete mini protease inhibitor cocktail tablets (Roche; cat 04693124001). Afterwards, 50 μg of protein was subjected to SDS-PAGE and transferred to Hybond ECL membrane (GE Lifesciences), which was probed with the following primary antibodies overnight at 4°C: CD26 (DPP4, clone H-270, rabbit polyclonal antibody, 1:500; Santa Cruz Biotechonology Inc. Santa Cruz, CA, USA); CD14 (clone M305, sc9150, rabbit polyclonal antibody 1:500; Santa Cruz); β-actin (monoclonal Anti-β-Actin antibody produced in mouse, clone AC-15, ascites fluid, A5441, 1:2000, Sigma-Aldrich) and then with appropriate secondary antibodies (170–6515 Goat Anti-Rabbit Ig-G (H+L) HRP conjugate; 170–6516 Goat Anti-Mouse Ig-G (H+L) HRP conjugate; Bio-Rad.). Protein expression was detected with ECL Plus Western Blotting Detection Reagents (GE Lifesciences) in a ImageQuant LAS 4000 (GE Healthcare).
Immunocytochemistry
Fibroblasts were grown on coverslips in the absence or presence of 1,25(OH)2D3 0.5nM for 24 hours. Samples were fixed in 4% paraformaldehyde and permeabilized with 0.5% Triton X-100/PBS, in case of intracellular targets. Blocking of unspecific binding was performed with 2% BSA/PBS. Afterwards, cells were incubated with the primary antibody (CD26, clone H-270 rabbit polyclonal antibody, anti-DPP4, 1:200, Santa Cruz; CA II, clone G-2, sc-48351 mouse monoclonal antibody, 1:100, Santa Cruz) overnight in humid chamber at 4°C and then with the secondary antibody conjugated with Alexa Fluor 488 (1:700, species specific: goat anti-rabbit IgG, n° A11008; goat anti-mouse IgG n° A11001; Molecular Probes) for 1 h at room temperature in the dark. DAPI was added for nuclear staining. Images were acquired in a Olympus fluorescence microscope DX-5AI, using an Image Pro-PLUS 6,0 software.
Immunohystochemistry
Breast cancer slices from seven patients (six samples cultured in the absence (control) or presence of 1,25(OH)
2D
3 100nM and one sample cultured in the presence of 1,25(OH)
2D
3 0.5nM) were available for analysis (from the patients described in Table
1). Sections of 3 μm thickness were cut from paraffin blocks and antigen retrieval was carried out in 10 mM citrate buffer at pH 6.0 in humid heat under pressure cooker. Staining with the following specific antibody took place overnight at 4°C: CD14, clone M-305, (sc-9150Santa Cruz Biotechnology) rabbit polyclonal IgG, 1:800. Reaction was revealed with Novolink Polymer Detection Systems (Leica Biosystems, Newcastle, UK, cat: RE7280-k), followed by analysis in a Olympus fluorescence microscope DX-5AI (40x objective) and acquisition with an Image Pro-PLUS 6,0 software.
Detection of soluble CD14 (sCD14) in culture medium of tumor samples
Tumor slices from another four post-menopausal patients (median age 56 years) diagnosed with invasive ductal carcinoma clinical stages I-II, HER2 negative and hormone receptor positive (except for one tumor triple negative) were cultured with or without 1,25(OH)2D3 0.5nM or 100nM for 24 hours and 100 μL of the conditioned medium was used for soluble CD14 (sCD14) quantitative determination, through an enzyme-linked immunosorbent assay (Quantikine ELISA Human sCD14 Immunoassay, R&D Systems, Minneapolis, MN, USA). For every sample, two analyses on the same plate were carried out and the mean value was used.
Statistics
Kolmogorov-Smirnov test was applied to check for normality of the data, followed by parametric or non-parametric tests, as appropriate. To detect an association between variables, Pearson chi-square or Fisher exact tests were used. A two-tailed p value ≤ 0.05 was considered significant. Analysis was undertaken using Instat (GraphPad Software, Inc., La Jolla, CA, USA) or SPSS (Chicago, IL, USA).
Discussion
The primary goal of this work was to evaluate the transcriptional responses of breast cancer samples to physiologically relevant concentrations of 1,25(OH)2D3, using a culture model that retains features of intact tumors, such as stromal-epithelial interactions. Microarray analysis identified nine genes that were significantly altered within 24 h of exposure to 1,25(OH)2D3 0.5nM, a concentration that is physiologically achievable in patients. Of these, the vitamin D target gene CYP24A1(which codes a cytochrome P450 enzyme, that hydroxylates 25(OH)D3 and 1,25(OH)2D3 to less active forms 24,25(OH)2D3 and 1,24,25(OH)3D3) was induced over 7-fold in microarray analysis and was validated in another set of tumor samples, clearly indicating activation of VDR signaling. Additional evidence for activation of the VDR pathway in this dataset was obtained by GSEA, which indicated a trend towards the enrichment of genes sharing DR3 binding sites, a consensus motif for VDR.
Comparison of microarray data from tumor slices cultured with 0.5nM vs. 100nM 1,25(OH)
2D
3 indicated a clear concentration effect, as the number of differentially expressed transcripts increased from nine at 0.5nM to 186 at 100nM (20 fold increment). Induction of CYP24A1 increased from 7-fold (at 0.5 nM) to 70-fold (at 100nM) - a 10 fold enhancement. In both datasets, the majority of genes (approximately 75%) were up-regulated rather than down-regulated by 1,25(OH)
2D
3, consistent with other array data from established cell lines cultured with high dose 1,25(OH)
2D
3
in vitro[
43‐
45].
In addition to
CYP24A1, five other genes were commonly up-regulated in tumor slices exposed to both low and high concentrations of 1,25(OH)
2D
3:
DPP4, KCKN3, EFTUD1, TKTL1 and
CA2. All, except
TKTL1 (transketolase-like 1) have been previously identified as VDR target genes in various model systems.
DPP4 (dipeptidyl-peptidase 4, also called CD26) was up-regulated in artery smooth muscle cells exposed to 1,25(OH)
2D
3[
46] and its overexpression in distinct cell types (melanocytes, non-small cell lung, prostate and neuroblastoma cells) triggered anti-tumorigenic effects including cell growth arrest, inhibition of cell migration and increased apoptosis [
47].
KCNK3 (potassium channel, subfamily K, member) was induced by 1,25(OH)
2D
3 in artery smooth muscle cells, and
EFTUD1 (elongation factor Tu GTP binding domain containing 1) in oral squamous carcinoma, breast cancer associated fibroblasts, immortalized prostate cells and lymphoblastoid cell lines [
40,
43‐
46].
CA2 (carbonic anhydrase II) mRNA appeared to be directly induced by 1,25(OH)
2D
3 in myelomonocytic cell lines but indirectly regulated in osteoclast progenitors, where the physical communication with stromal cells seems to be required [
48,
49].
CYP26B1 (cytochrome P450, family 26, subfamily b, polypeptide 1) which was up-regulated in samples treated with 1,25(OH)
2D
3 0.5nM, was previously identified as a vitamin D induced gene in immortalized non-transformed prostate epithelial and oral squamous carcinoma cell lines, and
in silico analysis has tentatively identified a VDR binding site at this genomic region [
43,
44].
Other authors have analyzed physiological concentration effects of vitamin D using animal models. Vitamin D supplemented diet as well as calcitriol injections were shown to stimulate the VDR pathway, mildly increasing CYP24A1 expression (x2) in MCF-7 xenografts in immunocompromised mice [
50]. Interestingly, vitamin D transcriptional effects may not overlap in tumor specimens and non-transformed mammary glands in the MMTV-neu transgenic mouse model of breast cancer, fed a high vitamin D diet [
51]. Comparison between cancer and normal cells is an interesting issue, as vitamin D potential effects in cancer prevention have also been claimed. In accordance with the previous work [
51], differences in transcriptional targets were also described for breast cancer associated fibroblasts (CAF) and normal adjacent fibroblasts (NAF) exposed to 1,25(OH)
2D
3 in a supra-physiological concentration. Among up-regulated genes 45.7% were commonly modulated in CAFs and NAFs, however, 36.4% were exclusively up-regulated in NAFs and 17.4% exclusively up-regulated in CAFs [
40]. In addition, looking at overlapping genes in the Venn diagram of vitamin D up-regulated transcripts in six works [
40,
43‐
46], only seven intersections were found in non-cancer cells:
AKR1B1, CRIP1, FZD8, MREG (in immortalized prostate cells and NAF),
BCAT1, GCLC (in coronary artery smooth muscle cells and NAFs) and
PRR6 (in immortalized prostate cells and coronary artery smooth muscle cells). Furthermore, it was reported that vitamin D response is blunted in transformed HME normal mammary cells as compared with parental normal cells [
52]. The last works evaluating vitamin D effects in normal cells however, were performed using supra-physiological concentrations of 1,25(OH)
2D
3 (10-100nM) or analogs and the role of physiological concentrations of the hormone in normal cells is not fully established.
At 100nM, 1,25(OH)
2D
3 exerted more extensive transcriptional effects, and at least 40 of the induced genes in breast cancer organotypic culture have already been reported as up regulated by the hormone, such as
ALCAM, ARRDC4, BMP2, BMP6, CA2, CD14, CLIC6, CILP, CLMN, CYP19A1, DCLDB1, EFTUD1, EHBP1, FAM20C, FOXF1, FRAS1, GOS2, GRK5, HBGEF, HSMPP8, IL1RL1, KCNK3, KIAA0500, PKD2, RGNEF, SEMA6D, SERPINB1, SLC1A1, THBD, TIMP1, TRIM56[
40,
43‐
46]. However, co-aggregation of paired samples (treated and untreated) upon cluster analysis suggests that an individual dominant transcriptional profile was maintained, regardless of treatment. These results were not unexpected, as a high degree of transcriptional similarity was also demonstrated for matched pre and post-neoadjuvant chemotherapy, even though the chemotherapy exerts a more pronounced acute cellular effect than hormonal treatments [
53‐
55].
Some of the genes induced by 100nM 1,25(OH)
2D
3 concentration are involved in TGF beta signaling pathway, in accordance with other authors [
56,
57]. Other genes are involved in regulation of leukocyte mediated immunity and positive regulation of alpha-beta T cell activation, including
CD14, which encodes a receptor to bacterial lipopolysaccharide, as previously reported in a variety of cells as mononuclear phagocytes, normal human epidermal keratinocytes, oral squamous carcinoma, immortalized non-transformed prostate epithelial cell lines and malignant breast cells [
43,
56,
58].
The present tumor slice model represents a heterogeneous combination of epithelial and stromal cells, in which the complex array of reciprocal interactions taking place in the tumor microenvironment, including cell-cell contacts and a variety of secreted factors, might modulate the overall response to 1,25(OH)2D3. Hence, after evaluating the hormone effects in tumor slices, the effects of 1,25(OH)2D3 0.5nM in defined populations of cancer associated fibroblasts and epithelial cells were compared. This data indicated that even though CYP24A1 was induced in both fibroblasts and epithelial cells, CD14, CA2, and IL1RL1 were primarily induced in epithelial cells. There was also a trend towards up-regulation of CA2, DPP4 and IL1RL1 in cancer associated fibroblasts.
One major strengthen of this work was the comparison of achievable versus supra-physiological concentrations of 1,25(OH)2D3 in breast cancer slices, a model that preserves the epithelial-mesenchimal interactionss, indicating that effects are much less intense in near physiological concentrations. However, a weakness of this work was the small number of samples used in microarray experiments. These effects however, were later confirmed in a larger number of tumor samples and cell lines, using RT-PCR, even though they were more difficult to detect at the protein level, in face of the discrete changes induced by 0.5nM 1,25(OH)2D3.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CM designed the study, participated in data acquisition, participated in data analysis and interpretation, drafted and revised the manuscript; MLHK designed the study, participated in data analysis and interpretation, drafted and revised the manuscript; ECL, designed the study and participated in data acquisition; JW participated in data analysis and interpretation, drafted and revised the manuscript; LTC participated in data acquisition; MMB participated in data analysis and interpretation, drafted and revised the manuscript; MSM was responsible for the study in Hosp. A. C. Camargo and participated in data acquisition; RAR participated in data acquisition; PRV participated in data acquisition; JCGSG was responsible for the study in IBCC and participated in data acquisition; SN participated in data acquisition; RET participated in data acquisition; MAAKF designed the study, participated in data analysis and interpretation, drafted and revised the manuscript. All authors read and approved the final manuscript.