FOXP3 Up-regulates p21 Expression by Site-Specific Inhibition of Histone Deacetylase 2/Histone Deacetylase 4 Association to the Locus Free

As a universal cyclin-dependent kinase inhibitor, p21 plays an important role in preventing cell cycle progression by acting at G₁ checkpoint (1–4). p21 is down-regulated in many types of cancers, including the majority of breast cancer (5–7). Absence of p21 has been shown to confer oncogenic properties to KLF4 (8). Moreover, p21 loss is causatively related to the tamoxifen-stimulated growth of breast cancer (9). Surprisingly, p21 mutation is rarely observed in cancer (10). Instead, p21 has emerged as a major downstream target of tumor suppressor genes, including p53 (1, 11, 12), BRCA1 (13), CHK2 (14), KLF4 (15, 16), and KLF6 (17). Although p53-mediated regulation has been established as a classic example, the lack of correlation between p53 protein levels (usually used as an indication of p53 mutation) and the down-regulation of p21 would argue strongly that p53 mutation is perhaps not the major underlying cause for p21 loss in breast cancer (5–7). Likewise, although it has been shown that BRCA1-mediated (13) and Chk2-mediated (14) tumor cell cycle arrest and senescence require p21 function, mutations of these two genes had not been established as the genetic cause for the lack of p21 in the tumors. On the other hand, epigenetic factors have been suggested as possible mechanisms of p21 silencing in breast cancer cells (18–21).

We reported recently that heterozygous FOXP3 mutation leads to spontaneous development of mammary tumors (22). The significance of FOXP3 mutation in human is shown by both widespread somatic mutation and deletion of the gene in human breast cancer samples (22). Ectopic expression of the FoxP3 gene caused profound growth inhibition for breast cancer cell lines both in vivo and in vitro. Because FoxP3 is a transcription factor, it is important to identify critical targets of FoxP3 that are responsible for the tumor suppressor activity of FoxP3. In this context, we have reported that FoxP3 is a repressor for the HER-2/ErbB2 and Skp2 oncogenes (22, 23). Alternatively, it is possible that FoxP3 may activate additional tumor suppressor genes. To test this hypothesis, we used a gene array analysis to identify genes affected by FOXP3. We uncovered several tumor suppressor genes that were induced >2-fold after induction of FOXP3. We focused on p21, as it is the most highly induced tumor suppressor and because of its unique role in breast cancer biology. Here, we report that FOXP3 is a potent inducer of p21 in both normal epithelial cells and malignant breast cancer cell lines. Our data provide a novel mechanism for FOXP3-mediated activation of tumor suppressor gene.

Mice. Rag2^−/−FoxP3^+/+ and Rag2^−/−FoxP3^sf/sf BALB/c mice have been described previously (24). Two-month-old virgin mice were used to analyze the effect of FoxP3 mutation on p21 expression and hyperplasia of mammary epithelia. All animal experiments were conducted in accordance with accepted standards of animal care and approved by the Institutional Animal Care and Use Committee of University of Michigan.

Cell culture. Breast cancer cell line MCF-7 and immortalized mammary epithelial cell line MCF-10A were purchased from the American Type Culture Collection. The HO15.19 cell line, which is the c-Myc null derivative of TGR-1 (25, 26), was a kind gift from Dr. John M. Sedivy of Brown University. A previously established tet-off FOXP3 expression system in the MCF-7 cells was also used (22, 23).

Microarray analysis of FOXP3-regulated genes. The FOXP3 tet-off MCF-7 cells (22, 23) were seeded in six-well plates and cultured with (2.0 μg/mL) and without doxycyclin in the culture media. After 48 h of incubation, cells were washed with ice-cold PBS twice and RNA extraction was performed with RNeasy Mini kit (Qiagen) according to the manufacturer's protocol. Contaminated genomic DNA was eliminated with DNase I (Invitrogen) according to the manufacturer's protocol. We conducted mRNA microarray analyses using HG-U133 Plus 2.0 (Affymetrix) according to the manufacturer's protocol. We used the most current version of ENTREZ gene-based CDFs, as of July 2008, that has been maintained at University of Michigan for accurate analysis (27). dChip software (University of California at Los Angeles Clinical Microarray Core) was used to make a heat map of microarray profiles according to the instruction of the software. Gene expression profiles of FOXP3 tet-off cells cultured with and without doxycyclin were compared. Differences of mRNA expression levels between FOXP3⁺ and FOXP3⁻ cells were calculated by Student's t test.

FOXP3, p21, and histone deacetylase silencing. Two FOXP3 short hairpin RNA (shRNA) constructs are FOXP3-993 shRNA and FOXP3-1355 shRNA (Genbank accession no. NM_014009). Oligonucleotides encoding small interfering RNA (siRNA) directed against FOXP3 are 5′-GCTTCATCTGTGGCATCATCC-3′ for FOXP3-993-shRNA [993–1013 nucleotides from transcription starting site (TSS)] and 5′-GAGTCTGCACAAGTGCTTTGT-3′ for FOXP3-1355 shRNA (1355–1375 from TSS). The selected shRNA oligonucleotides were cloned into pSIREN-RetroQ vectors (Clontech) to generate siRNA according to manufacturer's protocol. The human p21shRNA (CGCCTCTGGCATTAGAATTATT), human shHDAC2 (shHDAC2-1, CCGACGGTGATATTGGAAATTA; shHDAC2-2, CGGGCAGATATTTAAGCCTATT), human shHDAC4 (shHDAC4-1, ACGGCATGACTTTATATTGTAT; shHDAC4-2, AGACCGGCATGACTTTATATTG), and control lentiviral vectors were purchased from Open Biosystems.

Western blot. The anti-FOXP3 (Abcom, 1:1,000), anti-hFOXY (eBioscience, 1:100), anti-p21 (Cell Signaling, 1:1,000), and anti–β-actin (Sigma, 1:3,000) were used as primary antibodies. Anti-rabbit or mouse IgG horseradish peroxidase–linked secondary antibody at 1:3,000 to 1:5,000 dilutions (Cell Signaling) was used. To ensure equal loading of proteins, the membranes were stripped under the same conditions as described above. They were then incubated with enhanced chemiluminescence reagents (Amersham Biosciences) and exposed to X-ray film for 1 to 5 min.

Chromatin immunoprecipitation. Chromatin immunoprecipitation (ChIP) was carried out according to published procedure (28). Briefly, the FOXP3-transfected tet-off cells were sonicated and fixed with 1% paraformaldehyde. The anti-FOXP3, anti–acetyl-H3 (Cell Signaling), anti–histone deacetylase 1 (HDAC1), anti-HDAC2, anti-HDAC3, anti-HDAC4, anti-HDAC5, anti-HDAC7 (Cell Signaling), and anti-IgG (Santa Cruz Biotechnology) antibodies were used to pull down chromatin associated with FOXP3. The amounts of the specific DNA fragment were quantitated by real-time PCR and normalized against the genomic DNA preparation from the same cells. The ChIP real-time PCR primers are listed in Supplementary Table S3.

Quantitative real-time PCR. Relative quantities of mRNA expression were analyzed using real-time PCR (ABI Prism 7500 Sequence Detection System, Applied Biosystems). The SYBR (Applied Biosystems) green fluorescence dye was used in this study. The primer sequences are listed in Supplementary Table S3.

Immunohistochemistry. Immunohistochemistry was performed by the avidin-biotin complex method. Expression of FOXP3 in human breast cancer or normal tissue samples was determined using immunohistochemistry, as described (22, 23). The p21 mouse monoclonal antibody (Cell Signaling, 1:100) and biotin goat anti-mouse IgG (Santa Cruz, 1:200) were used as secondary antibodies. FOXP3 and p21 staining were scored double blind.

Statistical analysis. Data are shown as means ± SD. Statistical analysis was performed with Student's t test for means from two groups. ANOVA test was used for ANOVA between several groups. χ² test was used to compare the relationship between the expression of FOXP3 and p21.

p21 is up-regulated after FOXP3 induction and contributes to its tumor suppressor activity. We used the MCF-7 cell lines engineered to express FOXP3 in the absence doxycyclin. The cells cultured in the presence or absence of doxyclin for 48 hours were compared with five independent RNA isolates in each group by gene array analysis. A summary of the gene array data, depicting genes that are induced by >2-fold, is shown in Fig. 1A. The full data set is shown in Supplementary Tables S1 and S2, and the raw data are deposited to MIAExpress (accession no. E-MTAB-73).

Among the FOXP3-induced genes are several tumor suppressors, including p18, p21, LAT2, and ARHGAPS (Fig. 1A). We have chosen p21 as the prototype to study the mechanism by which FOXP3 activates tumor suppressors as the relevance of defective p21 in breast cancer is well established. In addition, real-time PCR showed induction of p18 is <2-fold (data not shown). We first used real-time PCR and Western blot to confirm the induction of p21 after the inducible expression of FOXP3. As shown in Fig. 1B, p21 transcripts were induced by 7-fold in the MCF-7-pBI-FOXP3/green fluorescent protein (GFP) cell line after removal of doxycyclin, but not the MCF-7-pBI-GFP/control cell line under the same culture condition. Western blot analysis confirmed that accumulation of p21 protein followed that of FOXP3 (Fig. 1C). To determine whether induction of p21 contributed to tumor suppression, we transfected the MCF-7 cell lines with either control vector or p21 shRNA. The transfectants were cultured in the absence of doxycyclin for 10 days and were stained by crystal violets. As shown in Fig. 2B, p21 shRNA specifically increased the number of colonies in the cell line that expressed FOXP3, but barely so for those that expressed GFP. Microscopically, the sizes of colonies were usually larger in the shRNA group, even for those that expressed GFP only, consistent with the notion that endogenous p21 in the MCF-7 cells limited its growth potential (Fig. 2A). Even in the p21-silenced group, FOXP3 transfection still reduces the number of colonies by nearly 60%, which is consistent with the contribution of other FOXP3 targets, including those that we have reported recently (22, 23). Nevertheless, the partial restoration of the colonies indicated that p21 induction contributes to the tumor suppressor activity of the FOXP3 gene.

FOXP3 maintains p21 levels in normal mammary epithelial cells. An important issue is whether FOXP3 expression contributes to expression of p21 in normal mammary epithelial cells. As shown in Fig. 2C, the FOXP3 protein can be identified by Western blot in immortalized human mammary epithelial cell line MCF-10A. To determine the role of FOXP3 in p21 expression, we used FOXP3 shRNA to silence FOXP3 expression and measured the levels of p21 transcripts. As shown in Fig. 2D, FOXP3 silencing caused a 5-fold to 10-fold reduction of the p21 transcripts, which revealed a critical role for FOXP3 in maintaining p21 expression in mammary epithelial cells. A similar effect was observed when FOXP3 was silenced in the early passage of primary human mammary epithelial culture (Supplementary Fig. S1).

To test the role of FoxP3 in p21 expression in vivo, we microdissected mammary epithelium from 2-month-old Rag2^−/−FoxP3^sf/sf and Rag2^−/−FoxP3^+/+ mice. The amounts of p21 transcripts were determined by real-time PCR. As shown in Fig. 3A, the FoxP3 mutation caused ∼6-fold reduction in p21 transcripts. Perhaps due to nonmediated decay caused by frameshift mutation, mammary epithelia from FoxP3 mutant mice lacked FoxP3 transcripts. Correspondingly, dramatically increased numbers of breast epithelial cells in the Rag2^−/−FoxP3^sf/sf mice have entered the cell cycle, as judged by Ki67 staining (Fig. 3B). H&E staining of the mammary tissue indicated extensive ductal hyperplasia in the Rag2^−/−FoxP3^sf/sf mice (Fig. 3C). These data showed that the FoxP3 mutation leads to reduced p21 expression and increased proliferation of normal epithelium in vivo. Because the young mice had yet to develop mammary tumor at this age, down-regulation of p21 is not due to the secondary effect of malignant transformation.

Correlation between expressions of FOXP3 and p21 in human breast cancer. The majority of breast cancer samples lack p21 and FOXP3 expression (5–7, 22, 23). An important issue is, therefore, whether the expression of the two genes is interrelated among human breast cancer samples. To address this issue, we analyzed 62 cases of breast cancer samples in TMA for expression of FOXP3. As shown in Fig. 4, among the FOXP3⁺ samples, 66% are also p21⁺. In contrast, only 30% of the FOXP3⁻ samples expressed p21. The strong correlation between FOXP3 and p21 expression suggests that FOXP3 down-regulation may be an important factor for the lack of p21 among breast cancer tissue.

Specific binding of FOXP3 to the p21 locus is essential for activation of p21. Two p21 mRNA isoforms (1, NM_078467 and 2, NM_000389) have been reported with different exon 1–exon 2 junctions. To properly align the genomic structure of the locus, we sequenced the p21 RNA from the MCF-7 cells after the induction of FOXP3. As shown in Supplementary Fig. S2, only isoform 2 was produced in FOXP3-transfected MCF-7 cells. This allowed us to assign the position of intron 1 for the p21 locus. As illustrated in Fig. 5A (top), a large number of forkhead binding motifs RYMAAYA (29, 30) and TRTKTRC (refs. 31, 32; R = A, G; M = A, C; Y = C, T; K = G, T) can be identified throughout the p21 gene. To identify the sites that bind to FOXP3, we induced FOXP3 by culturing the cell line in the absence of doxycyclin and then used ChIP to determine whether FOXP3 interact with the p21 locus. To normalize the efficiency of PCR primers, the products were compared with input DNA amplified by the same primers. As shown in Fig. 5A (middle), quantitative analysis show that peak binding activity localized at the forkhead/HNF-3 binding motif at 0.2 kb 3′ of TSS. Low but detectable levels of DNA are observed over an 8-kb fragment, which could be due to either the low resolution of ChIP or the existence of multiple weaker binding sites. To confirm the specific requirement for FOXP3 for the signal at 0.2 kb, we also compared the signal to uninduced pBI-FOXP3/GFP cell lines and pBI-GFP control cell lines cultured in the presence or absence of doxycyclin. As shown in Fig. 5A (bottom), the p21 region is precipitated if, and only if, FOXP3 was induced.

To directly show the function and specificity of the FOXP3-mediated induction of p21, we first produced three constructs consisting of overlapping fragments of the 5′ of the p21 locus (Fig. 5B). Using a dual-luciferase assay, we found that FOXP3-mediated induction of p21 requires sequences that are both 5′ and 3′ to TSS, with the maximal activity requiring −540 and +365 bp at 5′ and 3′, respectively. Further extension at 3′ significantly reduced the p21 induction (Fig. 5B). We, therefore, used the optimal reporter to confirm the function of the forkhead binding site at the 0.2 kb at 3′ of TSS. As shown in Fig. 5C, whereas wild-type (WT) reporter is induced by FOXP3 expression, mutation of the forkhead binding site abrogated induction. These data showed that the specific cis-element is essential for FOXP3-mediated activation of p21 locus.

It has been shown that c-Myc can target the p21 promoter and inhibit its expression (33–35). To determine whether the FOXP3 gene regulates p21 directly, we measured the effects of FOXP3 on the p21 promoter activity in the c-Myc knockout cell line. As shown in Fig. 5D, the promoter activity of p21 was significantly induced by FOXP3 in c-Myc knockout cells. The relatively low induction, compared with HEK 293 cells, is likely due to the drastically reduced transfection efficiency of the Myc-deficient cell line.⁴

Localized chromatin modification as a mechanism for FOXP3-induced expression of p21. Recent studies showed that FOXP3-mediated induction of gene expression is associated with histone acetylation (36). We, therefore, used anti–acetyl-H3 antibodies to monitor local chromatin changes associated with FOXP3 binding. As shown in Fig. 6A and B, in cells expressing FOXP3, H3 acetylation in the +0.2-kb site of p21 was increased by >2-fold. The increase in the neighboring areas mirrored what was observed with FOXP3 binding. These data showed that FOXP3 enhances the H3 acetylation of p21, especially at the 0.2-kb region. We carried out ChIP analysis using antibodies specific for HDAC1 to HDAC7. The MCF-7 cells with or without FOXP3 induction were compared. As shown in Fig. 6B, a generalized reduction of HDAC association to the p21 locus was observed after FOXP3 induction. However, by far, the strongest effect was observed at the 0.2-kb site, wherein FOXP3 associates with the p21 locus. Moreover, although a reduction of HDAC1 to HDAC7 was observed after FOXP3 binding, the most significant reduction was observed on HDAC2 and HDAC4, as these two HDACs showed the strongest association before FOXP3 induction.

To determine whether HDAC2 and HDAC4 are involved in p21 up-regulation, we used shRNAs to modulate their expression. As shown in Fig. 6C, two independent shRNAs were specifically silencing the expression of either HDAC2 or HDAC4. Correspondingly, the levels of p21 transcripts were increased by 2-fold to 3-fold after the silence of either gene. Note that some of the p21 protein induced by HDAC shRNAs had a molecular weight of 15 kDa rather than 21 kDa. This is likely due to the cleavage of p21 associated with a general increase in histone acetylation, as reported by others (37). To test if HDAC2 and HDAC4 are necessary mediators of p21 induction by FOXP3, we transfected FOXP3 into MCF-7 cells, in which HDAC2 and HDAC4 are silenced, and compared the levels of p21 in either FOXP3-tracnsfected or vector control–transfected cells. As shown in Fig. 6C, FOXP3-mediated induction of p21 is abrogated in the HDAC4-silenced cell lines. These data further support the notion that FOXP3-mediated induction of p21 is mediated by the disruption of HDAC2-mediated and HDAC4-mediated repression.

Mechanism of p21 regulation in normal and cancerous epithelial cells may hold the keys to the molecular mechanism of carcinogenesis. Here, we present several lines of evidence demonstrating a critical role of p21 as a downstream target of FOXP3, and its expression contributes to the FOXP3-mediated growth inhibition of a breast cancer cell line.

First, in confirming the cDNA microarray data, we showed that inducible expression of FOXP3 induced the p21 transcripts and protein in breast cancer cell line MCF-7. The induction is mediated by transcription regulation, as it is reflected in luciferase assay. ChIP analysis revealed that a specific site at 0.2 kb downstream of TSS is necessary for FOXP3-mediated induction by FOXP3. Moreover, the induction is not an artifact of FOXP3 overexpression, as shRNA silencing of the FOXP3 gene leads to a dramatic reduction of p21 in primary mammary epithelial cells.

Second, to determine whether induction of p21 contributes to growth inhibition of the tumor cell line, we tested whether blunting p21 induction by shRNA abrogate growth inhibition by FOXP3. Our data showed the significant, albeit incomplete, rescue of FOXP3-mediated growth inhibition. The significant rescue shows an important role of p21 induction in FOXP3-mediated growth inhibition in MCF-7 cell line. Other recent studies indicate that FOXP3 also inhibits growth by repressing the expression of HER-2 and SKP2 (22, 23). Thus, depending on tumor cell lines used, FOXP3-mediated inhibition of oncogenes and induction of tumor suppressor may work either independently or in concert to cause growth inhibition of breast cancer cell line.

Third, our analysis of 62 cases of breast cancer samples showed a significant correlation between expression of FOXP3 and p21. Nevertheless, not unlike other tumor suppressor targets, there was no 1:1 correlation between expression of FOXP3 and p21. For instance, ∼30% of cases that stained positive for FOXP3 still lack detectable p21. This is, in part, due to the fact that nearly one third of breast cancer samples show somatic missense mutation of FOXP3 (22, 23). Conversely, nearly one third of the FOXP3 negative tumor cells still express p21. This can be due to either the false-negative staining of FOXP3, perhaps relating to the quality of tumor tissues, or the levels of FOXP3 expression in the first place. In addition, because p53 can induce p21 expression, it is possible that the p21 expression in FOXP3-negative tumor samples was due to functional p53. The limited sample set used in this study cannot distinguish these possibilities. Regardless of how the discrepancies are explained, the positive association between p21 and FOXP3 in clinical samples, when viewed in the context of the data in mice with FoxP3 mutation and the in vitro analysis of normal and malignant tumor cells, made a compelling case that FOXP3 is a major regulator for p21 expression in breast cancer. Recent studies revealed an interesting role of p21 loss and tamoxifen-stimulated growth of breast cancer (9). It is of great interest to determine whether genetic lesion to FOXP3 may account for the p21 loss and, therefore, the unusual response to a widely used drug.

Finally, whereas a number of studies have addressed the mechanism of FOXP3-mediated gene repression, the mechanism by which FOXP3 directly induces gene expression remained largely obscure. A recent report showed association between FOXP3-induced gene activation and histone acetylation (36), although the mechanism and significance of such acetylation have not been addressed. Our data showed that FOXP3 binding to a specific site in intron 1 of p21 increased histone H3 acetylation by reducing binding of HDAC4 and HDAC2 to the same site (Fig. 6E). Gene silencing with shRNA confirmed the significance of these two HDACs in p21 expression, although FOXP3 did not repress expression of either HDAC2 or HDAC4 (Supplementary Fig. S4). Therefore, our data provide a novel mechanism for FOXP3-mediated transcription activation. Because FOXP3 has been shown to recruit histone acetyl transferases (38), it is of interest to investigate whether this interaction contributes, either directly or indirectly, to increased H3 acetylation in the p21 locus.

Taken together, our data showed p21 as a downstream target for FOXP3, the first X-linked tumor suppressor in breast cancer. Because p21 serves as an important target for all major tumor suppressor genes of breast cancer and because irreversible genetic lesion to p21 is relatively rare, it might be possible to reactivate p21 in cancer by inducing FOXP3. Whereas p21 induction can be achieved by a general silencing of HDAC2 and HDAC4, the induced p21 are rapidly degraded, presumably due to the simultaneous induction of other proteins involved in p21 cleavage (37). On the other hand, our data showed that p21 induced by FOXP3 remained intact and mediates tumor suppression. Therefore, reactivating FOXP3 may prove to be a more relevant approach.

No potential conflicts of interest were disclosed.

Grant support: NIH grant CA120910 (Y. Liu), Department of Defense grant W81XWH08-1-0537 (Y. Liu), and American Cancer Society grant RSG-06-072-01-TBE (P. Zheng).

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.