Presence of BRCA1 and BRCA2 proteins in human milk fat globules after delivery.

We evaluated BRCA1 and BRCA2 oncosuppressor protein expression in 26 milk samples in women just after delivery. The quantification of BRCA1 and BRCA2 proteins was performed in isolated milk fat globules using an affinity chromatography strategy. The amounts of BRCA1 and BRCA2 proteins were found to be similar. We explained the presence of BRCA1 and BRCA2 proteins in human milk fat globules by the fact that they are formed by exocytosis of lipids from epithelial cells of the mammary gland and are enveloped by plasma membrane from the apical part of the milk-secreting cells. This raises the possibility that BRCA1 and BRCA2 proteins are a protective response to proliferation and play a possible role in newborn nutrition.     

The last exon of SNAP-23 regulates granule exocytosis from mast cells

SNAP-25 and its ubiquitous homolog SNAP-23 are members of the SNARE family of proteins that regulate membrane fusion during exocytosis. Although SNAP-23 has been shown to participate in a variety of intracellular transport processes, the structural domains of SNAP-23 that are required for its interaction with other SNAREs have not been determined. By employing deletion mutagenesis we found that deletion of the amino-terminal 18 amino acids of SNAP-23 (encoded in the first exon) dramatically inhibited binding of SNAP-23 to both the target SNARE syntaxin and the vesicle SNARE vesicle-associated membrane protein(VAMP). By contrast, deletion of the carboxyl-terminal 23 amino acids (encoded in the last exon) of SNAP-23 does not affect SNAP-23 binding to syntaxin but profoundly inhibits its binding to VAMP. To determine the functional relevance of the modular structure of SNAP-23, we overexpressed SNAP-23 in cells possessing the capacity to undergo regulated exocytosis. Expression of human SNAP-23 in a rat mast cell line significantly enhanced exocytosis, and this effect was not observed in transfectants expressing the carboxyl-terminal VAMP-binding mutant of SNAP-23. Despite considerable amino acid identity, we found that human SNAP-23 bound to SNAREs more efficiently than did rat SNAP-23. These data demonstrate that the introduction of a "better" SNARE binder into secretory cells augments exocytosis and defines the carboxyl terminus of SNAP-23 as an essential regulator of exocytosis in mast cells.     

The role of exocytosis in the apocrine secretion of milk lipid globules in mouse mammary gland during lactogenesis.

Functional relations between exocytotic vesicle membranes, plasmalemma and milk fat globule membranes (MFGM) were studied during the final stages of mouse mammary gland differentiation, in the gland during full lactation and in the postpartum gland in which the synthesis of secretory products was partly inhibited by application of 2-Br-alpha-ergocryptine. Analysis of ultrathin sections, freeze-fracture replicas, scanning electron microscopy and application of a cytochemical marker filipin showed that the apocrine secretion of lipid globules was closely related to the exocytosis of milk proteins. During the last days of gestation the secretion of lipid globules resulted from many exocytotic events of the secretory vesicles that accumulated and fused around the cytoplasmic lipid droplets. Seldom the lipid droplet protruded partly into the gland lumen and a part of its surface became covered with the apical plasmalemma. Although apical plasmalemma became more important in the formation of MFGM in the postpartum period, we could still confirm a direct contribution of secretory vesicle membranes to the final detachment of the lipid globule. The application of 2-Br-alpha-ergocryptine hindered the apocrine secretion of the lipid globules and a situation similar to the situation in the prepartum gland was observed.     

Intracellular origin and secretion of milk fat globules

The cream or fat fraction of milk consists of fat droplets composed primarily of triacylglycerols that are surrounded by cellular membranes. In this review we discuss what is known about how these droplets are formed in and secreted by mammary epithelial cells during lactation. This secretion mechanism, which appears to be unique, is unlike the exocytotic mechanism used by other cell types to secrete lipids. Milk fat globules originate as small, triacylglycerol-rich, droplets that are formed on or in endoplasmic reticulum membranes. These droplets are released from endoplasmic reticulum into the cytosol as microlipid droplets coated by proteins and polar lipids. Microlipid droplets can fuse with each other to form larger cytoplasmic lipid droplets. Droplets of all sizes appear to be unidirectionally transported to apical cell regions by as yet unknown mechanisms that may involve cytoskeletal elements. These lipid droplets appear to be secreted from the cell in which they were formed by being progressively enveloped in differentiated regions of apical plasma membrane. While plasma membrane envelopment appears to be the primary mechanism by which lipid droplets are released from the cell, a mechanism involving exocytosis of lipid droplets from cytoplasmic vacuoles also has been described. As discussed herein, while we have a general overview of the steps leading to the fat globules of milk, virtually nothing is known about the molecular mechanisms involved in milk fat globule formation, intracellular transit, and secretion.     

Phosphorylation of SNAP-23 regulates exocytosis from mast cells

Regulated exocytosis is a process in which a physiological trigger initiates the translocation, docking, and fusion of secretory granules with the plasma membrane. A class of proteins termed SNAREs (including SNAP-23, syntaxins, and VAMPs) are known regulators of secretory granule/plasma membrane fusion events. We have investigated the molecular mechanisms of regulated exocytosis in mast cells and find that SNAP-23 is phosphorylated when rat basophilic leukemia mast cells are triggered to degranulate. The kinetics of SNAP-23 phosphorylation mirror the kinetics of exocytosis. We have identified amino acid residues Ser(95) and Ser(120) as the major phosphorylation sites in SNAP-23 in rodent mast cells. Quantitative analysis revealed that approximately 10% of SNAP-23 was phosphorylated when mast cell degranulation was induced. These same residues were phosphorylated when mouse platelet degranulation was induced with thrombin, demonstrating that phosphorylation of SNAP-23 Ser(95) and Ser(120) is not restricted to mast cells. Although triggering exocytosis did not alter the absolute amount of SNAP-23 bound to SNAREs, after stimulation essentially all of the SNAP-23 bound to the plasma membrane SNARE syntaxin 4 and the vesicle SNARE VAMP-2 was phosphorylated. Regulated exocytosis studies revealed that overexpression of SNAP-23 phosphorylation mutants inhibited exocytosis from rat basophilic leukemia mast cells, demonstrating that phosphorylation of SNAP-23 on Ser(120) and Ser(95) modulates regulated exocytosis by mast cells.     

Plasma membrane targeting of exocytic SNARE proteins

SNARE proteins play a central role in the process of intracellular membrane fusion. Indeed, the interaction of SNAREs present on two opposing membranes is generally believed to provide the driving force to initiate membrane fusion. Eukaryotic cells express a large number of SNARE isoforms, and the function of individual SNAREs is required for specific intracellular fusion events. Exocytosis, the fusion of secretory vesicles with the plasma membrane, employs the proteins syntaxin and SNAP-25 as plasma membrane SNAREs. As a result, exocytosis is dependent upon the targeting of these proteins to the plasma membrane; however, the mechanisms that underlie trafficking of exocytic syntaxin and SNAP-25 proteins to the cell surface are poorly understood. The intracellular trafficking itinerary of these proteins is particularly intriguing as syntaxins are tail-anchored (or Type IV) membrane proteins, whereas SNAP-25 is anchored to membranes via a central palmitoylated domain-there is no common consensus for the trafficking of such proteins within the cell. In this review, we discuss the plasma membrane targeting of these essential exocytic SNARE proteins.     

Production of low-lactose milk by ectopic expression of intestinal lactase in the mouse mammary gland

We have investigated, in mice, an in vivo method for producing low-lactose milk, based on the creation of transgenic animals carrying a hybrid gene in which the intestinal lactase-phlorizin hydrolase cDNA was placed under the control of the mammary-specific alpha-lactalbumin promoter. Transgenic females expressed lactase protein and activity during lactation at the apical side of mammary alveolar cells. Active lactase was also secreted into milk, anchored in the outer membrane of fat globules. Lactase synthesis in the mammary gland caused a significant decrease in milk lactose (50-85%) without obvious changes in fat and protein concentrations. Sucklings nourished with low-lactose milk developed normally. Hence, these data validate the use of transgenic animals expressing lactase in the mammary gland to produce low-lactose milk in vivo, and they demonstrate that the secretion of an intestinal digestive enzyme into milk can selectively modify its composition.     

Ternary SNARE complexes are enriched in lipid rafts during mast cell exocytosis

Lipid rafts are membrane microdomains rich in cholesterol and glycosphingolipids that have been implicated in the regulation of intracellular protein trafficking. During exocytosis, a class of proteins termed SNAREs mediate secretory granule-plasma membrane fusion. To investigate the role of lipid rafts in secretory granule exocytosis, we examined the raft association of SNARE proteins and SNARE complexes in rat basophilic leukemia (RBL) mast cells. The SNARE protein SNAP-23 co-localized with a lipid raft marker and was present in detergent-insoluble lipid raft microdomains in RBL cells. By contrast, only small amounts (<20%) of the plasma membrane SNARE syntaxin 4 or the granule-associated SNARE vesicle-associated membrane protein (VAMP)-2 were present in these microdomains. Despite this, essentially all syntaxin 4 and most of VAMP-2 in these rafts were present in SNARE complexes containing SNAP-23, while essentially none of these complexes were present in nonraft membranes. Whereas SNAP-23 is membrane anchored by palmitoylation, the association of the transmembrane protein syntaxin 4 with lipid rafts was because of its binding to SNAP-23. After stimulating mast cells exocytosis, the amount of syntaxin 4 and VAMP-2 present in rafts increased twofold, and these proteins were now present in raft-associated phospho-SNAP-23/syntaxin 4/VAMP-2 complexes, revealing differential association of SNARE fusion complexes during the process of regulated exocytosis.     

Mitosis in Milk Secreting Epithelial Cells of Mammary Gland An Ultrastructural Study

In all stages of lactation mitotic configurations were observed in mammary gland epithelial cells of rats. An electron microscopic study is presented which shows that ultrastructure of such mitotic stages is normal and that mitotic cells contain typical products of milk secreting cells such as casein micelle-containing vesicles and milk fat droplets. Such secretory products can even be observed in the immediate vicinity of the chromosomes and microtubules of the spindle apparatus. The endoplasmic reticulum of mitotic cells appeared altered in that it did not show typical cisternal stacks characteristic of interphase cells. While the numbers of such mitotic cells were very low, especially from the second week of lactation on (always less than 0.1% of the milk secreting epithelial cells encountered), the observations clearly demonstrate that differentiation for milk secretory activity and cells division are not mutually exclusive. We conclude that postpartum growth of mammary gland epithelium and replacement of epithelial cells lost during desquamation into the milk liquids can occur by division of existing differentiated milk secreting cells and does not require mitotic activity of non-lactating 'stem cells' which are not observed in lactating alveoli.     

Syntaxin isoform specificity in the regulation of renal H+-ATPase exocytosis

Intercalated and inner medullary collecting duct (IMCD) cells of the kidney mediate the transport of H+ by a plasma membrane H+-ATPase. The rate of H+ transport in these cells is regulated by exocytic insertion of H+-ATPase-laden vesicles into the apical membrane. We have shown that the exocytic insertion of proton pumps (H+-ATPase) into the apical membrane of rat IMCD cells, in culture, involves SNARE proteins (syntaxin (synt), SNAP-23, and VAMP). The membrane fusion complex observed in IMCD cells with the induction of proton pump exocytosis not only included these SNAREs but also the H+-ATPase. Based on these observations, we suggested that the targeting of these vesicles to the apical membrane is mediated by an interaction between the H+-ATPase and a specific t-SNARE. To evaluate this hypothesis, we utilized a "pull-down" assay in which we identified, by Western analysis, the proteins in a rat kidney medullary homogenate that complexed with glutathione S-transferase (GST) fusion syntaxin isoforms attached to Sepharose 4B-glutathione beads. The syntaxin isoforms employed were 1A, 1B, 2, 4, 5, and also 1A that was truncated to exclude the H3 SNARE binding domain (synt-1ADeltaH3). All full-length syntaxin isoforms formed complexes with SNAP-23 and VAMP. Neither GST nor synt-1ADeltaH3 formed complexes with these SNAREs. H+-ATPase (subunits E, a, and c) bound to syntaxin-1A and to a lesser extent to synt-1B but not to synt-1ADeltaH3 or synt-2, -4, and -5. In cultured IMCD cells transfected to express syntaxin truncated for the membrane binding domain (synt-DeltaC), expression of synt-1ADeltaC, but not synt-4DeltaC, inhibited H+-ATPase exocytosis. In conclusion, because all full-length syntaxins examined bound VAMP-2 and SNAP-23, but only non-H3-truncated syntaxin-1 bound H+-ATPase, and synt-1ADeltaC expression by intact IMCD cells inhibited H+-ATPase exocytosis, it is likely that the H+-ATPase binds directly to the H3 domain of syntaxin-1 and not through VAMP-2 or SNAP-23. Interaction between the syntaxin-1A and H+-ATPase is important in the targeted exocytosis of the proton pump to the apical membrane of intercalated cells.     

Mammary gland secretion: hormonal coordination of endocytosis and exocytosis

The mammary epithelium coordinates the uptake of milk precursors and the transport of milk components in order to produce milk of relatively constant composition at a particular stage of lactation, as long as the mammary gland is healthy. The mammary epithelial cell controls the uptake of blood-borne molecules at its basal side and the release of products into milk at its apical side, through mechanisms of internalization (endocytosis) and mechanisms of release (exocytosis). These events are strictly dependent on the physiological stage of the mammary gland. This review addresses the mechanisms responsible for these processes and points out new questions that remain to be answered concerning possible interconnections between them, for an optimal milk secretion.     

Septin Dynamics Are Essential for Exocytosis

Septins are a family of 14 cytoskeletal proteins that dynamically form hetero-oligomers and organize membrane microdomains for protein complexes. The previously reported interactions with SNARE proteins suggested the involvement of septins in exocytosis. However, the contradictory results of up- or down-regulation of septin-5 in various cells and mouse models or septin-4 in mice suggested either an inhibitory or a stimulatory role for these septins in exocytosis. The involvement of the ubiquitously expressed septin-2 or general septin polymerization in exocytosis has not been explored to date. Here, by nano-LC with tandem MS and immunoblot analyses of the septin-2 interactome in mouse brain, we identified not only SNARE proteins but also Munc-18-1 (stabilizes assembled SNARE complexes), N-ethylmaleimide-sensitive factor (NSF) (disassembles SNARE complexes after each membrane fusion event), and the chaperones Hsc70 and synucleins (maintain functional conformation of SNARE proteins after complex disassembly). Importantly, α-soluble NSF attachment protein (SNAP), the adaptor protein that mediates NSF binding to the SNARE complex, did not interact with septin-2, indicating that septins undergo reorganization during each exocytosis cycle. Partial depletion of septin-2 by siRNA or impairment of septin dynamics by forchlorfenuron inhibited constitutive and stimulated exocytosis of secreted and transmembrane proteins in various cell types. Forchlorfenuron impaired the interaction between SNAP-25 and its chaperone Hsc70, decreasing SNAP-25 levels in cultured neuroendocrine cells, and inhibited both spontaneous and stimulated acetylcholine secretion in mouse motor neurons. The results demonstrate a stimulatory role of septin-2 and the dynamic reorganization of septin oligomers in exocytosis.     

Role of SNAP-23 in trafficking of H+-ATPase in cultured inner medullary collecting duct cells

The trafficking of H+-ATPase vesicles to the apical membrane of inner medullary collecting duct (IMCD) cells utilizes a mechanism similar to that described in neurosecretory cells involving soluble N-ethylmaleimide-sensitive factor attachment protein target receptor (SNARE) proteins. Regulated exocytosis of these vesicles is associated with the formation of SNARE complexes. Clostridial neurotoxins that specifically cleave the target (t-) SNARE, syntaxin-1, or the vesicle SNARE, vesicle-associated membrane protein-2, reduce SNARE complex formation, H+-ATPase translocation to the apical membrane, and inhibit H+ secretion. The purpose of these experiments was to characterize the physiological role of a second t-SNARE, soluble N-ethylmaleimide-sensitive factor attachment protein (SNAP)-23, a homologue of the neuronal SNAP-25, in regulated exocytosis of H+-ATPase vesicles. Our experiments document that 25-50 nM botulinum toxin (Bot) A or E cleaves rat SNAP-23 and thereby reduces immunodetectable and (35)S-labeled SNAP-23 by >60% within 60 min. Addition of 25 nM BotE to IMCD homogenates reduces the amount of the 20 S-like SNARE complex that can be immunoprecipitated from the homogenate. Treatment of intact IMCD monolayers with BotE reduces the amount of H+-ATPase translocated to the apical membrane by 52 +/- 2% of control and reduces the rate of H+ secretion by 77 +/- 3% after acute cell acidification. We conclude that SNAP-23 is a substrate for botulinum toxin proteolysis and has a critical role in the regulation of H+-ATPase exocytosis and H+ secretion in these renal epithelial cells.     

Secretory membranes of the lactating mammary gland

Summary The lactating mammary gland is one of the most highly differentiated and metabolically active organs in the body. Membranes of the lactating mammary cell have important roles in transmitting from one membrane to another of hormonal information and in milk secretion, which is the final event. During milk secretion, the projection of the surface membrane into the alveolar lumen by enveloping intracellular lipid droplets with the apical plasma membrane is one of the most remarkable aspects of biological membrane action throughout nature.This review focuses on current knowledge about membranes in the lactating mammary gland. (1) Advances in the isolation and properties of membranes, especially the plasma membrane and Golgi-derived secretory vesicles, concerned with milk secretion from the lactating mammary gland are described. (2) Milk serum components are secreted by fusing the membranes of secretory vesicles that condense milk secretions with the plasma membrane in the apical regions. This occurs through the formation of a tubular-shaped projection and vesicular depression in a ball-and-socket configuration, as well as by simple fusion. (3) Intracellular lipid droplets are directly extruded from the mammary epithelial cells by progressive envelopment of the plasma membranes in the apical regions. (4) The balance between the surface volume lost in enveloping lipid droplets and that provided by fusion of the secretory vesicle and other vesicles with the apical plasma membrane is discussed. (5) The membrane surrounding a milk fat globule, which is referred to as the milk fat globule membrane (MFGM), is composed of at least the coating membrane of an intracellular lipid droplet, of the apical plasma membrane and secretory vesicle membrane, and of a coat material. Consequently, MFGM is molecularly different from the plasma membrane in composition. (6) MFGM of bovine milk is structurally composed of an inner coating membrane and outer plasma membrane just after segregation. These two membranes are fused and reorganized through a process of vesiculation and fragmentation to stabilize the fat globules. Hypothetical structural models for MFGM from bovine milk fat globules just after secretion and after rearrangement are proposed.Abbrevations MFGM milk fat globule membrane - HEPES N-2-hydroxylpiperazine-N-2-ethanesulfonic acid - INT 2-(p-indophenyl)-3-(p-nitrophenyl)-5-phenyltetrazolium - SDS-PAGE polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate - Sph sphingomyelin - PC phosphatidyl choline - PE phosphatidyl ethanolamine - PS phosphatidyl serine - PI phosphatidyl inositol - PAS periodic acid-Schiff reagent - CB Coomassie brilliant blue R-250 Dedicated to Professor Stuart Patton on the occasion of his 70th birthday.     

Tight coupling of the t-SNARE and calcium channel microdomains in adrenomedullary slices and not in cultured chromaffin cells

Regulated exocytosis involves calcium-dependent fusion of secretory vesicles with the plasma membrane with three SNARE proteins playing a central role: the vesicular synaptobrevin and the plasma membrane syntaxin1 and SNAP-25. Cultured bovine chromaffin cells possess defined plasma membrane microdomains that are specifically enriched in both syntaxin1 and SNAP-25. We now show that in both isolated cells and adrenal medulla slices these target SNARE (t-SNARE) patches quantitatively coincide with single vesicle secretory spots as detected by exposure of the intravesicular dopamine beta-hydroxylase onto the plasmalemma. During exocytosis, neither area nor density of the syntaxin1/SNAP-25 microdomains changes on the plasma membrane of both preparations confirming that preexisting clusters act as the sites for vesicle fusion. Our analysis reveals a high level of colocalization of L, N and P/Q type calcium channel clusters with SNAREs in adrenal slices; this close association is altered in individual cultured cells. Therefore, microdomains carrying syntaxin1/SNAP-25 and different types of calcium channels act as the sites for physiological granule fusion in "in situ" chromaffin cells. In the case of isolated cells, it is the t-SNAREs microdomains rather than calcium channels that define the sites of exocytosis.     

Prolactin-dependent expression of GD1 alpha ganglioside, as a component of milk fat globule, in the murine mammary glands

Lactation-associated expression of GD1 alpha ganglioside in murine mammary glands was found to be due to the increasing specific activities of Gg4Cer alpha2,3- and GM1b alpha2,6-sialyltransferases in the glands from 12th day of gestation. The gene for GM1b alpha2,6-sialyltransferase, mST6GalNAcV, which was not detected in nonpregnant glands, appeared at 12th day of gestation and increased in the following gestational and lactation periods. At 3rd day of lactation, the amounts of lipid-bound sialic acid (LSA) in the mammary glands and milk of HR-1 mice were 99.3 +/- 8.5 microg per gram of dried tissue and 2.9 microg per ml, GD1 alpha comprising 64.0% and 80.5% of the total LSA, respectively, and GD1 alpha in milk was found to be preferentially distributed in the fat globule fraction. When the mammary epithelial cells at 15th day of gestation were cultured in prolactin- and epidermal growth factor (EGF)-containing media, the synthesis of fat globules and casein, together with the enhanced synthesis of GD1 alpha, were observed in the cells in prolactin medium, indicating that synthesis of GD1 alpha occurs in association with milk production as a prolactin-dependent event. Thus, GD1 alpha ganglioside, which is characteristically distributed in the cerebellar Purkinje cells of the murine brain, is supplied to neonates through the milk of the mother.     

Syntaxin requirement for Ca2+-triggered exocytosis in neurons and endocrine cells demonstrated with an engineered neurotoxin

Botulinum neurotoxins cleave synaptic SNAREs and block exocytosis, demonstrating that these proteins function in neurosecretion. However, the function of the SNARE syntaxin remains less clear because no neurotoxin cleaves it selectively. Starting with a botulinum neurotoxin that cleaves both syntaxin and SNAP-25, we engineered a version that retains activity against syntaxin but spares SNAP-25. These mutants block synaptic release in neurons and norepinephrine release in neuroendocrine cells, thus establishing an essential role for syntaxin in Ca2+-triggered exocytosis. These mutants can generate syntaxin-free cells as a useful experimental system for research and may lead to pharmaceuticals that target syntaxin selectively.     

Expression of alpha-lactalbumin, alpha-S1-casein, and lactoferrin genes is heterogeneous in sheep and cattle mammary tissue.

We used 35S-labeled cRNA probes to localize the sites of alpha-lactalbumin, alpha-S1-casein, and lactoferrin mRNA synthesis in sheep and forcibly weaned cattle mammary tissue. Expression of alpha-lactalbumin was absent in three of four "virgin" glands studied, present in some alveoli of "pregnant" glands but not in others, despite a similar histological appearance. In the early lactating gland, expression was high in those alveoli with few fat globules in their cells and lumen and was absent in alveoli with abundant fat globules. These observations suggest either that alpha-lactalbumin gene expression is linked to the long-term secretory activity of cells and falls once cells are resting or regressing, or that there are cyclical variations in expression, or that in the lactating gland some groups of epithelial cells are synthesizing alpha-lactalbumin and some are synthesizing fat. Expression patterns of alpha-S1-casein were similar to those of alpha-lactalbumin. Lactoferrin, in contrast, was expressed almost exclusively in the "fatty alveoli" of both species. Our results show that dramatic variations in milk gene expression can occur throughout the mammary gland of sheep and cattle and that at no stage of pregnancy, lactation, or involution can the gland be considered metabolically homogeneous.     

Regulation and functional relevance of milk fat globules and their components in the mammary gland

Mammary gland and epithelial cells are unique to mammals and are under the control of lactogenic hormones such as prolactin. Recent findings indicated that major components of milk fat globule membrane (MFGM) are under the control of lactogenic hormones, and that the major components butyrophilin and xanthine oxidoreductase are indispensable for milk fat secretion. Further, prolactin signaling is negatively controlled by two highly related protein tyrosine phosphatases, PTP1B and TC-PTP. Milk fat globule EGF factor 8 (MFG-E8) is one of the major components of MFGM and is upregulated during lactation. MFG-E8 is further upregulated in the involuting mammary gland. MFG-E8 on exosome-like membrane vesicles in the milk recovered from post-weaning but not lactating mammary glands exhibits higher binding activity to phosphatidylserine and apoptotic mammary epithelial cells, and serves as a link between apoptotic mammary epithelial cells and phagocytes. Recent reports using MFG-E8 deficient mice support the view that MFG-E8 is indispensable for eliminating apoptotic mammary epithelial cells during involution.