Background
Three-dimensional (3D) culture of breast epithelial cells has become a widely accepted and highly relevant tool for the examination of mammary gland biology [
1,
2]. Immortalized breast epithelial cells in a 3D context have contributed to a deeper understanding of breast morphogenesis including the role of its microenvironment. In this latter regard, the composition of extracellular matrix [
3‐
5], its stiffness [
6,
7], as well as collagen fiber density and organization [
4,
7] are all relevant factors responsible for the behavior of individual cells, their intercellular communication, and their organization into complex multicellular structures.
From puberty onward, mammary gland structure and function are dependent on circulating hormones. Curiously, however, most mammary epithelial morphogenesis has been studied using hormone receptor-negative cell lines, such as MCF-10A. During the last decades, a number of immortalized mouse and tumorigenic human-derived mammary epithelial cell lines that express estrogen receptor alpha (ER) and display hormone sensitivity have been established and used for such a purpose; they include HC11 (mouse), MCF7, BT-474 ZR-75-B, MDA-MB-361 and T47D [
8,
9]. Cells of the latter established line have recently been shown to form normal-like structures in 3D culture in response to mammotropic hormones. Exposure to estradiol alone resulted in the formation of multicellular structures while further addition of promegestone (progesterone analog) or prolactin resulted in flattened branching structures and budding structures, respectively [
10]. While searching for a model that would closely resemble the human mammary tissue, we chose to test MCF-12A cells, a human mammary cell line allegedly reported to be non-tumorigenic and estrogen-responsive.
MCF-12A cells originally established at the Michigan Cancer Foundation are non-tumorigenic human cells that were derived from the reduction mammoplasty of a postmenopausal 63-year-old nulliparous woman [
11]. Tissues excised from the donor’s breast revealed non-malignant fibrocystic disease containing intraductal hyperplasia abutted by dense stroma. After they were dissociated, these cells were plated for long-term culture in DMEM/H containing < 0.06 mM calcium, later supplemented with 5% Chelex-treated equine serum and maintained for 1717 days. At this point, “passages one to 15 [were] exposed to 45 °C for as long as 72 h” due to an incubator malfunction. As a result of this event, most cells died. The few surviving cells were expanded and over the next 2 months sublines MCF-12A (adherent cells) and MCF-12F (floating cells) were maintained separately and became “established” [
12].
During their initial characterization, MCF-12A cells were implanted subcutaneously in athymic mice, some of which were also implanted with pellets containing 17β-estradiol (E2) but no tumors developed at the inoculation sites in either groups of mice [
11]. Given their origin and this outcome, the cell line was labeled as “normal”. Like MCF-10A cells, and unlike tumorigenic MCF7 cells, MCF-12A cells were also considered as ER-negative and non-tumorigenic. Later, Zeillinger [
13] described the expression of ER transcripts in MCF-12A as “extremely weak” and Subik [
14] was not able to identify any ER positive MCF-12A cells via immunohistochemistry. At least three other studies done using MCF-12A cells described them as ER negative [
15‐
17]. Notwithstanding, literature identifying MCF-12A cells as ER-positive gradually accrued [
18‐
27]. In sum, diverse groups using PCR, western blots, and immunostaining have reached conflicting conclusions regarding the ER status of these cells.
Marchese et al. [
28] grew MCF-12A cells in a Matrigel-based 3D model that resulted in the formation of acini and went on to show alterations in lumen formation following treatment with a variety of different estrogens. Western blots purportedly showed the expression of ER and ER-beta proteins by MCF-12A, and MCF7 cells. In addition, estradiol was reported to induce progesterone receptor (PGR) and pS2 in these cells; however, MCF-10A cells were used as a control in this experiment instead of the estrogen-responsive MCF7 cells. This more recent report suggested that MCF-12A cells were a desirable tool for the study of hormone-mediated epithelial morphogenesis.
In order to test the worthiness of the MCF-12A cell line to study hormone-mediated epithelial morphogenesis, we ran experiments from which we conclude that (a) that MCF-12A cells are not responsive to estrogens, (b) the heterogeneous morphology of MCF-12A cells is due to the expansion of tightly growing epithelial colonies which gradually release populations of motile cells that generate morphologically distinct subclones, and (c) when grown in a rat-tail collagen type I matrix, MCF-12A cells produce acini and ducts, resembling those seen in the human mammary gland. Implications of these data are further discussed below.
Discussion
Hormonal regulation of morphogenesis has been explored in a few established culture models that use immortalized cell lines such as MCF7 and T47D [
10,
35]. Initially, we were interested in employing the MCF-12A cells in our 3D culture model due to their normal, non-tumorigenic origin and their alleged estrogen sensitivity. To our knowledge, currently, no “normal”, ER-positive established cell lines are available. Claims of estrogen-sensitivity made about MCF-12A cells appealed to us for the exploration of the human mammary gland in both normal development and in pathological contexts, like carcinogenesis. It is acknowledged, however, that a given cell line often varies widely in hormone receptor content from one lab to the next [
36‐
38]. Thus, before embarking in an extended study of the subject it was necessary to accurately characterize the cell line to be used in such a project.
Morphology and cell-type markers
Unlike MCF-10A cells, which form epithelial plaques with smooth, defined boarders in 2D cultures, MCF-12A cells underwent a process of phenotypic changes during their propagation. Namely, cells at the edge of plaques lost the plasma membrane localization of E-cadherin and beta-catenin, altered their distribution of vimentin, and became highly motile. Morphologically, these cells appeared to be migrating away from the plaques due to their fan-like projections (Fig.
2). A possible explanation for the dynamic morphologies seen in MCF-12A cells would have been the co-expression of E-cadherin and vimentin. An inverse relationship between E-cadherin and vimentin in intact tissues is considered to be an indicator of epithelial–mesenchymal switching [
39]. When vimentin was overexpressed in MCF7 cells they adopted mesenchymal morphologies with vimentin localized within the cells in a pattern similar to that seen at the edges of MCF-12A cell colonies [
40]. Apparently, the mechanical and biochemical constraints imposed upon MCF-12A cells growing at the center of epithelial plaques are reduced on cells residing in the plaque’s periphery.
MCF-12A cells express gene products that are associated with both luminal and basal subtypes and have features of basal progenitor cells [
41]. Dual expression of E-cadherin and vimentin has been linked to highly aggressive tumors in other tissues including those in the lung and neck [
42,
43]. However, MCF-12A cells are considered non-invasive [
11]. Expression of p63 has been reported to be restricted to the basal layer of complex glands such as the mammary gland and prostate [
44,
45]. Decreasing p63 expression in these cells was linked to a need to replenish cells composing the luminal compartment. The finding that p63−/− mice lack stratified epithelium would support this notion [
46]. The transient expression of p63 in dividing epithelial-like MCF-12A and MCF-10A cells is likely to affect the behavior of their daughter cells. However, non-epithelial MCF-12A cells do not express p63 (Fig.
2), further questioning the cell type that MCF-12A cells more closely resemble in vivo.
Estrogenicity
Until now, the estrogen responsive profile of MCF-12A has been ambiguous. In the original report, MCF-12A xenografts failed to grow in the presence of E2 and additional findings, mentioned but not shown by the authors, suggested that the cells may have been ER-negative [
11]. The expression of ER is not sufficient to describe a cell line as being responsive to estrogens. Nevertheless, published reports claimed these cells to be ER positive and have therefore been used as normal controls to genuine estrogen-responsive cancer cells. Mischaracterizations such as these lead to contradictory conclusions and may have added to the increasingly poor reproducibility of biomedical studies [
47]. By using multiple assays, here we have now shown that MCF-12A cells are unresponsive to E2 in culture conditions, both in regard to the controls of cell proliferation and of the expression of estrogen-regulated genes.
Estrogen responsive cells remain quiescent in the presence of charcoal-stripped serum and the absence of E2 and proliferate in the presence of estrogenic compounds [
48]. The proliferative effect of estrogen on these cells only becomes effective when serum supplemented to the basic cell culture medium is CD stripped; in this condition, cells enter proliferative quiescence due to the effect of the serum-borne inhibitor, albumin [
33,
49]. Addition of physiological levels of estradiol neutralize the inhibitory effect of CD serum and thus the capacity of these cells to proliferate is restored. Instead, MCF-12A cells proliferated equally well regardless of the media’s estrogen content. This implies that MCF-12A cells are insensitive to CD-serum induced quiescence and thus are non-responsive to both the serum-borne inhibitor and to estradiol.
In order to explore whether the lack of proliferative response to E2 was due to absence of ER, we tested transcriptional activation resulting from estradiol-ER binding to estrogen responsive cells (MCF7) and compared it to that of MCF-12A cells. The binding of estrogenic chemicals to ER affected the induction and/or attenuation of different transcripts within MCF7 cells [
50]. We used qRT-PCR to investigate the induction of estrogen-responsive genes in MCF7 cells by adding estradiol concentrations that resulted in maximal proliferative responses. While the expression of both amphiregulin and progesterone receptor was significantly increased in MCF7 cells treated with estradiol, MCF-12A fail to significantly express either gene product following 48 h of estradiol stimulation. These results imply that, in addition to the lack of a proliferative response by estrogens, exposure to estrogens also failed to induce specific gene transcription in MCF-12A cells. Moreover, relative to MCF7 cells, MCF-12A cells express levels of ER transcripts similar to those seen in ER-negative cell lines. The basal-like transcriptional profile of MCF-12A cells further suggests that these cells are ER negative, as only luminal cells express ER in tissues [
51].
Growth in 3D
The formation of epithelial structures in 3D culture allows for the in vitro study and manipulation of the mammary epithelium. We have previously shown that some cell lines form both acini and ducts when grown in collagen with and without the addition of Matrigel [
30,
52]. Herein, we examined the capacity of MCF-12A cells to form normal epithelial tissue structures. In a 3D context, the formation of lumena is a hallmark of normal mammary phenotype [
5]. When grown on top of laminin-rich extracellular matrix for 4 days, MCF-12A cells formed rounded, non-lumenized structures [
41]. Using a similar cell model, these early round structures were shown to form growth-arrested acini consisting of polarized cells after 16 days in culture [
28]. After 14 days of being embedded in rat-tail type I collagen, in addition to lumenized acini and solid, round structures, MCF-12A cells in our growth conditions organized into large, lumenized, branching structures. The formation of duct-like structures by MCF-12A cells has previously been described only when grown on top of matrix derived from nulliparous rat mammary glands [
53]. This implies that rat-tail type 1 collagen recapitulates the mammary gland environment found in vivo better than in several other 3D models currently in use. The rounded clusters of concentric cells (Fig.
4c, arrow) never occurred outside of other structures implying that their formation is dependent on the interruption of interactions between those cells and the collagen matrix. MCF-10A cells grown in rat-tail type I collagen only form acini and short non-branching ducts without any “rosebud” structures.
We employed the SAMA package for the unbiased, unsupervised analysis of MCF-12A structures formed in collagen gels. SAMA allows for unsupervised, and hence unbiased and automated measurements of a host of biologically relevant physical parameters of epithelial structures, an ability that is useful in analyzing large number of structures. Another of the advantages of the SAMA software is the ability to filter the data based on both size and biological relevance of the structures analyzed. This software package enabled us to analyze hundreds of structures within a 3D space for geometric parameters that otherwise would have been difficult to perform.
The variability of structural shapes formed in other areas of these gels may be due to the distribution of MCF-12A cells migrating toward the perimeter of each gel. The ability of MCF-12A to drastically contract the gels (Fig.
5) and the heterogeneous collagen fiber organization may be responsible for this variability [
30]. Gel contraction of similar magnitude has been previously described when MCF-10A cells were grown in rat-tail collagen and resulted in a non-homogenous distribution of structures [
30]. The finding that fibroblast-like cell-depleted BF9 subclone contracted gels at a similar degree to that seen by the parental cell population implies that those non-epithelial-like cells do not contribute to gel contraction. If those cells were responsible for increased gel contraction, assuming their contributions to be additive, BF9 gels would be expected to contract to a degree intermediate between acellular and parental gels.
The results of SAMA analysis confirm that MCF-12A parental cells form spherical, acinar and duct-like structures. Upon removal of the MCF-12A cells which grow away from the epithelial plaques in 2D, the remaining cells form longer, flatter structures than those produced by the heterogeneous parental cell line, on average. The differences seen between parental and BF9 gels are not simply due to the exclusion of individual spindle-shaped cells but, rather, imply an interaction between the two cell types in 3D which results in distinctly different structural compositions. Human fibroblasts have been shown to speed up the morphogenesis of epithelial structures formed by MCF-10A cells and increase the number and length of ducts formed when seeded jointly in collagen gels; this effect is likely due to fibroblast-mediated formation of thicker, parallel collagen fibers [
4,
54]. However, the single cell population of MCF-12A acts differently than fibroblasts; this effect might be due to fibroblasts interfering with the ability of the epithelial cells to alter collagen organization leaving thinner, less organized fibers favoring the formation of rounded acinar structures.
Conclusions
We have documented that MCF-12A cells are a non-tumorigenic, heterogeneous, estrogen receptor-negative cell line that express a combination of epithelial and mesenchymal markers. In floating rat-tail type I collagen gels, MCF-12A cells form complex, lumenized acini and ducts, the characteristics and distribution of which are altered by either the presence or absence of different subpopulations. Based on the time-dependent progressive population changes within the parental cell line and the complex interactions between these two cell types in 3D, MCF-12A cells appear unsuitable for use in epithelial morphogenesis studies when compared to MCF10A cells.
Authors’ contributions
MS generated all figures and the body of the manuscript. CS and AS provided significant intellectual contributions to the design of the study and the interpretation of its findings. All authors read and approved the final manuscript.