DNA hypermethylation in prostate cancer is a consequence of aberrant epithelial differentiation and hyperproliferation

Abstract

Prostate cancer (CaP) is mostly composed of luminal-like differentiated cells, but contains a small subpopulation of basal cells (including stem-like cells), which can proliferate and differentiate into luminal-like cells. In cancers, CpG island hypermethylation has been associated with gene downregulation, but the causal relationship between the two phenomena is still debated. Here we clarify the origin and function of CpG island hypermethylation in CaP, in the context of a cancer cell hierarchy and epithelial differentiation, by analysis of separated basal and luminal cells from cancers. For a set of genes (including GSTP1) that are hypermethylated in CaP, gene downregulation is the result of cell differentiation and is not cancer specific. Hypermethylation is however seen in more differentiated cancer cells and is promoted by hyperproliferation. These genes are maintained as actively expressed and methylation-free in undifferentiated CaP cells, and their hypermethylation is not essential for either tumour development or expansion. We present evidence for the causes and the dynamics of CpG island hypermethylation in CaP, showing that, for a specific set of genes, promoter methylation is downstream of gene downregulation and is not a driver of gene repression, while gene repression is a result of tissue-specific differentiation.

Figure 1

GSTP1 is hypomethylated and highly expressed in undifferentiated basal prostate cancer cells. (a) Pyrosequencing methylation analysis of the GSTP1 promoter performed in prostate cell lines (bars=single CpG sites; n=3 technical replicates; mean±S.D.; line=average of 14 CpG sites, Positive Control=RC-165N/hTERT DNA methylated with SssI methyltransferase). (b) Pyrosequencing methylation analysis of GSTP1 performed in basal and luminal cells derived from BPH and CaP (each dot represents the average of 14 CpG sites analysed in a single sample; boxplots show minimum, 25%, median, 75% and maximum, hypermethylation threshold (dot-dashed line)=average methylation of BPH basal+2 S.D., P-values from Mann–Whitney test). (c) qRT-PCR analysis of GSTP1 expression relative to GAPDH in basal and luminal cells derived from BPH and CaP (boxplots show minimum, 25%, median, 75% and maximum; each dot represents a single sample, P-values from Mann–Whitney test). (d) qRT-PCR analysis of GSTP1 expression relative to GAPDH in primary prostate cancer xenografts generated in RAG2^−/− γC^−/− mice. (e) Pyrosequencing methylation analysis of GSTP1 performed in primary prostate cancer xenografts (left panel), MACS selected cells from disaggregated xenografts tumours (central panel), and matched xenografts and original tumour tissue (right panel) (bars=single CpG sites; n=3 technical replicates; mean±S.D.; line=average of 14 CpG sites)

Figure 2

GSTP1 expression and promoter methylation correlates with differentiation of hyperproliferating prostate epithelial cells. (a) RT-PCR analysis of GSTP1 and ACTB (β-actin) expression in 18 randomly selected clones of BPH-1 cells. Five independent preparations of the parental cell line were used as a control for the stability of GSTP1 expression and reliability of the technique (WP-1–WP-5). (b) Pyrosequencing methylation analysis of the GSTP1 promoter performed on the same clones normalized versus the parental cell line (WP) (bars=single CpG sites; n=3 technical replicates; mean±SD; line=average of 14 CpG sites). (c) GSTP1 expression plotted against promoter methylation in BPH-1 clones. GSTP1 expression was normalized to β-actin and calibrated against the average of the five WP samples (open circle) (each dot represent a single clone, dashed line=linear regression). (d) Western blot analysis for GSTP1 on three hypermethylating clones (B8, C1, C3), two hypomethylating clones (C6, C11) and two clones (C4, C9) with average methylation levels comparable to the parental cell line (WP). (e) Immunofluorescence analysis of Cytokeratin5 (KRT5) and GSTP1 levels in BPH-1 colonies. (f) Pyrosequencing methylation analysis of the GSTP1 performed on the DNA extracted from the colonies shown in e after the immunofluorescence pattern was recorded

Figure 3

Identification of DAH genes: a set of genes behaving similarly to GSTP1. (a) Schematic representation of the workflow undertaken to select genes hypermethylated in prostate cancer luminal-like cells, but actively expressed in prostate cancer basal cells (described in details in materials and methods section). (b, c) Expression analysis of DAH genes in CaP versus Normal (b), and Luminal versus Basal normal prostate cells (c) (differentially expressed= all probes with P<0.05 in a t-test comparing the two sample groups; P-values from Mann-Whitney test; Red-blue=difference of the mean log2 fold change for each set of genes). (d) Chromatin status of DAH promoters in H1-hESCs (data retrieved from ENCODE database). (e) Selection of representative DAH genes for further analysis

Figure 4

CCND2, CSTA, DKK3, LDHB, S100A14, SFN and THBS1 are hypomethylated and highly expressed in basal fraction of primary CaP. (a) qRT-PCR analysis of CCND2 (i), CSTA (ii), DKK3 (iii), LDHB (iv), S100A14 (v), SFN (vi) and THBS1 (vii) expression relative to GAPDH in basal and luminal cells derived from BPH and CaP (boxplots show minimum, 25%, median, 75% and maximum; each dot represents a single sample, P-values from Mann–Whitney test). (b) Pyrosequencing methylation analysis of CCND2 (i), CSTA (ii), DKK3 (iii), LDHB (iv), S100A14 (v), SFN (vi) and THBS1 (vii) in basal and luminal cells derived from BPH and CaP (each dot represent the average of all the CpG sites analysed in a single sample; boxplots show minimum, 25%, median, 75% and maximum, hypermethylation threshold (dot-dashed line)=average methylation of BPH basal+2 standard deviations, P-values from Mann–Whitney test, the two BPH basal samples that clearly hypermethylated SFN were excluded from the threshold calculation)

Figure 5

Differentiation of BPH-1 cells induces downregulation and transcriptional inactivation of differentiation-associated hypermethylated genes, but not promoter hypermethylation. (a, top) Representative image of BPH1-PPO cells grown in 2D (standard culture conditions) and 3D (differentiating conditions) showing induction of mOrange expression in differentiating conditions, indicative of successful differentiation. (a, bottom) Schematic representation of the viral vector used to generate BPH-1 luminal reporter cells. (b) qRT-PCR analysis of DAH candidate genes expression relative to GAPDH in BPH-1 cells grown in 2D and 3D. (c–f): ChIP-qPCR analysis carried out in BPH-1 cells grown in 2D and 3D with (c) anti-RNAPolII, (d) anti-H3K4Me3, (e) anti-acetylated H3 (f) anti-H3K27Me3 (data presented as % of immunoprecipitated DNA). (g) Pyrosequencing methylation analysis of DAH candidate genes performed on BPH-1 PPO cells grown in 2D and 3D (bars=single CpG sites; n=3 technical replicates; mean±S.D.; line=average of 14 CpG sites)

Figure 6

Expression and methylation of DAH genes is not related to induction of prostate tumorigenesis. (a) Diagram explaining the in vivo experimental design. (b) qRT-PCR for PSA and PAP in BPH-1 cells and renal grafts generated by recombining BPH-1 cells with patient matched normal prostate fibroblasts (BPH-1+NPF, n=3) or cancer-associated fibroblasts (BPH-1+CAF, n=3). qRT-PCR (c) and pyrosequencing methylation (d) analysis of DAH candidate genes in the same samples

Figure 7

Regulation of DAH genes in prostate cancer hierarchy. Schematic representation of the proposed downregulation and hypermethylation mechanism of differentiation-associated hypermethylated genes in prostate cancer