Introduction
Successful immunotherapeutic strategies in cancer treatment require an effective infiltration of the tumor by tumor-suppressive immune cells. Immunotherapeutic approaches can therefore be impeded by an unfavorable composition of the intratumoral immune milieu: while regulatory T cells (T
regs) and myeloid-derived suppressor cells (MDSCs) repress a successful immune intervention and promote tumor progression, natural killer (NK) cells and T helper (Th) 1 CD4
+/CD8
+ lymphocytes are potent mediators of anti-tumor activity [
1‐
3]. However, insufficient migration of this type of immune cells towards the tumor microenvironment will not allow attacking of cancer cells by these cells in patients with advanced tumors [
4,
5].
The CXCR3 chemokine receptor is preferentially expressed on the surface of NK cells or Th1 tumor-suppressive T lymphocytes and is responsible for their chemotactic recruitment into the tumor tissue [
6,
7]. Correspondingly, high intratumoral concentrations of the interferon (IFN)-γ-inducible chemokines CXCL9 and CXCL10, two of the CXCR3 ligands, are associated with increased immune infiltration and improved survival in patients with solid malignancies [
8‐
13]. In human breast cancer, we and others have shown that a high expression of the
CXCL9 mRNA correlates with an increased number of infiltrating lymphocytes and a better response to chemotherapy [
14,
15]. Furthermore, in a mouse model transfection of murine breast cancer cells with CXCL9 increases chemotactic T cell recruitment, impairs tumor growth, prevents lung metastasis formation, and prolongs survival [
16]. Raising the intratumoral concentration of CXCR3 ligands is therefore a feasible therapeutic option to improve immune intervention. Still, origin and regulation of CXCR3 chemokines in human breast cancer are poorly understood.
A conceivable way to shift the tumor microenvironment to a more tumor-suppressive Th1 milieu is modulation of the cyclooxygenase (COX) system. Two isoenzymes, constitutively expressed COX-1 and inducible COX-2, are found in human breast tumors. COX-2 overexpression is associated with reduced infiltration of tumor-suppressive immune cells, and COX inhibition in turn enhances immunosurveillance [
17‐
19]. Moreover, prostaglandin E
2 (PGE
2), the major product of COX in tumors, promotes tumor growth at least in part by reducing the activity of NK cells and expanding MDSCs and T
regs [
20,
21]. In breast cancer models, both COX inhibition and PGE
2 receptor antagonism suppress local tumor growth and metastatic spread in an IFN-γ and T cell- or NK cell-dependent manner [
22‐
24].
In light of these findings, we explored whether components of the COX pathway would be pharmacologic candidates to enhance CXCR3 ligand concentration in human breast cancer. We now demonstrate that PGE2 inhibits IFN-γ induced CXCL9 and CXCL10 secretion from breast cancer cells and that, conversely, the COX inhibitors acetylsalicylic acid (ASA) and indomethacin augment this release. Inverse correlation of COX expression and intratumoral CXCL9 concentration in human breast cancer samples indicate the relevance in vivo. The COX-2-specific inhibitor celecoxib, however, has the opposite effects at higher concentrations implicating that the choice of the appropriate COX inhibitor for clinical use seems to be decisive. In summary, our results provide a mechanistic link between the COX pathway and a reduced infiltration of tumor-suppressive lymphocytes in breast cancer through the modulation of intratumoral CXCR3 chemokine release.
Materials and methods
Reagents and cell lines
Dulbecco's modified Eagle's medium (DMEM), fetal calf serum (FCS), gentamycin, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and glutamine were from Gibco Life Technologies (Gaithersburg, MD, USA); recombinant human interferon gamma (IFN-γ) and recombinant human TNF-α were from PeproTech (Hamburg, Germany); prostaglandin E2, indomethacin, and celecoxib were from Cayman Chemicals (Ann Arbor, MI, USA) and reconstituted in dimethyl sulfoxide (DMSO); acetylsalicylic acid (ASA) was from Sigma (St. Louis, MO, USA) and reconstituted in phosphate buffered saline (PBS); bovine serum albumin was from Sigma (St. Louis, MO, USA). Antibodies to human antigens: monoclonal mouse anti-CXCL9 antibody (clone 49106, R&D Systems, Minneapolis, MN, USA); monoclonal mouse anti-COX-1 (clone COX111, Invitrogen, Camarillo, CA, USA); monoclonal mouse anti-COX-2 (clone CX229, Cayman Chemicals, Ann Arbor, MI, USA); monoclonal mouse anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH, clone 6C5, Millipore, Billerica, MA, USA); horseradish peroxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch, Burlington, ON, USA). All other chemicals were of analytical grade and obtained from Merck (Darmstadt, Germany) or Sigma (St. Louis, MO, USA).
MCF-7 and MDA-MB 231 human breast cancer cell lines (American Type Culture Collection, Manassas, VA, USA) were cultured in a humified 5% CO2 atmosphere at 37°C in DMEM supplemented with glutamine, 10% FCS, 10 mM HEPES, and 20 μg/mL gentamycin.
Human tissue samples and patient characteristics
Fresh-frozen tissues from 60 breast cancer patients who were treated at the Department of Gynecology and Obstetrics, Technical University Munich, between 1991 and 2004 were selected. The study was approved by the local ethics committee, all patients had given written informed consent. All tumors came from nodal-positive, non further metastasized tumors that were estrogen receptor positive (N+, M0, ER+).
Immunohistochemistry
Immunohistochemistry was performed on 4 μm thick paraffin sections from invasive breast cancer tissue specimens (n = 20) obtained from patients treated at the Department of Gynecology and Obstetrics, Technical University Munich. Briefly, sections were deparaffinized by treatment with xylene followed by a graded series of ethanol (100% to 70%) in distilled H2O and subjected to heat-induced epitope retrieval in citrate buffer (pH 6.0). Endogenous peroxidase was blocked using 3% H2O2 in distilled H2O, 20 minutes, room temperature, followed by antigen blocking with 5% goat serum in 50 mM Tris-HCl, pH 7.4, 10 minutes, room temperature. The sections were then incubated with mouse anti-CXCL9 diluted to a final concentration of 20 μg/mL in antibody diluent (DAKO, S2022), one hour, room temperature. For detection of primary antibody binding the ZytoChem Plus HRP Broad Spectrum Kit was employed (Zytomed Systems, Berlin, Germany) according to the manufacturer's instruction. Sections were washed thoroughly between incubations and counterstained with hematoxylin.
Determination of CXCL9 and CXCL10 secretion from breast cancer cells
MCF-7 and MDA-MB 231 cells were plated on 24-well culture plates and grown to 70 to 80% confluency before they were washed in PBS and starved for 24 hours in serum-free medium. The medium was then replaced by serum-free medium and test reagents added as indicated. In case of inhibition studies the cells were preincubated with prostaglandin E2 or cyclooxygenase inhibitors (indomethacin, ASA, celecoxib) for 30 minutes before addition of IFN-γ. After 24 hours the supernatants were collected and stored at -20°C until further use. In each experiment, at least six wells were stimulated. Subsequently the culture supernatants were subjected to ELISA for determination of chemokine concentrations. The DuoSet ELISA Kits DY392 and DY266 (R&D Systems, Minneapolis, MN, USA) were used for the determination of CXCL9 and CXCL10, respectively, according to the manufacturer's instructions.
MTT assay
Viability of cells subjected to cytokines, cyclooxygenase inhibitors, or prostaglandin E2 was assessed using the MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, Sigma-Aldrich, St. Louis, MO, USA). A sample of 5 × 103 MCF-7 or 7 × 103 MDA-MB 231 cells were plated into each well of a 96-well plate and cultured for 24 hours. After starving in serum-free medium for another 24 hours, cells were stimulated with the respective stimulants for 24 or 48 hours as indicated in the results section. Then, 20 μL of MTT was added to a final concentration of 200 μg/mL and the cells incubated at 37°C for two hours, followed by the addition of 100 μL DMSO. Absorbance was measured at 590 nm.
Quantification of CXCL9 concentration in tumor tissue extracts
Breast cancer tumor tissue specimens were obtained at surgery, inspected by a pathologist, and then stored in liquid nitrogen until further use. Tumor tissue homogenates were prepared as described previously [
25]. Total protein concentrations were determined applying the BCA Protein Assay Kit (Pierce, Thermo Scientific, Rockford, IL, USA). Tissue Extracts were diluted 1:5 in PBS/1% BSA and then subjected in duplicates to DuoSet ELISA kit (R&D Systems, Minneapolis, MN, USA).
Western blot analysis
After incubation of the tumor cells with stimulants culture medium was removed and cells washed with ice-cold PBS and immediately lysed in SDS-PAGE sample buffer containing 1% (w/v) Triton X-100, 150 mM NaCl, 1 mM EDTA, 50 mM HEPES, 1 mM sodium vanadate, and complete protease inhibitor cocktail (Roche, Mannheim, Germany). Samples were kept on ice for 20 minutes, subjected to ultrasound (2 × 10 seconds, 4°C), and stored at -20°C until further use. Total protein concentration was determined using the BCA Protein Assay Kit (Pierce, Thermo Scientific, Rockford, IL, USA). Equal amounts of protein (60 μg) were separated by 10% SDS-PAGE and blotted onto polyvinylidene fluoride (PVDF) membranes using a semi-dry transfer apparatus (Biometra, Göttingen, Germany). Blots were blocked with 5% (wt/vol) milk powder (Sigma, St. Louis, MO, USA) containing PBS/0.5% Tween 20 (w/v) (PBS-T) and then incubated at 4°C overnight with primary antibodies (anti-COX-1 0.5 μg/mL, anti-COX-2 0.5 μg/mL, anti-GAPDH 0.1 μg/mL). Following washing with PBS-T and incubation with horseradish peroxidase-coupled anti-mouse IgG (diluted 1:10.000 in 5% milk powder in PBS-T), one hour, room temperature, the blots were washed and the antibody reaction visualized by enhanced chemoluminescent detection (Amersham Biotech, Uppsala, Sweden). Blots were exposed to Kodak X-Omat AR-5 film (Eastman Kodak Co., Rochester, NY, USA).
Statistical analysis
Results were taken from at least three independent experiments and analyzed using the Mann-Whitney test (SPSS Statistics Software, Zurich, Switzerland, Version 17.0). Results are given as mean ± standard deviation of the mean, if not indicated otherwise. Statistical significance was defined as * P≤0.05, ** P≤0.005, or *** P≤0.0005.
Discussion
In this study, we have identified the COX pathway as a potential pharmacologic candidate to enhance the intratumoral accumulation of CXCL9 and CXCL10 levels in breast cancer. Earlier studies have highlighted the importance of CXCR3 ligands in recruiting NK cells, as well as CD4
+ or CD8
+ T lymphocytes to the tumor site [
7,
30,
35,
36]. Although the three CXCR3 ligands CXCL9-11 have redundant functions for the most part, for yet unknown reasons CXCL9 emerges as the preferred CXCR3 ligand, mediating lymphocytic infiltration and growth suppression of tumors [
30,
31]. This is in agreement with clinical studies in human breast cancer identifying CXCL9, rather than CXCL10 or CXCL11, as a potential biomarker for diagnosis of breast cancer and therapy response, suggestive of its protective role in breast cancer biology [
14,
15,
37]. These reasons led us to focus predominantly on CXCL9. Our immunohistochemical breast cancer studies localize CXCL9 to cancer cells (Figure
1). Datta et al. reported similiar findings for CXCL10 [
26], emphasizing that these cells are a major source of CXCR3 ligands in the breast tumor microenvironment. However, our results also show CXCL9 expression in endothelial cells. As endothelial cells seem to produce CXCR3 chemokines in response to similiar stimuli as cancer cells they might also participate in modulating immune infiltration in breast cancer [
38].
In our study, IFN-γ induced CXCL9 and CXCL10 secretion in a dose-dependent manner, whereas TNF-α induced CXCL10 only. Induction of CXCL10 by TNF-α is in agreement with results obtained in human eosinophils and corneal keratocytes [
39,
40]. TNF-α potentiated IFN-γ induction of both chemokines, an effect which has been described for other cell types as well and may be ascribed to the synergistic action of transcription factors such as STAT-1α (activated by IFN-γ) and NF-κB (activated by TNF-α) or to the action of transcription coactivators such as CREB binding protein [
41]. For the subsequent regulatory experiments on the effects of PGE
2 and COX inhibitors we decided to use IFN-γ stimulated cells as baseline controls, because it is the most potent inducer of both CXCR3 chemokines and a prerequisite for the immune-mediated tumor-suppressive effects of COX inhibitors in murine breast cancer models [
23].
Our results demonstrate that PGE
2 inhibits, whereas COX inhibitors induce the release of CXCL9 and CXCL10 from breast cancer cells. This is in line with similar results described for epidermoid tumor cells and various kinds of immune cells [
42‐
44]. Both effects were more pronounced in MCF-7 than in MDA-MB 231 cells, although the latter ones expressed higher levels of cyclooxygenases (Figure
4, [
32]), so that a higher effect of COX inhibition was expected. Although not the subject of the present study, one may assume that a higher activity of the PGE
2 prostanoid receptors (EP1-EP4) and their downstream targets in MCF-7 cells is responsible for this difference rather than the higher intrinsic PGE
2 production in MDA-MB 231 cells. Both cell lines do express all four EP prostanoid receptors [
45].
Our findings provide a mechanistic link between the COX pathway and CXCL9/CXCL10 chemokine secretion into the tumor microenvironment of human breast tumors. The inhibition of CXCR3 ligands might be added to the mechanisms by which PGE
2 promotes tumor escape from the immune system [
46]. Several mouse tumor models other than breast cancer have collectively demonstrated that COX inhibition enhances the efficacy of cancer vaccines and impairs tumor growth by raising the number of infiltrating Th1 lymphocytes [
17‐
19,
47]. Moreover, these studies have detected an increased expression of Th1 cytokine mRNA, including those of murine CXCL9 and CXCL10 homologs after COX inhibition. Fulton et al. reported that inhibition of COX-1 and COX-2 leads to reduced breast tumor growth and reduced metastatic spread in mice [
48]. This inhibition depends on CD4
+ and CD8
+ T cells in case of local tumor control and NK cell activity and IFN-γ in case of metastatic control [
23]. Preincubation of the cancer cells with COX inhibitors prior to injection into the animals is sufficient to cause anti-tumor actitvity, which supports our observation that the tumor cells are the source of the COX-mediated effect [
48]. Although determination of chemokine levels was not subject of these studies, our results offer an additional explanation in that enhanced CXCR3 ligand secretion from cancer cells contributes to this immune-mediated anti-cancer effect of COX inhibitors. Clinical data by Denkert et al. further support our interpretation which show a significant correlation between CXCL9 mRNA levels and infiltrating T lymphocytes with favorable chemotherapy response in breast cancer patients [
14].
Our findings demonstrate a differential regulation of CXCL9 and CXCL10 secretion by COX inhibitors. Although ASA and indomethacin increased the release of CXCL9 and CXCL10, the COX-2-specific inhibitor celecoxib had this effect only at low concentrations. The mechanisms underlying the inhibitory effect of high concentrations on CXCR3 ligand release remain unclear, but a possible explanation is that the COX inhibitory effects of celecoxib at high concentrations are superimposed by its COX-independent effects, particularly the inactivation of NF-κB signalling [
49]. Similiar findings have been reported for the NSAID sulindac, that, in contrast to ASA, impaired CXCL9 mRNA synthesis in mouse macrophages [
50]. Moreover, in a murine model of colorectal cancer, celecoxib showed COX-independent anti-tumor activity and even decreased the number of infiltrating lymphocytes [
51], consistent with an impairment of CXCR3 ligand release.
Elevated intratumoral COX and PGE
2 levels in breast cancer are known to be associated with poor outcome and development of distant metastases [
34,
52,
53]. We observed an inverse correlation between COX-2 overexpression and intratumoral CXCL9 concentration and a trend towards lower CXCL9 expression in COX-1 overexpressing breast cancer tissues. This is in line with preclinical data demonstrating that COX-2, rather than COX-1, is inducing PGE
2 synthesis in breast tumors [
54]. In this context, it might be interesting to analyze tumor tissue samples of breast cancer patients that took NSAIDs on a regular basis to see if CXCL9 expression is different from patients who did not, but such investigations were not in focus of the analysis presented.
In conclusion, our results show that COX inhibition is a feasible way to improve immunosurveillance in human breast cancer by inducing intratumoral CXCR3 binding chemokines. To this end, unselective NSAIDs such as indomethacin or aspirin might be more suitable than COX-2-specific agents. The need for COX inhibiton as a component of breast cancer therapy is further endorsed by supporting clinical data. Although the use of NSAIDs in the prevention of breast cancer has been debated for years, albeit with conflicting results [
55‐
58], recent retrospective analyses of large clinical trials have shown that NSAID intake during the course of the cancer disease is associated with a significantly decreased risk of disease recurrence and breast cancer related death [
59‐
61]. Yet, randomized prospective trials are needed to clarify the benefit of COX inhibition in breast cancer therapy and to discover the underlying mechanisms
in vivo.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
HB coordinated the studies, contributed to all of the experiments, and drafted the manuscript. SK and AS performed the stimulation experiments, western blot, and MTT assays. USB provided the patient collectives. CC and SA contributed to the immunohistochemical localization experiments. MK and MS participated in the design of the study, and supervised the research project. All authors read and approved the final manuscript.