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

The human breast cancer cell lines MCF-7 (ATCC® HTB-22™) and MDA-MB-231 (ATCC® HTB-26™), and the human colorectal adenocarcinoma cell line HT-29 (ATCC® HTB-38™) were used. For culture conditions see Additional file 1, Supplementary Materials and methods.

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.

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

Total RNA was extracted with Trizol reagent (Life Technologies, NY, USA) and cDNA was synthesized with SuperScript III First Strand kit (Life Technologies, NY, USA). N-Cadherin primers were 5´-CACCCAACATGTTTACAATCAACAATGAGAC-3 (forward) and 5´-CTGCAGCAACAGTAAGGACAAACATCCTATT-3 (reverse) [16]. Commercial β-actin primers were used (Stratagene, La Jolla, CA, USA). Quantitative PCR was performed with Power SYBR Green Master Mix (Applied Biosystems, NY, USA, 4367659) and commercial primers for HEY1, HES1, cMYC, NOTCH1, NOTCH4, SNAI1, ECAD, VIMENTIN and HPRT1 (Sigma, St. Louis, MO, USA) were used.

Hes1-Luc promoter activity [17] was measured in MCF-7 cells expressing N1ICD in a constitutive or inducible manner. The activity of the artificial promoter 10XCBF1 [18] was measured after transient transfection of MDA-MB-231 cells. Briefly, cells were co-transfected with the plasmid containing the promoter 10XCBF1-Luc and pcDNA3-CBF1-VP16 or pcDNA3-DN-CBF1/RBPJK. The plasmid pTK-RL (Promega, Madison, WI, USA)

was also included as a control of transfection efficiency. When indicated, cells were treated for the indicated period of time with DOXY 2 μg/ml or with the γ-Secretase Inhibitors DAPT (N-(N-(3,5-Difluorophenacetyl)-L-alanyl)-S-phenylglycine t-butyl ester, 10 to 50 μM; 565770, Calbiochem, Millipore, MA, USA) and RO4929097 ((2,2-dimethyl-N-(S)-6-oxo-6,7-dihydro-5H-dibenzo(b,d)azepin-7-yl)-N'-(2,2,3,3,3-pentafluoro-propyl)-malonamide)), 10 to 20 μM; S1575, Selleckchem, Houston, TX, USA) for 48 h. After transfection cells were lysed with passive lysis buffer (Promega, Madison, WI, USA) and firefly and renilla luciferase were measured with the "Dual-luciferase reporter assay" (Promega, Madison, WI, USA). The activity in MCF-7 clones or in MDA-MB-231 treated cells was referred to the activity in control cells or cells transfected with the empty vector (pcDNA3).

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

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

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.

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

All animal procedures were approved by the Institutional Committee for the Care and Use of Laboratory Animals of the Centro Nacional de Investigaciones Cardiovasculares (CNIC, Madrid, Spain) and Centro Nacional de Biotecnología (CNB-CSIC, Madrid, Spain). Animal procedures conformed to EU Directive 2010/63EU and Recommendation 2007/526/EC, regarding the protection of animals used for experimental and other scientific purposes, enforced in Spanish law under Real Decreto 1201/2005.

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 × 10⁶ 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 (a² × 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 × 10⁶ 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.

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.

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 myc-tagged 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 comparison with control, mock-transfected MCF-7 cells (Figure 1D). qPCR analysis revealed a marked up-regulation 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.

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).

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. MDA-MB-231 cells endogenously expressed N1ICD (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 RO-treated 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 up-regulation 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).

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.

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.