In situs were performed as described in supplementary materials photographed at 100 X magnification (except panel D). (A) Demonstrates labeling with an anti-sense probe of an ER negative breast cancer. (B) Demonstrates labeling with a sense probe of the same tumor as in A. (C) Demonstrates labeling of an ER+ tumor at a junction between normal breast tissue in the top half of the panel and less-intensely labeling tumor in the bottom half of the photograph. (D) Demonstrates a higher power magnification (400X) picture of the normal appearing breast tissue shown in C to demonstrate MFG-E8 labeling of ductal tissue throughout all layers. (E) Demonstrates an ER negative tumor scored as 3+. Photographic exposure times for this panel were used in subsequent panels F-L to allow comparison of the relative levels of expression in all other samples. (F) Demonstrates the most intensely staining ER+ tumor scored as 3+. Only this single ER+ tumor fell in the same range of staining as the ER negative group of tumors. (G) An ER negative tumor staining 3+. (H) The second most intensely staining ER+ tumor scored as 1+. (I) Shows the median most intensely staining ER negative tumor scored as 3+. (J) Shows median most intensely staining ER+ tumor scored as 0 staining. (K) Shows the least intensely staining ER negative tumor scored as 2+. (L) Shows an ER+ tumor scored as 0 staining. (M). In situ staining was scored by an investigator (D.S.) who was blinded to the ER status of the tumors and had not performed the actual in situ staining. Shown is a graph of the fraction of the tumors staining with various intensities, comparing the ER+ and ER- samples. The actual number of tumors staining at each value for each type is shown within the columns of the graph. In situ staining was scored as 0 (no staining), 1+ (weak positive), 2+ (moderate positive), and 3+ (strong positive). Comparing specimens staining 2+ or 3+ to those staining 0 or 1+, one ER+ tumor was among the top 13 MFG-E8 expressing tumors, and one ER negative tumors was among the bottom 12 expressing samples. MFG-E8 expression was highly correlated with the ER+ status of breast cancers across all staining intensities (p < 0.001 by Chi square analysis).

(A) p63 was conditionally knocked down in mouse mammary myoepithelial cells using K14CreERTam/p63^flox mice. Shown are the resulting decreases in p63 and MFG-E8 mRNA (normalized to GAPDH) in mouse mammary glands 5 and 9 weeks after knockdown. MFG-E8 changes are significant (p = 2.99 × 10^-4 at 5 weeks and p = 6.42 ×10^-4 at 9 weeks.) (B) Demonstrates decreased MFG-E8 mRNA in cells expressing a shp63 lentiviral construct. MDA-MB-468 cells were treated as described (15). Protein was harvested and analyzed by Western analyses 72 hours after infection. Tested cells included untreated controls (-), and cells transduced with shGFP (GFP) and shp63 (p63). An actin loading control is shown. Total mRNA was analyzed from the shGFP and shp63 transduced cells. Shown is the mean and standard error for four independent samples comparing the fold change in the shp63-expressing cells to the shGFP-expressing cells. (C) We performed ChIP-PCR for p63 binding as described in supplementary materials. anti-p63 ChIP-PCR signals are plotted as a percentage of the total input DNA for each reaction. We measured α-p63 ChIP DNA in triple negative MDA-MB468 and HCC1937 cells. Shown are the mean and standard deviation for signals at the actin promoter (negative control) versus reported p63 binding sites. Binding at the PUMA promoter was used as positive control (not shown).

(A) T47D cells were transfected with a siRNA oligonucleotide for MFG-E8 (MFG) or a scrambled control siRNA (Scr), and MFG-E8 and actin levels were analyzed 48 hours later using standard immunoblots to validate the siRNA. (B) Three breast cancer cell lines were transfected with scramble and MFG-E8 siRNAs. They were grown for 72 hours in the absence and presence of the RGB peptide that competitively inhibits ligand binding to α and β integrins and harvested for the MTT assay. The fold difference in proliferation was obtained by dividing the MTT measurement in the MFG-E8 siRNA samples by the MTT measurement in the scrambled oligonucleotide samples. Plotted are the means and standard deviations for each cell line. (C and D) We crossed MFG-E8 null mice with homozygous MMTV-erbB2 transgenic mice that express the non-mutated form of erbB2 in their mammary tissues. Resulting MFG-E8^+/- mice were then back-crossed to the MFG-E8^-/- mice and mice of four genotypes were identified using PCR-based genotyping: non-transgenic-MFG-E8^+/-, erbB2 transgenic-MFG-E8^+/-, non-transgenic-MFG-E8^-/-, and erbB2 transgenic-MFG-E8^-/-. Mice were evaluated weekly for palpable masses and scored as positive upon appearance of a tumor. Shown are photomicrographs of standard hematoxylin and eosin stained tumors that arose in compound erbB2/MFG-E8^-/- transgenic mice before fifteen months of age (40X). The tumors are high grade invasive ductal carcinomas that display a predominant solid pattern with focal gland formation. (E) At 15 months of age all remaining mice were sacrificed and mammary gland whole mounts were prepared. Tumors identified either in the whole mounts or in the mice that developed overt palpable tumors were scored as positive. We plotted the incidence of tumor formation for each genotype of mice and indicate the tumor number/number of mice evaluated within the bars of the graph. (p = 0.006 by Fisher's exact test)

(A-C) Shown are box and whisker charts for MFG-E8 mRNA using LCM-isolated tissues from human pathologic specimens where normal, ductal carcinoma in situ (DCIS) and invasive ductal carcinoma components could be isolated from the same patient. The y axis measures MFG-E8 levels in arbitrary units based on a standard curve of MFG-E8 PCR product used in each qRT-PCR run. The patient cohort was originally described in Ma et al (20). The median value is plotted as the junction of the shaded and empty rectangles. The mean value is plotted as the diamond symbol. (A) Demonstrates levels of expression of MFG-E8 comparing all available specimens. (B) Demonstrates MFG-E8 expression in the ER+ patients. (C) Demonstrates MFG-E8 expression in patients where a two-fold or greater increase in cyclin D1 was identified in the same specimen. (D) Shown are box and whisker charts for fold change in MFG-E8 mRNA using LCM-isolated tissues from human pathologic specimens where normal, ductal carcinoma in situ (DCIS) and invasive ductal carcinoma components could be isolated from the same patient whose tumor was erbB2+ (N = 10). (E) mRNAs from patients with sporadic basal breast cancers included in a previous study that described their X chromosomal abnormalities were generously provided by the Harvard Breast Cancer SPORE tissue bank. The details of the patients evaluated are described in Richardson et al (13). We measured MFG-E8 mRNA levels, which are plotted in box and whisker charts as in A-D. (F) A commercial MFG-E8 ELISA was used to measure MFG-E8 protein in patient sera. Shown are MFG-E8 protein levels in serum from patients with tumors of the indicated phenotypes plotted using box and whisker plots. N = 10 for each tumor phenotype. Differences between ER+ versus controls were not significant. Differences between control and erbB2 (p = 0.027) as well as triple negative cancers (p = 0.019) were significant by Mann-Whitney Rank Sum test. (The data were not normally distributed.)

(A-C) p63 expression changes differ among breast cancer subtypes. Shown are box and whisker charts for the TAp63 isoform mRNA using the tissues described in Figure 1. (Total p63 mRNA measurements gave similar results.) The y axis measures p63 levels in arbitrary units based on a standard curve of 63 PCR product. (A) Demonstrates levels of expression of p63 comparing ER+ specimens. (B) Demonstrates p63 expression in the erbB2+ patients. (C) Demonstrates differences in p63 expression between ER+/erbB2+ patients versus patients with basal tumors.

A diagram of luciferase reporter constructs containing different proportions of the MFG-E8 promoter is shown. Grey ovals indicate p63 binding sites. The reporter constructs were co-transfected with expression constructs containing p63 isoforms in combination with pCMV-Renilla luciferase. The ratio of firefly luciferase values in the presence and absence of the p63 expression constructs (normalized to the internal Renilla standard) were calculated for each condition. Shown are the means and standard errors for three replicas at each point. (A) Shown are the results for the four different p63 isoforms in T47D cells. (B) Shown are the results for the four different p63 isoforms in MDA-MB468 cells.

(A) Shown are MFG-E8 immunoblots for a variety of breast cancer cell lines whose ER, PR and erbB2 status are known. Cell lysates were harvested and Western blots were performed for MFG-E8. A commercial MFG-E8 ELISA was then used to measure MFG-E8 protein in the supernatants of media from the same cell lines. Supernatant media were harvested after 24 hours of incubation and ELISAs were performed according to kit instructions. The mean and standard error are shown for three replicas harvested from each cell line, which are plotted as ng/ml of MFG-E8 in the conditioned media. (B) An MFG-E8 immunoblot of cells transfected using scramble (Sc) and siMFGE8 (MFG) shows effective MFG-E8 knockdown. An actin control is shown. (C and D) Dose-response curves (MTT cell viability assay) of cells expressing the scramble (sc) or siMFGE8 (MFG) after treatment with cisplatin at the indicated concentrations (μM). Error bars show SD for ten experiments. Shown are MDA-MB-468 (C) and HCC-1937 cells (D). p < 0.01 for each point on the curves.