The cyclin E regulator cullin 3 prevents mouse hepatic progenitor cells from becoming tumor-initiating cells

Abstract

Cyclin E is often overexpressed in cancer tissue, leading to genetic instability and aneuploidy. Cullin 3 (Cul3) is a component of the BTB-Cul3-Rbx1 (BCR) ubiquitin ligase that is involved in the turnover of cyclin E. Here we show that liver-specific ablation of Cul3 in mice results in the persistence and massive expansion of hepatic progenitor cells. Upon induction of differentiation, Cul3-deficient progenitor cells underwent substantial DNA damage in vivo and in vitro, thereby triggering the activation of a cellular senescence response that selectively blocked the expansion of the differentiated offspring. Positive selection of undifferentiated progenitor cells required the expression of the tumor suppressor protein p53. Simultaneous loss of Cul3 and p53 in hepatic progenitors turned these cells into highly malignant tumor-initiating cells that formed largely undifferentiated tumors in nude mice. In addition, loss of Cul3 and p53 led to the formation of primary hepatocellular carcinomas. Importantly, loss of Cul3 expression was also detected in a large series of human liver cancers and correlated directly with tumor de-differentiation. The expression of Cul3 during hepatic differentiation therefore safeguards against the formation of progenitor cells that carry a great potential for transformation into tumor-initiating cells.

Figures

Figure 1. Loss of Cul3 leads to accumulation of hepatic progenitor cells in the liver.

(A) Livers from Cul3loxP/loxP AlfpCre mice were harvested at 4 and 8 weeks after birth. The livers showed a pale discoloration and firm consistency. Upon microscopic analysis, hepatocytes could be distinguished from small mononucleated cells. Arrow indicates small mononucleated cells in strands (compare inset) that surround hepatocytic fields. (B) LacZ stainings in Cul3loxP/loxP AlfpCre Rosa26-LacZ mice were used to analyze the functionality of the Cre recombinase. Scale bars: 50 μm; 20 μm (insets). (C) Immunofluorescence and immunohistochemical detection of progenitor-associated surface markers CK14, CD34 (first and third columns), and CD133. In wild-type livers, CK14-positive cells are primarily localized around bile ducts (compare arrows). CD34- and CD133-positive cells can only be found in Cul3-knockout livers (compare arrows that indicate positive cells in the strands). Cells also expressed the hepatocyte marker albumin. In the albumin staining, arrows indicate small mononucleated cells in the strands positive for albumin. Images in the second and fourth columns display the Dapi staining of the respective areas. Scale bars: 10 μm (top panels); 50 μm (bottom panels). (D) Quantification of hepatocytes in livers from Cul3loxP/loxP AlfpCre mice. H&E sections and β-catenin stainings (compare Supplemental Figure 1E) were used to quantitate the number of hepatocytes in the livers. For each quantification, 3–4 mice were analyzed. (E) Quantification of progenitor cells positive for CD34, CD133, and CK14 at 4 and 8 weeks after birth. For each quantification, 3–4 mice were analyzed. **P < 0.005.

Figure 2. Differentiation of Cul3-knockout hepatic progenitor cells leads to the induction of DNA damage and cellular senescence.

(A) Western blot analysis of known Cul3 substrates cyclin E, Aurora B, and p21 at 4 and 8 weeks after birth. Arrow indicates the cyclin E band. Actin was used as a loading control and was performed as noted in the methods section. Noncontiguous lanes from the same blot were spliced. Numbers indicate the different mice analyzed. (B) Immunofluorescence staining of γ-H2AX in Cul3-knockout and wild-type mice positive and negative for Cre recombinase. Arrows in Cul3loxP/loxP AlfpCre (Cul3 KO) livers indicate cells that accumulated DNA damage. In wild-type Cre+ and Cre– livers, no γ-H2AX (left panels) staining can be detected (compare Supplemental Figure 2C). DAPI staining is shown in the right panels. (C) Quantification of DNA damage in CD34-, CD133-, and CK14-positive cells (at least 100 cells from 3–4 mice were analyzed). The graph shows quantification of double-positive cells. (D) Representative pictures of round- and oval-shaped CK14-positive cells (left panels). DAPI is shown in the right panels. CK14-positive cells cannot be found in wild-type livers. (E) Western blot analysis for senescence-associated markers p15 and p16 in whole liver lysates in Cul3loxP/loxP AlfpCre livers at 4 and 8 weeks after birth in comparison with wild-type liver lysates, in which no p15 or p16 could be detected. Numbers indicate the different mice analyzed. Noncontiguous lanes from the same blot were spliced. (F) Quantification of the expression of p16 in different liver cells at 4 and 8 weeks (200 CD34- or CK14-positive cells in 3 independent experiments). Hepatocytes were identified by typical morphology (Supplemental Figure 1E). Scale bars: 10 μm. *P < 0.05, **P < 0.005.

Figure 3. In vitro differentiation of progenitor cells recapitulates the in vivo phenotype.

(A) RT-PCR analysis of progenitor cell– and hepatocyte-specific markers in asynchronous (as) cells and at different time points during differentiation (Diff.). Noncontiguous lanes from the same gel were spliced. (B) Quantitation of γ-H2AX staining by immunofluorescence at the indicated time points after the induction of differentiation. At least 3 independent experiments were analyzed. Asynchronously growing cells were set as 100%. (C) Quantitation of cells in S phase by BrdU uptake (experiments were performed in triplicate). (D) Hepatic progenitor cells induce senescence upon differentiation. Western blot analysis during differentiation shows a strong accumulation of the senescence-associated marker p15. Progenitor cells from Cul3loxP/loxP AlfpCre mice stain positive for β-galactosidase at pH 5.8 at 96 hours after the induction of differentiation. All experiments were repeated 3–4 times. Actin was used as a loading control and analysis performed as noted in Methods. Noncontiguous lanes from the same blot were spliced. (E) Analysis of proliferation of CD34-positive cells during differentiation by quantification of CD34/BrdU. Proliferation of CD34-positive cells in asynchronously growing culture is also shown (n = 2). (F) Immunofluorescence γ-H2AX staining of CD34-positive cells at 96 hours of differentiation. Squares indicate CD34-positive cells, showing that these do not accumulate DNA damage. Scale bars: 50 μm. *P < 0.05, **P < 0.005.

Figure 4. Accumulation of cyclin E in differentiating hepatic progenitor cells is responsible for the accumulation of DNA damage.

(A) Cyclin E half-lives during differentiation and in asynchronously proliferating cells. At 24 hours of differentiation, cyclin E accumulates (experiments were performed in triplicate). (B) Overexpression of myc-tagged cyclin E in asynchronously growing cells from Cul3loxP/loxP AlfpCre mice and quantitation of DNA damage (γ-H2AX). Immunofluorescence for myc tagged (MT)-cyclin E, γ-H2AX, and DAPI. Arrows indicate cyclin E–overexpressing cell positive for γ-H2AX (pink arrow) and an untransfected cell (white arrow) (n = 3). (C) Treatment of cells from Cul3loxP/loxP AlfpCre mice with H2O2 and analysis of DNA damage by γ-H2AX and pChk1 Western blots. (D) Measurement of cyclin E half-lives upon siRNA-mediated knockdown of Fbw7 and Cul1 in asynchronously proliferating cells. Loss of Fbw7 and Cul1 increases cyclin E protein stability (n = 3/experiment). CHX, cycloheximide. (E) Confirmation of siRNA-mediated knockdown by RT-PCR. GAPDH was used as control. Noncontiguous lanes from the same gel were spliced. (F) Increased DNA damage as measured by γ-H2AX staining upon knockdown of Fbw7 and Cul1. For identification of transfected cells, a labeled siRNA control was cotransfected. (G) siRNA-mediated knockdown of cyclin E during differentiation and determination of DNA damage (γ-H2AX). Knockdown of cyclin E was confirmed by Western blotting. The arrow indicates the cyclin E band. c., scrambled siRNA control. Actin was used as a loading control and analysis performed as noted in Methods. Noncontiguous lanes from the same blot were spliced. The graph shows quantification of γ-H2AX–positive cells transfected with the indicated siRNAs 96 hours after the induction of differentiation (n = 3). (H) RT-PCR analysis of Fbw7 and Cul3 Δ3–7 expression during differentiation of progenitor cells into hepatocytes at indicated time points. GAPDH was used as a control. *P < 0.05.

Figure 5. Maintenance of senescence by p53 is essential for preventing the generation of tumor-initiating stem cells.

(A) Quantitation of p16 expression in the indicated liver cells from Cul3/p53 double-knockout animals at 10–11 weeks after birth. Values are shown as difference relative to p16-positive cells in livers from Cul3 single-knockout animals (compare Figure 2F) (n = 3/group). (B) Quantitation of γ-H2AX expression in the indicated liver cells from Cul3/p53 double knockout animals at 10–11 weeks after birth. Values are shown as difference relative to γ-H2AX–positive cells in livers from Cul3 single-knockout animals (compare Figure 2C) (n = 3/group). (C) Quantitation of γ-H2AX staining by immunofluorescence at the indicated time points after the induction of differentiation. Asynchronously growing cells were set as 100%. (D) Quantitation of S-phase cells by BrdU uptake at the indicated time points after the induction of differentiation. (E) Comparative genomic hybridization analysis of Cul3 and Cul3/p53 double-knockout cells after induction of differentiation for 96 hours compared with undifferentiated cells of the identical genotype. *P < 0.05, **P < 0.005.

Figure 6. Loss of p53 in Cul3-knockout progenitor cells leads to generation of tumor-initiating cells in vivo.

(A) Transplantation of Cul3-knockout and Cul3/p53-knockout progenitor cells into nude mice. Shown is the average of 6 tumors from 3 mice after injection of 1 × 106 cells. P values for the increase in tumor size were calculated using the t test and are as follows: d 31/d 41, P = 0.0034; d 31/d 39, P = 0.744; d 39/d 48, P = 0.0022. (B) H&E staining of tumors derived from nude mice. CK14 and CD34 (bottom left panels) immunofluorescence staining to identify cells with stem cell characteristics. The bottom right panels show DAPI staining. Scale bars: 50 μm. (C) Primary hepatocellular tumors arise in Cul3/p53 double-knockout livers (area indicated by arrow), while no tumors can be found in Cul3-knockout livers. For comparison, WT livers are shown. Scale bars: 200 μm. (D) Analysis of Cul3 and cyclin E protein levels in normal human liver tissue (nt) and hepatocellular carcinoma (HCC-G3) by immunohistochemistry. Original magnification, ×40. (E) HCC tissues show a strong downregulation of Cul3 and induction of cyclin E expression as measured by immunohistochemistry in normal human liver, dysplastic nodes, and hepatocellular cancers. De-differentiation of tumors correlated with the decrease in Cul3 protein (r = –0.375; P = 0.0001). Expression ranges from 0 (no expression) to 3 (strong expression). Increase in cyclin E protein levels and decrease in Cul3 protein levels correlated significantly upon de-differentiation of tumors (r = –0.476; P = 0.046).