Introduction
NOTCH activation has been implicated in several malignancies; notably T-cell acute lymphoblastic leukemia, chronic lymphocytic leukemia, glioblastoma, and breast cancer [
1‐
11]. Overexpression of NOTCH receptors has been implicated in ductal carcinoma
in situ (DCIS) and invasive breast cancer [
12,
13], and high levels of the NOTCH ligand JAG1 appear to predict a poor overall survival [
14]. High NOTCH1-receptor levels have been linked with basal-like, triple-negative (estrogen receptor-, progesterone receptor-, and HER2-negative) breast cancer, and NOTCH1 levels correlate with abbreviated survival [
15]. More recently, silencing of Lunatic Fringe, the glycosylase that regulates NOTCH1 ligand activity, has been observed in patients with basal-like breast cancer, and increased levels of intracellular NOTCH1 are detected in these patients' cells [
16]. NOTCH1 activity levels have also been shown to correlate with the development of resistance to conventional as well as to targeted therapies [
17‐
20], leading us to hypothesize that NOTCH1 may contribute to therapeutic resistance and disease recurrence by regulating breast tumor-initiating cell activity.
NOTCH pathway activation is triggered on ligand-receptor interaction. Mammals possess four NOTCH receptors (NOTCH 1-4) and five ligands (JAG1, 2, DELTA-like (DLL) 1, 3, and 4). Ligand binding stimulates two sequential proteolytic cleavages; the first in the extracellular domain mediated by metalloproteases of the ADAM family, and the second within the transmembrane domain mediated by the gamma-secretase complex. The second cleavage allows the release and translocation of the intracellular domain of NOTCH into the nucleus, where it associates with the
CBF1/RBP-Jκ/
Suppressor of Hairless/
LAG-1 (CSL) repressor and on the recruitment of co-activators Mastermind-like 1 (MAML1) and CBP/p300 induces expression of NOTCH target genes, including HES1, HEY2, DELTEX1, and c-MYC [
21]. Gamma secretase inhibitors (GSIs) have been shown to inhibit Notch1 and to have antileukemia activity
in vivo [
22‐
24].
Constitutive Notch1 signaling in the normal mouse mammary stem cell (MaSC) has been shown to stimulate differentiation toward a luminal fate, whereas suppression of Notch signaling in MaSC via
CSL knockdown results in the expansion of the MaSC compartment [
25,
26]. These studies implicate Notch1 pathway activation in mouse luminal progenitor expansion and differentiation. NOTCH pathway activation has also been shown to enhance human mammosphere formation, which likely reflects NOTCH pathway effects on the human mammary stem or progenitor cells [
27].
In addition to the Notch receptor family, the gamma-secretase complex regulates the expression of ErbB4, CD44, and E cadherin, cell-surface receptors known to contribute to tumor growth, migration, and invasion [
28‐
30]. Thus, experiments that use GSIs to determine the effect(s) of Notch inhibition on tumor growth likely affect the stability of other substrates relevant to mammary gland transformation. Moreover, GSI studies fail to reveal which Notch receptor family member mediates the effects on tumor growth/survival.
To determine the specific effects of NOTCH1 activation/inhibition on bulk mammary tumor growth and on mammary tumor-initiating cells, we generated a mouse mammary tumor model in which human intracellular NOTCH1 expression is doxycycline regulated. Consistent with previous reports [
25,
26,
31‐
33], we demonstrated that NOTCH1 signaling stimulates luminal cell fate and results in luminal lineage transformation.
In vivo limiting-dilution analysis reveals that only a small percentage (~1/3,000) of NOTCH1-driven mammary tumor cells are capable of transplanting disease, revealing that mammary tumor-initiating cells contribute to disease pathogenesis in this model. We also demonstrated that NOTCH1 signaling is required for mammary tumor-initiating cell activity, as NOTCH1 inhibition results in rapid mammary tumor regression and delays and, in some cases, prevents disease recurrence. By using gene-expression profiling, we identified the embryonic stem cell transcription factor
Nanog as a novel NOTCH1-regulated gene in mammary tumor cells. These data demonstrate that NOTCH1 activation stimulate luminal lineage development and implicate NOTCH1 in the regulation of mammary tumor-initiating activity.
Mice histopathology and immunohistochemistry
The MMTV-tTA (C57BL/6) and tet-op-Notch
IC (FVB/N) mice were described previously [
34,
35] and generously provided by D. Tenen and A.J. Capobianco, respectively. Mice were maintained in mating pairs of MMTV-tTA/TOP-ICN1 females and MMTV-tTA males, and females were monitored weekly for signs of disease. Mice were killed when total tumor volume surpassed 1,000 mm
3, as determined by external measurement by using calipers. Tumors were fixed in 10% formalin for 4 hours at room temperature or overnight at 4°C, then transferred to 70% ethanol and maintained at 4°C until mounting in paraffin and sectioning. Sections were stained with antibodies against mouse keratin 5 (PRB-160P; Covance, Princeton, NJ), keratin 8/18 (GP11; Progen Biotechnik, Heidelberg, Germany), keratin 14 (Clone LL002; Thermo Scientific, Waltham, MA) or estrogen-receptor alpha (sc-8005; Santa Cruz Biotechnology, Santa Cruz, CA). All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Massachusetts Medical School.
Establishment of tumor-derived cell lines
Primary mouse mammary tumors were minced with a razor blade and digested in DMEM/F12 (1:1) media supplemented with 5% fetal bovine serum (FBS) and 2 mg/ml collagenase (Gibco, Grand Island, NY) for 2 hours at 37°C. Samples were spun down and washed 5 times in PBS supplemented with 5% FBS and then plated onto 10-cm collagen-coated plates (BD Biosciences, San Diego, CA) in DMEM/F12 (1:1) media supplemented with 2% FBS and penicillin/streptomycin. Cell clusters were left undisturbed for 3 days, with subsequent media changes every 3 days, gradually increasing FBS concentration to 10%. When confluent, cells were passaged at a 1:2 or 1:3 dilutions after a 5-minute incubation with Versene (Gibco, Grand Island, NY) and maintained on standard tissue-culture plates.
Mammary fat-pad transplants
For limiting-dilution studies, primary mammary tumors were digested as previously described [
36,
37]. In brief, tumors were sequentially digested at 37°C in 300 U/ml collagenase (Gibco, Grand Island, NY) plus 100 U/ml hyaluronidase (Sigma, St. Louis, MO) (2 hours), 0.25% trypsin (Gibco, Grand Island, NY) (2 minutes), and 5 mg/ml Dispase II (Roche, Indianapolis, IN) plus 0.1 mg/ml DNaseI (Sigma, St. Louis, MO) (5 minutes). Cells were filtered through a 40-μm mesh, counted, and enzymatic digestion was repeated until the suspension was >95% single cells. Serial dilutions were resuspended in 35 μl gelatinous protein mixture (Matrigel; BD Biosciences, San Diego, CA) and injected into the thoracic mammary fat pads of nude mice. For tumor-derived cell-line tumorigenicity assays, cultured cells were lifted by using Versene (Gibco, Grand Island, NY), followed by trypsin-EDTA treatment and filtering through a 40-μm mesh. Approximately 10
6 cells were resuspended in 35 μl Matrigel and injected into the thoracic mammary fat pads of nude mice.
Tumorsphere cultures
Single-cell suspensions (>95% single cells) were generated from primary mammary tumors as described earlier. Cells were then plated in 3 ml defined tumorsphere media [
38] plus 0.5% methylcellulose (R&D Systems, Minneapolis, MN) per well of a six-well ultra-low-attachment plate (Fisher Scientific, Pittsburgh, PA), at a concentration of 20,000 cells/ml. For doxycycline-treated samples, 2 μg/ml doxycycline was added to the culture media at the time of plating.
Flow cytometry
Tumor samples, thoracic and inguinal mammary fat pads from nulliparous females, or tumor-derived cell lines were subjected to enzymatic digestion to create a single-cell suspension as described earlier. Antibodies against mouse antigens were purchased from BD Pharmingen (San Diego, CA) unless otherwise noted, and included Ter-119-PE, CD31-PE, CD45R-PE, CD61-FITC, CD24-biotin, streptavidin-APC, and CD29-PE-Cy7 (eBioscience, San Diego, CA). Cells were stained in PBS at 4°C for 25 minutes and analyzed live. For cell-cycle analysis, cells were fixed in 70% ethanol, stained with propidium iodide (PI), and analyzed for DNA content.
Microarray analysis
Tumor-derived cell lines 8534 and 8542 were left untreated or treated with 2 μg/ml doxycycline for 24 hours. Cells were collected by scraping, and total RNA was isolated by using Trizol (Invitrogen, Grand Island, NY). After real-time PCR validation of NOTCH1 target-gene modulation, RNA samples were further purified by using the RNAeasy Mini kit (Qiagen, Valencia, CA) and hybridized to Affymetrix mouse genome 430A2.0 arrays (Affymetrix, Santa Clara, CA, USA). Raw data were processed with MAS5 analysis, and genes showing a >2.0-fold change in both cell lines were considered targets of interest. The data from these arrays have been deposited in the NCBI Gene Expression Omnibus and are accessible through GEO Series accession number GSE34146.
Quantitative RT-PCR
Total RNA from cells was extracted by using Trizol (Invitrogen, Grand Island, NY). cDNA was prepared with the Superscript First Strand Synthesis kit (Invitrogen, Grand Island, NY), and PCR was carried out with SYBR Green (Qiagen, Valencia, CA). The following primers were used in this study:
hey1, 5 -TGAGCTGAGAAGGCTGGTAC-3 (Forward) and 5 -ACCCCAAACTCCGATAGTCC-3 (Reverse);
deltex1, 5 -TGCCTGGTGGCCATGTACT (Forward) and 5 -GACACTGCAGGCTGCCATC-3 (Reverse); β
-actin, 5 -CGAGGCCCAGAGCAAGAGAG-3 (Forward) and 5 -CGGTTGGCCTTAGGGTTCAG-3 (Reverse);
c-myc, 5 -CTGTTTGAAGGCTGGATTTCCT-3 (Forward) and 5 -GTCGTGGCTGTCTGCGG-3 (Reverse);
hes1, 5 -AAGACGGCCTCTGAGCACA-3 (Forward) and 5 -CCTTCGCCTCTTCTCCATGAT-3 (Reverse);
nanog, 5 -TCTTCCTGGTCCCCACAGTTT-3 (Forward) and 5 -GCAAGAATAGTTCTCGGGATGAA-3 (Reverse). The
nanog primer set (PrimerBank ID 31338864a1) was obtained from PrimerBank [
39].
Western blotting
Protein was isolated from cells collected by using Versene, washed in PBS, and lysed in radioimmunoprecipitation assay (RIPA) buffer containing protease inhibitor tablets (Roche, Indianapolis, IN). Fifteen to twenty-five micrograms of total protein was resolved via 9% sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE), as previously described [
40]. Blots were probed with antibodies against intracellular NOTCH1 (provided by Jon Aster), active NOTCH1 (2421; Cell Signaling Technology, Danvers, MA), Nanog (AB9220; Millipore, Billerica, MA), cytokeratin 8/18 (GP11; Progen Biotechnik, Heidelberg, Germany), caspase-3 (Cell Signaling, Danvers, MA), and α-tubulin (T5168; Sigma, St. Louis, MO) or Erk1/2 (Cell Signaling, Danvers, MA) to control for equal loading. For MDA-MB-231 study, cells were left untreated or treated with the gamma-secretase inhibitor (GSI) Compound E (Axxora, Farmingdale, NY) at 10 μ
M for the times noted, before cells were collected, and analyzed as described earlier.
Mammary fat pad whole mounting
Inguinal fat pads were isolated from mice administered doxycycline-treated sugar water (10 μg/ml) for various time periods, spread on glass slides, and fixed in glacial acetic acid/ethanol (1:3) overnight. Samples were washed in 70% ethanol for 15 minutes, rinsed in distilled water for 5 minutes, and stained overnight in carmine alum solution. Samples were then dehydrated and transferred to xylene overnight for delipidation. Whole mounts were briefly air dried, and coverslips were mounted by using Permount (Fisher Scientific, Pittsburgh, PA).
MTT assay
MTT cell-viability assays were performed as previously described [
22], with the following modifications. In brief, approximately 10
4 cells/200 μl of a cell suspension were plated in a 96-well flat-bottom plate, ~16 hours before treatment. Cells were then left untreated or treated with 2 μg/ml doxycycline. After 72 hours, 20 μl of a 5-mg/ml 3-(4,5-dimethylthiazol-2-y1)-2,5-diphenyl tetrazolium bromide (MTT) solution (Sigma-Aldrich, St. Louis, MO) was added and incubated for 4 hours at 37°C. Media was then removed, and the reagent was solubilized with 100 μl dimethyl sulfoxide (DMSO) (Sigma-Aldrich, St. Louis, MO) and incubated for 10 minutes at room temperature. Plates were then analyzed at A595 wavelength. Data are plotted as absorbance and are the average of five independent experiments. The
P values were calculated by using a two-tailed distribution and paired Student
t test.
Tumorsphere immunofluorescence
For detection of Keratin 8/18 and Nanog protein expression, primary mammary tumor cells were grown under mammosphere culture conditions for 7 days. Spheres were collected by pipette under a dissection microscope, pooled, washed, and digested in 0.25% trypsin (Gibco, Grand Island, NY) at 37°C for 5 minutes, and the resulting cell suspension was cytospun onto coverslips. In parallel, tumor-derived cell line 8542 was plated onto glass coverslips and allowed to grow for 48 hours before processing. Cell line and primary samples were fixed in 4% paraformaldehyde and permeabilized for 10 minutes in phosphate-buffered saline (PBS) containing 0.2% Triton X-100. Cells were washed with PBS containing 0.02% Triton X-100 and 10% FBS, followed by incubation with α-Keratin 8/18 (Progen Biotechnik, Heidelberg, Germany) and α-Nanog (Millipore, Billerica, MA) antibodies or IgG controls (Santa Cruz, Santa Cruz, CA) for 1 hour at room temperature. Cells were stained with FITC- or rhodamine-conjugated secondary antibodies (Santa Cruz, Santa Cruz, CA), and coverslips were mounted with Permount (Fischer Scientific, Pittsburgh, PA) and photographed under ultraviolet illumination at a magnification of 600X.
For detection of cell-surface marker CD61, tumorspheres were grown, collected, and digested with trypsin as described earlier to break up the spheres into loose aggregates of cells. Cells were resuspended in PBS plus 5% FBS and stained with CD61-FITC (1:75, eBioscience, San Diego, CA) or IgG-FITC at 4°C for 1 hour in the dark, and washed. Cells were then cytospun onto glass coverslips and mounted with DAPI-containing mounting medium and photographed under ultraviolet illumination at a magnification of 600X.
Discussion
Notch1 has been shown to promote commitment of mouse mammary stem cells along the luminal lineage [
25,
58]. Consistent with these and other studies [
26,
31‐
33], we show that constitutive expression of intracellular NOTCH1 in the developing mouse mammary gland stimulates luminal fate, ultimately resulting in transformation of the mammary gland. The mammary tumors predominantly express the luminal lineage marker keratin 8/18. Interestingly, in our model expression of human intracellular NOTCH1 in the developing mouse mammary gland did not result in induction of diverse tumor types that regressed upon weaning. Nor did the transgenic females exhibit any difficulty nursing their young. This is in contrast to transgenic models that constitutively express mouse ICN1 driven by the Mouse Mammary Tumor Virus (MMTV) LTR. These mice are unable to nurse their young and they develop lactation-dependent papillary tumors that regress upon involution [
31,
32]. The reasons for the phenotypic differences could reflect transgene expression levels or the timing of transgene induction since our model is doxycycline-regulated. Alternatively, human ICN1 may not interact with a mouse co-factor essential for lactation.
Using an in vivo limiting-dilution assay, we provide evidence that NOTCH1-transformed mammary tumors are functionally heterogeneous and estimate the frequency of mammary tumor-initiating cells to be approximately 1/3000 cells. We demonstrate that doxycycline treatment or NOTCH1 inhibition in vivo prevents disease recurrence in 4 of 6 mice examined. However, disease recurred within 21 days in 2 tumor bearing mice treated with dox, suggesting that NOTCH1 inhibition in these tumors was not sufficient to eliminate the tumor-initiating cells. These relapsed mammary tumors may contain increased numbers of mammary tumor-initiating cells and/or harbor genetic changes that render the tumors NOTCH1 independent.
Consistent with the in vivo limiting-dilution analyses, a subpopulation of NOTCH1 transformed mammary tumor cells grow in an in vitro tumorsphere assay and importantly, doxycycline treatment significantly reduces sphere number and size. The tumorsphere assays revealed that NOTCH1 is required both for the initiation and maintenance of tumorspheres in vitro and potentially for mammary tumor-initiating activity in vivo.
GSI treatment of ERB2-induced mouse mammary tumors reduced tumorspheres
in vitro and interfered with the ability of the mammary tumor-initiating cells to induce disease in immunodeficient mice [
59]. These studies are consistent with our findings and collectively suggest that NOTCH inhibitors may target mammary tumor-initiating cells driven by other oncogenes and not be limited to mammary tumors that exhibit NOTCH pathway activation.
NOTCH pathway activation has also been implicated in human mammary tumor-initiating cell biology. GSI treatment or treatment with an anti-NOTCH4 monoclonal antibody significantly reduces human tumorsphere formation
in vitro [
60]. Studies in the human breast cancer cell lines MCF7 and MDA-MB-231 show that NOTCH1 or NOTCH4 silencing reduces tumorsphere formation and inhibits tumor growth
in vivo; however, NOTCH4 suppression appears to have the greatest inhibitory effect [
61]. Thus, its possible that NOTCH4 is the relevant NOTCH receptor in human breast cancer-initiating cells.
To identify NOTCH1-regulated genes that might mediate mammary tumor-initiating cell activity, we applied transcriptional profiling to two mammary tumor cell lines in the absence/presence of doxycycline. We found the expression of several NOTCH1-regulated genes such as
Hes1, Hey1, Deltex1 and
c-Myc significantly reduced upon doxycycline treatment. In addition to these target genes, NOTCH1 activation stimulates expression of embryonic stem (ES) cell pluripotency transcription factor
Nanog. The
Nanog-
Oct4-
Sox2 (NOS) transcription factors activate self-renewal and inhibit differentiation in human and mouse ES cells and the NOS signature is enriched in claudin-low and basal-like breast cancer subtypes [
62‐
64]. Consistent with these findings, we show that treatment of the ER-negative, basal-like human breast cancer cell line MDA-MB-231 with a gamma-secretase inhibitor reduces intracellular NOTCH1 and NANOG protein levels.
Like CD61, Nanog expression was not detected in the primary mouse mammary tumor tissue but was readily observed in the nuclei of the CD61-positive mammary tumor cell lines and tumorspheres. These data suggest that NOTCH1 regulation of Nanog may be cell-type or developmental stage specific. Thus, NOTCH1 may induce
Nanog expression in luminal progenitors and mammary tumor-initiating cells but not in the bulk differentiated tumor cells. Although CSL sites are present in the mouse
Nanog regulatory region, we were unable to demonstrate NOTCH1 or Mastermind-like 1 recruitment to the mouse
Nanog locus, leading us to speculate that NOTCH1 may indirectly regulate
Nanog expression in mammary tumor- initiating cells. Consistent with this hypothesis, ChIP-seq analysis has suggested that NOTCH1 binds the genome in association with the zinc finger protein ZNF143 [
65] and Nanog expression in mouse ES cells has been linked to
Znf143 regulation [
66]. Thus, Notch1 and Znf143 may co-regulate
Nanog expression in mammary tumor-initiating cells. Consistent with our findings in the mouse, siRNA studies have demonstrated that OCT4 and NANOG expression are required for human breast tumor-initiating activity [
62].
Authors' details
1MS, NH, and MK are members of the Department of Cancer Biology at the University of Massachusetts Medical School.
2RS is a member of the Program in Gene Function and Expression, University of Massachusetts Medical School. Worcester, MA, USA.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MS carried out all the experiments with assistance from NH. RS assisted with the chromatin immunoprecipitation experiments. MS and MK designed the experiments, analyzed the data, and wrote the manuscript. All authors read and approved the manuscript.