Differential Gene Regulation by the Two Progesterone Receptor Isoforms in Human Breast Cancer Cells*

The PR-A and PR-B isoforms of progesterone receptors (PR) have different physiological functions, and their ratio varies widely in breast cancers. To determine whether the two PR regulate different genes, we used human breast cancer cell lines engineered to express one or the other isoform. Cells were treated with progesterone in triplicate, time-separated experiments, allowing statistical analyses of microarray gene expression data. Of 94 progesterone-regulated genes, 65 are uniquely regulated by PR-B, 4 uniquely by PR-A, and only 25 by both. Almost half the genes encode proteins that are membrane-bound or involved in membrane-initiated signaling. We also find an important set of progesterone-regulated genes involved in mammary gland development and/or implicated in breast cancer. This first, large scale study of PR gene regulation has important implications for the measurement of PR in breast cancers and for the many clinical uses of synthetic progestins. It suggests that it is important to distinguish between the two isoforms in breast cancers and that isoform-specific genes can be used to screen for ligands that selectively modulate the activity of PR-A or PR-B. Additionally, use of natural target genes, rather than “consensus” response elements, for transcription studies should improve our understanding of steroid hormone action.

Progesterone receptors (PR) 1 are ligand-activated transcription factor members of the steroid hormone family of nuclear receptors. They exist naturally as two isoforms, PR-B and PR-A, transcribed from two promoters on a single gene (1). Human PR-B are 933 amino acids in length and contain a unique activation function AF3 (2). PR-A lack the 164 N-terminal residues that contain AF3 and are 769 amino acids in length. Both isoforms are physiologically important. Mice lacking both PR display pleiotropic reproductive abnormalities, incomplete mammary gland development, and impaired thymic function and sexual behavior (3), whereas those lacking only PR-A exhibit a subset of these phenotypes (4).
Clinically, PR are important therapeutic targets. Progestational agents are widely used for oral contraception, menopausal hormone replacement therapy (HRT), and to treat breast cancer and endometrial hyperplasia (5,6). Antiprogestins are in clinical trials for contraception, induction of labor, and the treatment of meningiomas, endometriosis, and endometrial cancers. In breast cancers, total PR levels are routinely measured as a guide to hormone therapy and as markers of disease prognosis (7)(8)(9)(10). Interestingly, whereas progestins added to HRT successfully decrease the incidence of endometrial cancer, they increase the incidence of breast cancer (11,12).
Little is known regarding the unique roles of the two PR isoforms in progesterone target tissues. In vitro, the two receptors have markedly different transcriptional effects on progestin-responsive promoters (2,(13)(14)(15)(16). The antiprogestin RU486 has partial agonist effects only on PR-B, whereas only PR-A inhibit PR-B and other steroid receptors including estrogen receptors (ER) (17)(18)(19). In vivo, the two PR isoforms are usually coexpressed in normal cells, yet their ratio varies dramatically in different tissues, physiological states, and in disease (20 -22). For example, in the estrogen-treated primate, the hypothalamus expresses an excess of PR-B, but the pituitary expresses an excess of PR-A (23,24). In human endometrium the levels and ratio of PR-A to PR-B vary extensively during the menstrual cycle (25)(26)(27)(28), and overexpression of PR-B is associated with highly malignant forms of endometrial, cervical, and ovarian cancers (29,30).
With regard to the mammary gland, in transgenic mice, 3:1 overexpression of PR-A over PR-B results in extensive epithelial cell hyperplasia, excessive ductal branching, and a disorganized basement membrane, all features associated with neoplasia (31). In contrast, overexpression of PR-B leads to premature ductal growth arrest and inadequate lobulo-alveolar differentiation (32). Interestingly, PR-A null mice, which express only PR-B, exhibit normal mammary gland development, yet the same mice display severe uterine hyperplasia and reproductive defects (4). Collectively, these data suggest that PR-A and PR-B have physiologically different functions in different tissues and that alterations in their ratios carry different consequences depending on the tissue.
Although PR levels are routinely measured in breast cancers for clinical decision making, only two studies have examined the levels of the two isoforms. An analysis of 202 PR-positive breast cancers by immunoblotting shows that expression levels of PR-A are higher than PR-B in 59% of tumors and are 4-fold or greater in 25% of tumors (33). In another study of 32 PRpositive breast cancers, excess PR-B correlated with the absence of Her-2/neu indicating a good prognosis, whereas excess PR-A correlated with a poorly differentiated phenotype and higher tumor grade (34). Overexpression of PR-A in cultured human breast cancer cells results in marked morphological changes and loss of adherent properties (35), suggesting, as do the transgenic mice data, that an excess of PR-A is particularly harmful in the breast.
Little is known at present about the molecular mechanisms that might explain these differences. We therefore undertook the first systematic, large scale comparison of gene regulation by the two PR, using a unique human breast cancer cell model for this purpose. Wild-type T47Dco breast cancer cells express equimolar levels of PR-A and PR-B in an estrogen-independent manner (36). To study differential gene regulation by the two PR isoforms independently, we isolated a PR-negative subline of T47Dco (designated T47D-Y cells) and then engineered the T47D-Y to stably express equivalent levels of either PR-B (T47D-YB cells) or PR-A (T47D-YA cells) (37). Because these are pure cell populations, and all of the cells have the same parental-cell background, the PR isoform-specific effects of progesterone on gene transcription can be quantitatively and reproducibly investigated in a tightly controlled manner.
Our data, based on triplicate determinations, demonstrate that in response to progesterone, PR-A and PR-B primarily regulate different subsets of genes, and although PR-B are transcriptionally more active, there are genes that are uniquely regulated by PR-A. These subsets include genes known to be involved in breast cancer and/or mammary gland development but not previously known to be progesterone targets. Progesterone regulation of many of these genes would be deleterious in breast cancers. A surprisingly large number of genes are targeted to the cell membrane or involved in membrane-initiated signaling. Other gene clusters are involved in metabolism, transcription, cell growth and apoptosis, and nucleic acid and protein processing. The results suggest that PR-A and PR-B have different molecular functions and that it may be important to quantify either the PR isoform content of breast cancers or their gene targets, rather than total PR.

EXPERIMENTAL PROCEDURES
Cell Culture-The PR-positive T47Dco breast cancer cell line, isolation of its PR-negative clonal derivative T47D-Y, and construction of T47D-YA and T47D-YB cells have been described (36,37). Cells are routinely cultured as described previously (38).
Atlas TM Human cDNA Expression Array-T47D-YA and T47D-YB cells were grown to ϳ70% confluence in Eagle's minimum essential medium (MEM) with Earle's salts as described previously except without G418, washed with serum-free MEM, and changed into MEM containing 5% charcoal-stripped fetal calf serum for 24 h, after which the cells were treated with 10 nM progesterone dissolved in ethanol or with ethanol alone, for 6 or 12 h. Total RNA was prepared from the 4 sets of cells using guanidinium isothiocyanate; poly(A) ϩ RNA was purified with the Oligotex mRNA kit (Qiagen, Valencia, CA), and 32 P-labeled cDNA was synthesized from 1 g of each sample using SuperScript II reverse transcriptase (Invitrogen). Labeled probes were hybridized to Atlas TM Human cDNA Expression Array (CLONTECH, Palo Alto, CA) nylon membranes onto which 588 cDNA fragments encoding known proteins are spotted in duplicate. After processing according to the CLONTECH User Manual (PT3140-1 PR91208), signals were detected by PhosphorImager TM (Molecular Dynamics, Sunnyvale, CA). Data were analyzed using Atlas TM Image 1.0 (CLONTECH) and normalized to signals from control housekeeping genes on the same filter. For selected genes, progesterone inducibility and PR isoform specificity were confirmed by reverse transcriptase (RT)-PCR, and/or Western blotting as described below.
Affymetrix GeneChip TM Experiments-T47D-Y, T47D-YA, and T47D-YB cells were grown to ϳ70% confluence in MEM without anti-biotic, washed with serum-free media, and changed into media containing 5% charcoal-stripped fetal calf serum for 24 h. Cells were then treated with 10 nM progesterone dissolved in ethanol, or in ethanol alone, for 6 h. Total RNA and poly(A) ϩ RNA were prepared from the 6 samples as described above. Poly(A) ϩ RNA was processed according to the Affymetrix Expression Analysis Technical Manual (P/N 700218 rev2). Briefly, first strand and second strand cDNA syntheses were performed, and biotin-labeled cRNA was generated using the EnZo BioArray TM High Yield Transcript Labeling Kit (Enzo Diagnostics, Inc., Farmingdale, NY). Unincorporated nucleotides were removed with RNeasy affinity columns (Qiagen, Valencia, CA). Purified, biotinylated cRNAs were quantified, and 20 g were subjected to a fragmentation reaction to randomly generate fragments ranging from 35 to 200 bases. HuGeneFL 6800 Array TM chips consisting of 5,600 full-length human genes from Unigene, GenBank TM , and TIGR data bases were used for hybridization. Thirty l of fragmented cRNA were added to a hybridization mixture together with control oligonucleotide B2 and control cRNA mixture, then washed, and stained with streptavidin/phycoerythrin. DNA chips were read at a resolution of 6 m with a Hewlett-Packard GeneArray Scanner. The entire experiment was performed three separate times with PR-positive T47D-YA and T47D-YB cells and two separate times with PR-negative T47D-Y cells. Each repeat was separated by ϳ1 month and was designed to be a true replicate taking into account experimental variability in cell culture conditions and sample preparation. To determine which progesterone-regulated genes are direct targets of PR, a separate experiment was performed in which T47D-YB cells were treated with cycloheximide (10 g/ml) 30 min before treatment with or without 10 nM progesterone. Cells were otherwise treated as described above, and RNA was derivatized and hybridized to microarray chips as above.
Data Analyses and Statistics-Detailed protocols for data analyses of Affymetrix microarrays and extensive documentation of the sensitivity and quantitative aspects of the method have been described (39). Briefly, MicroArray Suite 4.0 Expression Analysis Program TM (Affymetrix, Inc., Santa Clara, CA) was used for the first level of analysis, including the "present" or "absent" call, and pairwise comparisons. Each gene on the chip is represented by perfectly matched (PM) and mismatched (MM) oligonucleotides from 16 to 20 regions of the gene. The number of instances in which the PM hybridization signal is larger than the MM signal is computed along with the average of the logarithm of the PM:MM ratio (after background subtraction) for each probe set. These values were used to arrive at a matrix-based decision concerning the presence or absence of an RNA transcript. The average difference serves as a relative indicator of the level of expression of a transcript and is used to determine the change in the hybridization intensity of a given probe set among different experiments. Multiple experimental (minus versus plus progesterone) pairwise comparisons were performed. In addition, multiple control comparisons (all minus hormone samples and all plus hormone samples) were performed to serve as a measure of the variability among samples. Finally, we compared fold change in "minus" versus "plus" hormone sets in PRpositive cells to fold change in PR-negative controls.
The data were also analyzed using GeneSpring TM version 4.0 (Silicon Genetics, San Carlos, CA) to identify patterns of gene regulation in PR-A, PR-B, or PR-negative cells treated with or without progesterone. To normalize for staining intensity variations among chips, the average difference values for all genes on a given chip were divided by the median of all measurements on that chip. In addition, to scale the gene expression measurements so that they could be plotted on a reasonable y axis for visualization in GeneSpring TM 4.0, the average difference value for each individual gene was then normalized to itself by dividing all measurements for that gene by the mean of the expression values of the gene over all the samples. Normalized values below 0 were set to 0. Finally, to identify patterns of gene expression among cell lines and hormone treatments, k-means clustering was performed generating 24 clusters representing 53.8% explained variability. This generated clustergrams of genes regulated by progesterone in a PR isoform-dependent manner. However, because replicates were done for each cell line, additional statistical analyses were possible, and genes whose regulation was not statistically significant were discarded from the clusters. Statistical significance was assessed by one-way analysis of variance using a cut-off value of p Ͻ 0.05, followed by a Tukey multiple comparison test to determine whether the expression level in any individual cell line or hormone treatment was different from all other expression levels. The genes shown in the figures and listed in the tables were statistically significantly regulated by progesterone or, in the case of the ligand-independent effects, were significantly different in the presence versus the absence of PR.
Immunoblots-Cells were plated at 1 million per 100-mm 2 plates, treated with 10 nM progesterone for the times indicated, and harvested in RIPA buffer as described previously (38). Protein extracts were equalized to 150 g by Bradford assay (Bio-Rad), resolved by SDS-PAGE, and transferred to nitrocellulose. Equivalent protein loading was confirmed by Ponceau S staining. Following incubation with the appropriate antibodies and horseradish peroxidase-conjugated secondary antibodies, protein bands were detected by enhanced chemiluminescence (Amersham Biosciences). Primary antibodies were STAT5 C-17 (detects both STAT5A and 5B isoforms), p21 (C-18), C/EBP␤ (⌬ 198) (specific for LAP isoforms), and C/EBP␤ (C-19) (detects both LAP and LIP isoforms), all from Santa Cruz Biotechnology (Santa Cruz, CA). Cdk1/ckc2 (PSTAIR). The anti-PR antibodies used, AB-52 and B-30, were from our laboratories.

RESULTS
The Model System-PR immunoblots show that the two stable cell lines, T47D-YA and T47D-YB, contain equal amounts of PR-A or PR-B, respectively, and each isoform is expressed at levels comparable with its levels in the parental T47Dco cells (Fig. 1, left panel). These levels are half of the total PR in T47Dco. This was confirmed by ligand binding assays (not shown), in which the T47D-YA and T47D-YB cell extracts bind equivalent amounts of [ 3 H]R5020, at half the levels bound by T47Dco cell extracts. In addition, 6 h of progesterone generates the expected ligand-dependent phosphorylation and down-regulation (40) of both isoforms (Fig. 1, right panel) in a manner identical to that seen with wild-type T47Dco cells (not shown). This important ligand-dependent receptor down-regulation is required for strong transcriptional activity by progesterone (41).
Summary of Findings-To identify genes regulated by the two PR, replicate data points representing gene expression levels in T47D-YA or T47D-YB cells, and in PR-negative T47D-Y cells, treated with or without progesterone for 6 h, were analyzed by pairwise comparison. Genes that increased or decreased more than 1.8-fold in all three experiments and showed no significant variation among controls (PR negative cells, or triplicate "minus hormone" sets) were identified. Altogether, 94 genes of the 5,600 interrogated met these criteria (Table I). Fold changes are the average of triplicate experiments. In cases in which both receptors regulate the same gene, fold changes for each receptor are shown. Genes that were undetectable and called "absent" in one sample, but were detectable and called "present" in the other, are denoted with a tilde in Table I. (Note that the latter cannot be compared with fold changes for genes that were present in both samples, because the genes called absent were set to background levels.) All other genes represent ones that were present even in the absence of treatment but whose levels were altered by hormone. Genes indicated by Footnote a in Table I were identified with Atlas arrays; those indicated by Footnote c were identified using both Atlas and Affymetrix systems. All others were iden- tified with the Affymetrix chips.
In summary, there are six sets of progesterone-regulated genes as follows: (i) 59 genes uniquely up-regulated by PR-B; (ii) 4 uniquely up-regulated by PR-A; (iii) 19 up-regulated by both receptors; (iv) 6 uniquely down-regulated by PR-B; (v) 0 uniquely down-regulated by PR-A; and (vi) 6 down-regulated by both receptors. These data demonstrate that the two PR isoforms largely regulate different subsets of genes. The low number of genes regulated by both receptors was a surprising outcome. Based on progesterone-induced fold changes in gene expression levels in the presence of cycloheximide, over half of the progesterone-regulated genes (51 of 94) are direct targets of PR. These are indicated as Footnote b after the accession number in Table I.
Functional Categories, Known Progesterone-regulated Genes, and Breast Cancer/Mammary Gland Development Genes-The genes were organized into functional categories (Table I) based on GeneCard information and an extensive review of the literature. Categories containing multiple progesterone-regulated genes include the following: (i) a large number of membraneassociated proteins including cell adhesion and cytoskeletal proteins, cytokines, and cytokine receptors, chemokines, secreted proteins, calcium-binding proteins, and membrane signaling molecules; (ii) many steroid, lipid, and general metabolic proteins; (iii) nucleic acid and protein processing factors; (iv) proteins involved in cell growth and apoptosis; and (v) transcription factors. Together these genes draw a picture of progesterone as an important metabolic hormone, with many surprising cell surface effects.
Sixteen of the 94 genes found to be regulated by progesterone in the present study have been reported previously to be progesterone-responsive in either breast cancer cells or other hormone-responsive cell types or tissues (Table II). The independent confirmation of these 16 genes serves as an internal control and demonstrates the quality of our data. The data described here increase the number of known progesterone-regulated genes by ϳ6-fold. A set of 10 genes had been reported previously to be involved in either breast cancer or mammary gland development (Table III). Most, however, were not previously known to be progesterone-regulated.
Cluster Analysis and Confirmatory Studies-Average differences indicating relative intensities from replicate data sets were entered into GeneSpring TM . Gene expression measurements were scaled so that they could be plotted on a reasonable y axis for clustergram visualization. This was accomplished by normalizing each gene to itself (by dividing all measurements for each gene by the median of its expression values across all samples). Normalized values below 0 were set to 0. Because of these normalization procedures, the fold changes may appear different in the clustergrams than reported in Table I Table I. A, cluster of genes up-regulated by progesterone in PR-A-and PR-B-containing cells but not in the PR-negative cells. A highly progesterone-regulated gene, the leucine zipper transcription factor, EZF, is shown as a dashed line. B, gene expression pattern of EZF isolated from the cluster in A, showing standard error bars for replicate experiments (top). RT-PCR using cDNA generated from T47D-YA and -YB cells treated with 10 nM progesterone (ϩ) or vehicle (Ϫ) for 3, 12, and 24 h. Products were generated by RT-PCR using primers specific for EZF and ␤2MG. C, because the average difference value for each gene was normalized to itself by dividing all measurements for that gene by the median of the expression values of the gene over all samples, genes expressed at detectable levels in all samples have relative intensity ratios that cluster around 1.0. These genes are shown in a re-scaled version of A. D, gene expression pattern for S100P isolated from C, showing standard error bars for replicate experiments (top). RT-PCR from cDNA from T47D-YA or -YB cells treated with 10 nM progesterone (Prog) (ϩ) or ethanol vehicle (Ϫ) for 3, 12, and 24 h was performed with specific primers for S100P and ␤2MG.
Affymetrix algorithms. Within clusters, any one gene can be viewed individually, and standard error bars generated from the replicate experiments are then shown. Only genes regulated in a statistically significant manner are listed in Table I, and although clusters were generated by k-means, only those genes that were regulated in a statistically significant manner were left in the clusters. For several genes of interest, the array data were confirmed by measurement of the expressed transcripts by RT-PCR, or the proteins by Western blotting, to assess progesterone regulation at multiple time points. Additionally, for two differentially regulated genes, ITGA6 and BCL-X, the isoform specificity of the regulation was confirmed by in vitro transcription assays using their promoters, in a cell line other than T47D.
PR-A and PR-B Up-regulated Genes-Nineteen genes were up-regulated by both PR isoforms. Fig. 2A shows a cluster of such genes. They are up-regulated by progesterone in both PR-A and PR-B containing cells but not in the PR-negative cell line. The gene encoding the leucine zipper protein, EZF, shown as a dashed line in Fig. 2A and is isolated in Fig. 2B, top, to show the standard error for triplicate determinations. EZF was below detectable levels on the Affymetrix chips in all cell lines in the absence of progesterone and was detectable only in PR-positive cells in the presence of progesterone. EZF was detectable at 25 cycles by RT-PCR, however, and was upregulated after progesterone treatment at 3, 12, and 24 h in the presence of both PR-A and PR-B (Fig. 2B, bottom).
Most genes were expressed at detectable basal levels, even in the absence of progesterone, and for such genes, dividing by the mean over all samples results in relative intensity ratios of ϳ1.0, as shown when the data from Fig. 2A are re-scaled in Fig.  2C. Although some of the genes in Fig. 2A were more strongly up-regulated by PR-B than PR-A, others, as shown in Fig. 2C, are equally well regulated by both PR isoforms. An example of the latter is the gene encoding calcium-binding protein S100P (Fig. 2C, in red). S100P is up-regulated by both PR (Fig. 2D,  top) and is up-regulated as early as 3 h and remains elevated after 24 h of progesterone treatment, as shown by RT-PCR (Fig.  2D, bottom).
PR-B Up-regulated Genes-The majority of genes (59 of 94) are uniquely up-regulated by PR-B as illustrated by the cluster in Fig. 3A and in Table I. Tissue factor (F3), indicated by an arrow in Fig. 3A and isolated in Fig. 3B, top, is one example. Its differential regulation by PR-B was confirmed by RT-PCR (Fig.  3B, bottom). The tissue factor transcript is up-regulated 3 and 12 h after the start of progesterone treatment but only in cells expressing PR-B. It is undetectable 24 h after the start of progesterone, however, indicating that its regulation is transient.
Integrin ␣ 6 is also regulated only by PR-B. This was observed using Atlas TM arrays and was confirmed by RT-PCR at 6, 12, and 24 h after progesterone treatment (Fig. 3C, top). The integrin ␣ 6 promoter had been cloned and reported to be progesterone-responsive (42). We therefore used this promoter, linked to luciferase, to demonstrate the PR isoform specificity of its regulation in an exogenous transcription system and a different cell line (Fig. 3C, bottom). Indeed, the integrin ␣ 6 promoter was induced 9-fold by progesterone in HeLa cells transfected with PR-B but was not induced by PR-A. That the differential regulation of this promoter was recapitulated in an entirely different cell line and system validates the PR-B-specific regulation in T47D cells and provides a model where the mechanisms underlying the isoform specificity can be dissected.
STAT5A and C/EBP␤ are two important mammary gland regulatory proteins (43)(44)(45). Their expression levels are also controlled only by PR-B (Fig. 4). Fig. 4A shows regulation of Stat5a by PR-B isolated from the clustergram shown in Fig. 3A. Its preferential regulation by PR-B is confirmed by the immunoblot in Fig. 4B (black arrow). In the same experiment, p21 and cyclin D1, known progesterone-regulated genes (46 -48), are equally well regulated by either PR isoform (Fig. 4B).
Gene array data for C/EBP␤ regulation by progesterone are shown in Fig. 4C. The protein product of this gene is also uniquely regulated through PR-B as confirmed by the immunoblot (Fig. 4D), using antibody specific for the Lap isoforms of C/EBP␤. The C/EBP␤ Lip isoforms are also up-regulated by progesterone only through PR-B (not shown). Again, cyclin D1 and p21 are regulated by both PR isoforms. Note that cyclin D1 is up-regulated at 24 h (Fig. 4, B and D), but returns to control by 48 h (Fig. 4D). In contrast, p21 is still elevated at 48 h.
PR-A Up-regulated Genes-Only four genes were preferentially up-regulated by PR-A (Fig. 5 and Table I Bottom, autoradiographic image of [␣-32 P]dCTP incorporated into RT-PCR products generated using primers specific for tissue factor (TF) and ␤2MG, using cDNA generated from T47D-YA and -YB cells treated with 10 nM progesterone (ϩ) or vehicle (Ϫ) for 3, 12, and 24 h. C, top, [␣-32 P]dCTP-labeled products were generated by RT-PCR using primers specific for integrin ␣ 6 and ␤2MG, using cDNA isolated from T47D-YA and -YB cells treated with 10 nM progesterone (ϩ) or vehicle (Ϫ) for 6, 12, and 24 h. Bottom, HeLa cells transiently transfected with, from left to right, a promoter for integrin ␣ 6 linked to a luciferase reporter (pGL3 vector) and either PR-B or PR-A expressed in pSG5, the empty pSG5 vector with the integrin ␣ 6 promoter, or the empty pSG5 and empty pGL3 vectors, in triplicate dishes. Cells were treated with ethanol vehicle (open bars) or 10 nM progesterone (solid bars) for 24 h and harvested. Relative luciferase activity units corrected for transfection efficiency using the ␤-galactosidase expression plasmid pCH110 are shown with standard deviations for triplicate determinations.
(HEF1), is predominantly up-regulated by PR-A (Fig. 5A) as shown by the array data (top) and RT-PCR data (bottom). The gene encoding the orphan nuclear receptor, estrogen-related receptor ␣ (ERR␣), also is only significantly up-regulated by PR-A (Fig. 5B), as shown by the array data (top) and RT-PCR (bottom). Finally, the anti-apoptosis gene BCL-X L appears to be uniquely up-regulated by PR-A. This was first observed with the Atlas TM macroarrays (not shown). The Affymetrix microarray data were equivocal (Fig. 5C, top), as standard error bars were large. However, RT-PCR (Fig. 5C, bottom) clearly demonstrated preferential regulation by PR-A. To confirm the unique regulation of BCL-X L by PR-A, we isolated ϳ1000 bp of the human BCL-X promoter and cloned it in front of a luciferase reporter (Fig. 5D). The construct was transfected into HeLa cells together with one or the other PR isoform, and cells were treated with or without progesterone. Hormone-dependent regulation of the BCL-X promoter was observed only in the presence of PR-A (Fig. 5D).
Down-regulated Genes-In general, progesterone-induced gene down-regulation was uncommon, but 12 such genes were identified (Table I) by pairwise comparison of triplicate experiments using MicroArray Suite. Additionally, gene filtering using GeneSpring TM generated a clustergram of genes regulated in this manner (Fig. 6A). Progesterone-mediated downregulation of two of these genes (highlighted in red in Fig. 6A), monocyte chemotactic protein (MCP; open arrow in Fig. 6A and isolated in Fig. 6B, top) and bullous pemphigoid antigen (BPAG; closed arrow in Fig. 6A and isolated in Fig. 6B, bottom), was confirmed by RT-PCR, particularly at early time points (Fig. 6C). This down-regulation is in contrast to the gene encoding 11␤-hydroxysteroid dehydrogenase (HSD11␤2), which is up-regulated by progesterone in the same RT-PCR experiment (Fig. 6C). ␤2MG served as a loading control.

DISCUSSION
Overview-This study, the first global examination of PR regulated genes in any system, reveals the molecular basis for functional differences between the two PR isoforms. We demonstrate that in breast cancer cells, although some genes are regulated by progesterone through both PR isoforms, most genes are uniquely regulated through one or the other isoform and predominantly through PR-B. These studies were performed in homogenous tumor cell populations allowing quantitative, statistical analyses of replicate independent experiments and straightforward interpretation of the data. This is difficult to do in organs or tumors that contain mixed cell types, without or with malignant epithelium. The results are validated by recent studies (49) that classified cell lines from various types of cancers based on their gene expression patterns, and found strong correlations among cell lines, the primary tumors from which they were derived, and the normal tissue of origin. This indicates that adaptation for growth in culture does not overwrite the gene expression programs established during tissue differentiation in vivo. Thus, our findings in T47D cells regarding progesterone-mediated gene regulation will apply to other breast cancer cells and normal progesterone target tissues. Sixteen of 94 progesterone responsive genes identified here have been reported to be progesterone regulated in other tissues and models.
Practical Applications and PR Measurements in Breast Cancer-An excess of one or the other PR isoforms may result in tumors with different prognostic and hormone-responsiveness profiles than tumors that have equimolar levels of the two PR. If so, it would be clinically important to distinguish among these tumor subsets. Current immunohistochemical clinical PR assays are incapable of doing this. In fact, it has been discovered recently that several anti-PR antibodies used clinically fail to detect PR-B by immunohistochemistry, even if they can do so by immunoblotting (50). Therefore, current clinical assays may fail to measure what may be the biologically more active PR isoform in breast cancers (PR-B) and do not distinguish between the two isoforms. It is possible that in the future judicious selection, and measurement, of progesterone-regulated genes can substitute for measurement of the receptors.
The data can also serve as the standard against which future studies of progesterone action in other cell types and tissues will be compared. This may provide explanations for differences in function of the two PR isoforms in different tissues. The preponderance of genes regulated uniquely by PR-B in breast cancer cells is surprising. Until progesterone-regulated gene expression profiles are reported in other cells and tissues, we will not know whether this dominance by PR-B is, or is not, unique to breast cancer. Recall that in PR-A knockout mice, the FIG. 4. STAT5A and C/EBP␤ are uniquely up-regulated by progesterone only through PR-B. A, cells were treated and analyzed as described in Fig.  2. The gene expression pattern of Stat5a (isolated from the clustergram in Fig. 3A) is shown to demonstrate standard error bars generated from replicate experiments. B, T47D-YA and -YB cells were treated with 10 nM progesterone (Prog) (ϩ) or ethanol vehicle (Ϫ) for 24 h. Cells were harvested, and 100 g of whole cell lysates were resolved by SDS-PAGE and immunoblotted with AB-52 antibody, which recognizes both isoforms of PR, with antibody to total Stat5, which recognizes both STAT5A (solid arrow) and Ϫ5b (open arrow), or with antibodies to cyclin D1, p21 WAF1 , or to PSTAIR. C, the gene expression pattern of C/EBP␤ isolated from the clustergram in Fig. 3A, showing standard errors. D, T47D-YA and -YB cells were treated with 10 nM progesterone (ϩ) or ethanol vehicle (Ϫ) for 24 and 48 h, harvested as described above, and immunoblotted with antibodies that recognize C/EBP␤ lap isoforms, cyclin D1, or p21 WAF1 . virgin mammary gland develops normally, but the reproductive tract exhibits hyperproliferative anomalies consistent with failure of progesterone to oppose the actions of estrogens when only PR-B are present (4). It is possible that PR-A have a more important ER repressor function in the endometrium than they do in the breast. If so, this may be reflected in a different gene expression profile for PR-A in endometrial cells. Interestingly, we observe that ERR␣1 levels are preferentially up-regulated by PR-A (Fig. 5B). Because ERR␣1 can regulate some of the same target genes as ER␣ and interfere with the functional activity of ER␣ (51), this may be a molecular mechanism by which PR-A modulate the activity of ER␣ in vivo.
Genes Regulated in the Normal Breast and Breast Cancers-We show that STAT5A, MSX-2, and C/EBP␤ are upregulated only by PR-B (Table I and Fig. 4). The PR-B-specific regulation of these three proteins, known to be critical for normal mammary gland development (43-45, 52, 53), is consistent with data demonstrating that the mammary gland develops normally in PR-A knockout mice that contain only PR-B (4).
The following genes, newly found to be progesterone-regulated (Table I), are differentially expressed in breast cancer compared with normal breast (Table III). 1) Bullous pemphigoid antigen (BPAG) is down-regulated by progesterone through both PR isoforms. The protein, associated with hemidesmosome formation, is 12-fold overexpressed in normal compared with malignant breast epithelium (54). In the normal breast it may be involved in the regulation of milk secretion (55). Expression of hemidesmosome component proteins is lost in invasive breast and other cancers (56,57). We suggest that this deleterious effect may be exacerbated by progesterone. 2) Expression of the gene encoding calcium-binding protein S100P is up-regulated by progesterone through both isoforms. Overexpression of S100P is associated with immortalization of human breast epithelial cells in vitro and with early stage breast cancer development in vivo (58). Progesterone would exacerbate this deleterious effect. 3) The gene encoding EZF, a zinc finger transcription factor, is up-regulated by progesterone through both isoforms. EZF is up-regulated during breast cancer progression (59). Progesterone would exacerbate this deleterious effect. 4) The gene encoding tissue factor, a cell surface glycoprotein, is up-regulated by progesterone uniquely through PR-B. Tissue factor is associated with metastasis in breast and other cancers (60,61) and is regulated by progesterone in the endometrium during decidualization (62)(63)(64). Its up-regulation by progesterone in breast cancers might enhance metastasis. 5) The gene encoding GAS6, a ligand for the tyrosine kinase receptor, Axl receptor tyrosine kinase, is also uniquely regulated by PR-B. GAS6 is mitogenic in breast cancer cells (65) and promotes chemotaxis of vascular smooth muscle cells (66). Its up-regulation by progesterone in breast cancers might be deleterious. 6) The anti-apoptosis gene, BCL-X L , is up-regulated FIG. 5. Genes uniquely regulated by progesterone through PR-A. A, top, the expression profile of the isolated enhancer of filamentation 1 gene (HEF-1) is shown including standard error bars for triplicate determinations. Bottom, reverse image of ethidium bromidestained RT-PCR products generated using primers specific for HEF-1 and ␤2MG from cDNA isolated from T47D-YA and -YB cells treated with 10 nM progesterone (Prog) (ϩ) or vehicle (Ϫ) for 3, 12, and 24 h. B, top, isolated gene expression pattern of estrogen-related receptor (ERR␣1). Bottom, RT-PCR performed with primers specific for ERR␣1 and ␤2MG from cDNA prepared from T47D-YA and -YB cells treated with 10 nM progesterone (ϩ) or vehicle (Ϫ) for 3, and 12 h. C, top, isolated gene expression pattern of BCL-X L . Bottom, [␣-32 P]dCTP-labeled products generated by RT-PCR performed with primers specific for BCL-X L and GAPDH from cDNA isolated from T47D-YA and YB cells treated with 10 nM progesterone (ϩ) or vehicle (Ϫ) for 3, 12, and 24 h. D, HeLa cells were transiently transfected with the following constructs from left to right: a promoter for BCL-X L linked to the luciferase reporter (pA3Luc) isolated as described under "Experimental Procedures," together with either PR-B or PR-A expressed in pSG5; PR-B or PR-A with empty pA3LucLink; empty pSG5 with the BCL-X L promoter; or empty pSG5 and no promoter, in triplicate dishes. Cells were treated with ethanol vehicle (open bars) or 10 nM progesterone (solid red bars) for 24 h and harvested. Relative luciferase activity units normalized to ␤-galactosidase expression plasmid pCH110 are shown with standard deviations for triplicate determinations. only by PR-A. Resistance to apoptosis by preferential up-regulation of BCL-X L could explain the deleterious effect of PR-A overexpression in the mammary gland of transgenic mice (31). 7) HEF1, a docking protein associated with focal adhesion kinase (67), is also preferentially up-regulated by PR-A. HEF1 is related to BCAR1/p130 Cas , which is up-regulated in tamoxifen-resistant tumors (68,69). Are tumors overexpressing PR-A more resistant to apoptosis-inducing chemotherapeutic agents or to tamoxifen? Taken together, our data raise the possibility that physiological progesterone levels are harmful in breast cancer, and may explain recent HRT data that, unlike its effect in the uterus, progesterone is not protective in the breast and indeed increases breast cancer risk (11,12).
Progesterone and the Cell Membrane-The genes that we have discovered to be progesterone-regulated are involved in particular functional pathways as shown in Table I. It was previously known, for example, that progesterone regulates proteins involved in steroid biosynthesis and trafficking pathways (70 -72), so our confirmation of this role for the hormone is not surprising. However, the extensive number of genes involved in membrane-initiated events that we define as being progesterone-regulated is surprising (Table I). These include proteins involved in cell adhesion, membrane receptors, calcium-binding proteins, and signaling molecules including genes involved in G protein signaling. Together they represent almost half of all progesterone-regulated genes. These data clearly point to the membrane as an important target of progesterone action.
Mechanisms-Most normal progesterone target cells express both PR-A and PR-B. The studies described here define the gene regulatory properties of each isoform independently. This information is critical to understanding the more complex question: how the presence of one isoform influences gene regulation by the other. Our preliminary data (not shown), using T47Dco cells that contain both receptors, suggest that presence of PR-A can suppress up-regulation of some but not all PR-Bspecific genes. For example, transcripts for GAS6 and STAT5A are up-regulated 9.3-and 6.1-fold, respectively, in PR-B containing T47D-YB cells, but their levels are unaffected by progesterone in T47Dco cells. This suggests that in the T47Dco cells, PR-A suppress the effects of PR-B on these genes. Other genes, for example C/EBP␤ and the zinc finger transcription factor, AREB6, are up-regulated in both T47D-YB and T47Dco cells. Clearly, presence of PR-A does not suppress expression of these PR-B-specific genes in T47Dco cells. The underlying mechanisms for these differences will require studies of the specific gene promoters. To that end, we are isolating key promoters. We have also generated new inducible cell lines, in which the expression of each isoform as well as the PR-A to PR-B ratio can be controlled. These cells are also being used to confirm the apparent ligand-independent PR regulation of some genes.
Interestingly, the converse may be simpler. Genes up-regulated only by PR-A (Fig. 5), such as BCL-X L and ERR␣1, are also up-regulated in T47Dco cells, suggesting that PR-B lack the inhibitory properties of PR-A. We hypothesize that genes regulated only by PR-B require the AF3 function unique to PR-B. This would further suggest that genes regulated by both PR isoforms do not require AF3. If so, there might be three subsets of progesterone-regulated genes as follows: those regulated by the PR-B homodimer, those regulated by the PR-A homodimer, and those regulated by the PR heterodimer. We are now in position to test these ideas using the inducible cell lines and mutant PR-B that lack AF3 activity (73).
Concluding Remarks-The actions of progesterone in the breast are paradoxical because the hormone has both prolifer-ative and differentiative functions therein. This is in apparent contrast to the uterus, where its actions are mainly antiproliferative. Therefore, to protect the uterus, progestins are often added to estrogens for HRT. However, this regimen is controversial, because recent evidence suggests that the progestins in HRT increase the risk of breast cancer (11,12). Given that expression of one or the other PR isoform may be more or less beneficial in certain physiological states or tumors, it would be useful to have ligands that activate or suppress one of the isoforms preferentially. By using specific subsets of the genes we have identified here, together with cell lines that express only one or the other PR isoform, one can screen large libraries of candidate progestins and antiprogestins for isoform specificity. Along those lines we can now ask how gene regulation profiles compare when the ligand is progesterone, or one of the many synthetic progestins in widespread clinical use such as medroxyprogesterone acetate.