Activating Transcription Factor 2 and c-Jun–Mediated Induction of FoxP3 for Experimental Therapy of Mammary Tumor in the Mouse

The overwhelming majority of tumor suppressor genes are autosomal, and their inactivations involve two genetic events (1–4). The best defined two hits of the tumor suppressors in cancer cells are usually irreversible, such as deletion or mutations (3, 4). More recently, however, it has become increasingly clear that epigenetic inactivation of tumor suppressors plays a critical role in inactivating tumor suppressor genes (5). On the other hand, due to X-inactivation, the X-linked tumor suppressor genes are operatively hemizygous, and can be inactivated by a single hit (6). This notion has been substantiated by the recent identification of two tumor suppressor genes, FoxP3 for breast cancer (7) and WTX for Wilms' tumor (8). Because the majority of mutations and/or deletions associated with X-linked tumor suppressor genes are heterozygous (7, 8), most cancer cells have a wild-type (WT) allele that has not been irreversibly inactivated (7).

An important aspect of tumor therapy is how to restore the function of tumor suppressors. For those with two irreversible genetic changes, such as p53, this has been technically challenging, although a pharmaceutical restoration of mutant protein function has been reported (9). Because X-linked tumor suppressor genes are operationally hemizygous, only one allele is subject to genetic selection in cancer cells. The other allele is therefore genetically intact and can potentially be induced to suppress tumor growth. Based on this concept, we searched for a biochemical pathway that can be activated to induce FOXP3 expression in breast cancer cells. Here, we report that anisomycin, which is commonly used to induce stress responses of cells, induces expression of FOXP3 in breast cancer cell lines. Our biochemical and genetic analyses revealed that activating transcription factor (ATF)2, which was recently shown to be a tumor suppressor gene in the mouse (10), is essential for the induction of FoxP3 and FoxP3-mediated apoptosis. Moreover, ATF-2 forms heterodimer with c-Jun to activate transcription of FoxP3. These data show a novel function of ATF2 in the expression of FoxP3 in the epithelial cells and suggest a novel approach for the therapy of breast cancer.

Antibodies and reagents. Anti–ATF-2 (20F1), phospho-ATF2-(Thr71), and activated caspase-3 were purchased from Cell Signaling, Inc. Other vendors are as follows: anti-Foxp3 (eBioscience; #14-5779-82); and anti–β-actin (I-19), c-Jun (NX), and phosphor-c-Jun (KM-1; Santa Cruz Biotechnology, Inc.). Chemicals SP600125, SB203580, and PD98059 were purchased from CalBiochem, Inc., whereas anisomycin was purchased from Sigma, Inc.

Experimental animals. Mice heterozygous for Atf2 null mutation (Atf2^tm1Glm/Atf2⁺129S2/SvPas; ref. 11) were revived from the frozen embryo bank in The Jackson Laboratory. Heterozygous mice were crossed to produce Atf2^+/+ and Atf2^−/− littermates. BALB/c mice were purchased from Charles River through a National Cancer Institute Subcontract. All studies involving animal has been approved by University Committee on Use and Care of Animals at University of Michigan.

Preparation of mammary epithelial cell culture. Mouse mammary fat pads were removed from 6- to 8-wk-old virgin female mice and minced into small pieces. After collagenase digestion at 37°C in a shaking incubator in DMEM supplemented with 5% FCS, cells were sieved through a 70-μm cell strainer (BD Falcon) to obtain a single-cell suspension. The cells were cultured in DMEM supplemented with 10% fetal bovine serum and 10 ng/mL epithelial growth factor. At day 3 of culture, fibroblast cells were removed by a short digestion with 0.05% trypsin-EDTA as less adherent cells.

Measurement of FOXP3 transcripts. Total cDNA were prepared from breast cancer cell lines or epithelial cultures. The levels of FOXP3 mRNA were measured by reverse transcription-PCR (RT-PCR) under two conditions. Full-length encoding regions were analyzed by agarose gel electrophoresis, whereas shorter transcripts were quantitated using real-time PCR. All primers for PCR were listed in Supplementary Table S1.

ShRNA lentiviral vector. The lentivirus-based shRNA expressing vectors were created by introducing the murine U6 RNA polymerase III promoter and a murine phosphoglycerate kinase promoter (pGK)–driven enhanced green fluorescent protein expression cassette into a vector of pLenti6/V5-D-TOPO back bone without cytomegalovirus promoter. Hairpin shRNA sequence of FoxP3, JNK1, JNK2, and Atf2 (FoxP3: 5′-aagccatggcaatagttcctt-3′; FOXP3: 5′-gcagcggacactcaatgag-3′; JNK12: 5′-agaaggtaggacattcctt-3′ and 5′-aagcctagtaatatagtagt-3′; Atf2: 5′-cttctgttgtagaaacaac-3′ and 5′-agcacgtaatgacagtgtca-3′) were cloned into the lentiviral shRNA expressing vectors by restriction sites of ApaI and EcoRI.

Electrophoresis mobility shift assay. Anisomycin were added to 4T1 cells in conjunction with either vehicle control or kinase inhibitor SP600125 at a dose of 2 μg/mL for 2 h before cells lysed. The nuclear extracts were mixed with either WT or mutant probes in the presence of either control anti–c-Jun (Santa Cruz; sc-45X) or anti-ATF2 (Cell Signaling, #9226) antibodies, as indicated and analyzed by electrophoresis, as described (12).

Western blot. Protein samples for Western blot were prepared by lysing cultured cells in SDS sample buffer, resolved on 10% SDS-PAGE, and electroblotted onto nitrocellulose membranes. Membranes with transferred proteins were incubated with primary antibody followed by incubation with horseradish peroxidase conjugated to the secondary antibody. Chemiluminescence reaction using the enhanced chemiluminescence kit (Amersham Biosciences) was detected by film.

Chromatin immunoprecipitation (ChIP) was carried out according to a published procedure (13). Briefly, the vehicle or 2-h anisomycin-treated 4T1 cells were sonicated and fixed with 1% paraformaldehyde. The anti–phospho-c-Jun or anti–phosphor-ATF2 antibodies or control rabbit IgG were used to pull down chromatin associated with these proteins. The amounts of the specific DNA fragments were quantitated by real-time PCR and normalized against the genomic DNA preparation from the same cells.

Immunoflurensence staining. TSA cells were treated or left untreated with vehicle DMSO or 50 ng/mL of anisomycin for 24 h. After that, the cells were fixed by methanol, permeabilized with 0.3% Triton-X100, and stained with rabbit antibody against cleaved Caspase 3 (Cell signaling, #9661s) overnight. The stained cells were then visualized with Cy3-conjugated anti-rabbit IgG (Jackson ImmunoResearch Laboratory).

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) cell viability assay has been described in details (14).

DNA contents Anisomycin-treated or control cells were stained by Propidium Iodide using “PI/RNase Staining Buffer” from BD Biosciences according to the manufacturer's manual.

Induction of FoxP3 by anisomycin. We have recently shown that the expression of FoxP3 cDNA leads to rapid cell death of breast cancer cell lines (7, 15). These results raised the possibility that the induction of FoxP3 may represent a novel therapeutic approach for the treatment of breast cancer. When we screened for drugs that induce FoxP3 expression in mammary tumor cell lines, we observed a rapid induction of FoxP3 mRNA by anisomycin. As shown in Fig. 1A, significant levels of the FoxP3 transcripts were induced in a mouse mammary tumor cell line, TSA, as early as 4 hours after anisomycin treatment. Similar effects were observed in another mouse mammary tumor cell line 4T1 and human breast cancer cell line MCF-7 (Fig. 1B), which we showed to harbor the WT FOXP3 gene (7). In contrast, treatment of phorbol 12-myristate 13-acetate (PMA) did not result in any induction of FoxP3 (Fig. 1B). Because higher doses of anisomycin can inhibit translation, we tested induction of the FoxP3 protein by Western blot. As shown in Fig. 1C, low doses of anisomycin induced high levels of FoxP3 protein, which indicated that activation of FoxP3 locus can be achieved at doses that did not prevent translation of the FoxP3 protein. Importantly, the induction of FoxP3 protein can be prevented by FoxP3 shRNA. These data not only proved the specificity of Western blot but also confirmed that the accumulation of FoxP3 protein depends on induction of FoxP3 mRNA.

Involvement of ATF2, c-Jun-NH₂-kinase in FoxP3 induction. As a first step to identify the mechanism by which anisomycin induced FoxP3, we treated mammary tumor cell line with anisomycin in conjunction with inhibitors of overlapping specificity, including ATF-2/c-Jun-NH₂-kinase (JNK) inhibitor SP10096 (SP), p38α inhibitor SB203580 (SB), and p41/42 MAP kinase inhibitor PD9786 (PD). As shown in Fig. 2A, SP completely prevented the induction of FoxP3. On the other hand, SB and PD had little effect. These data raised the possibility that ATF-2 and JNK pathways may be involved in the induction of FoxP3 by anisomycin.

To test this hypothesis, we generated lentiviral vectors expressing shRNA for JNK1/2 or ATF2. The efficacy of shRNA silencing and the effect of the silencing on anisomycin-mediated induction of FoxP3 is shown in Fig. 2B. These data showed that silencing either JNK or ATF2 resulted in abrogation of the induction of the FoxP3 transcripts and protein by anisomycin. These data provide important genetic evidence for the involvement of JNK and ATF2 in anisomycin-induced FoxP3 expression.

Interestingly, a recent study showed that mice with heterozygous deletion of the ATF2 gene developed spontaneous mammary tumors (10). Because FoxP3 heterozygous mutants have the same phenotype, it is intriguing that ATF2 may be responsible for constitutive and/or inducible expressions of FoxP3 in mammary epithelial cells. To address this issue, we obtained ATF2^+/− mice from the frozen embryo bank of The Jackson Laboratory. The ATF2^+/+ and the ATF2^−/− mice were obtained by F1 cross. A previous report indicated that the only a small fraction of the ATF2^−/− mice survive to adulthood (11). We obtained two ATF2^−/− females, from which we obtained two independent primary mammary epithelial cell cultures (Fig. 2D). The epithelial origin of the cultures was shown by the expression of CK19. Because T cells are the major source of FoxP3 transcripts in vivo, we also confirmed that the primary culture has no T-cell contamination by the lack of CD3 transcripts (Fig. 2D). WT epithelial cultures expressed significant amounts of Foxp3 transcripts, which were further induced by the treatment of anisomycin. ATF2^−/− cells had no detectable FoxP3 transcripts and did not express FoxP3 after stimulation by anisomycin. These data revealed an essential role for ATF2 in both constitutive and inducible expressions of FoxP3 in normal epithelial cells. On the other hand, the thymocytes from the ATF2^−/− mice had normal number of CD4⁺FoxP3⁺ T cells (data not shown). Therefore, the function of ATF2 in FoxP3 expression seems to be epithelia specific.

Identification of the FoxP3 enhancer associated with ATF2 and c-Jun. To study the mechanism of ATF2/JNK-mediated induction of FoxP3, we carried out ChIP to identify an anisomycin-inducible binding site of the FoxP3 locus. To identify specific ATF2 binding sites, we treated the 4T1 cell line with or without anisomycin and carried out ChIP with either control IgG or anti–p-ATF2 antibodies. Because JNK regulate transcription by phosphorylation of c-Jun (16), we used anti–phospho-c-Jun antibodies in the ChIP. To identify the FoxP3 sequence associated with p-ATF2 and p-c-Jun, we first analyzed the 5′ sequence of the FoxP3 gene and identified 14 potential AP1 and cAMP-responsive element binding protein sites. PCR primers were designed across the 10.4 kb regions, and the amounts of each PCR product were normalized against that amplified from the input DNA. The quantitative real-time PCR results were shown in Fig. 3A, whereas the PCR products from two major peaks were shown in Fig. 3B. These data reveal two potential sites for ATF2/cJun interaction. The first is hereby called P2, which is 4.8 kb 5′ of exon 1. The second and the stronger binding site P10 is 4.2 kb 3′ of exon 1. Importantly, whereas the P2 ATF2/cJun association is not inducible by anisomycin, the P10 binding is enhanced by >2-fold by anisomycin. Moreover, comparison of mouse and human FoxP3 sequence revealed that the P10, but not the P2 site, is highly conserved (Supplementary Fig. S1). Therefore, we focused on the potential significance of P10 as the site for p-ATF2 and p-cJun interaction.

Sequencing comparison identified a typical AP1 site within the P10 (Supplementary Fig. S1). To directly show interactions of ATF2 and c-Jun to the FoxP3 promoter, we radiolabeled an oligonucleotide probe containing conserved AP1 site as well as two control oligos with mutations in the AP1 site and tested their binding to nuclear extracts. As shown in Fig. 3C, the nuclear extracts from anisomycin-treated, but not those from the untreated, 4T1 cells showed strong interaction with the WT P10 probe. The specificity was confirmed as mutations in the AP1 site significantly reduced the binding. Furthermore, the involvement of ATF2 and c-Jun was shown as their specific antibodies abolished the binding of nuclear extracts to WT probe. Thus, both ChIP and electrophoresis mobility-shift assay identify a specific activator protein (AP1) site with 4.2 kb 3′ of the TSS, which binds to both p-ATF2 and p-cJun by anisomycin-inducible fashion.

To test whether the P10 sequence was a functional FoxP3 enhancer, we generated a series of constructs consisting of the basal promoter and putative enhancer elements. As shown in Supplementary Fig. S2, a 265 bp sequence 5′s of TSS of the FoxP3 locus plus 50 bp down-stream of TSS is sufficient to convey a significant basal promoter activity. This fragment is therefore chosen to measure the enhancer activity. As shown in Fig. 4, addition of three copies of P2 fragment increased the promoter activity by ∼2-fold, which suggests that P2 is at best a weak enhancer. Inclusion of three copies of P10 sequences, however, increased the Foxp3 promoter activity by 10-fold. Surprisingly, this seems unidirectional as the inversion of the P10 fragment eliminated its enhancer activity. Moreover, the involvement of AP1 site in P10 was confirmed as a mutation of the AP1 site significantly reduced the enhance activity. Moreover, addition of P2 to P10 failed to further enhance the promoter activity. Taken together, our data showed that anisomycin induced ATF2/c-Jun interaction with a specific enhancer within the intron 1 of the FoxP3 gene.

A critical role for ATF2-FoxP3 pathway in anisomycin-induced apoptosis and the therapy of breast cancer. Our recent studies have shown that the induced expression of FoxP3 caused apoptosis of breast cancer cell lines (7, 11, 15). To determine whether anisomycin treatment caused apoptosis of breast cancer cells, we measured the cytotoxic effect of anisomycin on several of breast cancer cell lines, by MTT assay. As shown in Fig. 5A, both mouse (TSA) and human breast cancer cell lines (BT474, MCF-7) were highly susceptible to anisomycin, with an IC₅₀ between 50 and 100 ng/mL. The reduced viability is due to apoptosis as revealed by the increased expression of active caspase 3 in TSA cells (Fig. 5B,, top) with less than 2C DNA contents (Fig. 5B,, bottom). Given the critical role for ATF2 in FoxP3 induction, we tested the contribution of ATF2 to anisomycin-induced cell death by comparing the dose response to anisomycin in cells transfected with either vector alone or those with ATF2 shRNA. As shown in Fig. 5C  and D, ATF2 and FoxP3 shRNAs increased resistance to anisomycin by nearly 3-fold. These data show a critical role for the ATF2-FoxP3 pathway in anisomycin-induced cell death of breast cancer cells.

To test whether induction of FoxP3 by ATF2-FoxP3 pathway can be explored for breast cancer therapy, we injected the TSA cell line into the mammary pad. Seventeen days later, when the cancer cells established locally, the mice were treated with either vehicle control or anisomycin. As shown in Fig. 6, the growth of the TSA tumor cells in syngeneic mammary pad is abrogated by anisomycin. These data show the potential of ATF2-FoxP3 pathway in the therapeutic development for breast cancer.

FOXP3 is an X-linked gene that is subject to X-inactivation (7, 17). Our inquiry into the high incidence of spontaneous mammary tumors in mice heterozygous for the Scurfy mutation led to the identification of it as the first X-linked tumor suppressor for breast cancer in mouse and in woman (7). FOXP3 acts as a transcriptional repressor of oncogenes such as ErbB2 and SKP2 (7, 15). Moreover, ectopic expression of FOXP3 cause an apoptosis of breast cancer cells (7). These data showed that the induction of FOXP3 in the tumor cells may prove valuable for the treatment of breast cancer.

Because the one WT allele was not irreversibly inactivated in the overwhelming majority of breast cancer samples analyzed (7), it is theoretically possible to reactivate the expression of FOXP3 locus for the treatment of breast cancer. The data presented herein showed such reactivation by anisomycin.

We observed that anisomycin, a drug commonly used to activate MAP kinases, rapidly induced FOXP3 expression in multiple breast cancer cell lines tested. Using shRNA specific for Jnk and Atf2, we showed that the Jnk and Atf2 genes are required for the induction of FoxP3 expression. Moreover, biochemical analysis allowed us to identify critical cis-element involved in the induction of FoxP3 and that this cis-element interacts with c-Jun and ATF2 to cause the activation of the FoxP3 gene. It is worth considering whether the effect of anisomycin can be related to reactivation of X-inactivated FoxP3, given recent reports of chromatin modification-dependence of c-Jun–induced transcriptions (18, 19).

Using mammary epithelial culture isolated from WT and ATF2^−/− mice, we showed that the targeted mutation of Atf2 not only reduced the basal levels of FoxP3 transcripts in the mammary epithelial cells, but also eliminated its induction by anisomycin. Therefore, ATF2 plays an essential role in both constitutive and inducible expression of FoxP3. It is of great interest to note that mice heterozygous for Atf2-null allele spontaneously developed mammary tumors (10). The similarity in tumor onset suggests that the lack of FoxP3 expression may be an underlying cause for the spontaneous mammary tumors, although the Atf2^+/− mice available to us is in 129/SV background, which is not suitable to address this issue.

FOXP3 is expressed in both T cells and epithelial cells (7, 20, 21). Because the majority of the Atf2^−/− mice die shortly after birth (11), a systemic analysis of the effect on expression of FOXP3 in the T-cell lineage remained to be determined. However, our preliminary analysis suggested that Atf2 is not essential for FOXP3 expression in the thymocytes (data not shown). Therefore, ATF2 may play different roles in different lineages. Consistent with this notion, both cis-element and the trans-activating factors identified here differ from what were reported in FoxP3 induction in T cells (22–25).

Finally, an important but unresolved issue is how to reactivate tumor suppressor function in the tumor cells. Classic tumor suppressors are inactivated by two irreversible hits (2). Therefore, despite an elegant recent approach (9), restoring the function of classic tumor suppressors remains a major challenge for cancer therapy. On the other hand, recent data from our group and that of another indicated that X-linked tumor suppressor genes are subject to X-inactivation (7, 8). Because X-inactivated genes are not subject to selection during tumor growth and because deletion and mutation found in the majority of the cases are heterozygous (7, 8), it may be possible to reactivate FOXP3 for the treatment of breast cancer. In this regard, we have shown that anisomycin cause apoptosis in an ATF2- and FOXP3-dependent manner. Moreover, the doses used here do not interfere with protein translation and cause no obvious side effect, yet the drug causes dramatic inhibition of growth of established mammary tumors in syngeneic hosts. Although more work is needed to evaluate the potential for anisomycin for breast cancer treatment, our data suggest that one may be able to reactivate X-linked tumor suppressor genes for cancer treatment.

No potential conflicts of interest were disclosed.

Grant support: National Cancer Institute, Department of Defense, and American Cancer Society.

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 Lynde Shaw and Todd Brown for assistance.