FoxM1 regulates mammary luminal cell fate

Abstract

Elevated expression of FoxM1 in breast cancer correlates with an undifferentiated tumor phenotype and a negative clinical outcome. However, a role for FoxM1 in regulating mammary differentiation was not known. Here, we identify another function of FoxM1, the ability to act as a transcriptional repressor, which plays an important role in regulating the differentiation of luminal epithelial progenitors. Regeneration of mammary glands with elevated levels of FoxM1 leads to aberrant ductal morphology and expansion of the luminal progenitor pool. Conversely, knockdown of FoxM1 results in a shift toward the differentiated state. FoxM1 mediates these effects by repressing the key regulator of luminal differentiation, GATA-3. Through association with DNMT3b, FoxM1 promotes methylation of the GATA-3 promoter in an Rb-dependent manner. This study identifies FoxM1 as a critical regulator of mammary differentiation with significant implications for the development of aggressive breast cancers.

Figure 1. FoxM1 expression in tumor and normal tissue

(A) Oncomine, a publicly available microarray database was used to analyze FoxM1 expression in 200 samples of invasive ductal carcinoma. Samples were organized by grade and fold-change of FoxM1 RNA from normal was graphed using a box plot *p<10⁻⁶. (B) Representative immunohistochemistry of FoxM1 in normal mammary tissue as well as, grade 1, grade 2, and grade 3 breast carcinomas are shown. Scale bar represents 10μm. (C) Levels of FoxM1 RNA were determined by semi-quantitative RT-PCR. All samples were collected from inguinal mammary glands at various developmental stages: 5 weeks (puberty), 8 weeks (virgin adult), P6, P18 (early and late pregnancy), L1, L5, 10 (lactation), and I6 (involution). 4–7 mice were used for each stage. (D) Glands from each stage were sectioned and stained for FoxM1 expression using DAB and Hematoxylin counterstain. (E) Mammary glands from 8-week-old C57BL/6 mice were digested and stained for CD24, CD29, and CD61. Stem cells, luminal progenitors, and luminal cells were sorted. Cytokeratin 18, smooth muscle actin, and c-kit were used to assess the purity of the populations. RT-PCR shows relative expression of CK18, SMA, and FoxM1. Data is normalized to the stem cell population *p<10⁻⁴ **p<0.05.

Figure 2. FoxM1 loss results in defects in mammary development during pregnancy

(A) H&E staining of mammary gland sections from each time-point is shown. Control glands expressing only WAP-Cre are shown on right and glands lacking FoxM1 are shown in adjacent panels. A lower magnification of pregnancy day 6 glands is shown to better demonstrate details. Scale bar represents 10μM. (B) Representative carmine alum whole mount analysis of mice expressing WAP-Cre and either wildtype FoxM1 or FoxM1 FL/FL. Pictures taken at 3× magnification are shown in black and white to increase the clarity. 4–8 mice from both genotypes were analyzed at every stage. (C) Western blot of FoxM1, α-Casein, and β-Casein are shown. α-tubulin is provided as a loading control. (D) Immunohistochemistry using an antibody against mouse milk is shown in the lower panel. Staining is done using DAB with a hematoxylin counterstain.

Figure 3. FoxM1 deletion leads to an expansion of differentiated luminal cells

(A) Four 8- week-old WAP-rtTA-Cre and FoxM1 FL/FL, WAP-rTA-Cre mice were treated with doxycycline (2mg/ml) in their drinking water for five days. Stem, luminal progenitor, and luminal cells were sorted and FoxM1 expression was analyzed by RT-PCR *p<0.01 **p<10⁻³. (B) Whole mount of inguinal mammary glands 15 days after doxycycline treatment were visualized with carmine alum stain. Higher magnification (3X) is shown with lower magnification in the inset (1X). Images are shown in black and white in order to increase the clarity. (C) Hematoxylin and eosin staining as well as immunohistochemistry of FoxM1, cytokeratin 18, estrogen receptor alpha, and milk are shown after 15 days of treatment. Scale bar represents 10μm. (D) Flow cytometry analysis of stem cells, luminal progenitors, and differentiated luminal cells. The plots show CD29 and CD61 expression of cells that stained positive for CD24. Percentage of each cell type for both genotypes is shown below *p<0.04 **p<0.05 ***p<0.03. (E) RT-PCR of estrogen receptor alpha, amphiregulin, cytokeratin 18, and cadherin 11 are shown normalized to 18s RNA *p<10⁻³, **p<0.05, ***p<0.01.

Figure 4. Over-expression of FoxM1 in mammary gland results in an expansion of progenitors and a loss of differentiation markers

(A) Schematic representation of experimental design. (B) GFP whole mount imaging of mammary glands. Boxed areas are shown in the inset at lower magnification. (C) Hematoxylin and eosin and immunohistochemistry of GFP and FoxM1-GFP glands. Smooth muscle actin, cytokeratin 18, and estrogen receptor alpha immunostaining is shown. Arrows in SMA section are showing displaced SMA+ cells surrounded by luminal cells. Representative sections from six mice are shown. Scale bar represents 10μm. (D) CD61 immunohistochemistry is shown. Enlarged images of GFP and GFP-FoxM1 mice are displayed in the right panel. Arrowhead shows a cluster of CD61+ cells. (E) Analysis of mammary stem cells, luminal progenitor, and luminal cell pools was performed in GFP or FoxM1-GFP expressing mice. Glands were digested to generate single cell suspensions, stained, and examined by flow cytometry. Representative dot plots of CD24+ cells are shown with percentages listed in each box. The bottom panel provides quantification from four mice. The percentage of each population is shown relative to the GFP control in the same animal *p<0.03 **p<0.01 ***p<0.001. (F) RT-PCR analysis of GFP and GFP-FoxM1 glands for expression of estrogen receptor alpha, cytokeratin 18, amphiregulin, and cadherin 11 *p<10⁻⁴ **p<0.001 ***p<0.05.

Figure 5. FoxM1 is a negative regulator of GATA-3 in vivo

(A) Western blot analysis of FoxM1 and GATA-3 protein levels in WAP-rtTA-Cre, FoxM1 FL/+ (control) and WAP-rtTA-Cre, FoxM1 FL/FL as well as GFP (control) and GFP-FoxM1 expressing animals are shown in the top panel. Alpha tubulin is shown as a loading control. RT-PCR expression of glands is shown below, GATA-3 expression is normalized to 18S RNA. *p<0.05, **p<10⁻⁶ (B) Immunohistochemical staining of GATA-3 expression using DAB and hematoxylin counterstain. (C) Binding of FoxM1 to the promoter of GATA-3 was analyzed using an in vivo ChIP experiment. FoxM1 antibody was used to immunoprecipitate FoxM1 in glands of C57BL/6 mice. RT-PCR was used for three different regions of the GATA-3 promoter that have putative FoxM1 binding sites. Graph displays relative binding over an IgG control *p<10⁻⁵ **p<0.003 ***p<0.001. (D) Graph summarizing flow cytometry data from control, GATA-3, FoxM1, and FoxM1-GATA-3 expressing mice. Each group contains four mice and percentage of each cell type is graphed. p-values are calculated as compared to control animals *p<0.05, **p<0.01. Semi-quantitative RT-PCR of GATA-3 and FoxM1 expression is shown in graphs to the right. (E) Flow cytometry analysis of glands regenerated with either WT or FoxM1 FL/FL, WAP-tTA-Cre and either shRNA control or GATA-3 targeting shRNA. All mice were fed doxycyline for 14 days following the surgery. Glands were analyzed after 8 weeks. Data collected from five mice is shown in the graph. N.S. (not significant), *p<0.05, **p<0.001, ***p<0.01. Semi-quantitative RT-PCR of FoxM1 and GATA-3 expression is shown in graphs to the right.

Figure 6. FoxM1 transcriptional repression of GATA-3 is methylation dependent

(A) FoxM1 and GATA-3 expression in human breast cancers. Fold change from normal is graphed. Heat map of individual samples is shown above the graphs *p<10⁻³ **p<10⁻⁵ ***p<10⁻¹¹ (B) MDA-MB-453 ells were treated with either control or FoxM1 targeting siRNA. Chromatin immunopreciptation assay of FoxM1 binding to the GATA-3 promoter was performed. Quantitative PCR results are shown *p<0.01. (C) FoxM1 was transfected into MDA-MB-453 cells and 4 hours later, either vehicle (PBS) or 1uM of 5′azacytidine was added to each plate. Samples were collected 48 hours later and RT-PCR of GATA-3 expression is shown as normalized to GAPDH *p<0.01. (D) Cells were collected and immunopreciption was performed using DNMT3b antibody, western blot of FoxM1 is shown. (E) 72 hours after transfection with control siRNA or FoxM1 targeting siRNA cells were fixed and chromatin immunoprecipitated (ChIP) with DNMT3b antibody or IgG was performed. Relative binding of DNMT3b to the GATA-3 promoter sites are graphed. Samples have been normalized to IgG and relative binding is shown *p<0.01, **p<0.05.

Figure 7. Methylation by FoxM1 is Rb-dependent

(A) Tet-off shRNA cell lines were either treated with doxycycline or vehicle for 14 days. Control or FoxM1 constructs were transfected into lines expressing Rb or where Rb was silenced by shRNA. RT-PCR of GATA-3 expression is shown normalized to GAPDH *p<0.05 **p<0.001. (B) Cells were transfected with control or FoxM1 targeting siRNA. After 72 hours, cells were collected and fixed for ChIP assay. IgG or Rb antibody was used for immunoprecipitation. RT-PCR of Rb binding to the GATA-3 promoter is shown *p<0.05 **p<10⁻⁴. (C) Tet-off shRNA cell lines were used for methylation specific PCR analysis of the GATA-3 promoter in the presence and absence of FoxM1 expression *p<10⁻⁵, **p<0.01, ***p<10⁻³. (D) Flow cytometry of stem cells, luminal progenitors, and differentiated cells from mice expressing scrambled shRNA, Rb-targeting shRNA, FoxM1, or both FoxM1 and Rb-targeting shRNA. Panel to the right shows semiquantitative RT-PCR of FoxM1, GATA-3 and Rb expression. Cyclophilin is shown as a loading control *p<10⁻⁴ **p<0.01.