The transcription factor Runx2 is highly expressed in breast cancer cells compared with mammary epithelial cells and contributes to metastasis. Here we directly show that Runx2 expression promotes a tumor cell phenotype of mammary acini in three-dimensional culture. Human mammary epithelial cells (MCF-10A) form polarized, growth-arrested, acinilike structures with glandular architecture. The ectopic expression of Runx2 disrupts acini formation, and electron microscopic ultrastructural analysis revealed the absence of lumens. Characterization of the disrupted acini structures showed increased cell proliferation (Ki-67 positive cells), decreased apoptosis (Bcl-2 induction), and loss of basement membrane formation (absence of B 4 integrin expression). In complementary experiments, inhibition of Runx2 function in metastatic MDA-MB-231 breast cancer cells by stable expression of either short hairpin RNA-Runx2 or a mutant Runx2 deficient in subnuclear targeting resulted in reversion of acini to more normal structures and reduced tumor growth in vivo. These novel findings provide direct mechanistic evidence for the biological activity of Runx2, dependent on its subnuclear localization, in promoting early events of breast cancer progression and suggest a molecular therapeutic target.

BackgroundMCF-10A cells are near diploid and normal human mammary epithelial cells. In three-dimensional reconstituted basement membrane culture, they undergo a well-defined program of proliferation, differentiation, and growth arrest, forming acinar structures that recapitulate many aspects of mammary architecture in vivo. The pre-malignant MCF-10AT cells and malignant MCF-10CA1a lines were sequentially derived from the MCF-10A parental cell line first by expression of a constitutively active T24 H-Ras generating the MCF-10AT cell line. This was followed by repeated selection for increasingly aggressive tumor formation from cells recovered from xenograft tumors in immuno-compromised mice, generating the MCF-10CA1a cell line. When inoculated subcutaneously into the flanks of immuno-compromised mice, MCF-10AT cells occasionally form tumors, whereas MCF-10CA1a cells invariably form tumors with a shorter latency than MCF-10AT derived tumors.ResultsMCF-10AT cells grown in three-dimensional basement membrane culture form complex multi-acinar structures that produce a basement membrane but undergo delayed cell cycle arrest and have incomplete luminal development. MCF-10CA1a cells grown in three-dimensional basement membrane culture form large, hyper-proliferative masses, that retain few characteristics of MCF10A acini and more closely resemble tumors.ConclusionHere we report on the growth and differentiation properties of these three matched cell lines in three-dimensional basement membrane culture. Features of tissue morphogenesis were assessed, including proliferation, basement membrane formation, polarization of alpha-6 beta-4 integrin to the basement membrane, formation of cell:cell junctions, and apoptosis for luminal clearance. The matched series of normal MCF-10A, pre-malignant MCF-10AT, and malignant MCF-10CA1a cells offers a unique opportunity to study the mechanisms of malignant progression both in a three-dimensional microenvironment and in the same cell background.

MCF-10A human mammary epithelial cells cultured inside reconstituted basement membrane form acini that resemble the acinar structures of mammary lobules. This three-dimensional culture system has been used for identifying and characterizing the signal transduction pathways controlling cell proliferation and death, and for studying their disregulation in malignant progression. We have compared the ultrastructure of MCF-10A acini, MCF-10A cells grown in monolayer, and the acinar structures of human breast lobules. The tissue architecture of MCF-10A acini was formed by hemidesmosomes connected to a basement membrane and by abundant desmosomes between acinar cells. Intermediate filaments that joined into large and abundant filament bundles connected hemidesmosomes and desmosomes to sites at the nuclear surface. Fewer and thinner bundles of filaments were observed in monolayer MCF-10A cells and even fewer in breast tissue. Tight junctions were observed between cells in breast tissue but missing in MCF-10A acini. The cytoplasm of MCF-10A acinar cells had a polar organization similar to that observed in breast tissue, with centrosomes and the Golgi apparatus on the apical side of the nucleus. MCF-10A acinar nuclei had an irregular, frequently invaginated surface and had a single nucleolus. The distribution of heterochromatin was similar to that in the epithelial cells of breast tissue. The nuclei of monolayer MCF-10A cells had multiple nucleoli, a more regular profile, and less heterochromatin. Electron microscopy has the resolution required to survey features of MCF-10A cell and acinus architecture that may change with manipulations designed to induce malignant phenotypes.

Changes in nuclear morphology occur during normal development and have been observed during the progression of several diseases. The shape of a nucleus is governed by the balance of forces exerted by nuclear-cytoskeletal contacts and internal forces created by the structure of the chromatin and nuclear envelope. However, factors that regulate the balance of these forces and determine nuclear shape are poorly understood. The SWI/SNF chromatin remodeling enzyme ATPase, BRG1, has been shown to contribute to the regulation of overall cell size and shape. Here we document that immortalized mammary epithelial cells show BRG1-dependent nuclear shape changes. Specifically, knockdown of BRG1 induced grooves in the nuclear periphery that could be documented by cytological and ultrastructural methods. To test the hypothesis that the observed changes in nuclear morphology resulted from altered tension exerted by the cytoskeleton, we disrupted the major cytoskeletal networks and quantified the frequency of BRG1-dependent changes in nuclear morphology. The results demonstrated that disruption of cytoskeletal networks did not change the frequency of BRG1-induced nuclear shape changes. These findings suggest that BRG1 mediates control of nuclear shape by internal nuclear mechanisms that likely control chromatin dynamics.

The ATPase subunits of the SWI/SNF chromatin remodeling enzymes, Brahma (BRM) and Brahma related gene 1 (BRG1), can induce cell cycle arrest in BRM and BRG1 deficient tumor cell lines, and mice heterozygous for Brg1 are predisposed to breast tumors, implicating loss of BRG1 as a mechanism for unregulated cell proliferation. To test the hypothesis that loss of BRG1 can contribute to breast cancer, we utilized RNA interference to reduce the amounts of BRM or BRG1 protein in the nonmalignant mammary epithelial cell line, MCF-10A. When grown in reconstituted basement membrane (rBM), these cells develop into acini that resemble the lobes of normal breast tissue. Contrary to expectations, knockdown of either BRM or BRG1 resulted in an inhibition of cell proliferation in monolayer cultures that was enhanced in three-dimensional rBM culture. This inhibition was strikingly enhanced in three-dimensional rBM culture, although some BRM depleted cells were later able to resume proliferation. Cells did not arrest in any specific stage of the cell cycle; instead, the cell cycle length increased by approximately 50%. Thus, SWI/SNF ATPases promote cell cycle progression in nonmalignant mammary epithelial cells.