Background
Chronic inflammation is strongly associated with the development of cancer [
1]-–[
3]. One of the crucial mediators of inflammatory reaction is cyclooxygenase (COX). The COX family of enzymes comprises two members (COX-1 and COX-2) and is the main controller of eicosanoid biosynthesis. Studies of human breast tumor tissues demonstrate that upregulation of COX-2 has been detected in approximately 40% of human breast tumor tissues, as well as preinvasive ductal carcinoma
in situ lesions [
4]. Elevated expression of COX-2 is associated with large tumor size, advanced histologic grade, axillary node metastasis, and unfavorable disease-free survival [
4],[
5]. In addition, COX-2 expression also links with increased tumor angiogenesis [
6]. Epidemiologic investigations suggest that use of nonsteroidal antiinflammatory drugs or selective COX-2 inhibitors reduces breast cancer risk [
7],[
8].
Results of animal study also support an oncogenic role of COX-2. Transgenic COX-2 overexpression induces mammary tumor formation in mice [
9]. This tumorigenic transformation is highly dependent on PGE
2 production and angiogenic switch. In addition,
HER-2/Neu oncogene-induced mammary tumorigenesis and angiogenesis are dramatically attenuated in COX-2 knockout mice, suggesting a key role of COX-2 in breast cancer [
10]. Recent studies also show that COX-2 inhibitors exhibit antitumor and antiangiogenic activities
in vivo and exhibit chemopreventive effects against mammary carcinogenesis induced by 7,12-dimethyl-benz(a)anthracene in rats [
11]. All of the results suggest that COX-2 is involved in multiple steps of breast carcinogenesis and is a potential target for cancer prevention and therapy.
Interplay between breast cancer cells and cancer-associated fibroblasts (CAFs), the most abundant and active stromal cells, is crucial for tumor growth, progression, angiogenesis, and therapeutic resistance [
12]. Cancer cells release a number of factors to enhance the production of cytokines, chemokines, and matrix metalloproteinases (MMPs) from CAFs, which in turn facilitate cancer cell proliferation, migration, and invasion. Previous study demonstrated that stromal fibroblasts present in invasive breast carcinomas can secrete large amounts of stromal cell-derived factor 1 (SDF-1) to enhance tumor growth and angiogenesis [
13]. Co-injection of breast cancer cells and fibroblasts also promotes the progression of ductal carcinoma
in situ to invasive breast carcinoma by stimulating chemokine (C-X-C motif) ligand 14 (CXCL14) and chemokine (C-X-C motif) ligand 12 (CXCL12) production [
14]. However, most studies addressing the crosstalk between cancer and stromal cells focus on protein factors like cytokines and chemokines. Whether other small molecules such as lipids or metabolites participate in cancer-stromal cell interaction is largely unknown.
The tumor-promoting role of CAFs via upregulation of COX-2 in ductal carcinoma
in situ of the breast was first demonstrated by Hu
et al. [
15]. The authors showed that co-culture with fibroblasts increases COX-2 expression in breast cancer cells and subsequently induces MMP-9 and MMP-14 in these cells to promote invasion. They also elucidated the underlying mechanism by demonstrating that inhibition of nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and COX-2 activity reduces the invasion-promoting effect of fibroblasts. These data suggest that fibroblasts secrete some factors to activate NF-κB-mediated transcription of COX-2 in breast cancer cells to enhance tumor progression.
However, several issues remain elusive. First, does PGE2 generated by COX-2-expressing cancer cells also affect gene expression or behavior of stromal fibroblasts? Second, do CAFs secrete small molecules (other than proteins or peptides) to regulate cancer cell invasion? Finally, can the importance of cancer-stroma interaction in cancer progression be validated in clinical samples?
In this study, we address these questions and try to clarify the underlying mechanism.
Methods
Cell culture
Human breast cancer cell lines MCF-7 and MDA-MB-231 were purchased from the Bioresource Collection and Research Center (BCRC) and ATCC. Immortalized human breast fibroblasts, RMF-EG [
16], were kindly provided by Dr. Charlotte Kuperwasser (Tufts University, Boston, MA, USA). These cells were cultured in DMEM/F12 containing 10% fetal bovine serum (FBS). Other experimental materials and procedures are provided in Additional file
1.
Establishment of inducible COX-2-expression MCF-7 cell line
To establish an inducible COX-2-expression cell line, MCF-7 cells (1 × 106) were resuspended in buffer R containing 2 μg pCMV-Tet3G plasmid. Transfection was performed by using Neon microporation transfection system at room temperature with 1,250 V, 20 milliseconds, and two pulses. After 48 hours, the cells were selected with 1 mg/ml G418 for 2 weeks.
For the delivery of the second plasmid, pCMV-Tet3G stably transfected cells (1 × 106) were resuspended in buffer R containing 2 μg of pTRE-mCherry-COX-2 plasmid. Transfection was performed by using Neon microporation transfection system at room temperature with 1,250 V, 20 milliseconds, and two pulses. After 48 hours, the cells were subjected to selection with 100 μg/ml hygromycin B. The stable cell line harbors both pCMV-Tet3G and pTRE-mCherry-COX-2 plasmid was used for induction experiment. The cells were maintained at 37°C in a 5% CO2-humidified atmosphere and were incubated with doxycycline to induce COX-2 expression before co-culture assay.
Co-culture assay
In the co-culture system, 1 × 105 RMF-EG cells were grown in the bottom of a six-well plate in DMEM/F12 with 10% FBS, and 1 × 106 breast cancer cells were seeded on the 0.4-μm polyester membrane of a transwell insert in the same medium. MCF-7 cells were treated with or without doxycycline (1 μg/ml) for 72 hours. The conditioned medium, breast cancer cells, and RMF-EG cells were harvested for metabolomics and Western blotting analysis.
The proteins in the conditioned medium were removed by using 3-kDa ultracentrifugation filter devices. The metabolites in the filtered medium were extracted by using iced 50% methanol and were subsequently dried by a speedvac. Metabolite profiles were analyzed with the Metabolomics Core of National Health Research Institutes by using a high-resolution ultraperformance liquid chromatography (UPLC) coupled online to a triple-quadrupole time-of-flight mass spectrometry system, as described previously [
17]. Metabolite identity was predicted with Human Metabolome Database [
18].
RNA extraction and quantitative reverse transcription-PCR analysis
Total RNA was isolated from cells by using an RNA extraction kit (Qiagen, Valencia, CA, USA) and 1 μg of RNA was reverse-transcribed to cDNA. Target mRNAs were quantified by using real-time PCR reactions with SYBR green fluorescein, and actin served as an internal control. cDNA synthesis was performed at 95°C for 3 minutes, and the conditions for PCR were 28 cycles of denaturation (95°C/1 minute), annealing (60°C/1 minute) extension (72°C/1 minute), and 1 cycle of final extension (72°C/10 minutes). The primers used are tryptophan 2,3-dioxygenase (TDO)-forward: 5′-GGGAACTACCTGCATTTGGA-3′; TDO-reverse: 5′-GTGCATCCGAGAAACAACCT-3′; IDO-forward: 5′-GCGCTGTTGGAAATAGCTTC-3′; IDO-reverse: 5′-CAGGACGTCAAAGCACTGAA-3′; E-cadherin-forward: 5′-CCTGGGACTCCACCTACAGA-3′; E-cadherin-reverse: 5′-GGATGAACACAGCGTGAGAGA-3′; actin-forward: 5′-TGTTACCAACTGGGACGACA-3′; actin-reverse: 5′-GGGGTGTTGAAGGTCTCAAA-3′.
Immunoprecipitation and Western blot analysis
MCF-7or COX-2-overexpressing MCF-7 cells were treated with or without 100 μM kynurenine for 24 hours; the cells were harvested with an RIPA buffer (50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, 2 mM EDTA, and 50 mM NaF), and cellular lysates were incubated with anti-AhR antibody overnight at 4°C with rotation. Immunocomplexes were pulled down by Protein-G agarose bead, washed with RIPA buffer 3 times, and eluted with a sample buffer in boiled water for 10 minutes. The eluted samples were subjected to SDS-PAGE separation, and proteins were transferred to nitrocellulose membranes. Finally, the blots were probed with anti-E-cadherin or anti-Skp2 antibody and developed with enhanced chemiluminescence reagent.
Migration assay
Migration assays were conducted in transwells with 8-μm-pore filter inserts. Then 1 × 104 MCF-7 or COX-2-overexpressing MCF-7 cells were seeded in the upper chamber. The lower chambers were filled with DMEM medium containing 1% FBS and 100 μM kynurenine. After 24 hours, the cells on the upper surface were removed by wiping with a cotton swab, and the cells that migrated to the lower surface were fixed. The cells were stained with 4′,6-diamidino-2-phenylindole (DAPI), and the cell number in 15 randomly selected fields was counted under a microscope (100×). Experiments were performed independently at least 3 times.
Protein ubiquitination assay
MCF-7 cells treated with or without kynurenine were incubated with the proteasome inhibitor MG132 or the lysosome inhibitor chloroquine. The cells were harvested with a lysis buffer (20 mM Tris–HCl at pH 7.5, 150 mM sodium chloride, 1 mM calcium chloride, and 1% Triton X-100 and protease inhibitors), and cellular lysates were incubated with an E-cadherin antibody overnight at 4°C with rotation. Protein-G beads were added to the samples and incubated for another 1 hour at 4°C. Immunocomplexes were eluted and were subjected to SDS-PAGE separation, and proteins were transferred to nitrocellulose membranes. Finally, the blots were probed by using an anti-ubiquitin antibody to detect the ubiquitination status of E-cadherin.
Immunofluorescent staining and confocal microscopy
MCF-7 cells were treated with or without 100 μM kynurenine for 6 hours and fixed with 3.7% formaldehyde for 15 minutes at room temperature. Cells were washed twice with PBS and permeabilized by 0.1% Triton X-100 solution for 10 minutes. After permeabilization, cells were incubated with 0.05% BSA solution to block nonspecific binding. Anti-AhR mouse monoclonal antibody (1:80) or anti-E-cadherin goat polyclonal antibody (1:250) was added and incubated at room temperature for 1 hour. After extensive washing, Alexas Fluro 594 anti-mouse IgG or Alexas Fluro 488 anti-goat IgG was added and incubated for another 1 hour. Cell nuclei were stained with DAPI solution. Finally, coverslips were washed twice with PBS and subsequently placed in mounting solution. The fluorescent image was observed under a confocal microscope.
In vivoorthotopic animal study
MCF-7 or MCF-7-COX2 (8 × 106) cells were mixed with RMF-EG (6 × 106) cells in Hanks balanced salt solution and Matrigel (BD Transduction Laboratories, San Jose, CA, USA). Cells were inoculated into the fourth mammary fat pads of 6-week-old female BALB/cAnN.Cg-Foxn1nu/CrlNarl mice. Before the inoculation of the cancer cell/fibroblast mixture, all mice were primed with 6 mg/kg of 17β-estradiol twice a week for 3 weeks.
After inoculation, 17β-estradiol was continuously given to mice throughout the experiments. Measurement of tumor growth was begun at 4 weeks after injection, and tumor volume was calculated by using the equation: tumor volume = (length × width2)/2. After 10 weeks, mice injected with COX-2-overexpressing MCF-7 and RMF-EG produced tumors with volumes approximately 200 mm3 and were randomly divided into four groups that received vehicle (DMSO), NS-398 (10 mg/kg), L-1-methy-tryptophan (10 mg/kg), or both inhibitors 5 times per week.
Two weeks later, animals were killed, and tumors were isolated from mice. The statistical difference between experimental groups was evaluated with repeated-measures two-way ANOVA analysis. The animal-use protocol was approved by the Institutional Animal Care and Use Committee of National Health Research Institutes.
Patients and statistical analysis
Paraffin-embedded human breast tumor tissues were obtained from Chi-Mei Medical Center (Tainan, Taiwan) between 1998 and 2004. The slides were stained with anti-COX-2 or anti-IDO antibodies. The COX-2 and IDO stainings were interpreted by using the H-score, defined by the following equation: H-score = ΣPi (i + 1), as previously described [
19], where i is the intensity of the stained tumor cells (0 to 3+), and Pi is the percentage of stained tumor cells with various intensities. We classified tumors with cancer cells and stromal cells showing H-scores no less than the median of all scored cases as having high COX-2 and IDO expression, respectively.
The follow-up duration ranged from 5.4 to 143.6 months, with a mean of 87.1 months. Survival analyses for disease-specific and metastasis-free survival were performed by using Kaplan-Meier plots and compared by using the log-rank test. The correlation between COX-2 and IDO expression with clinicopathologic parameters was examined with χ2 test. P value < 0.05 was considered statistically significant. This study was approved by the Research Ethics Committee of National Health Research Institutes. Written informed consent was obtained from all patients participating in this study.
Discussion
Previous studies demonstrated that IDO overexpression increases the secretion of kynurenine to inhibit effect T cells to promote immune escape and tumor progression in various human cancers [
27]-–[
29]. The expression of IDO in cancer stroma has not been clarified. In addition, the clinical significance of stromal IDO is unclear.
In this study, we provide evidence that COX-2-overexpressing breasts cancer cells may secrete PGE
2 to induce IDO expression and kynurenine production in stromal fibroblasts. In addition, we show that kynurenine in the coculture-conditioned medium is produced mainly by CAFs because IDO is not induced by COX-2 overexpression in MCF-7 cells. An important upstream regulator of IDO is interferon-γ. Yoshida
et al. [
30] first reported that the pulmonary IDO was induced in the mouse after intraperitoneal administration of bacterial endotoxin or during
in vivo virus infection, and this induction was triggered by interferon-γ [
30]. Because interferon-γ exhibits antitumor activity on various cancers
in vitro and
in vivo, it is unlikely that COX-2-overexpressing cancer cells produce interferon-γ to stimulate stromal IDO. For the first time, we show that cancer cell-produced PGE
2 transcriptionally upregulates IDO expression through the EP4/STAT3 signaling pathway.
In vivo binding of STAT3 to
IDO gene promoter is confirmed by ChIP assay. Additionally, knockdown of STAT3 totally abolishes EP4 agonist-induced IDO expression. These data suggest that
IDO is a direct transcriptional target of STAT3.
An unresolved question is why PGE
2 stimulates IDO expression in stromal fibroblasts but not in breast cancer cells, because both cell types express EP4 receptor [
31] and data not shown]. We are aware that the binding of STAT1 to
IDO promoter is reduced by PGE
2 (Additional file
3: Figure S2); therefore, it is possible that the expression level of STAT1 and STAT3 and the competition between these two STATs may determine the response of cells to PGE
2 stimulation.
The concept of oncometabolite was established by the studies that mutations of isocitrate dehydrogenase 1 (IDH1) and IDH2 generate a novel metabolite 2-hydroxyglutarate (2-HG) that exhibits oncogenic activity in acute myeloid leukemia and glioma [
32],[
33]. Subsequently, 2-HG was shown to be a competitive inhibitor of α-ketoglutarate-dependent dioxygenases and inhibits histone demethylases like Tet methylcytosine dioxygenase 2 (TET2) to change promoter methylation and gene transcription [
34],[
35]. Kynurenine represents another oncometabolite, which acts as an immunosuppressor to create a favorable microenvironment for tumor formation and metastasis [
36]. A recent study demonstrated that the tryptophan catabolism enzyme TDO is overexpressed in human brain tumors, and elevated secretion of kynurenine promotes cell migration via an AhR-dependent pathway [
25].
However, the underlying mechanism by which kynurenine increases cell motility is still unclear. After screening of the EMT markers, we found that E-cadherin is decreased in kynurenine-treated breast cancer cells, and AhR is involved in this process. AhR has been shown to integrate as a component of a novel Cul4B ubiquitin E3 ligase complex and participated in the degradation of sex steroid receptors [
26]. We demonstrated that kynurenine increases the interaction between AhR and E-cadherin, and the AhR/E-cadherin complex also contains Skp2, an F-box protein of SCF E3 ligase. The formation of the E-cadherin/AhR/Skp2 complex and ubiquitination of E-cadherin induced by kynurenine is also detectable in A549 cells, indicating a general mechanism of kynurenine-induced proteolysis of E-cadherin in different cancer cells. Our results provide a novel oncometabolite function of kynurenine to enhance cancer cell migration by degrading E-cadherin.
The clinical validation of tumor COX-2 and stromal IDO in this study is important to verify the cancer-stroma interplay in cancer progression. Many histopathologic studies investigated the expression of two specific genes in the epithelial components of tumor tissues to show their association and to demonstrate the vertical regulation of these two genes. The correlation and clinical significance of genes separately expressed in tumor and stroma have received little attention.
However, the gene signatures in CAFs may provide more information than originally thought. West
et al. [
37] first classified two stromal gene signature from tumors with solitary fibrous tumor (SFT) and desmoids-type fibromatosis (DTF) features and showed that patients with the expression of DTF had a favorable clinical outcome. Their subsequent study by using public databases and immunohistochemical approaches suggested that DTF fibroblast signature is a common tumor stroma signature in different types of cancers [
38]. Mercier
et al. [
39] identified a hyperproliferative gene signature in CAFs and found that breast cancer patients with this signature had a poor prognosis with tamoxifen monotherapy and a great reduction in recurrence-free survival [
39]. By using a mouse model of squamous skin carcinogenesis, Erez [
40] demonstrated that carcinoma cells could educate CAFs to express proinflammatory genes to promote macrophage recruitment, neovascularization, and tumor growth. Additionally, this gene signature was also evident in mammary and pancreatic tumors in mice and in human cancers. By using metabolomics, molecular, and pathological approaches, we revealed that induction of stromal IDO by COX-2-overexpressing breast cancer cells promotes tumor progression and predicts poor patient survival.
Results of our animal study also clearly demonstrate the anticancer effect of COX-2 and IDO inhibitor on COX-2-overexpressing breast cancer in vivo. A combination of IDO and COX-2 inhibitor exhibits a more obvious effect on the inhibition of tumor growth. However, we did not find an additive effect. This can be because (a) the number of animals in each group is small, and (b) inhibition of COX-2 in cancer cells will attenuate stromal IDO expression, which reduces the anticancer activity of IDO inhibitor. Additional experiments are needed to clarify this issue.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
JY conducted cell culture, biochemical, and molecular biology assays and prepared the draft of the manuscript. CF did IHC study and pathological analysis and helped to draft the manuscript. CC performed metabolomics study and data analysis and helped to draft the manuscript. KK participated in the design of the study and helped to draft the manuscript. MF conceived of the study and participated in the preparation of primary cancer-associated fibroblasts. WC conceived of the study and participated in the design of the study and wrote the manuscript. All authors read and approved the final manuscript.