Smad Signaling Antagonizes STAT5-mediated Gene Transcription and Mammary Epithelial Cell Differentiation*

Both the transforming growth factor-β (TGFβ)/Smad and the prolactin/JAK/STAT pathway are critical to the proper development, maintenance, and function of the mammary epithelial tissue. Interestingly, opposing physiological effects between these two signaling pathways are prominent in the regulation of mammary gland development. However, the exact nature of the biological network existing between the Smad and STAT signal transduction pathways has remained elusive. We identified a novel regulatory cross-talk mechanism by which TGFβ-induced Smad signaling acts to antagonize prolactin-mediated JAK/STAT signaling and expression of target genes. Furthermore, we found activin, another member of the TGFβ family, to also efficiently block STAT5 signaling and β-casein expression in mammary epithelial cells. Our results indicate that ligand-induced activation of Smad2, -3, and -4 by activin and TGFβ leads to a direct inhibition of STAT5 transactivation and STAT5-mediated transcription of the downstream target genes, β-casein and cyclin D1, thereby blocking vital processes for mammary gland growth and differentiation. Finally, we unveiled the mechanism by which these two signaling cascades antagonize their effects, and we found that activated Smads inhibit STAT5 association with its co-activator CREB-binding protein, thus blocking STAT5 transactivation of its target genes and leading to inhibition of mammary gland differentiation and lactation.

Mammary gland growth and differentiation are complex processes regulated by steroids, polypeptide hormones, and growth factors. Among them, prolactin and TGF␤ 5 family members play a major role in the regulation of mammary gland development. Prolactin is required for lobuloalveolar formation and functional differentiation of mammary epithelial cells.
TGF␤ has an opposite effect, inducing apoptosis during mammary gland involution and inhibiting expression of the milk proteins (1,2). TGF␤ is expressed and plays critical roles in every phase of post-natal mammary gland development (3). TGF␤ has been shown to inhibit alveolar formation and synthesis of milk proteins and to induce apoptosis during involution of the mammary gland (4 -6). Together, these data suggest that TGF␤ would antagonize prolactin (PRL)-induced signals in mammary cells (7,8). The effect of activin, another member of the TGF␤ family, on the development of the mammary gland stems from the activin ␤b subunit knock-out mouse model. Deletion of the activin ␤b subunit, through ablation of three of the dimeric ␤ molecules (activin B, activin AB, and inhibin B), results in mice with the phenotype of incomplete mammary development and an absence of lactation, suggesting that activin/inhibin may play an important role in this process (9). In summary, although TGF␤ and activin clearly play important roles in mammary gland development, their mechanism of action in mammary epithelial functional differentiation has yet to be fully elucidated.
Prolactin signal transduction is induced by formation of a homodimeric complex of two molecules of prolactin receptor (PRLR), which lack intrinsic kinase activity but are constitutively associated with the intracellular tyrosine kinase JAK2 (10 -12). Prolactin-induced receptor homodimerization brings their associated JAK2 molecules in close proximity, allowing for their transactivation as well as phosphorylation of the PRLR (13,14). These phosphotyrosine residues on both JAK2 and the PRLR then create docking sites for the recruitment and activation of the transcription factor STAT5 via its Src homology 2 domain. Once phosphorylated, STAT5 homodimerizes and translocates into the nucleus to bind response elements on promoters of target genes such as those encoding milk proteins and cell growth regulators (15,16). The importance of STAT5 in mammary gland development is further highlighted by the STAT5a knock-out mouse, which shows no lobuloalveolar development during pregnancy and a complete absence of lactation (17). The physiological role and importance of STAT5 has been subsequently extended as many cytokines, including growth hormone, erythropoietin, thrombopoietin, granulocyte-macrophage colony-stimulating factor, and most interleukins, also signal through STAT5 (18 -23).
TGF␤ and activin transduce their signals by binding to serine/threonine kinase receptors (type I and type II). Following ligand binding to the type II receptor, the type I receptor is recruited to the complex and transphosphorylated by the type II kinase domain. This in turn will lead to activation and phosphorylation of the Smad proteins, the main downstream effector molecules for these receptors. The activated type I receptor phosphorylates receptor-regulated Smads (R-Smads), Smad2 and 3, on their carboxyl-terminal serine residues (SXS motif). Phosphorylation allows for heterotrimerization of two phosphorylated R-Smad subunits with one common partner Smad4 subunit (24 -26). Subsequently, the Smad heterotrimer translocates into the nucleus where it associates with various transcription factors, co-activators or co-repressors, to regulate expression of target genes in a cell-specific manner (27).
In this study, we identified a novel regulatory cross-talk mechanism by which both activin and TGF␤, through the Smad pathway, inhibit STAT5-mediated gene transcription and mammary epithelial cell differentiation. Furthermore, these antagonistic effects are initiated by activin/TGF␤-mediated inhibition of STAT5 binding to its co-activator CBP, thereby blocking STAT5 transactivation capacity.
Cell extracts were separated on polyacrylamide gels, transferred onto nitrocellulose, and incubated with specific antibodies as indicated. Immunoreactivity was revealed by chemiluminescence (Lumi-Light Plus Western blotting substrate, Roche Diagnostics) according to the manufacturer's instructions and measured using an Alpha Innotech Fluorochem Imaging system (Packard Canberra).
Northern Blot Analysis-HC11 cells were cultured to differentiate and then treated or not with HIP, 0.5 nM activin A, or both. Total RNA was extracted using TRIzol (Invitrogen). Each sample (20 g) was then separated on agarose gels (1% agarose in 0.04 M MOPS; 0.01 M sodium acetate; 10 mM EDTA, pH 8.0; and 2.5 M formaldehyde) and transferred to nylon membranes. A probe for mouse ␤-casein was labeled using the random priming kit (Roche Diagnostics) and added to the hybridization solution (0.5 mM NaPO 4 , pH 7.2, 1 mM EDTA, pH 8.0, 7% SDS, 1% bovine serum albumin, and 200 mg/ml salmon sperm DNA). Results were revealed using a PhosphorImager Cyclone Storage Phosphorscreen (Packard Canberra).
Reverse Transcription-PCR-Differentiated HC11 cells were treated for 24 h with HIP and subsequently with or without actinomycin D, actinomycin D and 0.5 nM activin, or actinomycin D and 100 pM TGF␤ from 0 to 24 h. Total RNA was extracted using TRIzol reagents (Invitrogen). Reverse transcription of total cellular RNA using random primers was carried out using Superscript Reverse Transcriptase (Invitrogen) as per the manufacturer's instructions. Subsequently, amplification of cDNA to obtain products for ␤-casein and GAPDH was performed. The PCR conditions were as follows: 30 cycles (94°C for 30 s, 60°C for 30 s, and 72°C for 1 min 30 s). The densitometry analysis was performed using an Alpha Innotech Fluorochem Imaging system (Packard Canberra) and Fluorochem 8000 software (Alpha Innotech) that allows normalization and quantitative analysis of chemiluminescence under nonsaturating conditions.
Chromatin Immunoprecipitation Assay-ChIP was performed as described previously (32). Differentiated HC11 cells were cross-linked with 1% formaldehyde for 10 min at room temperature. The cells were subsequently collected in ice-cold 1ϫ phosphate-buffered saline and then incubated at 30°C for 10 min. Cells were then centrifuged at 2000 ϫ g for 5 min. Cells were washed first with ice-cold phosphate-buffered saline, then with buffer I (0.25 M Triton X-100, 1 mM EDTA, 0.5 mM EGTA, 10 mM HEPES, pH 6.5) and finally with buffer II (200 mM NaCl, 1 mM EDTA, 10 mM HEPES, pH 6.5), containing protease and phosphatase inhibitors. Cells were the centrifuged at 2000 ϫ g for 5 min after each wash, after which they were lysed in lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1, plus protease inhibitors and 1 mM phenylmethylsulfonyl fluoride) for 10 min on ice. Sonication was performed on ice in 7-s bursts followed by 1 min of cooling on ice for a total sonication time of 21 s per sample, which resulted in DNA fragment sizes of 0.3-1.5 kb. Samples were then centrifuged at 14,000 rpm for 10 min at 4°C. Supernatants were diluted 5-fold in ChIP dilution buffer (1% Triton X-100, 2 mM EDTA, 15 mM NaCl, 20 mM Tris-HCl, pH 8.1, plus protease and phosphatase inhibitors) and precleared for 30 min at 4°C with 20 l of preimmune serum and 80 l of salmon sperm DNA/protein A/G-agarose slurry. Ten percent of total supernatant was saved as a total input control and processed with the eluted immunoprecipitates at the crosslinking reversal step. Then 5 g of CBP (Santa Cruz Biotechnology) was added to the chromatin solutions (with no antibody control included), and samples were incubated at 4°C overnight while rotating. Immunocomplexes were collected with 60 l of the salmon sperm DNA/protein A/G-agarose beads for 1 h at 4°C with rotation. Beads were then washed consecutively for 10 min each with low salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl, 20 mM Tris-HCl, pH 8.1), high salt wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 500 mM NaCl, 20 mM Tris-HCl, pH 8.1), and LiCl wash buffer (0.25 M LiCl, 1% Nonidet P-40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.1), and twice in 1ϫ TE buffer. Complexes were then eluted twice in 100 l of freshly made elution buffer (1% SDS, 0.1 M NaHCO 3 ). To reverse formaldehyde cross-links, 1 l of 10 mg/ml RNase and 5 M NaCl to a final concentration of 0.3 M was added to the samples, and they were incubated at 68°C overnight. The next day, 4 l of 0.5 M EDTA, 8 l of 1 M Tris, pH 6.5, and 1 l of 20 mg/ml proteinase K was added, and samples were incubated for 2 h at 45°C. DNA was recovered using the QIAquick spin columns (Qiagen, Valencia, CA) and eluted in 80 l of 10 mM Tris, pH 8.0. PCR analysis was performed using 315 bp of the mouse ␤-casein promoter containing the STAT5-binding elements with the following forward primer, 5Ј-CACTTGGCTGGAGG-AACATGTAGTT-3Ј, and reverse primer, 5Ј-ACATCTGAA-GTTCTTACCTTTAGTGGAGG-3Ј.
Statistical Analysis-Results are expressed as mean Ϯ S.D. of three or more separate independent experiments. statistical analysis was assessed by one-way ANOVA or the unpaired t test, as indicated in figure legends, using GraphPad Prism 4 software (GraphPad Software, Inc.). Statistical analyses were meant to compare fold induction (percentage of control) of TGF␤/activin-treated samples among themselves within each experiment. Additional post-ANOVA tests were performed when necessary to compare all data with HIP-treated control Bonferroni's Multiple Comparison test GraphPad Prism. For all statistical analyses and tests, a p value of Ͻ0.05 was considered significant and is indicated at the top of the error bars by an asterisk.

Activin and TGF␤ Inhibit Lactogenic Hormone-induced Milk
Protein Gene Expression-The signals that regulate mammary gland development and terminal differentiation of mammary epithelial cells also coincide with those that are involved in milk protein production (33). Originally derived from a mammary epithelial cell line isolated from mid-pregnant BALB/c mouse mammary gland, HC11 cells can be differentiated with lactogenic hormone stimulation, leading to induction of milk protein synthesis (34,35). HC11 cells have often been used as a model system to study functional differentiation of the mammary gland (15). To investigate the role of activin/TGF␤ on mammary gland differentiation, we examined the effects of these two growth factors on the regulation of the expression of the milk protein ␤-casein. HC11 cells were differentiated and subsequently treated or not with either prolactin, activin, or a combination of both for different periods of time. ␤-Casein mRNA levels were assessed using Northern blot analysis. As shown in Fig. 1A, prolactin induced a strong up-regulation of ␤-casein mRNA levels at all time points analyzed, although activin alone had no effect. However, when cells were treated with both activin and prolactin simultaneously, ␤-casein mRNA levels were strongly suppressed (Fig. 1A). The same results were obtained in cells treated with TGF␤ (data not shown).
To then investigate whether the decrease in prolactin-induced ␤-casein mRNA levels upon activin treatment was followed by a decrease in ␤-casein protein levels, HC11 cells were treated or not with prolactin, activin, or a combination of both factors, and total cell lysates were analyzed by Western blot using an anti-␤-casein antiserum. As shown in Fig. 1B (upper panel), prolactin significantly increased ␤-casein protein levels, although this effect was almost completely blocked in the presence of activin. Reprobing of the membrane with an anti-Erk1/2 antibody showed equal loading of the proteins (Fig. 1B, lower panel). Previous studies have shown that optimal in vitro STAT5-mediated ␤-casein gene expression is achieved upon treatment of mammary epithelial cells with a combination of lactogenic hormones containing a glucocorticoid (such as hydrocortisone), insulin, and prolactin (36,37). As such, we treated differentiated HC11 cells with hydrocortisone, insulin, and prolactin (HIP) for 24 h, in either the presence or absence of TGF␤ or activin, and subsequently examined the effect upon ␤-casein protein expression. HIP induced a robust increase in ␤-casein protein levels, whereas treatment with both activin and TGF␤ completely inhibited its expression (Fig. 1C, upper panel). The membrane was stripped and reprobed with anti-Erk1/2 antibody showing equal loading of the proteins (Fig. 1C, lower panel). Together these results indicate that activin and TGF␤ are strong inhibitors of prolactin and HIP-mediated ␤-casein gene expression in mammary epithelial cells, suggesting the existence of a direct cross-talk mechanism between these two growth factor family signaling pathways.
JAK/STAT Expression Levels and ␤-Casein Protein Stability Are Not Affected by Activin/TGF␤-To rule out that the inhibitory activin/TGF␤ effect observed on prolactin-induced ␤-casein expression was merely because of a change in JAK or STAT expression levels, we examined the expression of JAK2 and STAT5 in HC11 cells upon activin/TGF␤ treatment from 4 to 24 h. As shown in Fig. 1D, we observed no change in their expression, thus indicating that activin/TGF␤ regulation of ␤-casein expression is not a result of changes in JAK2 or STAT5 expression levels.
Lactogenic hormone treatment can augment the rate of ␤-casein mRNA transcription 2-4-fold. However, this is not sufficient to explain the drastic increase in ␤-casein mRNA, which can rise 17-25-fold above basal levels (38). This net increase is because of direct transcriptional activation of the ␤-casein gene coupled to stabilization of ␤-casein mRNA (39).
To then examine whether activin/TGF␤ could also affect ␤-casein mRNA stability, HC11 cells were treated for 24 h with lactogenic hormones (HIP) to increase ␤-casein mRNA levels. Subsequently, cells were treated with the transcriptional inhibitor actinomycin D, and the ␤-casein mRNA levels were then examined by reverse transcription-PCR analysis. As shown in Fig. 2A (left panel), control HC11 cells treated with actinomycin D alone showed a significant decrease in ␤-casein mRNA levels after 16 h, whereas the levels remained constant at earlier time points. Activin ( Fig. 2A, middle panel) or TGF␤ (right panel) treatment of the cells did not affect ␤-casein mRNA stability, as the degradation rate was identical to that observed in the control cells (left panel). GAPDH mRNA levels were analyzed in parallel as a negative control ( Fig. 2A). Fig. 2B represents a semiquantitative densitometry analysis of three independent experiments, where ␤-casein mRNA expression level were standardized over GAPDH expression level. In conclusion, these results indicate that activin/TGF␤ does not affect ␤-casein mRNA stability but rather acts at the transcriptional level. Activin/TGF␤ Does Not Affect Lactogenic Hormone-induced Phosphorylation and Nuclear Translocation of STAT5-Following phosphorylation by the JAK2 kinase, STAT5 rapidly accumulates in the nucleus to regulate transcription of target genes. To determine whether activin/TGF␤ signaling could affect STAT5 phosphorylation and nuclear translocation, HC11 cells were treated or not with HIP in the presence or absence of activin for 0 -60 min. Cytoplasmic and nuclear fractions were purified and analyzed by Western blot analysis using a specific phospho-Stat5 antibody. As shown in Fig. 3A (upper panel), HIP rapidly induced STAT5 phosphorylation, and this effect was unaltered by activin treatment. Equal loading of the proteins was assessed by reprobing the membrane with an anti-STAT5 antibody (Fig. 3A, bottom panel). Similarly, STAT5 nuclear translocation in response to HIP treatment was not antagonized by activin (Fig. 3B). The nuclear fraction membranes were reprobed with an anti-TBP for equal loading (Fig. 3B, lower panel). Proper activin signaling in these cells was verified by reprobing of the membrane with a phospho-Smad3 antibody. As shown in Fig. 3C, activin rapidly induced Smad3 phosphorylation in a time-dependent manner. These results indicate that activin does not affect HIP-induced phosphorylation and nuclear accumulation of STAT5, suggesting that the antagonistic effect observed for activin on HIP-induced ␤-casein expression takes place at a step further downstream within the nucleus.
Activin/TGF␤ Treatment Does Not Affect STAT5 Binding to the ␤-Casein Gene Promoter-␤-Casein gene expression is mediated through direct binding of STAT5 to the ␤-casein gene promoter. Using EMSA, we then investigated whether activin/ TGF␤ could interfere with STAT5 DNA binding. HC11 cells were treated from 0 to 120 min with HIP, activin, or both; the nuclear fractions were isolated, and EMSA was performed using the STAT5-binding site (TTCNNNGAA) of the ␤-casein promoter as a probe. As illustrated in Fig. 4A, HIP induced a very rapid and transient DNA-protein complex formation, although activin alone had no effect.
Interestingly, co-treatment of the cells with HIP and activin did not affect HIP-induced STAT5 DNA binding. The same results were observed in response to TGF␤ (Fig. 4B). The presence of STAT5 in the HIP-induced DNA-protein complex was confirmed by a supershift experiment using a polyclonal antibody against STAT5. As shown in Fig. 4B (5th and 6th lanes), addition of anti-STAT5a resulted in a supershift of the complex in HIP-treated extracts. Together, these results indicate that the decrease of ␤-casein mRNA and protein levels by activin/ TGF␤ is not because of an inhibition of STAT5 binding to the ␤-casein gene promoter. Activin/TGF␤ Inhibits Activation of the ␤-Casein Gene Promoter in Response to Lactogenic Hormones-In trying to understand the mechanism by which activin/TGF␤ act to inhibit ␤-casein synthesis, we investigated their effects on the ␤-casein gene promoter. For this, HC11 cells stably expressing the ␤-casein gene promoter fused to the luciferase gene were treated from 4 to 24 h with either lactogenic hormones, activin, TGF␤, or a combination of both. As shown in Fig. 5, whereas HIP increased ␤-casein luciferase activity in a time-dependent man-ner, this effect was strongly inhibited by activin and TGF␤ (Fig.  5). Activin and TGF␤ treatments alone did not affect ␤-casein luciferase activity (Fig. 5). These results demonstrate that activin and TGF␤ robustly block lactogenic hormone induction of the ␤-casein gene promoter activity.
STAT5-mediated Transcription Is Blocked by Activin/TGF␤ Treatment-The ␤-casein gene promoter construct encompasses the proximal part of the promoter that includes STAT5binding sites, but also contains regulatory regions for other transcription factors such as the CCAAT enhancer-binding protein C/EBP␤, Oct1 (40 -42). To investigate whether the changes in ␤-casein expression are a result of activin/TGF␤-induced alterations specifically in STAT5-driven activation of the ␤-casein gene promoter, we made an artificial promoter containing five tandem . STAT5 phosphorylation and translocation to the nucleus is unaffected by activin/TGF␤ signaling. A and B, HC11 cells were treated with HIP or HIP/activin from 0 -60 min and cytoplasmic and nuclear fractions were isolated. Western blots were performed using specific antibodies against phosphoSTAT5, STAT5, and TBP. C, HC11 cells were treated for the indicated time periods with activin or HIP/activin. Nuclear fractions were analyzed by Western blot using specific antibodies against phospho-Smad3 and TBP.
EMSA Free probe FIGURE 4. Activin/TGF␤ signaling do not inhibit STAT5 binding to the STAT5 binding element on the ␤-casein promoter. A, nuclear extracts from HC11 cells were isolated after treatment with HIP, HIP/activin or activin from 0 to 120 min. Electromobility shift assays were performed, using a 21 base pairs double stranded oligonucleotide encompassing the STAT5-binding site of the bovine ␤-casein promoter as a probe. B, EMSA on nuclear extracts treated or not with HIP, HIP/TGF␤ or TGF␤ for 20 min were performed. Supershifts were performed with a polyclonal antibody against STAT5a (lanes 5 and 6).
repeats of the STAT5-response elements fused to the luciferase gene (5XSTAT5-luc) and generated HC11 stable cell lines expressing this 5XSTAT5-luc (HC11-5XSTAT5-luc). As shown in Fig. 6A, the three positive clones that were selected all responded strongly to HIP treatment, as measured by luciferase activity. Interestingly, although TGF␤ alone had no effect on this reporter construct, it strongly inhibited HIP-induced 5XSTAT5 luciferase activity, demonstrating that TGF␤ antagonizes STAT5mediated transcriptional activity (Fig. 6A).
To further extend our results to human mammary epithelial cells, we transfected MCF10A cells with the 5XSTAT5-luc reporter construct and examined the TGF␤ response on HIP-induced STAT5 transactivation. As shown in Fig. 6B, HIP treatment induced a strong increase of 5X-STAT5-luc promoter activity, and this effect was also robustly antagonized by TGF␤. Together, these results indicate that TGF␤ signaling directly impedes HIP-induced STAT5 activation in both mouse and human mammary epithelial cells, thereby leading to inhibition of milk gene protein expression.
Activin/TGF␤-mediated Inhibition of STAT5 Signaling Requires Smad Proteins-To address the molecular mechanisms by which activin/TGF␤ antagonizes STAT5 signaling, we investigated the role of the main regulatory pathway downstream of activin/TGF␤, the Smad pathway. For this, CHO cells were co-transfected with the ␤-casein luciferase reporter construct, cDNAs encoding STAT5 and PRLR, in the presence or absence of varying combinations of cDNAs encoding Smad2, Smad3, and Smad4. Cells were then stimulated overnight with HIP or TGF␤/activin alone or in combination. As shown in Fig.  7A, both TGF␤ and activin significantly repressed HIP-induced ␤-casein (top left and top right panels, respectively) and 5XSTAT5 (bottom left and bottom right panels) luciferase activity. Interestingly, overexpression of Smad3 alone or in combination with Smad4 significantly potentiated the TGF␤ and activin's inhibitory effect on HIP-induced ␤-casein and 5xSTAT5 luciferase activities (Fig. 7B). Similar results were obtained in response to Smad2 overexpression (Fig. 7C). Together, these data indicate that  Having demonstrated that overexpression of the Smads could potentiate activin/TGF␤-mediated inhibition of prolactin/STAT5 ␤-casein gene promoter activation, we next examined the effect of blocking Smad signaling on the activin/TGF␤ response. For this we used dominant negative Smad2 and dominant negative Smad3, in which the carboxylterminal serine residues were mutated to alanine and hence cannot be phosphorylated by the receptor (28). Likewise, we also used a mutant form of the activin type I receptor (ALK4mL45) in which the three critical residues for interaction with Smad2 and Smad3 (Asn-265, Asp-267, and Asn-268 within the Leu-45 loop of the receptor) were mutated to alanine. The resulting mutant receptor failed to recruit the Smads and was unable to mediate activin signaling (28).
The efficiency of the DNSmads and the mutant receptor (ALK4mL45) was first assessed using the activin/ TGF␤-responsive promoter construct, 3TPluc. As shown in Fig. 7D, DNSmad2, DNSmad3, and the mutant receptor (ALK4mL45) all strongly inhibited activin-induced 3TPluc activity. We then assessed the effect of overexpressing DNS-mad2, DNSmad3, or ALK4mL45 in CHO cells transfected with cDNAs encoding PRLR, and STAT5, along with ␤-casein or 5XSTAT5-luc reporter constructs. Although in the control transfected cells HIP/activin treatment led to inhibition of ␤-casein-luc and 5XSTAT5-luc luciferase activity, overexpression of DNSmad2, DNSmad3, and ALK4mL45 completely reversed these effects, clearly highlighting the important role played by the Smads in mediating the inhibitory effect of activin/TGF␤ on STAT5driven promoters (Fig. 7E).
Smad signaling in response to activin/TGF␤ is centrally controlled by the common partner Smad4. Hence to further demonstrate the   (43). Inactivation of Smad4 expression has been shown to prevent activin/TGF␤ signaling and target gene transcriptional activation (43,44). Wild type and Smad4 null mutant MEFs were co-transfected with either ␤-casein promoter or 5XSTAT5 promoter, STAT5, and PRLR. Subsequently cells were treated or not with HIP, TGF␤, or HIP/TGF␤. In the wild type MEFs, HIP-induced ␤-casein promoter and 5XSTAT5 luciferase activity were antagonized by TGF␤. This effect was completely abolished in the Smad4-deficient MEFs (Fig. 7F). Restoring Smad4 expression in the Smad4 null mutant MEFs restored the TGF␤ inhibitory effect on HIP-induced ␤-casein gene expression and STAT5 activation, clearly indicating the importance of the Smad pathway in repressing STAT5-mediated gene activation (Fig. 7F).
To further examine the effect of TGF␤-induced Smad signaling on the inhibition of STAT5 promoter activity, we used specific siRNAs to block Smad2 and Smad3 expression in HC11 mammary epithelial cells. As shown in Fig. 7G (top panels), the TGF␤ inhibitory effect on ␤-casein and 5XSTAT5 gene promoter activity was reversed when expression of endogenous Smad2 and Smad3 was blocked with the specific siRNAs. The efficiency of the siRNAs was verified by Western blotting (Fig.  7G, lower panels). To then determine whether blocking endogenous Smad2 and Smad3 expression could affect TGF␤-mediated inhibition ␤-casein protein levels, HC11 cells were transfected with the different siRNAs against Smad2 and Smad3 or a scrambled sequence as control, and the cells were then treated or not with HIP alone or a combination of HIP and TGF␤. As shown in Fig. 7H, although TGF␤ strongly inhibited HIP-induced endogenous ␤-casein expression levels in HC11 cells, this effect was reversed when Smad2 or Smad3 expression levels were knocked down using the siRNAs. Combined, these results highlight the central and critical role played by Smad signaling in mediating the   The Transactivation Capacity of STAT5 Is Potently Repressed by the Smads-The transactivation domain of STATs is essential to drive transcription (45). To evaluate the involvement of activin/TGF␤ signaling via the Smads on STAT5 transactivation ability, we co-transfected CHO cells having a GAL4-responsive promoter fused to the luciferase gene (pXPAL7) with a chimeric construct containing both the GAL4 DNA binding domain and the carboxyl-terminal STAT5 transactivation domain (GAL4-STAT5a) (46). The cells were also transfected with or without increasing concentrations of Smad3/4. As shown in Fig. 8, TGF␤ induced a 27% decrease in GAL4-STAT5a-induced luciferase activity. Increasing concentrations of Smad proteins led to a parallel increased inhibition of the luciferase activity by TGF␤, reaching a 55% repression at the highest Smad concentration. This finding demonstrates that Smad signaling efficiently blocks the transactivation ability of STAT5, leading to inhibition of milk gene protein production.
STAT5 Target Gene Cyclin D1 Is Inhibited by Activin/TGF␤-In the mammary gland, STAT5 regulates milk gene protein production and cell differentiation but also controls expression of genes involved in cell cycle progression and survival. Interestingly, prolactin/STAT5 and activin/TGF␤/Smad signaling have opposite effects on mammary epithelial cell growth and survival. Having shown that activin/TGF␤-induced Smad signaling inhibits STAT5 transactivation and ␤-casein expression, we next investigated whether Smad signaling could also antagonize STAT5-mediated gene expression of other target genes, particularly those involved in STAT5-induced cell growth, such as cyclin D1. Indeed, prolactin signaling via STAT5 has been shown to induce cyclin D1 proximal promoter and increase cyclin D1 protein expression in the mammary epithelial cells PRE-1 (16). To investigate whether the lactogenic stimulation (HIP) could also induce an increase in cyclin D1 protein expression in HC11 cells, we treated differentiated HC11 cells from 0 to 8 h with HIP. As shown in Fig. 9A, HIP induced a rapid and transient increase in cyclin D1 protein levels, peaking at 4 h and returning back to basal levels at 8 h. The membrane was then probed with anti-␤-tubulin as a loading control. We then sought to evaluate the effect of activin and TGF␤ on HIP-induced cyclin D1 protein levels. Differentiated HC11 cells were treated for 2 h with or without HIP, TGF␤, or activin or combinations of HIP/ TGF␤ and HIP/activin. As shown in Fig. 9B, HIP-induced increase in cyclin D1 protein, was completely blocked when cells were co-treated with either TGF␤ (Fig. 9B, left   panel) or activin (Fig. 9B, right panel). To then evaluate whether the TGF␤ antagonistic effect was mediated at the transcriptional level, we used the proximal cyclin D1 promoter fused to luciferase (cyclinD1-944), which has previously been shown to be activated by prolactin in a STAT5-dependent manner (16). CHO cells were transfected with the cyclinD1-944 reporter construct, STAT5, PRLR, and ␤-galactosidase cDNAs and stimulated overnight with HIP, TGF␤, or both. As shown in Fig.  9C, HIP treatment induced cyclin D1 gene promoter activity, and this effect was significantly reversed in the presence of TGF␤. Finally, to evaluate the role of the Smad pathway in antagonizing STAT5-mediated cyclin D1 gene promoter activation, CHO cells were transiently co-transfected with cyclinD1-944, STAT5, PRLR, ␤gal with or without a combination of Smad2, Smad3 and Smad4. As shown in Fig. 9D, in the absence of overexpressed Smad proteins TGF␤ induced a 20% reversal effect of HIP-induced cyclin D1 gene promoter activity. Interestingly, overexpression of the Smad proteins resulted in a clear potentiation of the TGF␤ antagonistic effect to ϳ50% reversal of HIP-induced cyclin D1 gene promoter activity (Fig. 9D). These results demonstrate that activin/TGF␤-inhibition of STAT5induced cyclin D1 protein expression requires the Smad pathway. Altogether, these findings reveal that another STAT5 target gene, cyclin D1, activation is repressed by both activin and TGF␤ mediated Smad signaling.
Activin/TGF␤ Signaling Reduces STAT5 Interaction with CBP-Next, we wanted to elucidate the mechanism by which activin/TGF␤-mediated Smad signaling impedes STAT5 transactivation of its target genes. It has been shown previously that the co-activator CBP/p300 physically interacts with STAT5 to regulate STAT5-mediated gene expression (47). In fact, CBP/ p300 binds the transactivation domain of various STAT family members and is essential for their activation (48 -51). Henceforth, to determine whether activin/TGF␤ regulates STAT5 interaction with CBP, CHO cells were transfected with PRLR, JAK2, STAT5, and HA-CBP. Cells were subsequently treated with or without HIP, TGF␤, or both before being lysed and immunoprecipitated with an anti-HA antibody, affinity-conjugated to matrix beads, and revealed by Western blot analysis with the anti-STAT5 antibody. As shown in Fig. 10, HIP treatment of the cells induced increased association between HA-CBP and STAT5. However, when the cells were treated with TGF␤, this interaction was significantly diminished. Equal protein levels were ensured by reprobing membranes with anti-HA and anti-STAT5 antibodies (Fig. 10A). Furthermore, total cell lysates were probed with anti-phospho-Stat5 and anti-phospho-Smad3 to ensure proper ligand stimulation (Fig. 10A). These results indicate that activin/TGF␤ inhibits STAT5 interaction with the co-activator CBP. To investigate the involvement of the Smad pathway in activin/TGF␤ negative regulation of STAT5-CBP interaction, 293-LA cells were transfected with PRLR, STAT5, and HA-CBP with or without Smad3 and Smad4. 293-LA cells, stably expressing the JAK2 kinase, were used in these experiments to limit the number of transfected cDNAs (13). Cells were then stimulated with or without HIP. When Smad3/4 was overexpressed, STAT5-CBP association was completely abolished (Fig. 10B). To evaluate the effect of activin signaling on STAT5 and CBP interaction, cells were transfected with or without a constitutively active activin type I receptor, Alk4T⌬D. This constitutively active receptor allows for continuous signaling and mimics the activin effects (52). As shown in Fig. 10C, in the presence of Alk4T⌬D, STAT5 no longer interacted with CBP upon HIP treatment. The results observed in Fig. 10, B and C, denote the importance of both activin and TGF␤ signaling pathway in negatively regulating STAT5 binding to CBP.
To explore whether TGF␤ affected CBP recruitment to the ␤-casein gene promoter in vivo, ChIP assays were performed. Briefly, HC11 cells were treated or not with HIP, TGF␤, or both, and cell lysates were sonicated and cross-linked before being immunoprecipitated with an anti-CBP antibody. Following elution of the DNA-protein complexes and reversal of the crosslinking, amplification of the ␤-casein gene promoter was performed by PCR. As shown in Fig. 10D, upon lactogenic hormone treatment of HC11 cells, CBP is recruited to the ␤-casein gene promoter. However, when cells are stimulated with the combination of HIP and TGF␤, the interaction between CBP and the ␤-casein promoter is no longer observed. In Fig.  10D, the lower band observed in both gels is probably because of the excess of primers. These results indicate that CBP binding to the ␤-casein promoter is negatively regulated by TGF␤ signaling.
The above results suggest that the intracellular pool of CBP is limiting, and there exists a competition for CBP between Smads and STAT5. To confirm this hypothesis, we wanted to determine whether transfection of CBP abolishes the effects of activin and TGF␤ on lactogenic hormone-induced STAT5 signaling. For this, HC11-␤-casein-luc cells transfected (CBP) or not (mock) with the cDNA encoding CBP were treated with HIP, TGF␤, HIP/TGF␤, activin, and HIP/activin. As shown in Fig. 10E (left panel), in mock-transfected cells, both TGF␤ and activin strongly antagonized HIP-induced ␤-casein gene promoter activity. However, the TGF␤ and activin inhibitory effects were reversed in the presence of overexpressed CBP (Fig. 10E, right panel), thus supporting the hypothesis that CBP is a limiting factor in these cells and that STAT and Smad signaling pathways compete for CBP. All together, these findings suggest that activin/TGF␤ signaling may regulate STAT5 gene activation by inhibiting its binding to the co-activator CBP and hence blocking CBP binding to its target promoter.
Lactogenic Hormones Inhibit TGF␤ Signaling-TGF␤-mediated inhibition of STAT5 and CBP interaction suggests that the Smads may compete with STAT5 for CBP/p300 binding, thereby inhibiting STAT5 transactivation. Moreover, previous work reported that TGF␤ signaling is suppressed by sequestration of p300 or CBP by other transcription factors, including STATs (53). Hence, we tested whether a reciprocal effect of lactogenic hormones on activin/TGF␤ signaling could be observed as a result of a competition for p300/CBP between Smads and STATs. For this, we transfected CHO and MEF ϩ/ϩ cells with the well documented activin/TGF␤-responsive promoter 3TPluc and stimulated the cells with or without TGF␤, HIP, or both (Fig. 11). Interestingly, whereas TGF␤ treatment led to 3TPluc activation, HIP led to the repression of TGF␤-induced promoter activity in two different cell lines. Similar effects were seen with activin (data not shown). These results imply that sequestration of the intracellular pool of CBP may be the major mechanism by which these two major signaling pathways antagonize each other's effects.

DISCUSSION
Mammary gland growth and differentiation is regulated by an assembly of signaling networks in response to various hormones and growth factors. Alterations within these signaling cascades represent underlying causes of a variety of human diseases, such as breast cancer. Two critical signaling cascades within the mammary epithelial tissue are the TGF␤/Smad signaling cascade and the JAK/STAT pathway. These two pathways oppose their physiological effects and are eminent in the regulation of mammary gland development and breast carcinogenesis. However, the exact nature of the biological network existing between the Smad and STAT signal transduction pathways remains elusive. Our results underline a novel cross-talk mechanism in which signal transduction by TGF␤ family members, including activin and TGF␤ itself, inhibit STAT5-regulated gene expression. These antagonistic effects take place in the nucleus, and although activin/TGF␤-mediated Smad signaling does not prevent STAT5 phosphorylation, translocation, and DNA binding, it blocks STAT5 association with the co-activator CBP, thereby inhibiting STAT5 transactivation of its target genes.
Cross-talk mechanisms between distinct signaling pathways are key in providing an integrated response and regulating homeostasis. A previous study indicated that interferon-␥ signaling through the JAK/STAT pathway could down-regulate TGF␤ signaling (54). In fact, the existence of a direct effect of the JAK/STAT pathway toward TGF␤ signaling is further supported by the fact that prolactin inhibits TGF␤-induced apoptosis in primary mammary epithelial cells (55). Inhibition of TGF␤-induced apoptosis is carried out by induction of the kinase Akt in a phosphatidylinositol 3-kinase-dependent manner (55). However, to date, the reciprocal effect of the TGF␤/ Smad signaling on STAT signaling has remained inconclusive. For instance, in mouse lymphocytes, TGF␤ immunosuppressive effects are partly mediated through the negative inhibition of IL-2-and IL-12-induced phosphorylation of JAK/STAT proteins (56,57). However, in human T cells and natural killer cells, TGF␤ had no effect on IL-2-and IL-12-induced JAK/STAT phosphorylation (58).
All three isoforms of TGF␤ are up-regulated during mammary gland involution (1,59,60). In addition, mice overexpressing the whey-acidic protein promoter-driven TGF␤ displayed increased apoptosis in pregnant and lactating mammary glands associated with decreased lobuloaveoli formation and lactation (32,33,61). Our findings indicate that these effects of TGF␤/Smad signaling pathway are mediated by tight regulation of the JAK/STAT signaling cascade through a direct crosstalk mechanism leading to inhibition of STAT5 transactivation of its target genes.
Our results also show that JAK/STAT-induced cyclin D1 expression is blocked by activin/TGF␤ signaling. Cyclin D1 is an important regulator of lobuloalveolar development, as illustrated by the phenotype of the cyclin D1 knock-out mice that fail to undergo proper lobuloalveolar formation during pregnancy (61), a phenotype shared by PRL and PRLR knock-out mice (62,63). This is consistent with the fact PRL regulates cyclin D1 gene promoter through STAT5 signaling (64). The antagonistic effect of activin/TGF␤ on lactogenic hormone induction of cyclin D1 implicates that they play a regulatory role in mammary gland development and allows for the differentiation phase of late pregnancy to occur by regulating proliferation of the lobuloalveoli. Moreover, activin/TGF␤ Smad signaling inhibition on cyclin D1 suggests that the effect of activin/TGF␤ Smad signaling is broad, potentially encompassing all STAT5 target genes and biological systems.
We observed that activin and TGF␤ signaling significantly reduce STAT5 association with the co-activator CBP. Interestingly, CBP/p300 can bind Smad2, Smad3, and Smad4 and contribute to their full activation (65)(66)(67). As we seen in Figs. 10 and 11, our data suggest Smad signaling inhibits STAT5 transactivation by competing with STAT5 for CBP/p300 binding. Such a mechanism is reminiscent of the STAT1␣ competition with Smad3 for CBP/p300 binding. STAT1␣ binding to CBP/p300 allows interferon-␥ to block TGF␤-induced transcription of type I collagen by inhibiting Smad activities (53). The competition model by which the Smads sequester CBP/ p300 away from STAT5 can potentially be applied to other STAT molecules where CBP/p300 binding is essential for their activation.
The STAT pathway is prominent in the regulation of mammary gland development. Our study indicates that the intracellular mediators of activin/TGF␤ signaling, the Smad proteins, strongly inhibit STAT5-regulated gene induction by the inhibition of STAT5 transactivation. Interestingly, STAT5 is known to contribute to mammary epithelial cell proliferation (68). Moreover, constitutively active STAT5 expression in the mammary gland leads to an increase in cellular proliferation during pregnancy in mice, accompanied by shrunken alveoli during late pregnancy and early lactation (69). The phenotype of the forced activation of STAT5 in mammary gland of transgenic mice underlines the importance and critical role played by the inhibitory pathways that regulate STAT5 activity in the mammary gland in order to prevent the development of abnormal phenotypes in that tissue. Furthermore, this also emphasizes the significance of activin and TGF␤ signaling in maintaining proper STAT5 gene activation to prevent overproliferation of the mammary epithelium and premature lactation. In future studies, it will be interesting to examine whether STAT5 is hyperactivated or whether the STAT5/CBP interaction is affected in mouse models, such as the activin ␤b and TGF␤ knock-out mice or a Smad knock-out mice model, where Smad gene expression is specifically deleted in the mammary tissue.
Our findings unveiled strong antagonistic cross-talk mechanism between the Smad and STAT signaling pathways in regulating mammary gland growth and morphogenesis. The potent tumor suppressor and growth inhibitory effects of the activin/ TGF␤ pathway are well established in mammary epithelial and breast cancer cells (70, 71). TGF␤ displays a dual role in cancer where, as the tumor progresses, TGF␤ tumor-suppressive effects switch to tumor promotion, cell invasion, and metastasis (54). Interestingly, prolactin also exhibit such a dual role, as in later stages of tumor development, whereas TGF␤ promotes metastasis, prolactin acts as an invasion suppressor hormone (72). Thus, our results showing that these two signaling cascades oppose their effects are consistent with these observations and suggest that the antagonistic cross-talk mechanism existing between the two pathways not only regulates differentiation of mammary epithelial cells and lactation but may also affect tumor formation and breast cancer metastasis.