The Integrin α_vβ_3-5 Ligand MFG-E8 Is a p63/p73 Target Gene in Triple-Negative Breast Cancers but Exhibits Suppressive Functions in ER⁺ and erbB2⁺ Breast Cancers

Tumor progression results from accumulation of genetic changes that permit autonomous growth of malignant cells (1). While evaluating the genetics of tumor progression, we identified downregulation of milk fat globule—EGF8 (MFG-E8) mRNA in a microarray analysis of invasive murine tumors (2), but we were puzzled by its original description as a breast cancer antigen (3). A monoclonal antibody raised against human mammary epithelial cells was originally used to demonstrate elevated circulating levels of a 46-kD protein (BA46) in patients with metastatic breast tumors (4). BA46 radioimmune assays accurately monitored tumor burden and the α-BA46 antibody slowed tumor growth in xenotransplantation studies (5, 6). However, cDNA cloning of BA46 revealed that it was the normal breast protein milk fat globule factor 8 (MFG-E8)/lactadherin (7, 8). Importantly, MFG-E8 is the ligand for α_vβ_3-5 integrins (8), which mediates apoptotic cell phagocytosis (9). Likewise, homozygous MFG-E8 loss impairs two mammary developmental stages; its loss blocks both branching morphogenesis (10) and clearance of apoptotic cells during involution (11). Finally, loss of integrins β3, β5, or β3β5 accelerates MMTV–erbB2 tumor formation (12).

In contrast, microarray studies have shown that MFG-E8 mRNA increases in a diagnostic gene cluster in basal breast cancers (13, 14). p63 gene expression is generally restricted to the basal myoepithelial cell layer of mammary glands and p63/p73 regulation plays a role in the biology of tumors arising from these cells (15). Recent developments in our understanding of BRCA1's functions have suggested new therapeutic strategies incorporating platinum chemotherapy and poly(ADP-ribose) polymerase (PARP) inhibitors for triple-negative tumors, which encompass the basal subtype distinguished by its unique gene signature (16). The use of cisplatin as a targeted therapy is based on findings that BRCA1 defective cells are particularly susceptible to its effects (17). Studies of the p53/p63/p73 protein network provided additional support for this approach. Importantly, both p63 and p73 control of the p53 apoptosis program is a necessary and sufficient contributor to the effects of cisplatin. Additionally, a recent publication identified consensus p53/p63/p73 binding sites in the MFGE8 promoter, which control transcriptional responses to p63/p73 in skin (18). Here, we show that MFG-E8 is a potential biomarker for triple-negative breast cancers due to its upregulation by p63/p73, which contrasts with its downregulation in ER⁺ and erbB2⁺ breast cancers.

MFG-E8 null mice obtained from Barry Shur (Emory University; ref. 19) were crossed with nonmutant MMTV-erbB2 mice (Jackson Labs). Tumor incidence was evaluated by routine histology or by counting masses in mammary gland whole mounts at 15 months of age. Conditional p63 knockdowns using p63^flox mice are described in Supplementary Materials. Animal experiments followed approved standards of the MGH Animal Advisory Committee.

The patient cohort used for laser capture microdissection (LCM) specimens was previously described (N = 36; ref. 20). Additional erbB2⁺ patients were added (total N = 10). Tumor samples were used to compare MFG-E8 and p63 mRNA levels in the basal versus other tumors that were described in Richardson et al. (13). Unpaired 2-tailed Student's t tests were used to determine statistical significance of all data unless otherwise noted.

In situ hybridization procedures are described in the Supplementary Methods. MFG-E8 expression was scored by an investigator (D.S.) who was blinded to the estrogen receptor (ER) status of the tumors and who had not performed the in situ staining. In situ staining was scored as 0 (no staining), + (weak positive), ++ (moderate positive), and +++ (strong positive).

MFG-E8 was measured in stage I–III sera obtained prior to treatment from 10 healthy donors, 10 ER and/or PR⁺ patients, 10 erbB2⁺ patients, and 10 triple-negative patients. Sera were obtained from the Dana-Farber Cancer Institute (DFCI) with support from the NCI Breast SPORE program. Written consent was obtained from all subjects under institutional review board approval. MFG-E8 protein was measured using a commercial ELISA (Cusabio Biotech Co.; CSB-E12637h).

T47D, MCF7, ZR-75-1, BT474, SKBR3, HCC 1937, MDA-MB468, MB231, HCC1143, BT20, and HS578T cells were grown in conditions indicated by the supplier (ATCC). Secreted MFG-E8 was measured in culture supernatants 24 hours after changing media by using the commercial ELISA.

A published lentiviral small inhibitory RNA for p63 (shp63; ref. 15) was transduced, and total RNA and protein were harvested 3 days later for MFG-E8 regulatory studies (15).

To study antiproliferative effects of MFG-E8, MFG-E8 (Ambion; ID numbers 1436, 1531, and 1621) and scramble (Dharmacon D-1205-20) siRNAs were transfected using Oligofectamine (Invitrogen). Ten wells of 96-well plates were seeded at 6,000 cells per well for each treatment condition for MTT assays performed 72 hours after transfection. Two hours after siRNA transfection, RGD blocking peptide (7) or its control peptide from ENZO Life Science was added to the cell culture medium to a final concentration of 2 μg/mL. For cisplatin sensitivity the same siRNAs were transfected, cells were harvested 48 hours after transfection in varying doses of cisplatin (Sigma-Aldrich), and MTT assays were performed as previously described (15).

Antibodies used for standard Western blots included anti-MFG-E8 MAB2767 (R&D Systems), anti-p63 (H-129; Santa Cruz Biotechnology Inc.), and anti-actin MAB1501R (Chemicon).

We performed p63 chromatin immunoprecipitation (ChIP) experiments as detailed in Supplementary Materials. MFG-E8 reporter plasmids were provided by Dr. I. Katoh (Ikawa Laboratory, RIKEN, Wako, Japan; ref. 18) and luciferase reporter assays are detailed in Supplementary Materials.

Comparing mRNA expression changes between invasive tumors and preinvasive mammary tissues, we initially identified MFG-E8 as a uniformly downregulated gene during tumor progression in murine transgenic erbB2–, ras–, and cyclin D1–induced tumors (Supplementary Fig. S1A; ref. 2). An analysis of available SAGE expression data suggested that similar changes also occur in human cancers (Supplementary Fig. S1B). We therefore evaluated MFG-E8 expression changes in LCM-isolated human specimens where the patient's own normal and neoplastic tissues could be directly compared (20; ref. Fig. 1). We found that MFG-E8 mRNA decreased 3.3-fold from normal to ductal carcinoma in situ (DCIS) and from normal to invasive ductal carcinoma (IDC) in a mixed breast cancer population (Fig. 1A). This result contrasted with the original description of MFG-E8 as the breast cancer antigen BA46 so we next considered whether MFG-E8 changes might differ among different breast cancer subtypes.

We first found that MFG-E8 decreased 3.9-fold in ER⁺ DCIS and 4.8-fold in ER⁺ IDC samples (Fig. 1B). This was especially true for cyclin D1⁺ tumors (Fig. 1C). A similar decrease in MFG-E8 expression was found in LCM-isolated erbB2⁺ tumors during tumor progression from normal to DCIS to IDC (Fig. 1D). Published microarray data suggested that basal and BRCA-mutated breast cancers likely had increased MFG-E8 mRNA levels (Supplementary Fig. S2), which might explain the original description of MFG-E8 as a metastasis-associated tumor marker in this poor prognosis tumor class. We therefore assessed MFG-E8 in these BRCA1 associated and sporadic basal cancers (13), finding a significant increase in MFG-E8 levels in them compared with ER⁺ and erbB2⁺ tumors (Fig. 1E). In addition, we compared MFG-E8 protein in sera from 10 controls, 10 ER⁺, 10 erbB2⁺, and 10 triple-negative breast cancer patients (Fig. 1F). Control sera contained 318 ± 96 pg/mL of MFG-E8. MFG-E8 levels increased slightly to 1,700 and 1,730 pg/c in ER⁺/PR⁺ and erbB2⁺ patient sera, respectively. In contrast, a mean of 3,900 pg/mL of MFG-E8 was found in sera from triple-negative breast cancer patients. This increase was due in large part to 4 of the 10 patients whose MFG-E8 ranged from 3,000 to 50,700 pg/mL, markedly exceeding levels seen in the other tumor types.

To delineate cell types expressing MFG-E8 and to validate the results in ER⁺ breast cancers, we used in situ analyses using specimens from a separate cohort of breast cancer patients (Fig. 2, ref. 21). We found high-level expression of MFG-E8 in an ER-negative breast cancer (Fig. 2A) where a sense control probe showed no hybridization (Fig. 2B). Moreover normal-appearing breast tissues showed robust MFG-E8 expression, which was decreased in adjacent tumor tissue in an ER⁺ breast cancer (Fig. 2C). We confirmed that MFG-E8 is expressed in both juxta-lumenal and myoepithelial cells in normal ductal tissues (Fig. 2D), as previously described (10). We then performed our in situ analyses using 13 randomly selected ER⁻ and 12 ER⁺ breast cancers to evaluate the significance of our observations. Characteristic in situ results for ER⁺ versus ER⁻ tumors are demonstrated in Figure 2E–L. One ER⁺ tumor was among the top 13 MFG-E8-expressing tumors, and one ER⁻ tumor was among the bottom 12 expressing samples, showing that low MFG-E8 expression is highly correlated with the ER⁺ status of breast cancers (Fig. 2M, P < 0.001 by chi-square analysis).

Abnormalities in p63/p73 regulation are an important feature of triple-negative (basal) breast cancers (15). Recently, MFG-E8 was found to be a p63/p73 target gene in skin cells (18). To determine whether p63/p73 regulatory changes in triple-negative breast cancer cells might account for their increased MFG-E8 levels, we compared p63 levels in LCM-isolated normal, DCIS, and IDC breast tissues in ER⁺, erbB2⁺, and triple-negative cancers (Fig. 3A–C). mRNA for the transcriptionally active (TA) form of p63 (TAp63) increased in triple negative but decreased in ER⁺ and erbB2⁺ cancers.

We next evaluated the effect of conditional loss of p63, finding that 90% knockdown of p63 in myopithelial cells in p63^flox mice targeted by a K14 Cre recombinase system decreased mammary MFG-E8 levels by 40% (Fig. 4A). We then used p63 siRNA to knock its expression down in a triple-negative breast cell line. The p63 siRNA knockdown caused a 2.5-fold decrease in MFG-E8 mRNA (Fig. 4B). Using two triple-negative cell lines, we performed anti-p63 ChIP experiments, finding that p63 is present in the p63 binding region in the MFGE8 promoter in triple-negative breast cancers (Fig. 4C). We compared p63 binding at the MFG-E8 promoter to actin promoter binding as a negative control. As a positive control, we found that p63 bound to the PUMA promoter as previously described (not shown; ref. 15).

Finally, we used a previously described luciferase reporter assay (18) that reports activity at −1,198, −742, −370, and −27 p63 MFG-E8 promoter binding sites in T47D (Fig. 5A) and in MDA-MB468 cells (Fig. 5B). Reporter activity increased 12.3-fold in response to cotransfection with TAp63γ at the upstream site and 5.3-fold at the −877 site in T47D cells. This changed to 9-fold for TAp63α at the upstream site. p63 cotransfections had similar but lesser effects in MDA-MB468 cells.

Since MFG-E8 is the ligand for integrins α_vβ₃ and α_vβ₅ that promotes apoptosis (22–24), we evaluated effects of MFG-E8 RNAi on cell proliferation in ER⁺ breast cancer cells (Fig. 6A and B). We also compared the effect of the RGD peptide that blocks integrin–ligand interactions (7) on siMFG−E8-induced cell proliferation. MFG-E8 knockdown increased cell numbers for ER⁺ breast cancer cell lines and the RGD peptide blocked this stimulation.

These data identified potentially antiproliferative effects of MFG-E8-integrin β3/β5 signaling in ER⁺ breast cancer cell lines. Such antiproliferative effects have been shown in vivo for the integrin β3/β5 receptor by acceleration of tumor formation in transgenic MMTV-erbB2 mice lacking integrins β3 and/or β5 (12). To assess MFG-E8 in vivo, we crossed MFG-E8 knockout mice to erbB2 transgenic mice. Several tumors developed in erbB2⁺/MFG-E8 null mice before 1 year of age, but none developed in either the erbB2/MFG-E8^+/− or mice that were solely MFG-E8 null (Fig. 6C and D). However, erbB2-induced tumor onset was relatively slow in these crosses compared to the usual kinetics of tumor formation in inbred erbB2 mice. Consequently, we sacrificed remaining mice at 15 months of age and evaluated tumor formation using whole mounts to assess tumor incidence at 15 months of age (Fig. 6E). Tumor incidence caused by the MMTV-nonmutant-erbB2 transgene on its own was lower than published experiments using inbred strains. However, mice that were erbB2⁺ and MFGE8 null developed tumors at 3 to 7 times the rate seen in mice that were singly erbB2⁺ or MFG-E8 null alone.

We also evaluated MFG-E8 protein levels in 10 breast cancer cell lines (Fig. 7A top). High levels of MFG-E8 expression were only seen in triple-negative breast cancers. We measured secreted MFG-E8 levels in culture supernatants, finding that secreted MFG-E8 matches intracellular expression (Fig. 7A bottom). We evaluated the functional significance of these changes by knocking MFG-E8 down in two triple-negative breast cancer cells. The MFG-E8 siRNA effectively knocked its protein levels down (Fig. 7B), which led to increased sensitivity of both triple-negative breast cancer cell lines to the inhibitory effects of cisplatin (Fig. 7C and D).

BA46/MFG-E8 was first identified using an antibody (BA46) that could monitor metastatic breast cancer burden in patient samples. Subsequently, MFG-E8/lactadherin physiologic role was shown in mammary development, complicating the earlier findings. Here, we clarify potentially opposing roles MFG-E8 may play in breast cancers of different types. We first show that MFG-E8 is downregulated in ER⁺ and erbB2⁺ breast cancer (Figs. 1 and 2) where MFG-E8 is antiproliferative (Fig. 6). In contrast, we found that MFG-E8 is expressed at high levels in the basal/triple-negative set of breast cancers where it apparently functions as a functional target gene of the p63/p73 pathway. We present evidence that p63 directly regulates MFG-E8 in those cells (Figs. 3–5). These descriptions of MFG-E8 suggest that its interactions with its α_vβ_3,5 receptor offer new insights for the diagnosis and treatment of different breast cancer types.

We started by investigating gene regulation during breast cancer progression. We were surprised to find MFG-E8 decreases in mouse models and in human breast cancers, given its initial description. This was especially surprising since we separately found that estradiol induces MFG-E8 and inhibitors of erbB2 downregulate its expression (not shown). Evidently these effects are secondary to regulation by p63 given its control of MFG-E8 as shown in Figures 3–5. Decreases in MFG-E8 that accompany decreased p63 in ER⁺ or erbB2⁺ tumors may provide a selective advantage in these cancer types to evade immune clearance during tumor progression since MFG-E8 acts as the ligand for the phagocytic clearance pathway (9).

In contrast, MFG-E8 expression was increased in a different set of tumors-the triple-negative subset. These tumors are the most refractory to treatment, and recent studies are investigating novel treatment strategies for them. By evaluating microarray databases, a triple-negative–specific association with MFG-E8 expression in breast cancer was readily apparent. We confirmed this increased expression using cell lines and in a patient cohort, thus showing that the initial descriptions of BA46 potentially focused on patients with this type of tumor. We confirmed reports that p63/p73 regulates MFG-E8, and we showed this is functionally significant since an MFG-E8 siRNA increased cisplatin sensitivity. These results suggest that antagonists of MFG-E8-integrin signaling should be investigated in triple-negative breast cancers and further suggest that circulating MFG-E8 levels might be used to monitor the clinical response of some patients where p63 dysregulation is a major feature of their triple-negative breast cancer. MFG-E8 levels vastly exceeded those seen in any other tumor patient in three patients (3,000, 5,000, and 50,000 pg/mL). While these patients represent only 30% of the triple-negative patients tested, markers such as alpha fetoprotein and carcinoembryonic antigen are useful when found in similar proportions of patients with other cancers. One erbB2 tumor patient had a serum level of 3,670 pg/mL. This level could be due to misclassification, to a mixed tumor phenotype, to co-occurrence of another disease with elevated MFG-E8 levels such as systemic lupus erythematosus (25) or to secretion by inflammatory cells in the tumor itself. Importantly, healthy donors exhibited no MFG-E8 levels higher than 366 pg/mL, again suggesting that further explorations of MFG-E8 levels as a serum tumor marker are warranted.

Taken together, our results show that MFG-E8 joins a large group of ligands with context-specific functions in cancer. For example, these dual roles are not unlike the dual functions of transforming growth factor-β. Our results provide additional evidence that triple-negative breast cancers are phenotypically distinct from other breast cancers and these interactions suggest that integrins might be a druggable target whose ligands need to be evaluated in context as integrin-related therapies are advanced in the clinic.

No potential conflicts of interest were disclosed.

We thank Andrea Richardson for help in obtaining the basal breast cancer mRNA samples from the Harvard Breast Cancer SPORE tissue bank. We also thank Dr. I. Katoh (Ikawa Laboratory, RIKEN, Wako, Japan) for the kind gift of the MFG-E8 promoter luciferase plasmids.

This work, C. Yang, L. Chen, and E. V. Schmidt were supported by a grant from the National Cancer Institute of the National Institutes of HealthRO1 CA69069 and by the Harvard Breast Cancer SPORE grant P50 CA89393. C. Yang received additional support from R01 CA112021, and E. V. Schmidt received support from U01 AI07033. D. Sgroi and C. Li were supported by the Harvard Breast Cancer SPORE grant P50 CA89393 and NIH RO1 CA112021. NIH/National Cancer Institute grant CA89138 supported S. Maheswaran and T. Hayashida. K.S. Anderson was supported by NCI U01 CA117374. The Tracey Davis Memorial Fund supported Nicole Forster and Leif Ellisen received support from the Avon foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.