Decorin, an exercise-induced secretory protein, is associated with improved prognosis in breast cancer patients but does not mediate anti-tumorigenic tissue crosstalk in mice

Purpose: Regular exercise can reduce incidence and progression of breast cancer, but the mechanisms for such effects are not fully understood. The purpose of this study was to examine the mechanisms behind the protective effects of exercise. Methods: We used a variety of rodent and human experimental model systems to determine whether exercise training can reduce tumor burden in breast cancer and to identify mechanism associated with any exercise training effects on tumor burden. Results: We show that voluntary wheel running slows tumor development in the mammary specific polyomavirus middle T antigen overexpression (MMTV-PyMT) mouse model of breast cancer but only when mice are not housed alone. We identify the proteoglycan decorin as a contraction-induced secretory factor that systemically increases in patients with breast cancer immediately following exercise. Moreover, high expression of decorin in tumors is associated with improved prognosis in patients, while treatment of breast cancer cells in vitro with decorin reduces cell proliferation. Notwithstanding, when we overexpressed decorin in murine muscle or injected recombinant decorin systemically into mouse models of breast cancer, elevated plasma decorin concentrations did not result in higher tumor decorin levels and tumor burden was not improved. Conclusion: Exercise training is anti-tumorigenic in a mouse model of luminal breast cancer, but the effect is abrogated by social isolation. The proteoglycan decorin is an exercise-induced secretory protein, and tumor decorin levels are positively associated with improved prognosis in patients. The hypothesis that elevated plasma decorin is a mechanism by which exercise training improves breast cancer progression in humans is not, however, supported by our pre-clinical data since elevated circulating decorin did not increase tumor decorin levels in these models.