Phosphoproteins

We assessed protein expression levels in tumor and adjacent-normal tissue according to the three PA levels using one-way ANOVA. Regression models were only performed for protein expression in tumor tissue because the number of patients with the adjacent-normal tissue would not provide sufficient statistical power. Linear regression was performed to estimate the difference in H-scores for mTOR between the categories of PA levels because it was normally distributed. For the phosphoproteins, because a high proportion of tumors were negative for expression (H-score = 0 for 12% of p-mTOR staining, 27% of p-AKT, and 21% of p-P70S6K; Supplementary Table S1), gamma hurdle models were used to model the positive (nonzero) versus negative (zero) data with a logistic model and then the positive data with a gamma model with a log-link. ORs and percentage differences were converted from the regression coefficients of the respective parts of the gamma hurdle models. The covariates included age (continuous), race (Black or White), educational level (≤high school, some college, or ≥college graduate), menopausal status (premenopausal or postmenopausal), BMI (continuous), history of diabetes (yes or no; defined as using any oral or injection medication for diabetes), breast cancer molecular subtype (HR⁺/HER2⁻, HR⁺/HER2⁺, HR⁻/HER2⁺, or HR⁻/HER2⁻), tumor grade (low, intermediate, or high), tumor size (<1.0, 1.0–1.9, or ≥2.0 cm), and disease stage (American Joint Committee on Cancer staging system stage 0/I, II, or III/IV). Because BMI and history of diabetes can be intermediate variables on the causal pathway between PA and mTOR signaling, we also performed models without each of these two variables as a sensitivity analysis. Because dietary caloric intake is associated with PA levels and can modulate p-mTOR and p-AKT activities (34 ), total energy intake was additionally adjusted in a sensitivity analysis. Furthermore, to examine specific PA intensity in association with mTOR signaling pathway activity, we estimated the associations for (i) moderate PA (3 to <6 METs of exercise for at least 150 minutes/week defined as sufficient) among patients without vigorous PA and (ii) vigorous PA (≥6 METs of exercise for at least 75 minutes/week defined as sufficient) regardless the level of moderate PA. Exploratory stratification analyses were performed by race, BMI, menopausal status, history of diabetes, total energy intake levels, breast cancer stage, tumor grade, ER status, tumor size, and lymph node status because these variables were associated with energy balance or the mTOR pathway signaling. All tests of statistical significance were two sided; a P value less than 0.05 was considered statistically significant. All analyses were a priori, and the results were not adjusted for multiplicity.

All experiments were performed in biological triplicates for different treatments (CK, Cold, and EBR-Cold), and sequencing libraries were constructed independently. Libraries from triplicate samples were combined and sequenced in two separate experiments. The lysis buffer of phosphoprotein extraction contained phosphatase inhibitors (PhosSTOP, Roch, Germany), and the extraction method was the same as in Protein extraction.