BTRT (A) and SKRT (C) cells were grown in 3D Matrigel followed by treatment with serial doses of AMD3100 (Materials and Methods). Photographs were taken on day 13. The total acini area was quantitatively analyzed with AlphaView SA. The data were analyzed using one-way ANOVA (B, D). BTRT (E) and SKRT (G) cells were seeded at low density and treated with different doses of AMD3100. The plates were scanned on day 18. Colony numbers were quantitatively analyzed using AlphaView SA. The data were analyzed using one-way ANOVA (F, H). BTRT (I) and SKRT (K) cells were co-cultured with BCAFs in 96-well “U” bottom unattached plates and treated with AMD3100 (2.5μM; Materials and Methods). Dynamic changes of the spheres were monitored and photographed. At the end of the study, viability of the cells in monoculture or co-culture was detected using CellTiter-Glo 3D viability assay kit. The cell viability ratio of treated with AMD3100 to vehicle was analyzed using one-way ANOVA (J, L). BTRT (M) and SKRT (N) cells were co-cultured with BCAFs (two lines) or with BCAFs and PBMCs (three lines), followed by treatment with trastuzumab (20 μg/ml) and/or AMD3100 (2.5μM) as illustrated in Fig S1. At the endpoint, cell viability was detected using CellTiter-Glo 3D viability assay kit and analyzed using one-way ANOVA. The data are reported as mean ± SD of triplicates, representing two independent experiments (*P < 0.01, **P < 0.001, ***P < 0.0001 compared with vehicle).