Novel Progesterone Target Genes Identified by an Improved Differential Display Technique Suggest That Progestin-induced Growth Inhibition of Breast Cancer Cells Coincides with Enhancement of Differentiation*

Progesterone is an important regulator of normal and malignant breast epithelial cells. In addition to stimulating development of normal mammary epithelium, it can be used to treat hormone-dependent breast tumors. However, the mechanism of growth inhibition by progestins is poorly understood, and only a limited number of progesterone target genes are known so far. We therefore decided to clone such target genes by means of differential display polymerase chain reaction. In this paper, we describe an improved differential display strategy that eliminates false positives, along with the identification of nine positive (TSC-22, CD-9, Na+/K+-ATPase α1, desmoplakin, CD-59, FKBP51, and three unknown genes) and one negative progesterone target genes (annexin-VI) from the mammary carcinoma cell line T47D, which is growth-inhibited by progestins. None of these genes have been reported before to be progesterone targets. Regulation of desmoplakin, CD-9, CD-59, Na+/K+-ATPase α1, and annexin-VI by the progestin suggests that progesterone induces T47D cells to differentiate. Three of these genes were repressed by estradiol and up-regulated by the progestin. Estradiol treatment of T47D cells also leads to formation of lamellipodia and delocalization of two cell adhesion proteins, E-cadherin and α-catenin. All these effects were reversed by the progestin. These data suggest that estradiol dedifferentiates T47D cells, while progestins have the opposite effect. This may be linked to the capacity of progestins to inhibit tumor growth.

The steroid hormone progesterone has important and complex effects in female sex organs. In two of its major target organs, the uterus and the mammary gland, it strongly influences proliferation and differentiation. In mouse mammary epithelium, progesterone stimulates proliferation, which eventually culminates in glandular development. In the endometrial epithelium, progesterone inhibits estrogen-mediated growth but also induces differentiation (reviewed in Ref. 1). In line with this, knock-out mice lacking PR 1 have, apart from defects in the ovaries and in sexual behavior, uninhibited proliferation in the endometrium upon estrogen and progestin treatment but severely impaired mammary gland development (2).
Progestins also have strong effects on human hormone-responsive endometrial and breast cancers. Both in endometrial and mammary carcinomas progestins inhibit estrogen-mediated growth, which is, for endometrial carcinoma, associated with increased differentiation. Progestins are therefore used to treat these cancers (reviewed in Ref. 1). The breast tumorderived cell line T47D exhibits in vitro a similar inhibition of estrogen-or growth factor-mediated growth by progestins, although initially a brief growth stimulation is seen upon progesterone addition (3). The precise mechanism responsible for this biphasic response is not known (reviewed in Ref. 1). The influence of progestins on processes essential in tumor progression, like tumor invasion and metastasis, is poorly understood. For endometrial tumor cells, both in vivo and in vitro studies suggest that estrogens stimulate invasion, while progestins inhibit this process (4,5). For breast tumor cells, estrogens appear to stimulate invasion (Ref. 6 and reviewed in Ref. 7), while an effect of progestins has not been reported.
Progestins bind inside the cell to the PR, which belongs to the nuclear receptor superfamily of transcription factors (reviewed in Ref. 8). This receptor subsequently homodimerizes and activates gene transcription after binding to progesterone response elements (PRE) in promoters (reviewed in Ref. 9). PREs have the same consensus sequence as response elements for the related glucocorticoid, androgen, and mineralocorticoid receptors. PR cannot only activate, but also repress genes, for example by negative cross-talk with transcription factors of the AP1 and NF-B families (10,11). To better understand progesterone effects in human breast epithelial cells, we wanted to identify a larger number of progesterone target genes, as only a limited number of established target genes for this hormone in breast tumor cells is known. Negative target genes are the estrogen receptor (ER) (12) and PR itself (13), while strong positive targets are fatty acid synthetase (14), methallothionein-II A (15), alkaline phosphatase (16), pepsinogen C (17), epidermal growth factor, and the epidermal growth factor receptor (3). Transient induction of c-fos and c-myc has been reported and might be involved in the initial growth stimulatory effect (3). Induction of none of these genes can explain the observed * This work was supported by a grant from N. V. Organon, Oss, The Netherlands. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Recently, a new technique for cloning differentially regulated genes has been published, designated as differential display or RNA fingerprinting (18,19). This technique has substantial advantages over earlier techniques, e.g. subtractive libraries or hybridization, since it does not have such a strong bias toward highly abundant genes and is much more versatile, enabling cloning of down-regulated as well as up-regulated genes and allowing a rapid comparison between multiple samples. For identifying progesterone target genes we used a differential display technique as modified by Van Belzen et al. 2 adding a number of improvements. In this paper we present, apart from 10 new progesterone target genes, a fast procedure to identify the most promising genes. Some of the genes identified suggest that the T47D breast tumor cell line differentiates upon progestin treatment. Three of these were repressed by E 2 treatment and subsequently up-regulated by the progestin. By means of immunofluorescence detecting three markers for epithelial differentiation, we could also show that E 2 addition resulted in delocalization of the marker proteins, while additional progestin treatment reversed this effect. From this we conclude that E 2 dedifferentiates the T47D cells, while progestins increase the differentiation state of estrogen-treated T47D cells.

EXPERIMENTAL PROCEDURES
Cell Culture-A phenol red-free 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium (DF) was obtained from Life Technologies, Inc. Trypsin and the EDTA used for cell culture were bought from Flow Laboratories (Irvine, Scotland). Fetal calf serum (FCS) was purchased from Integro (Linz, Austria), E 2 , cycloheximide, dexamethasone, bFGF, human transferrin, and bovine were from Sigma. Org2058 was kindly provided by Organon (Oss, The Netherlands). Human TGF␤1 was kindly provided by Dr. A. J. M. van den Eijnden-van Raaij. DCC-FCS was prepared by treating FCS with DCC to remove steroids, as described (20). T47D cells were a kind gift of Dr. R. L. Sutherland (Garvan Institute of Medical Research, Sydney, Australia).
RNA Isolation-Cells were plated at a density of 4 ϫ 10 6 cells/175cm 2 dish or 1.5 ϫ 10 6 /60 cm 2 dish in DF medium containing 5% (v/v) DCC-FCS, 30 nM selenite, 10 g/ml transferrin, and 0.2% (w/v) bovine and allowed to attach for 24 h. Medium was refreshed, and hormones were added as indicated. Total RNA was isolated using the guanidine isothiocyanate method according to Ref. 21. Poly(A) ϩ RNA was isolated with PolyATtract® beads (Promega, Madison WI).
Differential Display-The differential display was performed according to Van Belzen et al. 2 with some modifications. RNA was isolated from cells that were treated with 10 Ϫ9 M E 2 (48 h before harvesting) and 10 Ϫ8 M Org2058 or vehicle (24 h before harvesting). 20 g of total RNA was digested with RNase-free RQ1 DNase according to the manufacturer's protocol (Promega), then phenol-extracted and precipitated (22). 1 g was used for cDNA synthesis with the Superscript kit (Life Technologies, Inc.) together with 100 ng of T 12-18 primer. The cDNA was purified with Qiaquick spin columns according to the manufacturer's conditions (Qiagen, Hilden, Germany) and eluted with 100 l of TE (10 mM Tris, 1 mM EDTA, pH 8.0). The differential display PCR was performed with 2 l of cDNA, 200 ng of primer, 1.5 mM MgCl 2 , 1 ϫ Goldstar reaction buffer, 2 M dNTPs, 1 Ci of [ 33 P]dATP and 4 units of Goldstar Taq polymerase (Eurogentec, Seraing, Belgium) per reaction of 20 l. The primers used were on average 20 bases long, with 50% GC content. Primers already available from other experiments were used after checking whether they gave a clear differential display pattern under the conditions used (sequences of primers available upon request). Cycles were performed as follows: 3 min, 96°C; 2 ϫ (1 min, 96°C; 5 min, 37°C; 5 min, 37-72°C, 5 min, 72°C); 30 ϫ (1 min, 96°C; 2 min, 55°C; 5 min, 72°C); 10 min, 72°C; soak at 4°C. 5 l of the PCR was mixed with an equal volume of 96% formamide loading buffer, boiled for 2 min, and 6 l was electrophoresed on a Sequagel XR (Biozym, Landgraaf, The Netherlands) sequencing gel until the xylene cyanol marker was at the bottom. After immobilizing the gel on Whatman No. 3MM paper and drying for 30 -60 min under vacuum at 80°C, the gel was exposed against Biomax MR films (Eastman Kodak Co.). Positive bands were excised and immersed in 100 l of TE. The DNA was eluted by freezing the vial twice in liquid N 2 , thawing at 70°C, and vortexing. DNA was purified by Qiaquick spin columns, and 5-10 l was amplified using standard PCR conditions and the primer used for the display (annealing temperature: 55°C). The mixture was loaded on an agarose gel, bands were excised, purified with glassmax (Life Technologies, Inc.), and cloned in pGEM-T (Promega). If necessary, the PCR was repeated.
Procedure for Elimination of False Positive Clones-DNA was isolated from 6 to 8 colonies per excised band using the cetyltrimethylammonium bromide miniprep procedure (23). DNA was cut with PstI-ApaI, RsaI, and TaqI and analyzed on agarose gel. Clones with different restriction patterns were sequenced and reamplified from the plasmid by standard PCR with the primers used in the differential display. The PCR mix was dot-blotted in duplicate on two Hybond C extra filters (Amersham Corp., Little Chalfont, United Kingdom; Ref. 24). Clones identified with the same primer were spotted on the same two filters. With this primer, two differential display PCRs were carried out with RNA from progesterone treated and untreated cells. These reactions were performed with 5 Ci of [ 32 P]dCTP instead of 1 Ci of [ 33 P]dATP, precipitated with ammonium acetate and ethanol, and then used as a probe on the two filters. Filters were hybridized and washed as described (25) in 2-ml Eppendorf reaction vials. A schematic outline of the strategy is depicted in Fig. 1.
Northern Hybridization-Northern blotting was carried out as described (22). The inserts were cut from pGEM-T with ApaI-PstI or SstII-PstI, run on agarose gel, and purified. Labeling was performed with the Rediprime random primer labeling kit and 50 Ci of [ 32 P]dCTP (Amersham), hybridization, and washing was as described (25).
Immunofluorescence Microscopy-Cells were plated at a density of 4 ϫ 10 4 cells (for 3-day incubations) or 1 ϫ 10 4 cells (for 6-day incubations) per coverslip and allowed to attach for 24 h. Medium was refreshed, and hormones were added at the concentrations indicated. After 3 days, medium and hormones were refreshed. Cells were fixed using 2% paraformaldehyde and immunolabeled following standard procedures. As anti-E-cadherin antibody DECMA-1, a kind gift from Dr. R. Kemler (Freiburg, Germany) was used (1:200); anti-␣-catenin was from Sigma (1:2000). As second antibody, species-specific CY-3-labeled antibodies from Jackson (West Grove, PA) were used in a dilution of 1:250.

RESULTS
Differential Display-To clone PR target genes we used cDNA prepared from total RNA isolated from control T47D cells in parallel with cells treated for 24 h with Org2058 (a synthetic progestin). In both cases cells were grown in the presence of E 2 for 48 h before harvest to obtain maximal PR induction (26). For each primer the differential display was performed in duplicate with cDNA derived from two different batches of RNA per treatment. In total, 34 primers were used. We isolated 57 clones with different inserts, derived from 38 differentially regulated bands. Most of them represented upregulation in the presence of the progestin. All clones were partially sequenced and compared against EMBL-44, EMBL-NEW 11, and UGenBank 91_44 data bases, while apparent open reading frames were compared against PIR 46 and Swissprot 31 using IGsuite software.
Initially a number of clones was analyzed on Northern blots using total RNA and poly(A) ϩ RNA Northern blots. Although some clones were found to be up-regulated, several clones gave no signal on Northern blots. In addition, very frequently different clones were isolated from the same band, indicating that the DNA fragments present in the excised gel slices were impure. Therefore, we designed an assay that could potentially discriminate false positives and contaminations from genuine progesterone targets. In Fig. 1 the strategy is illustrated. Because hybridization and washing can be performed in 2-ml Eppendorf tubes, all steps permit simultaneous handling of multiple samples. This allowed us to analyze all samples in one experiment. In this assay, ten positive clones (Ptg-1 to Ptg-10, for progesterone target gene) could be easily distinguished by visual comparison and quantification (Fig. 1). After verification by Northern blot analysis, all of these genes were found to be strongly up-regulated by the progestin (Fig. 2A). Ptg-1, Ptg-5, and Ptg-10 were also clearly up-regulated after 6 h by the progestin in the presence of cycloheximide, a protein synthesis inhibitor. This indicates that these genes are direct progesterone receptor targets, since protein synthesis is not required. To check if PR target genes had been missed with this selection procedure, 18 from the 47 remaining clones were checked on Northern blots. Five gave no signal at all, ten appeared to be false positives, and three (clones Ptg-11 to Ptg-13) were slightly regulated by the progestin (Fig. 2B). In conclusion, with this assay we did not obtain false positives. Analysis of the remaining clones suggested that only some weakly regulated clones were missed (3 out of 13 giving a signal on Northern blots). Therefore, this method is a very effective selection procedure, particularly for strongly induced mRNAs.
Novel Progesterone Target Genes-The ten clones identified in the Dot blot assay, Ptg-1 to Ptg-10, were derived from seven different genes: CD-9/MRP-1, a gene known to have effects on tumor metastasis and cell motility (27,28); CD-59/protectin, a gene coding for a protein that protects cells from lysis by complement factors, acting as a ligand for CD-2 and which has been shown to bind to G i␣ (29 -31); TSC-22, a gene encoding a putative transcriptional repressor (32,33); desmoplakin, a desmosomal protein; FKBP51, an immunophilin (34,35); Na ϩ /K ϩ -ATPase subunit ␣1 (cloned three times), and a fragment with no homology to known genes (cloned two times) (see Table I).
The two clones that were not identified in the Dot blot assay but showed slight up-regulation on Northern blots, Ptg-11 and Ptg-12, were not represented among the sequences present in the EMBL and GenBank™ data bases. However, their open reading frames show homology to splicing factors of the SR protein family and a putative C 2 H 2 -type zinc finger protein from Caenorhabditis elegans, respectively (see Table I). This first open reading frame contains five copies of VTRRRSRSRTSP as a repeat. This repeat shows homology FIG. 1. Dot blot assay to identify most prominent clones. A, outline of the strategy used to identify the most prominent clones. Two identical Dot blots of the isolated clones are prepared, of which one is probed with the differential display product of ϩprogestin RNA and the other with the differential display product of Ϫprogestin RNA. B, examples of Dot blot analysis of the clones Ptg-1, Ptg-2, and Ptg-8. Dot blots were hybridized with differential display product from RNA treated with Org2058 or vehicle for 24 h. All clones are spotted in duplicate on the same filter next to each other. The other dots that are spotted on the same filter as the Ptg clones originate from false positive or weakly regulated clones, amplified with the same primer as the Ptg clone. Fold induction was corrected for input. Ptg clones are indicated by arrows.

FIG. 2. Northern blot analysis of novel progesterone target genes.
A, Northern blots for clones that were identified with the Dot blot assay. 0ch, 6-h cycloheximide (10 g/ml) treatment without progestin; 6ch: 6-h progestin (10 Ϫ8 M Org2058) and cycloheximide treatment; 0, no progestin treatment, 24, 24-h progestin treatment. Clones Ptg-1 to Ptg-10 were hybridized on 20-g total RNA blots. GAP ϭ GAPDH signal, to correct for unequal loading. Indicated are also the positions of the 18 and 28 S ribosomal RNAs. B, Northern blots for Ptg-11, Ptg-12 and Ptg-13, which were not picked up with the Dot blot assay. Clones Ptg-11 and Ptg-12 were hybridized on a 1-g poly(A) ϩ RNA blot, clone Ptg-13 on a 20-g total RNA Northern blot. The abbreviations used are the same as in A.
(56% amino acid identity) to the arginine-serine domain of SR proteins, although also other arginine-serine-rich proteins have high scores (like protamines and viral DNA-binding proteins). The only clone that was found to be down-regulated (Ptg-13), was annexin-VI. This gene encodes a calcium-dependent phospholipid binding protein possibly involved in intracellular endosomal trafficking (37,38).
The expression pattern of four genes (FKBP51, Na ϩ /K ϩ -ATPase ␣1, CD-9, and desmoplakin) was further characterized (Fig. 3). Time-dependent expression in the presence of the progestin and E 2 was investigated to determine whether the response was transient or sustained. All four genes were still strongly up-regulated after 48 h. Furthermore, regulation of these genes by the progestin in the additional presence of different growth stimuli (E 2 and bFGF) was examined, and in the mammary carcinoma cell line MCF-7, which is also strongly responsive to glucocorticoids and the growth inhibitor TGF␤. FKBP51 was strongly up-regulated by both the progestin and the synthetic glucocorticoid dexamethasone under all conditions tested, agreeing with the assumption that FKBP51 is a direct target gene.
The three other genes showed more complex expression patterns. The Na ϩ /K ϩ -ATPase ␣1 was up-regulated by the progestin in T47D cells under all conditions tested (Fig. 3A); in MCF-7 cells, the gene was up-regulated by the progestin but downregulated by dexamethasone, which is not to be expected for a direct progesterone target gene (Fig. 3B). Also the regulation by TGF␤ was complex with up-regulation in the presence of E 2 and down-regulation in its absence. CD-9 was only strongly up-regulated in T47D cells by the progestin in the presence of E 2 (Fig. 3A), while in MCF-7 cells, this gene was activated constitutively (Fig. 3B). Desmoplakin was clearly induced in T47D cells by the progestin in the absence of additional hormones or in the presence of estradiol, but not in the presence of bFGF (Fig. 3A). In MCF-7 cells, this gene was only slightly up-regulated by the progestin, dexamethasone, and TGF-␤ in the presence of estradiol (Fig. 3B). Interestingly, we noted that Na ϩ /K ϩ -ATPase ␣1, CD-9, and desmoplakin were all downregulated in T47D cells by E 2 , while the progestin clearly reversed this effect (Fig. 3A).
Estradiol and Progestins Have Opposing Effects on Differentiation of T47D Cells-Interestingly, five of the genes identified suggested that the T47D cells differentiate upon progestin treatment. These are: desmoplakin, which is a marker for epithelial differentiation and is expressed higher in well differentiated mammary carcinoma cell lines (39); CD-9, which is expressed higher in well differentiated leukemias and lower in highly metastatic breast tumors (28,40); CD-59, which is ex-pressed higher in well differentiated colorectal carcinomas (41); Na ϩ /K ϩ -ATPase ␣1, which is a very important protein in secretory epithelial cells (42); and the negative target gene annexin-VI, which is down-regulated in developing mammary gland secretory epithelium (43)(44)(45). Note that all three positive  progesterone target genes that were repressed by E 2 as compared with no hormonal treatment belong to this group. Because several of the target genes identified were in line with the induction of differentiation by progestins, we studied if this could be confirmed by studying other markers of differentiated epithelia. We observed that upon hormonal treatment clear morphological changes of the T47D cells were apparent. Upon E 2 treatment, cells become more pointed and lamellipodia are visible, already after 24 h (Fig. 4). These lamellipodia clearly contain, as determined by immunocytochemistry, paxillin (data not shown), a cell-matrix adhesion protein commonly associated with lamellipodia (46). In the presence of the progestin, with or without E 2 , hardly any lamellipodia are induced, and cells become much more rounded, consistent with a more differentiated phenotype. We also investigated the localization of two adherens junction proteins, E-cadherin and ␣-catenin. The expression of these genes is often reduced in tumor cells and correlates with differentiation state (39,47). After 6 days in the presence of E 2 , E-cadherin and ␣-catenin localization was clearly reduced at cell borders and much was present in lamellipodia (Fig. 4). The localization of these proteins in lamellipodia is already visible after 24 h and does not occur in the presence of cycloheximide, indicating that changes in protein expression are required (data not shown). This situation is completely reversed when a progestin is added. In that case these proteins were again localized at the cell borders, which is consistent with the localization in fully differentiated epithelial cells (42). These results confirm that the progestin induces differentiation in T47D cells.

Cloning of Genes Using an Improved Differential Display
Protocol-Differential display is a very useful technique for identifying regulated genes. However, a major problem with this technique is the occurrence of a large number of false positives, contaminations, and inefficient probes for detection on Northern blots, as has been described for the Liang and Pardee method (48 -50). Hence, we devised a powerful strategy to identify the most promising clones rapidly and efficiently. During the preparation of this manuscript a comparable technique was published (51). There are some differences with our procedure. Vögeli-Lange et al. (51) recommend to pool multiple reaction mixtures of previous differential display reactions for use as a probe, while in our procedure, fresh differential display reactions are performed. Using the same reaction mixture has the disadvantage that when a false positive band is cloned, again a false positive signal will be generated in the Dot blot procedure. Moreover, we use 32 P instead of 33 P as label in the Dot blot procedure, which gives a stronger signal and circumvents the need to pool reactions.
In the present case we have identified mostly up-regulated genes. Possibly, there are more genes up-regulated than downregulated after progestin treatment in T47D cells. A different explanation could be that genes which are up-regulated show larger differences in expression level upon progestin treatment and therefore are identified more easily. Clones derived from the positive target Na ϩ /K ϩ -ATPase ␣1 were even collected independently three times. Probably this is due to the fact that the expression of this gene upon stimulation is very high. One other gene was selected twice (Ptg-1 and Ptg-3), but this is presumably not a result of abundance of the mRNA. Both clones had the same primer incorporated at the same site of the gene at one end, while at the other end the primer had annealed at different sites. The clones were therefore of different size and cut out twice. All the genes identified here have never been reported to be progesterone targets, although the Na ϩ / K ϩ -ATPase subunit ␣1 and TSC-22 have been shown to be up-regulated by dexamethasone (32,36), a synthetic activator of the glucocorticoid receptor (note that the glucocorticoid response element is identical to the PRE). This could suggest that there are still more progesterone target genes to be identified.
Novel Progesterone Target Genes-The genes that were identified can be divided into two groups. The first group consists of the four genes that might play a role in regulating gene expression. These are TSC-22, which is a putative transcriptional regulator, Ptg-12, possibly encoding a zinc finger protein, Ptg-11, which has homology with members of the SR protein family of splicing factors, and FKBP51, an immunophilin. The immunophilins FKBP52, FKBP54, and CYP40 have been shown to bind to steroid hormone receptor complexes (52)(53)(54), and FKBP52 is required for nuclear transport of the glucocorticoid receptor (55). FKBP51 might have a similar function and therefore directly influence transcriptional activity of steroid hormone receptors. Very clear was the strong up-regulation of this gene by the ligand-bound PR (also in the presence of cycloheximide) and glucocorticoid receptor, which could indicate the presence of one or more PRE/glucocorticoid response elements in its promoter.
The second group consists of genes that are suggestive for differentiation of T47D cells upon progestin treatment. Low expression of CD-9/MRP-1 (motility-related protein 1) correlates in leukemias with poor differentiation (40), in lung cancer with poor prognosis (56), and in breast cancer with stronger metastatic potential and poor prognosis (28,57). CD-59/protectin is strongly expressed in well and moderately differentiated colorectal carcinomas but to a lower extent in metastasizing tumors (41). A third gene that suggests differentiation to be induced by progesterone is desmoplakin. The product of this gene connects intermediate filaments with the desmosomes, and in this way a continuous filamentous network is formed between cells (58). Desmoplakins are markers for epithelial differentiation. The expression of desmoplakin, along with other epithelial markers, in mammary carcinoma cell lines has been correlated negatively with invasiveness and positively with the differentiation state (39). In normal breast tissue, progesterone stimulates glandular development (1). This is consistent with the induction of the Na ϩ /K ϩ -ATPase ␣1 subunit. This gene is likely to be important for vectorial transport in epithelial cells (42) and is indeed up-regulated during development of the intestinal epithelium (59). Furthermore, we identified the negative target gene annexin-VI, which is downregulated in developing mammary gland secretory epithelium (43)(44)(45). Three positive target genes of this group all showed not only an up-regulation upon progestin treatment in the presence of E 2 , but also an estrogen-induced down-regulation. Not many negative estrogen receptor target genes are known so far, and the ER is mostly considered as a positive transcription factor.
Progestins Reverse Dedifferentiating Effects of Estrogens-Apart from opposing E 2 -induced suppression of gene expression, progestins oppose morphological changes (cell shape, appearance of lamellipodia) induced by E 2 . This is also clear from our findings concerning the localization of ␣-catenin and Ecadherin. A vast literature is available on the significance of expression and distribution of E-cadherin during differentia-tion and invasion of carcinoma cells. E-cadherin expression is high in well differentiated breast or other tumors, is considered to be an invasion suppressor, both in vivo and in vitro, and correlates with good prognosis (60 -63). ␣-Catenin expression is often reduced in breast tumors, and this is probably also important in invasion (47,64).
Our data suggest that the human mammary carcinoma line T47D dedifferentiates after estrogen treatment, while a differentiation program is activated upon adding the progestin. Progestins can be used in breast cancer treatment and can induce regression of tumors through unknown mechanism (reviewed in Ref. 65). In contrast, estrogens are strong stimulants of breast cancer proliferation (reviewed in Ref. 66). Possibly, the influence of estrogens and progestins on breast tumor cell differentiation plays an important role in growth modulation of breast tumors.
In general, a higher state of differentiation is linked to a lower metastasizing capacity (reviewed in Ref. 67). The dedifferentiating action of E 2 correlates well with the reported stimulation of invasion by estrogens (6,68). A clear picture for progestins has not emerged yet. In literature, progestins have been shown to up-regulate invasion-associated proteins, like laminin receptors and proteinases (7,17,69). However, to our knowledge in in vitro or in vivo invasion assays no stimulatory effect of progestins has been reported. Our data rather suggest that progestin-treated cells will be less invasive, at least in the presence of E 2 . This is based on the progestin-induced expression of the invasion suppressors E-cadherin, CD-9, and ␣-catenin and from the effects on lamellipodia, which are instrumental in locomotion and tumor cell migration.
The mechanism by which the progestin antagonizes E 2 in our experiments is not clear. The PR has been shown to downregulate ER activity in many ways. As possible mechanism down-regulation of ER levels, the induction of E 2 degrading enzymes, or competition for cofactors that bind to both receptors can be envisaged (reviewed in Ref. 1;Ref. 70). Although this down-regulation of ER activity seems to play a role in progestin action in T47D cells, some of the differentiationassociated genes are already up-regulated by the progestin in the absence of E 2 . Therefore, the progestin on its own has a clear effect on differentiation. The influence of steroid hormones on breast tumor differentiation and the role of this differentiation on growth inhibition and metastasis clearly represents an important area for further study.