Supplementary Figure 4: A G-CSF responsive gene signature (G-CSF-sig) links the mTOR activity to MDSC infiltration in human breast cancer.

(a) Heat maps show the expression of G-CSF signature as a single score (top), and TCR pathway components (bottom) in the EMC-MSK dataset. The red sticks above the heat maps indicate tumour samples whose gene expression level exceeds the mean of all tumours by more than 2 × s.d. n = 615 patients. (b–d). Three human datasets were analysed for the relationship between expression of G-CSF signature and T-cell activation signature in tumour biopsies. n = 615, 409, and 1992 patients for (b), (c), (d), respectively. (e) As a negative control, the analysis was applied to a panel of tumour cell lines (n = 51, available from GSE12777), showing no correlation as expected. For (b–e), P values were determined based on two-side Student’s t tests for Pearson correlation coefficients. (f) Multivariate (MV) analysis of METABRIC and EMC-MSK datasets using the Cox Proportional-Hazards model to estimate the hazard ratios (HR) of G-CSF-sig. The P values were computed based on the Multivariate Cox Proportional-Hazards model. (g) MDSC quantity in P53N-C tumour-bearing hosts that are either wild type (WT) Balb/c mice or athymic nude (Nu) mice. TF (−) controls are shown for comparison. ‘n.s.’: no significant difference. n = 5 animals per group. Error bars indicate s.e.m., and P values are calculated by two-tailed Student’s t tests. (h) Representative Western blotting of pS6K(T389) and quantitative PCR assay of G-CSF in PDX tumours. Dotted lines indicate the matching between the two types of data across different PDX lines. Error bars indicate s.e.m.