Lactoferrin expression in mammary epithelial cells is mediated by changes in cell shape and actin cytoskeleton

Analysis of the mechanisms by which the expression of tissue-specific genes is induced and maintained is necessary for full understanding of normal tissue development and differentiation. Mammary epithelium provides an accessible model for such studies because it continues to cycle through different well characterized developmental stages in the adult animal. Mammary epithelial cells can transduce microenvironmental cues that regulate cellular function from a variety of sources, including soluble hormones and cytokines, cell-cell interactions, and interactions with extracellular matrix (for review see Stoker et al., 1990; Roskelley et al., 1995). In mammary epithelia, signals from the basement membrane, in conjunction with hormones, are required for maintenance of complex three-dimensional alveoli and milk secretion. The extracellular matrix itself influences cellular function by a variety of mechanisms including classical ligand-receptor interaction, promotion of cell-cell cohesion, and modulation of cell shape (Mooney et al., 1992; Adams and Watt, 1993; Juliano and Haskill, 1993; Lin and Bissell, 1993; Frisch and Francis, 1994). Such biochemical and mechanochemical signals from extra-cellular matrix are integrated with those from soluble factors and other cells to coordinate differentiated function.

A culture model which utilizes murine mammary epithelial cells from pregnant or lactating tissues has shown that extra-cellular matrix induces such cells to form complex, highly polarized three-dimensional structures that can vectorially secrete milk specific products (Li et al., 1987; Barcellos-Hoff et al., 1989; Streuli, 1993). Each differentiated function appears to depend on a distinct set of microenvironmental determinants (Lin and Bissell, 1993). Recent evidence obtained using this system has revealed that appropriate cell configuration is a prerequisite for the expression of both β-casein and whey acidic protein (WAP) in response to basement membrane. Transcription and synthesis of β-casein, one of the major products of functionally differentiated murine mammary epithelial cells, is maintained by contact with basement membrane, specifically by laminin binding to β1 integrins (Streuli and Bissell, 1991; Roskelley et al., 1994; Streuli et al., 1995). In addition, cell rounding in response to basement membrane induces a physical signal that is a prerequisite for β-casein expression (Roskelley et al., 1994).

One important parameter of mammary differentiation that is poorly understood is the expression of the iron binding protein lactoferrin (LTF). This evolutionarily conserved glycoprotein is a member of the transferrin gene family. Originally identi-fied in milk, LTF’s iron binding properties are thought to endow it with growth promoting, iron transport and bacteriostatic properties (Chrichton, 1990; Kijlstra, 1991). Unlike β-casein and WAP, however, significant quantities of LTF mRNA and protein are synthesized in normal mammary glands of non-pregnant, non-lactating mammals, including humans (Teng et al., 1989; Campbell et al., 1992; Schanbacher et al., 1993). Furthermore, the expression of LTF mRNA and protein is often decreased or absent in premalignant and malignant human breast tissue compared to normal tissue (Campbell et al., 1992).

In order to evaluate the role of microenvironmental determinants in the regulation of LTF gene expression in mammary epithelial cells, we have utilized the murine mammary epithelial cell line, CID-9 (Schmidhauser et al., 1990) and a clonal derivative, SCp2 (Desprez et al., 1993), both selected from COMMA-1D cells originally derived from mammary tissue of mid-pregnant mice (Danielson et al., 1984). We report that cell-extracellular matrix interactions can regulate LTF gene expression in these cells, but that unlike the requirements for β-casein, the extracellular matrix does not provide a specific regulatory signal. Instead, changes in cell shape (manifested by rounding in these cells) are sufficient for LTF expression. That this is the case is now demonstrated experimentally by induction of rounding in the absence of extracellular matrix and/or cell-cell interaction. Adherence and spreading on unyielding substrata causes active suppression of LTF expression. This effect can be demonstrated in the murine mammary epithelial cell lines as well as in early passage normal human mammary epithelial cells derived from a non-pregnant, non-lactating individual (Stampfer and Yaswen, 1993).

Mouse mammary epithelial cells of the heterogenous cell line CID-9 and clonal derivative, SCp2, were maintained in culture as previously described (Schmidhauser et al., 1992; Desprez et al., 1993). The mouse cells were routinely plated at approximately 3×10⁴ cells/cm², and maintained in DME/F12 medium supplemented with 5% fetal calf serum and insulin (5 μg/ml) unless stated otherwise.

EHS matrix was prepared from EHS ascites tumors passaged in C57BL mice and used as a substratum as described (Barcellos-Hoff et al., 1989; Streuli et al., 1991; Blaschke et al., 1994). In some exper-iments, commercially prepared EHS matrix was used (Matrigel, Collaborative Research, Bedford, MA). Glass coverslips were coated with dried EHS matrix by applying Matrigel diluted 1:5 with serum-free medium, then drying overnight at 37°C in a tissue culture incubator. Purified laminin, poly 2-hydroxyethylmethacrylate (PolyHEMA) and poly-L-lysine were obtained from Sigma Chemical Co. (St Louis, MO.). Laminin was added to the cells as an extracellular matrix overlay at 50 μg/ml as described (Streuli et al., 1995). For suspension cultures, tissue culture plates were coated with 0.8 mg/cm² PolyHEMA dissolved at 12 mg/ml in 95% ethanol and air dried as described (Folkman and Moscona, 1978). Cells in suspension cultures were mounted for immunohistochemistry on glass coverslips coated with 1 mg/ml poly-L-lysine (4×10⁵ kDa).

In experiments in which the concentration of extracellular calcium was varied, calcium chloride was added to calcium-free minimal essential medium (S-MEM, Joklik modified; Life Technologies; Gaithersburg, MD) to give final concentrations of 2 mM or 50 μM. SCp2 cultures were trypsinized, then washed in calcium-free medium with 5% Chelex (Bio-Rad; Richmond, CA)-treated fetal calf serum. The cells were resuspended in serum-free media containing either 2 mM or 50 μM CaCl₂ and either insulin (5 μg/ml) alone, hydrocortisone (1 μg/ml), or insulin plus hydrocortisone. Adherent cultures were seeded at 1-2×10⁶ cells per 100 mm plastic dish, while suspension cultures were seeded at 2-4×10⁶ cells per 100 mm polyHEMA-coated dish. The SCp2 cells were plated at higher density on the polyHEMA coated dishes to compensate for the differences in growth rates between adherent and suspension cultures during the course of the experiments.

In experiments utilizing cytochalasin D, cells were first trypsinized and plated on tissue culture plastic or glass coverslips in medium supplemented with 5% fetal calf serum for at least 24 hours. Cytocha-lasin D (Sigma) was then added to the medium at concentrations ranging from 0.1 to 0.3 μg/ml from a 250 μg/ml stock in 50% dimethyl sulfoxide. Cells were incubated for 12 hours in the presence of cytochalasin D before fixation for analysis by immunofluorescence or harvest for RNA analysis. In some experiments, actinomycin D (Sigma, 10 μg/ml) was employed as an inhibitor of mRNA synthesis.

Human mammary epithelial cells were obtained from reduction mammoplasty tissues. The tissue was mechanically and enzymatically dissagregated to yield epithelial organoids which were subsequently stored frozen. Monolayer cultures were established from organoids in MM medium as previously described (Stampfer, 1985). Normal human mammary epithelial cell cultures were used within the first five passages for experiments.

Cells from suspension or adherent cultures were fixed in 2% paraformaldehyde at ambient temperature for 20 minutes. Cells were immunostained with affinity-purified polyclonal antibody to mouse LTF (Teng et al., 1989) as described (Streuli et al., 1991). The anti-LTF antibody did not cross react with purified transferrin on western blots (data not shown). For visualization, fluoroscein isothiocyanate-conjugated goat anti-rabbit secondary antibodies were applied. Control coverslips were stained with secondary antibodies alone. Cellular nuclei were visualized by staining with 4,6-diamidino-2-phenylindole (Sigma) at 0.1 μg/ml. Actin filaments were visualized by incubation of the fixed cells with 2.5 μg/ml tetramethylrhodamine B isothiocyanate-conjugated phalloidin (Sigma) in phosphate buffered saline.

Analysis of cell growth was performed by visualization of tritiated thymidine or bromodeoxyuridine incorporation into cellular DNA. Cells on coverslips were labelled for 24 hours with 1.0 μCi/ml of [³H]thymidine prior to fixation for immunocytochemistry. After immunocytochemistry, the coverslips were coated with Kodak NTB2 emulsion and exposed for 7-10 days in a dark box with dessicant. After development, the coverslips were mounted in Vectashield mounting medium (Vector Labs, Burlingame, CA) for observation. Labelling and immunostaining for 5-bromo-2′-deoxyuridine (BrdU) incorporation into cellular DNA was performed according to the protocol supplied with the BrdU Labelling and Detection Kit I (Boehringer Mannheim; Indianapolis, IN). Cells were fixed in 70% ethanol, 50 mM glycine, pH 2.0 at −20°C for 20 minutes. Staining reactions for LTF and BrdU were carried out sequentially.

Attached cultures were lysed directly in buffered guanidine thio-cyanate solution, while suspension cultures were centrifuged prior to lysis in the same solution. Total RNA was purified and northern blots prepared as described (Yaswen et al., 1992). The blots were hybridized to the following ³²P-labelled cDNA probes: (a) the 2.2 kb mouse LTF EcoRI insert from pT267 (Pentecost and Teng, 1987); (b) the 540 bp mouse β-casein PstI insert from pBR322-mu-casein (Gupta et al., 1982); (c) the 2.1 kb human LTF EcoRI insert from pHLF1212 (Panella et al., 1991); (d) the 210 bp murine histone 3.2 SalI insert from pm His 3.2 (provided by William Marzluff, U. Florida); (e) the 1.3 kb myc EcoRI-ClaI insert from pBLJ (Battey et al., 1983). Efficiency of gel loading and transfer of different samples was judged by staining with ethidium bromide and hybridization to a ribosomal RNA probe. Quantitative comparisons of relative RNA abundance were performed using ImageQuant software (Molecular Dynamics, Sunnyvale, CA).

SCp2 cells at 50% confluence in a 100 mm plate were co-transfected with 30 μg of the 2.6 mL14-CAT plasmid, containing 2.6 kb of the 5′ flanking sequence from the lactoferrin gene cloned into the polylinker region of pCAT-Basic (Liu and Teng, 1991), along with 3 μg of pSV2Neo (Southern and Berg, 1982). The cells were selected in growth medium containing 200 μg/ml G418. Approximately fifty G418 resistant clones were pooled. The stable transfectants were plated at approximately 3×10⁴ cells/cm² overnight with 2% fetal calf serum on either plastic or polyHEMA-coated (0.03 mg/cm²) dishes. Under these conditions, the cells on plastic attached and spread, while cells on polyHEMA-coated dishes formed attached clusters of rounded cells. In each case, the cells were then switched to serum-free conditions for defined times prior to harvest and preparation of cell extracts. Cells were lysed in 25 mM Tris, 40 mM KCl, 3.5 mM NaCl, 0.5% NP-40, and further solubilized by 2 cycles of freeze-thawing. Soluble protein was collected after centrifugation (10 minutes, 16,000 rpm, room temperature). Lysates were then heated for 10 minutes at 60°C to destroy endogenous CAT activity. Protein concentration was determined by a modified Bradford procedure (Bio-Rad). CAT activity in 10 μg of each extract was determined by standard assay (Maniatis et al., 1982), and quantitated by liquid scin-tillation counting.

Tissue-specific expression of milk protein genes in primary and secondary mouse mammary epithelial cells has previously been shown to depend on lactogenic hormones and contact with reconstituted basement membrane. To begin to determine the specific factors which influence LTF gene expression in mammary epithelium, the CID-9 mouse mammary epithelial cell line, previously demonstrated to be competent for functional differentiation, was cultured in the absence or presence of reconstituted basement membrane (EHS matrix) with or without the mammotrophic hormone, prolactin. LTF mRNA was not observed in total RNA harvested from medium density cultures of CID-9 attached to tissue culture plastic, but was easily demonstrated in RNA from the same cells attached to EHS matrix (Fig. 1A). Unlike the caseins, the expression of LTF was independent of prolactin. This latter finding is in apparent conflict with an earlier study on mouse mammary gland explants (Green and Pastewka, 1978), but is consistent with previous observations on calf mammary epithelial cells cultured on collagen gels (Schanbacher et al., 1993; Talhouk et al., 1993). A possible explanation for this discrepancy between observations on tissue explants and isolated cells could be the presence of paracrine factors in heterogeneous tissue explants which are not present in the isolated cultured cells.

Further characterization of the matrix effect was performed with a clonal derivative of CID-9 cells, SCp2. Growing cultures of SCp2 cells, initially negative for LTF mRNA expression, were deprived of fetal calf serum for 72 hours, then given 50 μg/ml of purified laminin, a major basement membrane component, or 5% fetal calf serum in the culture medium for an additional 6 or 48 hours prior to harvest. The cells exposed to laminin aggregated into clusters of round, refractile cells, as previously described (Desprez et al., 1993; Roskelley et al., 1994; Streuli et al., 1995). In the cultures exposed to fetal calf serum, the SCp2 cells achieved confluence by 48 hours. As illustrated in Fig. 1B, incubation with either laminin or fetal calf serum was sufficient to induce LTF mRNA expression by 48 hours. This result was in striking contrast to that observed for β-casein mRNA, which specifically required laminin for expression. The fetal calf serum result suggested that changes in cell-cell interactions, cell growth, and/or cell shape could replace the conditions for LTF induction provided by basement membrane.

Having determined that confluent cultures of SCp2 cells on plastic express LTF mRNA, we examined the roles of cell-cell interaction and cell shape changes in this inductive process by culturing SCp2 cells on polyHEMA-coated substrata for 48 hours in the absence of serum or exogenous extracellular matrix. The neutrally charged plastic polymer, polyHEMA, prevented cell-substratum attachment and spreading. Under these conditions, SCp2 cells aggregated to form rounded cell clusters in suspension, and LTF mRNA (Fig. 1C) as well as protein expression (Fig. 2) was induced. LTF mRNA was expressed as long as insulin was present in the serum-free medium. Hydrocortisone, while not absolutely necessary for LTF mRNA expression, elevated the level of expression when present.

To determine whether LTF expression could be induced in SCp2 cells in the absence of cell-cell interactions, the levels of extracellular Ca²⁺ were lowered. In suspension cultures, lowering the extracellular calcium concentration from 2 mM to 50 μM decreased the number of large cell clusters and increased the proportion of single cells. In adherent cultures, lowering the extracellular calcium concentration to 50 μM reduced the extent of cell spreading and gave the cells a more refractile appearance. In both suspended and attached cultures, lowered calcium levels were accompanied by increased LTF mRNA expression, suggesting that calcium dependent cell-cell cohesion is not necessary for LTF expression.

To show conclusively that cell-cell contact is not required and that cell rounding alone is sufficient to trigger LTF expression, we asked whether LTF protein expression could be induced in single, suspended mammary epithelial cells in the absence of either extracellular matrix or intercellular contact. LTF protein expression was compared immunocytochemically in suspended SCp2 cells present as single cells or in clusters in low calcium medium. This was accomplished by immobilizing the suspended cells by unit gravity sedimentation onto polylysine-coated coverslips immediately prior to fixation and staining. LTF protein expression was apparent in virtually all cells present as single cells or small clusters harvested from polyHEMA cultures (Table 1).

To determine the timecourse of LTF mRNA induction following release from a flattened, adherent state, SCp2 cells cultured on plastic were trypsinized and re-plated back onto either plastic or into suspension on polyHEMA for varying times. Whereas trypsin treatment disrupted the adherence of the cells to plastic within minutes, the SCp2 cells required 48 hours of culture on polyHEMA to achieve significant levels of LTF mRNA expression (Fig. 3). The relatively slow induction of LTF mRNA may reflect time required for synthesis or degradation of associated regulatory factors. Cells replated onto plastic exhibited slight transient LTF mRNA expression at the 12 and 24 hour time points, but levels were again undetectable by 48 hours. Hybridization of the same northern blot to a histone probe, which is used as a marker of proliferation, showed an inverse correlation with LTF expression (Fig. 3). Thus, LTF and growth-dependent gene expression may be mutually exclusive.

Spreading and growth of SCp2 cells on plastic or glass could be largely prevented by removing fetal calf serum from the medium within 12 hours after plating. Although the cells were able to bind to the plastic or glass under these conditions, the majority remained round and exhibited LTF protein (data not shown). DNA synthesis and LTF expression were measured in individual cells maintained on glass coverslips under serum containing and serum-free conditions. Under both conditions, there was an inverse correlation between DNA synthesis and LTF expression (Table 2). In order to determine whether cessation of cell growth, by itself, was sufficient for induction of LTF expression, the SCp2 cells were incubated under a condition which maintained the cells’ spread morphology, while inhibiting growth. Such a condition occurred when the cells were plated onto a substratum composed of air-dried EHS matrix. EHS matrix subjected to dehydration was found to be irreversibly altered so that cells placed in contact with it were no longer able to pull the matrix over them. When plated on this substratum in the presence of fetal calf serum, the SCp2 cells spread as on glass (Fig. 4). Unlike cultures bound directly to glass, however, cultures on dried EHS matrix did not round up when deprived of serum. In order to efficiently block growth on dried EHS matrix, it was necessary to remove insulin as well as fetal calf serum. Such cultures did exhibit decreased rates of DNA synthesis, but did not exhibit increased LTF protein expression. In parallel cultures maintained on fully hydrated EHS matrix in the same medium within the same dishes, the majority of cells exhibited the round, refractile phenotype and abundant LTF protein. This indicates that cessation of growth, per se, is not sufficient, and that change in cell shape may also be required for induction of LTF protein expression in the SCp2 cells. Furthermore, the experiment suggests that a non-diffusible aspect of EHS matrix is capable of substituting for the insulin requirement of LTF expressing cells during short-term culture.

Separation of cell-matrix interactions, cell-cell interactions, and cessation of growth from induction of LTF expression left the possibility that changes in cell shape might more directly influence LTF expression. Dynamic changes in microfilament abundance and orientation are known to accompany cell shape changes during cell growth and differentiation. The involvement of microfilaments in regulation of LTF expression was examined by treating SCp2 cells growing in serum-containing medium on plastic with cytochalasin D, an inhibitor of microfilament assembly. Twelve hours of treatment of spread cells with cytochalasin D disrupted microfilaments, caused most cells to round up, and increased expression of LTF protein (Fig. 5). Concentrations of cytochalasin D between 0.1 and 0.3 μg/ml caused a dose-dependent increase in the level of LTF mRNA expression in parallel cultures (Fig. 6). Removal of the drug from the cells and continued culture for twelve hours in medium containing 5% fetal calf serum reversed the morphology and allowed the cells to reenter the cell cycle (data not shown).

To determine whether the shape-dependent changes in LTF mRNA levels were mediated by differences in rates of transcription, a chimeric gene, containing 2.6 kb of the 5′-LTF promoter region fused to a bacterial chloramphenicol acetyltransferase (CAT) reporter gene (Liu and Teng, 1991), was stably transfected into SCp2 cells. When pooled transfectants were switched from flattened, spread conditions on plastic to rounded conditions on polyHEMA, a slow increase in the level of CAT activity was recorded (Fig. 7). At 48 hours, the increase was approximately threefold in two separate experiments. To determine whether increased mRNA stability contributes to LTF mRNA abundance when cells are induced to round up, RNA synthesis inhibitor actinomycin D was added at the time of cytochalasin D exposure or serum deprivation. In each case, after 12 hours of exposure to the rounding stimulus, levels of LTF mRNA were higher in the presence of actinomycin D (Fig. 8). This was not a general effect on cellular mRNAs, as c-myc mRNA decayed normally in the presence of the inhibitor. A possible explanation for the increased abundance of LTF mRNA in the presence of an inhibitor of RNA synthesis is that transcription of an unstable factor is required for degradation of LTF mRNA. In the absence of continued transcription, the LTF mRNA that has already been made persists, whereas in the presence of continued transcription, the short-lived regulatory factor continues to cause LTF mRNA degradation.

While the previous experiments suggested that expression of LTF is regulated at the mRNA level by microfilament-dependent changes in cell shape, the possibility remained that the phenomenon observed was peculiar to the immortal murine mammary cell line used in the studies. In human mammary epithelial cells from normal tissues, as in the murine system, basement membrane has been shown to regulate several aspects of development and differentiation, including formation of three-dimensional alveolar-like structures, basal deposition of endogenous basement membrane and apical secretion of sialomucins (Petersen et al., 1992; Stampfer and Yaswen, 1992; Howlett et al., 1995). In order to investigate whether the signalling requirements for LTF expression in normal finite lifespan human mammary epithelial cells were similar to those observed in the murine cell line, we compared RNA from breast organoids, isolated directly from surgically removed reduction mammoplasty tissue, and from cells in early passage, cultured on plastic, EHS-matrix and polyHEMA. LTF mRNA was abundantly expressed in organoids, but was down-regulated in human mammary epithelial cells cultured on plastic (Fig. 9). In contrast, under identical medium conditions and in the absence of added prolactin, fourth passage human mammary epithelial cells plated on EHS-matrix showed robust expression of LTF mRNA. In the absence of EHS-matrix, LTF mRNA was also expressed in fifth passage human mammary epithelial cells cultured in suspension on polyHEMA coated dishes. LTF mRNA expression was also observed in RNA from cultures of human mammary epithelial cells allowed to grow past confluence (data not shown). Unlike the early passage human mammary epithelial cells grown in serum-containing medium, human mammary epithelial cells selected for long term growth potential in serum-free medium (Stampfer, 1985; Stampfer and Yaswen, 1993) were not inducible for LTF expression. Thus, the early passage normal human mammary epithelial cells show regulation of LTF mRNA similar to that demonstrated for the murine mammary cell line, although post-selection human mammary epithelial cells, which may represent specific cell lineages, do not appear to be capable of expressing LTF mRNA under any of the culture conditions employed.

The experiments presented here indicate that although exogenous extracellular matrix is capable of supporting LTF expression, it is possible to mimic its action by experimentally manipulating cell shape and actin cytoskeleton. Modulation of cell shape is known to regulate other cellular functions (reviewed by Watt, 1986; Ben-Ze’ev, 1991; Ingber, 1993; Singhvi et al., 1994). Early evidence that cell shape changes have a functional significance came from experiments designed to force cell shape changes independently of exposure to extra-cellular matrix (Folkman and Moscona, 1978). Under these circumstances, changes in morphology caused by depriving cells of substratum contact were shown to directly influence both cellular growth and differentiation (Allan and Harrison, 1980; Shannon and Pitelka, 1981; Glowacki et al., 1983). More recently, it was shown that extracellular matrix controls the growth of endothelial and hepatic cells by regulation of cell spreading (Ingber, 1990; Mooney et al., 1992; Chen et al., 1997), and that it selectively stimulates the transcription of hepatocyte-specific genes by promoting cell rounding (DiPersio et al., 1991). Cell rounding also initiates tissue specific gene expression in keratinocytes, steroidogenic cells, retinal pigmented epithelial cells and osteoblasts (Watt et al., 1988; Opas, 1989; Bidwell et al., 1993; Roskelley and Auersperg, 1993).

Both extracellular matrix and cells themselves carry information which governs the specificity of particular cell shapes. Distinct cell types adopt different shapes even when presented with the same extracellular ligand: Sertoli cells, for example, assume a columnar morphology while Schwann cells become elongated on laminin substrata (Kleinman et al., 1984). Alternatively, the same cell type may adopt different shapes in response to different extracellular matrix ligands; chondrocyte morphology is polygonal on plasma fibronectin but elongated on cellular fibronectin (West et al., 1979, 1984). The shapes of mammary epithelial cells, like other adherent cell types, are very sensitive to changes in surrounding extracellular matrix and cell density. Plasma fibronectin in the growth medium or cellular fibronectin secreted by the mammary epithelial cells themselves (Stampfer et al., 1981; Hynes and Yamada, 1982) promote cell spreading and the formation of focal contacts (Burridge et al., 1988). EHS matrix or purified laminin, along with cell crowding, promote rounding in the same cells (Desprez et al., 1993; Roskelley et al., 1994; Streuli et al., 1995).

LTF is synthesized in rounded mammary epithelial cells in the absence of all other exogenous microenvironmental factors with the exception of insulin, which may be a general survival factor (Merlo et al., 1995). This ‘default’ state apparently does not require positive external regulation and thus may represent the lowest level of an extracellular matrix-mediated signalling hierarchy that is operative during each cycle of pregnancy, lactation and involution (Roskelley et al., 1995). Unlike other proteins found in milk, LTF is made in abundance by mammary epithelial cells throughout mammary development, including non-pregnant, non-lactating and involuting stages. LTF expression may be actively suppressed only at times when the cellular synthetic apparatus is needed for other processes such as growth. Whether cell shape changes play a role in LTF suppression in vivo remains to be determined. However, changes in extracellular matrix composition and cell density do occur during wound healing and tissue remodeling, and could affect cell shape-dependent processes (Ruoslahti, 1997).

The suppression of LTF expression in spread cells may be dependent upon redistribution of specific integrins. Like receptors of many soluble ligands, the physical clustering of integrins causes allosteric interactions resulting in enhanced intracellular enzymatic activity (Schwartz, 1993). Unlike many growth factor receptors, integrins themselves have no intrinsic enzymatic activity, and their ability to mediate signaling depends on their ability to assemble cytoskeletal components and intracellular signaling networks (Damsky and Werb, 1992). In many cases, integrin-dependent responses are inseparable from hormone or growth factor requirements. Integrin-dependent expression of β-casein, for example, requires the presence of prolactin (Schmidhauser et al., 1992). As a result, it has been difficult to distinguish the elements of integrin-dependent signal transduction from those of prolactin. Since neither positive nor negative regulation of LTF in cultured mammary epithelial cells requires additional soluble factors such as prolactin, this gene may provide an unambiguous readout for analysis of cell shape-dependent signal transduction. Our current hypothesis is that spreading of mammary epithelial cells on endogenous or exogenous fibronectin, vitronectin, or dehydrated EHS matrix suppresses LTF expression through integrindependent changes in cytoskeleton structure. The fact that overlay with purified laminin is able to alter the shape of such cells and relieve suppression of LTF expression suggests that there is crosstalk among the internal signal transduction pathways which respond to different extracellular matrix components.

Shape directed changes in cytoskeletal structure may influence LTF mRNA turnover in a number of ways. Organization of microfilaments facilitates the transport of soluble factors, such as MAP kinases (Chen et al., 1994) and c-abl (Lewis et al., 1996), into the nucleus where they phosphorylate and alter the activities of specific transcription factors. In addition, restructuring of the cytoskeleton can lead to changes in the structure of insoluble nuclear components which govern accessibility of chromatin to transcription apparatus (Pienta and Coffey, 1992; Bidwell et al., 1993; Lelievre et al., 1996). Suppression of LTF expression in spread cells is mediated by transcriptional and post-transcriptional mechanisms. Our data indicate that: (a) LTF promoter activity is increased in clustered, rounded cells in comparison to flattened, spread cells, and (b) LTF mRNA is preferentially stabilized in the presence of an inhibitor of transcription. The latter data suggest that an unstable activity results in destruction of LTF mRNA. Post-transcriptional regulation of mRNA stability plays an important role in the expression of a number of proteins. Cis-acting elements involved in regulating mRNA turnover rates have been described, and vary greatly in both sequence and location. Such an mRNA destabilization element has been described in the 3′ untranslated region of bovine mammary LTF mRNA (Pattanajitvillai et al., 1994). In the presence of RNA synthesis inhibitors, stabilization of transcripts containing such elements sometimes occurs. For example, actinomycin D stabilizes chimeric β-globin transcripts containing mRNA destabilizing elements from c-fos (Shyu et al., 1989), interferon-β (Whittemore and Maniatis, 1990), or tissue factor (Ahern et al., 1993) transcripts. Like LTF mRNA therefore, these elements require ongoing transcription for destabilization. The effect of actinomycin D could be due to repression of transcription of a labile mRNA encoding a labile protein involved in mediating mRNA turnover. Very little is known about targeting of RNases to specific transcripts. Further analysis of the cell shape dependent regulation of LTF expression may reveal important information about how cells alter gene expression in response to growth or differentiation requirements.

We thank Dr Mina Bissell for her thoughtful guidance and support, and Dr Carolyn Larabell for assistance with confocal imaging exper-iments. This work was supported by the Health Effects Research Division, Office of Health and Environmental Research, US Department of Energy under contract DE-ACO3-76SF00098, the US Army Medical Research and Development, Aquisition and Logistics Command (Prov.), National Cancer Institute grants CA-24844 and CA-57621, a grant from the Canadian Breast Cancer Research Initiative, and a postdoctoral fellowship from the Ligue National Francaise contre le Cancer and the Association Claude Bernard. Opinions, interpretations, conclusions and recommendations are those of the authors and are not necessarily endorsed by the US Army.