Notch activation stimulates migration of breast cancer cells and promotes tumor growth

Introduction Dysregulated NOTCH receptor activity has been implicated in breast cancer but the mechanisms by which NOTCH contributes to transformation are not yet clear, as it has context-dependent effects on the properties of transformed cells. Methods We have used various in vitro and in vivo carcinogenic models to analyze the impact of Notch signaling in the onset and progression of breast tumors. Results We found that ectopic expression of the Notch1 intracellular domain (N1ICD) in MCF-7 breast adenocarcinoma cell line caused reduction and delocalization of E-CADHERIN levels and increased migratory and invasive abilities. Notch inhibition in the invasive breast cancer cell line MDA-MB-231 resulted in increased E-CADHERIN expression and a parallel reduction in their invasive capacity. The growth of subcutaneous xenografts produced with MCF-7 cells was boosted after N1ICD induction, in a cell autonomous manner. In vivo Notch1 activation in the mammary gland using the MMTV-Cre driver caused the formation of papillary tumors that showed increased Hes1 and Hey1 expression and delocalized E-cadherin staining. Conclusions These results confirm NOTCH1 as a signal triggering epithelial-mesenchymal transition in epithelial cancer cells, which may have implications in tumor dissemination, metastasis and proliferation in vivo. The identification of specific factors interacting with NOTCH signaling could thus be relevant to fully understanding the role of NOTCH in breast neoplasia.


Introduction
Notch is a fundamental signaling pathway that regulates embryonic cell fate specification, proliferation and patterning [1,2]. In addition to its central role in development, Notch signaling is deregulated in a number of cancers [3]. Notch1 mutations lead to oncogene expression in certain T cell acute lymphoblastic leukemias [4] and a subset of breast carcinomas [5]; deregulated Notch activity might also affect cell transformation [6], regulation of the cell cycle [7], progenitor/stem cell maintenance [3] and the outcome of breast cancer [8].
The mammalian Notch proteins (Notch1 to 4) are membrane-bound type I receptors with a large extracellular domain involved in ligand binding, and a cytoplasmic domain responsible for signal transduction. The Notch ligands Delta-like 1, 3 and 4 and Jagged 1 and 2 are also membrane-bound. Ligand-receptor interactions between neighboring cells trigger Notch signaling, which leads to a sequence of proteolytic cleavage events in the receptor. The last of these is mediated by γ-secretase activity, generating the Notch intracellular domain (NICD), which translocates to the nucleus and binds the CSL transcription factor. The NICD/CSL complex induces expression of target genes, including those of the hairy/enhancer of split (Hes) family [1,2], the cell cycle regulator p21 [9] and cyclin D1 [7].
Many studies focus on the role of Notch1 in mammary tumorigenesis. Hyperactivated Notch1 signaling was first implicated in mammary tumorigenesis in studies of the MMTV model, which showed that N1ICD expression in MMTV-Neu mammary tumors is due to an MMTV insertion in the Notch1 locus [10]. Other reports indicated that transgenic activation of N1ICD in mammary glands leads to development of lactation-dependent tumors that regress at weaning [11,12]. These findings link aberrant Notch activation in the murine mammary gland to adenocarcinoma. Experimental evidence shows that altered Notch1 signaling leads to direct transcriptional regulation of c-myc, which is crucial in MMTV-N1ICD-induced murine mammary tumorigenesis [13]. NOTCH1 is also involved in human mammary tumorigenesis as a downstream effector of oncogenic Ras [14].
Here we used various in vitro and in vivo models to analyze the impact of Notch signaling in breast tumor onset and progression. We find that stable or inducible N1ICD expression in the poorly invasive MCF-7 breast adenocarcinoma cell line causes a reduction and delocalization of E-CADHERIN levels, suggesting a disassembly of adherens junctions that correlates with enhanced cell migratory and invasive abilities. These properties may be extended to other epithelial tumor cell lines as we have made similar observations in the colon cancer cell line HT-29 stably expressing N1ICD. To the contrary, Notch inhibition in the highly invasive cell line MDA-MB-231 resulted in increased E-CADHERIN expression and a parallel reduction in their invasive capacity. Notch1 activation in the mouse mammary gland using the MMTV-Cre driver caused the formation of papillary tumors that showed increased Hes1 and Hey1 and delocalized E-cadherin expression. We also found that the growth of subcutaneous xenografts produced with MCF-7 cells was boosted after N1ICD induction, in a cell autonomous manner. These results confirm Notch1 as an epithelialto-mesenchymal transition (EMT) inducer in breast cancer cells, which may have implications in tumor dissemination and metastasis.

Transfection of MCF-7 and HT-29 cells
A cDNA fragment encoding the active version of mouse Notch1 (N1ICDΔ OP ) was used [15]. The Tet-Off system was employed to obtain transfectants of MCF-7 with inducible N1ICD expression. In this system, gene expression is turned on when doxycycline (DOXY; a tetracycline derivative) is removed from the culture medium. For details see Additional file 1, Supplementary Materials and methods.

Western blot analysis
For details see Additional file 1, Supplementary Materials and methods.

Immunofluorescence and immunohistochemistry
For details see Additional file 1, Supplementary Materials and methods.

In situ hybridization
In situ hybridization was performed as described in [19]. Details of probes will be provided on request.

In vitro cell chemotaxis
Cell migration was performed in Transwell (Corning, Tewksbury, MA, USA) with 8 μm pore filters coated with 20 μg/ml collagen type IV (Sigma, St. Louis, MO, USA). Cells were pretreated for the time indicated with DOXY (MCF-7) or DAPT/RO4929097/DMSO (MDA-MB-231), trypsinized and added to the upper chamber in basal medium with 0.5% BSA and the additives. The lower chamber was replenished with basal medium with BSA and the chemoattractant (IGF-1, 50 ng/ml, R&D Systems, Minneapolis, MN, USA or SDF1α, PeProTech (New Jersey, USA). After 18 h incubation, the upper chamber was emptied and cells remaining are removed. Cells in the filter are fixed with PFA and then stained with violet crystal (Sigma-Aldrich). Cell counts were obtained by counting two (MDA-MB-231) or four (MCF-7) grids using a microscope fitted with a grid eyepiece at a total magnification of 100X.

Flow cytometry
For details see Additional file 1, Supplementary Materials and methods. Tumorigenic assays MCF-7/TetOff and B12, M5 and M20 derivatives' clones, growing in culture without DOXY for 25 days, were inoculated s.c in both flanks (1.5 or 2.4 × 10 6 cells) in BALBc/SCID mice treated with 17α-ethylenestradiol 1 μg/ml (Sigma) provided in the drinking water from one week before cells were injected. Tumor size was monitored weekly and tumor volume estimated with a caliper by measuring the width (a) and the length (b) and applying the formula (a 2 × b)/2. Once finished with the period of treatment, mice were sacrificed and tumors were extracted for further analysis. MCF-7/TetOff and B12 were transduced with recombinant retrovirus to express luciferase activity. Plasmid pRV-luc-IRES-CopGreen was used to obtain the retroviral supernatants (Genetrix S.L., Madrid, Spain) and transduced cells were sorted according to the associated green fluorescence. BALBc/SCID mice (Harlan Laboratories, Indianapolis, IN, USA) were injected in the two inguinal mammary glands with 2.5 × 10 6 cells and mice were treated as above. After injection, half of the mice were treated also with DOXY 2 mg/ml provided in the drinking water. Tumoral growth rate was analyzed by bioluminescence at different weeks after cell inoculation. Briefly, mice were injected with luciferin with the general anesthetic and luciferase activity expressed by cells was detected with a CCD camera placed in a dark box (Hamamatsu Photonics, Shizuoka, Japan). Images were processed with the software provided and luminescence units were represented. Tumor size was estimated as above and once finished with the period of treatment, tumors were excised, weighted and preserved adequately to make further analysis.

Transgenic N1ICD expression in the mammary gland
The transgenic lines MMTV-Cre [20] and Rosa26N1ICD [21] were bred to generate MMTV-Cre/+; Rosa26N1ICD/ + double transgenic mice. For primers and conditions of mouse genotyping see [20,21]. Mice were subjected to several rounds (a median of four) of pregnancy and lactation, and when a breast tumor arose, mice were euthanized and the breast tumor excised and processed for further analysis. Tumor samples were fixed with 10% buffered formalin (Sigma-Aldrich) for 48 h and afterward were paraffin-embedded. Staining of Hes1, ERα, p63, E-cadherin and Ki67 was performed in 5 μm sections of paraffin samples following standard techniques. For details see Additional file 1, Supplementary Materials and methods.

N1ICD expression enhances the invasive capacity of the breast cancer cell line MCF-7
To gain an insight into the role of Notch in breast cancer we used the breast cancer cell line MCF-7 that has several features of differentiated mammary epithelium [22]. These cells show low levels of N1ICD expression by Western blot when compared with the metastatic breast cancer cell line MDA-MB-231 ( Figure 1A). This observation fits with the idea that high NOTCH signaling is associated with the expression of basal breast cancer markers [23].
We generated MCF-7 clones stably expressing a myctagged N1ICD version ( Figure 1B). MCF-7 cells have a typical cobblestone phenotype (not shown) and express the epithelial cell marker E-CADHERIN ( Figure 1B, C). N1ICD expression caused a reduction in total E-CADHERIN levels in MCF-7 clones E8 and F7 but not in clone F5 ( Figure 1B, C). The levels of Notch activity in MCF-7/N1ICD cells measured by a luciferase reporter assay using a fragment of the mouse Hes1 promoter [17], revealed an evident activation of the Notch pathway in  Figure 1D). qPCR analysis revealed a marked upregulation of the NOTCH target genes HEY1, HES1 and C-MYC while the epithelial marker E-CADHERIN was down-regulated and SNAIL1 and VIMENTIN were not significantly changed ( Figure 1E). We also examined NOTCH1 and NOTCH4 expression because of their role in mouse breast cancer malignancy [24] and their overexpression in triple-negative breast cancer subtypes [25]. NOTCH4 expression was almost undetectable in MCF-7 cells (Additional file 2, Figure S1A) while NOTCH1 expression was unaffected (Additional file 2, Figure S1B and not shown), as previously reported [26]. Semi-quantitative RT-PCR analysis of various MCF-7-N1ICD expressing clones revealed no variation in JAG1 and TWIST1 expression (Additional file 2, Figure S1B). The lack of response of TWIST1 to N1ICD expression is in agreement with previous findings showing that during developmental EMT, Twist1 is induced by Bmp2 [27] but does not respond to Notch [28]. Immunofluorescence analysis confirmed that forced N1ICD expression caused a reduction in membranous E-CADHERIN staining (Figure 1Fa-i). Cells with strong nuclear N1ICD staining showed mostly nuclear E-CADHERIN expression, suggesting a disassembly of adherens junctions [29], which contrasted with its accumulation in the membrane at the cell-cell contacts of MCF-7 cells that did not express N1ICD (Figure 1Fe). Concomitant to the reduction in E-CADHERIN levels, N1ICD expression endowed MCF-7 cells with increased chemotactic ability towards IGF-1 ( Figure 1G, H), a chemo-attractant for this cell line [30]; some N1ICD-expressing MCF-7 clones showed an increased migratory capacity, even in basal medium (Figure 1Gb, c). To test if the reduction of E-CADHERIN upon N1ICD expression could be extended to other epithelial tumor cell lines, we transfected N1ICD into the HT-29 colon adenocarcinoma cell line and generated stable clones. We chose HT-29 cells because, similarly to the mammary gland, they derive from a tissue in which NOTCH has an oncogenic role [31,32]. Additional file 3, Figure S2A, B shows that HT-29 cells stably expressing N1ICD down-regulate E-CADHERIN expression.

Inducible N1ICD expression in MCF-7 cells leads to E-cadherin down-regulation and increased migratory capacity
To study more precisely the effect of Notch expression in MCF-7 cells, we generated N1ICD-inducible clones using the Tet OFF system, so that gene expression was induced when the antibiotic doxycycline (DOXY) was removed from the culture medium. Figure 2A shows N1ICD-myc staining of three inducible clones (B12, M5 and M20) cultured in the presence (OFF condition) or absence (ON condition) of DOXY. There was some leaky N1ICD expression in the OFF condition in these three clones (especially for B12; Figure 2Aa-f), but 48 h after DOXY withdrawal there was a clear induction of N1ICD expression (Figure 2Ag-l), although not in 100% of the cells. Examination after seven days of induction also revealed a non-homogenous N1ICD expression in these MCF-7 clones (data not shown).
The effect of DOXY retrieval in N1ICD induction was measured by luciferase assay upon transfection of a Hes1 reporter. There was clear reporter activation after N1ICD induction in the different clones studied, especially in clone B12 at 48 h ( Figure 2B). This enhanced N1ICD-induced transcriptional activity correlated with the increase of N1ICD expression in the different clones upon doxycycline withdrawal ( Figure 2C). Concomitantly to N1ICD induction, there was an increase in the migratory capacity of these cells ( Figure 2D, E).
To investigate the possible correlation between the enhancement in the migratory capacity of MCF-7-Tet OFF -N1ICD clones and the reduction of E-CADHERIN expression in these cells, we performed an immunofluorescence analysis of clones M5, M20 and B12 after 20 days of culture in the absence of DOXY. Induction of N1ICD expression coincided with a reduction of membranous E-CADHERIN staining in these clones ( Figure 3A-D). Western blot analysis after 20 days of induction revealed a marked reduction in E-CADHERIN expression in clone B12 that was not so apparent in clones M5 and M20 ( Figure 3E). We then carried out a time-course of N1ICD repression/induction in clone B12 and its effect on E-CADHERIN expression. As Figure 3F shows, there was an inverse correlation between N1ICD and E-CADHERIN expression. MCF-7-Tet OFF -N1ICD cells of clone B12 cultured seven days in the absence of DOXY showed strong N1ICD expression ( Figure 3F). After one day in culture in the presence of DOXY, N1ICD expression was progressively reduced and in parallel, E-CADHERIN expression was increased throughout seven days of culture ( Figure  3F). To the contrary, when cells from clone B12 were grown in the absence of DOXY, N1ICD expression was increased and E-CADHERIN expression was reduced, and this effect was clear after seven days of culture without DOXY ( Figure 3F).
These data indicated that N1ICD expression in MCF-7 cells, either in an inducible or stable manner, leads to a reduction in E-CADHERIN levels, suggesting that these cells began to lose their epithelial phenotype. During EMT there is a progressive cadherin switch, such that E-cadherin expression is reduced and N-cadherin expression is increased [33]. Figure 3G shows a semi-quantitative RT-PCR analysis of N-CADHERIN expression in clone B12. Upon DOXY withdrawal there was a progressive increase in N-CADHERIN expression that was at a maximum after seven days of culture ( Figure 3G, H), suggesting that B12 cells acquired a mesenchymal phenotype. Moreover, we found that N1ICD induction enhanced VIMENTIN expression although it did not change Twist1 mRNA levels (Additional file 4, Figure S3).

NOTCH inhibition reduces the migratory ability of MDA-MB-231 cells
To gain further insights on the role of NOTCH in breast tumor progression and the potential use of NOTCH inhibitors as therapeutic agents against breast cancer, we inhibited NOTCH activity in the adenocarcinoma cell line MDAMB-231 [34], which is more tumorigenic and invasive than the MCF-7 cell line.  Figures 1A and 4A) and E-CADHERIN levels in MDA-MB-231 were reduced compared to those of MCF-7 cells (Additional file 5, Figure S4). We used the γ-secretase inhibitors DAPT [35] and RO4929097 [36], to prevent the generation of NICD and thus inhibit Notch activity. After 48 h we could readily detect a drastic reduction in N1ICD levels in DAPT-and ROtreated cells ( Figure 4A). This effect could also be measured by the reduction in the activity of a CBF1 reporter ( Figure 4B). NOTCH inhibition also resulted in an upregulation of E-CADHERIN paralleled by a reduction in HES1 expression ( Figure 4C). The migratory response of MDA-MB-231 cells towards IGF-1 was reduced around 80% when cells were cultured with DAPT or RO ( Figure  4D, E). MDA-MB-231 cells were also transfected with a dominant-negative version of CBF1 (DN-CBF1); we observed a 25% reduction in the migratory capacity of DN-CBF1-expressing MDA-MB-231 cells compared to controls ( Figure 4F). The weaker inhibition of migration in comparison with the DAPT or RO treatments was likely due to the fact that MDA-MB-231 cells were transiently transfected and only a subpopulation expressed the DN-CBF1 construct (data not shown).

Inducible Notch1 activation in MCF-7 cells stimulates tumor growth in vivo
To examine the tumorigenic ability of MCF-7-N1ICD inducible clones we focused on the B12 clone as it showed a clear phenotypic change as a consequence of N1ICD induced overexpression. Growth of MCF-7 xenografts is estrogen-dependent [37], so we injected MCF-7-Tet-Off-N1ICD B12 cells and added 17α-ethynyl estradiol to the drinking water. When small size tumors were evident in all injection points, mice were divided into two groups. In the first group N1ICD expression was "turned on" (no DOXY) and in the second one, N1ICD expression was turned "Off" by addition of DOXY to the drinking water. As Figure 5A shows, the tumors in which N1ICD expression was turned off did not grow significantly, while tumors in which N1ICD expression was maintained for 12 weeks continued growing and were significantly larger than those generated by control MCF-7 (data not shown) or by un-induced cells ( Figure 5A). After 12 weeks, treatments were switched between both groups of mice ( Figure  5A, arrow), and tumors were monitored for seven weeks more. The result was a reduction in the differences between both groups, leading to similar tumor size ( Figure 5A). Thus, tumor growth was clearly dependent on turning "On" or "Off" N1ICD expression.
The growth curves shown in Figure 5A, B suggest also that estrogens may be a limiting factor in Notch-mediated tumor formation and growth. Western blot analysis of xenografts generated after 12 weeks of N1ICD induction revealed strong N1ICD expression ( Figure 5C).
Next, we analyzed growth of orthotopic tumors formed by the clone B12 transduced with a luciferase-expressing retrovirus to monitor tumor evolution by chemoluminiscence. After cell injection, mice were separated in two groups, receiving (N1ICD off) or not (N1ICD on) DOXY in the drinking water. In agreement with our previous results ( Figure 5A), clone B12 with induced N1ICD expression gave rise to tumors significantly larger than those generated when N1ICD was not expressed ( Figure  5D, E). DOXY treatment did not affect the growth of tumors formed by control MCF-7 cells (Figure 5D, F). These results suggested that N1ICD induction might be directly responsible for MCF-7 tumor formation.

N1ICD expression in the mammary gland leads to tumor formation and reduction in estrogen receptor and Ecadherin expression
To test in vivo the effect of NOTCH activation in the mammary gland we bred the mouse mammary tumor virus (MMTV)LTR-Cre transgenic line (MMTV-Cre) [20] with the Rosa26N1ICD line that expresses the active form of Notch1 (N1ICD) in a conditional manner [21]. With the MMTV-Cre driver we targeted N1ICD expression to the secretory epithelium of the mammary gland of pregnant and lactating females. MMTV-Cre/+;N1ICD/+ double transgenic mice developed normally and were born at Mendelian ratios (data not shown). Adult MMTV-Cre/+; N1ICD/+ females showed high incidence of papillary breast carcinoma (>90%; Additional file 6, Table S1). Figure 6A shows a wildtype lactating breast, with greatly expanded secretory lobules composed of multiple distended acini. Figure 6B, C shows papillary tumors developed in lactating transgenic females (V004 and V006) after three to four rounds of pregnancy and lactation. Tissue architecture was disorganized and large necrotic areas were observed ( Figure 6B). Also, frequent mitotic figures and cytological atypia were common ( Figure 6C, inset). We analyzed the expression of the Notch targets Hes1 and Hey1, estrogen receptor, the myoepithelial marker p63, which stains basal/myopithelial preserved cells in normal or non-malignant breast tissue [38], the epithelial marker E-cadherin and the cell proliferation marker Ki67. Hes1 expression was low and restricted to a few cells in normal breast epithelial tissue ( Figure 6D), like that of estrogen receptor ( Figure 6E), while p63 stained around 30% of cells, as expected for a normal tissue ( Figure 6F). E-cadherin was strongly expressed in the membrane of normal breast epithelial cells ( Figure 6G). Ki67 was expressed only in a few cells in the normal breast ( Figure 6H) and Hey1 was undetectable ( Figure 6I). In the tumors generated in double transgenic MMTV-Cre;   N1ICD mice (Additional file 6, Table S1) three patterns could be distinguished. The first one, represented by female V006 showed moderate but widespread Hes1 expression ( Figure 6J), moderate estrogen receptor staining ( Figure 6K), undetectable p63 expression ( Figure 6L), normal E-cadherin distribution (Figure 6M), and a marked up-regulation of Ki67 and Hey1 expression ( Figure 6N, O) in the tumor area. The second one, exemplified by female V015, showed strong and widely distributed Hes1 expression ( Figure 6P), relatively high estrogen receptor expression ( Figure 6Q) and undetectable p63 expression ( Figure 6R) in the tumor. E-cadherin was found in both membrane and cytoplasm ( Figure 6S), while Ki67 and Hey1 expression was strong ( Figure 6T, U). The third pattern was represented by female V093, in which Hes1 expression was widespread but not as much as in the V015 female ( Figure 6V), moderate estrogen receptor expression ( Figure 6W), very low or absent p63 staining ( Figure 6X), with relatively normal E-cadherin expression ( Figure 6Y) and Ki67 and Hey1 expression was moderate ( Figure 6Z, A'). In general, our data showed that Hes1 and Hey1, as markers of Notch1 activation in epithelial tissues, coexisted with higher estrogen receptor expression. These results are compatible with those obtained with the xenografts of inducible MCF-7-N1ICD cells, whose growth was dependent on N1ICD and estrogens ( Figure 4A, B). This observation, together with the fact that tumors were only observed after three to four pregnancies, suggested that the involvement of Notch in breast tumor formation was strongly dependent on estrogens and that Notch expression could lead to changes in estrogen receptor expression.

Discussion
There is considerable recent interest in understanding how NOTCH signaling affects the development or progression of breast cancer. Notch is critical in mammary gland development, probably by regulating mammary stem cell function [39]. In addition, NOTCH activity has been associated with a number of pro-tumorogenic activities in breast cancer cell lines, and could cause mammary hyperplasia and carcinogenesis in mice [11,[40][41][42]. This evidence strongly pinpoints NOTCH receptors and/or ligands as targets in breast cancer.
Here we used three different in vitro and in vivo models to analyze the impact of NOTCH signaling in the onset and progression of breast tumors. We found a positive association of NOTCH activity with cancer growth or initiation. In agreement with others [43][44][45][46], one of the most consistent observations along our study was the association between NOTCH1 activity and E-CADHERIN down-regulation. First, stable expression of activated Notch (N1ICD) was associated with a reduction and delocalization of E-CADHERIN in most of the MCF-7 cell clones analyzed; second, the data with the inducible MCF-7-N1ICD clone B12 clearly established a causal relationship between active NOTCH1 and reduced E-CADHERIN levels; third, inhibition of endogenous NOTCH activation with DAPT in MDA-MB-231 cells, a highly invasive cell line that expresses high NOTCH1 levels, resulted in an increase of E-CADHERIN expression; and fourth, papillary tumors raised in MMTV-Cre/+; N1ICD/+ transgenic mice expressing high levels of Hes1 also showed, at least, a delocalization of E-cadherin in the epithelium. Repression and/or delocalization of E-CADHERIN is usually associated with adherens junctions disassembly [29] and enhanced cell invasiveness [33]. Concurrently, we observed that N1ICD-induced E-CADHERIN repression correlates with enhanced motility in transwell assays, whereas inhibition of Notch signaling, either by DAPT or RO inhibitors treatment or DN-CBF1 overexpression, reduced the motility of the invasive MDA-MB-231 cells. Collectively, these results indicate that NOTCH1 activation could induce EMT in epithelial tumor cells and, consequently, to favor tumor metastasis [47].
The role of Notch as a critical inducer of EMT has been demonstrated during the formation of the cardiac valve primordium [48]. In this process, Notch activates Snail expression that in turn down-regulates VE-cadherin [48]. We analyzed Snail1 expression in the MCF-7 clones expressing N1ICD, and found no clear correlation among N1ICD expression, E-CADHERIN down-regulation and SNAIL1 expression. We observed a markedly increased HES1 and HEY1 expression in response to NOTCH1 activation that correlated with a reduction in E-CADHERIN expression in both our cellular and animal models. Interestingly, both HES1 and HEY1 have been implicated as part of the hypoxic response associated to breast cancer progression [45].
We also analyzed whether NOTCH1 affects the growth and/or the onset of breast tumors. NOTCH signaling regulates the balance between cell proliferation, differentiation and apoptosis [2] and different reports have demonstrated that NOTCH triggers the proliferation of breast cancer cells [46,49]. In line with these observations, we found that the growth of subcutaneous and orthotopic xenografts produced with MCF-7-B12 cells was boosted after the induction of N1ICD ( Figure 5). This effect was cell autonomous, since silencing of N1ICD-expressing tumors by administration of DOXY stopped the growth of tumors whereas induction of N1ICD by DOXY withdrawal boosted tumor growth, with kinetics compatible with N1ICD induction.
Although induction of N1ICD in MCF-7 fosters tumor growth, this effect was only observed when mice were treated with estrogens; indeed, MCF-7 xenografts did not grow in the absence of estrogens, independently of the induction of N1ICD. These results suggest that N1ICD cooperates with the estrogen receptor (ER) on tumor growth, as recently reported [50]. In agreement with this conclusion, MMTV-Cre/+;N1ICD/+ mice only developed mammary tumors after repeated pregnancies. It is noteworthy to mention that these breast tumors appeared in the lactating gland and regressed after gland involution; the regression was independent of N1ICD activity as determined by the Hes1 expression level. Our results thus resemble those obtained by Kiaris et al. [11], and contrast with the formation of non-regressing mammary neoplasm in transgenic mice expressing the active forms of Notch1, -3 and -4, reported by others [12,41,42]. In summary, our results confirm NOTCH1 as an EMT inducer in breast cancer cells, which may have implications in tumor growth, dissemination and metastasis. The identification of specific factors interacting with NOTCH signaling could thus be relevant to fully understand the role of NOTCH in breast neoplasia.

Conclusions
NOTCH1 activation attenuates E-CADHERIN expression and favors the motility and invasive ability of epithelial human breast cancer MCF-7 cells in vitro. In xenografts and in transgenic mice, NOTCH1 activation caused tumors whose increased growth is NOTCH-and estrogen receptor-dependent. To the contrary, NOTCH inhibition leads to increased E-CADHERIN expression and attenuates the migratory properties of invasive MDA-MB-231 breast cancer cells. Our findings in these mammary tumor models point to NOTCH1 as a potential therapeutic target in breast cancer onset and progression.

Additional material
Additional file 1: Supplementary Materials and Methods.
Additional file 2: Figure S1. NOTCH4 expression in the stable MCF-7 clones E8 and F7 compared to the mock MCF-7 cells, measured by qPCR. cDNA of HUVEC cells was used as positive control of NOTCH4 expressing cells. Data are mean ± SEM of triplicates in two independents experiments (***P<0.001 determined by student's t test).