HDAC inhibitors induce tumor-cell-selective pro-apoptotic transcriptional responses

Abstract

The identification of recurrent somatic mutations in genes encoding epigenetic enzymes has provided a strong rationale for the development of compounds that target the epigenome for the treatment of cancer. This notion is supported by biochemical studies demonstrating aberrant recruitment of epigenetic enzymes such as histone deacetylases (HDACs) and histone methyltransferases to promoter regions through association with oncogenic fusion proteins such as PML-RARα and AML1-ETO. HDAC inhibitors (HDACi) are potent inducers of tumor cell apoptosis; however, it remains unclear why tumor cells are more sensitive to HDACi-induced cell death than normal cells. Herein, we assessed the biological and molecular responses of isogenic normal and transformed cells to the FDA-approved HDACi vorinostat and romidepsin. Both HDACi selectively killed cells of diverse tissue origin that had been transformed through the serial introduction of different oncogenes. Time-course microarray expression profiling revealed that normal and transformed cells transcriptionally responded to vorinostat treatment. Over 4200 genes responded differently to vorinostat in normal and transformed cells and gene ontology and pathway analyses identified a tumor-cell-selective pro-apoptotic gene-expression signature that consisted of BCL2 family genes. In particular, HDACi induced tumor-cell-selective upregulation of the pro-apoptotic gene BMF and downregulation of the pro-survival gene BCL2A1 encoding BFL-1. Maintenance of BFL-1 levels in transformed cells through forced expression conferred vorinostat resistance, indicating that specific and selective engagement of the intrinsic apoptotic pathway underlies the tumor-cell-selective apoptotic activities of these agents. The ability of HDACi to affect the growth and survival of tumor cells whilst leaving normal cells relatively unharmed is fundamental to their successful clinical application. This study provides new insight into the transcriptional effects of HDACi in human donor-matched normal and transformed cells, and implicates specific molecules and pathways in the tumor-selective cytotoxic activity of these compounds.

Figure 1

Matched normal and tumor cells are differentially sensitive to vorinostat-mediated apoptosis. (a, b) BJ and BJ LTSTERas cells were treated with 0, 2.5, 5, 10, 25 or 50 μM doses of vorinostat for 24, 48 or 72 h. Following treatment, all cells were harvested and (a) stained for annexin-V-APC binding and propidium iodide uptake, and analyzed by flow cytometry; (b) fixed with 50% (v/v) ethanol in PBS, stained with propidium iodide and analyzed by flow cytometry. The percentage of annexin-V-APC, propidium iodide double positive cells or cells with sub-diploid DNA content is shown in panels (a, b), respectively. Data are presented as mean±standard error of at least three independent experiments. P-values indicate statistical significance of cell type and dose response interactions as measured by two-way ANOVA, where P<0.05 is considered to be significant. (c, d) BJ (top panels) and BJ LTSTERas (bottom panels) cells were treated with 25 μM vorinostat or DMSO equivalent for 0–32 h. Protein extracts were analyzed by western blotting for: (c) acetylation of histone H3 with a pan-acetylated Histone H3 antibody, expression of α-Tubulin was used as a loading control; (d) acetylation of α-Tubulin with an anti-acetylated tubulin antibody, expression of β-Actin served as loading control. Western blots shown are representative of two independent experiments. (e) BJ and BJ LTSTERas were treated with 0, 2.5, 5, 10, 25 and 50 nM concentrations of romidepsin for 48 h. Following treatment, all cells were harvested and stained for annexin-V-APC binding and propidium iodide uptake (top panel) and analyzed by flow cytometry or fixed with 50% (v/v) ethanol in PBS, stained with propidium iodide (bottom panel) and analyzed by flow cytometry. The percentage of annexin-V-APC, propidium iodide double positive cells or cells with sub-diploid DNA content is shown. Data are presented as mean±standard error of at least five independent experiments. The difference in apoptotic sensitivity of BJ and BJ LTSTERas (interaction between cell type and romidepsin dose) was statistically significant (P<0.05), as determined by two-way ANOVA. (f) BJ and BJ LTSTERas fibroblasts were treated with 50 nM romidepsin for 0–24 h or with 25 μM vorinostat for 6 h (V). Immunoblots were probed with antibodies against acetylated histone H3 (AcH3), acetylated α-Tubulin (Ac-α-Tubulin) and total α-Tubulin to control for protein loading. Images are representative of two independent experiments

Figure 2

Vorinostat-mediated apoptosis requires de novo protein synthesis. (a, b) BJ and BJ LTSTERas cells were pre-treated with 0, 5, 50, 250 and 500 ng/ml CHX to inhibit new protein synthesis and incubated with 25 μM vorinostat for 48 h. Cells were harvested and analyzed for cell death by flow cytometry. (a) Cells were stained with annexin-V-APC and propidium iodide. The percentages of cells staining positively for both markers are shown. (b) Cells were fixed with 50% (v/v) ethanol in PBS, stained with propidium iodide and analyzed for DNA content. The percentages of cells with sub-2N DNA content are shown. All data are presented as mean±S.E.M. of three independent experiments. *P<0.05, **P<0.01 as determined by a two-tailed Student's t-test, comparing 25 μM vorinostat treatments in the presence and absence of CHX. (c, d) BJ and BJ LTSTERas cells were treated with 25 μM vorinostat (red) or DMSO (black) for 0, 2, 4, 8, 16 and 24 h. Expression of CDKN1A in (c) BJ and (d) BJ LTSTERas cells was analyzed by qRT-PCR. Messenger RNA levels were calculated relative to that of transcripts from the non-HDACi-regulated control gene RPL32. Fold expression changes at each time point are plotted relative to time 0 h (untreated cells)

Figure 3

Regulation of gene expression in BJ and BJ LTSTERas by vorinostat. (a) The number of probe sets that were significantly altered in BJ and BJ LTSTERas cells in response to treatment with vorinostat. Probe sets that were induced and repressed by at least 1.5-fold at any of the 4, 12 and 24 h time points relative to time 0 h are shown. (b) Genes that responded differently to vorinostat treatment in matched normal and transformed cells (contrast genes) were analyzed using the IPA tool. The associations of various molecular and cellular functions with contrast genes are plotted in decreasing order of statistical significance, according to −log2 (P-value). P-values were adjusted for multiple testing using the method of Benjamini and Hochberg. The orange line indicates the threshold for statistical significance (P<0.05)

Figure 4

Ingenuity pathways analysis. Interaction map of the death receptor pathway was created using the ingenuity pathways analysis tool (a) BJ and (b) BJ LTSTERas fibroblasts. (c, d) An interaction map of the BCL-2 family of apoptotic regulators was created using the ingenuity pathways analysis tool, and overlaid with transcriptional responses to vorinostat in (c) BJ and (d) BJ LTSTERas fibroblasts. For (a–d), the direction and magnitude of vorinostat-mediated expression changes following 24 h of treatment, relative to time 0 h are indicated by pseudo-coloring. Induced (red), repressed (green) and non-responding probe sets (white) are shown. Log2 expression values of the strongest responding probe sets for each vorinostat-responding pathway component are shown. Interactions between BH3-only and pro-survival proteins are based upon Biacore analyses described in reference . Interactions between multi-domain proteins BAX/BAK and other BCL-2 family members were based on ingenuity pathway information

Figure 5

Validation of vorinostat-induced transcriptional responses by qRT-PCR. Transcriptional responses of BJ and BJ LTSTERas fibroblasts following DMSO (black) and 25 μM vorinostat treatment (red) were validated by qRT-PCR using the same template RNA as that used in the microarray study. Log2 transformed expression intensities of HG U133 plus 2.0 probe sets corresponding to (a) BAK1, BAD, BIK and BCL2A1 and (b) BCL2L11 and BMF are shown. Gene names and probe set IDs are indicated in the title of each panel. Normalized probe set intensity values (Y-axis) are presented, which allows direct comparison of basal expression (time 0 h), and log2 transcriptional responses over time between cell types (BJ and BJ LTSTERas). Data represent the summarized probe set intensities from three independent microarray time course experiments. qRT-PCR verification of selected genes is shown. Transcript expression was calculated relative to the control gene RPL32 (that does not respond to vorinostat or DMSO treatment), and fold-change mRNA expression over time was calculated relative to the 0 h time point. Data are presented as the mean±S.E.M. of three independent experiments. For both Affymetrix and qRT-PCR results, data sets for BJ cells are in the left panels while data sets for BJ LTSTERas are in the right panels. (c) Whole cell protein extracts from BJ and BJ LTSTERas fibroblasts treated with DMSO or 25 μM vorinostat for 0–48 h were separated by SDS-PAGE and analyzed for the expression of BMF, BIK and BIM by western blotting. The expression of α-Tubulin was analyzed to confirm equal protein loading in each lane. The asterisk (*) denotes a non-specific band in the BIK western blot, which acts as a positive control for activity of the BIK antibody. The arrow denotes the location of BIK. Images are representative of two independent experiments

Figure 5

Validation of vorinostat-induced transcriptional responses by qRT-PCR. Transcriptional responses of BJ and BJ LTSTERas fibroblasts following DMSO (black) and 25 μM vorinostat treatment (red) were validated by qRT-PCR using the same template RNA as that used in the microarray study. Log2 transformed expression intensities of HG U133 plus 2.0 probe sets corresponding to (a) BAK1, BAD, BIK and BCL2A1 and (b) BCL2L11 and BMF are shown. Gene names and probe set IDs are indicated in the title of each panel. Normalized probe set intensity values (Y-axis) are presented, which allows direct comparison of basal expression (time 0 h), and log2 transcriptional responses over time between cell types (BJ and BJ LTSTERas). Data represent the summarized probe set intensities from three independent microarray time course experiments. qRT-PCR verification of selected genes is shown. Transcript expression was calculated relative to the control gene RPL32 (that does not respond to vorinostat or DMSO treatment), and fold-change mRNA expression over time was calculated relative to the 0 h time point. Data are presented as the mean±S.E.M. of three independent experiments. For both Affymetrix and qRT-PCR results, data sets for BJ cells are in the left panels while data sets for BJ LTSTERas are in the right panels. (c) Whole cell protein extracts from BJ and BJ LTSTERas fibroblasts treated with DMSO or 25 μM vorinostat for 0–48 h were separated by SDS-PAGE and analyzed for the expression of BMF, BIK and BIM by western blotting. The expression of α-Tubulin was analyzed to confirm equal protein loading in each lane. The asterisk (*) denotes a non-specific band in the BIK western blot, which acts as a positive control for activity of the BIK antibody. The arrow denotes the location of BIK. Images are representative of two independent experiments

Figure 6

Regulation of BCL2 family gene expression by the HDACi romidepsin and the DNA damaging agent etoposide. BJ (left panels) and BJ LTSTERas (right panels) fibroblasts were treated for 0, 4, 12 and 24 h with 50 nM romidepsin (red), 50 μM etoposide (blue) and analyzed for the expression of BAK1, BCL2L11, BIK, BMF and BCL2A1 by qRT-PCR. Expression of each gene was calculated relative to the non-romidepsin- and non-etoposide-responding control gene RPL32 and fold-change expression over time was calculated relative to time 0 h. Gene expression changes resulting from vehicle treatment (black) are also shown. Data are presented as the mean expression fold change from two independent time course experiments

Figure 7

Forced expression of BFL-1 protects BJ LTSTERas fibroblasts from vorinostat-mediated apoptosis. (a) Proteins were extracted from exponentially growing BJ LTSTERas fibroblasts, and MSCV and MSCV FLAG–BFL-1 transduced variants. Proteins were separated on 10% polyacrylamide gels by SDS-PAGE and analyzed for the expression of FLAG–BFL-1 by western blotting with an anti-FLAG antibody. The expression of α-Tubulin levels served as a control for protein loading. (b) BJ, BJ LTSTERas, BJ LTSTERas MSCV and BJ LTSTERas FLAG–BFL-1 cells were incubated with 0, 2.5, 5, 10, 25 and 50 μM vorinostat for 24, 48 and 72 h. Cells were harvested and stained with annexin-V-APC and propidium iodide as an apoptosis readout or stained for loss of mitochondrial outer membrane potential (MOMP, ΔΨm) using the mitochondrial dye tetramethylrhodamine ethyl ester (TMRE). Cells were then analyzed by flow cytometry. The percentage of annexin-V-APC, propidium iodide double positive cells are shown in the left panels while cells that had lost the ability to bind TMRE (ΔΨm) are in the right panels. Data are presented as mean±S.E.M. of three independent experiments

Figure 8

Model for the induction of tumor-selective apoptosis by vorinostat. The tumor-selective induction of apoptosis following vorinostat treatment may be due to acute alterations in the expression of pro-survival and pro-apoptotic BCL2 family genes. The vorinostat-mediated transcription changes observed in BJ LTSTERas cells are hypothesized to tip the balance of pro-survival and pro-apoptotic gene expression in a direction that favors cell death. Although pro-survival and pro-apoptotic expression changes were observed in normal cells, the net change of expression in terms of ability to ‘tip the balance' is relatively neutral. Genes are pseudo-colored according to the direction of gene expression change at the 24-h time point, relative to time 0 h: induced (red) and repressed (green), and are placed on the balance according to whether the expression change is expected to confer a pro-death or pro-survival response