A) An ECM transcription network was identified by co-expression network analysis. Hsp47 (Serpinh1) is a hub of the ECM network. (B-D) Scatter plot of correlated mRNA levels between Hsp47 and (B) Col1A1, (C) Col4A1, and (D) FN1 in normal and malignant breast tissues (n=593). Expression data were obtained from the TCGA microarray dataset (Oncomine). (E) Box plots showed the mRNA levels of the ECM network genes in breast cancer tissues (n=532) and normal tissues (n=61), p<0.001). (F) Box plots showed the mRNA levels of Hsp47 in breast cancer tissues (n=532) and normal tissues (n=61), p<0.001). Data were obtained from the TCGA microarray dataset. (G) Western blot showed expression of Hsp47 in breast cancer cells (BT549, MDA-MB 231, T4-2, and SUM-140) and non-malignant cells (S1 and MCF10A). (H) Quantitative RT-PCR analyzing expression of Hsp47 gene in basal type of breast cancer (red, n=7) and luminal cancer cells (gray, n=8).

(A) T4-2 cells were infected with lentivirus containing control shRNA and Hsp47 shRNA. The knockdown efficiency was assessed by Western blot. (B) Phase and confocal images showed the morphology and α6-integrin staining (basal marker, green) of control and Hsp47-silenced T4-2 cells. (C) Colony size of the control and Hsp47-silenced T4-2 cells in 3D culture, n=50, ** p<0.01. (D, E) Control and Hsp47-silenced MDA-MB-231, BT549, and Hs-578T cells were cultured in 3D for 4 days. (D) Phase images showed the organoid morphology in 3D culture. (E) Quantitative data showed that the branching structures was significantly reduced in Hsp47-silenced MDA-Mb 231 cells, n=35, ** p<0.01. (F) Cell proliferation was examined by the 5-ethynyl-2′-deoxyuridine (EdU) labeling assay. Bar graph represents the ratio of EdU positive cells in control and Hsp47-silenced T4-2 and MDA-MB 231 cells, n=3; * p<0.05. (G) Cell invasion was assessed by the Transwell assay. Results indicate that silencing Hsp47 significantly inhibits cell invasion in T4-2 and MDA-MB 231 cells, n=4; * p<0.05, ** p<0.01. (H) Phase and confocal images showed the morphology and α6-integrin staining (basal marker, green) of the control and Hsp47-expressing S1 cells. The cells were cultured in 3D Matrigel™ for 4 days.

(A) The secreted and cellular levels of collagen I (Col I), IV (Col IV), fibronectin (FN) and were analyzed by Western blot in control and Hsp47-silenced T4-2 cells. (B) Confocal images showed the deposition of collagen IV and fibronectin in control and Hsp47-silenced T4-2 cells. (C) Expression of collagen IV and fibronectin (FN) in S1 and T4-2 cells were analyzed by Western blot. (D, E) Phase images (D) and quantification (E) of the colony size of control and Col4A1-silenced T4-2 cells in 3D culture, n=100; ** p<0.01. (F) Basal polarity in control and Col4A1-silenced T4-2 cells was assessed by α6-integrin staining; the percentage of polarized colonies was determined, n=40; ** p<0.01. (G) Bar graph showing the percentage of EdU positive cells in control and Col4A1-silenced T4-2 cells, n=3; ** p<0.01. (H, I) Phase images (H) and quantification (I) of the colony size of control and FN1-silenced T4-2 cells in 3D culture, n=100, ** p<0.01. (J) Phase images and quantification of invasive branch structures in control and FN1-silenced MDA-MB 231 cells. (K, L) Cell migration (K) and invasion (L) of Hsp47-silenced MDA-MB 231 cells were analyzed in the presence and absence of exogenous fibronectin (10 μg/ml) (FN), *p<0.05; ** p<0.01. (M) Phase images of invasive phenotypes of Hsp47-silenced MDA-MB 231 cells in 3D culture. Matrigel was mixed with fibronectin (12.5 μg/ml) (FN) or same volume of PBS.

(A) Co-expression network analysis showed that miR-29b and 29c were associated with the ECM transcription network. (B) Sequence alignment showed the miR-29 binding site in the 3′UTR region of Hsp47 gene (SerpinH1) (Upper panel). The target site of miR-29 in Hsp47 3′-UTR is highly conserved among mammalian species (Lower panel). (C) Hsp47 protein levels were assessed by Western blot in control and miR-29 mimic-transfected T4-2 cells. (D) Hsp47 protein levels were assessed by Western blot in control and miR-29 inhibitor-transfected MCF10A cells. (E) Quantitative RT-PCR analysis of the expression of Hsp47 and the ECM network genes in control and miR-29 mimic-transfected cells; n=4, ** p<0.01; * p<0.05. (F) Luciferase reporter analysis showed that miR-29b inhibited Hsp47 expression by targeting the seed complementary sequence in the 3′UTR region. pGL3-Hsp47-3′-UTR reporter plasmid in which the luciferase-coding sequence had been fused to the 3′-UTR of firefly luciferase was cotransfected into HEK293 cells with negative control (white columns) or miR-122 duplex (red columns). Hsp47-Mu indicates the introduction of alterations into the seed complementary sites; n=4, *p< 0.05. (G, H) Phase images (G) and quantification (H) of the colony size of the control and miR-29b-expressing T4-2 cells in 3D culture. (I, J) Phase images (I) and quantification (J) of the invasive branch structures in control and miR-29b-expressing MDA-MB 231 cells. (K, L) Spearman correlation analysis showed the negatively correlated expression of Hsp47 and miR-29b, 29c in human breast cancer tissues, n=97.

(A) Spearman correlation analysis showed the correlated expression of Hsp47 and TGFB3 in human breast cancer tissues, n=97. (B) Hsp47 expression was measured by quantitative RT-PCR in control and TGF-β treated S1 and MCF10A cells. The cells were treated with TGF-β (5ng/ml) for 48 hours, n=4, *, p<0.05. (C) Hsp47 expression was assessed by Western blot in control and SB431542treated MDA-MB 231 cells. Cells were treated with SB431542, a TGFBR inhibitor, for 48 hours. (D) MiR-29b expression in control and TGF-β treated MCF10A cells was analyzed by quantitative PCR analysis, n=4, ** p<0.01. (E) Quantitative PCR analysis of miR-29b expression in the control and SB431542 treated MDA-MB 231 cells, n=4, * p<0.1. (G) MDA-MB 231 cells were treated with SB431452 for 12 hours, and then transfected with miR-29b or 29c inhibitors. Hsp47 expression was assessed by Western blot. (H) Spearman correlation analysis showed a negative correlation between miR-29b and TGFB3 levels in human breast cancer tissues, n=97.

(A) Control or Hsp47-silenced T4-2/Luciferase cells were subcutaneously injected into SCID mice. IVIS images showed the tumor size in each group after 6 weeks. (B) The curves illustrated tumor growth of control and Hsp47-silenced T4-2 cells in SCID mice over time; n=6, * p<0.05. (C) Quantification of Ki67 staining, indicative of cell proliferation, in control and ShHsp47-expressing tumors; ** p<0.01. (D) Masson's trichrome staining of tumor sections (blue, collagen; black, nuclei; red, cytoplasm); scale bar, 200μm. The bar graph illustrated the collagen staining- area in tumors formed by control and Hsp47-silenced T4-2 cells. (E) Control or Hsp47-silenced MDA-MB 231/Luciferase cells were injected into the fat pads of SCID mice. IVIS images showed tumors in representative mice from each group. The bar graph showed that tumor volume formed by Hsp47-silenced cells was significantly reduced compared with control cells; n=6, ** p<0.01.

(A, B) Kaplan-Meier survival analysis showed the association of Hsp47 expression with the overall survival (A) and the recurrent-free survival in human breast cancer patients (B). The tumor samples were classified as low (n=137), high (n=138), and moderate (n=138) based on the mRNA levels of Hsp47 in the published microarray datasets. Significant differences in survival time were calculated using the Cox proportional hazard (log-rank) test. (C) Bar graph showing the mRNA levels of Hsp47 in different stages of breast cancer. Data were from Dr. Chin's microarray dataset. (D, E) Kaplan-Meier survival analysis showed the association of miR-29b (D) and 29c (E) expression with the overall survival in human breast cancer patients. The tumor samples were classified as low (n=33) and high (n=66) based on the mRNA levels of miR-29b and 29c in the published microarray datasets. Significant differences in survival time were calculated using the Cox proportional hazard (log-rank) test.