Fig 2: Stromal-derived IGFBP2 is sufficient to inhibit MM231 breast cancer cell invasion.(A and B) Representative images (A) and quantification (B) of MM231 breast cancer cell invasion in inverted collagen/fibronectin matrices in the presence of conditioned medium from HUVECs transfected with siRNAs against IGFBP2 (siIGFBP2_1 and siIGFBP2_2) or siNTC. Scale bars, 50 µm. n = 3 biological replicates performed in triplicate, with three stacksper transwell; one-way ANOVA with Tukey correction; **P < 0.01 and ***P < 0.001; ns, not significant. (C and D) Representative Western blot (C) and densitometry analysis (D) of TIFs stably overexpressing mTurquoise2 (mT2) or IGFBP2 (n = 4; one-sample t test; *P < 0.05). (E) Enzyme-linked immunosorbent assay (ELISA) for human IGFBP2 in concentrated conditioned media from TIFs overexpressing mT2 or IGFBP2, measured in duplicate (n = 3 biological replicates; two-tailed Student’s t test; **P < 0.01). (F) Representative images and quantification of fibroblast-contracted 3D collagen invasion assays of MM231 cancer cell invasion in mT2 or IGFBP2 TIF-contracted matrices after 14 days and stained for Pan-CK (brown). Scale bars, 100 µm. n = 3 biological replicates, triplicate matrices, eight regions per condition per replicate; one-way ANOVA with Tukey correction; ***P < 0.001. (G and H) Representative hematoxylin and eosin (H&E) images (G) and quantification (H) of MM231 cell invasion into the mouse dermis from subcutaneous xenografts of MM231 co-injected with TIFs overexpressing either mT2 (control) or IGFBP2. Scale bars, 500 µm; insets, 50 µm. n = 10 mice (mT2) or 12 mice (IGFBP2).

Fig 3: IGFBP2 expression during breast cancer progression.(A) Representative H&E-stained breast tissue samples from healthy patients, patients with DCIS, or patients with IDC. Scale bars, 200 µm. (B) Quantification of adipocytes per section from healthy (n = 8 patients, three to eight sections per patient), DCIS (n = 6 patients, one to four sections per patient), and IDC (n = 3 patients, one to two sections per patient). Kruskal-Wallis test using Dunn’s test to correct for multiple comparisons (**P < 0.01 and ***P < 0.001). (C) Representative image from a patient sample stained for IGFBP2 (magenta) and counterstained with DAPI (cyan) and keratin-8/keratin-14 (KRT8/14; red). Scale bar, 100 µm. n = 8 normal reduction mammoplasty patient samples. Autofluorescence is given in green. (D) Representative image of human DCIS breast cancer tissue stained for IGFBP2 (magenta) and counterstained with DAPI (cyan) and KRT8/14 (red). Scale bar, 100 µm (n = 4 DCIS patient samples). Autofluorescence is given in green. (E) Representative image of human IDC breast cancer tissue stained for IGFBP2 (magenta) and counterstained with DAPI (cyan) and KRT8/14 (red; n = 3 IDC patient samples). Scale bar, 100 µm. Autofluorescence is given in green. The brightness of IGFBP2 staining was increased for display purposes only. (F) Quantification of IGFBP2 per adipocyte from healthy (n = 8 patients, 143 to 280 adipocytes per patient), DCIS (n = 6 patients, 39 to 410 adipocytes per patient), and IDC (n = 3 patients, 75 to 283 adipocytes per patient). Kruskal-Wallis test using Dunn’s test to correct for multiple comparisons (***P < 0.001).

Fig 4: Breast adipocytes secrete the anti-invasive factor IGFBP2.(A) Three representative images [(i) to (iii)] of human breast tissue stained for IGFBP2 (magenta) and counterstained with 4',6-diamidino-2-phenylindole (DAPI) (cyan). Autofluorescence (AF) signal is green. Scale bars, 100 µm; insets, 10 µm. n = 2 adjacent healthy breast tissue sections from ductal carcinoma in situ (DCIS) patient samples. (B) Representative images of primary human breast pre- and mature adipocytes with immunofluorescence staining for IGFBP2/PPAR?/actin or Nile Red/vimentin. Scale bars, 50 µm; insets, 10 µm. n = 4 normal reduction mammoplasty patient samples from which preadipocytes were isolated, cultured, and differentiated into mature adipocytes. (C) Fold change in gene expression between pre- and mature adipocytes, normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and detected by quantitative reverse transcription polymerase chan reaction [adipocytes from n = 4 patient samples, differentiated as in (B), in triplicate; one-sample t test; **P < 0.01 and ***P < 0.001]. PPARG, peroxisome proliferator activated receptor gamma; LIPE, lipase E; FABP4, fatty acid binding protein 4; CFD, complement factor D. (D) ELISA for human IGFBP2 from pre- and mature adipocyte conditioned media, compared to the adipocyte culture media [AM-1; adipocytes from n = 4 patient samples processed as in (B) and media collected; two-tailed Student’s t test with Welch’s correction; ***P < 0.001]. (E and F) Representative images (E) and quantification (F) of MM231 cell invasion in inverted collagen/fibronectin matrices in the presence of concentrated conditioned media from adipocytes or adipocyte growth medium, AM-1, or full culture medium with an equivalent volume of PBS added. Scale bars, 50 µm. adipocytes from n = 4 patient samples, processed as in (B) and media collected, performed in triplicate with three stacks per transwell; one-way ANOVA with Tukey correction; **P < 0.01. (G and H) Representative images (G) and quantification (H) of a fibroblast-contracted 3D collagen matrix cancer cell invasion assay monitoring MM231 cancer cell invasion ± mature adipocyte coculture and stained for either Pan-CK or the proliferation marker Ki67. Scale bars, 100 µm. n = 3 biological replicates, triplicate matrices, eight regions per condition per replicate; one-way ANOVA with Tukey correction; ***P < 0.001).

Fig 5: IGFBP2 acts through depletion of IGF-II from the cancer microenvironment.(A) Schematic of the proteomics experimental setup. LC-MS/MS, liquid chromatography–tandem mass spectrometry. IP, immunoprecipitation. (B and C) Representative Western blot (B) and densitometry analysis (C) after silencing of IGF2 (IGF-II gene) using siRNAs in MM231 cells (n = 4 biological replicates; one-sample t test; *P < 0.05 and ***P < 0.001). (D and E) Representative images (D) and quantification (E) of MM231 cell invasion in inverted collagen/fibronectin matrices after IGF2 silencing. Scale bars, 50 µm. n = 3 biological replicates, performed in duplicate with three stacks per transwell; one-way ANOVA with Tukey correction; ***P < 0.001. (F and G) Representative images (F) and quantification (G) of MM231 cell invasion in inverted collagen/fibronectin matrices treated with PBS, IGF-II (10 ng/ml), IgG1? (10 µg/ml), or anti–IGF-II (10 µg/ml). Scale bars, 50 µm. n = 3 biological replicates, performed in duplicate with three stacks per transwell; one-way ANOVA with Tukey correction; ***P < 0.001. (H and I) Representative IGF-II Western blot (H) and quantification (I) of TIF, MM231, MM468, and HCC1937 cells (n = 6 biological replicates; one-sample t test; ***P < 0.001). (J) ELISA for human IGF-II in conditioned media from TIF, MM231, MM468, and HCC1937 cells (n = 4 biological replicates; one-way ANOVA with Dunnett’s correction; *P < 0.05 and ***P < 0.001). (K and L) Matrigel invasion assays for MM231 (K) and MM468 (L) cells treated with IgG1? or anti–IGF-II (n = 3, eight fields of view (FOVs) per chamber, two to three invasion chambers per condition per replicate; two-tailed Student’s t test with Welch’s correction; **P < 0.01). Scale bars, 100 µm. (M) Schematic of the proposed mechanism for IGFBP2 inhibition of invasion through disruption of breast cancer IGF-II autocrine signaling.