(A) Hazard ratios of SRS status and ER status on late-onset bone metastasis when these parameters are fit into a bivariate Cox proportional hazard regression model. Various cutoffs from 0 year to 5 years were used to define “late-onset” bone metastasis. Error bars, standard error as determined by the statistical model. p values assess the significance of the hazard ratio’s difference from 1 (no prognostic value). * p < 0.05; ** p < 0.001; *** p < 10^-5. Error bars indicate ±SEM.
(B) Kaplan-Meier representation of bone metastasis-free survival. Tumors are categorized according to both ER status and SRS status. p values based log rank tests.
(C) Histograms of bone metastasis onsets in the indicated categories of tumors.

(A) Growth of the indicated cell lines after direct implantation into the marrow of tibia. Quantitative bioluminescence was done at day 35. Averages ±SEM (n=10 per group).
(B) Representative bioluminescence imaging and H&E staining of bone lesions of each experimental group in (A) at day 35.
(C) BoM-1833 cells were injected into left cardiac ventricle of mice, and animals were treated with vehicle control (Mock) or dasatinib (10 mg/kg) daily starting on days 0, 7 or 14 after inoculation. The plot shows quantitative bioluminescence values of the hind limb region at day 35. Data are averages ±SEM (n=10 per group).
(D) Vectors encoding wild type c-Src or dasatinib-resistant mutant c-Src were used to rescue the expression of c-Src in BoM-1833 cells that also expressed c-Src shRNA. On day 7 after intraventricular inoculation, animals were treated with vehicle control or dasatinib. Day 35 average bioluminescence ±SEM (n=10 per group).
(E) Size-matched femoral metastases from mice inoculated with control, c-Src knockdown, or c-Src-rescued BoM-1833 cells, or with BoM-1833 cells and treated with dasatinib (starting on day 7), were extracted on Day 35 (control and c-Src-rescued groups) or Day 56 (dasatinib and Src RNAi groups). Samples were subjected to TUNEL staining. Four or more randomly picked fields were quantified and the percent of TUNEL-positive cells ±SEM is plotted
(F) The samples in (E) were subjected to staining and quantification of Ki-67 proliferation marker. n.s., not significant. Data are averages ±SEM.
(G) The samples in (E) were subjected to TRAP staining in order to identify presumptive osteoclasts at the tumor-bone matrix interface. Arbitrary units were assigned to represent the proportion of red TRAP staining-cells. n.s., not significant. Data are averages ±SEM.
(H) Representative TRAP staining. Arrows indicate positive staining.

(A) Microarray gene expression analysis of 58 human breast cancer metastasis samples revealed 17 cytokines whose expression was up-regulated in a majority of bone metastases but not in a majority of metastases to lung, liver or brain.
(B) Cytokine expression in metastases from mice inoculated with MDA231 cells that are metastatic to either bones, brain or lungs. Lesions were extracted from mice on day 28 and qRT-PCR was performed for mouse genes encoding Cxcl12, Trail, Igf1, Pegfa, Vegfc and Tgfb1. Expression levels relative to the mean of the values in bone metastasis are shown for each sample. Data are averages ± SEM.

(A) Kaplan-Meier representation of the probability of cumulative overall metastasis-free survival in 615 breast cancer cases according to the estrogen receptor α (ER) status. The numbers of tumors in each category, the total metastasis events, the late-onset (> 5 yrs) metastasis events and the corresponding p values (log rank test) are shown in the embedded table.
(B) Same as in (A), but bone metastasis-free survival.
(C) Same as in (B), but cases are categorized according to their SRS status.
(D) Hierarchical clustering of 615 primary tumors with known bone metastasis outcomes, by ER status, SRS status, and molecular. Red marks above the heatmap indicate tumors that do develop bone metastasis.

(A) Metastasis organ-tropism and c-Src gene signature activity of in vivo-selected MDA231 derivatives cell lines. The blue-white-red shows a single value score of SRS activity determined by principal component analysis of the expression of the 159 genes that constitute the SRS. Cell lines were ordered by unsupervised hierarchical clustering with the SRS genes (refer to Figure S2).
(B) Representative MDA231 derivatives were subjected to western immunoblotting with the indicated antibodies. Organ-tropism: 1833, bone; 831, brain; 4175, lung; and 1834, adrenal.
(C) Knockdown of c-Src in BoM-1833 as confirmed by western immunoblotting analysis.
(D) Survival of mice after intracardiac injection (3×10⁴ cells) with BoM-1833 transduced with a control vector (Control), a c-Src shRNA vector (Src RNAi), or c-Src shRNA and shRNA-resistant c-Src expression vectors (Rescued). (n=15-20 per group)
(E) Bioluminescent, radiographic and H&E analysis of bone lesions from representative mice in each group, at the indicated times after inoculation. In the X-ray images, areas of bone lysis are indicated by dotted lines. In the H&E staining, asterisks indicate tumor.
(F) Normalized bioluminescence signal of bone metastases in the hind limbs of mice inoculated with the indicated cell lines. The signal intensities were normalized to Day 0, which was set arbitrarily as 100. Data are averages ±SEM. p value calculated using Student’s t-test with Welch correction.
(G) Left panel: the indicated cell lines (5×10⁵ cells) were injected into the cleared fourth mammary fat pad of mice. Tumor sizes were measured at day 35 (n=7-10 in each group). Right panel: Quantification of lung metastasis burden originated from the orthotopic mammary tumors. Data are averages ±SEM.
(H) LM2-4175 cells (2×10⁵ cells) expressing a control vector (LM2 control) or a c-Src shRNA vector control (LM2-Src RNAi) were injected into the tail vein of mice. Lung colonization was assayed by weekly bioluminescence imaging. Plots show normalized photon flux in the lung over time (n=5 per group). Representative H&E stained lungs 5 weeks after xenografting are shown. c-Src nockdown was confirmed by western immunoblotting. Data are averages ±SEM.

(A) c-Src and activated (Y416 phosphorylated) c-Src protein levels in parental, bone-tropic (BoM2) and brain-tropic derivatives (BrM2c) of CN34 cells, as determined by western immunoblotting of cell lysates.
(B) Normalized bioluminescence signal intensity at the hind limbs of mice that were intracardially inoculated with the indicated cell lines (1×10⁵ cells). Data are averages ±SEM (n=8-10 per group).
(C) Normalized bioluminescence signal intensity at the upper back region of the same mice, to capture lymph node metastasis signal. Data are averages ±SEM.
(D) Schematic of assay to determine the survival of breast cancer cells in the bone marrow.
(E) Samples (5×10⁴ cells) were intracardially injected in 7-week old mice. Surviving tumor cells were extracted and grown as shown in panel D. Images show immunofluorescence staining of representative colonies with the indicated antibodies. Scale bar = 50μm.
(F) Quantification of latency-derived human breast cancer cell colonies. Data are averages ±SEM (n=3 per group).

(A) The indicated cell lines were incubated with or without CXCL12 for 30 min. Immunoblots using the indicated antibodies were performed on whole-cell extracts. P-AKT, phosphorylated (activated) AKT.
(B) TUNEL assays were performed 3 days after culturing the indicated cell lines in serum-free medium supplemented with CXCL12 or TRAIL. Averages ±SEM (n=6)
(C) Cell viability assays were performed by culturing BoM-1833 cells in serum-free medium supplemented with CXCL12 (for 5 days) or TRAIL (for 3 days). Data are averages ±SEM (n=6).
(D) Immunoblots using the indicated antibodies were performed on whole-cell extracts from the indicated BoM-1833 derivatives after incubation with or without TRAIL in the media.
(E) Upper panel: Schematic representation of the role of c-Src in the survival of breast cancer cells that infiltrate the bone marrow. Breast tumors that disseminate c-Src-activated cancer cells have an advantage for long-term survival in the bone marrow microenvironment. c-Src influences the responsiveness of breast cancer cells to specific bone metastasis microenvironment factors, CXCL12 and TRAIL. CXCL12 (a.k.a. SDF1) binding to its receptor CXCR4 triggers AKT activation, and we show that c-Src is required for this activation and its associated pro-survival effects. c-Src activity is also required for the resistance of breast cancer cells to the cell-death effect of TRAIL. Lower panel: schematic representation of the course of breast cancer metastasis. After disseminated from the primary tumors, cancer cells may infiltrate different organs. Disseminated cancer cells may survive in the form of latent disease for decades before eventually gaining competence to outgrow and colonize the host tissue through the production of the osteoclastogenic factors.