Fig 2: DKK3 is upregulated in the stroma of breast, colon and ovarian cancers. a Tukey boxplots showing z-score values of DKK3 mRNA expression in normal and cancerous stroma from breast, colorectal and ovarian cancers (Breast: normal, n = 6; cancer, n = 53. Colon: normal, n = 4; cancer, n = 13. Ovary: normal, n = 8; cancer, n = 31). b Representative images of DKK3 staining in breast, colorectal and ovarian cancers and normal tissues. Scale bar, 100 µm. c Tukey boxplots showing quantification of DKK3 staining (Histoscore) in breast, colorectal and ovarian cancers and normal tissue counterparts (Breast: normal/adjacent, n = 9; cancer, n = 109. Colon: normal/adjacent, n = 14; cancer, n = 107. Ovary: normal/adjacent, n = 8; cancer, n = 138). d Tukey boxplots showing DKK3 Histoscore in non-invasive breast cancers (Stage 1&2), invasive breast-cancers (Stage 3&4) and normal tissue counterparts. Left graph shows all cancers irrespective of their subtype (normal, n = 9; Stage 1&2, n = 74; Stage 3&4, n = 74). Middle graph shows ER-negative breast cancers (normal, n = 9; Stage 1&2, n = 40; Stage 3&4, n = 40). Right graphs shows ER-positive breast cancers (normal, n = 9; Stage 1&2, n = 21; Stage 3&4, n = 21). e Disease-free survival of breast cancer patients stratified on stromal DKK3 gene expression (GSE9014, ER-negative patients). f Images show DKK3 (green), vimentin (VIM; red) and DAPI (blue) staining of two representative human breast cancer tissues. Scale bar, 50 µm. g Tukey boxplot shows Dkk3 mRNA expression levels (relative to Gapdh) in Cancer cells (Epcam+), immune cells (Cd45+), endothelial cells (Cd31+) and fibroblasts (Pdgfra+) from MMTV-PyMT tumorus (n = 4). h Graphs show correlations between the expression of DKK3 and ACTA2, FAP and COL1A2 in normal and cancerous stroma from mammary gland (GSE9014), colorectal (GSE35602) and ovarian (GSE40595) human tissues. Pearson correlation coefficient (r) is shown. Each dot represents z-score values from individual patients. i Kaplan–Meier curves of recurrence-free survival, disease-specific survival and progression-free survival of ER-negative breast cancer, colorectal cancer and ovarian cancer patients, respectively, based on DKK3 and DKK2 gene expression. Where indicated, individual p values are shown; alternatively the following symbols were used to describe statistical significance: *P < 0.05; **P < 0.01; ***P < 0.001; #P < 0.0001; n.s., non-significant

Fig 3: Bulk RNA-seq analysis of cells sorted by surface marker.a In situ hybridization of E16.5 mouse embryo showing dermis. eYFP, Pdgfra, and Thy1 signals were observed at dermis while only DAPI (blue color) was observed at epidermis. b Section staining of E16.5 mouse embryo showing dermis. eYFP, Pdgfra, and Crabp1 signals were observed at dermis while only DAPI (blue color) was observed at epidermis. c Section staining of E16.5 mouse embryo showing muscle fiber and brown adipose tissue. While DAPI and YFP signal were observed at both locations, muscle and brown adipose tissue were respectively stained by MF20 and Pparg antibodies. d In vitro culture assay of E14.5 YFP+ cells sorted by the combination of Pdgfra and Thy1 antibodies. Scale bar, 100 µm. e Bar plot of Oil Red O intensity between cultured samples. Student’s t-test was performed between two culture conditions. Values are means ± SEM, n = 3; ns not significant, ***P < 0.001. f Culture schematic and RT-qPCR result of E14.5 Pdgfra+/Thy1- cells (LP) before and after TGFß1 ligand treatment. Two-way ANOVA was performed followed by Dunnett’s multiple comparisons test. Values are means ± SEM, n = 3; ns, not significant; **P < 0.01,***P < 0.001, ****P < 0.0001. g Venn diagrams of highly variable gene sets between sorted and cultured samples. Highly variable genes between two populations were selected. Key genes found in the corresponding sets are listed. Genes with normalized expression data > 0.3 were kept as set in the diagram.

Fig 4: Flow cytometry analysis of the identity of post-Passage 3 culturesPassage 3 cryopreserved primary fibroblasts were thawed and analyzed 24 h later by flow cytometry. Data was analyzed with Flowlogic v8.(A) Gating strategy applied to a representative culture of primary fibroblasts (derived from a young 4 months-old male). Debris is first gated out from the FSC-A vs. SSC-A plot of the entire population. Single cells are then gated based on the FSC-A vs. FSC-H plot. From this singlet population, CD45- cells are obtained by gating on the CD45 vs FSC-A plot. Finally, we use this CD45- population to gate for CD90.2+ and CD140a+ cells.(B) Same gating strategy as in (A) applied to a population of peritoneal lavage cells, which should contain a large fraction of immune (CD45+) cells.(C) Fibroblast cultures from three 4 months-old males and females were analyzed by flow cytometry, and a beeswarm plot was generated for the percentage of cells staining for combinations of immune (CD45) and fibroblast (CD90.2 and CD140a) markers.(D) The same plot as in (C) but for a population of peritoneal lavage cells as a CD45+ control population.

Fig 5: Wnt5a inhibits adipogenic gene expression in embryonic multipotent progenitors.a Violin plot of Wnt5a expression in Smart-seq2 and 10× datasets. b Scheme of Wnt5a treatment on embryonic progenitor cells. E14.5 embryos were harvested, and both Pdgfra+/Thy1− and Pdgfra+/Thy1+ populations were sorted. Cells were treated with 2.5 ng/mL bFGF to reach confluency. Wnt5a was then applied and 48 h after treatment, cells were harvested for analysis. c Embryonic cells were harvested and subjected to Western blot assay. Ponceau S staining was used as loading controls. d RT-qPCR analysis of Wnt5a-treated embryonic progenitor cells in b. One-way ANOVA was performed on Pdgfra+/Thy1− cultured cells, followed by Dunnett’s multiple comparisons test. Student’s t-test was performed in Pdgfra+/Thy1+ cultured cells. Values are means ± SEM, n = 3; ns, not significant; **P < 0.01, ***P < 0.001.