The nuclear coactivator amplified in breast cancer 1 maintains tumor-initiating cells during development of ductal carcinoma in situ.

The key molecular events required for the formation of ductal carcinoma in situ (DCIS) and its progression to invasive breast carcinoma have not been defined. Here, we show that the nuclear receptor coactivator amplified in breast cancer 1 (AIB1) is expressed at low levels in normal breast but is highly expressed in DCIS lesions. This is of significance since reduction of AIB1 in human MCFDCIS cells restored a more normal three-dimensional mammary acinar structure. Reduction of AIB1 in MCFDCIS cells, both before DCIS development or in existing MCFDCIS lesions in vivo, inhibited tumor growth and led to smaller, necrotic lesions. AIB1 reduction in MCFDCIS cells was correlated with significant reduction in the CD24-/CD44+ breast cancer-initiating cell (BCIC) population, and a decrease in myoepithelial progenitor cells in the DCIS lesions in vitro and in vivo. The loss of AIB1 in MCFDCIS cells was also accompanied by a loss of expression of NOTCH 2, 3 and 4, JAG2, HES1, GATA3, human epidermal growth factor receptor 2 (HER2) and HER3 in vivo. These signaling molecules have been associated with differentiation of breast epithelial progenitor cells. These data indicate that AIB1 has a central role in the initiation and maintenance of DCIS and that reduction of AIB1 causes loss of BCIC, loss of components of the NOTCH, HER2 and HER3 signaling pathways and fewer DCIS myoepithelial progenitor cells in vivo. We propose that increased expression of AIB1, through the maintenance of BCIC, facilitates formation of DCIS, a necessary step before development of invasive disease.

Introduction

There are >50,000 new cases of Ductal Carcinoma In Situ (DCIS) diagnosed every year in the USA. In DCIS lesions the lumen of the mammary ducts are filled with proliferative malignant cells that have not invaded beyond the basement membrane of the duct into the adjacent stroma. Based on epidemiology, pathology and assessment of shared genetic changes between DCIS and adjacent invasive lesions, DCIS has been identified as the precursor lesion for invasive breast cancer[1-3]. Extensive comparative gene expression analyses to date have not defined the critical driver pathways that are needed for the development and maintenance of DCIS and its further progression to invasive carcinoma[4,5].However, DCIS and invasive disease share a similar gene expression profile, whereas the expression profile of normal breast and DCIS are significantly different[6]. This suggests that many of the important signaling changes that are required for the development of invasive and even metastatic disease occurred during the formation of DCIS.

AIB1 is the third member of the Nuclear Coactivator (NCOA-3) and p160 Steroid Receptor Co-activator-3 (SRC-3) family. The AIB1 oncogene is located on chromosome 20q, a region frequently amplified in breast cancer[7]. AIB1 is a transcriptional co-activator that promotes the transcriptional activity of multiple nuclear receptors such as the estrogen and progesterone receptors[7] and a number of other transcription factors, including E2F-1, AP-1, NFκB, and STAT6[8]. Multiple studies have shown that the AIB1 gene is amplified and overexpressed in human breast cancer[9]. High levels of AIB1 mRNA or protein predict significantly worse prognosis and overall survival in breast cancer patients[9]. Transgenic mice expressing high levels of the human AIB1 transgene developed mammary hyperplasia and tumors of the mammary gland[10]. Conversely, loss of AIB1 expression in the mammary gland prevents RAS- or HER2-induced mammary tumor development[11,12]. Transgenic mice expressing high level of an AIB1 isoform in the mammary gland resulted in the development of DCIS and fibrosis[13]. In the current study we determined that AIB1 is highly expressed in human DCIS samples. Based on these observations, we hypothesized that AIB1 could be important in the early stages of breast cancer notably in the development and maintenance of DCIS. We have examined the role of AIB1 in a well-described model of human DCIS, MCFDCIS.com (MCFDCIS) cells, which are a derivative of premalignant MCF-10AT cells, that when implanted in mice produce discrete DCIS lesions that rapidly progress to invasive disease[14]. The gross pathology of the lesions and the gene expression changes observed in the DCIS stage of MCFDCIS after implantation mirror those seen in human DCIS samples[15]. Using this model, we demonstrate that AIB1 is critical for the formation and maintenance of DCIS. AIB1 is required to maintain the BCIC population and the myoepithelial progenitor cell population in the MCFDCIS cells in vitro and in vivo. In parallel, AIB1 also maintains levels of NOTCH and HER2/HER3 signaling molecules in the DCIS lesions. These pathways are known to be involved in maintaining BCIC populations. Based on these data, we propose that preventive strategies aimed at reducing AIB1 in the mammary gland could be selective and effective in reducing DCIS incidence and consequently decreasing the overall incidence of invasive breast cancer.

Results

AIB1 expression in human DCIS

To determine the expression levels of AIB1 in human DCIS, we obtained paraffin-embedded sections of DCIS patient samples. We determined that the AIB1 levels in normal breast ducts, detected by immunohistochemistry (IHC), were very low (. In contrast, strong AIB1 expression was observed in the luminal cells of DCIS lesions from patients with invasive breast cancer (. Estrogen receptor positive (ER+), human epidermal growth factor receptor 2 positive (HER2+) and ER−, HER2− and progesterone receptor negative (PR−) ‘triple negative’ invasive breast cancer samples all showed strong AIB1 staining (. Strong AIB1 staining was also observed in breast tissue samples from DCIS patients that had no evidence of invasive disease (. The staining for AIB1 in the DCIS samples was high in the nuclear or cytoplasmic compartments but patterns of AIB1 subcellular distribution did not appear to be associated with a particular breast cancer subtype. Previous reports have shown that AIB1 and its isoform can traverse between different cellular compartments depending on, stage of cell cycle and cellular expression of other coregulators[16-18]. Overall, our IHC results indicate that AIB1 protein is highly expressed at very early stages of human breast cancer progression prior to development of Stage 1 invasive disease, and this suggests that AIB1 could function in an early stage of breast cancer formation.

To determine the role of AIB1 in early stage breast cancer we utilized the MCFDCIS cell model of breast cancer[14]. When implanted subcutaneously MCFDCIS cells give well defined DCIS lesions that progress to invasive disease within 3-5 weeks[19](. The ductal lesions that develop after 3 weeks implantation of MCFDCIS cells resemble high grade comedo DCIS ( and both luminal and myoepithelial cell layers of DCIS lesions are generated by the progenitor MCFDCIS cells[14]( Invasive carcinoma develops from the DCIS lesions between weeks 3-5 after implantation of cells (. Similar to the human DCIS samples strong AIB1 staining was found in the MCFDCIS lesions 3-weeks after implantation (. The presence of AIB1 protein was confirmed by Western blot analysis of MCFDCIS cells in vitro ( and in MCFDCIS tumor extracts (. AIB1 was also detected in the invasive lesions that developed from the MCFDCIS cells in vivo although the staining was less intense than in the DCIS lesions (. FISH analysis revealed that the MCFDCIS cells do not have amplified copies of either AIB1 ( or HER2 (. The MCFDCIS lesions are PR-negative ( and have low levels of ERα (. MCFDCIS tumor growth is estrogen-independent since it is unaffected by ovariectomy ( or by treatment of mice with the estrogen antagonist fulvestrant (. These observations are consistent with a previous classification of the MCFDCIS lesions as a basal form of human DCIS[15].

AIB1 has a role in the maintenance of DCIS multi-acinar structures in a 3D matrix

We first examined the impact of changing AIB1 levels in mammary acinar 3D structures. Normal mammary epithelial cells (MEC) cells, such as MCF-10A, grown in Matrigel™, recapitulate features of normal breast, including polarized acini, and deposition of basement membrane characterized by laminin V. After a number of days of growth, the acini develop a hollow lumen characteristic of the lumen seen in mammary acini in vivo[20](. The overexpression of oncogenes, such as HER2, disrupts normal mammary acinar structure and instead larger, multi-acinar structures are observed, with filled lumen and irregular deposition of basement membrane proteins[21]. The MCFDCIS cells grown in Matrigel™ showed multi-acinar structures, with no lumen development and with deposition of laminin throughout the multi–acinar structure (. To determine the impact of AIB1 on maintenance of the DCIS phenotype in Matrigel™, we infected MCFDCIS cells with a lentiviral vector expressing two distinct AIB1 shRNAs (shAIB1-1 and shAIB1-2) that lowered the endogenous AIB1 levels by 40% and 70% respectively (. Reduction in AIB1 levels after shAIB1-1 infection showed some changes in laminin V distribution to the periphery of the acini (, but with the more effective shRNA AIB1-2 a more normal spheroid acinar structure was observed (. The average spheroid size of MCFDCIS shAIB1-2 spheroids was also significantly smaller than MCFDCIS shCTRL and similar to the average size seen in MCF-10A ( Consistent with this, overexpression of AIB1 in MCF-10A ( caused loss of polarization and an increase in highly disorganized acini in 3D culture (. Interestingly, in the MCFDCIS shAIB1 spheroids we observed Caspase 3-positive apoptotic cells especially in the center of the sphere ( and an increased expression of the apoptotic marker BIM EL and the cell cycle inhibitor p21 (. Changes in expression of these molecules have been associated with resumption of a more normal mammary acinar structure with the hollow lumen being created by increased apoptosis[20]. Overall, the data indicate that high AIB1 levels in the MCFDCIS cells are required to maintain a disorganized multi-acinar structure in Matrigel™ and reduction of AIB1 levels can partially restore normal mammary acinar structure in MCFDCIS cells.

AIB1 is required for the formation of DCIS lesions in vivo

We next determined if the loss of AIB1 in MCFDCIS cells could affect MCFDCIS tumor development in vivo. To examine this, we infected the MCFDCIS cell line with a lentiviral vector (. In these pooled selected cells, AIB1 mRNA and protein levels are constitutively depleted (. Cells were injected subcutaneously into nude mice and the tumors monitored for up to 24 weeks or to the time that they exceeded the size of 500 mm3 (. The control-infected cells grew rapidly and all of the tumors had reached 500 mm3 volume by 7 weeks after implantation (. In contrast the AIB1 shRNA infected cells grew slowly and 20% of the tumor injection sites had not developed lesions>500 mm3 by 24 weeks (. Tumors from the control animals had the characteristic comedo DCIS lesions (. DCIS lesions were also observed in the shAIB1 cells, but they were of a significantly smaller area (, with less overall prevalence (, and more necrosis than the control (. We also observed a decrease in the number of proliferative cells in the MCFDCIS AIB1 shRNA lesions as determined by the quantitation of overall PCNA staining ( and the percentage of p21-positive cells (. These data indicate that MCFDCIS cells with low AIB1 levels will not progress to invasive tumors as rapidly as the control cells. Furthermore, the DCIS lesions that develop in the presence of lower levels of AIB1 will be smaller and more necrotic.

AIB1 is required for the maintenance of DCIS lesions in vivo

We next asked whether reduction of AIB1 in MCFDCIS cells would impact on the maintenance of DCIS lesions once they have formed. We developed a MCFDCIS derivative line infected with lentivirus that harbored the control scrambled shRNA or shRNA AIB1-2 as well as the turbo red fluorescent protein (tRFP) under the control of a tetracycline inducible promoter (. Thus, doxycycline (Dox) induction of tRFP and shRNA expression can be detected in MCFDCIS cell culture () and by in vivo imaging ( The administration of Dox causes reduction of AIB1 mRNA and protein in vitro (. After implantation of these cells subcutaneously into nude mice, DCIS lesions were allowed to develop for 2 weeks and then doxycycline was included in the diet (. The level of AIB1 expressed in the tumors was determined by IHC. The AIB1 levels in the shCTRL tumors was unaffected by doxycycline treatment (. All the MCFDCIS control cells grew rapidly irrespective of doxycycline treatment (. In contrast induction of AIB1 shRNA in MCFDCIS by doxycycline resulted in reduced AIB1 protein expression in MCFDCIS shAIB1 tumors ( and a decrease in size in 50% of the tumors (. The tumors that developed from the AIB1-depleted MCFDCIS cells had smaller and more necrotic lesions ( as well as an overall decrease in number of PCNA-positive cells compare to the controls (. Overall the data in both the constitutive and conditional AIB1 shRNA expressing systems suggest that AIB1 is required for the establishment and maintenance of DCIS.

Loss of AIB1 causes reduction in the BCIC population of MCFDCIS and affects MCFDCIS cell differentiation

To investigate how AIB1 affects initiation and progression of DCIS, we compared global gene expression changes in MCFDCIS cells +/−AIB1 shRNA in vitro using cDNA array analysis. Notable among the most significant expression changes between the two groups were genes that are differentially expressed in BCIC such as CD24[22] and members of the NOTCH signaling pathway (DLL1 and DLL3)[23](. Luminal progenitor marker expression, such as Mucin 1 (MUC 1)[24], was also increased (. To determine if these expression changes impacted the population of BCIC, we FACS sorted MCFDCIS +/− AIB1 shRNA cells for CD24−/CD44+ (. Reduction of AIB1 levels caused the percentage of CD24−/CD44+ cells in the MCFDCIS population to drop substantially from 18% to <3 % (. Consistent with a loss of BCIC, the CD44 and CD49f levels were also significantly reduced in MCFDCIS AIB1shRNA cells (. In contrast, the level of expression of CD24 and ESA on their cell surface was increased (. There was also an increase in overall expression of MUC1 mRNA in the AIB1shRNA MCFDCIS cells ( CD24 and ESA, like MUC1 are considered luminal cell markers and the gain in their expression after AIB1 reduction suggested that there were changes in relative levels of progenitor luminal vs. progenitor myoepithelial cells in MCFDCIS population. Consistent with a role for AIB1 in BCIC maintenance, overexpression of AIB1 in MCF-10A cells led to increased sphere formation and increased expression of CD44 and CD49f (.

To determine if the changes in BCIC cell populations were observed after AIB1 reduction in vivo we stained MCFDCIS tumors obtained from the constitutive and conditional mouse models for CD44, CK18 and p63 expression (. A significant reduction in overall CD44 staining in the smaller lesions that developed in MCFDCIS AIB1 shRNA cells implanted in vivo is also consistent with BCIC loss (. Interestingly, in the MCFDCIS AIB1 shRNA lesions, there was a gain in vivo in the CK18 staining, a marker of differentiated luminal epithelium (. In contrast, myoepithelial progenitor cells were significantly reduced in AIB1 shRNA, cells, indicated by loss of overall expression of αSMA mRNA in vitro ( and loss of p63-positive cells in vivo (.

In summary, the loss of AIB1 in MCFDCIS cells in vitro and in vivo reduces the BCIC and myoepithelial progenitor populations and also prevents the progression of the remaining luminal cells to invasive cancer.

Disruptions in NOTCH and HER signalings are observed in MCFDCIS AIB1 shRNA cells and tumors

cDNA array analysis of gene expression changes induced by loss of AIB1 in MCFDCIS cells indicated that NOTCH ligands (DLL1 and DLL3) were significantly downregulated (. This was of interest since the NOTCH signaling pathways has been shown to be involved in BCIC cell maintenance and differentiation[23]. We therefore examined the impact of reduction in AIB1 expression in MCFDCIS cells in vitro and in vivo on the gene expression pattern of a panel of NOTCH signaling molecules. Downregulation of DLL1, DLL3, JAG 1 and JAG2 mRNA levels was confirmed by real-time PCR in MCFDCIS cells upon shRNA AIB1 expression (. Consistent with this, overexpression of AIB1 in MCF-10A results in an increase in DLL1, DLL3, NOTCH 2,JAG1 and JAG 2 mRNA levels (. In the MCFDCIS tumors we also observed JAG2, NOTCH 2, 3 and 4 mRNA levels downregulated ( as well as the NOTCH target genes HES1 and GATA3 (. Previous reports have indicated that HER2 and HER3 control the level of NOTCH signaling in BCIC[23]. Thus it was possible that changes in AIB1 levels were affecting HER2 and HER3 signaling and this in turn was altering expression of members of the NOTCH pathway. We determined that HER2 and HER3 mRNA and protein were detectable by PCR and Western blot analysis in MCFDCIS cells and tumors (. A significant reduction in HER2 and HER3 mRNA and protein levels are observed in the MCFDCIS tumors after reduction in AIB1 levels ( while EGFR protein levels are unchanged ( Furthermore, a clear positive correlation between AIB1 mRNA and HER2 and HER3 mRNA levels was found in the tumors (. Consistent with this, we demonstrated that HER2 mRNA and protein are upregulated in MCF-10A cells infected with lentivirus expressing AIB1 (. In summary, reduction in AIB1 levels in MCFDCIS tumors in vivo leads to loss of HER2, HER3 as well as loss of NOTCH signaling pathway members. These changes in gene expression are paralleled by a significant loss in BCIC population and myoepithelial progenitor cells.

Discussion

This is the first report that defines a role for AIB1 in the development and maintenance of human DCIS. Although AIB1 is known to be involved in the proliferation of a number of different epithelial tumors[25], the role of AIB1 in the maintenance of BCIC has not been reported previously. The loss of the BCIC population in MCFDCIS tumors with reduced levels of AIB1, most likely leads to the reduced size of DCIS lesions and the loss of overall MCFDCIS tumor burden due to increased necrosis. While increased necrosis is often seen in the lumen of DCIS lesions in vivo, the necrosis we observed was more associated with the inability to sustain the DCIS cells thus inhibiting their progression to invasive lesions. Interestingly reduction of AIB1 after formation of DCIS lesions also causes formation of smaller lesions, reduced tumor burden and increased necrosis. This suggests that sustaining the BCIC population is required throughout DCIS development, not just after initial implantation of the cells in vivo. This has important therapeutic implications since it suggests that a reduction of AIB1 or inhibition of AIB1 signaling pathways involved in the maintenance of BCIC would hamper development of DCIS. It should be noted that only 20-30% of human DCIS is thought to progress to invasive disease[26] and this raised the question of the role of AIB1 in progression of DCIS to invasive disease. Although MCFDCIS cells can give rise to invasive lesions in vivo, they have low motility in a Boyden chamber invasion assay ( and low invasion potential in an endothelial layer in an ECIS assay compared to invasive MDA-MB-231 cells (. It is also known that MCFDCIS cells have low cell velocity in the wound healing assay[27]. Consistent with this low motile potential of MCFDCIS, no micrometastases were observed after tumor formation in vivo (. We found that reduction in AIB1 did not affect these minimal phenotypes of MCFDCIS cells in either the in vitro invasion or motility assays (. However, AIB1 may have an additional role in epithelial cell invasion in the context of stroma that is not apparent in these assays. Since AIB1 is highly expressed in the patient DCIS samples that we have examined to date, and its loss prevents MCFDCIS invasive tumor development, we can conclude that AIB1 overexpression is necessary primarily for development and maintenance of DCIS through preserving the BCIC population, but further genetic changes are most likely necessary for the development of the invasive phenotype in a subset of DCIS.

Our analysis of gene expression controlled by changes in AIB1 levels in MCFDCIS cells revealed that members of the NOTCH pathways are regulated by AIB1. This was confirmed by analysis of the in vivo samples although interestingly the patterns of changes in NOTCH signaling molecules in vivo did not parallel the changes seen in vitro. Although NOTCH1 was not regulated by AIB1 in vitro or in vivo, notably NOTCH 2, 3 and 4, and JAG 2 were significantly downregulated in vivo. We conjecture that the regulation of NOTCH by AIB1 in vivo requires crosstalk with stromal components that are vital to the maintenance of DCIS in vivo. NOTCH signaling has been implicated in both restricting mammary stem cell expansion and has been associated with commitment of mammary stem cells to the luminal lineage, leading to hyperplasia and tumor formation[28]. In implanted MCFDCIS in which AIB1 was lowered we did observe a decrease in hyperplasia and a decrease in the number of tumors although we did not see eradication of luminal cells which might be expected with a significant reduction in NOTCH signaling. Interestingly, recent reports have demonstrated that the NOTCH pathway can be controlled by HER family signaling in breast cancer stem cells[29] and in DCIS[30]. AIB1 is required for HER2 mediated mammary tumor formation in vivo[11,12] and high levels of AIB1 and HER2 have been associated with worse prognosis and tamoxifen resistance in human breast cancer[9]. Overexpression of HER2 is seen in significant portion of DCIS cases estimated at 50-80%[31]. However, an association between AIB1 and HER2 in the development of DCIS and its progression to invasion has not been established. In our gene expression analysis of +/−AIB1 in MCFDCIS cells in vitro, major changes in expression of HER family members HER1, 2 and HER3 were not observed. However, in vivo a reduction in AIB1 led to loss in HER2 and HER3 expression, again suggesting the importance of stromal crosstalk in determining levels of gene expression in DCIS. HER2 signaling can regulate mammary stem cell renewal[32] and loss of HER2 signaling controlled by loss of AIB1 could explain the significant loss of BCIC in vivo. Interestingly, HER3 has recently been shown to play a role in maintaining luminal epithelium cells and loss of HER3 signaling causes a switch to mammary basal epithelium gene expression patterns[33]. Thus, the loss of HER3 signaling in vivo may also play a role in overall loss of luminal cells, although it would not explain the decrease in the myoepithelial progenitor layers in the AIB1 depleted lesions. We tried to rescue BCIC loss caused by reduced AIB1 with overexpressed HER2 but were unable to increase HER2 expression in the shAIB1 cells (, presumably because the CMV promoter regulating HER2 expression in the lentiviral vector is also dependent on AIB1. Nevertheless, the data in and strongly implicate a connection between AIB1, HER2/3 and NOTCH in development and maintenance of DCIS.

Previous studies in transgenic mice indicated that overexpression of AIB1 can lead to DCIS[13] and we now show that human DCIS has high expression of AIB1 at the protein level. 5-10% of breast cancers harbor an AIB1 amplicon[7] although high AIB1 levels can be observed in the absence of gene amplification in up to 60% of breast cancer. AIB1 protein levels are regulated by multiple post-transcriptional mechanisms that control steady state levels of AIB1 during the transformation process[8]. AIB1 levels in cells can be downregulated by estrogen[34]. However, loss of estrogen regulation does not seem to be the explanation of high AIB1 in DCIS since we observed high AIB1 levels in ER+ and ER− DCIS as well as HER2+ and HER2-DCIS. A previous study examining ERα knock out mice demonstrated that ER signaling was necessary for the formation of DCIS[35]. Our group also published a study in which development of DCIS and invasive mammary cancer was correlated with over expression of AIB1 with ERα in transgenic mice[13]. The present study extends these observations demonstrating that AIB1 over-expression can also contribute to the development of the basal form of DCIS. Thus the contribution of AIB1 over-expression to DCIS development and maintenance is not breast cancer sub-type specific. In conclusion, regardless of the mechanism of upregulation of AIB1 in human breast epithelium, our data indicate a critical role for this event in the development and maintenance of DCIS lesions in vivo through preservation of BCIC in part through NOTCH, HER2 and HER3 signaling. Our data suggest that selective degradation of AIB1 in early breast cancer lesions could be a useful therapeutic approach to prevent the development and maintenance of DCIS, thus reducing the overall incidence of invasive breast cancer.

Materials and Methods

Human Tissue Samples and Mice

De-identified formalin-fixed, paraffin-embedded tissues from normal breast reduction mammoplasty and DCIS were obtained from Georgetown University tissue bank shared resource. Athymic nude mice were obtained from Harlan Laboratories, were maintained in the Georgetown University's animal facility and animal experiments were conducted in accordance with procedures approved by the Institutional Animal Care and Use Committee.

Cell lines

We thank Dr Fred R. Miller of the Karmanos Cancer Institute for the gift of the MCFDCIS and MCF-10A cell lines. These cells were maintained as described previously[36,37].

Histology analysis

Hematoxylin and eosin, immunohistochemistry and immunofluorescence analyses were performed on paraffin-embedded 5 μm sections using standard protocols described elsewhere[12].

Western Blot

Whole-cell extracts were isolated from cultured MCFDCIS, MCF-7 and MCF-10A cells or DCIS tumor samples, and immunoblot assays were performed as described previously[38].

Fluorescence In Situ Hybridization (FISH) analysis

FISH hybridization and analysis were performed in the MCFDCIS cells using a standard protocol previously described[39,40].

Short Hairpin RNA Constructs and lentivirus Infection

For gene down-regulation studies, short hairpin RNAs (shRNAs) against AIB1 gene (shRNA#1 5’GCAGTCTATTCGTCCTCCATA3’, shRNA#2: 5’TGGTGAATCGAGACGGAAACA 3’,) were cloned into PLKO.1 lentiviral vector (Addgene) as described in[41] or into XhoI and EcoRI restriction sites in Doxycycline-inducible lentiviral expression vector pTRIPZ (Open Biosystems). A scrambled shRNA was used as a control (Addgene). For gene overexpression, the coding sequence of human AIB1 was cloned into AscI and PacI restriction sites in the lentiviral expression vector pCDF1-MSC2-EF1-Puro (System Biosciences). Lentivirus production and infection were performed as described elsewhere[36,42].

Xenografts

MCFDCIS xenograft models were generated as detailed in and following the protocol described presviously[15].

Analysis of three-dimensional Matrigel™ cultures by Bright-field and Confocal Microscopy

3D culture assays were performed as previously described[36,37]. For confocal microscopy, immunostaining of the acinar structures was carried out as described elsewhere[43].

Quantitative RT-PCR

Real-time quantitative RT-PCR analyses of total cellular RNA from MCFDCIS cells, MCF-10A cells and DCIS xenografted tissues was carried out as previously described[41]. All of the primers used for quantitative RT-PCR are given in . The results were calculated by the comparative CT method, with relative transcript levels determined as 2-ΔΔCT and were normalized using actin gene[44].

Flow Cytometry Analysis

MCFDCIS cells were stained with the following antibodies: anti-CD24-488, anti-CD44-PE, anti-CD49f-APC, anti-ESA-488, and anti-CD10-PE (BioLegend). FACS analysis was performed in the Lombardi Comprehensive Cancer Center Flow Cytometry Shared Resource.

Statistics

All experiments have been carried out at least three times. Statistical differences and linear regression analysis were performed with GraphPad Prism software (Graph-Pad Software, Inc., San Diego, CA, USA). The significance of changes was assessed by the application of an unpaired Student's t-test, with p<0.05 considered statistically significant unless stated otherwise.