We have previously shown that BMP4 reduces proliferation and increases migration of breast cancer cells in vitro. As these results were derived from cells grown in 2D monolayer culture, we set out to analyze the effect of BMP4 in a more physiological setting by employing 3D culture systems. We approached this issue by using both a biological gel (Matrigel, the standard 3D culture environment) and a synthetic material with RGD peptides and MMP-degradable peptide links (PEG gel).
The two materials studied provided dissimilar 3D environments as first evidenced by differences in the morphology of the normal and cancer cell clusters. The MCF-10A normal mammary epithelial cells had a polarized acini structure in Matrigel, as previously shown
, while in PEG gel the cells formed irregular non-polarized structures. Similarly, the morphology of the different cancer cells varied between the two 3D models, with the structures formed in Matrigel again corresponding to those previously reported
. On a functional level, the growth response of cells to BMP4 treatment in PEG gel mirrored the 2D data, whereas in Matrigel more diverse effects were observed. These data could be explained by several factors. Matrigel contains multiple biologically active molecules, such as laminin, collagen IV and many growth factors
, that are likely to impact the results obtained. Of these biologically active molecules, e.g. laminin-1 has been shown to be essential for correct polarization of primary luminal epithelial cells in collagen gels
. It has also been reported that 50 mM RGD peptide is an optimal concentration for acinar growth of MCF-10A cells in polyethylene glycol tetravinyl sulfone (PEG-VS) gel
. A lower concentration of RGD (50 μM) was present in the PEG gel used here, possibly explaining the lack of acinar formation. In addition, the stiffness and elasticity of the matrix is known to influence the cellular phenotype, including proliferation, differentiation and migration, in 3D environments
[31–33]. To summarize, the differences in cell morphology and BMP4 response between the two materials tested demonstrate that the mere 3D architecture is not sufficient to mimic the biological effects of tissue environment. Based on the morphological characteristics, Matrigel seems to provide a more appropriate milieu for breast epithelial cells. While many synthetic 3D materials are entering the market, they should be used cautiously until their biological properties have been explored.
Previous data from us and others
[6, 10] clearly demonstrate that BMP4 reduces the proliferation of breast cancer cells in 2D culture, and similar results have been reported in other tumor types
[5, 34–37]. Here we extend these findings and first show the same growth suppressive effect of BMP4 in MCF-10A normal immortalized breast epithelial cells both in 2D and 3D environment. The 3D data from the breast cancer cell lines were more diverse. In PEG gel, BMP4 administration led to reduced cell proliferation for all cell lines tested, whereas in Matrigel two out of four cell lines (MDA-MB-231 and MDA-MB-361) did not display growth inhibition upon BMP4 treatment. In the case of MDA-MB-361, the very slow growth rate of the cells in 3D may have contributed to these findings, although the difference between responses in PEG gel and Matrigel implies an actual effect triggered by the different environments. Furthermore, the growth suppressive action of BMP4 seen in MDA-MB-231 cells in 2D
 disappeared in 3D Matrigel and was overcome by a migratory phenotype. The response of the cells to biological molecules is known to change drastically in 3D, for example, many anticancer drugs are less effective in 3D culture
. Our data now suggest that the ability of BMP4 to reduce cell growth in 3D strongly depends on the material used. Nevertheless, cell line specific differences also exist and further highlight the importance of testing the impact of biological factors, including BMP4, in a proper environment.
BMP4 has been reported to induce G1 cell cycle arrest in cancer cells
[10, 39–41]. We now show for the first time that the mechanism behind this cell cycle arrest in breast cancer cells is the increased expression of the cell cycle inhibitor p21. This result is in concordance with previous reports in 2D culture of various normal and neoplastic cells
[41–45]. Additionally, BMP2 has been shown to induce p21 expression in breast cancer cells
[39, 40, 46]. Interestingly, BMP4 induced p21 expression in MDA-MB-231 and MDA-MB-361 cells in 3D even in the absence of growth inhibition, suggesting that p21 alone is not sufficient to induce growth arrest in these cells in 3D. Furthermore in MCF-10A cells, p21 induction and G1 cell cycle arrest were not evident until day 5 in 2D culture, even though a significant growth reduction was seen already at day 3. Likewise, in MCF-10A 3D culture no p21 induction was observed even after 7 days of BMP4 treatment. Therefore it seems likely that other factors are involved in the BMP4-mediated growth regulation in MCF-10A cells. Examination of a panel of cell cycle regulators in T-47D and MDA-MB-361 cells in 2D showed that BMP4 influenced the expression of multiple cell cycle proteins, including pCDC2, Cyclin B1 and Cyclin B2. These or other cell cycle regulators could thus contribute to the observed growth inhibition in MCF-10A cells as well. Previous studies have reported dysregulation of several cell cycle associated proteins, including Cyclin B1, CDC2, Rb, and E2F, after different stimuli in MCF-10A cells
[47, 48], emphasizing the fact that multiple factors may be simultaneously involved. Further research is needed to identify the specific cell cycle regulators influenced by BMP4 treatment in MCF-10A cells.
In most cases, BMP4 had no effect on the morphology of the cells grown in 3D environment, with the exception of MDA-MB-231 cells and MCF-10A cells. In PEG gel, MCF-10A cells formed irregular structures with small protrusions, the number of which increased upon BMP4 stimulation, indicating increased migration and/or invasion. This is consistent with previous results showing BMP4-induced invasive properties in mouse mammary epithelial cells in collagen gels
. In Matrigel, MDA-MB-231 cells formed stellate, branching structures in response to BMP4, which is in concert with previous observations of increased migration and invasion in 2D experiments
[6, 10]. Such structures were not observed in PEG gel, highlighting again the variation between the different 3D materials.
The MDA-MB-231 cells are known to be triple negative and represent the so-called basal subtype, whereas the remaining breast cancer cell lines used in this study are of luminal type
. We thus speculated whether the molecular subtype could explain the migratory response to BMP4 treatment seen only in MDA-MB-231 cells. To address this issue, we examined another triple negative basal breast cancer cell line, MDA-MB-436. However, the MDA-MB-436 cells were inherently migratory in Matrigel and BMP4 did not induce any additional effects (data not shown). Thus we conclude that the effects of BMP4 cannot be simply explained by the molecular subtype of the cell line. Neither could we link the BMP4-induced phenotypes to other known cell line characteristics, such as the histological type, mutational status, or tumorigenicity
The BMP antagonist Gremlin was able to reverse the MDA-MB-231 stellate phenotype, demonstrating that the effect is truly due to the action of BMP4. Similarly, a broad spectrum MMP inhibitor Batimastat was able to inhibit the BMP4-induced branching of the MDA-MB-231 cells, indicating that the phenomenon required the action of matrix metalloproteinases (MMPs). Unexpectedly, Batimastat also reduced the growth of the cells, both with and without BMP4. MMPs have been shown to cleave intracellular or transmembrane proteins, thereby releasing factors that regulate cell proliferation, apoptosis, invasion and angiogenesis
[51–54]. MMP9 has been particularly shown to possess growth-promoting effects
[55, 56]. Shon et al.
 found BMP4 to suppress the activity of MMP9 in MDA-MB-231 cells, albeit in 2D culture, but in our 3D experiments the expression level of MMP9 was too low to allow accurate measurements and thus MMP9 is unlikely to explain the growth suppressive effects of Batimastat. Nevertheless, examination of the expression of MMPs targeted by Batimastat revealed upregulation of MMP3 and MMP14 in BMP4-treated compared to vehicle-treated cells. Similar induction of MMP3 or MMP14 expression was not seen in the non-migratory BT-474 cells, further suggesting a mechanistic link between these MMPs and the stellate phenotype in MDA-MB-231 cells. A recent study also showed that BMP4 induces the expression of multiple MMPs, including MMP3 and MMP14, in mouse mammary fibroblasts and it also modestly induces the expression of MMP3 in cancer associated human mammary fibroblasts and to a greater degree in normal human mammary fibroblasts
. In contrast, Otto et al.
 found BMP4 to inhibit MMP3 mRNA and protein expression in C3H10T1/2 stem cells, and this inhibition was related to adipogenetic differentiation. These opposing results are likely to reflect cell-type and context-specific differences.
The exact mechanisms behind MMP3 and MMP14 induction upon BMP4 treatment in MDA-MB-231 cells remain to be revealed. MMP3 has in its promoter a binding element for AP-1, which is in turn known to be regulated by BMP4
[59, 60], thereby representing a likely link between BMP4 and MMP3. However, previous data from other BMP/TGF-β family members suggest that additional signaling pathways may also contribute to the MMP induction. In MDA-MB-435 melanoma cells, TGF-β-induced upregulation of MMP14 has been shown to be dependent on the ERK1/2, PI3K, and JNK pathways
 and in MDA-MB-231 cells TGF-β induced the expression of many MMPs, including MMP14, through the p38 MAP kinase
. Similarly, BMP2 has been shown to increase the expression of MMP9 in gastric cancer cells through AKT, ERK and NF-κB signaling cascades
. Taken together, multiple signaling pathways may be involved in the BMP4-induced upregulation of MMP expression.