Fig. 6

The fibroblast compartment changes dramatically from a healthy fat pad to a breast tumour. a-b SPICE visualisation of Boolean CAF subsets (within the CAF+ population). a 4T1 tumours and b 4T07 tumours. The arcs around the pie charts corresponds to each of the 6 CAF markers. In this way, each pie slice corresponds to one theoretical Boolean subset of CAFs, subset S32 is marked on each pie with a black border as an example. It is evident that not all of the theoretical subsets are present in the tumours, and that the overall composition of CAF subsets change as tumours matures. Furthermore, the initial difference between tumour types seen at D7 is translated into a difference at the latest stage at D21. Statistical p-values are from permutation testing comparing the CAF subset composition within each pie to that of another pie. Data in A-B are from three independent repeats combined (128 tumours in total), 4T1 D7 n = 21, 4T1 D14 n = 24, 4T1 D21 n = 24, 4T07 D7 n = 24, 4T07 n = 21, 4T07 D21 n = 14. c Model. In normal mammary fat pads fibroblasts expressing PDGFRα are the most abundant, while PDGFRβ+ fibroblasts are very scarce. Once a tumour forms this picture is reversed. As the tumour progresses the heterogeneity decreases and CAFs expressing combinations of CD26, FAPα and PDGFRβ become the dominant subpopulations