Interactions between luminal epithelial cells and their surrounding microenvironment govern the normal development and function of the mammary gland. Alterations of these interactions can induce abnormal intracellular signaling pathways that affect the development and progression of breast tumors. One critical component of mammary gland development, as well as breast cancer progression, is the expression of estrogen receptors. In a previous study using cultured nonmalignant mammary epithelial cells, we found that the basement membrane molecules, laminin-1 and collagen-IV, were involved in maintenance of estrogen receptor (ER) α expression, and that this response could be interfered with by disrupting cell-extracellular matrix adhesion. Here we use phenotypically normal mammary epithelial SCp2 cells to dissect the promoter region of the ERα that is involved in the selective response to basement membrane. We also analyze the alteration of this response in SCg6 cells, a malignant cell line that shares a common lineage with the SCp2 cells, to provide insight into the relative overexpression of ERα and the unresponsiveness to basement membrane regulation found in those malignant cells. Evidence is presented to show the relevance of the cross-talk between different signaling pathways in the constitution of a functional tissue organization and how this integration may be disrupted in the malignant phenotype.
Background and Results
Involvement of Steroid Receptors in Mammary Gland Development and Carcinogenesis.
Steroid receptors play key roles in mammary gland development. Experiments in transgenic mice have revealed that estrogen receptor (ER) α is needed for ductal elongation of the mammary gland during puberty (1) , and that progesterone receptor (PR) is required for lobuloalveolar growth (2) . The distribution of ER and PR in the mature mammary gland is heterogeneous both in rodents and in humans where between 4% and 20% of luminal epithelial cells are positive for ER and ∼ 50% for PR (3) . Aside from epithelial ER, stromal ER is necessary for estrogen-induced epithelial proliferation in rodents but not in humans, whereas myoepithelial cells are negative for ER both in rodents and in humans. Expression of these receptors fluctuates in conjunction with the cyclic changes in estrogen and progesterone associated with the estrus cycle, pregnancy, lactation, and age, indicating that expression levels and distribution of ER and PR in the mammary gland are tied to the developmental fate of the epithelial cells (3) .
However, signaling through the ER has long been implicated as a factor in the induction and/or promotion of carcinogenesis. This is attributed to receptor-mediated stimulation of proliferation that confers an autonomous and selective growth advantage to premalignant breast epithelial cells (4) . This increased proliferation contributes to the accumulation of genetic damage and stimulates the synthesis of additional growth factors that act on mammary epithelial cells via autocrine and paracrine pathways (5) . ER is present in the mammary gland as two isoforms, ERα and ERβ, and the relative levels and activities of these isoforms may affect the etiology of hormone-dependent cancers, the growth responses to estrogenic substances, and the response to hormonal therapies (6 , 7) . Thus, the signaling pathways that govern the expression of ER isoforms in mammary epithelial cells have been an intensive topic of investigation. Here we focus on ERα and, its regulation by extracellular matrix (ECM) in murine cell lines because this isoform is indispensable for normal mammary growth and differentiation, and is a key effector of invasive and in situ ductal carcinoma (8) .
Relevance of Tissue Architecture to the Biology of the Mammary Gland: Application to Cell Culture Studies.
Regulation of normal mammary gland function is closely related to the preservation of the tissue architecture, where the correct integration of signaling pathways associated with integrins, growth factor receptors, and steroid receptors plays a critical role (9, 10, 11) . Either alterations in the microenvironment or altered perception of the microenvironment by the epithelial cells can predispose to tumorigenesis. In analyzing these processes, it is important to consider that the mammary gland is composed of several distinct cell types with many overlapping reciprocal signaling interactions. Luminal epithelial cells are the major target for neoplastic transformation, and this cell type has been the focus of most cancer investigations. However, myoepithelial cells also are critical determinants of both normal breast structure and breast cancer status (12) . In vivo, myoepithelial and most luminal epithelial cells are in contact with the basement membrane (BM), a specialized form of the ECM. It has been shown that appropriate communications between luminal epithelial cells and BM, either through direct interactions or indirectly through intervening myoepithelial cells, are essential for normal mammary differentiation (9 , 12 , 13) . Furthermore, the breast stroma, composed of adipocytes, fibroblasts, vascular cells, and immune effector cells, is an active participant in mammary gland development as well as in cancer progression (14 , 15) . The relative proportion of stromal versus epithelial cells can vary as a function of mammary development and carcinogenesis, and also in response to hormones and/or growth factors. Indeed, differentiation of the mammary gland during lactation involves changes in the predominant cell type from nonepithelial to epithelial (3) , in the expression of different combinations of ECM components and integrin receptors (11) , and in the enzymes responsible in ECM remodeling, the matrix metalloproteinases, which also cooperate with hormonal stimuli to induce morphological and functional changes required at various developmental stages (16) . All of these changes could affect estrogen response of the luminal epithelial cells during differentiation of the mammary gland.
When nonmalignant rodent or human mammary epithelial cells are cultured on conventional tissue culture plastic in the absence of ECM, normal signaling pathways are affected and the typical patterns of morphology and gene expression become disrupted, including steroid hormone receptor expression (10 , 17) . By contrast, human breast epithelial cells retain normal levels of ER and PR expression when cultured in a mixture of collagen-I and Matrigel (a laminin-rich reconstituted BM; lrBM) or after transplantation into nude mice where the epithelial cells can interact with the host stroma (18) . Likewise, Parmar et al. (14) have cultured human mammary epithelial cells in vivo under the renal capsule of female nude mice in combination with stromal cells in a collagen-I gel or in lrBM. The reconstitution of epithelial-stromal interactions allows the maintenance of normal organization of the organoids, leads to branching morphogenesis and functional differentiation under appropriate hormonal conditions, and sustains ER and PR levels and steroid hormone responsiveness of the epithelial cells. Similarly, we have shown that the re-establishment of cell-BM interactions in culture is crucial for the maintenance of ER expression and function in isolated mouse mammary epithelial cells (19) . However, this regulatory effect of BM or its components, laminin-1 and collagen-IV, on ERα levels is restricted to the epithelial cells because stromal mammary fibroblasts isolated from mouse mammary glands do not respond to lrBM under the same conditions (19) . Furthermore, when culture dishes were coated with the nonadhesive substratum poly(2-hydroxyethyl methacrylate) to prevent attachment to plastic and to maintain epithelial cells as cell aggregates, we found that the addition of lrBM was still necessary to maintain ERα levels in primary mammary epithelial cells in culture (Fig. 1)<$REFLINK> . These observations demonstrate that loss of ER in cells cultured on plastic is not because the cells are attaching to an unnatural substratum, but rather is a consequence of separation of the epithelial cells from the stroma and myoepithelial cells, which are the physiological sources of the BM in the mammary gland (12) .
Collectively, these data emphasize that the relationships between luminal epithelial, myoepithelial, and stromal cells, and ECM components are a critical component of steroid hormone responsiveness, i.e. the right microenvironment and context are crucial for hormone responsiveness.
Cross-Talk among Steroid, Growth Factor, and ECM Receptors Differ between Normal and Malignant Breast Cells.
The foregoing brief summary shows that reciprocal interactions are constantly in place between steroid hormones, steroid hormone receptors, ECM, and ECM receptors, generating an orchestrated functional regulation of the mammary gland. Several lines of research suggest that any process required for the correct interaction of the cells with their microenvironment, and for the normal function of the mammary gland, could potentially go awry during the process of tumor progression. Such changes include: loss of cell-cell contact (20) , altered steroid hormone and growth factor response (21 , 22) , quantitative and qualitative modifications in ECM composition and ECM-receptor expression (23, 24, 25) , altered signaling pathways (22 , 26) , and loss of functional differentiation (9 , 27) .
There is now a growing appreciation that cross-talk between these different signaling pathways is also critical for the normal function of the mammary gland and for development of tumors; for example, integrins lack intrinsic tyrosine kinase activity, but integrin-ECM interactions recruit a number of proteins involved in signal transduction, including focal adhesion kinase, p42/44 mitogen-activated protein kinase (MAPK) and Rho-GTPases (28) . We have shown a bidirectional cross-modulation of β1 integrin and epidermal growth factor receptor (EGFR) signaling through MAPK (26) and phosphatidylinositol 3′-kinase (PI3K) (29) signaling pathways associated with morphological and functional reversion to a normal phenotype of human breast cancer cells cultured in a three-dimensional BM.
There is evidence for cross-talk between growth factor and steroid hormone receptors. Akt activation in response to HER-2 signaling has been proposed as one of the mechanisms involved in the kinase-induced estrogen-independent activation of ER, and this mechanism is thought to be involved in the development of hormone resistance in certain breast cancers (22) . Integrins and steroid hormone receptor signaling pathways also show cross-talk. Zutter et al. (24) have found that the promoter of integrin α2 contains an ER-response element, and Nishida et al. (30) reported that the promoter of integrin α6 contains PR binding sites, so PR and ER may play a role in regulating integrin expression. Consistent with this possibility, ER has been correlated with α2β1 expression in breast carcinomas, and a high proportion of ductal carcinomas that lack ER show low or absent α2β1 expression (31) and correspond to less differentiated tumors with poorer clinical outcome. Our studies have shown that α2 integrin subunit is a mediator of the regulatory effect of laminin-1 and collagen-IV on ERα expression in functionally normal mouse mammary epithelial cells (19) .
Finally, matrix metalloproteinase expression and consequent ECM remodeling are regulated by growth factors and cytokines, by integrins, and by steroid hormones (32) . Steroid hormone receptors can also regulate matrix metalloproteinase expression and/or activity, as an imbalance in the ratio of PR-A:PR-B isoforms affects mammary gland development by disrupting BM organization, a process which may lead to induction of mammary tumors (33) .3
Tumor Cells Become Unresponsive to the Extracellular Context: The Implications for Misregulation of Steroid Receptors.
Clearly, interactions between ECM signaling and steroid receptor expression involve many different levels of regulation. Using an exogenous construct containing 5 kb of the 5′-untranslated region of the mouse ER-promoter, we asked how these pathways may differ in normal and malignant cells.
We used functionally normal and malignant mouse mammary cell lines that share a common derivation (34) . The phenotypically normal epithelial cell line, SCp2, shows a cobblestone morphology (Fig. 2)<$REFLINK> and can be induced to express milk proteins by exposure to lrBM and lactogenic hormones, whereas the malignant cell line, SCg6, displays spindle-like morphology (Fig. 2)<$REFLINK> , has invasive behavior in culture, and is tumorigenic when the cells are transplanted into athymic nude mice (35) . Unlike SCp2 cells, the malignant SCg6 cell line did not show any detectable cell shape change when exposed to lrBM, and expressed higher levels (∼5-fold) of ERα than SCp2 cells. Finally, SCp2 cells responded to BM by up-regulating ERα (19) , whereas SCg6 cells were not responsive (Fig. 2)<$REFLINK> . These characteristics made SCp2 and SCg6 cell lines useful models for investigating how microenvironmental control of hormone receptor function may become compromised in malignancy.
Dissecting BM-Mediated Transcriptional Regulation of ERα.
We analyzed the regulatory effect of BM on ERα at the transcriptional level in SCp2 cells. We used pools of cells stably transfected with a β-galactosidase reporter construct controlled by the 5′-untranslated region of the ERα promoter (referred to as MER II, Fig. 3A<$REFLINK> ) using a −5 kb construct that is sufficient to direct organ-specific expression of the transgene in mice (35) . When stably transfected into SCp2 cells, this construct showed a significant increase in β-galactosidase activity 24 h after adding 2% lrBM, and this effect persisted for 36 h (Fig. 3B)<$REFLINK> . Activation of MER II was equally effective with two individual components of the BM (laminin-1 and collagen-IV) but not by components of the stromal ECM, such as collagen-I (data not shown). There was a slight increase in β-galactosidase activity with extended culture of SCp2 cells, even in the absence of added lrBM, probably due to endogenous synthesis of BM components in high cell densities.
To define the BM-responsive elements in the ERα promoter, we used a series of constructs with sequential deletions. We transfected SCp2 cells with constructs of variable length of the 5′-untranslated region of the ERα promoter (Fig. 3A)<$REFLINK> . MER 1 has ∼5 kb but lacks the first intron (involved in maturation of pre-mRNA), MER 2 has ∼2.5 kb, and MER 3 has ∼1 kb. Whereas MER 1 had activity comparable with MER II, MER 2 and MER 3 were unresponsive to lrBM (Fig. 3C)<$REFLINK> , suggesting that the portion of the ERα promoter that extends from −5 kb to −2.5 kb is the region responsive to regulation by ECM.
We transfected SCg6 cells with the same MER II construct used in SCp2 cells and found a complete lack of lrBM-induced effect (Fig. 3D)<$REFLINK> , suggesting that the response to lrBM is altered in these malignant cells. When SCg6 cells were transfected with the previously defined ECM-responsive element, BCE-1, found in the β-casein gene and controlling a luciferase reporter construct (37) , addition of lrBM did not regulate luciferase activity.4 These results indicate that unlike SCp2 cells, SCg6 cells may lack ECM-mediated responses all together.
Analysis of the BM-responsive sequence localized in the −5 kb to −2.5 kb promoter region of the ERα gene revealed consensus Stat5 and CAAT/enhancer binding protein β (C/EBPβ) sites, localized at −4508 and −4954 bp, respectively (Fig. 4)<$REFLINK> . The β-casein enhancer BCE-1, located at −1760 bp also contains a Stat5 site and a C/EBPβ site (37) . Stat5 transcription factor has been shown to functionally interact with the ER and interfere with steroid hormone-induced transcription (38) . Furthermore, Stat5 sites have been shown to be required in the regulation of ERα and ERβ transcription in the rat corpus luteum and deciduas during pregnancy (39) . This portion of the ERα promoter is then a good candidate to localize the ECM-responsive region of the ERα gene; additional investigations performing directed mutagenesis as was done for BCE-1 (37) would allow us to identify the minimal enhancer element in this gene as well. Given that BM responsiveness is lost in malignant SCg6 cells, identification of this minimal element may provide insight into the mechanisms involved in the initiation and/or progression of tumorigenesis in the mammary gland.
We have determined that the effect of BM on ERα regulation applies to primary cultures of normal mammary epithelial cells as well as to established cell lines, but that malignant cells whereas expressing higher levels of ERα are unresponsive to ECM. Thus, context-dependent regulation of ER activity appears to be a specific property of normal mammary epithelial cells that is lost in malignancy. These data provide a possible explanation for the misregulation of ER expression that occurs during breast tumor progression (6) : as premalignant mammary epithelial cells progress to malignancy, their interactive response to BM becomes disrupted, affecting the intracellular signals that maintain the normal regulation of ER expression. Our findings presented here, of a regulatory region that impacts cell transformation status-specific ER expression, provide a novel tool for investigating these processes.
Dr. Myles Brown: How practical would it be to routinely culture tumor cells? This is something that needs to be visited in that those models are the progeny of clonogenic cells. How hard would it be to take a fresh tumor biopsy and put it into collagen? What do the numbers look like if they do that on 20 patients? How many would we get?
Dr. Bissell: It is practical, but it would be expensive and time-consuming. However, in all honesty, I think it needs to get done.
Dr. Carlos Arteaga: In your primary cultures, are you getting rid of the fibroblasts?
Dr. Bissell: In the initial cases, we did partial purification of the epithelial cells because if we don’t do that, the other cells are going to take over, and so in the later sets that I showed you, we have actually magnetically sorted where we pulled the different cell types out. With our collaborators in Denmark (12) , we now have made a double layer tube of the breast. In the future, we want to come in with the fibroblasts and put those in the bottom, and what we want to do is take fibroblasts from tumor patients and fibroblasts from normal patients.
Dr. James Ingle: Your data on integrin and tumor reversion were very interesting. Have you done any further studies with that, in vitro or in vivo studies?
Dr. Bissell: I have a clinical collaborator, Dr. Cathy Park, and we have begun to use β1 integrin antibodies to treat tumors in mice. When you take these reverted cells that are treated with inhibitors and you inject them into mice as aggregates, you get zero tumors. If you separate them, and then treat with inhibitors, you get a 70% reduction in tumors and a very long latent period. So, we have begun doing studies with the mouse, to form the tumors first and then treat. The initial results are quite encouraging. My usual statement is that the structure of the tissue is dominant over the genome, and I think that the little bit of data that I showed you support the conclusion well.
Dr. Ingle: Are there other ECM receptors that you’ve found that are important?
Dr. Bissell: α6 β4 is extremely important as well. If you mess up this integrin, the polarity goes, and once the polarity goes, then the whole lot of other things go as well. So there are a number of integrin receptors that appear to be important [Weaver et al., Cancer Cell 2(3): 205–216, 2002]. With all of these phenotypic and signaling changes, I argue that it is balance that is important and not the absolute levels. For example, there are at least 4 or 5 different classes of ECM receptors. One important class is dystroglycan, which is not only in muscles, but is actually also in epithelial tissues, and we have shown that the ratio of β1 to dystroglycan is important. So, if you bring up dystroglycan when β1 is up, you keep the structure, but if you change the ratio, it becomes problematic.
Dr. Rachel Schiff: Can you go half way to this 3D model and just use laminin, cover plastic with laminin?
Dr. Bissell: There are certain things that laminin signaling will give you, if you have the cells on tissue culture, but if you coat the dish with laminin and let it dry, it will not do much. On the other hand, if you drip it in the medium so there is a different presentation of the components to the cells, you get a very different result. Each cell, believe it or not, is like a tiny embryo. They are able to do this amazing stuff, but they have to have a malleable substratum to allow them to reorganize. In fact, most of the work I showed you with the ER receptor was with the dripping of basement membrane components into the media. That can give you estrogen receptor upregulation, but it will not give you any of the changes in pathways that I talked about for chemotherapy resistance. For that, we need laminin and 3D structures.
Dr. Schiff: Can you configure cells to make laminin?
Dr. Bissell: Actually, 99% of luminal epithelial cells in culture make laminin, but when these cells are in the breast they do not. You take the cells from reduction mammoplasty, you separate them, you put them in culture; within 24 hours they change. They are in this very strange microenvironment, they need to have laminin for polarity, which is important for survival. They start making it. They become hybrid, between luminal epithelial cells and myoepithelial cells. The trick would be to make tumor myoepithelial cells in vivo regain their ability to make laminin. We have not yet figured out how to do this. Gene therapy, of course, is an option, but it needs to be very targeted.
Dr. Arteaga: You showed some data that implied very different gene expression profiles as a function of 3D versus 2D arrangement of the mammary cells. This result suggests the possibility that some signaling program outputs may vary as a function of the architectural arrangement of the cells and/or the composition of the ECM, both of which could well be heterogeneous within a tumor. Thus, has anybody shown, for example, whether ER transcription varies spatially as a function of these variables? Such variability would explain at least in part the heterogeneous responses to antiestrogens that are sometimes observed.
Dr. Bissell: Absolutely. This is precisely the point of the review we wrote for the first issue of Nature Reviews: Cancer [Nat. Rev. Cancer, 1: 46–54, 2001]. One argument I make is that the tumor itself is like an organ, and it constantly remakes itself. I think that’s why we need to pay attention to the context that tumors are in. That’s why we need to come up with chemotherapies that are multiple compounds, and this model can actually be used for testing combinations of chemotherapeutic drugs. The problem, of course, is that even this model, which is so much better than monolayers, is not good enough, because in vivo you have other cells such as the stromal cells and vascular endothelium. In fact, there are a number of us who want to get together and actually model this. A few years ago we modeled tumor cells in 3D and put fibroblasts from tumor cells together with the tumor cells in culture. They looked exactly like a primary tumor. But, on the other hand, you have to have all the components and by no means are we anywhere near having a model that approximates the in vivo situation, but clearly that is the goal.
Dr. Arteaga: So are there examples of oncogenes that sensitized tumor cells to extracellular stimuli?
Dr. Bissell: A paper I did in collaboration with Senthil Muthuswamy and Joan Brugge a couple of years ago shows that “oncogenes” also behave differently in 2D vs. 3D [Nat. Cell Biol., 3: 785–792, 2001]. The problem with the transgenic models is that you turn this on from day one, and you don’t know how this mass develops because different things are in the wrong places. The beauty of this model is that you can allow the cells to become an acini and then you could conditionally turn on ErbB1 or ErbB2. What we showed was that when these cultures have either ErbB1 or ErbB2 they grow, but when you put the cells in 3D, ErbB1 (EGFR) didn’t touch the cells. ErbB2, on the other hand, just fills up these acini and makes something analogous to carcinoma in situ, because under these conditions ErbB2 completely changes the ratio of growth and apoptosis.
Presented at the Third International Conference on Recent Advances and Future Directions in Endocrine Manipulation of Breast Cancer, July 21–22, 2003, Cambridge, MA.
Grant support: Contract DE-AC03-76F00098 from the United States Department of Energy, Office of Biological and Environmental Research, by the Innovator Award DAMD 17-02-1-0438 from the United States Department of Defense, Breast Cancer Research Program (M. J. B.), by NIH Grant CA57621 (Zena Werb and M. J. B.), and by the Italian Association for Cancer Research and the FIRB Grant RBNE0157EH of the Italian Ministry for Education, University and Research (A. W.). Virginia Novaro was supported by the postdoctoral fellowships from the Department of Defense Breast Cancer Research Program (Award No. DAMD17-01-1-0293) and from The Susan G. Komen Breast Cancer Foundation (Award No. PDF-2000-223). Derek C. Radisky was supported by the American Cancer Society Postdoctoral Fellowship (Award No. PF-02-009-01-DDC). Nancy E. Ramos Castro was supported by the Biology Transfer Consortium Program of the University of California in Berkeley.
Requests for reprints: Mina J. Bissell, Life Sciences Division, Lawrence Berkeley National Laboratory, Berkeley CA 94720. Phone: (510) 486-4365; Fax: (510) 486-5586; E-mail:
↵3 M. Simian, M. H. Barcellos-Hoff, M. J. Bissell, and G. Shyamala, unpublished observations.
↵4 Unpublished data.