Stromelysin-1 as a Regulator of Stromal-Epithelial Interactions During Mammary Gland Development, Involution and Carcinogenesis

by Zena Werb

Summary
An intact basement membrane is essential for the proper function, differentiation and morphology of many epithelial cells. The disruption or remodeling of the basement membrane occurs during normal development as well as in the disease state. Stromelysin-1 (SL-1), a member of the matrix metalloproteinase (MMP) family, was one of the first proteinases found to be associated with cancer. In this review we describe the role of MMPs in normal mammary gland involution. To examine the importance of basement membrane during development in vivo, we altered the MMP and tissue inhibitor of metalloproteinases (TIMP) balance in mammary gland. Inhibition of MMP synthesis by glucocorticoids or implants or transgenic overexpression of TIMP-1 delays matrix degradation and the involution process after weaning. The mammary glands from transgenic mice that inappropriately express autoactivating isoforms of SL-1 are both functionally and morphologically altered throughout development. Transgenic mammary glands have supernumerary branches, and show precocious development of alveoli that express fl-casein expression and undergo unscheduled apoptosis during pregnancy. This is accompanied by progressive development of an altered stroma, which resembles that of a wound site or a tumor, and becomes fibrotic after postweaning involution, and by development of neoplasias. These data suggest that MMPs and disruption of the basement membrane may play key roles in branching morphogenesis of mammary gland, apoptosis, and stromal fibrosis as well as in induction and progression of breast cancer. These observations suggest that SL-1 and other MMPs may be useful targets for therapeutic intervention in cancer.

Extracellular matrix (ECM) provides not only a structural framework for tissues but also plays an important regulatory role for cell proliferation, apoptosis, migration and differentiation. Proteolysis of ECM can, therefore, regulate all these processes. Proteinases play a major role in tissue remodeling events such as ovulation, implantation, wound healing, and involution of the mammary gland, uterus and prostate. Proteinases that degrade the extracellular matrix (ECM), including the serine proteinases and matrix metalloproteinases (MMPs) have been implicated in various pathological states such as inflammation, rheumatoid arthritis, and all stages of tumor progression including growth, invasion, metastasis, and angiogenesis. Although the MMPs are directly responsible for most ECM degradation, proteinases from the other enzyme classes may contribute to this process, primarily through the activation of the MMPs. MMPs have long been implicated as having a role in tumor invasion and metastasis (reviewed in 1). Repeatedly, it has been shown that malignant tumors express MMPs more frequently and at higher levels than normal tissues. In fact, several MMPs were first cloned from tumor cell lines: Gelatinase A, gelatinase B, SL-1, matrilysin, stromelysin-2, membrane-type matrix metalloproteinase (MT-MMP) and as metastasis-specific genes from metastatic tumors (stromelysin-3, collagenase-3) (1-5). Stromelysin-3 and collagenase-3 have been shown to be expressed by breast carcinomas (5,6), MT-MMP and gelatinase B by lung carcinomas (4,7), and matrilysin by colon carcinomas (8,9).

Tumor invasion and metastasis requires cells to cross multiple anatomical barriers such as basement membranes, the surrounding stroma, penetration into the circulation system (intravasation), departure from the vessel walls (extravasation), followed by colonization in the surrounding tissue (metastasis). The ability of tumor cells to cross tissue boundaries may be a result of misregulation of the proteinases relative to their inhibitors (1,10). When active proteinases are more abundant than their inhibitors net degradation ensues; if the activity of the inhibitors is enough to effectively neutralize the proteinases then net ECM accumulation occurs. MMP expression correlates strongly with invasive behavior of tumor cells in culture and with metastasis in animal models (reviewed in 10). The expression of MMPs in human tumors is outlined in Table 1. Of particular interest is the fact that the majority of these proteinases are expressed by the reactive stroma surrounding the tumors rather than by the tumor cells themselves (11). Their significance in this pathological process is underlined by the fact that MMP inhibitors are effective in blocking tumor growth in animals (reviewed in 12). These observations raise the question of how MMPs contribute to altered stromal-epithelial interactions.

Expression and function of matrix metalloproteinases
The MMPs are a family of more than eleven zinc-dependent endopeptidases whose expression is tightly controlled by growth factors, hormones, oncogenes, phorbol esters and cytokines. This family can be subdivided into four main groups: interstitial collagenases, type IV collagenases (gelatinases), stromelysins, and the furin family-activated proteinases (MT-MMP and related enzymes). There are four characterized members of the stromelysin family: SL-1, stromelysin-2, matrilysin, and metalloelastase. The stromelysins have a broad substrate specificity, catalyzing the degradation of various ECM components including proteoglycan core protein, laminin, fibronectin, elastin, entactin, galectin and collagen types IV, V, IX, and X (reviewed in 12), as well as a1-proteinase inhibitor non-ECM substrates such as TNF-a precursor. The activity of the stromelysins can be effectively inhibited by the three members of the tissue inhibitor of matrix metalloproteinases (TIMP) family (13). The stromelysins are secreted as inactive multidomain proenzymes with the active-site zinc blocked by an unpaired cysteine within the propeptide, which is removed upon activation. Like the collagenases, SL-1, stromelysin-2 and matrilysin may be activated by plasmin and other proteinases with trypsin-like substrate specificity, which cleave short basic sequences located in the middle of the propeptide. SL-1, stromelysin-2, metalloelastase and matrilysin are all closely related, being members of a multigene family located on the same chromosome (14). Stromelysin-3, despite its name, is derived from a more primitive gene, and unlike the members of the stromelysin family is activated intracellularly by a furin-related mechanism (15), similar to MT-MMP (4). Stromelysin-3 demonstrates poor proteolytic capabilities; a1-proteinase inhibitor is the only documented substrate for the full length form, although the truncated molecules have weak proteolytic activity on laminin (16,17). The stromelysin genes are primarily expressed in the stromal cells of most tissues. SL-1, stromelysin-2 and stromelysin-3 are localized to the stromal cells of human endometrium, while matrilysin is expressed only in the epithelial component of the endometrium (18-20). Metalloelastase expression is largely confined to macrophages (21). SL-1 expression has been localized to the mouse heart and lung, while stromelysin-2 is found in the heart and kidney (22).

The expression of SL-1 and its role in influencing mammary gland development has been extensively studied in regard to branching morphogenesis, cell proliferation, tissue-specific gene expression and epithelial apoptosis. The role of proteinases in regulating the interaction of epithelial cells with their underlying stroma has been elucidated in normal mammary gland development and in abnormal mammary glands produced as a result of ectopic expression of a SL-1 transgene as described in the following paragraphs.

Matrix metalloproteinases are regulated during mammary gland development
The mammary gland is an excellent model to investigate the role of proteinases in tissue remodeling because the mammary gland undergoes cycles of growth, differentiation, morphogenesis and apoptosis throughout development. A variety of secreted proteinases and proteinase inhibitors are expressed during this cycle. Among the proteinases are SL-1, stromelysin-3 and the 72 kDa gelatinase A (23-27). The MMPs are suppressed during lactation, compared with pregnancy, but are highly expressed during involution. The time course of induction of proteins involved in ECM remodeling correlates with loss of expression of fl-casein during involution. During lactation, fl-casein mRNA is expressed abundantly, whereas mRNAs for proteinases and inhibitors are barely detectable. During involution, the basement membrane is degraded and the mammary epithelial cells lose the capacity to express milk protein genes and are shed into the lumina, where they die. The degradation of the ECM appears to involve cooperation between epithelial and mesenchymal cells.

Mammary gland shows significant proteinase activity during virgin gland development, pregnancy and postweaning involution when ECM remodeling is required. SL-1 expression correlates with ductal elongation and SL-1 mRNA is localized to the stromal fibroblasts along the length of the advancing ducts (23,24) as well as around the tip of growing mammary end buds (24). As determined by RNA blotting analysis, the mRNA for SL-1 is most abundant between 5-10 weeks of development in the juvenile virgin mouse mammary glands, when the ducts are most actively growing and branching (23). SL-1 mRNA is expressed in the stromal cells of the mammary glands of pregnant mice as early as 6 days of pregnancy as branching morphogenesis occurs in preparation for lactation (24). The transition from a fully functional lactating gland to an involuting gland is characterized by three major events. First, during lactation and the early stages of involution, ECM-degrading proteinases are expressed at low levels. However, 3-4 days after weaning, SL-1, stromelysin-3, tPA, uPA, gelatinase A, collagenase-3, and the 26 kDa caseinase are upregulated (7,8,10). This expression reaches a maximum around day 5-6 and remains high for at least 10 days into involution. SL-1 is highly expressed during the degradation of the basement membrane ECM that occurs during involution of the mammary gland following weaning (25-27). During this process of involution the basement membranes are degraded by proteinases, the expression of milk protein genes fall, the alveolar structures collapse, the secretory luminal epithelial cells, the endothelial cells, and the myoepithelial cells are removed by apoptosis, and most epithelial cells are replaced by adipose tissue (25,27-33). The peak of expression of MMPs in involution is associated with a loss-of-function phenotype during which major tissue remodeling is characterized by basement membrane degradation, alveolar regression, apoptosis of epithelial cells, as well as a biosynthetic phase of stromal ECM synthesis, angiogenesis and adipocyte differentiation.

When extracts from involuting mammary glands are analyzed for expression of ECM-degrading proteinases, gelatinase A (72 kDa and its 62 kDa active form) and a 130 kDa gelatinase, which is not inhibited by TIMP-1, account for the major gelatinolytic species (25-27). These gelatinases are vectorially secreted in the direction of the basement membrane by mammary epithelial cells in culture, suggesting their involvement in the ECM-remodeling events of involution in vivo. Stromelysin-1 is expressed in periductal and peri-alveolar stromal cells in involuting mammary glands ((27). The major casein-degrading enzymes with apparent molecular weights of 26 kDa, 35 kDa, 92 kDa and >100 kDa are not secreted but instead are cell-associated. Tissue-type and urokinase-type plasminogen activator (tPA and uPA, respectively) are also expressed during involution of the mammary gland, and the extensive remodeling of the basement membranes and stroma during mammary gland involution requires cooperation of both serine proteinases and MMPs.

In the normal process of mammary gland involution, the expression of SL-1 is temporally and spatially controlled; however, when the expression of SL-1 is misregulated, mammary gland function is profoundly affected as in the case of transgenic mice in which rat SL-1 is overexpressed in epithelial cells (26). In the SL-1 transgenic mice the endogenous mRNA for SL-1 is highly expressed in the stromal fibroblasts surrounding the ducts and in the fibroblasts, but not in the myoepithelial cells of the intralobular unit of the mammary gland in virgin (23), mid-pregnant, and involuting glands (24). The above pattern of expression was similar in both transgenic and normal mice, except that in the mid-pregnant transgenic mice the expression of the SL-1 mRNA extended further into the stroma thereby resembling the pattern seen during involution of a normal mammary gland (24). In addition to the disruption of the basement membrane which leads to development of a reactive stroma in SL-1 transgenic mice, an increase in vascularization occurs with an apparent increase in the number of blood vessels found between the alveoli and in the periductal stroma compared to normal mice (24). This increase in vascularization may be due in part to the liberation of ECM-bound angiogenic factors such as basic fibroblast growth factor (bFGF) or vascular endothelial growth factor (VEGF) (34).

Overexpression of SL-1 in the developing virgin mammary gland demonstrates that this inappropriate expression results in premature branching morphogenesis, differentiation, and hyperproliferation. The SL-1 transgenic mice develop a reactive stroma even in virgin mice and are predisposed to forming mammary gland tumors (24,26,35). A reactive stroma is characterized by an accumulation of collagen, recruitment of inflammatory cells (36), increased vascularization (37), and an upregulation of proteinases (1). Such a stromal phenotype resembles that found in wound sites and in involuting glands. It is likely that the dramatic alteration in the interaction of the stroma with the epithelial cells directly or indirectly leads to tumor development (24). An upregulation of MMP expression in the reactive stroma surrounding malignant tumors is a key feature of tumor progression (38). While SL-1 plays a role in the normal involution of the mammary gland following the cessation of lactation (25), the overexpression of this proteinase can result in tumor formation. When the SL-1 transgene is overexpressed during lactation the transgenic mammary glands contain smaller, collapsed alveoli and resemble an early involution phenotype. Moreover, during pregnancy the epithelial cells undergo unscheduled apoptosis (39,40). Therefore, the inappropriate expression of SL-1 results in profound changes in the stromal-epithelial interactions in the mammary gland and can profoundly affect gland function and promote tumor formation.

Stromelysin-1 regulates stromal-epithelial interactions
Mammary gland involution is associated with remodeling of the ECM, down regulation of tissue-specific gene expression, and a massive wave of epithelial cell apoptosis. When mammary epithelial cells (CID-9) are cultured in vitro with a suitable ECM, b-casein expression is induced (41,42) and apoptosis is suppressed (39). Overexpression of SL-1 can degrade the ECM leading to induction of apoptosis (39). Expression of interleukin-1b converting enzyme (ICE), a mammalian homologue of the Caenorhabditis elegans ced-3 gene, correlates with the loss of ECM, and induction of epithelial cell apoptosis, while inhibitors of the ICE family of proteinases prevent apoptosis in these cells (39). When SL-1 is overexpressed in transgenic mice the precocious expression of SL-1 during pregnancy induces an early involution phenotype (24,39,40). In contrast, inhibition of SL-1 gene expression by glucocorticoid treatment prevents the proteolytic phase of involution, suppresses apoptosis in the epithelial cells, while the tissue-specific gene expression of b-casein was maintained (27,43). Therefore, SL-1 production by the mammary stromal cells can regulate epithelial cell survival and function through interactions with the ECM.

Proteinase-to-inhibitor ratio determines ECM remodeling
An example of the concept that it is the proteinase-to-inhibitor ratio that determines ECM remodeling was first demonstrated by placing implants of TIMP-1 into involuting mouse mammary glands and delaying involution (25). When TIMP-1 is overexpressed in transgenic mice involution is slightly delayed (40). Crossing TIMP-1 transgenic mice with SL-1 transgenic mice can quench the premature involution phenotype that is induced during pregnancy by the ectopic expression of SL-1 (40). Native or recombinant TIMP-1 can inhibit in vitro invasion of human amniotic membranes (44,45) and in vivo metastasis in animal models (45,46). Expression of a transfected TIMP-1 gene lowers the metastatic potential of melanoma cells (47), while ablation of the TIMP-1 gene results in enhanced invasiveness of normal differentiated cells (48) and altered metastasis of co-isogenic transformed cells (Soloway P, Alexander CM, Werb Z, Jaenisch R, unpublished data). Therefore, a critical balance between the ECM-degrading proteinases and their inhibitors regulates epithelial cell function and morphology during involution of the mammary gland.

Stromal-epithelial interactions during tumor formation
In human tumors, MMP mRNAs have been found in stromal fibroblast and vascular cells adjacent to clusters of malignant tumor cells, as well as in some tumor cells (Table 1), depending on the enzyme in question and the type of tumor (1,11). On the other hand, MMP proteins are often associated with tumor cells, suggesting that proteinases made by stromal cells near the invasive front may localize to the tumor cells because of cell surface receptors or local activation. Thus activation at the cell surface may link MMP expression by stromal cells with invasion of tumor cells, and may actually provide the most significant control point in MMP activity. Taken together, previous observations suggest that MMPs may be produced by the tumor itself, by the reactive stromal cells, by endothelial and associated cell of the vasculature responding to the angiogenic stimulus, and by macrophages and other inflammatory cells. Therefore, MMPs may play a role in any one of multiple critical events in tumor evolution including: a) tumor growth, b) growth factor availability by the release of factors such as bFGF from ECM reservoirs, c) angiogenesis by stimulating release of factors such as VEGF from the ECM, d) generation of a reactive stroma, and e) promotion of tumor cell invasion of basement membranes as the enter and exit the vasculature during metastasis. Overexpression of SL-1 in transgenic mice can lead to ductal hyperplasia, the development of hyperplastic nodules, and in some cases adenocarcinomas (35). Therefore, alterations in the stromal epithelial interactions transduced via changes in the integrity of ECM can promote tumor formation. An unanswered question is whether SL-1 or other MMPs actually act as oncogenes, or whether when they are misexpressed they negatively regulate tumor suppressor genes that then facilitate the genomic alterations leading to preneoplastic lesions.

Stromelysin-1-induced degradation of the ECM may lead to increased invasiveness
Extracellular proteolysis of the ECM by MMP may increase MMP expression through feedback mechanisms involving generation of bioactive ECM fragments and liberation of growth factors (1,10,26,49,50). Misregulation of this feedback loop may allow for disruption of the basement membrane and surrounding interstitial tissue thereby facilitating cell infiltration into surrounding tissues. This proteinase-induced turnover of the basement membrane may be signaled through integrin-ECM contacts. Moreover, the repertoires of integrins and of ECM-degrading proteinases change when either normal or tumorigenic cells become invasive. The tumor cell phenotype has been associated with altered contacts between ECM and integrins (reviewed in 2). Several investigators have observed that tumorigenic cell lines are altered either in their patterns of integrin expression or in the expression or pericellular organization of the ECM ligands (generally fibronectin). In some cases more normal growth properties have been obtained by restoring normal levels of the integrins (51-53). Therefore, invasive activity may be regulated by the expression of particular integrins and their interaction with the ECM.

Control of apoptosis
These studies clearly demonstrate that coordinated expression and a critical balance between ECM degrading proteinases and their inhibitors regulate epithelial cell function and morphology during involution of the mammary gland. As the ratio of ECM-degrading proteinases to inhibitors increases, basement membrane is degraded, mammary epithelial cells lose their ability to express fl-casein and undergo apoptosis, and alveolar regression becomes evident (23-29,31-33,39,40). Our in vivo studies of the mouse mammary gland have led us to propose a model for the series of events that occur during, and regulate the process of, involution (Fig. 1). After cessation of milk removal from the gland, milk continues to be secreted and accumulates in the alveolar lumina for 1-2 days. At this time, local factors induce the expression of ECM-degrading proteinases and their inhibitors in a coordinated and temporal pattern. We believe that events that regulate cell-ECM interactions occur in a microenvironment at the individual cell level within an alveolus. Each cell or group of cells within an alveolus is part of a milieu with various amounts of either the ECM-degrading proteinases or the inhibitors. The ECM-degrading proteinases and their inhibitors are secreted either by the epithelia themselves or by neighboring fibroblasts in the underlying stroma. The postulated events that occur at the level of an individual alveolus are as follows. As the lumen of a particular alveolus is engorged owing to cessation of milk removal, a local signal instructs certain cells in the alveolus or in the immediate microenvironment (myoepithelial cells or fibroblasts) to secrete ECM-degrading proteinases. Local concentrations of ECM-degrading proteinases then degrade the basement membrane in the immediate vicinity and alter cell-ECM interactions. This leads to the generation of bioactive ECM fragments, and eventually to detachment of a cell from its degraded basement membrane and from neighboring cells within the alveolus. At some time during this process the cells lose their ability to express milk proteins. The detachment of the cells from the basement membrane could trigger signals for programmed cell death or apoptosis, including the induction of ICE gene expression. Apoptotic events commence as the ratio of ECM-degrading proteinases to inhibitors increases during involution of the mammary gland, or in an unscheduled way in the glands from midpregnant autoactivating SL-1 transgenic mice. After the apoptotic cell detaches from its underlying degraded basement membrane and from neighboring cells in the alveolus, the remaining cells, which rest on an intact basement membrane, join together. As involution proceeds, the net result is a smaller alveolus with a continuous basement membrane. It is not clear what regulates these events at the individual alveolus, and why certain cells within an alveolus are destined to die before others. It is possible that only cells that are progressing through the cell cycle are able to die. We have postulated that the local concentrations of inhibitors bound to the basement membrane protect surviving cells within an alveolus and temporally regulate programmed cell death and alveolar regression.

Many recent studies suggest a role for apoptosis in maintaining homeostasis in normal development, such as embryonic morphogenesis, endometrial cycling, and immune cell selection. In these and other systems, apoptosis offers a physiological way to limit the number of cells, remove cells that have completed a given function, or eliminate premalignant cells. It is clear from our studies that the degradation of ECM by proteinases plays an important role in the apoptosis of MMEC of the involuting mouse mammary gland. However, in the midpregnant glands from transgenic mice, it is not clear if the epithelial cells undergoing apoptosis are indeed normal, leaving premalignant cells that escape apoptosis.

The initiation of apoptosis has been correlated in several cell types with aberrant regulation of the cell cycle and proliferation. Analysis of a variety of cells in tissue culture has shown that at least one of the cyclins is induced in the absence of the rest of the normal regulatory pathway for proliferation. The timing of this induction of cyclin expression is closely associated with the cell's commitment to apoptosis (56-63). Inhibiting the induction of the cyclin or the activation of the associated kinase suppresses apoptosis (56,57). Suppression of apoptosis by stable overexpression of bcl-2 prevents the inappropriate expression or activation of the cell cycle regulatory molecules (60,61), in one case by allowing induction of the normal complement of cyclins (61,62). The specific cell cycle regulatory molecules that are induced during apoptosis differ between cell types; specific examples are cyclin D1 in mammary epithelial cells (62) and cultured neurons (63), cyclin A in HeLa cells (60) and embryonic fibroblasts (61) and B-type cyclins in hematopoietic cells (56,57,59). It is not yet clear whether misregulation of the cell cycle is a necessary component of apoptosis in all cases, as there are several examples in which there are as yet no obvious aberrations in cell cycle regulation (64,65). The observations that link apoptosis and misregulation of the cell cycle have not yet been extended into intact tissues. It is interesting to note that in the involuting mammary gland, apoptosis is induced at 4 days postweaning and is largely confined to the epithelial cells that line the alveoli. Bromodeoxyuridine labeling during this period has demonstrated that there is a contemporaneous increase in the number of cells entering S phase that is also largely confined to the epithelial cells of the alveoli (A. MacAuley, L. Lund and Z. Werb, unpublished observation). In addition, the precocious apoptosis that occurs in the SL-1 transgenic mice during pregnancy, when the epithelial cells are actively proliferating, ceases during lactation, when these cells exit the cell cycle (40). It seems possible that induction of proliferation during the radical changes in the cellular environment that accompany involution may result in misexpression of cell cycle regulatory components in many of the epithelial cells, leading to apoptosis.

Conclusion
Taken together, the observations on mammary glands in which the balance of MMPs and inhibitors has been perturbed indicate that in the context of a normal gland, ECM remodeling is both necessary and sufficient to alter the lactational phenotype of mammary epithelial cells (see Figure). The experiments on cultured cells described above suggest one mechanism for this, namely, laminin-based signaling through a fl1 integrin. However, it remains an outstanding challenge in this system to link the in vivo results with observations on cultured mammary cells. Meeting this challenge will require a more detailed understanding of integrin-mediated signaling in mammary cells. It may also entail designing a cell culture model of mammary involution and/or genetically manipulating the mammary gland in live animals to alter signal transduction systems or cell-ECM contacts.

A model of the critical role played by ECM acting through ECM receptors (ECM-R) in the control of tissue-specific gene expression, cell cycle and apoptosis in mammary gland. MMPs negatively regulate ECM function by proteolysis. MMPs, in turn, are inhibited by TIMPs. ECM then regulates the cell cycle both negatively and positively. ECM also suppresses the apoptotic program, seen as the upregulation of proteinases of the interleukin-1b- converting enzyme (ICE) family. Inhibition of ICE by the viral serpin CrmA, inhibitory derivatives of aspartic acid (BAMCK) or Bcl-2 also suppress the apoptotic program. Chronic alteration of the ECM environment of cells, through modification of the cell cycle and cell cycle checkpoints may be permissive for genomic instability and promote multistage carcinogenesis.

The current body of evidence suggests that a complex pattern of mutual regulation between integrin-mediated signaling and matrix proteolysis has profound influences on a variety of cell behaviors. Thus, ECM regulates cellular function by signaling through integrins, which in turn regulate the turnover of ECM, and these intersecting events may alter both cellular proliferation and programmed cell death, as well as cell migration and invasion, differentiation, and the maintenance of tissue-specific functions. This mechanism operates in the mammary gland, and it is likely that ECM turnover in other tissues is the critical event regulating morphogenesis and differentiation both in developmental tissue regeneration and in pathogenic situations. Because ECM, in turn, signals to the cells through integrins and results in the regulation of MMPs and TIMPs, this interplay serves to choreograph migratory, invasive and differentiative events in multicellular organisms.

Acknowledgements
We thank our colleagues and collaborators past and present not cited as authors (Mina J. Bissell, Caroline Alexander, Ole Behrendtsen, AndrČ Lochter, Leif Lund, Carolyn Sympson, Rabih Talhouk and Nicole Thomasett) for their contributions to the development of this story, and Rick Lyman for help in preparing the manuscript.

References
1. Stetler-Stevenson WG, Aznavoorian S & Liotta LA (1993). Tumor cell interactions with the extracellular matrix during invasion and metastasis. Annual Review of Cell Biology, 9: 541-573.

2. Ashkenas J, Damsky CH, Bissell MJ & Werb Z (1992). Integrins, signalling, and the remodeling of the extracellular matrix. In: Cheresh D & Mecham R (Editors), The Integrins. Academic Press, New York, 79-109.

3. Birkedal-Hansen H, Moore WGI, Boden MK, Windsor LJ, Birkedal-Hansen B, DeCarlo A & Engler JA (1993). Matrix metalloproteinases: A review. Critical Reviews in Oral Biology & Medicine, 4: 197-250.

4. Sato H, Takino T, Okada Y, Cao J, Shinagawa A, Yamamoto E & Seiki M (1994). A matrix metalloproteinase expressed on the surface of invasive tumor cells. Nature, 370: 61-65.

5. Freije JM, Diez-Itza I, Balbin M, Sanchez LM, Blasco R, Tolivia J & Lopez-Otin C (1994). Molecular cloning and expression of collagenase-3, a novel human matrix metalloproteinase produced by breast carcinomas. Journal of Biological Chemistry, 26: 16766-16773.

6. Basset P, Bellocq JP, Wolf C, Stoll I, Hutin P, Limacher JM, Podhajcer OL, Chenard MP, Rio MC & Chambon P (1990). A novel metalloproteinase gene specifically expressed in stromal cells of breast carcinomas. Nature, 348: 699-704.

7. Canete-Soler R, Litzky L, Lubensky I & Muschel RJ (1994). Localization of the 92kd gelatinase mRNA in squamous cell and adenocarcinomas of the lung using in situ hybridization. American Journal of Pathology, 144: 518-527.

8. Witty JP, McDonnell S, Newell KJ, Cannon P, Navre M, Tressler RJ & Matrisian LM (1994). Modulation of matrilysin level in colon carcinoma cell lines affects tumorigenicity in vivo. Cancer Research, 54: 4805-4812.

9. Yamamoto H, Itoh F, Hinoda Y, Senota A, Yoshimoyo M, Nakamura MH, Imai K & Yachi A (1994). Expression of matrilysin mRNA in colorectal adenomas and its induction by truncated fibronectin. Biochemical and Biophysical Research Communications , 201: 657-664.

10. Mignatti P & Rifkin DB (1993). Biology and biochemistry of proteinases in tumor invasion. Physiological Reviews, 73: 161-195.

11. Crawford HC & Matrisian LM (1994). Tumor and stromal expression of matrix metalloproteinases and their role in tumor progression. Invasion and Metastasis, 14: 234-245.

12. Birkedal-Hansen H (1995). Proteolytic remodeling of extracellular matrix. Current Opinion in Cell Biology, 7: 728-735.

13. Leco KJ, Khokha R, Pavloff N, Hawkes SP & Edwards DR (1994). Tissue inhibitor of metalloproteinases-3 (TIMP-3) is an extracellular matrix-associated protein with a distinctive pattern of expression in mouse cells and tissues. Journal of Biological Chemistry, 269: 4352.

14. Pendas AM, Matilla T, Estivill XA & Lopez-Otin C (1995). The human collagenase-3 (CLG-3) gene is located on chromosome 11q22.3 clustered to other members of the matrix metalloproteinase gene family. Genomics, 26: 615-618.

15. Pei D & Weiss SJ (1995). Furin-dependent intracellular activation of the human stromelysin-3 zymogen. Nature, 375: 244-247.

16. Pei D, Majmudar G& Weiss SJ (1994). Hydrolytic inactivation of a breast carcinoma cell-derived serpin by human stromelysin-3. Journal of Biological Chemistry, 269: 25849-25855.

17. Murphy G, Segain JP, O'Shea M, Cockett M, Ioannou C, Lefebvre O, Chambon P & Basset P (1993). The 28-kDa N-terminal domain of mouse stromelysin-3 has the general properties of a weak metalloproteinase. Journal of Biological Chemistry, 268: 15435-15441.

18. Rodgers WH, Osteen KG, Matrisian LM, Navre M, Guidice LC & Gorstein F (1993). Expression and localization of matrilysin, a matrix metalloproteinase in human endometrium during the reproductive cycle. Journal of Obstetrics and Gynecology, 168: 253-60.

19. Rodgers WH, Matrisian LM, Guidice LL, Dsupin B, Cannon P, Svitek C, Gorstein F & Osteen KG (1994). Patterns of matrix metalloproteinase expression in cycling endometrium imply differential functions and regulation by steroid hormones. Journal of Clinical Investigation, 94: 946-953.

20. Wilson CL, Heppner KJ, Rudolph LA & Matrisian LM (1995). The metalloproteinase matrilysin is preferentially expressed by epithelial cells in a tissue-restricted pattern in the mouse. Molecular Biology of the Cell, 6: 851-869.

21. Shapiro SD, Griffin GL, Gilbert DJ, Jenkins NA, Copeland NG, Welgus HG, Senior RM & Ley TJ (1992). Molecular cloning, chromosomal localization, and bacterial expression of a murine macrophage metalloelastase. Journal of Biological Chemistry, 267: 4664-4671.

22. Gack S, Vallon R, Schaper J, Ruther U, Angel P (1994). Phenotypic alteration in Fos-transgenic mice correlate with changes in Fos/Jun-dependent collagenase type I expression. Journal of Biological Chemistry, 269: 10363-10369.

23. Witty JP, Wright JH & Matrisian LM (1995). Matrix metalloproteinases are expressed during ductal and alveolar mammary morphogenesis, and misregulation of stromelysin-1 in transgenic mice induces unscheduled alveolar development. Molecular Biology of the Cell, 6: 1287-1303.

24. Thomasset N, Sympson CJ, Leif LR, Werb Z & Bissell MJ (1994). Overexpression of stromelysin-1 modifies mammary gland stroma in transgenic mice. Molecular Biology of the Cell, 1994, 5(suppl): 181a.

25. Talhouk RS, Bissell MJ & Werb Z (1992). Coordinated expression of extracellular matrix-degrading proteinases and their inhibitors regulates mammary epithelial function during involution. Journal of Cell Biology, 118: 1271-1282.

26. Sympson CJ, Talhouk RS, Alexander CM, Chin JR, Clift SM, Bissell MJ & Werb Z (1994). Targeted expression of stromelysin-1 in mammary gland provides evidence for a role of proteinases in branching morphogenesis and the requirement for an intact basement membrane for tissue-specific gene expression. Journal of Cell Biology, 125: 681-693.

27. Lund LR, Romer J, Thomasset N, Solberg H, Pyke C, Bissell MJ, Dano K & Werb Z (1996). Two distinct phases of apoptosis in mammary gland involution: proteinase-independent and -dependent pathways. Development, 122: 181-193.

28. Helminen HJ & Eriksson JLE (1968). Studies on mammary gland involution. I. On the ultrastructure of the lactating mammary gland. Journal of Ultrastructural Research, 25: 193-213.

29. Lascelles AK & Lee CS (1978). Involution of the mammary gland. In: Larson BL & Smith VR (Editors), Lactation: A Comprehensive Treatise. Vol. 4. Academic Press, New York, 115-77.

30. Marti A, Jehn B, Costello E, Keon N, Ke G, Martin F & Jaggi R (1994). Protein kinase A and AP-1 (c-Fos/JunD) are induced during apoptosis of mouse epithelial cells. Oncogene, 9: 1213-1223.

31. Strange R, Li F, Saurer S, Burkhardt A & Friis RR (1992). Apoptotic cell death and tissue remodeling during mouse mammary gland involution. Development, 115: 49-58.

32. Walker NI, Bennett RE & Kerr JFR (1989). Cell death by apoptosis during involution of the lactating breast in mice and rats. American Journal of Anatomy, 185: 19-32.

33. Williams WL (1942). Normal and experimental mammary involution in the mouse as related to the inception and cessation of lactation. American Journal of Anatomy, 71: 1-41.

34. Klagsbrun M (1992). Mediators of angiogenesis: the biological significance of basic fibroblast growth factor (bFGF)-heparin and heparin sulfate interactions. Seminars in Cancer Biology, 3: 81-87.

35. Sympson CJ, Talhouk RS, Bissell MJ & Werb Z (1995). The role of metalloproteinases and their inhibitors in regulating mammary epithelial morphology and function in vivo. Perspectives in Drug Discovery and Design, 2: 401-411.

36. van den Hooff A (1988). Stromal involvement in malignant growth. Advances in Cancer Research, 50: 159-196.

37. Weidner N, Folkman J, Pozza F, Bevilacqua P, Allred EN, Moore DH, Mei S & Gasparini G (1992). Tumor angiogenesis: A new significant and independent prognostic indicator in early stage breast carcinoma. Journal of the National Cancer Institute, 84: 1875-1887.

38. Sieweke MH & Bissell MJ (1994). The tumor promoting effect of wounding; a possible role for TGF-fl-induced stromal alterations. Critical Reviews in Oncogenesis, 5: 297-311.

39. Boudreau N, Sympson CJ, Werb Z & Bissell MJ (1995). Suppression of ICE and apoptosis in mammary epithelial cells by extracellular matrix. Science, 267: 891-893.

40. Alexander CM, Howard EW, Bissell MJ, & Werb Z (1996). Rescue of mammary epithelial cell apoptosis and entactin degradation by a TIMP-1 transgene. Journal of Cell Biology, in press.

41. Roskelley CD, Desprez PY & Bissell MJ (1994). Extracellular matrix-dependent tissue-specific gene expression in mammary epithelial cells requires both physical and biochemical signal transduction. Proceedings of the National Acadamy of Science, U.S.A., 91: 12378-12382.

42. Streuli CH, Schmidhauser C, Bailey N, Yurchenco P, Skubitz APN, Roskelley C & Bissell MJ (1995). Laminin mediates tissue-specific gene expression in mammary epithelia. Journal of Cell Biology, 129: 591-603.

43. Feng Z, Marti A, Jehn B, Altermatt HJ, Chicaiza G & Jaggi R (1995). Glucocorticoid and progesterone inhibit involution and programmed cell death in the mouse mammary gland. Journal of Cell Biology, 131: 1095-1103.

44. Mignatti P, Robbins E & Robbins DB (1986). Tumor invasion through the human amnion membrane: requirement for a proteinase cascade. Cell, 47: 487-498.

45. Schultz RM, Silberman S, Persky B, Bajkowski AS & Carmichael DF (1988). Inhibition by human recombinant tissue inhibitor of metalloproteinases of human amnion invasion and lung colonization by murine B16-F10 melanoma cells. Cancer Research, 48: 5539-5545.

46. Alvarez OA, Carmichael DF & DeClerck YA (1990). Inhibition of collagenolytic activity and metastasis of tumor cells by a recombinant human tissue inhibitor of metalloproteinases. Journal of the National Cancer Institute, 82: 589-595.

47. Khokha R (1994). Suppression of the tumorigenic and metastatic abilities of murine B16-F10 melanoma cells in vivo by the overexpression of the tissue inhibitor of the metalloproteinases-1. Journal of the National Cancer Institute, 86: 299-304.

48. Alexander CM & Werb Z (1992). Targeted disruption of the tissue inhibitor of metalloproteinases gene increases the invasive behavior of primitive mesenchymal cells derived from embryonic stem cells in vitro. Journal of Cell Biology, 118: 727-739.

49. Werb Z, Tremble PM, Behrendtsen O, Crowley E & Damsky CH (1989). Signal transduction through the fibronectin receptor induces metalloproteinase gene expression. Journal of Cell Biology, 109: 877-89.

50. Tremble PM, Chiquet-Ehrismann R & Werb Z (1994). The extracellular matrix ligands, tenascin and fibronectin collaborate in regulating collagenase gene expression in fibroblasts. Molecular Biology of the Cell, 5: 439-453.

51. Giancotti FG & Ruoslahti E (1990). Elevated levels of the a5b1 fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell, 60: 849-859.

52. Schwartz MA (1993). Signalling by integrins: Implications for tumorigenesis. Cancer Research, 53: 1503-1506.

53. Hynes RO & Plantefaber LC (1991). Integrin receptors for extracellular matrix and their involvement in oncogenic transformation. In: Brugge J, Curran T, Harlow E & McCormick, F (Editors), Origins of Human Cancer: A Comprehensive Review. Cold Spring Harbor Laboratory Press, Plainview, NY. 297-307.

54. Steller H (1995). Mechanisms and genes of cellular suicide. Science 267: 1445-1449,

55. Ashkenas J & Werb, Z (1996). Proteolysis and the biochemistry of life-or-death decisions. Journal of Experimental Medicine, 183: 1-5.

56. Shi L, Nishioka WK, Th'ng J, Bradbury EM, Litchfield DW & Greenberg AH (1994). Premature p34cdc2 activation required for apoptosis. Science 263: 1143-1145.

57. Fotedar R, Flatt J, Gupta S, Margolis RL, Fitzgerald P, Messier H & Fotedar A (1995). Activation-induced T-cell death is cell cycle dependent and regulated by cyclin B. Molecular and Cellular Biology, 15: 932-942.

58. Li CJ, Friedman DJ, Wang C, Metelev V, Pardee AB (1995). Induction of apoptosis in uninfected lymphocytes by HIV-1 Tat protein. Science, 268: 429-431.

59. Shimizu T, O'Connor PM, Kohn KW & Pommier Y (1995). Unscheduled activation of cyclin B1/Cdc2 kinase in human promyelocytic leukemia cell line HL60 cells undergoing apoptosis induced by DNA damage. Cancer Research, 55: 228-231.

60. Meikrantz W, Gisselbrecht S, Tam SW & Schlegel R (1994). Activation of cyclin A-dependent protein kinases during apoptosis. Proceedings of the National Academy of Sciences U.S.A., 91: 3754-3758.

61. Hoang AT, Cohen KJ, Barrett JF, Bergstrom DA, Dang CV (1994). Participation of cyclin A in Myc-induced apoptosis. Proceedings of the National Academy of Sciences U.S.A., 91: 6875-6879.

62. Boudreau N, Werb Z & Bissell MJ (1996). Suppression of apoptosis by basement membrane requires three-dimensional tissue organization and withdrawal from the cell cycle. Proceedings of the National Academy of Sciences U.S.A., in press.

63. Freeman RS, Estus S & Johnson EM Jr (1994). Analysis of cell cycle-related gene expression in postmitotic neurons: selective induction of cyclin D1 during programmed cell death. Neuron, 12: 343-355.

64. Furuya Y, Lin XS, Walsh JC, Nelson WG & Isaacs JT (1995). Androgen ablation-induced programmed death of prostatic glandular cells does not involve recruitment into a defective cell cycle or p53 induction. Endocrinology, 136: 1898-1906.

65. Oberhammer FA, Hochegger K, Froschl G, Tiefenbacher R, Pavelka M (1994). Chromatin condensation during apoptosis is accompanied by degradation of lamin A+B, without enhanced activation of cdc2 kinase. Journal of Cell Biology, 126: 827-837.

Data are summarized from refs 1, 4-12. Development 121, 2079-2090.

Contributed By