key: cord-0005057-9wfco7o9
authors: Kirshner, Julia; Hardy, John; Wilczynski, Sharon; Shively, John E.
title: Cell–Cell Adhesion Molecule CEACAM1 is Expressed in Normal Breast and Milk and Associates with β1 Integrin in a 3D Model of Morphogenesis
date: 2004
journal: J Mol Histol
DOI: 10.1023/b:hijo.0000032360.01976.81
sha: 6c4b47fb804c7b7ffeccd621903d2b154b63c23b
doc_id: 5057
cord_uid: 9wfco7o9

CEA cell adhesion molecule-1 (CEACAM1) is a cell–cell adhesion molecule that, paradoxically, is expressed in an apical location in normal breast epithelium. Strong lumenal membrane staining is observed in 100% of normal glands (11/11), low in atypical hyperplasia (2/6), high in cribiform ductal carcinoma in situ (DCIS) (8/8), but low in other types of DCIS (2/15). Although most invasive ductal carcinomas express CEACAM1 (21/26), the staining pattern tends to be weak and cytoplasmic in tumours with minimal lumena formation (grades 2–3), while there is membrane staining in well-differentiated tumours (grade 1). The 'normal' breast epithelial line MCF10F forms acini with lumena in Matrigel with apical membrane expression of CEACAM1. MCF7 cells that do not express CEACAM1 and fail to form lumena in Matrigel, revert to a lumen forming phenotype when transfected with the CEACAM1-4S but not the -4L isoform. CEACAM1 directly associates with and down-regulates the expression of β1-integrin. Immuno-electron microscopy reveals numerous vesicles coated with CEACAM1 within the lumena, and as predicted by this finding, CEACAM1 is found in the lipid fraction of breast milk. Thus, CEACAM1 is a critical molecule in mammary morphogenesis and may play a role in the absorption of the lipid vesicles of milk in the infant intestinal tract.

CEA cell adhesion molecule-1 (CEACAM1) is a member of the carcinoembryonic antigen (CEA) gene family that was originally described in the bile ducts of liver as biliary glycoprotein (BGP) and shown to cross-react with anti-CEA antibodies (Svenberg 1976 ). Later, it was shown to be apically expressed in intestinal epithelia (Svenberg et al. 1979) . Studies in rat hepatocytes demonstrated that it was a cell-cell adhesion molecule and it was given the name Cell-CAM-105 (Ocklind & Obrink 1982) . The human gene was cloned by Hinoda et al. (1988) and Barnett et al. (1989) who showed that it was alternatively spliced into at least eight isoforms (Barnett et al. 1993) . The rat gene was originally cloned as a Ca 2+ /Mg 2+ ecto-ATPase (Lin & Guidotti 1989) , based on its ability to confer this activity to transfected COS cells, and it has been cloned as a bile transporter gene (Sippel et al. 1993) , but neither of these activities have been reproducibly assigned to the protein. CEACAM1 is a receptor for murine hepatitis virus in the mouse (Dvesksler et al. 1993) and for opa+ Nisseriae strains (Virji et al. 1996 , Gray-Owen et al. 1997 , Bos et al. 1998 ) and H. influenzae (Virji et al. 2000) in man. CEACAM1 associates with the insulin receptor in rat liver (Najjar et al. 1993) , is rapidly phosphorylated upon binding of insulin to its receptor (Najjar et al. 1995) , and upon mutation of Ser-503 in the cytoplasmic domain of CEACAM1, confers insulin resistance in transgenic mice (Poy et al. 2002a,b) . CEACAM1 is down-regulated early in >90% of colon cancers (Neumaier et al. 1993) , in prostate cancer (Kleinerman et al. 1995a) , and to a lesser extent, in breast cancer (Riethdorf et al. 1997 , Huang et al. 1998 . Transfection of the rat gene into prostate (Hsieh et al. 1995) , bladder (Kleinerman et al. 1996) , or breast cancer (Luo et al. 1997 ) cells renders the tumours less tumourigenic in nude mice.

The morphology of normal mammary gland is characterized by an inner layer of polarized epithelial cells with secretory surfaces facing a central lumen and basal surfaces surrounded by an outer layer of myoepithelial cells. When these cells are cultured ex vivo in Matrigel, a source of extracellular matrix (ECM), they form alveolar structures resembling their in vivo morphology (Gomm et al. 1997) . Other in vitro models that utilize the same Matrigel culture system have shown that normal mammary epithelial cells differentiate into acini or tubular structures (Barcellos-Hoff et al. 1989 , Petersen et al. 1992 , Bergstraesser & Weitzman 1993 , Gomm et al. 1997 . Upon contact with ECM proteins such as laminin and collagen IV, normal mammary epithelial cells become polarized, form spherical acini or branched tubules, and in the presence of prolactin initiate β-casein secretion (Muschler et al. 1999) . In contrast to normal cells, tumour mammary epithelial cells form large, non-polarized, undifferentiated colonies without lumena when grown in Matrigel (Petersen et al. 1992) .

A variety of growth factors such as hepatocyte growth factor (Brinkmann et al. 1995 , Soriano et al. 1995 , epidermal growth factor (Gomm et al. 1997) , keratinocyte growth factor (Hirai et al. 1998) , and fibroblast growth factor (Li & Shipley 1991) promote branching morphogenesis of cells grown on ECM. In addition to growth factors, integrins play an important role in establishing aveolar differentiation while aberrant integrin expression confers a tumourigenic phenotype. Integrins α1 and α2 confer invasive behaviour by regulating stromelysin-1 (matrix metalloproteinase) expression, and α6-integrin has been shown to regulate cell motility . Integrins α6 and β1 cause metastatic invasion in mouse mammary carcinoma cells and are required for cell movement in 2-dimensional (2D) cell assays on reconstituted basement membranes. In agreement with these findings, antibodies to α1, α2, α6, and β1-integrins were able to block cell motility . Moreover, blocking antibodies which inhibit expression of β1-integrin or EGFR in tumour cells grown in Matrigel have been shown to induce down-regulation of β1 and EGFR, and cause growth arrest and reversion to normal morphogenesis as indicated by the formation of acinus-like structures (Li et al. 1997 , Weaver et al. 1997 .

Assembled lipid droplets in the mammary gland are secreted from the cytoplasm and enveloped by cellular membranes. Milk lipids originate as small droplets of triacylglycerol, synthesized in or on the surface of rough endoplasmic reticulum membrane and are released into the cytoplasm with a surface coat of protein and polar lipid. Droplets of varying size are transported to the apical cytoplasm, and when secreted from the cell, are coated with an outer membrane bilayer. Thus, the cream fraction of milk comprises lipid droplets coated with cellular membranes. Two possible mechanisms for lipid secretion have been proposed: an apical mechanism, in which lipid droplets are enveloped with apical plasma membrane, and a secretoryvesicle mechanism, in which fat droplets are surrounded by secretory vesicles in the cytoplasm and are released from the surface by exocytosis from intracytoplasmic vacuoles (Mather & Keenan 1998 , Mather et al. 2001 , Oshima et al. 2002 . A number of proteins are found on the surface of secreted milk-fat globules including mucin 1 (Mather et al. 2001 ) and milk fat globule-EGF factor 8, a peripheral membrane glycoprotein abundantly expressed in lactating mammary glands and thought to play a role in vesicle secretion (Oshima et al. 2002) . The presence of sialyl-Lewis ligands on milk oligosaccharides and their known role as selectin ligands suggests that they could either play a role in inflammatory processes (Rudloff et al. 2002) or milk uptake in the infant intestinal tract. Interestingly, mastitis milk has the same anti-inflammatory components and characteristics of normal milk, with elevations in selected components/activities that may help protect the nursing infant from developing infections due to feeding on mastitis milk (Buescher 2001) . High levels of cell adhesion molecules ICAM-1 and VCAM-1 are also found in milk, which could further augment the immune system of the neonate (Xyni et al. 2000) . Moreover, multimeric α-lactalbumin, a potent Ca 2+ -elevating and apoptosis-inducing agent with broad, yet selective, cytotoxic activity was found in milk. Multimeric α-lactalbumin killed all transformed, embryonic, and lymphoid cells tested, but spared mature epithelial elements. These findings raise the possibility that milk contributes to mucosal immunity not only by furnishing antimicrobial molecules but also by activating lymphocytes and epithelium (Hakansson et al. 1995) .

We are currently studying the role of CEACAM1 in normal and malignant breast. Previously, we demonstrated that human CEACAM1 is expressed in the normal breast cell line MCF10 and is essential for lumen formation when the cells are cultured in Matrigel (Huang et al. 1999) . Blocking CEACAM1 expression either with an anti-sense gene or anti-CEACAM1 antibodies reverts the cells to a more malignant phenotype. Conversely, MCF7 breast cancer cells fail to express CEACAM1 or produce lumena when grown in Matrigel. When MCF7 breast cancer cells were transfected with CEACAM1-4S, the short cytoplasmic domain isoform that predominates in normal breast (Huang et al. 1998) , reversion to a more normal phenotype was evidenced by lumen formation when the cells were grown in Matrigel (Kirshner et al. 2003) . In the study presented here, we examine the expression of CEACAM1 in a wide variety of cancer cell lines grown in Matrigel, in normal, pre-malignant, and malignant breast tissue, and for the first time, show that CEACAM1 is secreted in breast milk on the membrane of lipid vesicles and in alveolar cells.

With institutional review board approval, tissues were obtained from the anatomic pathology file at the City of Hope National Medical Center. Normal mammary epithelial cells (HMEC, MCF10A, MCF10F), mammary carcinoma cells (MCF7, MB-MDA-231, MB-MDA-468, ZR-75, T47D, SKBR3, BT474), cervical carcinoma (HeLa), and colon carcinoma (HT29) cells were obtained from ATCC. MCF7-CEACAM1-4S-ectoGFP cells were generously provided by Dr. Charng-Jui Chen and were made as described previously (Kirshner et al. 2003) .

MCF10Fe cells were grown in mammary epithelial cell growth medium (MEGM) supplemented with SingleQuotes (10 ng/ml hEGF, 5 µg/ml insulin, 0.5 µg/ml hydrocortisone, 50 µg/ml gentamicin, 50 ng/ml amphotericin-B, and 52 µg/ml bovine pituitary extract) (Clonetics). MCF10Fm cells were grown in DMEM-F12 (Gibco-BRL) with 5 µg/ml insulin, 5 µg/ml transferrin, 20 ng/ml hEGF, 5 × 10 −8 M dexamethasone and 5% FBS (Huang et al. 1999) . MCF7 and HeLa cells were grown in MEM with Earle's salts and 10% FBS (Irvine Scientific). MCF7, HeLa, HT29 cells were grown in RPMI 1640 (Irvine Scientific) and supplemented with 10% FBS (Irvine Scientific), l-glutamine (GibcoBRL), and penicillin/streptomycin (GibcoBRL). MB-MDA-231, MB-MDA-468, ZR-75, T47D, SKBR3, HBL-100, and BT474 were grown per ATCC instructions.

Cells (2.5 × 10 5 ) were plated in a thick layer (1 mm) of Matrigel (Collaborative Biomedical Products) in two well chamber slides (Nunc). Solidified Matrigel was overlaid with MEGM. To harvest cells, the gels were incubated in MatriSperse solution (Collaborative Biomedical Products) for 1 h at 4 • C. Cells grown in Matrigel were scored for the presence (acinus formation) or absence of lumena and compared by Student's t-test or the chi-squared test.

GEArray Q Series Human Apoptosis and Human ECM and Adhesion Molecules gene arrays were purchased from SuperArray. Hybridization with the RNA from MCF/vector and MCF7/CEACAM1-4S cells grown on tissue culture plastic or in Matrigel for 4 days was performed according to the manufacturer's instructions. For cDNA positions and identification lists, see http://www.superarray.com.

For the detection of cell surface integrins 1 × 10 6 cells were incubated with 1 µg/ml of anti-integrin mAbs (Chemicon) for 1 h, washed with phosphate buffered saline (PBS), and then incubated with Alexa 488 conjugated goat anti-mouse IgG (Molecular Probes) for 1 h followed by washes with PBS. Cells were analysed on a MoFlo (Cytomation) FACS.

Matrigel cultures were fixed with 10% neutral buffered formalin (NBF) (10% formalin in PBS) for 10 min, embedded in 3% agar, and treated with 10% NBF overnight, followed by 70% ethanol until processed. Paraffin-embedded sections were prepared and used for haematoxylin and eosin staining and immunohistochemistry. Monoclonal antibody 4D1C2 (1 µg/ml) (Drzeniek et al. 1991 ) was used to detect CEACAM1. Immunohistochemical staining was performed as described by Huang et al. (1999) .

Cells were grown in Matrigel and stained with MitoSensor JC-1 cationic dye (Clontech), annexin V-FITC, or antiintegrin mAbs (Chemicon) according to manufacturer's instructions. Confocal microscopy was performed on a Zeiss Model 310 confocal microscope.

Transmission electron microscopy (TEM) was performed according to (Huang et al. 1999) . In short, MCF10F, MCF7/pHβ, and MCF7/CEACAM1-4S cells were grown in Matrigel for 5-12 days or isolated from human breast milk by centrifugation at 2000 rpm for 10 min. Acini were isolated from Matrigel as previously described, fixed for 3 h with 1% glutaraldehyde in 0.2 M cacodylate buffer, washed 3× in PBS, and post-fixed with 1% osmium tetroxide for 1 h at 4 • C. Following three PBS washes, the cells were placed in microcentrifuge tubes for graded ethanol rinses. The samples were treated with propylene oxide, embedded in Eponate with araldite, and polymerized for 72 h at 70 • C. Sections (50 nmthick) were cut on an LKB Ultrotome III, placed on 300 mesh, uncoated copper grids, and stained with 5% uranyl acetate for 10 min, followed by 1 min in Sato's lead solution. TEM was performed on a Philips CM 10 TEM.

MCF10Fe and MCF7/CEACAM1-4S cells were grown in Matrigel for 5-12 days, resulting acini were isolated with MatriSperse, fixed for 2 h with 4% paraformaldehyde-0.01% glutaraldehyde in 0.2 M cacodylate buffer, pH 7.2, and washed with cacodylate buffer. Following cryoprotection with 2.3 M sucrose overnight, the specimen was frozen by immersion into liquefied nitrogen and sectioned at 70 nm on a Leica Ultracut UCT equipped with EMFCS cryoattachment unit at −100 • C. The sections were picked up and transferred to formvar coated, carbon coated, glow discharged 300 mesh nickel grids. The grids were placed on drops of sucrose, washed with 3× on 50 µl drops of PBS-6% BSA for 5 min each and incubated for 45 min at room temperature on 30 µl drops of PBS-6% BSA (negative control) or mAb T84.1 (1 µg/ml) diluted with PBS-6% BSA. Subsequently, the grids were washed 3× on 50 µl drops of PBS-6% BSA for 2 min each and then incubated for 1 h on 50 µl drops of 1 : 100 Protein G gold (20 nm) diluted with PBS-6% BSA. After washing with PBS-6% BSA 5× for 2 min each, grids were incubated on a drop of PBS-1% glutaraldehyde for 5 min, washed 5× on nano-pure water, stained on a drop of methylcellulose-uranyl acetate for 30 s, dried with filter paper, and observed and photographed with the Philips CM 10 TEM.

Fractionated or whole human milk samples were applied to reducing gradient SDS-PAGE gels (Invitrogen) in SDS loading buffer with 100 mM DTT. Following semidry electrotransfer to PVDF membranes (BioRad) and blocking with TBS (100 mM Tris-HCl, pH 7.5, 150 mM NaCl) buffer containing 5% non-fat dry milk or 10% BSA (for phosphoprotein blotting) overnight, the blots were incubated for 1 h with mAb T84 or pAb 22.9 at a final concentration of 1 µg/ml, Following three TBS washes, the blots were incubated with goat anti-mouse IgG-HRP or goat anti-rabbit IgG-HRP at a 1 : 3000 dilution (BioRad) for 1 h and washed again. Detection was carried out using the SuperSignal Chemiluminescent Substrate and Enhancer Kit (Pierce).

Mammary alveolar cells were isolated from milk by centrifugation and RNA was isolated using Totally RNA kit (Ambion) as per the manufacturer's instructions. Reverse transcriptase-polymerase chain reaction (RT-PCR) was performed with sense and anti-sense primers P1 and P2 as described previously (Takahashi et al. 1993 ) using One Step RT-PCR kit (QIAGEN) per manufacturer's instructions.

CEACAM1 was found to be down-regulated in 30% of breast cancers (Riethdorf et al. 1997 , Huang et al. 1998 ). Riethdorf et al. (1997) assessed the pattern of CEACAM1 expression in ductal and lobular carcinomas, and found predominantly membrane staining in the less differentiated, and occasional lumenal staining in well-differentiated specimens. We have now compared normal, pre-malignant, and invasive breast cancers for their patterns of CEACAM1 expression in more detail (Table 1 and Figures 1 and 2) . Strong lumenal staining was observed in 11/11 normal glands examined ( Figure 1A) . Similarly, strong lumenal staining was observed for proliferative breast disease (13/15, Figure 1B ). Most atypical ductal hyperplasia specimens were negative (2/6, Figure 1C and D). All cases of ductal carcinoma in situ (DCIS) with a cribriform pattern were positive (Figure 2A ), but those with a solid, comedo, and micropapillary patterns were negative or had only weak cytoplasmic staining (3/15, Figure 2B ). Although most invasive ductal carcinomas showed expression of CEACAM1 (21/26) the strength and pattern varied with morphologic features of the tumours. Those tumours with well-formed tubules (Grade 1 and a few Grade 2 tumours) showed strong lumenal membrane expression ( Figure 2C ). Tumours growing as sheets and cords with little if any tubule formation (most Grade 2 and all Grade 3 tumours) had predominately weak to moderate cytoplasmic staining ( Figure 2D ) with only occasional membrane positivity. A rare high-grade tumour had strong cytoplasmic expression ( Figure 2E ). Invasive lobular carcinomas were negative ( Figure 2F ) with only one of four showing weak staining of signet ring cells. These results suggest that in ductal breast cancers, the altered pattern of cellular expression of CEACAM1 reflects the tubular-differentiation state. Lower grade tumours retain the ability to localize the protein to the apical surface while in the more poorly differentiated ductal and in lobular carcinomas the protein is cytoplasmic.

Previously, we examined the pattern of CEACAM1 isoform expression in normal and malignant breast and found that patients either produced the CEACAM1-3S isoform only, or a combination of the -4L, -4S, and -3S isoforms (Huang et al. 1999) . While the reason for these differences in isoform production are not clear, we noted that the -4S or -3S isoforms predominate. Later, we showed that transfection of the -4S but not the -4L isoform into MCF7 cells was sufficient to restore lumen formation in these cells when grown in Matrigel (Kirshner et al. 2003) . Thus, the -S isoform appears to be the biologically relevant isoform in the human mammary gland. In the majority of breast cancer patients examined, the combined isotype expression pattern (-4L, -4S, -3S) was observed (Huang et al. 1999) . It remains to be seen if the isoform expression pattern is changed during malignancy. 

We examined the pattern of CEACAM1 expression in a number of mammary epithelial cell lines grown in Matrigel to determine if CEACAM1 expression confers lumen formation across a wide range of cells. Both mammary epithelial cells obtained from breast reduction surgery (HMECs) and MCF10 cells, that are considered to mimic normal mammary epithelial cells, expressed CEACAM1 and formed acini with lumena, when grown in Matrigel (Figure 3 , Table 2 ). The pattern of expression was identical to that of normal mammary gland, namely, strong membranous staining surrounding the lumena. In the case of the breast cancer cell lines, four out of seven formed acini in Matrigel, and none of those had lumena. However, two of the four that form acini expressed CEACAM1, demonstrating that expression of CEACAM1 alone is not sufficient to confer lumen formation. We conclude that CEACAM1 expression is necessary but not sufficient for lumen formation. In the breast tumour cell lines that expressed CEACAM1, the sub-cellular distribution was similar to that of the mammary tumours, where the staining was distributed throughout the cytoplasm and not found exclusively at the membrane. Since CEACAM1 is a type 1 membrane glycoprotein, we speculate that it remains trapped in endosomes in these cells, and is prevented from transmitting its cell-cell signals that are crucial for lumen formation (Kirshner et al. 2003) .

We also tested the expression of CEACAM1 in other types of cancer cell lines grown in Matrigel and its potential effect on lumen formation. All cells tested formed colonies in Matrigel, but none formed lumena (Table 2) . However, only the colon cancer cell line HT29 expressed CEACAM1, thus making it hard to judge if these other cell lines behave like the breast cancer lines. We were especially interested in the prostate cell line, since down-regulation of CEACAM1 expression has been correlated with the progression from normal to malignant (Kleinerman et al. 1995a) , and normal prostate cells can form acini with lumena when grown in a 3D culture (Hudson et al. 2000) . Studies are now underway to determine if CEACAM1 transfected LNCap cells form lumena when cultured in Matrigel.

Previously, we showed that transfection of CEACAM1 into MCF7 cells restores lumen formation when the cells are grown in Matrigel and lumen formation requires apoptosis of the central cells in the colonies (Kirshner et al. 2003) . In those studies, we hypothesized that expression of CEACAM1 triggers apoptosis. For example, when MCF7 cells transfected with a CEACAM1-GFP fusion protein are grown in Matrigel and stained with the mitochondrial dye, JC1, CEACAM1-GFP is seen between the cells at 7 d and in the lumena at 12 d ( Figure 4A and B). Since the JC-1 dye stains the mitochondria of healthy cells red and apoptotic cells green, one cannot distinguish whether the central green cells are apoptotic, just express CEACAM1-GFP or both. However, when MCF7 cells are transfected with CEACAM1 (no GFP fusion protein) and stained with JC1, the central cells stain green ( Figure 4C ). Further evidence that the central cells are apoptic is shown by annexin V staining superimposed on the phase contrast photo of an isolated 12 d colony ( Figure 4D ). Based on these published results, we ask why the central cells apoptose while the peripheral cells do not? One answer consistent with a large body of data on the function of integrins and ECM is that only the peripheral cells contact ECM and thus are spared. The literature suggests that loss of contact of epithelial cells to ECM is sufficient to cause death or anoikis; however, control cells, that do not express CEACAM1 and are located in the centre of colonies, are perfectly healthy ( Figure 3B ). Thus, a major function of CEACAM1 appears to be induction of apoptosis in the context of the presence or absence of ECM. The mechanism of interaction between integrin subunits and CEACAM1 is addressed in a later section. The lumenal epithelium of the mammary gland is polarized with microvilli and budding vesicles facing the apical surface of the gland (Gomm et al. 1997) . The polarized pattern with microvilli and budding vesicles is faithfully reproduced when normal mammary epithelial cells are cultured in Matrigel (Huang et al. 1999) . Similarly, when MCF7 cells transfected with CEACAM1 are grown in Matrigel and the lumen is examined by EM, microvilli with budding vesicles are observed ( Figure 5A ). Since Frangsmyr et al. (1995) had previously demonstrated CEACAM1 expression on the tips of microvilli in colonic epithelia, we speculated that CEACAM1 would also be localized to the microvilli of polarized breast epithelial cells. Applying immuno-electron microscopy (immuno-EM) staining to MCF/CEACAM1 cells grown in Matrigel during the course of lumen formation, CEACAM1 was found between the cells of the 5-d acinus ( Figure 5B) , and redistributed to the lumenal surface by 12 d in culture, where it was found exclusively on microvilli ( Figure 5C ). Furthermore, vesicles staining positive for CEACAM1 were found both in the cytoplasm of the epithelial cells and in the central lumena ( Figure 5B -D, and see Figure 6D ). No CEACAM1 staining was found in the MCF7/vector cells grown in Matrigel ( Figure 5E ). Thus, we conclude that CEACAM1 plays a role in cell-cell adhesion in the early stages of mammary morphogenesis, and shifts to a purely apical pattern of expression once mature glands are formed. In mature alveolar cells CEACAM1 expression is limited to the microvilli that appear to produce small, micron sized vesicles that have abundant amounts of CEACAM1 on their surfaces. Based on this evidence, we predicted that CEACAM1 should be found in the lipid fraction of breast milk.

Breast milk was centrifuged to remove cells and the supernatant separated into the lipid and non-lipid fraction. Breast milk is known to contain abundant alveolar cells that are shed into the milk (Cross & Mercer 1993) . Staining of these cells revealed an intact nucleus and one or more membrane enveloped lipid vesicles ( Figure 6A-D) . The lipid vesicle found in these cells was often as large as the nucleus and was filled with triglycerides. RT-PCR analysis of the 

IH -immunohistochemistry, IB -immunoblotting. n -normal, t -tumour; m -membrane, c -cytoplasmic staining pattern. 1 50% make colonies. 2 Myoepithelial phenotype. 3 Loose colonies. mRNA isolated from the alveolar cells revealed that both the CEACAM1-4L and -4S isoforms were present ( Figure 7A and B). Western blot analysis of the lipid and non-lipid fractions of milk from a second donor revealed that CEACAM1 was found only in the lipid fraction ( Figure 7C ), in agreement with the prediction that it is on the surface of lipid vesicles. We concluded that only the CEACAM1-4 isoform was present in this specimen based on a molecular size comparison to CEACAM1 isoforms expressed in Jurkat cells. However, it was also possible that CEACAM1 was cleaved from its transmembrane and cytoplasmic domains leaving a soluble form somehow associated with the lipid vesicles. To test this possibility, soluble forms of CEACAM1-3 and -4 were expressed in NSO cells, purified, and compared to the CEA-CAM1 found in milk. It can be seen that the molecular size of CEACAM1 from milk aligns with CEACAM1-4S and not with CEACAM1-4L or either of the soluble isoforms ( Figure 7C ). The identity of the isoform was further confirmed by immunoblotting with an antibody that recognizes only the long cytoplasmic domain of CEACAM1, where the lack of signal indicates the absence of the long cytoplasmic domain ( Figure 7C ). The lack of CEACAM1-4L expression in the milk specimen and its presence in the alveolar cells could be due to variations among individual donors. Indeed, in our previous study we noted a short only versus a short plus long isoform polymorphism in the breast tissue specimens (Huang et al. 1998) . Further studies are required to determine if this is due to a genetic polymorphism.

The function of the CEACAM1 coated lipid vesicles is worth a speculation. It is known that the membrane of enveloped lipid vesicles in milk acts as a site of attachment to the newborn intestine (Cross & Mercer 1993) . A simple mechanism of attachment of these vesicles to the newborn epithelium could involve a homophilic cell-cell adhesion molecule like CEACAM1. Since CEACAM1 is expressed on both the lipid vesicles and the epithelium of the gut, it is in position to play this role. Furthermore, bacterial cells which bind to CEACAM1 on epithelial cells are rapidly internalized (Chen et al. 2001 , Muenzner et al. 2001 , suggesting that CEACAM1 is an internalizing receptor. Further studies are required to establish if our hypothesis is correct.

Previous studies on the 3D model of mammary morphogensis have focussed on the role of ECM-integrin interactions (for a review see Bissell et al. 2002) . It seems likely that the integration of these signals with the cell-cell adhesion signals conveyed by molecules such as CEACAM1 (or E-cadherin) are essential for establishing the complete morphogenic programme. An obvious possibility is a direct association between integrins and CEACAM1 receptors on the cell surface. Indeed, Brummer et al. (2001) have shown that when the phosphorylated form of the long cytoplasmic domain of CEACAM1 is used as bait for interacting proteins in neutrophil lysates, β3 integrin was isolated. However, extension of this finding to breast epithelial cells, to other integrin subunits, and especially to the short cytoplasmic domain isoform of CEACAM1 is lacking. As a first step in assessing the level of CEACAM1 involvement in integrin-ECM communication, a gene array analysis was performed comparing the expression of cell adhesion and ECM proteins in MCF7/vector and MCF7/CEACAM1-4S cells grown on plastic versus Matrigel (Table 3 ). After normalization for housekeeping genes, no gene regulation (up or down) was seen in CEACAM1 versus vector transfected cells grown on either plastic or Matrigel. Aside from several housekeeping genes, the gene array showed that several genes like CEA and E-cadherin are highly expressed in MCF7 cells. However, the mRNA levels for these genes were unchanged before and after growth of control and transfected cells on Matrigel. Based on the limited selection of genes probed, it appears that CEACAM1 alone is responsible for reverting breast cancer cells to a normal phenotype.

Since the role of integrins in a 3D model of mammary morphogenesis is well established, including changes in the surface levels of the β1 and β4 subunits (Weaver et al. 1997) , we speculated that either the protein or its cell surface levels must be regulated in CEACAM1 transfected MCF7 cells. To examine this possibility, cell surface levels of various integrins was measured on control and CEACAM1 transfected cells grown on plastic or in Matrigel ( Figure 8A ). Quite surprisingly, no or very low cell surface expression of integrins was seen in either control or CEACAM1 transfected cells grown on plastic ( Figure 8A , blue and purple traces); however, when the cells were grown in Matrigel, the cell surface levels of all the integrins tested increased dramatically ( Figure 8A , yellow and red traces). Since mRNA levels of the integrins are unchanged, we conclude that exposure of the cells to the ECM in Matrigel induces their cell surface expression. Comparison of the control versus transfected cells grown in Matrigel revealed that integrins α1 and β1 were down-regulated in MCF7/CEACAM1-4S compared to the vector controls (yellow vs. red traces in Figure 8A ), while all other subunits showed no changes. Weaver et al. (1997) had previously shown that the down-regulation of β1 and up-regulation of β4 integrin correlates with the induction of lumen formation in T4-2 cells arising from spontaneously transformed HMT-3522 cells, and that β1 function blocking antibodies restore lumen formation to the transformed T4-2 cells grown in Matrigel. While the corresponding α subunit was not identified in their studies, it is interesting to speculate that it may be α1, since it is down-regulated together with β1 in our study. Integrin β1 has been shown to be critical for murine mammary morphogenesis (Klinowska et al. 1999) where implanted blocking antibodies prevented mammary gland development. Since down-regulation of β1 correlates with restoration of lumen formation in mammary tumour cells grown in Matrigel, we speculate that the absolute levels of β1 are the critical factor, especially in terms of competition with other β subunits for paired α subunits. In this respect, α1β1 binds laminin-1 and α2β1 binds collagen IV, the major components of basement membrane that surround the alveolar structures.

Confocal microscopy was performed to evaluate the possible direct association of CEACAM1 with the above Figure 8 . Effects of ECM on integrin expression in MCF/CEACAM1-4S cells. (A) MCF7/vector and MCF7/CEACAM1-4S cells were grown on plastic or in Matrigel (2 d) and cell surface expression of integrins was assessed using FACS analysis. On the histograms, black peaks correspond to unstained MCF7 cells used as a negative control, and blue to MCF7/vector cells grown on plastic and yellow, in Matrigel. The purple peaks correspond to MCF7/CEACAM1-4S cells grown on plastic and red, in Matrigel. The yellow peaks (indicated by arrows) are under the red peaks for α2 and α3. (B) MCF7/CEACAM1-4S-ectoGFP cells were isolated from a 2 day Matrigel culture and stained with anti-integrin antibodies (red channel). Colocalization is indicated by the overlap of the two channels (yellow) on the periphery of the colonies where contact with ECM is established. integrin subunits ( Figure 8B) . A strong co-localization of CEACAM1 with α2 and β1, and a weaker association with α1, and α3 was observed. The co-localization occurred on the periphery of the colonies where contact with the basement membrane occurs. Even though the predominant location of CEACAM1 is between the cells during the early stages of morphogenesis, clearly it also has a basal-lateral localization that allows direct interaction with the integrin subunits and ECM. This suggests that CEACAM1-4S and β1 integrin participate in a common pathway leading to lumen formation. These findings further support our hypothesis that CEACAM1-4S communicates with the ECM during the initial stages of mammary gland development.

CEACAM1 is constitutively expressed in the epithelium of a variety of normal tissues, including colon, bladder, prostate, and breast. The early down-regulation of CEA-CAM1 in >90% of colon cancers (Neumaier et al. 1993) and its ability to render prostate, bladder, and breast cancer cells less tumourigenic (Kleinerman et al. 1995b , Luo et al. 1997 have led some investigators to propose that CEACAM1 is a tumour suppressor gene. However, the gene is only down-regulated in about 30% of breast cancers (Riethdorf et al. 1997 , Huang et al. 1998 , casting some doubt on this assignment. This discrepancy prompted us to re-examine the pre-malignant and malignant breast tissues for further clues. Strong CEACAM1 expression on the lumenal surfaces of hyperplastic ductal cells in benign proliferative breast disease is similar to that seen in the normal breast. Minimal expression was seen in the limited number of atypical hyperplasia cases examined. Cribriform DCIS, which retains lumenal structures stain strongly for CEACAM1, while those types that do not (solid and comedo types of DCIS) showed only cytoplasmic staining. Most invasive ductal carcinomas (80%) were positive and only those that had tubular formation expressed the protein on lumenal or membrane surfaces, otherwise cytoplasmic staining predominated. We conclude that, while a drop in CEACAM1 expression and alteration of its staining pattern correlates with the differentiation state of both the pre-malignant and invasive lesions, by classical definition, CEACAM1 is not a tumour suppressor gene. It is especially noteworthy that no genetic changes involving the CEACAM1 gene have been reported.

The finding of CEACAM1 in breast milk was a direct result from our ultrastructure studies on breast epithelial cells grown in Matrigel. These polarized cells produce abundant vesicles that are secreted from the tips of microvilli into alveolar lumena. As predicted by these results, the CEACAM1 positive vesicles are found in the lipid fraction of milk and the molecular size of CEACAM1 is identical to that found in intact cells. Others have speculated that lipid droplets in milk are surrounded by membranes to increase their absorption in the infant intestinal tract (Cross & Mercer 1993) . If this is correct, CEACAM1 is an ideal molecule for this purpose, since it is found on intestinal epithelia, is a cellcell adhesion molecule, and is know to cause internalization upon ligation with bacteria (Chen et al. 2001 , Muenzner et al. 2001 . Further studies involving uptake of milk lipid vesicles by cultured intestinal epithelial cells are planned.

The finding that transfection of CEACAM1 into CEACAM1 negative breast cancer cells restores lumen formation in Matrigel culture is exciting. It appears that CEACAM1 plays an essential role in the morphogenic process. However, we found several breast cancer cell lines that express CEACAM1, but fail to form a lumen when cultured in Matrigel. This means that CEACAM1 is essential but not sufficient for lumen formation. It is possible that these cells are either lacking another essential component of the programme or express an additional gene that overrides the programme. In the first case, other candidate essential genes may include the integrins which bind ECM and help to establish polarity. While this possibility has not been fully explored, we have shown that CEACAM1 associates with the β1 integrin subunit that previously has been shown to be essential for murine mammary gland development (Klinowska et al. 1999) . CEACAM1 also lowers the cell surface expression level of the β1 integrin subunit in transfected MCF7 cells, a finding that correlates well with studies on reversion of malignant breast cancer cells with anti-β1 integrin antibodies (Weaver et al. 1997) . A possible explanation of these apparently contradictory results is that the β1 subunit may compete for other essential α subunits. Since α1 was also down-regulated in our study, we believe that there may be a competition among α1β1 and α2β1 which bind laminin 1 and collagen IV, respectively, both components of the basement membrane that surrounds the alveolar cells and establishes polarity. The other possibility to be explored is that additional genes expressed suppress the morphogenic program. A likely candidate gene is Her2 which overrides the growth inhibitory signalling in breast cells. In support of this idea, both CEACAM1 and Her2 are expressed in BT474 cells (Table 2) , and studies with Her2 transfected cells demonstrate continued growth disrupting lumen formation (Muthuswamy et al. 2001) .

In summary, CEACAM1 expression, found in an apical location in normal breast glands, is altered in both pre-malignant and malignant disease. It is secreted into milk on vesicles, and while its function there is unknown, it may play a role in intestinal absorption of lipid. Using a 3D model of morphogenesis, we have shown that CEACAM1 is required for lumen formation in CEACAM1 positive MCF10F cells, and transfection of CEACAM1 into the MCF7 breast cancer cell line restores lumen formation. The mechanism of lumen formation involves apoptosis of the central cells and requires both ECM-integrin and CEACAM1 interactions.

Functional differentiation and alveolar morphogenesis of primary mammary cultures on reconstituted basement membrane

Human biliary glycoprotein gene: Characterization of a family of novel alternatively spliced RNAs and their expressed proteins

Carcinoembryonic antigens: Alternative splicing accounts for the multiple mRNAs that code for novel members of the carcinoembryonic antigen family

Culture of normal and malignant primary human mammary epithelial cells in a physiological manner simulates in vivo growth patterns and allows discrimination of cell type

The organizing principle: Microenvironmental influences in the normal and malignant breast

CD66 receptor specificity exhibited by neisserial Opa variants is controlled by protein determinants in CD66 N-domains

Hepatocyte growth factor/scatter factor induces a variety of tissuespecific morphogenic programs in epithelial cells

Cis interaction of the cell adhesion molecule CEACAM1 with integrin beta(3)

Anti-inflammatory characteristics of human milk: How, where, why?

The CGM1a (CEACAM3/CD66d)-mediated phagocytic pathway of Neisseria gonorrhoeae expressing opacity proteins is also the pathway to cell death

Cell and Tissue Ultrastructure

Identification of membrane antigens in granulocytes and colonic carcinoma cells by a monoclonal antibody specific for biliary glycoprotein, a member of the carcinoembryonic antigen family

Several members of the mouse carcinoembryonic antigen-related glycoprotein family are functional receptors for the coronavirus mouse hepatitis virus A-59

Cell and region specific expression of biliary glycoprotein and its messenger RNA in normal human colonic mucosa

Separated human breast epithelial and myoepithelial cells have different growth factor requirements in vitro but can reconstitute normal breast lobuloalveolar structure

Differential Opa specificities for CD66 receptors influence tissue interactions and cellular response to Neisseria gonorrhoeae

Apoptosis induced by a human milk protein

Molecular cloning of a cDNA coding biliary glycoprotein I: Primary structure of a glycoprotein immunologically crossreactive with carcinoembryonic antigen

Epimorphin functions as a key morphoregulator for mammary epithelial cells

Tumor suppressive role of an androgen-regulated epithelial cell adhesion molecule (C-CAM) in prostate carcinoma cell revealed by sense and anti-sense approaches

Essential role of biliary glycoprotein (CD66a) in morphogenesis of the human mammary epithelial cell line MCF10F

Expression of biliary glycoprotein (CD66a) in normal and malignant breast epithelial cells

Proliferative heterogeneity in the human prostate: Evidence for epithelial stem cells

CEACAM1-1S, a cell-cell adhesion molecule, mediates apoptosis and reverts mammary carcinoma cells to a normal morphogenic phenotype in a 3D culture

Suppression of human bladder cancer growth by increased expression of C-CAM1 gene in an orthotopic model

Consistent expression of an epithelial cell adhesion molecule (C-CAM) during prostate development and loss of expression in prostate cancer: Implication as a tumor suppressor

Application of a tumor suppressor (C-CAM1)-expressing recombinant adenovirus in androgen-independent human prostate cancer therapy: A preclinical study

Laminin and beta1 integrins are crucial for normal mammary gland development in the mouse

Mammary-derived signals activate programmed cell death during the first stage of mammary gland involution

Expression of multiple species of basic fibroblast growth factor mRNA and protein in normal and tumorderived mammary epithelial cells in culture

Cloning and expression of a cDNA coding for a rat liver plasma membrane ecto-ATPase

Alpha1 and alpha2 integrins mediate invasive activity of mouse mammary carcinoma cells through regulation of stromelysin-1 expression

Suppression of tumorigenicity of breast cancer cells by an epithelial cell adhesion molecule (C-CAM1): The adhesion and growth suppression are mediated by different domains

The distribution of MUC1, an apical membrane glycoprotein, in mammary epithelial cells at the resolution of the electron microscope: Implications for the mechanism of milk secretion

Origin and secretion of milk lipids

Pathogenic Neisseria trigger expression of their carcinoembryonic antigen-related cellular adhesion molecule 1 (CEACAM1; previously CD66a) receptor on primary endothelial cells by activating the immediate early response transcription factor, nuclear factor-kappaB

Division of labor among the alpha6beta4 integrin, beta1 integrins, and an E3 laminin receptor to signal morphogenesis and betacasein expression in mammary epithelial cells

ErbB2, but not ErbB1, reinitiates proliferation and induces luminal repopulation in epithelial acini

) pp120/ecto-ATPase, an endogenase substrate of the insulin receptor tyrosine kinase, is expressed as two variably spliced isoforms

Insulin-stimulated phosphorylation of recombinant pp120/HA4, an endogenous substrate of the insulin receptor tyrosine kinase

Biliary glycoprotein, a potential human cell adhesion molecule, is downregulated in colorectal carcinomas

Intercellular adhesion of rat hepatocytes. Identification of a cell surface glycoprotein involved in the initial adhesion process

Secretion of a peripheral membrane protein, MFG-E8, as a complex with membrane vesicles

Interaction with basement membrane serves to rapidly distinguish growth and differentiation pattern of normal and malignant human breast epithelial cells

Shc and CEACAM1 interact to regulate the mitogenic action of insulin

CEACAM1 regulates insulin clearance in liver

Differential expression of CD66a (BGP), a cell adhesion molecule of the carcinoembryonic antigen family, in benign, premalignant, and malignant lesions of the human mammary gland

Detection of ligands for selectins in the oligosaccharide fraction of human milk

The rat liver ecto-ATPase is also a canalicular bile acid transport protein

Hepatocyte growth factor stimulates extensive development of branching duct-like structures by cloned mammary gland epithelial cells

Carcinoembryonic antigen-like substances of human bile. Isolation and partial characterization

Immunofluorescence studies on the occurrence and localization of the CEA-related biliary glycoprotein I (BGP I) in normal human gastrointestinal tissues

Differential regulation of carcinoembryonic antigen and biliary glycoprotein by gamma-interferon

Carcinoembryonic antigens are targeted by diverse strains of typable and non-typable Haemophilus influenzae

Carcinoembryonic antigens (CD66) on epithelial cells and neutrophils are receptors for Opa proteins of pathogenic neisseriae

Reversion of the malignant phenotype of human breast cells in three-dimensional culture and in vivo by integrin blocking antibodies

Soluble form of ICAM-1, VCAM-1, E-and L-selectin in human milk