Spontaneous and carcinogen-induced mammary cancer in Foxp3sf/+ female mice
The mutant BALB/c mice we used for the initial study carried mutations in two closely linked X-chromosome genes, Foxp3sf
. During the course of the study, a spontaneous segregation of Otcspf
allowed us to obtain a BALB/c Otcspf/+
strain. Meanwhile, we obtained an independent line of Scurfy mice that had never been crossed to the Spf
mutant mice and we backcrossed the Scurfy
mutant allele (Foxp3sf
) for more than 12 generations into the BALB/c background (Chang et al., 2005
). Female mice with only one copy of the Foxp3
gene survived to adulthood and appeared normal within the first year of life (Godfrey et al., 1991
) with normal T cell function (Fontenot et al., 2003
; Fontenot et al., 2005
; Godfrey et al., 1994
). Our extended observations of the retired breeders for up to two years revealed that close to 90% of the Foxp3sf/+Otcspf/+
mice spontaneously developed malignant tumors. Cancer incidences in the littermate controls and a line of congenic mice with a mutation in Otc
but not Foxp3
were comparable with each other (). About 60% of the tumors were mammary carcinomas (), although other tumors, such as lymphoma, hepatoma, and sarcoma were observed. Histological analyses revealed lung metastasis ( lower panels, based on expression of ER and/or PR, data not shown) in about 40% of the mice with mammary cancer. More than a third of the tumor-bearing mice had multiple lesions in the mammary glands. Most, although not all, mammary carcinomas expressed the estrogen receptor (ER+
, 14/18) and progesterone receptor (PR+
Fig. 1 Increased susceptibility to breast cancer in mice heterozygous for Foxp3sf. A. Representative breast cancers developed in female Foxp3sf/+Otcspf/+ mice. The top panel shows the gross anatomy while the lower panel shows the histology of local and metastatic (more ...)
In order to focus on mammary cancer, we treated the mice with a carcinogen, 7,12-dimethylbenz [a] anthracene (DMBA), in conjunction with progesterone. Mice heterozygous for Foxp3sf, but not those heterozygous for Otcspf, showed substantially increased susceptibility to mammary cancer, as revealed by earlier onset, increased incidence () and multiplicity (data not shown) of the breast tumors. These data demonstrate that a mutation of Foxp3, but not Otc results in a major increase in susceptibility to mammary carcinoma.
Foxp3 expression in normal and cancerous mammary tissues
Since expression of Foxp3
has not been reported in mammary tissue, we isolated normal and cancerous cells by laser-capture microdissection (Supplemental Fig. S1A
) and compared expression of Foxp3
by real-time RT-PCR and histochemistry. The complete absence of the cd3
transcripts (Fig. S1B
) indicated that the micro-dissected samples were devoid of T cells, the main cell types known to express Foxp3
(Fontenot et al., 2005
). A representative profile and summarized data of Foxp3
expression in Foxp3sf/+Otcspf/+
mice and age-matched WT
control mice are shown in . Foxp3
mRNA was detected in normal mammary epithelium from both the WT and Foxp3sf/+Otcspf/+
mice, but not in mammary cancer cells from the same Foxp3sf/+Otcspf/+
mice. Immunohistochemical staining () confirmed the loss of expression of Foxp3
in the mammary carcinoma generated from the Foxp3sf/+Otcspf/+
Fig. 2 Inactivation of the WT Foxp3 allele in mammary cancer cells. A. Defective Foxp3 expression in breast cancer. RNA extracted from the cells isolated by LCM was subjected to quantitative real-time RT-PCR using primers specific for Foxp3, Hprt and CK19. In (more ...) Foxp3
is an X-linked gene that is subject to X-chromosomal inactivation (Fontenot et al., 2005
). We carried out an anchored RT-PCR and cloned the low levels of Foxp3
mRNA in the breast tissues. We sequenced the cDNA clones from pooled samples after ruling out potential T cell contamination (based on a lack of T-cell specific cd3
transcripts, Fig. S1B
). As shown in , 100% of the Foxp3
transcripts in the cancerous tissues were from the mutant alleles, which indicate that the wild-type allele was silenced in the tumor cells. In contrast, the transcripts from the mutant allele constituted 15% of the transcripts in the normal mammary samples from the same mice. Thus, the expression pattern of Foxp3
fulfills another criterion for a tumor suppressor gene.
FOXP3 is a repressor of ErbB2 transcription
Our characterization of the mammary tumors in the mutant mice revealed wide-spread up-regulation of ErbB2, in contrast to those rare tumors from WT mice, as shown in and supplemental Table S1
. Using real-time RT-PCR, 8–12-fold more ErbB2
mRNA was found in the cancer cells than in normal epithelium (). There was also more ErbB2
mRNA in the Foxp3sf/+spf/+
epithelium than in that of the WT
female mice (), which indicates a potential gene dosage effect of Foxp3
on the regulation of ErbB2
expression in vivo. Transfection of the TSA cell line with Foxp3
cDNA repressed ErbB2
levels on the TSA cell line ().
Fig 3 Foxp3 represses the expression of ErbB2. A. Over-expression of ErbB2 in mouse mammary cancers. The left panels show immunohistochemical staining of a non-cancerous mammary gland and an adjacent adenocarcinoma from a Foxp3sf/+Otcspf/+ mouse using anti-ErbB2 (more ...)
Analysis of the 5′ sequence of the ErbB2 gene revealed multiple binding motifs for the forkhead domain (). To test whether Foxp3 interacts with the ErbB2 promoter, we used anti-V5 antibody to precipitate sonicated chromatin from the TSA cells transfected with the Foxp3-V5 cDNA and used real-time PCR to quantitate the amounts of the specific ErbB2 promoter region precipitated by the anti-V5 antibodies in comparison to those that bound to mouse IgG control. As shown in , the anti-V5 antibodies pulled down significantly higher amounts of ErbB2 promoter DNA than the IgG control, with the highest signal around 1.6 kb 5′ of the transcription starting site.
To test whether the binding correlated with the suppression by Foxp3, we produced luciferase reporter using the 1.8, 1.2 and 0.8 Kb upstream of the ErbB2
TSS and tested the ability of Foxp3 to repress ErbB2
promoter activity. In three separate cell lines, we observed that the region with the strongest ChIP signal was required for optimal repression by Foxp3 (). Furthermore, we deleted two potential Foxp3
-binding sites based on intensity of ChIP signal, abundance of consensus binding sites and conservation between mouse and human (Supplemental Fig. S2
) by site-directed mutagenesis and measured the effect on Foxp3-mediated repression. As shown in , deletion of either binding site substantially increased the ErbB2
promoter activity in the presence of Foxp3 and thus alleviated Foxp3-mediated repression.
Since the region deleted in mut B is 100% conserved between mouse and man and since this deletion completely wiped out repression, we used an electrophoretic mobility shift assay (EMSA) to determine whether the forkhead DNA-binding motifs in region B bound to Foxp3. As shown in , the nuclear extracts from the Foxp3-expressing cells specifically retarded migration of the WT but not mutant 32P-labeled probes compared with control cells. While mutant cold probes did not affect Foxp3 binding activities, WT cold probes significantly diminished them, establishing that the binding of these complexes is specific to forkhead DNA-binding motifs. We therefore carried out site-directed mutagenesis to replace the 12 nucleotides (mut C) within the ErbB2 promoter and compared the promoter activity and Foxp3-repression by luciferase assays. While the wild-type promoter was repressed by Foxp3, no repression by Foxp3 was observed when the mutant promoter was used. Moreover, in contrast to the deletional Mut B, the mutations had no impact on the basal activity of the ErbB2 promoter (). Taken together, our new data make a compelling case that Foxp3 represses the ErbB2 promoter via specific forkhead binding motifs.
FOXP3 defects in human breast cancer
We analyzed the levels and isoforms of the FOXP3
transcripts in a panel of normal human mammary epithelial cells (HMEC), an immortalized but non-malignant cell line (MCF-10A), and 10 malignant breast cancer cell lines differing in ER/PR and HER-2 status. Early passage of HMEC with no methylation in the CpG island of the P16
promoter (supplemental Fig. S3
) was used to avoid effects associated with P16
inactivation in post-senescence HMEC cultures (Romanov et al., 2001
). As shown in , similar levels of FOXP3
transcripts were observed in two independent isolates of HMEC and in the immortalized cell line MCF-10A. Each of the 10 tumor cell lines had a different degree of reduction in FOXP3
mRNA levels in comparison to HMEC and MCF-10A. Among them, 2 were completely devoid of FOXP3 mRNA, while the others had a 1.5–20 fold reduction. We then used anchored primers spanning exons 1–12 to amplify the FOXP3 transcripts, and then we sequenced the PCR products. As shown in , none of the tumor cell lines expressed full-length FOXP3
transcripts. The HMEC expressed the same two isoforms as observed in the T cells, while MCF-10A expressed the exon 3-lacking. The same isoform was also found in 4 tumor cell lines at much lower levels. In addition, 3 tumor cell lines expressed an isoform lacking both exons 3 and 4. The alternative splicing resulted in a frame-shift beginning at codon 70 and an early termination at codon 172. Furthermore, 2 tumor cell lines expressed a FOXP3 isoform lacking exons 3 and/or 8. Exon 8 encodes the leucine-zipper domain that is frequently mutated in IPEX patients (Ziegler, 2006
). Thus, FOXP3 is abnormal in breast cancer cell lines. Consistent with a role for FOXP3
in repressing HER-2
expression, the majority of the breast cancer cell lines had higher levels of HER-2 in comparison to normal HMEC (, lower panel). However, additional changes are also likely required for HER-2
over-expression, as three cell lines did not over express HER-2
even though the FOXP3
transcripts were greatly reduced.
Fig. 4 Characterization of FOXP3 transcripts in primary, immortalized and malignant mammary epithelial cells. Relative levels and isoforms of FOXP3 (the upper panel) and HER-2 (the lower panel) mRNA. After normalizing against endogenous GAPDH, the amounts of (more ...)
We took three approaches to determine whether the findings in the mutant mice and human breast cancer cell lines are relevant to the pathogenesis of human breast cancer. First, we used immunohistochemistry to determine expression of FOXP3 in normal vs. cancerous tissue. As shown in , while more than 80% of the normal breast samples expressed FOXP3 in the nuclei of the epithelial cells, less than 20% of the cancerous tissue showed nuclear staining. Second, we used fluorescence in situ hybridization (FISH) to determine whether the FOXP3
gene was deleted in the breast cancer samples. The minimal common region of deletion was identified using flanking p-telomeric and centromeric clones. Out of 223 informative samples, we observed 28 cases (12.6%) with deletions in any of the three loci. Interestingly, deletion of the FOXP3
locus was found in all of the 28 cases ( and supplemental Table S2
). These data suggest that FOXP3
is likely within the minimal region of deletion in the Xp11 region studied. Although all deletions were heterozygous, the FOXP3 protein was undetectable in 26/28 cases. Thus, it appears that for the majority of the breast cancer samples, LOH alone was sufficient to inactivate the locus, perhaps due to X-chromosomal inactivation. The two cases with both deletion and FOXP3 expression had X-polysomy with 3 and 4 X-chromosomes respectively (Table S2
). Thirdly, we isolated DNA from matched normal and cancerous tissues (50 cases with formalin fixed samples and 15 cases of frozen samples) from patients with invasive ductal carcinoma and amplified all 11 coding exons and intron-exon boundary regions by PCR. Two independent PCR products were sequenced in order to confirm the mutations. Unless the bulk sequencing data were unambiguous (Supplemental Fig S4a & c
), the PCR products were cloned and 5–10 independent clones from each reaction were sequenced (Supplemental Fig. S4b
). Among the formalin fixed samples, we only used the cases in which the normal tissue samples gave unambiguous sequencing data that matched the wild-type FOXP3
sequence. When the cancerous tissues were compared with normal tissues from the same patient, 36% (18/50 formalin-fixed samples and 5/15 frozen samples) showed somatic mutations (Table S3
). Loss of the wild-type allele was found in 6/23 cases (38%) of cancer samples with somatic FOXP3
mutations (See supplemental Fig. 4c
for an example). The other cases had heterozygous mutations (Supplemental Fig. S4a
). Eighteen mutations resulted in the replacement of amino acids. Most are likely to be critical for FOXP3 function, as judged from the pattern of mutation in IPEX patients (Ziegler, 2006
) or in the conserved zinc finger domain that has so far not been implicated ().
Fig. 5 FOXP3 defects in human breast cancer. A. Down-regulation of the FOXP3 protein among human breast cancer cells. Photographs in the top panels show anti-FOXP3 staining of normal and carcinoma tissues from the same patient, with specificity control shown (more ...)
Although most samples had a single mutation of the FOXP3
gene, we did observe two cases with multiple mutations. In the first sample (supplemental Fig S4b
, case 3 in Table S3
), the two mutations occurred in consecutive codons, resulting in two nonconservative replacements of amino acid residues. Clonal analysis revealed that both mutations occurred in the same clone (Fig. S4b
). In the second sample (Table S3
, case 16), three mutations occurred in intron 11. Since this case lacked a WT allele (Supplemental Fig. S4d
), it is likely that all of the mutations occurred in the same allele. The possibility of a mismatch in the cancer and normal samples was ruled out by comparing the normal and cancer samples for polymorphism of two unrelated genes (data not shown).
To directly test whether FOXP3 mutations affect the repressor activity for the HER-2 gene, we chose two representative somatic FOXP3 mutants isolated in the cancer cells and tested their repressor activity for the HER-2 promoter. One mutation (338P>L) resided in the signature forkhead domain which is often mutated in the IPEX patient, while the other double mutation (204C>R205E>K) was from the zinc finger domain that has not been implicated in IPEX patients. As shown in , both mutations significantly reduced the repressor activity of FOXP3. The reduced repression of the HER-2 promoter correlates with a significantly reduced inhibition of HER-2 mRNA ().
Four cases had mutations in introns that may potentially affect RNA splicing. We used laser-guided micro-dissection to isolate normal and cancerous epithelial cells from one case with a mutation in intron 6 (case 23, Supplemental Table S3
). RNA was isolated and tested for the potential effects of the mutation on RNA splicing (using primers on exons 5 and 8) and total FOXP3
transcript, as quantitated by real time PCR using primers spanning exons 10–12. Tissues from another patient with a mutation in exon 7 were used as control. As show in left panel, primers spanning exons 5 and 8 failed to detect FOXP3
mRNA from the cancerous tissue of case No. 23. Furthermore, primers spanning exons 10–12 also failed to detect any FOXP3
transcripts. Substantial levels were detected in the normal epithelial cells of the same patients as well as in normal and cancerous tissues from case No. 22. Since the wild-type allele had been lost in the cancer cells of case No. 23, it is likely that the mutation in intron 6 inactivated FOXP3.
With an intron of 944 nucleotides, a mutation that prevented splicing of intron 6 would cause premature-termination codon-mediated RNA decay, which is operative in the FOXP3
gene (Chatila et al., 2000
FOXP3 defects and HER-2 over-expression
To demonstrate a role for FOXP3
defect in HER-2
overexpression, we first silenced the FOXP3
gene in early passage of primary HMEC (Supplemental Fig. S3
) using a lentiviral vector expressing FOXP3
siRNA. As shown in , the FOXP3
siRNA reduced FOXP3
expression by more than 100-fold while increasing HER-2
mRNA by 7-fold. A corresponding increase in cell surface HER-2 was also observed (). These results implicate FOXP3
as a repressor of HER-2
in human breast epithelial cells.
Fig. 6 FOXP3 is an important HER-2 repressor. A. Silencing of FOXP3 resulted in the up-regulation of HER-2 in primary human mammary epithelial cells (HMEC). Early passage of HMEC were transduced with lentiviral vector for either control sequence or FOXP3 siRNA. (more ...)
Second, since a major mechanism for HER-2 up-regulation in breast cancer is gene amplification (Kallioniemi et al., 1992
), an intriguing issue is whether FOXP3
is capable of repressing HER-2 in cancer cells with an amplified HER-2 gene. We produced a Tet-off line of BT474, a breast cancer cell line known to have HER-2
gene amplification (Kallioniemi et al., 1992
), and transiently transfected it with a pBI-EGFP-FOXP3
- vector. After drug selection, the cells were cultured either in the presence or absence of doxycycline. While the cells cultured with doxycycline did not express FOXP3 (data not shown), removal of doxycycline resulted in induction of FOXP3 in a significant fraction of the cancer cells, which allowed us to compare HER-2 levels in the FOXP3+
cells in the same culture by flow cytometry. As shown in , FOXP3−
cells had about a 5–10-fold higher level of the HER-2 protein on the cell surface in comparison to the FOXP3+
Thirdly, we compared the expression of FOXP3 with HER-2 expression in breast cancer tissues. As shown in , down-regulation of FOXP3 was strongly associated with the over-expression of HER-2, which supports a role for FOXP3 inactivation in HER-2 over-expression in breast cancer. Nevertheless, since many of the FOXP3− cells remained HER-2−, it is likely that dis-regulation of FOXP3 is insufficient for HER-2 up-regulation. On the other hand, since only 3/82 FOXP3+ cancer cells expressed high levels of HER-2, FOXP3 inactivation is likely important for HER-2 up-regulation under most circumstances.
Fourth, we divided breast cancer samples based on their HER-2 gene copy numbers and compared the FOXP3+ and FOXP3− cancer samples for the relative amounts of cell surface HER-2 expression. As shown in , in each of the gene dose categories, FOXP3+ samples had reduced HER-2 scores in comparison to the FOXP3− samples. These results strongly suggest a critical role for FOXP3 in repressing HER-2 expression even in the cases of HER-2 gene amplification.
Fifth, of the 223 informative samples among the 238 that we screened for Xp11.2 deletions, those with deletions encompassing the FOXP3
locus had significantly higher HER-2 scores compared to those without deletions (P=0.03) (Supplemental Table S4
). Likewise, we compared the relative HER-2 scores among the 50 samples in which we had sequenced all FOXP3
exons. As shown in Supplemental Table S5
, the mutations in the FOXP3 gene correlated with higher levels of HER-2 (P=0.0083).
Foxp3/FOXP3 inhibits tumorigenicity of breast cancer cells
To test whether the Foxp3 gene can suppress the growth of breast cancer cells, we transfected the empty vector or the vectors carrying either Foxp3 (mouse or human origin) or Otc cDNA into three breast cancer cell lines, including mouse mammary tumor cell line TSA or human breast cancer cell lines MCF7 (ER+HER-2low, no HER-2 amplification) and SKBr3 (ER-HER-2high with HER-2 amplification). The untransfected cells were removed by a selection with G418. While the vector-transfected cells grew rapidly, the Foxp3-transfected cell lines seldom grew into large colonies. The Foxp3-transfected culture had a drastic reduction in both the size and the number of the drug-resistant colonies. No effect was observed when the Otc cDNA was used ().
Fig. 7 Foxp3 inhibits the growth and tumorigenicity of multiple breast cancer cell lines. A. Breast cancer cell lines MCF-7, SKBr3 and TSA were transfected with equal concentrations of either vector alone (Vector), Foxp3 or Otc cDNA. After 3 weeks of G-418 selection, (more ...)
To test whether the somatic mutations uncovered from cancerous tissues ablated their growth inhibition, we transfected WT and two mutant Foxp3 cDNA into SKBr3 and MCF7 cell lines. As shown in , in both cell lines, the mutants had a greatly reduced ability to suppress tumor growth.
To test whether repression of ErbB2 explains the tumor suppressor activity of the Foxp3 gene in the ErbB2+ cancer cell line, we transfected TSA cells with mouse CMV promoter-driven ErbB2 cDNA cloned into the pcDNA6 vector and evaluated their susceptibility to Foxp3-mediated growth suppression. In this setting, the expression of ErbB2 was resistant to Foxp3-mediated repression (data not shown). If repression of endogenous ErbB2 is critical for Foxp3-mediated tumor suppression, ectopic expression of ErbB2 should alleviate the growth inhibition by Foxp3. As shown in , while the pcDNA6-vector-transfected TSA cells remained susceptible to Foxp3-mediated repression, the ErbB2-transfected TSA cells were completely resistant. In contrast, transfection of c-Myc barely alleviated the growth inhibition by FOXP3 (). These results suggest that Foxp3 suppresses TSA growth by repressing transcription of ErbB2.
We transfected TSA cells with either empty vector or V5-tagged Foxp3 cDNA. The stable transfectant cell lines were selected by G-418. The vector and Foxp3-V5-transfected cell lines were injected into syngeneic BALB/c mice, which were then observed for tumor growth and mouse survival. As shown in , Foxp3-transfectants showed reduced growth in vivo. The mice that received TSA-vector cells became moribund earlier with higher incidence, while about 50% of the mice that received the Foxp3-V5-transfected cells survived more than 7 weeks (). Similarly, Foxp3-transfected 4T1, a mouse mammary cancer cell line with ErbB2 over-expression, also showed reduced tumorigenicity in vivo (data not shown).