Introduction
Epithelial cancers, such as breast cancer, are being more frequently identified at the early pre-invasive stage of tumor development [
1]. These pre-invasive mammary lesions originate from the luminal epithelial cells that line the ducts and lobules of the mammary glandular epithelium and have a disrupted epithelial architecture characterized by hyperproliferative cells occupying the normally hollow luminal spaces of the ducts and lobules [
2,
3]. The amplification and overexpression of the receptor tyrosine kinase ErbB2 is observed in approximately 50% of pre-invasive lesions; however, in most cases, the genetic and epigenetic abnormalities that promote pre-invasive tumor growth are poorly understood [
4].
Since such a wide range of molecular perturbations can induce and enhance tumor growth, there are probably shared molecular signaling modules that integrate biochemical signals from the suite of genetic contexts found in epithelial tumors [
5]. To explain how normal cells become tumorigenic, a molecular framework that underpins the pre-invasive stage of tumor growth must be established. Such a molecular framework can assist in the identification of patients amenable to targeted therapeutics, in the development of novel therapeutics to treat pre-invasive cancer, and, in the future, in the introduction of preventative treatment [
6]. Attempts to identify the core signaling modules that promote these pre-invasive growth characteristics through the analysis of genetic abnormalities and gene expression patterns of pre-invasive tumor lesions have to date been unsuccessful [
7‐
9].
The Raf–MEK1/2–ERK1/2 mitogen-activated protein kinase signal transduction module transmits extracellular and oncogenic stimuli, resulting in cellular responses [
10]. In this module, Raf isoforms phosphorylate their primary substrates, the dual-specificity kinases MEK1/2. Once activated, MEK1/2 phosphorylate ERK1/2 on tyrosine and threonine residues, substantially increasing ERK1/2 catalytic activity [
11]. The Raf–MEK1/2–ERK1/2 module is activated by growth factors and proteins overexpressed in human breast cancer epithelium, by cytokines and hormones produced by fibroblasts and macrophages in the mammary stromal compartment, and by increased tissue stiffness observed during tumor progression [
10,
12]. In addition, the sequencing of breast cancer patient genomes suggests that infrequent mutations may drive tumor progression through known signaling pathways, such as the Raf–MEK1/2–ERK1/2 cascade [
5]. Considering the array of stimuli known to activate the Raf–MEK1/2–ERK1/2 module, it may be complicit in tumorigenesis in a variety of contexts.
Consistent with a role for the Raf–MEK1/2–ERK1/2 module in mammary carcinogenesis, ERK1/2 are activated in primary breast cancer tissue and in associated lymph node metastases [
13,
14]. The activation of ERK1/2 is not associated with a specific genetic signature, however, as ERK1/2 is active in ER-positive breast cancer, HER2-positive breast cancer and in triple-negative breast cancer [
15]. ERK1/2 phosphorylate transcription factors, kinases, proteases and non-enzymatic regulatory proteins, thus potentially integrating the Raf–MEK1/2–ERK1/2 module into a range of cellular activities associated with tumorigenesis [
16]. Accumulating evidence, however, has shown that results obtained in one cell type should not be generally applied across all classes of cancer without experimental validation [
6]. For example, the K-Ras2 oncogene has distinct effects on tumor progression depending on both the cell type of origin and the genetic context in which it is mutated [
6]. In addition, extrapolating the role of protein kinases in promoting breast cancer progression based on either their known substrate profile or biological behaviors induced in two-dimensional culture models has proven to be unreliable [
17,
18]. For example, the chemically induced homodimerization of the epidermal growth factor receptor (EGFR) is sufficient to induce focus formation in Rat 1 cells and the proliferation of MCF-10A mammary epithelial cells in monolayer cultures [
17,
19]. EGFR homodimerization of EGFR, however, is not sufficient to induce the proliferation of differentiated MCF-10A cells grown in organotypic culture [
17]. Considering the uncertainty in predicting the response of cells to the activation of a signaling pathway, determining the response of differentiated mammary epithelial cells to Raf–MEK–ERK activation can better define the early events of mammary tumorigenesis.
Three-dimensional organotypic culture models have been indispensable tools in deciphering the molecular and cell biological mechanisms underlying the disruption of differentiated epithelial architecture that is characteristic of pre-invasive mammary epithelial lesions. In organotypic culture models, individual mammary epithelial cells plated on reconstituted basement membrane proliferate to form a hollow sphere of polarized, growth-arrested cells (termed acini), thus recapitulating the salient features of the mammary gland [
20,
21]. Since the mammary epithelial cells differentiate and form a hollow monolayer of cells, organotypic cultures provide a more accurate reconstitution of the biochemical and cell biological growth restraints found in mammary glandular epithelium than is achieved using traditional two-dimensional cell culture models [
22]. Once cells become proliferative, they are confronted with similar local environmental selection pressures to those found during tumorigenesis. Namely, cells are required to become resistant to cell death triggered by the induction of either apoptosis or autophagy when cells enter the luminal space [
23,
24]. Organotypic culture models therefore provide both the biochemical signaling barriers that must be overcome for initial proliferation to occur, and the microenvironmental context in which pre-invasive tumor cells must survive and propagate.
We have previously developed a method for imaging cells in Raf:ER-induced acini at single-cell resolution through imaging a histone–green fluorescence protein (GFP) correct fusion protein, H2B-GFP [
25]. Using this unbiased discovery approach we have found that Raf:ER activation induces a disruption of epithelial architecture through promoting a non-invasive form of motility, cell proliferation and the survival of cells in the lumen. These findings suggest that ERK1/2 activation can promote the early events of tumorigenesis and that the induction of motility can, in principle, occur before tumor cell invasion. To determine how ERK1/2 signaling promotes the early events of tumorigenesis we have examined the intracellular signaling pathways that promote proliferation, cell survival and motility in response to ERK1/2 activation in mammary epithelial acini.
Materials and methods
Cell culture and reagents
MCF-10A human mammary epithelial cells were obtained from the American Type Tissue Culture Collection. Cells were cultured in DMEM/F12 (Gibco, Invitrogen, Carlsbad, CA, USA) supplemented with 5% horse serum (Gibco, Carlsbad, CA, USA), 10 μg/ml insulin (Research Diagnostics, Inc., Concord, MA, USA), 20 ng/ml epidermal growth factor (Research Diagnostics, Inc.), 500 ng/ml hydrocortisone (Sigma, St Louis, MO, USA), 100 ng/ml cholera toxin (Calbiochem, San Diego, CA, USA) and cyprofloxacin (Cellgro, Carlsbad, CA, USA). The growth-factor-reduced Matrigel (BD Biosciences, Franklin Lakes, NJ, USA) used in these experiments had protein concentrations between 10 and 12 mg/ml. 4-Hydroxytamoxifen (4-HT), LY294002, U0126 and AG1478 were from Calbiochem. Antibodies recognizing Ki-67 (Zymed, San Francisco, CA, USA), c-Fos, estrogen receptor alpha and cyclin B1 (Santa Cruz, Santa Cruz, CA, USA), phosphorylated AKT (S473), cleaved caspase 3, Bim and Bim (IF specific) (Cell Signaling, Beverly, MA, USA), p27 (BD Transduction Labs) and phosphorylated ERK2 (T183, Y185) (Sigma) were used. Secondary antibodies for immunofluorescence staining were labeled with Alexa fluor 488, 568 and 647 (Molecular Probes, Invitrogen, Carlsbad, CA, USA).
Three-dimensional morphogenesis assay and cell lines
MCF-10A cells plated in eight-well chamberslides (Falcon, San Jose, CA, USA) were cultured as described previously [
25]. The vector pBABE-Raf:ER was a gift from Michael White and Ron Bumeister (University of Texas Southwestern Medical Center, Dallas, USA), pBABE-GFP-Raf:ER was a gift from Martin McMahon (University of California San Francisco, USA) and pCLNRX-H2B:GFP was a gift from Ee Tsin Wong and Geoff Wahl (Salk Institute, La Jolla, CA, USA). VSVG-pseudotyped virus was generated by transfecting HEK293 cells stably expressing Gag and Pol with VSVG and pBABE-Raf:ER or pCLNRX-H2B:GFP. Cells were cultured in 500 ng/ml puromycin or 400 μg/ml G418 to create stable pools of pBABE-Raf:ER MCF-10A cells or pCLNRX-H2B:GFP MCF-10A cells. The GFP-Raf:ER MCF-10A cells did not undergo drug selection.
Immunoblot analysis and immunofluorescence staining
The acini were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors as described elsewhere [
25], and protein levels were normalized using Cyto-tox One (Promega, Madison, WI, USA) according to the manufacturer's instructions. Immunoblots were visualized using an Odyssey infrared scanner (LI-COR, Lincoln, NE, USA). Cultures were fixed in 2% formalin (Sigma Aldrich, St. Louis, MO, USA) for 20 minutes and were permeabilized with 0.5% Triton X-100 in PBS for 10 minutes at room temperature. Immunostaining was performed as described previously [
25].
Images were acquired on a Leica SP2 AOBS confocal microscope (Bannockburn, IL, USA) using Leica software in TIFF format. Images were arranged using Adobe Photoshop 7.0 and Keynote, and are representative of at least three independent experiments. For quantification of immunofluorescence images, either three or more Ki-67-positive cells per acinus or two or more phospho-AKT-positive cells per acinus were used as thresholds, as has been previously reported [
17,
24,
26]. These thresholds reproducibly distinguish between control acini with normal architectures and Raf:ER-induced acini with disrupted architectures from experiment to experiment.
Real-time imaging
Organotypic cultures were grown in eight-well chambered coverglass slides (Thermo Fisher Scientific, Pittsburgh, PA, USA) as described above and previously [
25]. Cultures were imaged with a spinning disk confocal scanhead (QLC100; Yokagowa, Newnan, Georgia, USA) enclosed in a 37°C chamber supplemented with humidified carbon dioxide (Solent, Segensworth, UK) and a CCD camera (C9100-02 EM-charged coupled device; Hamamatsu, Hammamatsu, Japan). Images were acquired with a 40×/0.60 objective (HCX Plan Fluor; Leica) using SimplePCI software (Compix, Hammamatsu, Japan) and were analyzed with Imaris software (Bitplane, Zurich, Switzerland). At least six different
x,
y coordinates with three or more z-slices over 20 μm for each condition were imaged in parallel for three independent experiments.
Discussion
We have demonstrated that the persistent activation of the Raf–MEK1/2–ERK1/2 mitogen-activated protein kinase module promotes the development of pre-invasive mammary lesions from differentiated epithelium in organotypic culture. This finding indicates that persistent ERK1/2 activation in luminal epithelial cells might contribute to the development of mammary tumors. It is known that ERK1/2 is activated by oncogenes, such as ErbB2; however, our results demonstrate that persistent activation of ERK1/2 can induce growth and survival in the absence of receptor tyrosine kinase mutation or overexpression. It is possible that unidentified genetic abnormalities, or combinations of abnormalities, promote activation of ERK1/2 in mammary epithelium. This conclusion is supported by the observation that persistent ERK1/2 activation is found in a wide range of patient-derived mammary tumor cell lines, many of which do not harbor amplified expression of ErbB2 [
43] and the sequencing of breast cancer tumor genomes [
5]. Furthermore, by uncoupling the activation of the Raf–MEK1/2–ERK1/2 module from a specific oncogenic lesion, our results suggest that the inappropriate expression of growth factor receptor ligands could promote tumorigenesis through the sustained stimulation of ERK1/2.
The number of ductal carcinoma
in situ (DCIS) cases identified in the United States annually has risen from 4,800 in 1983 to over 50,000 today [
4]. After identification, DCIS lesions are surgically removed with a breast-conserving excision and patients may undergo either a course of adjuvant therapy targeted to block the action of the hormone estrogen or receive gamma irradiation to kill the remaining proliferating tumor cells [
4]. The risk of a recurrent growth developing 15 years after lumpectomy is between 16 and 19%, and thus patients are required to undergo continual surveillance [
44]. One-half of recurrent growths are invasive breast cancer, which is more difficult treat and pose a much greater threat of metastasis [
44‐
46]. It is likely that early-stage epithelial tumors, such as DCIS, are susceptible to new and more efficacious diagnostic tests and forms of therapy. Our results demonstrate that ERK1/2 activation is sufficient to promote proliferation and cell survival in the lumens of mammary epithelial acini, which are characteristic behaviors required for recurrent tumor growth after lumpectomy.
These findings warrant further investigation of the activity level of the ERK1/2 signaling pathway in patient samples to determine the frequency of ERK1/2 activation in early-stage breast cancer and whether there is a correlation between ERK1/2 activation and recurrent growth after lumpectomy. In the event that a positive connection between ERK1/2 activation and recurrent growth is revealed, there are a number of inhibitors of MEK1/2, the direct upstream activators of ERK1/2, that have undergone various stages of in clinical testing and could be tested as adjuvant therapy in the clinic [
47].
Bim and c-Fos of targets of ERK1/2 signaling in differentiated mammary epithelial acini
We have identified c-Fos and Bim as downstream effectors of ERK1/2 that can contribute to the proliferation and survival of differentiated mammary epithelial cells in the lumens of epithelial acini. These targets of ERK1/2 signaling are worthy of investigation in patient samples to determine whether ERK1/2 signaling promotes early-stage human breast cancer progression through similar mechanisms to those observed in organotypic culture.
In addition to promoting c-Fos expression and Bim degradation, ERK1/2 directly phosphorylates a vast array of proteins that are also likely to contribute to the observed phenotypes. For instance, p90 RSK1/2 are activated by direct ERK phosphorylation on serine 363, in the linker between the N-terminal and C-terminal catalytic domains, and threonine 573, in the activation loop of the C-terminal catalytic domain, resulting in autophosphorylation at serine 380 and creation of a docking site for PDK1, which then phosphorylates serine 239 [
48,
49]. Once activated, p90 RSK1/2 promotes transcription through direct phosphorylation of transcription factors including the serum response factor and c-Fos [
50‐
52]. The transcriptional co-activator CREB binding protein is also a target for p90 RSK [
53]. Furthermore, p90 RSK can promote cell survival through the phosphorylation and inactivation of the Bcl-2-associated death promoter protein and the activation of the mammalian target of rapamycin protein by phosphorylating and inactivating tuberous sclerosis complex 2 [
54,
55].
This is just one of many examples of the molecular mechanisms by which ERK1/2 can promote pre-invasive tumor growth. The identification of the ERK1/2 substrates that are required to promote cell growth and survival will further provide a molecular framework with which to understand pre-invasive tumor development.
PI-3K activity is necessary for ERK1/2-stimulated proliferation
We have shown that the persistent activation of ERK1/2 increases the activity of the parallel PI-3K–AKT signaling module, but in a stochastic manner in cells within an acinus. The activity of the PI-3K, and possibly AKT, is necessary for the progression of MCF-10A cells through the cell cycle, as has been previously demonstrated in fibroblasts [
42]. The identity of the signaling circuit connecting ERK1/2 to PI-3K in epithelial organotypic culture is not known. Interestingly, autocrine activation of EGFR was not necessary for AKT activation in our organotypic culture model, which is in contrast to results that were obtained when Raf:ER was induced in MCF-10A cells grown as two-dimensional monolayers [
30]. This discrepancy could be due to subtle variations between MCF-10A cell lines or differences in the expression level of the Raf:ER protein. Alternatively, a distinct mechanism by which ERK1/2 signaling activates PI-3K could be present in organotypic culture, and possibly
in vivo. For example, although EGFR activation
per se is not necessary for proliferation of Raf:ER-induced acini, we do not rule out a role for autocrine growth factors in Raf:ER-stimulated proliferation or PI-3K activation in organotypic culture. This is because Raf:ER activation promotes the autocrine production of FGF-2 and VEGF, which act on non-EGFR receptor tyrosine kinases, and of heparin-binding EGF, which can elicit heterodimerization of ErbB4 with ErbB2 [
30]. Each of these factors activates receptors or receptor combinations that are capable of activating PI-3K, and thus one or more of these autocrine ligands could promote the phosphorylation and activation of PI-3K and AKT in our model.
PI-3K activity is necessary for ERK-stimulated motility
Our understanding of how cells become motile in response to ERK1/2 activation is limited. ERK1/2 can phosphorylate myosin light-chain kinase to promote myosin contraction and can also phosphorylate calpain to promote the severing of integrin attachment to substratum in fibroblasts [
56,
57]. We have shown that ERK1/2 promotes MLC2 phosphorylation through myosin light-chain kinase in mammary epithelial acini; however, a pharmacological inhibitor of calpain has had no effect on cell motility in our model (GW Pearson, unpublished observation). The targets of ERK1/2 signaling that regulate cell motility in general or in mammary epithelial acini are therefore a mystery. We have discovered that PI-3K signaling is upregulated by ERK1/2, and that PI-3K activity is necessary for cell motility in mammary epithelial acini. Although PI-3K and the phospholipid products of PI-3K activity can be elevated through mutation of the catalytic domain of PI-3K or deletion of the phosphatase and tensin homolog lipid phosphatase or amplification and activation of transmembrane receptor proteins, the activation of PI-3K in breast cancer does not require these mutagenic events [
58‐
60]. It is then possible that ERK1/2 activity could drive cell movement, in part, through the activation of PI-3K in some breast cancers.
PI-3K activity is necessary for cell motility in mammary epithelial acini
How cells become motile in mammary epithelial acini is not well understood. We have recently determined that cells can become motile in the absence of invasion [
25]. This finding has potential clinical relevance, because motile cells could be present in pre-invasive lesions, such as DCIS, and thus portend a greater risk of future invasive growth. Whether there are indeed motile cells in pre-invasive lesions is not yet known. A step towards determining how cells become motile during tumorigenesis is the identification of the intracellular signaling pathways that are necessary or sufficient to induce cell movement in these multicellular structures. We have already found that ERK1/2 activation is sufficient to induce movement and that this ERK1/2-driven motility requires MLC2 phosphorylation and a reduction in E-cadherin expression [
25]. We have now determined that PI-3K activity is necessary for the induction of motility induced by ERK1/2 signaling in mammary epithelial acini.
The requirement of PI-3K activity for Raf:ER-stimulated cell motility is independent of MLC2 phosphorylation or E-cadherin expression, which suggests that PI-3K regulates at least one additional process that is necessary for cells to become motile in mammary epithelial acini. PI-3K signaling has been extensively studied in the regulation of chemotaxis in the slime mold
Dictyostelium and neutrophils [
61]. In these model systems, PI-3K contributes the production of phosphatidylinositol (3,4,5)-triphosphate at the leasing edge of the cell, which is necessary for the polarization of the cell and the directional migration towards a chemoattractant [
61]. PI-3K activity is necessary for the chemotaxis of additional cell types, including some patient-derived breast cancer cell lines, possibly through an analogous mechanism [
62]. Whether cells in epithelial acini are moving by chemotaxis is not known. In fact, cells move in different directions within an acinus – which suggests that chemotaxis, and by extension a requirement for sustained polarization of cells, is not necessary for the movement observed. Considering this possibility, PI-3K activity probably regulates motility in mammary epithelial acini through a mechanism distinct from the polarization necessary for chemotaxis observed in other model systems. In the future, determining how PI-3K regulates movement in mammary epithelial acini will serve to further explain how cells become motile during breast cancer progression.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
GWP conceived of the study, designed and performed the experiments, and wrote the manuscript. TH wrote the manuscript. Both authors read and approved the final manuscript.