Disruption of 3D MCF-12A Breast Cell Cultures by Estrogens – An In Vitro Model for ER-Mediated Changes Indicative of Hormonal Carcinogenesis

Introduction Estrogens regulate the proliferation of normal and neoplastic breast epithelium. Although the intracellular mechanisms of estrogens in the breast are largely understood, little is known about how they induce changes in the structure of the mammary epithelium, which are characteristic of breast cancer. In vitro three dimensional (3D) cultures of immortalised breast epithelial cells recapitulate features of the breast epithelium in vivo, including formation of growth arrested acini with hollow lumen and basement membrane. This model can also reproduce features of malignant transformation and breast cancer, such as increased cellular proliferation and filling of the lumen. However, a system where a connection between estrogen receptor (ER) activation and disruption of acini formation can be studied to elucidate the role of estrogens is still missing. Methods/Principal Findings We describe an in vitro 3D model for breast glandular structure development, using breast epithelial MCF-12A cells cultured in a reconstituted basement membrane matrix. These cells are estrogen receptor (ER)α, ERβ and G-protein coupled estrogen receptor 1 (GPER) competent, allowing the investigation of the effects of estrogens on mammary gland formation and disruption. Under normal conditions, MCF-12A cells formed organised acini, with deposition of basement membrane and hollow lumen. However, treatment with 17β-estradiol, and the exogenous estrogens bisphenol A and propylparaben resulted in deformed acini and filling of the acinar lumen. When these chemicals were combined with ER and GPER inhibitors (ICI 182,780 and G-15, respectively), the deformed acini recovered normal features, such as a spheroid shape, proliferative arrest and luminal clearing, suggesting a role for the ER and GPER in the estrogenic disruption of acinar formation. Conclusion This new model offers the opportunity to better understand the role of the ER and GPER in the morphogenesis of breast glandular structure as well as the events implicated in breast cancer initiation and progression.


Introduction
In recent years, three dimensional (3D) cultures of immortalised breast cells have gained immense support, as they provide a unique opportunity to model the architecture of epithelium in vivo, in an in vitro system [1,2]. Unlike monolayer cultures, immortalised mammary epithelial cells grown in 3D recapitulate numerous features of the breast epithelium in vivo, including the formation of growth arrested polarised acini with hollow lumen and basal deposition of basement membrane components [3,4]. These wellorganised acinar structures can also reproduce important features of malignant transformation and breast cancer development in vivo, such as loss of apico-basal polarity, increased cellular proliferation and filling of the lumen [5] For these reasons, 3D epithelial cultures in a reconstituted basement membrane matrix (Matrigel) provide the appropriate structural and functional context for studying the events involved in the maintenance of the morphology and function of the glandular epithelium, as well as the events that trigger disruption of these structures and potentially lead to breast cancer.
Epidemiological and clinical data has provided evidence that the estrogen receptor (ER) plays an important role in the development and progression of breast cancer. Over 50% of all breast cancers express ERa and approximately 70% of those respond to therapy with anti-estrogens, such as tamoxifen, which makes the ERa an important predictor of the likelihood of a good response to hormonal therapy and overall prognosis. A role for ERa in breast cancer initiation and progression has also been suggested, as high levels of this receptor found in benign breast epithelium have been correlated with increased risk of breast cancer [6][7][8]. Nonetheless, a clear understanding of the involvement of the ER in breast epithelium formation, breast carcinogenesis and cancer progression is still missing, as well as the role of other estrogen-responsive receptors, such as the membrane associated G-protein coupled estrogen receptor 1 (GPER, formerly GPR30). Although its function in the breast is still unclear, the presence and activation of GPER have been recently associated with development of tumour resistance to tamoxifen [9].
These considerations emphasise the importance of developing an in vitro model where the involvement of estrogen responsive receptors on breast epithelial formation and subsequent tumourigenic transformation can be studied. Establishing a system where many features of the breast epithelium can be recapitulated and a connection between ER activation and carcinogenicity can be investigated is essential to clarify the role of the ER (in particular ERa) on breast carcinogenesis, as well as the mechanisms of hormonal carcinogenesis associated with endogenous and synthetic estrogens. However, such a model has been lacking so far.
To date, investigations of the effects of estrogens in the breast in an in vitro 3D setting have concentrated on cultures of nontumorigenic ERa negative/ERb positive breast epithelial MCF-10F cells, which were derived from the floating population of the culture that also originated MCF-10A cells and share many of their characteristics [10,11]. This MCF-10F cell line has been used to investigate the effects of 17b-estradiol (E2) and its metabolites on the formation of 3D structures which characterise normal breast development. Work conducted by Russo and colleagues [12][13][14] has revealed that E2-treated cells lose their ability to form 3D duct-like structures in a collagen matrix, have high invasiveness and form tumours when injected into immunodeficient mice, all indicative of a cancerous phenotype. Similar observations were also reported for environmental contaminants with estrogenic activity (xenoestrogens), such as bisphenol A (BPA) and butylbenzyl phthalate (BBP) [14] and shown to derive from genomic and epigenetic changes. However, the role of ERa could not be evaluated as it is lacking in these cells.
Here, we describe an in vitro 3D model for breast glandular structure development, using non-transformed breast epithelial MCF-12A breast cells [15]. Unlike the alternative 3D model with MCF-12F cells mentioned above [11,14], MCF-12A cells are ERa, ERb and GPR30 competent. This offers the opportunity to study the involvement of these receptors in breast morphogenesis, as well as the impact of ER agonists, such as estrogens and estrogen-like chemicals, on mammary gland formation, disruption and, potentially, carcinogenesis.
We observed that MCF-12A grown in matrigel under normal, control conditions formed organised, growth arrested, spheroid acini, with deposition of basement membrane components and hollow lumen. Conversely, treatment of these cells with E2 disrupted the morphology of the acini and interfered with lumen formation in a concentration-dependent manner. Interestingly, the same magnitude of effects was not observed in 3D cultures of ERa negative MCF-10A breast cells also treated with the hormone.
A similar effect to E2 was found with two xenoestrogens: BPA and the cosmetic additive n-propylparaben. Exposure of MCF-12A 3D cultures to 10 mM of these chemicals for 16 days resulted in large, misshapen, highly disorganised acini, with extensive lumen filling. The potential involvement of estrogen receptors in the described effects was evaluated by combining the test chemicals with inhibitory agents, such as the antiestrogen ICI 182,780 and the GPER antagonist G-15. Results from these coexposures revealed that both the nuclear and the transmembrane receptors play a role in the estrogenic disturbance of acini formation, as the antagonists reversed some of effects induced by the estrogens, restoring some of the phenotypes of normal, control structures.

Characterisation of receptor status of MCF-10A and MCF-12A cells and evaluation of ER target gene expression
Although the ER status of MCF-10A and MCF-12A cells has been previously published [10,16], it is widely accepted that receptor expression profiles can vary between different lots and batches of cells. As the main aim of this work was to study the effects of estrogenic chemicals in ER and GPER competent cells, it was essential to confirm the status of both receptors in the cells used here.
Our data ( Figure 1A) indicated that both ER receptors (a and b) were expressed at relatively high levels in MCF-12A. In contrast, MCF-10A cells did not express measurable levels of ERa, while only very low levels of ERb were detected. Both cell lines expressed the transmembrane GPER. Similar, confirmatory observations were made in terms of expression of the relevant genes (ERa and b and GPER) using real-time PCR (data not shown).
In order to confirm the presence of a functional ER and the response of this receptor to estrogenic treatment, the regulation of two well characterised ER-target genes was studied in both cell lines, in the presence of E2. The first gene, TFF1 (coding for the protein pS2) has an imperfect palindromic estrogen responsive element (ERE) in its promoter region and is directly up-regulated by the activated ER. PGR (coding for the progesterone receptor), on the other hand, is up-regulated in the presence of estrogens due to the presence of a half-ERE site in proximity to an Sp-1 binding site [reviewed in 17].
As can be seen in figure 1B, 1 nM E2 led to a significant upregulation of both TFF1 and PGR in MCF-12A cells. The effect of the hormone was reduced in the presence of the ER inhibitor ICI 182,780, an indication of the involvement of the ER in the effects reported. Conversely, no significant regulation of either gene was observed in MCF-10A cells, which further confirms the absence of the ER in this cell line.

3D acini formation in MCF-10A cells in the presence of E2
We first investigated the effects of E2 on acini formation in MCF-10A cells, which are commonly used for 3D culture studies, but, as shown, lack ERa and present low ERb levels.
The biological processes that contribute to the development of acinar structures in a basement membrane gel have been shown to include initial cell proliferation, deposition of the basement membrane and differentiation into two populations (outer and inner cells). This is followed by the death of the centrally located cells by apoptosis, which coincides with luminal formation [18]. Similar to previous observations [18], we found that immortalised non-malignant MCF-10A breast cells cultured for 16 days formed growth-arrested spheroid acini. These consisted of a single layer of cells ( Figure 2A -cell nuclei stained blue) attached to a basement membrane (Figure 2A -basement membrane component laminin V stained red) and surrounding a hollow lumen. Some residual apoptotic cells (stained positive for activated caspase-3 -green) were still present in the centre of the lumen.
When MCF-10A cells were incubated for the same period of time with E2 (10 nM) no significant changes in the glandular structures were observed, compared to solvent controls ( Figure 2). The acini were round and well organised, with a single layer of cells surrounded by a basement membrane. The lumen was completely formed, with very few residual apoptotic cells. However, in contrast to untreated cells, a small proportion of acini presented some slight filling of the lumen ( Figure 2B), which seems to indicate a mild disruption of acini formation or maintenance.

Effects of E2 on acini formation in MCF-12A cells -doseresponse studies
We repeated the experiment described above, but this time using ER competent MCF-12A cells.
In contrast to the findings with MCF-10A cells, E2 was capable of disrupting the 3D structure of MCF-12A acini in a concentration dependent fashion after treatment lasting for 16 days ( Figure 3). All tested concentrations of the hormone disrupted the formation of a hollow lumen in the majority of acini. The lowest tested concentration (1 pM) affected acini morphology only weakly. At this concentration, there were generally normally shaped acini with a clear basement membrane and some lumen hollowing. At the concentrations of 10 pM and above the disrupting effects of the hormone became progressively more pronounced, with filled lumens and no apparent apoptotic cells. At 10 nM E2, the acinar structures were severely misshapen and disorganised with completely filled lumen.

Effects of E2 and estrogen-like agents on acini formation -time course analysis
We were interested in establishing the time course for MCF-12A acini morphogenesis and investigating how the different stages in this biological process were affected by hormonal treatment. To this end, MCF-12A cells were cultured in 3D for 4, 8, 12 and 16 days in the presence and absence of E2 (1 nM) ( Figure 4). Similar experiments were also carried out in the presence of the environmental estrogenic agents BPA and propylparaben (10 mM each).
In solvent-treated cultures (negative controls) the cellular organisation into spheroid structures with deposition of the basement membrane occurred almost from the onset. A full basement membrane was visible within 4 days of incubation. At this stage, acini were still small and the internal organisation was not clear, but structures were round and well formed. Within 8 days, control acini were larger in size, measuring an average of 96.465.2 mm in diameter ( Figure 5), well organised and presented a single layer of cells in contact with the basement membrane. There was also clear evidence of lumen hollowing, as a large percentage of luminal cells stained positive for activated caspase-3, a marker for apoptosis. Further enlargement and luminal formation occurred at day 12 in control acini, when these increased in size to 102.064.8 mm on average. Finally, acini reached their maximum size at 16 days of incubation, with a diameter of 120.564.4 mm ( Figure 5). At this time point, structures were also completely hollow and growth arrested, as no further enlargement was observed with longer incubation periods (data not shown).
In E2 treated samples, the progress of acini formation was clearly disrupted from very early on. At 4 days, E2 treated acini appeared organised and well formed, but were larger than those seen in the controls. After 8 days, E2 treated acini were not too dissimilar from controls, as the size of the spheroid structures, as well as the number of cells per acini, did not differ significantly from those found in untreated samples ( Figure 5). However, there was an indication that luminal clearing was being affected or delayed, as the relative number of central apoptotic cells was  reduced. As seen in figure 6, the percentage of apoptotic cells in relation to the total number of cells in the acinar structures significantly decreased from 10.761.8, under control conditions, to 2.4660.66 in the presence of E2. This effect on luminal formation became even more evident after 12 days incubation. At this time, apoptosis was inhibited, as there was little or no evidence of activated caspase-3 in the central cells and the percentage of apoptotic cells had decreased to 1.660.4 (compared to 7.660.99% in controls).Finally, at 16 days the full impact of E2 treatment was seen, with the hormone inducing the formation of large, severely deformed acini with completely filled lumen. Here, acini measured an average of 142.163.9 mm and contained 62.365.9 cells, as opposed to 120.564.4 and 37.863.9, respec-tively, in controls. This, in combination with the distorted morphology of the structures formed, suggested excessive proliferation and indicated that cells had escaped the proliferative arrest phenotype characteristic of a control situation.
BPA and propylparaben, both tested at a concentration of 10 mM, had a dramatic impact on acini morphogenesis ( Figure 4 and 5). Similar to the natural hormone, this effect became noticeable within 8 days of incubation, when the majority of the structures completely lacked apoptotic cells within the lumen and presented no evidence of luminal clearing. At this stage, both treatments significantly reduced the percentage of apoptotic cells found in acinar structures ( Figure 6), resulting in 0.7960.6 and 0.4660.26% of apoptotic cells, respectively, in comparison to 10.761.8 in untreated controls.The differences between control and treatments became even more obvious at 12 and 16 days incubation. The resulting structures still presented deposition of basement membrane, but were significantly larger than controls (see Figure 5) and no longer had a regular spheroid shape. Cells within the acini were disorganized, there was no evident single layer of cells attached to the basement membrane and lumen were completely filled. Similar to E2, acini were not growth arrested, as they continued increasing in size throughout the treatment.

Impact of ER and GPER inhibitors on estrogen-mediated acini disruption
As all three tested compounds are known ER agonists and are able to activate GPER, we became interested in investigating whether blocking the signal from these receptors would prevent the disruptive effects of estrogens. 3D cultures of MCF-12A cells were treated with the test chemicals (E2, 1 nM; BPA, 10 mM; propylparaben 10 mM) in the presence of the ERa antagonist ICI 182,780 (1 mM), the GPER inhibitor G-15 (10 nM), or a combination of both inhibitors (ICI 182,780 1 mM + G-15 10 nM) for 16 days. The inhibitors were also tested on their own and it became apparent that they, themselves, had a small impact on acini morphology and caused some degree of acini filling ( Figure 7A). Nevertheless, it was also observed that the presence of both antagonists reverted to a considerable extent the acini disruption caused by all three estrogens. As seen in figure 7, ICI 182,780 allowed E2-treated acini to regain a well organized, small, spheroid shape and some luminal clearing. A similar observation was made with G-15, the GPER inhibitor, where acini appeared slightly smaller and partially hollowed, with apoptotic luminal cells. As shown in figures 7B and 7C, although the presence of the inhibitors failed to have a dramatic impact on the size of acini (although G-15 yielded a significant effect), both inhibitors significantly decreased the number of cells found in each acini, which is reflected in the ratio cell number/acini size ( Figure 7D).  This ratio was significantly reduced for samples treated with E2 + ICI 182,780 and E2 + ICI 182,780 + G-15, in relation to those treated with E2 alone. This indicates that the reduction in cell number was not proportional to the reduction in acini size, suggesting it was related to increased hollowing and not just a consequence of smaller acini. These inhibitors also seemed to affect the deregulation of acinar formation caused by the environmental estrogens. In the case of BPA, both inhibitors caused a reduction of acini size in comparison to those resulting from BPA treatment alone ( Figure 7B). Structures were also more organised and spheroidal, with some evidence of lumen hollowing, which was demonstrated by a strong reduction in cell number within acinar structures ( Figure 7C) and a reduction of the ratio cell number/acini size ( Figure 7D). With propylparaben, structures formed in the presence of ICI 182,780 and G-15 were smaller ( Figure 7B), but still presented a completely filled lumen and no signs of apoptosis. In this case, it seemed clear that the reduction in cell number was a consequence of a reduction in acini size, and not luminal clearing, as the ratio cell number/acinar size remained similar to that determined for propylparaben alone ( figure 7D).
Interestingly, the combination of ICI 182,780 and G-15 had a greater impact on the effects of E2 and BPA, yielding structures that resembled control acini in many ways. These were structurally organised and partially hollow, suggesting the inhibition of both ER and GPER reduced the disrupting effects of both chemicals. This was further supported by the ratio cell number/acini size, which decreased significantly for both chemicals in the presence of both inhibitors (Figure 7D), indicating a lower number of cells and, consequently, acinar hollowing. On the other hand, although the malformations caused by propylparaben were reduced by the inhibitors (especially in relation to size), apoptosis seemed largely unaffected, as acini were still completely filled and the ratio cell number/acini size remained unaltered. This highlighted the possibility that other factors might be at play here.

Discussion
For the first time, we show that the endogenous hormone E2 and the environmental estrogens BPA and propylparaben are able to disrupt the correct formation of acini in 3D cultures of ERa, ERb and GPER competent MCF-12A normal breast cells. The extent of these abnormalities differed depending on the compound and concentration, but overall they were characterised by the formation of aberrant, highly proliferative and disorganised acini, with filled lumen. As demonstrated previously, [4,[18][19][20] normal breast epithelial cells grown in 3D basement membrane cultures share properties with the glandular breast epithelium in vivo as they form hollow, organised spherical structures, with basal deposition of basement membrane and apoptotic clearance of the luminal cells. The development of large, misshapen acini with filled lumen and unregulated proliferation, as seen here in the presence of our test compounds, resembles the events that occur during the early stages of breast cancer in humans [3,21] and could, potentially, be indicative of neoplastic transformations.
Interestingly, a similar outcome was not observed in ERa negative MCF-10A cells treated with E2, where the hormone induced only mild effects, hardly distinguishable from those observed in controls. As shown here, MCF-12A differ from MCF-10A cells in that they express the ERa, raising the possibility that the marked acini disruption caused by E2 in MCF-12A cultures is somewhat dependent on the presence of ERa. This hypothesis is further supported by our observations that blocking of receptor signalling by using a combination of ICI 182,780 or G-15 with the test compounds reverted the disrupting effects of the estrogens to a large degree in MCF-12A cultures. The evidence that these inhibitors prevented, to a large extent, the damaging effects of E2, BPA and propylparaben to the 3D glandular structures, highlights a role for these receptors in the actions of estrogens (natural and man-made) in the breast glandular morphogenesis and maintenance.
Work conducted previously with MCF-10A cells has shown that the correct functioning and organisation of 3D acini depends on a fine balance between cell proliferation and cell death (apoptosis). An overstimulation or perturbation of these processes will lead to aberrations in glandular structure, such as formation of large misshapen acini with filled lumen and no apico-basal polarity [20,[22][23][24][25]. To date, a number of mechanisms responsible for cell proliferation and apoptosis have been identified as playing an important role in the development and maintenance of acini. These include receptor tyrosine kinases, such as type I insulin-like growth factor receptor (IGFIR) and epidermal growth factor receptor (EGFR), signal transduction pathways, such as PI3K/Akt and Ras/Raf/Erk and cell cycle and apoptosis regulating genes, like cyclin D1 and Bcl-X L [23][24][25][26]. Often, deregulation of acinar structures arises when one or more of these pathways are overactivated or blocked. For example, in 3D cultures of MCF-10A cells, hyperactivation of the PI3K/Akt cascade induces deregulation of acini polarity [27] as well as the disruption of cell size and proliferation, resulting in the formation of large, deformed structures. Also, when combined with an upregulation of the oncoprotein cyclin D1, the hyperactivation of PI3K/Akt or EGF signaling [22,25] promote uncontrolled cell proliferation and evasion from proliferative arrest in MCF-10A spheroids. Additionally, when these strong proliferative signals occur simultaneously with apoptosis inhibition, either by hyperactivation of anti-apoptotic genes, such as Bcl-2 or by the reduced expression of pro-apoptotic genes, such as BIM, the unregulated growth of acini is accompanied by filling of the lumen [23,26].
From the data presented here, it can be noted that the tested chemicals had an impact on both biological processes, cell proliferation and apoptosis, within the acinar structures and the blocking of the estrogen receptors (ER and GPER) seemed to preserve the original features of the spheroids, also in relation to both processes. The formation of large, deformed structures, sometimes presenting multiacinar phenotypes, indicated that the treated cells escaped proliferative arrest and were actively growing as cultures progressed. On the other hand, the reduction of cell death was evidenced by a reduced number of apoptotic cells as soon as day 8 and the completely filled lumen seen at the end of the experiment. Interestingly. The presence of the inhibitors ICI 182,780 and G-15 led to the formation of acini which were smaller in size and spheroid in shape, suggesting proliferative arrest had been restored. Moreover, especially in in the cases of E2 and BPA, luminal clearance was clearly visualised following co-incubation with ICI182,780 and ICI 182,780 + G-15. However, and in spite of the dramatic reduction in cell number in the presence of the inhibitors, these antagonists failed to completely revert the acinar filling induced by the estrogens.
In addition to the well documented direct activation of the nuclear ER and subsequent regulation of ER-target genes (genomic events), estrogens (endogenous and exogenous), such as E2 and BPA, also rapidly and transiently activate signal transduction pathways, such as the PI3K/Akt, Src/Ras/Erk and PKC cascades, via a cell membrane associated ER (non-genomic effects) [28] and GPER [29][30][31]. It is believed that these genomic and non-genomic pathways converge to influence cell proliferation and apoptotic processes [32]. Significantly, this cross-talk between ER and growth factor signalling has also been linked to genomic instability in cancer cells [33] and the development of resistance to hormone therapy [34].
As well as the second-messenger signalling cascades mentioned above, a number of other factors known to be involved in acini maintenance and formation have also been reported to be targets of estrogenic action. For example, cyclin D1, which has been identified as an important regulator of acinar development and proliferation arrest, is also regulated by estrogens, both via the genomic and the non-genomic routes in breast cancer cells. We and others have shown that the transcription of CCDN1 (the gene coding for cyclin D1) can be up-regulated by direct activation of the nuclear ER and in an ER-independent fashion by E2 and estrogen-like chemicals [32,35]. This culminates in an upregulation of the cyclin D1 mRNA and protein, leading, consequently, to an increase in cell division and proliferation [36]. Similar reported targets of estrogenic activity are apoptosis regulating factors, such as Bcl-xL and Bcl-2 [37]. Estrogens increase the expression of anti-apoptotic Bcl-2 and Bcl-xL in normal and breast cancer cells, directly by interacting with the ER [38,39], and indirectly via signalling cascades, such as PI3K [40]. Environmental contaminants, such as BPA, have also been shown to activate Bcl-2 in breast cancer cells in a similar way to E2 [41]. Based on these considerations, it is plausible that the interaction of estrogens with the ER, GPER and signalling cascades in 3D cultures of normal breast cells results in similar outcomes, inducing increased cell growth and decreased cell death. The changes in this balance between cell death and survival would ultimately result in disruption of acini formation and the deformed spheroid structures reported here.
One other important point to consider is the fact that we observed that both ICI 182,780 and G-15 alone induced a small, but noticeable disruption of the normal development of acini. Toxicity of these chemicals can be ruled out, as the concentrations used were tested in monolayer cultures of MCF-12A cells and no evident cell death or toxicity was observed (data not shown). Therefore, we suspect that these effects are possibly related to interactions of the inhibitors with targets other than their cognate receptors. For example, ICI 182,780, is a 'pure' antagonist of the ER, but also has been shown to act as an agonist potential for GPER [42]. In turn, this transmembrane receptor can interfere with signalling events controlled by EGF, which is present in the medium at low levels in order to support acini formation. As mentioned above, EGF-responsive signalling pathways, such as the Erk1/2 signalling cascade, are involved in breast structure formation and maintenance. Therefore, it is possible that by indirectly interfering with these cascades' normal function, inhibitors are causing a small degree of acinar aberrations, such as filling of the lumen, mimicking some of the responses caused by the hormones. Nevertheless, the effects caused by the inhibitors alone are minimal compared to those seen with the estrogens and do not seem to mask the reduction of acini disruption by the hormones in the co-incubation scenarios.
In light of our findings and the documented importance of signalling pathways in breast glandular maintenance, we hypothesise that the effects seen with MCF-12A cells in 3D cultures are the result of a disruption of the balance between cell proliferation and apoptosis, which, in turn, is a consequence of the effects of estrogens on the ER, GPER and extranuclear signalling cascades. Moreover, as highlighted previously by Thompson et al. [43], it is likely that ERa interplays with ERb and membrane associated receptors, such as the GPER, to influence breast glandular maintenance, as well as mediate breast tumour development and growth [42]. This might, ultimately, disrupt acini morphogenesis and lead to the abnormal structures characteristic of early stages of breast cancer.
Work conducted by Russo and colleagues [12][13][14] revealed that E2 is able to induce neoplastic phenotypes in 3D cultures of MCF-10F cells. These are normal, immortalised ERa negative/ERb positive breast cells, which are able to form 3D duct-like structures in a collagen matrix. When treated with E2, these cells lose their ductulogenic capacity, show increased invasiveness and form tumours when subsequently injected in immunodeficient mice. The treatment of the same cells with xenoestrogens, such as BPA and BBP, induced comparable effects, suggesting these compounds also have carcinogenic capabilities and could contribute to the development and progression of breast cancer. These phenotypes were shown to be associated with genetic and epigenetic changes, namely chromosomal aberrations (chromosomal loss and amplification) and DNA methylations [14]. Overall, our study provides further novel and valuable information on the effects of the endogenous hormone and xenoestrogens in 3D cultures of normal breast cells. In this context, our data supports the possibility that their interaction with the ER and possible GPER also plays a significant role, although other factors, such as genomic and chromosomal aberrations, and epigenetic changes are also likely to be associated with some of the carcinogenic characteristics of estrogenic compounds.
To this date, no major advances have been made in developing an in vitro assay that could link estrogenicity and ER activation with potential carcinogenicity. For example, widely used reporter gene assays for ER activation and breast cancer proliferation assays [44][45][46] are of great importance in screening for estrogens and defining ER affinity, but cannot be related to neoplastic events or cancer development. Although the changes in acini development and morphology reported here share many similarities with those observed in the early stages of breast cancer, it is still unclear if the abnormal acini seen in MCF12A are, in fact, representative of a neoplastic phenotype and result from neoplastic transformations induced by estrogenic compounds. Further work, fully validating this 3D in vitro model and characterising the nature of MCF-12A acini changes following treatment is now of utmost importance. The optimisation of a 3D system in ER competent cells, where neoplastic phenotypes can be defined and investigated, would present a valuable framework to dissect the role of the ER in breast morphogenesis and cancer development, both at cellular and tissue level. This would include investigating interaction between different types of receptors and the cross-talk between estrogens and other signalling pathways. Moreover, such a system would also provide a platform to test novel steroidal estrogen receptor modulators (SERMS) and anti estrogens, as well as study mechanisms of acquired resistance, which are still largely unknown and pose a major problem to breast cancer anti-hormonal treatment.
All reagents were purchased from Sigma-Aldrich unless stated otherwise.
Blots were then washed in TBS-T three times for 5 min, incubated with secondary antibody (mouse anti-rabbit antibody, Cell Signaling Technology, New England Biolabs, Hitchin, Hertfordshire, UK, 1:2000 in 5% milk/TBS-T) and washed again following the same procedure. Detection was carried out using the ECL system according to the manufacturer's protocol (New England Biolabs) and luminescence captured by ECL hyperfilm (Amersham).
Cell treatment for RNA isolation MCF-12A and MCF-10A cells were seeded in 25 cm 2 flasks at a density of 300,000 cells per flask and grown for 24 hours in 5 ml culture medium. Cells were then washed with 5 ml 1% Hanks' balanced salt solution without phenol-red (HBSS, Invitrogen) and changed into phenol red-free DMEM/F12 supplemented with 2% HS, 0.5 mg/ml hydrocortisone, 100 ng/ml cholera toxin, 10 mg/ ml insulin, 0.22 mg/ml L-glutamine and 50 mg/ml Pen/Strep and 5 ng/ml EGF, where they were kept for further 24 hours. Medium was replaced with fresh medium containing E2 (final concentration 1 nM) or E2 (1 nM) + ICI 182,680 (1 mM), and cells treated for 24 hours. The final ethanol concentration did not exceed 0.5% (v/v). Controls were obtained by treating the cells with 5% absolute ethanol. Following the dosing period cells were harvested by the addition of 1 ml 0.25% trypsin-EDTA and incubation at 37uC for 5 minutes. The obtained cell suspension was used for RNA isolation.

RNA isolation, reverse transcription and real-time PCR
Total RNA was isolated from cell suspensions using Nucleospin RNA II (Macherey-Nagel, Abgene, Epson, UK) following the manufacturer's instructions and approximately 2.5 mg of RNA was reverse transcribed into cDNA as previously described (45). Realtime PCR was performed using iQ SYBR Green Supermix (Bio-Rad) on the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). For every amplification reaction, melting curves were generated to ensure a single amplification product was obtained for each gene target. Primer pairs (Table 1) were optimised as described before (46) and purchased as high quality, purified oligos from Eurogentec Ltd (Hampshire, UK). b-Actin was used as internal control (reference gene) and the relative gene expression levels were calculated using comparative 2 2DDCt method (47). Results were expressed as fold increase over controls, standardised such that values obtained in cells treated with vehicle only (ethanol) were set to 1. The expression ratio results were tested for statistical significance by a non-parametric Mann Whitney test.
Three-dimensional Matrigel culture MCF-10A and MCF-12A cells were cultured in Matrigel following the protocol described by Debnath and colleagues [17], with slight modifications.
Wells in an eight-chamber slide were pre-coated with a layer of 100% growth factor reduced (GFR) Matrigel (45 ml/well) and allowed to polymerise for 15 minutes. A single cell suspension was prepared in DMEM/F12 supplemented with 2% HS, 0.5 mg/ml hydrocortisone, 100 ng/ml cholera toxin, 10 mg/ml insulin, 0.22 mg/ml L-glutamine and 50 mg/ml Pen/Strep, 5 ng/ml EGF and 2% GFR matrigel (assay medium), containing E2 (1 pM, 10 pM, 100 pM, 1 nM and 10 nM), BPA (10 mM) or propylparaben (10 mM) without exceeding the final concentration of 0.5% ethanol (solvent). For co-incubations, cells were treated with assay media containing a combination of the individual test agents (E2 1 nM, BPA 10 mM or propylparaben 10 mM) with either ICI 182,780 (1 mM), G-15 (10 nM) or a combination of both inhibitors at the same concentrations. Solvent controls were treated in assay media with ethanol at 0.5% or 0.5% ethanol + 0.2% DMSO for the treatments containing G-15. Cells were incubated for periods of up to 16 days for both cell lines, and the medium replaced every 4 days with fresh assay medium containing the test compounds. At the end of the incubation, cells were fixed and slides analysed as described below.

Immunofluorescence and confocal microscopy
Following the incubation period, the assay medium was removed from the chamber slides and cells fixed with freshly prepared 2% paraformaldehyde (in PBS, pH 7.4) for 20 mins, at room temperature. Next, cells were permeabilised with 0.5% Triton-X in PBS for 10 min at 4uC and washed three times with 100 mM glycine (in PBS) at room temperature. Acini were then blocked with immunofluorescence blocking buffer (IF buffer; 130 mM NaCl, 7.7 mM NaN3, 0.1% BSA, 0.2% Triton X-100, 0.05% Tween-20) containing 10% of goat serum for 1.5 hours at room temperature. This was followed by a secondary blocking step, where cells were incubated with IFBB containing 10% goat serum and 20 mg/ml goat anti-mouse F(ab')2 fragment for 30 mins. Primary antibodies were diluted 1:200 in the secondary block solution, incubated overnight at 4uC and rinsed three times with IF buffer. The primary antibodies used were: Mouse antilaminin V (ABcam; 1:200 in IF buffer) and rabbit anti-cleaved caspase-3 (Cell Signaling Technology, 1:200 in IF buffer). Incubation with secondary antibody was in IF buffer containing 10% goat serum for 50 mins Dilution of 1:200), followed by three washes at room temperature. Secondary antibodies were: Alexa Fluor 488 goat anti-rabbit and Alexa Fluor 555 goat anti-mouse (Molecular Probes, Invitrogen). Cell nuclei were counterstained

Acini quantification and statistical analysis
Acini formation and disruption were quantified by analyzing confocal images of acini and measuring the size, as well as the total number of cells in a representative number of acini for each treatment. A minimum of 10 randomly selected acini were analysed for each independent experiment.
For experiments with ICI 182,780 and G-15, the ratio cell number/acini size was also calculated. This allowed for the differentiation between reduction in cell number due to acini hollowing (increased apoptosis) or due to a decrease in acini size (proliferative arrest). If the presence of an inhibitor leads to reduction in cell number due to lumen hollowing, then the ratio cell number/acini size should decrease, whereas if the number of cells decreases solely due to smaller acini, than this ratio should not suffer a major change.
Finally, the percentage of apoptotic cells was calculated as the number of cells positive to activated caspase-3 (stained green) over the total number of cells in each acini. Like before, at least 10 randomly selected acini were analysed for each independent experiment.
Data analysis was performed using the Prism Software (GraphPad Software, version 5.04, La Jolla, CA, USA). ANOVA followed by Dunnett's post-hoc test was applied to compare treatments against negative controls, while ANOVA and Bonferroni's post-hoc test was used to compare paired samples (treatments and inhibitors).