Fig 2: CXCL1 secreted by mouse tumour cells attracts CXCR1-expressing ASCs.(a) Left: RT–PCR analysis of CXCL1 messenger RNA expression in mouse ASCs and indicated mouse tumour cell lines. Data were normalized to 18S RNA. Right: CXCL1 protein concentration in medium conditioned by mouse ASCs and indicated tumour cell lines measured by ELISA. ~ values above 5 ng ml-1. (b) Immunofluorescence (IF) analysis of tumours grown in non-obese and obese mice with anti-CXCL1 antibodies indicating CXCL1 expression increased in obesity. Endothelium is counterstained with isolectin B4 (IB4). Nuclei are blue. Scale bar, 100 µm. (c) IF analysis of mouse WAT with antibodies against CXCR1 (left) and CXCR2 (right), and endothelial cell marker CD31. Note CXCR1 expression in perivascular ASCs (arrows), while CXCR2 is expressed only in circulating leukocytes and macrophage crown structures (arrowheads). Scale bar, 100 µm. (d) RT–PCR analysis of CXCR1 and CXCR2 expression in mouse ASCs and RM1 tumour cells. Data were normalized to 18S RNA. (e) Primary mouse ASCs were subjected to migration through 8-µm pores towards indicated concentrations of CXCL1 in a transwell chamber. In a,d and e, graphs show mean±s.e.m. for technical triplicates; *P<0.05 versus control (Student's t-test). Experiments in b and c were performed for three mice with similar results; representative images are shown.

Fig 3: CXCR1/CXCR2 silencing inhibits ASC migration to tumours.(a) Immortalized GFP-expressing mouse ASCs transduced with an untargeted shRNA (control-sh), or shRNA targeting either only CXCR1 (CXCR1-sh) or both CXCR1 and CXCR2 (CXCR1/2-sh) were subjected to immunoblotting with CXCR1 and CXCR2 antibodies. Experiment was performed twice with similar results; representative scans are shown. (b) Cells from a were subjected to transwell chamber migration through 8-µm pores toward medium with or without 5 ng ml-1 CXCL1. Plotted are relative migrated cell numbers normalized to migration of SC ASCs towards 10% FBS (100%). Shown below are representative bright-field micrographs of crystal violet-stained transwell membranes corresponding to each treatment group. Scale bar, 100 µm. (c) Cells from a were subjected to transwell chamber migration through 8-µm pores towards CM from control-shRNA or CXCL1-shRNA RM1 cells supplemented with antibodies (abs) against CXCR1 or CXCR2 where indicated. (d) Representative sections of tumours formed by control-sh-RM1 and CXCL1-sh RM1 cells grafted into the nude mice (n=6) that received subcutaneous contra-lateral injections of control-sh, CXCR1-sh or CXCR1/2-sh GFP-expressing ASCs. Arrows indicate tumour-engrafted ASCs (green), recruitment of which is reduced by both CXCL1 and CXCR1 silencing. Scale bar, 100 µm. Nuclei are blue. Representative flow cytometric analyses of suspended tumour cells enumerating injected ASCs (GFP+) among viable (4,6-diamidino-2-phenylindole (DAPI) negative) cells. Note ASC frequency reduction by CXCL1 and CXCR1 silencing. For b–d, experiments were performed in technical triplicate. Graphs show mean±s.e.m.; *P<0.05 (Student's t-test) versus control-sh CM.

Fig 4: Human chemokine/receptor expression in prostate cancer and WAT.(a) Immunofluorescence (IF) analysis of representative tumour samples from identified patients belonging to the specified groups with antibodies against CXCL1, CXCL8 and CXCR1 with red and green fluorophore-conjugated secondary antibodies. Yellow colour indicates protein co-expression. Arrows, stroma; E, epithelium. Blue, nuclei. Scale bar, 100 µm. (b) Analyses of protein expression in the tumour based on the data for 37 patients (Supplementary Fig. 1c). Per cent of patients presenting indicated protein expression within each group is shown. (c) IF analysis of representative human periprostatic WAT samples with antibodies against CXCR1 and CXCR2 counterstained with endothelium-specific isolectin B4 (IB4), showing that perivascular stroma (arrows) expresses CXCR1, while CXCR2 is only expressed in leukocytes (arrowheads) observed inside blood vessels. A, adipocytes. Scale bar, 100 µm. (d) IF analysis of a representative human tumour/PP WAT junction showing CXCR1 expression by the stroma of tumour and WAT (arrows) and perilipin-1 expression by adipocytes. Scale bar, 100 µm. (e) Bright-field micrographs of adherent ASCs from SC and PP WAT of an obese prostate cancer patient. Scale bar, 100 µm. (f) RT–PCR analysis quantifying CXCR1 and CXCR2 messenger RNA expression in ASCs isolated from SC and PP WAT of obese and non-obese patients. Data are normalized to 18S RNA. (g) ASCs isolated from SC and PP WAT of obese and non-obese patients were subjected to migration through 8-µm pores towards increased indicated concentrations of CXCL1 in a transwell chamber. Plotted are relative migrated cell numbers normalized to migration of SC ASCs towards 10% FBS (set at 100%). Experiments in (a,c–e) were performed for three patients; representative images are shown. In f and g, graphs show mean±s.e.m. for technical triplicates; *P<0.05 (Student's t-test) versus control (no CXCL1).

Fig 5: CXCL1 silencing inhibits ASC trafficking to tumours and tumour growth.(a) Proliferation of control-shRNA and CXCL1-shRNA RM1 cells growing in 1or 10% FBS assessed by MTT assay. Plotted is relative fluorescence unit (RFE) increase above day 0 value from technical triplicates. (b) RM1 cells transduced with an untargeted shRNA (control-sh) or shRNA-targeting CXCL1 (CXCL1-sh) were subcutaneously grafted into non-obese or obese mice (n=5 per group), and tumour volume was measured weekly. (c) Growth of RM1 grafts in mice treated or not treated with reparixin (n=5 per group) in lean and obese mice. In a–c, graphs show mean±s.e.m.; *P<0.05 (Student's t-test). (d) Representative flow cytometric gating to separate tumour CD45- (stromal, vascular and malignant) and CD45+ (haematopoietic) cells. Gating of CD45-CD34+Sca1+ ASCs based on side scatter (SSC) and CXCR1/2 expression reveals increased frequency of CXCR1+ ASCs in control-shRNA, but not in CXCL1-shRNA tumours in obese mice. Gating of CD45- leukocytes based on CD11b and Gr1 expression reveals reduced frequency of CD11b+Gr1+ MDSCs in CXCL1-shRNA tumours in both lean and obese mice. % of total viable cells is shown. Analyses were performed twice with similar results.