Inhibition of Cdk5 induces cell death of tumor-initiating cells

Abstract

Background: Tumour-initiating cells (TICs) account for chemoresistance, tumour recurrence and metastasis, and therefore represent a major problem in tumour therapy. However, strategies to address TICs are limited. Recent studies indicate Cdk5 as a promising target for anti-cancer therapy and Cdk5 has recently been associated with epithelial-mesenchymal transition (EMT). However, a role of Cdk5 in TICs has not been described yet.

Methods: Expression of Cdk5 in human cancer tissue was analysed by staining of a human tissue microarray (TMA). Functional effects of Cdk5 overexpression, genetic knockdown by siRNA and shRNA, and pharmacologic inhibition by the small molecule roscovitine were tested in migration, invasion, cell death, and tumorsphere assays and in tumour establishment in vivo. For mechanistic studies, molecular biology methods were applied.

Results: In fact, here we pin down a novel function of Cdk5 in TICs: knockdown and pharmacological inhibition of Cdk5 impaired tumorsphere formation and reduced tumour establishment in vivo. Conversely, Cdk5 overexpression promoted tumorsphere formation which was in line with increased expression of Cdk5 in human breast cancer tissues as shown by staining of a human TMA. In order to understand how Cdk5 inhibition affects tumorsphere formation, we identify a role of Cdk5 in detachment-induced cell death: Cdk5 inhibition induced apoptosis in tumorspheres by stabilizing the transcription factor Foxo1 which results in increased levels of the pro-apoptotic protein Bim.

Conclusions: In summary, our study elucidates a Cdk5-Foxo1-Bim pathway in cell death in tumorspheres and suggests Cdk5 as a potential target to address TICs.

Figure 1

Expression of Cdk5 in human cell lines and cancer tissue.(A) The immunoblots show levels of Cdk5 in breast epithelial cells MCF10A, and cancer cell lines MDA-MB-231, MCF7 and T24. Actin indicates equal loading (n=3). (B) Representative immunostainings from a human breast cancer TMA show Cdk5 expression in healthy mamma tissue and breast cancer tissues. The table displays the evaluation of Cdk5 expression in tumour tissues (n=198) and healthy tissues (controls, n=7) according to the IRS score (intensity × positivity). (C) Immunoblots show Cdk5 expression in HMLE cells that have (+) or have not (−) undergone EMT. Tubulin indicates equal loading (n=3). (D) Cdk5 activity in HMLE cells that have (+) or have not (−) undergone EMT is shown by phosphorylated histone H1 (p-H1; n=3).

Figure 2

Cdk5 inhibition impairs cancer cell growth.(A) Proliferation of non-targeting (nt) or Cdk5 shRNA-transduced cells is shown (mean±s.e.m., *P<0.001, n=3). Immunoblots of non-targeting (nt) or Cdk5 shRNA-transduced T24 cells for Cdk5 and β-actin (loading control) proof Cdk5 knockdown. (B) Proliferation of MCF10A, MCF7, T24, MDA-MB-231, epithelial and mesenchymal HMLE cells treated with roscovitine for 72 h at indicated concentrations is shown. EC50 values for the various cell lines are indicated. (C) Viability of MCF10A, MCF7, T24, MDA-MB-231, epithelial and mesenchymal HMLE cells treated with roscovitine for 24 h at indicated concentrations is shown. (D) Colony formation of non-targeting (nt) and Cdk5 shRNA T24 cells after 7 days is shown. Bar graph shows quantification (mean±s.e.m., *P<0.001, n=3). (E) Colony formation of T24 cells treated with roscovitine for 24 h before freshly seeding at low density and cultivation for further 7 days is shown. Bar graph shows quantification (mean±s.e.m., *P<0.001, n=3).

Figure 3

Cdk5 inhibition reduces cancer cell motility.(A) FCS-induced migration of non-targeting (nt) or Cdk5 shRNA-transduced T24 cells is shown (mean±s.e.m., *P<0.001, n=3). (B) FCS-induced invasion of T24 cells transduced with non-targeting (nt) or Cdk5 shRNA is shown (mean±s.e.m., *P<0.001, n=3). (C) FCS-induced migration of T24 cells with/without treatment with roscovitine is shown (mean±s.e.m., *P<0.001, n=3).

Figure 4

Cdk5 regulates sphere formation and tumour establishment.(A) Tumorsphere formation of non-targeting (nt) and Cdk5 shRNA T24 cells is shown (mean±s.e.m., *P<0.05, n=3). Immunoblots of non-targeting (nt) or Cdk5 shRNA-transduced T24 cells for Cdk5 and β-tubulin (loading control) proof Cdk5 knockdown. (B) Tumorsphere formation after pretreatment of T24 cells with roscovitine for 24 h before resuspension in fresh sphere-formation medium and cultivation for further 10 days in presence of roscovitine is shown (mean±s.e.m., *P<0.001, n=3). (C) Tumorsphere formation after pretreatment of MCF7 cells with roscovitine for 24 h before resuspension in fresh sphere-formation medium and cultivation for further 10 days in presence of roscovitine is shown (mean±s.e.m., *P<0.05, n=3). (D) Sphere formation of non-tumorous MCF10A cells overexpressing empty vector (ev) or Cdk5/p35 (Cdk5) is shown (mean±s.e.m., *P<0.05, n=3). The immunoblot of MCF10A cells overexpressing either empty vector (ev) or Cdk5/p35 (Cdk5) for Cdk5 and β-actin (loading control) proofs Cdk5 overexpression. (E) Cdk5 inhibition impairs tumor establishment in vivo. The tables indicate the time of tumour establishment and the number of established tumours of mice injected with non-targeting (nt) shRNA and Cdk5 shRNA tumour cells (1 × 10⁵ cells).

Figure 5

Cdk5 knockdown does neither affect EMT related signalling nor common cell survival pathways or DNA damage.(A) Bar graphs show mRNA levels of the EMT markers E-cadherin, N-cadherin and vimentin in non-targeting (nt) and Cdk5 shRNA-transduced cells (n=3). (B) Immunoblots show protein levels of the EMT markers vimentin, snail, β-catenin, N-cadherin and claudin-1 from non-targeting (nt) and Cdk5 shRNA cells. β-actin and β-tubulin indicate equal loading. (n=3). (C) Immunoblots for Notch1 and Notch4 intracellular domain (N1-ICD, N4-ICD) and the Notch downstream target c-Myc in non-targeting (nt) and Cdk5 knockdown cells are shown. β-actin and β-tubulin indicate equal loading. (n=3). (D) Histogram plot from FACS analysis from CD44 surface expression is shown. Bars represent quantification of CD44-positive cells in non-targeting (nt) and Cdk5 shRNA cells (mean±s.e.m., NS=not significant, n=3). (E) MMP-2 and MMP-9 immunoblots are shown. β-actin and β-tubulin indicate equal loading. (n=3). (F) Immunoblots show different proteins related to cell survival in non-targeting (nt) and Cdk5 shRNA cells. HIF1α, total and phosphorylated Stat3 (S727), ERK1/2, AKT (S473) and Retinoblastoma protein (807/811) are shown. β-actin indicates equal loading (n=3). (G) The immunoblots show phospho-histone H2aX (γH2aX) in non-targeting (nt) and Cdk5 shRNA cells. β-actin indicates equal loading (n=3).

Figure 6

Cdk5 knockdown induces apoptosis in tumorspheres by increasing the pro-apoptotic protein Bim.(A) Cdk5 knockdown induces apoptosis in tumorspheres. Specific apoptosis in sphere-forming T24 cells is shown (mean±s.e.m., *P<0.01, n=3). (B) Cdk5 inhibition induces apoptosis in tumorspheres. Specific apoptosis of MCF7 tumorspheres after pretreatment of cells with roscovitine for 24 h before resuspension in fresh sphere-formation medium and cultivation for further 10 days in presence of roscovitine is shown (mean±s.e.m., *P<0.05, n=3). (C) The immunoblot shows levels of the pro-apoptotic protein Bim in non-targeting (nt) and Cdk5 shRNA cells. Equal loading is indicated (n=3). (D) Immunoblots show Bim protein in mitochondrial fractions of non-targeting (nt) and Cdk5 shRNA T24 cells at detachment of 3 h and 24 h. The mitochondrial marker COX IV indicates equal loading (n=3). (E) The bar graph displays Bim mRNA levels of non-targeting shRNA (nt) and Cdk5 shRNA cells (mean±s.e.m., *P<0.05, n=3). (F) Immunoblots indicate increased Foxo1 protein levels in Cdk5 knockdown T24 cells. Whole-protein bands indicate equal loading (n=3). (G) Immunoblots indicate increased Foxo1 protein levels in mesenchymal HMLE cells treated with roscovitine (24 h). Whole-protein bands indicate equal loading (n=3). (H) Cdk5 knockdown leads to an increase of Foxo1. Immunoblots show Foxo1 protein of cytosolic and nuclear fractionation of T24 cells. cAMP response element-binding protein (CREB) serves as marker for the nuclei fraction. Whole-protein bands indicate equal loading (n=3). (I) Immunostainings show Foxo1 protein (green) and nucleus (Hoechst33342, blue) of non-targeting (nt) and Cdk5 shRNA T24 cells (n=3).