Epigenetic reprogramming modulates malignant properties of human liver cancer

Abstract

Reversal of DNA hypermethylation and associated gene silencing is an emerging cancer therapy approach. Here we addressed the impact of epigenetic alterations and cellular context on functional and transcriptional reprogramming of hepatocellular carcinoma (HCC) cells. Our strategy employed a 3-day treatment of established and primary human HCC-derived cell lines grown as a monolayer at various cell densities with the DNMT1 inhibitor zebularine (ZEB) followed by a 3D culture to identify cells endowed with self-renewal potential. Differences in self-renewal, gene expression, tumorigenicity, and metastatic potential of spheres at generations G1-G5 were examined. Transient ZEB exposure produced differential cell density-dependent responses. In cells grown at low density, ZEB caused a remarkable increase in self-renewal and tumorigenicity associated with long-lasting gene expression changes characterized by a stable overexpression of cancer stem cell-related and key epithelial-mesenchymal transition genes. These effects persisted after restoration of DNMT1 expression. In contrast, at high cell density, ZEB caused a gradual decrease in self-renewal and tumorigenicty, and up-regulation of apoptosis- and differentiation-related genes. A permanent reduction of DNMT1 protein using short hairpin RNA (shRNA)-mediated DNMT1 silencing rendered HCC cells insensitive both to cell density and ZEB effects. Similarly, WRL68 and HepG2 hepatoblastoma cells expressing low DNMT1 basal levels also possessed a high self-renewal, irrespective of cell density or ZEB exposure. Spheres formed by low-density cells treated with ZEB or shDNMT1 displayed a high molecular similarity which was sustained through consecutive generations, confirming the essential role of DNMT1 depletion in the enhancement of cancer stem cell properties.

Conclusion: These results identify DNA methylation as a key epigenetic regulatory mechanism determining the pool of cancer stem cells in liver cancer and possibly other solid tumors.

Fig. 1

Cell density-dependent effects of a transient ZEB exposure on CSC properties. (A) Venn diagram (top) and heatmaps (bottom) of differentially expressed genes in Huh7 cells plated at low (LD-ZEB) and high (HD-ZEB) density and treated with ZEB for 3 days after normalization to the corresponding untreated cells (n=3 for each condition, 10,000 repetitions in Bootstrap ANOVA with contrast tests and a threshold cut-off of 2-fold change, P<0.05). (B) The IPA analysis of LD-ZEB and HD-ZEB gene signatures (score >40). (C) Representative images of Ki67 and EpCAM immunofluorescence staining in Huh7 primary spheres. Nuclei counterstained with DAPI. Scale bar, 20 μm. (D) Quantification of Ki67-positive cells (n=3, **P<0.01, ***P<0.001 versus LD-ZEB by ANOVA test with contrast test after inverse normal transformation). (E) Immunoblotting of indicated proteins in Huh7 spheres. β-actin used as a loading control. (F) Quantitative RT-PCR (qRT-PCR) analysis of c-KIT expression. GAPDH used as internal control (n=3, ***P<0.001 versus LD-ZEB by Bootstrap t-test with 10,000 repetitions). (G) qRT-PCR analysis of indicated genes in spheres formed by Huh7 cells grown under different culture conditions. The data are means ± SEM (n=3, *P<0.05 versus LD-ZEB cells by Bootstrap t-test with 10,000 repetitions). (H) Effect of ZEB on sphere-forming potential. The data are means ± SEM (n=6, Huh7; n=3, other tumor cell lines, Poisson GLM with multiple comparison test. P-values refer to LD-ZEB versus HD-ZEB).

Fig. 2

Transient DNMT1-inhibition causes sustained functional and molecular alterations in liver cancer cells. (A) Analysis of self-renewal. Primary spheres were passaged every 10 days (generations G1 to G5). The data are means ± SEM (n=3, *P<0.05, **P<0.01, ***P<0.001 by Poisson GLM with multiple comparison test, P-values refer to LD-ZEB versus HD-NT). (B) Gene expression of pluripotency, cancer stem/stem and EMT-related markers by q-RT-PCR (n=3). Data are means ± SEM (n=3, *P<0.05, **P<0.01, ***P<0.001 versus LD-ZEB-G1 by Bootstrap t-test with 10,000 repetitions). (C, D) Immunoblotting of indicated proteins in LD-ZEB spheres at different sphere generations. β-actin and H3-total (H3tot) were used as loading controls in C and D, respectively.

Fig. 3

Transcriptomic characteristics of tumor-spheres formed by Huh7 (A-D) and PLC/PRF/5 (E-H) cells. (A,E) Heat maps of differentially expressed genes in G1 spheres formed by Huh7 (A) and PLC/PRF/5 (E) cells grown in different conditions. Bootstrap ANOVA with contrast t test (n=3, P<0.05, 10,000 repetitions. (B,F) Supervised hierarchical cluster analysis of LD-ZEB-specific gene signatures identified by a comparison with HD-NT and HD-ZEB in Huh7 (B) and PLC/PRF/5 (F). (C,G) Bioequivalence test of similarities of LD-ZEB specific gene signatures between four consecutive sphere generations in Huh7 (C) and PLC/PRF/5 (G). Data were evaluated at fold change ≤1.5 and P<0.05. (D,H) Kaplan-Meier plot of overall survival of HCC patients using Huh7-LD-ZEB 249-gene signature (D) and PLC/PRF/5-LD-ZEB 108-gene signature (H).

Fig. 4

DNMT1 sh-RNA knockdown in Huh7 cells increases self-renewal potential. (A) Immunoblotting of DNMT1 and DNMT3a. HeLA nuclear extract served as a positive control for DNMT1 expression. β-actin used as a loading control. (B) Sphere frequency at G1-G5 generations. The data are means ± SEM (n=3, ***P<0.001 by Poisson GLM with multiple comparison test). (C) Limiting dilution analysis. The frequency of tumor initiating cells (TIF) and confidence interval (CI95%) were calculated based on the number of resulting tumors/injection site at 10 weeks. (D) Self-renewal of WRL68 cells expressing low basal levels of DNMT1 protein. The data are means ± SEM (n=3, Poisson GLM with multiple comparison test). (E) Immunoblotting of DNMT1. β-actin used as a loading control.

Fig. 5

Effects of a transient DNMT1 Inhibition on tumorigenicity. (A) The frequency of sphere-cell-generated tumors at 10 weeks after subcutaneous injection of 100 cells into NOD/SCID mice. Shown are the numbers of tumors per injection site. Fisher's exact test was applied to evaluate statistical significance. (B) Representative H&E and PCNA immunofluorescence staining on paraffin-embedded tumor sections. (C) Quantification of necrotic areas on tumor sections stained with H&E (n=4, *P<0.05 by Bootstrap t-test with 10,000 repetitions after inverse normal transformation). (D) Representative immunofluorescence images of CD34 staining. (E) Immunoblotting of CD31. β-actin used as a loading control. (F) Representative immunofluorescence staining with anti-E-cadherin and double immunofluorescence staining with anti-vimentin and anti-Ki67. (G) Western blot analysis of E-Cadherin and β-Catenin. β-actin used as a loading control. Scale bar, 50 μm. Nuclei counterstained with DAPI (B, upper images, D and F).

Fig. 6

Increased tumorigenicity with sphere passaging. (A) Kinetics of tumor growth in NOD/SCID mice after subcutaneous injection of 100 tumor cells isolated from the indicated spheres at generations G1, G3 and G4. P-values evaluated by log rank test refer to LD-ZEB versus HD-NT. (B) Representative immunofluorescence images of CD34 and SNAIL staining. Nuclei counterstained with DAPI. Scale bar, 50 μm. (C) Immunoblotting of indicated proteins. β-actin used as a loading control.