Background
Stromal-epithelial interactions play important roles during mammary gland development. This intense epithelial-mesenchymal crosstalk has been well documented during the process of mammary bud invasion of the underlying mesenchyme [
1,
2]. Comparable reciprocal interactions are also evident during tumor initiation and progression as highlighted in tissue recombination experiments. By recombining stroma and epithelium under different experimental conditions it was demonstrated that normal mammary epithelial cells were "instructed" to become tumor cells only when the stroma was treated with a carcinogen [
3,
4]. Moreover, transgenic mice that over-expressed the extracellular matrix (ECM)-degrading enzyme stromelysin-1 (also known as matrix metalloproteinase 3 or MMP3) in the mammary stroma, developed mammary epithelial tumors at approximately 3-4 months of age [
5,
6]. Equally significant, normal stroma induced epithelial tumor cells to form normal mammary ducts [
7]. Furthermore, mouse mammary tumor cells cultured
in vitro with embryonic mouse mammary mesenchyme normalized their phenotype and formed mammary ducts [
8]. These phenotypic reversions have also been documented in the liver where hepatocarcinoma cells acquired normal liver cell phenotypes when injected into a healthy rat hepatic parenchyma [
9,
10].
Changes in the stroma have been reported to accompany cancer progression leading to metastases [
11‐
13]. Such stromal changes include the induction or upregulation of a variety of molecules such as growth factors, matrix degrading enzymes, angiogenic factors and cytokines in stromal cells [
14]. In addition, tumor-derived fibroblasts produced a collagen-rich matrix [
15]. Using an animal model, Provenzano
et al. observed that increased collagen density promoted mammary tumor initiation and progression [
16].
Because of the complexity of
in vivo experiments, three-dimensional (3D) cell cultures are being increasingly used as surrogate models in order to explain how cell-matrix interactions affect the morphology, differentiation, proliferation, and apoptosis of breast cancer epithelial cells. Whereas cell morphology of normal and breast cancer epithelial cells is similar in 2D, non-malignant and malignant cells can be reliably distinguished when grown in 3D reconstituted basement membrane (rBM) (commonly known as Matrigel) cultures [
17]. While non-malignant cells organize into polarized, growth-arrested colonies, malignant cells - both established and primary tumor cells - show a loss of tissue polarity, a disorganized architecture without lumen and failure to arrest growth [
17‐
20]. Furthermore, sensitivity to chemotherapeutic agents dramatically decreased in 3D cultures compared to the effects shown by the same drugs in 2D [
21‐
24]; thus 3D cultures appear to be better models for the cytotoxic evaluation of anticancer drugs [
25].
Breast cancer cells derived from either established cell lines or from primary tumors formed disorganized cell clusters without lumen when grown in rBM [
17‐
19]. A comparison between the morphology and gene expression profile of several breast cancer cell lines grown in 3D cultures using rBM as a matrix showed that cells expressing similar phenotypes clustered together based on similarities in their gene expression [
18]. The commonly used human breast cancer MCF7 cell line formed round colonies with disorganized nuclei and absent lumen when cultured for 4 days using only rBM as the matrix. In addition to rBM and collagen gels, polylactic acid (PLA) polymers, chitosan scaffolds and poly(lactide-co-glycolide) (PLG) scaffolds have been shown to support growth of MCF7 cells [
21,
25,
26]. Recently, a 3D mammalian cell perfusion-culture system using microfluidic channels was shown to support 3D cell-cell and cell-matrix interactions using MCF7 cells [
27]. Although much is known about the important role of the stroma in determining the epithelial phenotype, research continues to be focused mostly on the epithelial cells while disregarding the influence of their microenvironment. Few papers have reported the use of co-cultures of breast cancer cells with other cells found in the
in vivo tumor microenvironment including adipocytes [
28] and osteoblasts found in metastatic sites [
29].
Herein, we report the role of two stromal components, namely human breast fibroblasts, and ECM composition in determining the morphology, survival and differentiation of the human breast cancer MCF7 cell line in a long-term 3D tissue morphogenesis model.
Methods
Chemicals and cell culture reagents
Methyl salicylate, Mayer's hematoxylin, HEPES and Carmine Alum were purchased from Sigma-Aldrich (St. Louis, MO). Dulbecco's modified Eagle's medium (DMEM) was purchased from MP Biomedicals (Solon, OH). DMEM/F12 and penicillin-streptomycin solution were obtained from Gibco/Invitrogen (Carlsbad, CA). Fetal bovine serum (FBS) was purchased from HyClone (Logan, UT). Bovine type I collagen was purchased from Organogenesis (Canton, MA). Matrigel™ and rat tail type I collagen were purchased from BD Biosciences (San Jose, CA). Formalin was obtained from Fisher Scientific (Atlanta, GA).
Cell maintenance
All cells were maintained and expanded in cell culture plastic flasks (Corning, Corning, NY). MCF7 cells were grown in DMEM containing 5% FBS. Human mammary fibroblasts obtained from reduction mammoplasties (RMF) were purchased from ScienCell (Carlsbad, CA). RMF were routinely grown in DMEM containing 10% FBS, 15 μM HEPES and penicillin-streptomycin. All cells were incubated at 37°C and 6% CO2. For co-culture experiments, a combined medium (1 part of MCF7 medium and 1 part of RMF medium) was used. The combined medium was tested in tissue culture flasks containing each cell type alone, i.e. either MCF7 cells or RMF, to assure proper growth and behavior of cells.
3D multicellular culture
Type I collagen was used at a concentration of 1 mg/ml according to Paszek et al. [
30]. Collagen was neutralized according to the manufacturer's instructions. Collagen gels were prepared using bovine collagen; mixed Matrigel™ and collagen gels were prepared using a 1:1 volume ratio of Matrigel™ and type I collagen keeping the final collagen concentration at 1 mg/ml. In co-cultures, 300,000 MCF7 cells and 100,000 RMF were used to mimic the
in vivo ratio of epithelial cells to fibroblasts in the human breast [
31]. The same number of cells was seeded for each cell type cultured independently. Cells were suspended in 3 ml collagen or a 3 ml Matrigel™-collagen mixture and seeded into 35 mm well inserts of a six-well plate (Organogenesis, Canton, MA) as previously described [
32]. The gels were allowed to solidify for 30 minutes at 37°C before adding combined medium onto each gel (2 ml) and into each well (10 ml). Cultures were maintained for one to six weeks and the medium was changed every two to three days.
Gel processing
On the harvest day, the circular gels were cut into two pieces: one half was fixed overnight in 10% phosphate-buffered formalin, paraffin-embedded and used for histological analysis, and the second piece was whole-mounted onto a slide and fixed overnight in 10% phosphate-buffered formalin for morphometric analysis and confocal microscopy. The whole-mounted gels were stained with Carmine Alum overnight as described previously [
33]. After staining, the whole mounts were progressively dehydrated in 70%, 95% and 100% ethanol, cleared in xylene and mounted with Permount™ (Fisher Scientific, Atlanta, GA).
Confocal microscopy
Whole-mounted gels were analyzed using a Zeiss LSM 510 system (Carl Zeiss MicroImaging Inc, Thornwood, NY). The HeNE 633 nm/5 mW laser was used for data acquisition due to the autofluorescence of Carmine dye at this wavelength. The epithelial structures formed in collagen and mixed Matrigel™-collagen gels were scanned with a 20× objective lens, and 8 bit depth images with a resolution of up to 2048 × 2048 pixels were taken. The data were three-dimensionally reconstructed using Zeiss software.
Immunohistochemical analysis
All primary antibody concentrations and supplier information are listed in Table
1. Apoptotic cells were stained using a commercially available TUNEL staining kit (Roche Applied Science, Indianapolis, IN). An antigen-retrieval method using microwave pretreatment and 0.01 M sodium citrate buffer (pH 6) was used for all antibodies except laminin 5 [
33]. The antigen-antibody reaction was visualized using the streptavidin-peroxidase complex, with diaminobenzidine tetrahydrochloride (Sigma-Aldrich, St. Louis, MO) as the chromogen. Counterstaining was performed with Mayer's hematoxylin. Images were captured using a Zeiss Axioscope 2 plus microscope (Carl Zeiss MicroImaging Inc, Thornwood, NY).
Table 1
List of primary antibodies used in immunohistochemical analyses.
Mouse anti keratin 18 | Luminal epithelial cells | Sigma-Aldrich (St. Louis, MO) | 1:500 |
Rabbit anti Ki67 | Proliferation | Vector (Burlingame, CA) | 1:3000 |
Mouse anti E-cadherin | Cell-cell adhesion | Novocastra (Newcastle, UK) | 1:50 |
Mouse anti laminin 5 | Basement membrane | Chemicon/Millipore (Bedford, MA) | 1:200 |
Mouse anti sialomucin | Apical proteins | Abcam (Cambridge, MA) | 1:800 |
Mouse anti-CEACAM | Cell-cell adhesion | Abcam (Cambridge, MA) | 1:50 |
Morphometric analysis
The area of the epithelial structures was measured at one, two and three weeks in culture. The analysis was done in both matrix conditions and in mono- and co-cultures. Three to four experiments for each condition were analyzed, and for each experiment, three arbitrarily chosen fields at 10× magnification per section were examined; two sections for each condition were used. There were between 10 and 20 colonies per field; the number of colonies per field decreases with time due to their increase in size. The sections were an accurate representation of the gel. Images were captured using a Zeiss AxioCam camera attached to a Zeiss Axioscope 2 plus using a 10× objective and 3900 dpi, and analyzed with the Zeiss Axiovision version 4.4 software (Carl Zeiss MicroImaging Inc, Thornwood, NY). The following parameters were measured per section: total number of structures, area of clusters, total number of cells per structure, and number of Ki67 positive cells within those structures. Proliferating epithelial cells were expressed as %Ki67 positive cells per total epithelial cell number. Morphometric analysis was carried out in collagen and mixed Matrigel™-collagen gels.
Statistics
SPSS software package 15.0 (SPSS Inc., Chicago, IL) was used for all statistical analyses. ANOVA was used to compare the area of epithelial structures across time points within a single treatment. Dunnett's T3 post-hoc tests were used to determine which time points were significant within each individual treatment. All results are presented as mean ± s.e.m. Chi square tests were used to compare each condition in the Ki67 data, because proliferation index is expressed as a proportion. For all statistical tests, results were considered significant at p < 0.05.
Discussion
Understanding normal development and carcinogenesis of the breast requires a multidisciplinary approach because the continuous cross-talk between the tissue compartments,
i.e. stroma and epithelium, involves both biochemical and biomechanical cues. In an effort to understand some of these interactions and how they modulate the epithelial phenotype, we have further characterized an
in vitro 3D breast cancer model involving the co-culture of MCF7 cells, a widely used human breast cancer cell line, and primary human breast fibroblasts grown within two different extracellular matrices. This promising model revealed that both the extracellular matrix composition and the presence of fibroblasts determined the fate of the epithelial cells and their phenotype. These findings are comparable to those described for the non-tumorigenic breast epithelial MCF10A cells grown in similar conditions [
32]. The addition of stromal cells, the utilization of extracellular matrices that more closely resemble the breast tissue microenvironment, and the long-term cultures have revealed the remarkable plasticity of the MCF7 cells under different experimental conditions. Our findings show that MCF7 cells are able 1) to reverse their tumor phenotype almost completely, and 2) to continue growing into large masses that resemble tumors depending on the environment surrounding them and the length of the experiment. This plasticity is precisely the reason why data gathered from 3D culture models ought to be interpreted in the context of each experimental design.
In regards to the reversion of the tumor phenotype, the formation of lumen in breast epithelial structures is one of the hallmarks of the normal mammary phenotype.
In vivo, the fetal mammary epithelial cords form a lumen at the same time as the mesenchymal cells undergo differentiation [
35]. Accordingly, the bi-directional biochemical and biomechanical signals are highly regulated. Lumen formation (associated with a well-differentiated phenotype) and lumen filling (associated with a tumor phenotype) were described in various 3D breast cell culture models; both phenomena are commonly explained as a result of differential gene expression in the epithelial cells. For instance, the pro-apoptotic factor BIM correlates with lumen formation by inducing apoptosis in MCF10A cells [
36]. Similarly, MCF10A cells filled the lumen when transfected with ERBB2, CSF-R1 or v-SRC genes [
36‐
38]. Regarding MCF7 cells, the cell-cell adhesion molecule CEACAM1-4 S, which is not expressed in MCF7 cells, is reported to be necessary for apoptosis to occur and lumen to form in a rBM-based 3D culture [
34].
Little attention has been given to the role of the extracellular matrix composition and/or the presence of stromal cells in modulating lumen formation in 3D cultures. Here, we showed that MCF7 cells organized into structures containing a lumen in the presence of RMF and/or rBM, and in the absence of CEACAM1 expression. Furthermore, in the mixed gels, apoptosis was observed in the center of the cell clusters both when MCF7 cells were grown alone and in co-culture with RMF. This is a remarkable phenomenon in breast cancer cells grown in 3D conditions and in collagen-based matrices. Interestingly, lumen formation has not been mentioned in studies reporting phenotypic reversion of cancer cells [
39]. The only instance in which MCF7 cells did not form a lumen was when the cells were cultured alone in type I collagen matrices. In all cases in which a lumen was observed, apoptosis in the center of the epithelial structures was accompanied by cell polarization, indicating that the latter may be required prior to lumen formation. This is also the case in the developing breast and in the non-tumorigenic breast MCF10A cells grown in 3D cultures [
32,
40]. Our findings suggest that complex interactions of the matrix components with the epithelial cells are necessary to induce lumen formation, and that such phenomena are not restricted to normal epithelial cells. This correlates well with a report showing that non-tumorigenic mammary epithelial cells filled the lumen and continued to proliferate as a result of changing only the compliance of a 3D collagen matrix, another example of the multifaceted epithelial-stromal crosstalk [
30].
Another phenotypic feature of reversion is a reduced proliferation rate. We have observed that the percentage of cells expressing Ki67 decreased over time in type I collagen gels. In addition, the proliferation index was significantly lower in the epithelial structures formed in the mixed gels compared to those formed in the collagen gels after one and two weeks; this correlates well with the hypothesis that cellular organization into defined structures such as alveoli and ducts is accompanied by a decrease in proliferation. It has been reported that normal breast epithelial cells significantly reduced their proliferation rate after polarized alveoli and ducts are formed in 3D cultures [
32,
41] whereas primary breast carcinoma cells and tumorigenic breast cell lines formed large, disorganized and non-polarized colonies that failed to undergo growth arrest [
41]. These observations are consistent with our findings that MCF7 cells in mixed gels undergo near complete phenotypic reversion.
Previously, we reported that RMF co-cultured with the non-tumorigenic human breast MCF10A cells accelerated the initial formation of epithelial structures in collagen gels [
32]. Similarly, we now show that RMF provided a "survival" signal and considerably improved the organization of MCF7 cells cultured in the same matrix. Moreover, RMF induced the formation of both round and elongated structures containing polarized cells and lumen. In contrast, many MCF7 cells grown alone in the same collagen matrix underwent apoptosis, and after a lag period, the surviving cells only formed round, disorganized cell clusters. It is worth mentioning that these structures evolved throughout the six week incubation period and that the phenotype of the tissue structures, and the conclusions drawn from these experiments, would have varied had the cultures been stopped at the usual two weeks or earlier [
18,
19,
42].
The addition of Matrigel™ to the collagen gels provided additional evidence for the role of the ECM proteins in modulating the epithelial phenotype. The morphology of the structures formed by the MCF7 cells cultured in mixed gels containing collagen and rBM proteins was different from that observed in collagen gels regardless of whether or not RMF were present in the culture. After one and two weeks in culture, MCF7 cells grown alone already formed alveolar structures containing polarized cells and lumen, indicating that rBM provided "survival" and patterning signal(s) for the MCF7 cells. In the presence of RMF in these gels, MCF7 cells formed both alveolar- and duct-like structures consisting of polarized cells.
In an extensive report, Kenny and colleagues studied the growth of different breast cancer cell lines in an rBM-based 3D culture. These researchers classified the cell lines into four categories depending on their epithelial morphology. MCF7 cells were part of the "mass class", meaning that the cells formed round structures with disorganized nuclei and no lumen [
18], which appears comparable to what we observed when MCF7 were grown in collagen gels. However, our findings that MCF7 cells growing alone or in co-culture with RMF were able to organize and polarize in mixed gels indicate that a matrix containing both type I collagen and rBM provided a more advantageous environment for the epithelial cells to differentiate. In addition, because the experiments described by Kenny
et al. only lasted four days [
18], it is difficult to compare our respective data considering the much longer length of our experiments (
i.e. up to 6 weeks).
It has been previously reported that tumor cells can reverse their malignant phenotype to a normal one both in 3D cultures and
in vivo [
7,
39,
43,
44]. The phenotype reversion in rBM 3D cultures has been defined by the presence of small, growth arrested, well-differentiated and polarized acinar structures [
39]. It is worth noting that lumen formation was not included as one of the features of reversion. Using the same model, it has been shown that reversion occurs through modulating cell-adhesion proteins such as beta1 and alpha2 integrins, dystroglycan1, CEACAM1, gap junctions, and EGFR that are aberrantly expressed in tumors cells [
19,
45‐
47], and inhibition of PI3K and MAPK pathways [
39]. Our results suggest that changes in either matrix composition alone and/or the addition of stromal cells are sufficient to lead to a reversed phenotype that was near normal,
i.e. organized and polarized epithelial structures, lumen formation, and low proliferation index. In our model, in the presence of RMF, MCF7 cells formed organized acini containing polarized cells and lumen, as well as elongated structures with features that resembled ducts. However, it should also be noted that in all these conditions MCF7 cells were unable to secrete their own basement membrane proteins indicating that the reversal was incomplete.
Despite dramatic changes in epithelial cell morphology due to the presence of RMF and/or different matrices, the continuous, although low, proliferation of the MCF7 cells observed during the six weeks in culture suggests that such a microenvironment was insufficient to completely reverse the MCF7 cell tumor phenotype throughout the length of the experiment, and that more research is necessary to understand how cell proliferation is controlled under these conditions. In order to observe a complete reversion as has been shown with cancer cells
in vivo [
7,
48,
49], this and other 3D culture models have the potential to be improved by increasing their complexity while adding either different types of stromal cells, various extracellular matrix components and/or modifying the biomechanical properties of the gels. Taking into consideration this and other reports, it becomes apparent that the fate of normal and neoplastic epithelial cells and their organizational phenotype is susceptible of being modified by manipulating a wide array of cellular and extracellular parameters.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SK performed all the experiments and analysis and drafted the manuscript. MVM participated in the study design, histological analysis and co-wrote the manuscript with SK. AMS and CS participated in the study coordination, design and interpretation of results and revised the manuscript. All authors read and approved the final manuscript.