Inhibition of Peroxisome Proliferator-Activated Receptor γ Increases Estrogen Receptor–Dependent Tumor Specification Free

Peroxisome proliferator-activated receptor γ (PPARγ) is a nuclear receptor that regulates gene transcription associated with intermediary metabolism, adipocyte differentiation, and tumor suppression and proliferation. To understand the role of PPARγ in tumorigenesis, transgenic mice were generated with mammary gland–directed expression of the dominant-negative transgene Pax8PPARγ. Transgenic mice were phenotypically indistinguishable from wild-type (WT) mice, but mammary epithelial cells expressed a greater percentage of CD29^hi/CD24^neg, CK5⁺, and double-positive CK14/CK18 cells. These changes correlated with reduced PTEN and increased Ras and extracellular signal-regulated kinase (ERK) and AKT activation. Although spontaneous tumorigenesis did not occur, transgenic animals were highly susceptible to progestin/7,12-dimethylbenz(a)anthracene–induced mammary carcinogenesis, which in contrast to WT mice resulted in a high tumor multiplicity and, most importantly, in the appearance of predominantly estrogen receptor α–positive (ER⁺) ductal adenocarcinomas. Tumors expressed a similar PTEN^lo/pERK^hi/pAKT^hi phenotype as mammary epithelium and exhibited high activation of estrogen response element–dependent reporter gene activity. Tumorigenesis in MMTV-Pax8PPARγ mice was insensitive to the chemopreventive effect of a PPARγ agonist but was profoundly inhibited by the ER antagonist fulvestrant. These results reveal important new insights into the previously unrecognized role of PPARγ in the specification of mammary lineage and the development of ER⁺ tumors. [Cancer Res 2009;69(2):687–94]

The peroxisome proliferation-activated receptor (PPAR) nuclear receptor subfamily consists of the PPARα, PPARγ, and PPARδ/β isotypes, which regulate multiple metabolic pathways controlling fatty acid β-oxidation, glucose use, cholesterol transport, energy balance, and adipocyte differentiation (1–3). These and other receptor-mediated functions pertain to their use as targets for chemopreventive agents (2, 4–6). PPARγ agonists inhibit 7,12-dimethylbenz(a)anthracene (DMBA)–induced preneoplastic lesions in the mammary gland (7) and delay the development of mammary carcinogenesis (8–10). Chemopreventive activity correlates, in part, with the ability of PPARγ to increase transcription of tumor suppressor genes, such as PTEN (11) and BRCA1 (12), through their respective PPAR-responsive element (PPRE) promoter regions. In contrast, PPARγ haploinsufficiency increases mammary carcinogenesis (13), and the dominant-negative Pax8PPARγ fusion protein (14, 15) increases proliferation and transformation and reduces expression of the Ras tumor suppressor gene NORE1A in thyroid tissue (16, 17).

These findings led us to examine the function of PPARγ in the mammary gland and in mammary carcinogenesis by generating transgenic mice that expressed Pax8PPARγ (14, 18). This animal model has lent itself to determining not only whether PPARγ agonists act directly or indirectly on mammary epithelium but also whether negative regulation of PPARγ affects cell lineage within the normal gland and developing tumors (19). These studies have revealed a previously unknown linkage between PPARγ signaling, stem and progenitor cell expansion, and tumor lineage specification pertaining to the induction of estrogen receptor (ER) expression that may have significant implications for the treatment of breast cancer.

Transgenic mice. Transgenic MMTV-Pax8PPARγ mice were generated by pronuclear injection into FVB mouse embryos using standard techniques by the Transgenic Shared Resource, Lombardi Comprehensive Cancer Center (LCCC). The Pax8PPARγ cDNA (14) was provided by Dr. Todd Kroll (University of Chicago, Chicago, IL) and subcloned into the EcoRI site in plasmid MMTV-SV40-Bssk provided by Dr. William Muller (McMaster University, Hamilton, Ontario, Canada). The MMTV-Pax8PPARγ construct was digested with SalI and SpeI, purified, and used for microinjection. Founder lines were confirmed by Southern blot and Western analysis (see Supplementary Fig. S1), two independent founders of similar phenotype were generated, and one was used for subsequent studies.

Whole-mount preparation. Female mice were euthanized using carbon dioxide, and mammary glands were harvested, placed on a dry silanized glass slide, and fixed overnight in 1 part glacial acetic acid:3 parts 100% ethanol. Tissues were rehydrated through successive incubation with 70% ethanol followed by distilled water, and stained with Carmine Red alum overnight. Tissues were then dehydrated by successive incubation in graded ethanol followed by mixed xylenes, and mounted in Permount (Fisher Scientific; ref. 20).

Involution studies. On the day of weaning, nursing females were sacrificed and mammary glands were harvested for whole-mount preparation on days 0, 3, 7, and 10 following induced involution by the teat sealing procedure (20).

Mammary carcinogenesis. Mammary carcinogenesis was initiated in MMTV-Pax8PPARγ and wild-type (WT) mice with medroxyprogesterone (Depo-Provera) and DMBA as described previously (8, 21). GW7845 (provided by GlaxoSmithKline) was administered in Purina rodent chow 5001 at a concentration of 0.005% (w/w) immediately after the last dose of DMBA (8, 21).

Primary cell culture. Mammary glands or mammary tumors were excised and minced into 1-mm pieces under sterile conditions. Tissue was digested in DMEM/F12 medium supplemented with 5% fetal bovine serum (FBS), 100 units/mL collagenase I, 100 units/mL hyaluronidase, 100 units/mL penicillin, 100 units/mL streptomycin, 10 μg/mL gentamicin, and 2.5 μg/mL amphotericin B. After incubation at 37°C for 16 h, tissue was harvested and centrifuged at 1,000 rpm for 5 min. The supernatant was discarded and the cell pellet was washed twice with PBS and resuspended in DMEM/F12 medium supplemented with 5% FBS, 10 ng/mL epidermal growth factor (EGF; Upstate Biotechnology), 5 μg/mL insulin (Biofluids), 100 units/mL penicillin, 100 units/mL streptomycin, 10 μg/mL gentamicin, and 2.5 μg/mL amphotericin B and incubated at 37°C under 5% CO₂. The cell culture was trypsinized for 1 to 2 min in 0.05% trypsin-EDTA (Biosource) every 3 d to remove fibroblast cells. Confluent cells were harvested between days 7 and 10 for further analysis. Mammary tumor cell lines MC and 437T were derived from medroxyprogesterone/DMBA-induced mammary adenocarcinomas formed in WT FVB and MMTV-Pax8PPARγ mice, respectively, and have maintained a stable phenotype for more than 30 passages.

Fluorescence-activated cell sorting analysis. Cells were prepared as single-cell suspensions, washed twice with PBS, and blocked for 15 min with PBS supplemented with 3% FBS. Cells were incubated for 1 h at room temperature with the following antibodies: FITC-Sca-1 (1:200 dilution, clone E13-161.7; BD Biosciences), PE-CD24 (1:200; BD Biosciences), biotinylated CD29 (1:200; AbD Serotec), FITC-CD133 (1:200; BD Biosciences), and FITC-CD49f (1:200; BD Biosciences). PE-Cy5–conjugated streptavidin (eBioscience) was used as secondary antibody for biotinylated CD29 at 1:200 dilution. IgG of the same isotype was used as a control. Cells were analyzed using a FACSort (BD Biosciences) and analyzed by FACExpress De Novo software by the Flow Cytometry/Cell Sorting Shared Resource, LCCC.

Western blot analysis. The cell pellet or pulverized tissue frozen in liquid nitrogen was mixed with lysis buffer containing 0.1% SDS, 0.5% NP40, phenylmethylsulfonyl fluoride, 1 mmol/L sodium vanadate, 50 mmol/L NaF, 10 mmol/L β-glycerophosphate, 5 mmol/L sodium pyrophosphate, and a protease inhibitor cocktail (Roche; ref. 22). Following incubation on ice for 30 min, lysates were cleared by centrifugation at 13,000 × g at 4°C for 15 min. Protein concentration was determined by the bicinchoninic acid protein assay (Pierce), and 50 μg of lysate were separated in a 4% to 12% NuPAGE Bis-Tris gel (Invitrogen). After wet transfer, membranes were blocked for 1 h at room temperature in 5% nonfat dry milk in TBS (pH 7.4) containing 0.1% Tween 20. Primary antibody was incubated for either 1.5 h at room temperature or overnight at 4°C. Secondary antibody was incubated for 1 h at room temperature, and proteins were visualized with either SuperSignal West Pico or SuperSignal West Dura (Pierce). The following antibodies and their dilution were used: PPARγ and PTEN (1:300; Santa Cruz Biotechnology); phosphorylated β-catenin (pβ-catenin), β-catenin, glycogen synthase kinase 3β (GSK3β), phosphorylated GSK3β (pGSK3β), phosphorylated extracellular signal-regulated kinase 1/2 (pERK1/2), pT308AKT, and AKT (1:1,000; Cell Signaling); and β-actin (1:5,000; Sigma-Aldrich).

Real-time PCR analysis. RNA was extracted from either mammary glands or cell cultures using Trizol reagent (Invitrogen). For reverse transcription, 2 μg total RNA was transcribed with SuperScript II reverse transcriptase (Invitrogen). Equal amounts of cDNA were used for the PCR under the following conditions: 94°C for 30 s, 56°C for 30 s, and 72°C for 1 min for 30 cycles. For ER mRNA, 38 cycles of amplification were necessary to measure levels in WT mammary gland and MC tumor cells. The following primers were used for mRNA analysis: ERα, 5′-TGATCAACTGGGCAAAGA-3′ (forward) and 5′-CAGGAGCAGGTCATAGAG-3′ (reverse); Pax8PPARγ, 5′-AACCTCTCGACTCACCAG-3′ (forward) and 5′-GATGGCATTATGAGACATCCC-3′ (reverse); and β-actin, 5′-AGAGGGAAATCGTGCGTGAC-3′ (forward) and 5′-CAATAGTGATGACCTGGCCGT-3′ (reverse). Samples were separated in a 1% agarose gel.

Gene microarray. RNA was prepared from MC and 437T tumor cells using Trizol and processed for array analysis as previously described (8, 21). Hybridization was carried out by the Macromolecular Analysis Shared Resource, LCCC, with an Affymetrix Mouse Genome 430A 2.0 GeneChip representing 14,000 well-annotated genes. Gene array analysis was evaluated by comparing differences between paired samples and ranking changes by their log₂ ratio. Only differences in signal ratio log₂ >3.0 and log₂ <−3.0 were ranked and are listed in Supplementary Table S1. Each cRNA was prepared from equal amounts of RNA pooled from three samples.

Quantitative real-time PCR. Total RNA was extracted using an RNeasy Mini kit (Qiagen) according to the manufacturer's protocol. Genomic DNA was digested by incubation with RNase-free DNase for 15 min at room temperature. RNA (1 μg) was reverse transcribed in a total volume of 20 μL using the Omniscript RT kit (Qiagen). PCR was performed in triplicate in an ABI Prism 7700 instrument (Applied Biosystems) using SYBR Green I detection (Qiagen) according to the manufacturer's protocol (8, 21). Amplification using the appropriate primers (see Supplementary Table S2) was confirmed by ethidium bromide staining of the PCR products on an agarose gel. The expression of each target gene was normalized to the expression of 18S RNA and is presented as the ratio of the target gene to 18S RNA calculated by 2^−ΔCt, where ΔC_t = C_t^Target-C_t^18s (see Supplementary Fig. S2).

Immunohistochemical staining. Paraffin sections of mammary tissue and tumors were prepared by the Histopathology and Tissue Shared Resource, LCCC. Slides were baked at 60°C for 1 h, deparaffinized in xylene for 15 min, and rehydrated in 100%, 95%, and 70% ethanol for 5 min each. Antigen retrieval was achieved by steaming slides for 20 min in 10 mmol/L citrate buffer (pH 6.0). Slides were washed thrice in PBS and blocked for 1 h in a buffer containing 10% goat serum in PBS and then incubated at room temperature for 1 h with the following antibodies: ERα (1:1,000; Santa Cruz Biotechnology), CK5 (1:100; Chemicon), CK14 (1:25; NeoMarkers), CK18 (1:25; Epitomics), and CK19 (1:100; NeoMarkers). Slides were washed thrice in PBS and incubated with biotinylated secondary antibody for 30 min and washed thrice in PBS, and antigen was visualized with Vectastain avidin-biotin complex method and 3,3′-diaminobenzidine as substrate (Vector Labs). Slides were counterstained with Harris-modified hematoxylin (Fisher Scientific) and mounted in Permount.

Reporter gene assays. Reporter assays were carried out with either 3XPPRE-TK-luciferase (provided by Dr. Mitchell Lazar, University of Pennsylvania, Philadelphia, PA) to measure PPARγ-dependent activity, pERE-luciferase (provided by Dr. Anna Riegel, Georgetown University; ref. 23) to measure ERα-dependent activity, or TopFlash (Upstate Biotechnology) to measure β-catenin/T-cell factor (TCF)–dependent activity. Cells were cultured in 24-well plates at 20,000 per well and transfected using Lipofectamine Plus (Invitrogen). A plasmid expressing Renilla luciferase was cotransfected as an internal control. Cells were harvested after 48 h and luciferase activity was determined using the Dual Luciferase Assay (Promega) according to the manufacturer's protocol. All samples were run in triplicate and repeated thrice. For PPARγ-dependent activity, cells were incubated overnight in medium containing delipidated stripped serum (Sigma-Aldrich) and treated for 24 h with 1 μmol/L GW7845. For ER-dependent activity, cells were grown for 24 h in phenol red–free medium containing stripped serum and transfected with pERE-luciferase. Cells were then treated with 10 nmol/L 17-β-estradiol, and luciferase activity was measured 24 h later. TopFlash activity was determined in cells grown in 5% FBS in DMEM.

Ras activation assay. Ras activity was determined in MC and 437T cells grown in 5% FBS in DMEM for 24 h in 100-mm dishes and then incubated for 24 h in serum-free DMEM. Cells were then incubated for 5 min with 10 ng/mL EGF (Invitrogen), and cell lysates were assayed with the Ras Activation Assay kit (Cytoskeleton, Inc.) according to the manufacturer's protocol. Each reaction contained 250 μg protein, and the Raf-1 pull-down product was separated in a 4% to 12% NuPAGE gel (Invitrogen) and analyzed for activated Ras by Western blotting.

DNA methylation assay. Genomic DNA was extracted from mammary gland and tumor cells in 600 μL DNA lysis buffer (10 mmol/L Tris, 0.4 mol/L NaCl, 0.6% SDS, 0.1 mol/L EDTA, 20 μg/mL protease K) at 56 °C for 5 h. The lysate was mixed with 150 μL of 6 mol/L NaCl and centrifuged at 12,000 rpm for 10 min. The supernatant was removed and mixed with an equal volume of 100% ethanol and centrifuged at 12,000 rpm for 5 min, and the pellet was washed once with 75% ethanol. DNA was dissolved in 300 μL nuclease-free water, extracted with phenol/chloroform/isoamyl alcohol (25:24:1; Fluka), centrifuged at 12,000 rpm for 5 min, and washed twice with 75% ethanol. DNA (2 μg) was then digested overnight at 37°C with EcoRI and purified with the QIAquick Nucleotide Removal kit (Qiagen). DNA (45 μL) was mixed with 5 μL freshly prepared 3 mol/L NaOH and incubated at room temperature for 10 min. Cytosine methylation was determined using the EZ DNA Methylation-Gold kit (Zymo Research Corp.) according to the manufacturer's protocol. The ERα promoter region of untreated DNA and bisulfite-oxidized DNA was amplified by PCR and sequenced. The primers used were as follows: bisulfite-treated DNA, 5′-TTGGAGTTTTTTTTAGGAATGTTGATTTTAG-3′ (sense) and 5′-AACCAATCCTACCCTACTAATTCAAAAAC-3′ (antisense); untreated DNA, 5′-CTGGAGTTTCTTCTAGGAATGCTGATTCTAG-3′ (sense) and 5′-AGCCGATCCTACCCTGCTGGTTCAAGAGC-3′ (antisense). The 345-bp PCR product was cloned into pCR2.1 (Invitrogen) and sequenced (see Supplementary Fig. S3).

Statistical analyses. Analysis of two variables was conducted by Student's two-tailed t test and of multiple independent variables by Fisher's exact test. Survival curves were analyzed by Wilcoxon's rank test. P < 0.05 was considered to be statistically significant.

MMTV-Pax8PPARγ transgenic mice. Mice expressing MMTV-Pax8PPARγ were developed to first assess the influence of endogenous PPARγ activation on mammary carcinogenesis and, second, to determine if the chemopreventive effects of PPARγ agonists were a result of their direct action on mammary epithelium. Pax8PPARγ was chosen as the transgene because it functions as an effective dominant-negative receptor that completely blocks ligand-dependent PPARγ activation (14, 18), in contrast to “dominant-negative” variants that act as low-affinity receptors (24). Transgenic mice did not exhibit gross morphologic changes in mammary gland structure in either virgin or pregnant mice after lactation and involution (see Supplementary Fig. S1). To ascertain that Pax8PPARγ functioned as a true dominant-negative receptor, primary mammary epithelial cell cultures were transfected with PPRE-luciferase in the presence and absence of the PPARγ agonist GW7845 (see Supplementary Fig. S1B). Transgenic mice did not express endogenous PPARγ-dependent reporter gene activity, in contrast to WT mice, which did express activity.

Gene profiling of mammary epithelium from MMTV-Pax8PPARγ revealed changes in genes associated with transport, the cell cycle, adhesion, and receptor signaling, such as 3-phosphoinositide–dependent protein kinase-1, phosphatidylinositol 3-kinase (PI3K) p110α, Ras-activating protein, Fos, Jak1, RGS5, cyclin D2, ABCA1, and PGC-1α, a coactivator of ERα (25), ERRα (26), and PPARγ (see Supplementary Table S1 and Supplementary Fig. S2; ref. 27). Up-regulation of proliferative gene expression was complemented by a reduction of the cyclin kinase inhibitor p27^Kip1, the Notch inhibitor Deltex, and an increase in the antiapoptotic gene Birc4. Expression of the ER-responsive genes XBP1, cyclin D1, and IL-6 signal transducer (28) were all significantly increased in MMTV-Pax8PPARγ mice.

Increased stem/progenitor cell expansion in MMTV-Pax8PPARγ. To determine if PPARγ affected cell lineage, primary cell cultures were evaluated for CD antigen expression (Fig. 1A). CD29^hi/CD24^neg cells were more abundant in transgenic mice in comparison with WT mice (9.4% versus 0.6%). In addition, immunostaining for cytokeratin expression revealed a greater percentage of CK5⁺ cells in the transgenic mammary gland (Fig. 1B) and an increased number of double-positive CK14/CK18 cells (Fig. 1C). These results suggest increased stem and progenitor cell expansion downstream of Pax8PPARγ.

β-Catenin and ERK signaling are increased in transgenic mice. Tissue lysates from MMTV-Pax8PPARγ mice revealed a marked reduction in PTEN and pGSK3β and an increase in pERK1 and pAKT (Fig. 2A). Measurement of β-catenin/TCF–dependent reporter gene activity in primary cell cultures indicated a moderate increase in activity in cells from transgenic mice in comparison with those from WT animals (Fig. 2B). These findings suggest that Pax8PPARγ increases Wnt signaling through reducing PTEN expression, increasing pERK and pAKT, and reducing pGSK3β to promote β-catenin/TCF transactivation.

Pax8PPARγ reduces PTEN and up-regulates pERK and pAKT in mammary tumors. Because spontaneous tumorigenesis did not occur in MMTV-Pax8PPARγ mice over their life span, their sensitivity to mammary carcinogenesis (8) was determined (Fig. 3A). MMTV-Pax8PPARγ mice were more sensitive to carcinogenesis and exhibited a median tumor latency of 40 days versus 70 days in WT mice. Tumor multiplicity was greater in MMTV-Pax8PPARγ mice, and ductal adenocarcinoma formation was significantly greater and squamous cell carcinoma formation was significantly less than in WT mice (Table 1).

To assess the effect of PPARγ activation on mammary tumorigenesis, animals were maintained on a diet supplemented with PPARγ agonist GW7845 (Fig. 3A). Whereas GW7845 delayed median tumor latency by approximately 1 month in WT mice as reported previously (21), transgenic mice were insensitive to the PPARγ agonist. These results show for the first time that a PPARγ agonist acts directly on mammary epithelium and not indirectly through its effects on stromal tissue.

To determine if mammary adenocarcinomas recapitulated the signaling phenotype present in the mammary gland of transgenic mice, mammary adenocarcinoma cell lines were developed from WT (MC cells) and transgenic (437T cells) animals (Fig. 3B). PTEN was markedly reduced in 437T cells, and pERK and pAKT were increased coincident with reduced β-catenin phosphorylation in comparison with MC cells. 437T cells were also less sensitive to the MAP/ERK kinase (MEK) inhibitor U0126, which was consistent with their greater pERK expression (Fig. 3C). To determine if ERK activation was associated with Ras activation, “pull-down” assays were carried out with the Raf domain recognizing activated Ras-GTP (Fig. 3D). In response to EGF stimulation, Ras activation was greater in 437T versus MC cells. In addition, β-catenin/TCF–dependent reporter gene activity was increased to a greater extent in 437T cells than in MC cells (Fig. 3E), which were consistent with the findings in primary cultures from transgenic mice (Fig. 2B).

Pax8PPARγ increases ER⁺ mammary carcinogenesis. Because ductal adenocarcinomas were the predominant histopathologic phenotype in MMTV-Pax8PPARγ mice following progestin/DMBA-induced carcinogenesis (Table 1), tumors were analyzed for ER expression (Fig. 4A). Significantly, tumors from transgenic animals were predominantly ER⁺, whereas little or no ER expression was present in adenocarcinomas from WT animals (Fig. 4A). ER mRNA was increased in tumors and primary mammary gland cultures from transgenic animals (Fig. 4B). In this regard, transduction of Comma-1D mouse mammary epithelial cells with Pax8PPARγ induced ER expression (Fig. 4B). Importantly, tumor formation in transgenic mice was inhibited by the ER antagonist fulvestrant when administered on a weekly basis beginning after the last dose of DMBA (Fig. 4C), indicating that tumor growth is functionally dependent on ER signaling. ER-dependent reporter gene activity was further assessed in MC and 437T cells (Fig. 4D). 437T cells exhibited ∼10-fold increase in reporter gene activity in comparison with MC cells. To determine if ER activation was associated with hypomethylation of the ER promoter, genomic analysis of the CpG motifs in the ER promoter region of MC and 437T cells was determined (see Supplementary Fig. S3). The ER promoter was not methylated in either tumor cell line, indicating that this mechanism is not responsible for ER induction in Pax8PPARγ-expressing cells.

Tumorigenesis involves the activation of oncogenic pathways in concurrence with repression of “tumor suppressor” pathways that ultimately influence not only tumor initiation and progression but also the lineage characteristics of malignant cells derived from transformed stem and/or early progenitor cells. Here, we present evidence that inhibition of PPARγ through a dominant-negative receptor promotes a stem and progenitor cell phenotype that results in the development of ER⁺ ductal carcinomas following carcinogenesis (Fig. 5). These results suggest a previously unknown link between PPARγ suppression and cell fate determination in modulating mammary gland differentiation along the luminal lineage.

The increase in proliferation and transformation of thyroid cells and fibroblasts by Pax8PPARγ (17) is consistent with its putative role as an oncogene in follicular thyroid cancer (14). Although Pax8PPARγ was not oncogenic in the mammary gland, it increased the percentage of CD29^hi/CD24^neg, CK5⁺, and CK5/CK19 and CK14/CK18 double-positive cells. CD29^hi/CD24^lo/neg cells have been reported to lack ER expression (29, 30). CD29^hi/CD24^neg cells are enriched in multipotent stem cells, are regulated by the Wnt pathway (31), and have a high regenerative capacity in the mammary fat pad (30, 32). Thus, despite the absence of an overt developmental phenotype in MMTV-Pax8PPARγ mice, which resembles PPARγ-null mice (33), this model indicates that PPARγ regulates mammary lineage specification and likely accounts for the histopathologic differences in tumors arising in these animals, such as increased ductal carcinomas and a reduction in heterologous metaplasia (i.e., squamous cell carcinomas; Table 1). These findings are consistent with the ability of PPARγ to act as a developmental switch to control differentiation, as shown previously in adipocytes and osteoblasts (34, 35).

The Wnt pathway increases expansion of Sca-1⁺ mammary tumor cells (36), progenitor cells (37), and stem cells (38). This phenotype is reminiscent of the association between PPARγ haploinsufficiency, increased β-catenin expression, and colon carcinogenesis (39). PPARγ associates with and targets WT β-catenin (40), but not oncogenic variants of β-catenin (41), for proteasomal degradation. β-Catenin activation (42) and transformation (22) are closely regulated by PI3K signaling, where loss of PTEN suppression increases stem and progenitor cell expansion (Fig. 5; ref. 43). The PI3K pathway is crucial for embryonic stem cell proliferation and tumorigenicity (44) and for the maintenance of stem cell pluripotency by AKT activation (45). It has also been shown that PI3K regulates ERK activation downstream of several oncogenes (46), that Ras and ERK mediate the transforming activity of the PI3K p110β and p110γ catalytic subunits (47), and that PI3K inhibition blocks ERK activation mediated by integrin signaling (48). These results are concordant with our findings that loss of PTEN, inhibition of GSK3β, activation of ERK and AKT, and increased β-catenin/TCF transactivation activity result from Pax8PPARγ expression in the mammary gland and are consistent with the ability of the PI3K-Ras pathway to promote Wnt pathway activation (49). In this context, Pax8PPARγ is known to suppress expression of NORE1A (16), an inhibitor of the ERK pathway (50), and agrees with the ability of Pax8PPARγ to stimulate and of WT PPARγ to inhibit ERK activation (51). Because ERK inhibits PPARγ and targets it for proteasomal degradation (52–55), this effect may also account for the reduction of PPARγ in the mammary gland of pregnant transgenic mice (Supplementary Fig. S1A).

MMTV-Pax8PPARγ mice expressed predominantly ER⁺ ductal adenocarcinomas. The high sensitivity of tumor formation to fulvestrant indicates that tumor growth is functionally dependent on ER signaling, although the precise mechanism by which inhibition of PPARγ induces ER expression is unknown. One possibility consistent with increased ER expression and ER-dependent transcriptional activity is the increase in expression of PGC-1 (Supplementary Table S1), a PPARγ coactivator (27) that preferentially coactivates ER (25), especially under conditions where PPARγ is repressed by Pax8PPARγ. Pax8PPARγ may also interfere with the ability of PPARγ to induce proteasomal degradation of ER and cyclin D (56, 57), an action that would be further sustained by Ras and ERK activation. An alternate but not necessarily exclusive possibility is that Pax8PPARγ interferes with the reported competition between PPARγ and ER for promoter occupation in estrogen response element (ERE)-responsive genes (58–60). However, we have not found this to be the case for ERE-dependent reporter gene activity when both receptors are coexpressed in 293T cells.³

Lastly, Pax8PPARγ could act as a dominant-negative regulator of Pax8 transcriptional activity. Pax8 transactivates the Wilm's tumor suppressor gene WT-1, which is known to inhibit ER function (61, 62); however, Pax8 and WT-1 expression were not detectable by real-time PCR (RT-PCR) in either the mammary gland or mammary tumors.³ Thus, by relieving the negative regulatory effects of PPARγ on ER stability and PTEN tumor suppressor activity, Pax8PPARγ would promote ER stability and coactivation by PGC-1 through ERK- and AKT-dependent phosphorylation (Fig. 5).

In summary, Pax8PPARγ induced mammary stem and progenitor cell expansion in the context of reduced PTEN and increased Ras, ERK, and AKT activation. Transgene expression resulted in a high tumor multiplicity and, most significantly, predominantly ER⁺ ductal adenocarcinomas following carcinogenesis, which was insensitive to the chemopreventive effect of a PPARγ agonist but profoundly inhibited by the ER antagonist fulvestrant. These results reveal important new insights into the previously unrecognized role of PPARγ in the specification of mammary lineage and the development of ER⁺ tumors, including potential intervention modalities.

No potential conflicts of interest were disclosed.

Grant support: National Cancer Institute, NIH grant R01 CA11482 and contracts N01 CN43309 and N01 CN43330. This investigation was conducted in a facility constructed with support from Research Facilities Improvement Grant C06 RR14567 from the National Center for Research Facilities, NIH.

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.

We thank Aaron Foxworth (Georgetown University Animal Research Resource) for assistance with the carcinogenesis studies.