a, 5 × 10⁵ LM2 cells expressing individual miRNAs or the control hairpin, as well as the parental MDA-MB-231 cells, were injected into the mammary fat pads of immunodeficient mice and tumour volumes were measured over time. n = 5 (MDA-MB-231, LM2/miR-126, LM2/miR-206) and n = 10 (LM2, LM2/miR-335); error bars indicate s.e.m.; P-values based on a one-sided Student's t-test at day 32. b, 5 × 10⁴ LM2 or primary breast cancer line CN34-BoM1 expressing miR-126, miR-335 or miR-206 and control cells were seeded in triplicate and viable cells were counted at 5 days after seeding. n = 3; error bars represent s.e.m.; P-values obtained using a one-sided Student's t-test. c, Parental MDA-MB-231 cells and lung metastatic LM2 cells expressing miR-126, miR-335 or miR-206 were seeded onto glass slides. Cells were stained with the actin marker phalloidin and confocal images were obtained. d, 2.5 × 10⁴ LM2 and CN34-BoM1 cells were transduced with the indicated miRNAs or a control hairpin, and trans-well migration was assessed. Images of cells that had migrated through trans-well inserts were obtained and analysed in automated fashion using Metamorph software. n = 3; error bars represent s.e.m.; P-values obtained using a one-sided Student's t-test.

a, The miR-335 metastasis signature consists of genes downregulated by miR-335 and are also overexpressed in bone as well as lung metastatic MDA-MB-231 derivatives. The heatmap depicts the variance-normalized expression values for each gene averaged across two MDA-MB-231 samples, four LM2 samples and two LM2/miR-335 samples. The scale bar depicts standard deviation change from the mean for the expression value of each gene. b, UTR reporter assays of miR-335 metastasis genes in LM2 and MDA-MB-231 cells. Reporter constructs consisting of the luciferase sequence fused to the 3′ UTRs of the miR-335 metastasis genes as well as the control gene UBE2F were transfected into the LM2 and parental MDA-MB-231 cell lines. Luciferase activity of cells was assayed at 32 h after transfection and the values were normalized to the LM2 cell line. n = 3; error bars represent s.e.m.; P-values were obtained using a one-sided Student's t-test. c, Luciferase activity of MDA-MB-231 cells co-transfected with 39 UTR reporter constructs with, or without, miR-335 antagomir was assayed at 32 h after transfection and normalized to control cells. n = 3; error bars represent s.e.m.; P-values obtained using a one-sided Student's t-test. d, Schematic diagram depicts seed sequence (red box) in SOX4 UTR. The 60 bp containing this sequence or the miR-335 seed target site mutant were subjected to UTR reporter assays in LM2 cells with restored miR-335 expression. Luciferase activity of cells was also assayed in the presence or absence of the miR-335 antagomir. n = 3; error bars represent s.e.m.; P-values obtained using a one-sided Student's t-test. e, SOX4 expression in LM2 and MDA-MB-231 cells was obtained through real time qRT-PCR in cells expressing miR-335 or transfected with the miR-335 antagomir. n = 3; error bars represent s.e.m.; P-values derived using a one-sided Student's t-test.

a, Hierarchical clustering of normalized miRNA expression levels for the 20 miRNAs that displayed the highest coefficients of variation clusters the MDA-MB-231 derivatives into three groups comprised of the bone metastatic BM2 lines (2287 and 1833), the lung metastatic LM2 lines (4180, 4173, 4175, and 4142) and the parental MDA lines. The heatmap highlights a set of miRNAs whose expression decreases across all lung and bone metastatic derivatives (red bracket along the right). The scale bar across the bottom depicts standard deviation change from the mean. b, Bioluminescence imaging of lung metastasis by lung metastatic breast cancer cells with restored expression of specific miRNAs. 1 × 10⁴ LM2 cells expressing individual miRNAs or the control hairpin, as well as the parental MDA-MB-231 cells, were inoculated intravenously into immunodeficient mice. Shown are representative mice corresponding to the LM2 set (top) and the MDA-MB-231 set (bottom) at day 50. Lung colonization was measured by bioluminescence and quantified. n = 5; error bars represent s.e.m.; asterisk, P < 0.05. c, Human-vimentin-stained images of representative lungs that emitted the median luciferase signal for each cohort. Lungs were extracted at 8 weeks after xenografting and representative images of sections were obtained at whole-field magnification. d,2 × 10⁴ BM2 cells expressing individual miRNAs or the control hairpin were inoculated into the arterial circulation via intracardiac injection of athymic mice. Bone metastasis was measured by bioluminescence and quantified as the normalized hindlimb photon flux. Shown are representative mice corresponding to marked data points. n = 6-10; horizontal line represents median signal for each cohort; P-values based on a one-tailed rank-sum test. e, Representative whole-field magnification images of haematoxylin- and eosin-stained femurs extracted from representative mice 7 weeks after intraventricular injection of indicated cancer cells. f, 2 × 10⁵ primary human cancer derivative CN34-LM1 cells expressing individual miRNAs or the control vector were inoculated intravenously into immunodeficient mice. Lung colonization was measured by bioluminescence, quantified and normalized. n = 8; error bars represent s.e.m.; P-values based on a one-sided rank-sum test. g, 2 × 10⁵ primary human cancer derivative CN34-BoM1 cells expressing individual miRNAs or the control vector were inoculated into the arterial circulation via intracardiac injection of immunodeficient mice. Bone metastasis was measured by bioluminescence, quantified and normalized. n = 9-10; error bars represent s.e.m.; P-values based on a one-tailed rank-sum test.

a, miR-335, miR-126, miR-122a and miR-199a expression was assessed in a set of 20 primary breast tumour samples through qRT-PCR. Kaplan-Meier curves depict metastasis-free survival of patients whose primary tumours contained low or high levels of the indicated miRNAs. P-values were obtained using a log-rank test. b, The parental MDA-MB-231 cells were transfected with antagomirs targeting endogenous miR-335, miR-199a* or a control antagomir. Four days after transfection, 1 × 10⁴ LM2 cells from each cohort were inoculated intravenously into immunodeficient mice. Lung colonization was measured by bioluminescence at day 35 and quantified. n = 5; error bars represent s.e.m.; P-values based on a one-sided rank-sum test.

a, 5.0 × 10⁴ LM2 cells were transduced with short hairpin control vector, either of two shRNAs targeting SOX4, a hairpin targeting TNC, or an siRNA targeting TNC, and invasion of cells through a trans-well matrigel-coated membrane insert was quantified. n = 6; error bars represent s.e.m.; P-values based on a one-sided Student's t-test. b, 2 × 10⁵ LM2 cells transduced with either a control short hairpin, a short hairpin targeting SOX4 or one targeting TNC were inoculated intravenously into immunodeficient mice. Lung colonization was measured by bioluminescence and quantified. n = 7; error bars represent s.e.m.; P-values based on a one-tailed rank-sum test. c, Haematoxylin-and-eosin-stained images of representative lungs that were extracted at 4 weeks after cell inoculation. Representative images were obtained at 10× magnification. d, Kaplan-Meier curves for the combined Memorial Sloan Kettering and Erasmus Medical Center breast tumour cohorts (368 tumours) depicting metastasis-free survival of patients whose primary tumours expressed the miR-335 six-gene signature (positive) and those that did not (negative). n = 368; P-value based on the Mantel-Cox log-rank test.