Modulation of CXCR3 ligand secretion by prostaglandin E2 and cyclooxygenase inhibitors in human breast cancer

Introduction: In murine breast cancer models, the two interferon-gamma (IFN-gamma) inducible chemokines and CXC-chemokine receptor 3 (CXCR3) receptor ligands, monokine induced by gamma-interferon (CXCL9) and interferon-gamma-inducible protein-10 (CXCL10) impair tumor growth and metastasis formation through recruitment of natural killer (NK) cells and tumor-suppressive T lymphocytes. In human breast cancer, CXCL9 mRNA overexpression correlates with the number of tumor infiltrating lymphocytes and predicts response to different chemotherapeutic regimens. Raising the intratumoral CXCR3 ligand concentration is therefore a possible way to enhance immune intervention in breast cancer. Little is known, however, about expression levels and regulation of these chemokines in human breast cancer. Since the inhibition of cyclooxygenases (COX) has been shown to reduce tumor growth and incidence of metastases in a lymphocytic and IFN-gamma dependent manner, we argued that COX isoenzymes are a pharmacologic target to increase intratumoral CXCR3 ligand concentration in human breast cancer. Methods: CXCL9 was visualized in breast cancer specimens by immunohistochemistry, expression levels of CXCL9 and cyclooxygenases were determined by ELISA and western blotting, respectively. For regulation studies, Michigan Cancer Foundation-7 (MCF-7) and M. D. Anderson - Metastatic Breast 231 (MDA-MB 231) breast cancer cells were stimulated with IFN-gamma with or without prostaglandin E-2 (PGE(2)) or COX inhibitors (indomethacin, acetylsalicylic acid (ASA), celecoxib). CXCR3 ligand release from cells was measured by ELISA. Results: Within the tumor microenvironment, cancer cells are the major source of CXCL9. PGE(2) impairs IFN-gamma mediated CXCL9 and CXCL10 release from MCF-7 and MDA-MB 231 cells, and inhibition of endogenous cyclooxygenases by indomethacin or ASA correspondingly increases this secretion. Otherwise, high concentrations of the Cyclooxygenase-2 (COX-2) specific antagonist celecoxib have opposite effects and impair CXCL9 and CXCL10 release. In human breast cancer tissue specimens there is an inverse correlation between COX-2 overexpression and CXCL9 concentration, suggesting that the observed in vitro effects are of importance in vivo as well. Conclusions: Suppressing endogenous PGE(2) synthesis by cyclooxygenase inhibition increases CXCL9 and CXCL10 release from breast cancer cells and is therefore a pharmacologic candidate to enhance intratumoral immune infiltration. Yet, to this end the unselective COX inhibitors ASA and indomethacin seem preferable to celecoxib that at higher concentrations reduces CXCR3 ligand release most probably due to COX independent mechanisms.