FoxM1 Promotes Stemness and Radio-Resistance of Glioblastoma by Regulating the Master Stem Cell Regulator Sox2

Abstract

Glioblastoma (GBM) is the most aggressive and most lethal brain tumor. As current standard therapy consisting of surgery and chemo-irradiation provides limited benefit for GBM patients, novel therapeutic options are urgently required. Forkhead box M1 (FoxM1) transcription factor is an oncogenic regulator that promotes the proliferation, survival, and treatment resistance of various human cancers. The roles of FoxM1 in GBM remain incompletely understood, due in part to pleotropic nature of the FoxM1 pathway. Here, we show the roles of FoxM1 in GBM stem cell maintenance and radioresistance. ShRNA-mediated FoxM1 inhibition significantly impeded clonogenic growth and survival of patient-derived primary GBM cells with marked downregulation of Sox2, a master regulator of stem cell phenotype. Ectopic expression of Sox2 partially rescued FoxM1 inhibition-mediated effects. Conversely, FoxM1 overexpression upregulated Sox2 expression and promoted clonogenic growth of GBM cells. These data, with a direct binding of FoxM1 in the Sox2 promoter region in GBM cells, suggest that FoxM1 regulates stemness of primary GBM cells via Sox2. We also found significant increases in FoxM1 and Sox2 expression in GBM cells after irradiation both in vitro and in vivo orthotopic tumor models. Notably, genetic or a small-molecule FoxM1 inhibitor-mediated FoxM1 targeting significantly sensitized GBM cells to irradiation, accompanying with Sox2 downregulation. Finally, FoxM1 inhibition combined with irradiation in a patient GBM-derived orthotopic model significantly impeded tumor growth and prolonged the survival of tumor bearing mice. Taken together, these results indicate that the FoxM1-Sox2 signaling axis promotes clonogenic growth and radiation resistance of GBM, and suggest that FoxM1 targeting combined with irradiation is a potentially effective therapeutic approach for GBM.

Fig 1. FoxM1 knockdown increases, and overexpression decreases, differentiation-associated marker expression in primary GBM cells.

(A) Representative immunofluorescence images stained with anti-FoxM1 antibody in NS07-448 cells that were cultured under sphere culture condition or 5% serum culture condition. Scale bar = 20 μm. (B) Western blotting analysis of FoxM1, Sox2, and GFAP (astroglial differentiation marker) using nuclear (N) or cytoplasmic (C) protein lysates from sphere cultures and serum cultures. As loading controls, α-tubulin (cytoplasm) and TBP (TATA-box binding protein, nuclear) were used. (C) FoxM1 expression in FoxM1-knockdown cells. β-actin was used as a loading control. (D) Differentiation marker expression in the control and FoxM1-shRNA expressing cells under differentiation-inducing conditions (0.1% serum culture condition for 7 days). Cells were stained with antibodies for neural cell differentiation markers (GFAP, astroglial marker; O4, oligoglial marker; TuJ1 and NeuN, neuronal marker). Scale bar denotes 100 μm. (E) FoxM1 expression in FoxM1-overexpressing cells. (F) Differentiation marker expression in the control and FoxM1-overexpressing cells under differentiation-inducing conditions. Ectopic expression of FoxM1 was achieved by electroporation with a FoxM1 expression vector. The cells were stained with the same markers used in (D). Scale bar, 50 μm.

Fig 2. FoxM1 knockdown decreases and Sox2 overexpression partially rescues clonogenic growth of GBM cells.

(A) FACS analysis of GBM cells to determine co-localization of FoxM1 and Sox2. (B) Four different GBM cells stained with anti-FoxM1 and anti-Sox2 antibodies were analyzed by flow cytometry. (C) Sox2 mRNA expression in Sox2-overexpressing cells. GBM11-096 GBM cells were transduced with lentivirus either expressing the mock control (NT) or Sox2. Expression levels of Sox2 mRNA were determined by real-time PCR. (D) Sphere formation limiting dilution assays using GBM cells with FoxM1 knockdown, Sox2 overexpression, or both. (E) Estimated frequencies of clonogenic cells in each group were determined by ELDA analysis.

Fig 3. FoxM1 binds to Sox2 promoter region and regulates its expression.

(A) Diagram of potential FoxM1-binding sites on Sox2 promoter region. RNAPII was used as a positive control. (B) CHIP-PCR analysis to determine the FoxM1 binding site on Sox2 promoter. Lysates of NS07-448 cells were incubated with IgG, FoxM1 or RNAP II (RNA polymerase II) and processed for CHIP-PCR. (C) CHIP-PCR analysis using NS07-448 and 387 cells treated with a small-molecule FoxM1 inhibitor Siomycin A (2 μM) for 24 hours. (D) Quantitation of PCR bands shown in (C).

Fig 4. FoxM1 and Sox2 are upregulated in GBM cells after irradiation.

(A) Western blot analysis of FoxM1 and Sox2 using the irradiated NS07-448 (left) and 387 (right) GBM cells. Cells were harvested with course of time after irradiation. β-actin was used as a loading control. (B) Representative IF images of NS07-448 GBM cells with or without irradiation using anti-FoxM1 and anti-Sox2 antibodies. Scale bar, 20 μm. (C and D) Western blot analysis of FoxM1 and Sox2 in GBM cells after irradiation. Cytoplasmic (C) and nuclear (N) fractions of protein lysates were prepared to determine FoxM1 and Sox2 changes 6 and 12 hours after irradiation. As loading controls, α-tubulin (cytoplasm) and TBP (nuclear) were used. Quantitation of these protein bands were shown in (D).

Fig 5. FoxM1 knockdown sensitizes GBM cells to irradiation.

(A) LDA analysis of the control and FoxM1 knockdown NS07-448 GBM cells with or without in vitro irradiation. (B) Change in the clonogenicity of GBM cells by FoxM1 knockdown and/or in vitro irradiation was evaluated by the limiting dilution assay. Sphere frequencies were calculated using ELDA. (C) Western blot analysis of FoxM1, Sox2 and apoptosis markers (cleaved PARP and Bax) in GBM cells with FoxM1 knockdown and irradiation treatments. β-actin was used as a loading control. (D) Cell cycle analysis of GBM cells with FoxM1 knockdown, irradiation, or both. Cell cycle distribution was determined by flow cytometry. P-values were calculated using the Fisher’s exact test (p<0.001). (E) Representative TUNEL assay images from the cell groups (top). Scale bar, 100 μm. Quantitation of TUNEL-positive apoptotic cells in each group (bottom). *** and *, p<0.001 and <0.05 compared to NT shRNA, respectively; #, p<0.05 compared to NT shRNA+RT; +++, p<0.001 compared to shFoxM1.

Fig 6. FoxM1 knockdown combined with in vivo irradiation significantly prolonged the survival of tumor bearing mice.

(A) Kaplan-Meier survival curves of orthotopic tumor bearing mice. Control or FoxM1 expressing 387 GBM cells were intracranially injected into the brain of mice. Two weeks later tumor implantation, in vivo irradiation directed to the brains was performed (2 Gy daily does for 5 days). (B) Summary of survival and statistical significance. (C) Representative immunohistochemical images using FoxM1 and Sox2 antibodies on xenograft tumor sections. The brain sections were paraffin-embedded and stained with anti-FoxM1 and anti-Sox2 antibodies. H&E staining was used to visualize tissue histology.

Fig 7. FoxM1 expression levels in Clinical glioma specimens and association with patient survival.

(A) Representative images of FoxM1 staining analysis using tissue microarray (TMA) sections. FoxM1 protein level of each glioma section was quantified according to the IRS method. (B) FoxM1 protein levels of each WHO glioma grade were compared using the IRS score (High = 4–12, Mild = 1–3, Low = 0). P-values were calculated using the Fisher’s exact test. (C and D) Kaplan-Meier survival curves and statistical analysis of glioma patients that were categorized based on the level of FoxM1. FoxM1 negative/low (blue) and high (red) groups were plotted. Overall survival (OS, left) and Progression-free survival (PFS, right) were compared by the Log rank test.