Gene Expression Profiling of Multiple Histone Deacetylase (HDAC) Inhibitors: Defining a Common Gene Set Produced by HDAC Inhibition in T24 and MDA Carcinoma Cell Lines

Introduction

Gene expression is in part regulated by differential acetylation of nucleosomal histones resulting in either transcriptional activation (hyperacetylation) or repression (hypoacetylation; Refs. 1, 2). This phenomenon is regulated tightly by the balance of histone acetyltransferase and HDAC ² activities (1, 2). One mechanism that promotes carcinogenesis is the repression of tumor suppressor genes (2). Known repressors are multiprotein complexes that contain DNA binding proteins (e.g., NCoR, SMRT, MEF, MeCP2, sin3A, and so forth) that commonly use HDACs to repress transcription and block the function of the tumor suppressor. Inhibitors of the HDACs are able to de-repress these genes resulting in antiproliferative effects in vitro and antitumor effects in vivo (3–6). The deacetylase superfamily can be divided into three distinct classes based on their structure (7). The HDACs comprise the first two classes and consist of class I (HDACs 1, 2, 3, 8, and 11) and class II (HDACs 4, 5, 6, 7, 9, and 10) enzymes (7–10). The class II enzymes are distinguished by a large NH₂-terminal domain or second catalytic site (e.g., HDAC 6) in the these enzymes (11). The class III enzymes, SIRTs (sirtuins) or Sir2-related proteins, are HDACs in yeast (12–14); however, their role in vivo in mammalian cells appears to involve deacetylation of other proteins or transcription factors, e.g., p53, rather than histones (15, 16).

The role of chromatin remodeling in carcinogenesis is based primarily on experiments with HDAC inhibitors, e.g., sodium butyrate and TSA. HDAC inhibitors induce the hyperacetylation of nucleosomal histones in cells resulting in the expression of repressed genes that produce growth arrest, terminal differentiation, and/or apoptosis in carcinoma cells (2, 3, 17). These effects coincide with the induction of p21^Cip1/Waf1 as well as markers of differentiation such as gelsolin (4, 18–20). Several HDAC inhibitors exhibit antitumor effects in preclinical animal models including methylnitrosourea-induced breast carcinogenesis (4, 21, 22). The role of inappropriate recruitment of HDACs or repressors in the development of a malignant disease is exemplified by RA insensitivity in a subset of acute promyelocytic leukemia patients (23, 24). This form of acute promyelocytic leukemia arises from recruitment of HDAC1 to a chimeric PLZF:RA receptor α receptor (23, 25). This maintains the receptor/transcription complex in a repressed state thereby preventing the differentiating effects of RA. In at least one such patient, combination therapy of RA and sodium phenylbutyrate generated disease remission (25). Western blot analysis revealed a time-dependent increase in histone acetylation in bone marrow mononuclear cells from this patient consistent with a mechanism involving inhibition of HDACs (25). These pharmacological properties of the HDAC inhibitors have generated significant interest in HDACs as targets for anticancer therapy (3, 6, 17, 26).

Because inhibitors of HDAC are known to alter gene expression, it was of interest to determine the extent of gene changes in carcinoma cells produced by HDAC inhibition and whether a common set of genes could be identified that defines the gene expression profile of small molecule HDAC inhibitor(s). These small molecule HDAC inhibitors include the hydroxamate compounds TSA (27) and SAHA (5) that are potent nanomolar inhibitors of HDAC enzymes and MS-275 a nonhydroxamate micromolar inhibitor of HDACs (4). Of these compounds, SAHA (3, 6) and MS-275 (3) are currently under clinical investigation as antitumor agents. Earlier studies with TSA suggested that <2% of the genes were regulated using differential display in human lymphoid cell lines (28). Similar studies of TSA in SNU1 gastric cancer cells demonstrated that of 2400 genes 5% of the total genes were regulated, 93 genes were up-regulated, and 27 were down-regulated significantly (29). We analyzed the gene expression patterns of three different HDAC inhibitors in three different cell lines using microarray studies and identified a core set of genes that are regulated by all three of the HDAC inhibitors in all of the cell lines evaluated. This core set of genes contains 8 up-regulated and 5 down-regulated genes. Each of these regulated genes was confirmed by an additional microarray experiment in T24 cells, Q-PCR, and some of the genes by Western analysis.

Materials and Methods

Cell Culture and Reagents.

Human breast carcinoma cells, MDA 468 and 435, or human bladder carcinoma cells, T24 (all of the cell lines were purchased from American Type Culture Collection, Bethesda, MD), were grown according to American Type Culture Collection guidelines. For experiments, cells were trypsinized and seeded at 1 × 10⁶ cells/100 mm dish and allowed to grow overnight at 37°C with 95%/5% air/CO₂ and 80% relative humidity. Cells were treated with various concentrations of the HDAC inhibitors, MS-275, SAHA (synthesized at Abbott Laboratories for investigational use), or TSA (Biomol, Plymouth Meeting, PA; final DMSO concentration in medium of 0.01%) for 24 h. The 24 h time point was used as determined for optimal p21 protein induction. Cells were rinsed 1× with PBS, harvested by scraping into PBS, and isolated by centrifugation at 700 × g for 10 min, and cell pellets were stored at −80°C until processing.

RNA Processing and DNA Array.

RNA was isolated using Qiagen RNeasy Mini kit according to the manufacturer’s protocol. The total RNA (10 μg) was used to make the cDNA preparation. Double-stranded cDNA was prepared using the SuperScript Choice System according to the manufacturer’s protocol. Biotinylated cRNA was prepared using the BioArray High Yield RNA Transcript Labeling kit (ENZO). After in vitro transcription, the unincorporated nucleotides were removed using RNeasy columns (Qiagen). Then, 15 μg of the biotinylated cRNA was fragmented at 94°C for 35 min in a fragmentation buffer containing 40 mm Tris-acetate (pH 8.1), 100 mm potassium acetate, and 30 mm magnesium acetate. Before hybridization, the fragmented cRNA in 6× saline-sodium phosphate-EDTA-T hybridization buffer was heated to 95°C for 5 min and subsequently to 40°C for 5 min before loading onto the Affymetrix human 6800 FL gene chip. Before loading the 6800 FL chip an Affymetrix test chip was hybridized to test the integrity of the cRNA preparation. The biotinylated cRNA was then hybridized to the Affymetrix Human Full Length gene chip containing sets of 20mer oligonucleotides to 6800 known genes. The gene chip was incubated for 16 h at 45°C at constant rotation (60 rpm). The washing and staining procedure was performed in the Affymetrix Fluidics Station. The arrays were then scanned at 560 nm using a confocal laser-scanning microscope with an argon ion laser as the excitation source (Hewlett Packard GeneArray Scanner G2500A). The gene array was analyzed by the Affymetrix Gene Expression Analysis Software. Data were processed using Spotfire, GeneMath, and Rosetta Resolver software by hierarchical cluster analysis.

Q-PCR Analysis of Core Genes.

Total RNA was isolated using Qiagen RNeasy kit as per the manufacturer’s instructions. From each sample, 100 ng of total RNA was used per reaction. Added to each experimental sample was 100 nm of primers (Table 1) TaqMan probe per reaction as per TaqMan EZ RT-PCR Core Reagent instructions (Applied Biosystems, Branchburg, NJ). Amplification and fluorescence measurements were made with an iCycler iQ real-time PCR detection system from Bio-Rad. The PCR program was designed as follows: hold 3 min at 50°C, hold 30 min at 60°C, hold 5 min at 94°C, and 40 cycles were then run for 20 s at 94°C followed by 1 min at 60°C. The iCycler software constructs amplification plots from the extension phase for fluorescent emission data collected during the PCR amplification. The SD was determined from the data points collected from the baseline of the amplification plot. C_T (threshold cycle number) values were calculated by determining the point at which the fluorescence exceeds a threshold limit, usually 10 times the SD of the baseline. Fold change was determined using C_T values normalized for the expression of the 28S ribosomal protein.

Electrophoresis and Western Blotting.

SDS-PAGE was performed using 4–12% Novex NuPAGE gels using the 4-morpholinepropanesulfonic acid buffer system (Invitrogen, San Diego, CA). Proteins were prepared at the appropriate concentrations in sample buffer (4× Novex NuPAGE sample buffer), heated at 70°C for 10 min, and loaded onto gels. After electrophoresis the proteins were transferred to polyvinylidene difluoride membranes (Novex) and blocked for 2 h with 10% nonfat milk in TBS. Primary antibodies were diluted in 10% nonfat milk in TBS [anti-p21 (1 μg/ml) and antihyperacetylated histone H4 (0.5 μg/ml) were both from UBI, Lake Placid, NY], and antithymidylate synthetase (1:4000 dilution) was from Sigma and incubated with membranes overnight at 4°C. The membranes were washed 3× with 0.1% Tween 20 in TBS before addition of the secondary horseradish peroxidase-conjugated antibody. Secondary (horseradish peroxidase-conjugated) antibodies were diluted in 10% nonfat milk in TBS (1:5000) and incubated for 2 h with shaking. The membranes were then washed 10 × 5 min in 0.1% Tween 20 in TBS. Proteins were visualized by enhanced chemiluminescence with the Pierce Dura SuperSignal substrate (Pierce, Rockford, IL).

Results

Induction of Gene Changes in Carcinoma Cells by HDAC Inhibitors.

The initial event after HDAC inhibition in cells is the accumulation of hyperacetylated histones. Fig. 1A demonstrates that TSA, SAHA, and MS-275 (Table 2) caused the accumulation of hyperacetylated histone H4 after 4 h of treatment. HDAC inhibitors produce profound morphological changes, a differentiated phenotype, through the altered expression or repression of sets of genes that control these events. p21, a cyclin-dependent kinase inhibitor, is upregulated in most cells exposed to HDAC inhibitors (4, 5, 18–20). Fig. 1B shows the effect of the reference HDAC inhibitors on the expression of p21 in T24 cells after 24-h exposure. An inactive analogue of SAHA (812; Table 2) was included as a negative control. Because the elevation of p21 correlates with many of the changes seen with HDAC inhibitors on cell cycle and proliferation, the 24-h time point was used to analyze changes in gene expression. Single concentrations of the HDAC inhibitors were used that had been demonstrated to produce robust hyperacetylation of histones and induced similar levels of p21 expression (Fig. 1B) without overt toxicity to the cells (measured as lactate dehydrogenase release; data not shown). For SAHA and MS-275 the concentrations chosen produced similar antiproliferative activity in the cell lines evaluated, whereas the concentration of TSA was 10–100-fold greater than its antiproliferative concentration; but as shown in Fig. 1B, this was the concentration of TSA required to produce equivalent p21 expression. Initial gene chip studies demonstrated that these three HDAC inhibitors up-regulated as many genes as were down-regulated (each gene selected at P ≤ 0.01 and >2-fold or <−2-fold; Table 3). Overall the number of genes regulated >2-fold was <10% of the genes on the Affymetrix FL 6800 gene chip. For SAHA, only 8% of genes were regulated in T24, MDA435, and MDA468 cells.

Cluster Analysis of Gene Expression Patterns.

The heat map generated by Rosetta Resolver Agglomerative clustering (P < 0.01) of genes regulated by various HDAC inhibitors is shown in Fig. 2A. The heat map demonstrates the similarities between these structurally distinct HDAC inhibitors (Table 2). The cluster tree generated using Rosetta Resolver Agglomerative clustering (Fig. 2A) demonstrates that each cell line formed an independent cluster. These results (gene change intensities observed in the heat map) also illustrate the dependency of gene changes on cell line. Fig. 2B shows the GeneMath hierarchical cluster heat map for a subset of genes regulated by HDAC inhibitors. This heat map also demonstrates the cell line-dependent effects that HDAC inhibitors generate (compare top left and right quadrants between T24 and MDA cells in Fig. 2B). Similar cluster patterns for cell lines and treatments were obtained using K-means and self-organizing maps (P < 0.01) from the Rosetta Resolver analysis (data not shown). Affymetrix analysis and hierarchical clustering using Spotfire Pearson’s correlation also demonstrated similar clustering patterns for these compounds (data not shown). Using these different clustering algorithms, MS-275, a benzamide HDAC inhibitor (4), produced a gene expression pattern significantly different from the hydroxamate HDAC inhibitors SAHA and TSA (R² value of 0.59 for MS-275 versus SAHA and TSA compared with 0.91 for SAHA versus TSA in T24 cells; Fig. 2A). This different gene expression pattern for MS-275 was also evident in the repeat gene array experiment in T24 cells (data not shown). This difference in clustering is consistent with differences observed in cellular effects. MS-275 did not cause the accumulation of acetylated α-tubulin in T24 (Fig. 3) or in HeLa S3 (data not shown) cells when compared with the other HDAC inhibitors such as SAHA and TSA. The two negative controls used in this experiment, 812, an inactive analogue of SAHA, and 377, an inactive analogue of MS-275 (Table 2), clustered separately from those of their HDAC active counterparts (R² of 0.03–0.14). (Fig. 2A) These analogues have no activity against isolated HDACs or cellular effects such as histone hyperacetylation or inhibition of proliferation (Table 2). Although 812 regulated a significant number of genes (Table 3), they are largely independent of the genes regulated by active HDAC inhibitors. This is consistent with its differential cellular effects (Fig. 1). These results also suggest that the gene changes observed with HDAC inhibitors are related to their enzyme inhibitory activity (mechanism) rather than the general chemical structure of the compound.

Histone hyperacetylation and p21 expression are observed in carcinoma cells treated with SAHA at 5 μm. Yet, in NHDFs this concentration induced hyperacetylation of histone H4 but did not induce p21 expression (Fig. 4A). Consequently the set of genes perturbed by 5 μm SAHA in NHDFs clustered outside the other HDAC inhibitors and looked more similar to the inactive controls. However, it should be noted that at this concentration of SAHA there was <50% effect on cell growth (Fig. 4B), whereas in carcinoma cells this concentration of SAHA usually produces >50% inhibition of cell growth. These results suggest that induction of p21 is necessary to observe the full spectrum of gene changes induced by HDAC inhibitors. These results also highlight the differences observed between normal cells (dermal fibroblasts) and carcinoma cells (30, 31).

Defining a Core Set of Genes Regulated by HDAC Inhibitors.

Using a cutoff of 2-fold change in expression level as being significant, a common (core) set of genes was identified that were positively or negatively regulated by all of the HDAC inhibitors in all of the cell lines tested. Thirteen genes were identified as the core set; 8 genes were found to be up-regulated and 5 genes down-regulated (Table 4). Each was a single experiment, and the studies in this analysis comprised a total set of 13 chips including controls. Genes most dominantly up-regulated were p21, Hep27 (a SCAD), and TRPM-2 or clusterin. The most prominent genes down-regulated by HDAC inhibitor treatment were thymidylate synthetase and CTP synthase, both of which are involved in DNA synthesis and which correlate with cell cycle arrest induced by HDAC inhibitors. This gene expression profile was confirmed in a second array experiment in T24 cells (Table 5). Several genes in this second array did not confirm from the first set of arrays including α-tubulin, metallothionein 1L, and transformation-related protein. All of the genes in the core set were then investigated in dose-response analyses using Q-PCR.

Confirmation of Gene Expression by Q-PCR and Western Blotting.

The core genes were evaluated by Q-PCR to determine the effects of SAHA and MS-275 on specific gene mRNA levels in T24 bladder carcinoma cells. Because these were evaluated in T24 cells, gelsolin was included as one of the genes evaluated. Gelsolin expression was not included in the core gene set, because in MDA breast carcinoma cell lines gelsolin levels are relatively high (Western blot data not shown). As shown in Fig. 5, A and C, between 1.5 and 15 μm, SAHA produced a dose-dependent increase in gene expression for p21, TRPM, gelsolin, Hep27 (Fig. 5A), α-fucosidase, α-tubulin, glutaredoxin, and metallothionein 1L (Fig. 5C). Hep27 and TRPM were highly induced by 15 μm SAHA demonstrating increases of 62- and 38-fold, respectively. Comparison of the induction of these genes at 5 μm between Q-PCR and microarray demonstrates a quite similar level of induction with the exception of p21 (Table 6). Likewise, most of the genes that demonstrated down-regulation from the gene array studies were similarly down-regulated in Q-PCR analysis (Fig. 5, B and D). Thymidylate synthetase was the most prominently down-regulated gene achieving 40–100-fold down-regulation by Q-PCR at 5 and 15 μm SAHA, respectively. MS-275 also demonstrated a good correlation between genes up-regulated and down-regulated by gene array and Q-PCR (Fig. 6, A and B). Table 6 summarizes the data for gene array and Q-PCR at the 5 μm dose of SAHA. Two genes did not show a >2-fold change in gene expression by Q-PCR, importin β and α-tubulin, and one gene had an opposite response, histone H2B. The results for importin β and α-tubulin are supported by the repeat microarray study in T24 cells (Table 4); however, the levels of histone H2B are consistently up-regulated in SAHA-treated cells. This discrepancy may be resolved by investigating the timing of regulation of histone 2B mRNA.

To additionally confirm the regulation of these genes by HDAC inhibitors, Western blot analysis was performed for p21 (Fig. 1B), gelsolin (Fig. 7A), and thymidylate synthetase (Fig. 7B). Gelsolin is up-regulated at the protein level by multiple HDAC inhibitors in the T24 bladder carcinoma line. Experiments were performed to demonstrate that a reduction in gene expression also correlated with a reduction in protein amount for thymidylate synthetase. Thymidylate synthetase protein expression was evaluated using TSA at 300 nm, SAHA and MS-275 at 5 μm; all of the treatments demonstrated a down-regulation of thymidylate synthetase protein expression in T24 carcinoma cells after a 24-h treatment.

Discussion

HDAC inhibitors have provided valuable tools to study the role of histone acetylation and the regulation of gene transcription. Through their ability to modulate gene transcription, HDAC inhibitors have shown promise as cancer therapeutics where dysregulation of normal transcriptional events has been widely used by cancers to escape apoptosis in the face of high genetic instability (2, 3). To this end we have attempted to identify the overlapping set of genes regulated in three cell lines by several reference small molecule HDAC inhibitors. To determine this set we used two hydroxamate HDAC inhibitors, TSA (27) and SAHA (5), and a benzamide inhibitor MS-275 (4), and compared their activity in three different cell lines, T24, MDA435, and MDA468. The available literature suggests that these inhibitors do not show any selectivity between the different isoforms of the HDAC family (3, 5, 32). Only 13 genes were modulated by >2-fold (P ≤ 0.01) by all of the inhibitors in all three of the cell lines.

The fact that so few genes form the core set of HDAC inhibitor-regulated genes is surprising considering the general mechanism of chromatin remodeling being interrupted by these inhibitors. Although chromatin remodeling is a basic mechanism to control gene transcription, these results emphasize the intricate and multifaceted processes that actually control transcription and expression of proteins. Other researchers have evaluated HDAC inhibitors such as TSA on gene transcription by differential display in lymphoid cell lines and observed only ∼2% (8 of 340 genes) of the genes evaluated being regulated by the induction of histone hyperacetylation (28). In colon carcinoma cells ∼7% of genes analyzed were regulated by sodium butyrate treatment (256 genes elevated and 333 genes repressed of 8063 genes; Ref. 33). Our results in bladder carcinoma and breast carcinoma cells demonstrate that approximately 8–10% of genes are regulated on the Affymetrix 6800 gene chips by various HDAC inhibitors. These results emphasize the fact that disruption of the deacetylation component of chromatin remodeling does not cause global changes in gene expression.

Inhibition of histone deacetylation results in the hyperacetylation of core nucleosomal histones, primarily H3 and H4. Our results demonstrate that three structurally distinct HDAC inhibitors generate this phenotype in cells. The resulting increase in nucleosomal acetylation would generally be thought to result in the increased transcription of various genes; however, HDAC inhibitors tend to down-regulate as many genes as are up-regulated in these cell lines. Repression of gene transcription by inhibitors of HDACs is rather counterintuitive yet has been observed by many different investigators (28, 29, 33). The down-regulation of a gene could result from the direct effect of acetylation of histones that are naturally unacetylated and, therefore, results in blocking of the necessary transcription machinery; alternatively, hyperacetylation could result in the transcription of a regulatory factor that negatively regulates gene transcription. Evaluation of expression profiles using larger genome chips could also help identify additional core genes as well as the genes that may be altered resulting in repression of genes on the 6800 gene chip that was observed in these experiments.

Various studies have probed the gene expression profiles of a single HDAC inhibitor usually in a single cell line (26, 28, 29, 34). The experiments presented herein provide the analysis of multiple HDAC inhibitors, SAHA, TSA, and MS-275, in multiple cell lines, one bladder and two breast carcinomas. The analysis of this gene expression data demonstrates that, as one might expect, the gene expression profiles of various HDAC inhibitors, TSA, SAHA, and MS-275, are very similar. However, these data also clearly demonstrate that differences in these HDAC inhibitors can be uncovered by gene expression profiling, because MS-275 produces a distinctly different expression profile when compared with the two hydroxamate HDAC inhibitors, TSA and SAHA. This may be related to their distinct differences in potency against the nuclear HDAC enzymes; TSA = 3 nm, SAHA = 30 nm, and MS-275 is ∼5 μm against nuclear HDAC preparations. However, potency at the cellular level is essentially equivalent between SAHA and MS-275 (0.5–5.0 μm range in tissue culture). The difference observed for MS-275 is consistent within multiple experiments, and when using different algorithms to analyze and cluster the data generated by microarray. Another example of the difference of MS-275 can be taken from the microarray studies of sodium butyrate and TSA in HT29 colon carcinoma cells (34), and comparison with the data herein. A gene involved in control of the cell cycle, tob-1, is up-regulated 10–12-fold in HT29 cells (34); similarly, in our experiments in T24 cells tob-1 is up-regulated by TSA and SAHA 4–5-fold, but is not regulated by MS-275 in the same experiment (data not shown). Unlike the other HDAC inhibitors evaluated in all of our studies, MS-275 does not cause the accumulation of acetylated α-tubulin in cells. HDAC 6 has been described recently as the deacetylase that regulates the acetylation level of tubulin in mammalian cells (35). A difference in the effects of MS-275 on HDAC 6 and/or the tubulin acetylation complex could be similar to the lack of effect of sodium butyrate and trapoxin B in vitro on the HDAC 6 enzyme (32) but has not been demonstrated for MS-275. Therefore, the gene expression cluster analysis appears to be supported by the cellular phenotype generated by MS-275. This demonstrates the utility of gene expression profiling in the differentiation of HDAC inhibitors of different structural classes and/or possibly selectivity profiles versus HDACs or HDAC complexes in cells.

Even with these differences seen with MS-275, a core or common set of genes that are regulated by HDAC inhibitors was established from these studies. It is important to note that although this is data from three different cell lines and three different inhibitors, this core set is still defined by those parameters. Extrapolation of the results from these or other in vitro, two-dimensional culture conditions to tumor models or primary tumors where cells grow in a three-dimensional matrix containing other cellular constituents would be at best tenuous; however, studies are ongoing to address this issue using a flank tumor xenograft model where the response to HDAC inhibitors for these 13 core genes can be studied in a three-dimensional matrix environment. It is interesting that of the 8–10% of genes regulated by these HDAC inhibitors, only 13 are regulated in common. When these genes are confirmed by Q-PCR, 8 of the 13 genes appear to be regulated significantly. Therefore, the total number of commonly regulated genes confirmed is only 1–2% of the total genes regulated by HDAC inhibitors. The significance of these results is affirmed by the lack of correlation to genes altered by the inactive analogues of the HDAC inhibitors used in these studies. The inactive analogue of SAHA, 812, is only subtly different on a structural basis and results in total loss of HDAC inhibitory activity. None of the original 13 genes of the core set are regulated significantly by 812 (>2-fold). The fact that such small structural changes in the molecule alter its biological function so dramatically emphasizes the mechanistic basis for the gene changes observed with SAHA. The MS-275 analogue, 377, alters just the orientation of the amide bond connection to the anilide moiety, again resulting in loss of HDAC inhibitory activity and a substantially different expression profile. The gene tob1 is an example of one of the genes that is regulated by SAHA in different cell types, also demonstrating a 4-fold induction in T24 cells but is not regulated by the other HDAC inhibitors (e.g., MS-275) to classify it as a core gene (34). Likewise, gelsolin is a gene highly regulated by HDAC inhibitors in T24 cells and other cells that have down-regulated its expression as a component of malignant progression but is normally expressed in MDA cells, and its expression at the protein level is not regulated by HDAC inhibitors. Therefore, because it is not regulated in MDA cells, it is not included in this core set of genes regulated by HDAC inhibitors. These results demonstrate additionally the cell line dependency of HDAC inhibitor effects.

In our experiments the core set of genes defined by microarray analysis comprise genes that are involved in cell cycle progression, DNA synthesis, and apoptosis. These genes would be expected to be regulated in these cells, as the HDAC inhibitors produce a profound cell cycle arrest and induce a differentiated phenotype and/or apoptosis in most carcinoma cell lines (17). The cyclin-dependent kinase inhibitor, p21, is markedly up-regulated by HDAC inhibitors in most cell types, both mRNA and protein (18–20, 36). p21 is a common gene regulated by TSA and sodium butyrate in most other gene microarray analyses, usually 2–5-fold depending on cell line and exposure time (29, 33, 34, 37). These changes in gene expression are supported consistently by induction of p21 protein by Western blot analysis. Another cell cycle regulator, p27^KIP1, has also been demonstrated to be elevated in some carcinoma cells by HDAC inhibitors, but this is not as common as p21 induction (38). Regulation of p21 seems crucial for the activity of HDAC inhibitors on cells, and the ability to substantially up-regulate p21 may govern the fate of the cell, cytostasis or apoptosis, after HDAC inhibition. The tob1 gene is another interesting cell cycle regulator induced by HDAC inhibitors (34). In our experiments tob1 is up-regulated by TSA (4.7-fold) and SAHA (3.7-fold) but not by MS-275 in T24 cells (data not shown). Therefore, tob1 is one of the genes that distinguishes MS-275 from the hydroxamate HDAC inhibitors. One possible function of tob1 is to induce cell cycle arrest in the absence of a functional p21 as demonstrated by the arrest of p21^−/− cells by butyrate. Therefore, HDAC inhibitors have the ability to alter cell cycle progression at multiple levels in cancer cells. It has been demonstrated that HDAC inhibitors produce cell cycle arrest in a p53-independent manner (36). This p53 independence results from induction of p21 and/or tob1 by HDAC inhibitors through a direct effect on the Sp1 site in the p21 promoter (18). Because most cancer cells have lost p53 or Rb or both, and, therefore, have lost the G₁/S DNA damage checkpoint, the induction of p21, p27^Kip1, and/or tob1 by HDAC inhibitors produces an aberrant cell cycle arrest (checkpoint), and may lead to apoptosis in carcinoma cells.

HDAC inhibitors also up-regulate Hep27, a SCAD first described as a nuclear protein induced by sodium butyrate (39–41). Hep27 has an unknown function in the cell but may be involved in the metabolism of nuclear hormones that regulate cell cycle progression or induction of differentiation, a function of other known SCADs (40). This gene has also been described in other gene array experiments as being up-regulated by TSA (42). Additional studies on the role this novel alcohol dehydrogenase plays in the cell response to HDAC inhibition may shed light on the normal function of Hep27. Similarly, another gene regulated by HDAC inhibition and also reported by other investigators is metallothionein 1L (42). Metallothionein 1L is a putative member of the metallothionein 1 family that binds divalent cations in the cell. However, the sequence for metallothionein 1L possesses a stop codon at position 26 in the sequence and, therefore, may not be expressed as a functional protein. Why HDAC inhibition would result in the regulation of these genes is unknown; however, one could speculate that HDAC inhibition results in changes that alter both nuclear hormone responses and transcription factors or the necessary cofactors that regulate the function of these factors. These responses are all components of the cell cycle arrest associated with alteration of chromatin function in tumor cells.

Cell cycle arrest or inhibition of proliferation can be caused by inhibition of the machinery necessary for DNA synthesis, e.g., thymidylate synthetase and CTP synthase, or through the induction of apoptotic genes, e.g., TRPM-2. The apoptotic gene clusterin, TRPM-2, is significantly up-regulated by HDAC inhibition in these cells. The nuclear form of TRPM-2, clusterin N, may be a major factor in the induction of apoptosis in these carcinoma cells (43). By down-regulation of genes involved in DNA synthesis, thymidylate synthetase (44), and CTP synthase, HDAC inhibition resembles the effects of antimetabolites commonly used in chemotherapy, which target these enzymes, e.g., 5-fluorouracil. Therefore, most of the genes commonly regulated by HDAC inhibitors are involved in regulation of cell cycle progression, proliferation, and apoptosis of cells according to the major phenotypic change observed with HDAC inhibitors in carcinoma cells.

Using gene expression profiling, our results demonstrate that these expression profiles can be used to differentiate structurally distinct HDAC inhibitors, e.g., hydroxamate- versus nonhydroxamate-type inhibitors. In each cell line evaluated, T24 and MDA 435, MS-275 clearly produced a different expression pattern than either of the hydroxamate HDAC inhibitors, TSA or SAHA. These observations on gene regulation are supported by a functional difference observed with MS-275, lack of acetylation of α-tubulin. Our studies also demonstrate that in three different cell lines and using three different HDAC inhibitors, <10% of genes (of 6800 evaluated) were regulated either positively or negatively by inhibition of HDACs. Therefore, in the face of global changes in histone acetylation generated by these HDAC inhibitors only a small fraction of genes is regulated. A somewhat counterintuitive observation is that as many genes are down-regulated as are activated in the face of histone hyperacetylation, highlighting the complexity of the regulation of gene transcription through chromatin remodeling. The overlapping set of genes that are up- or down-regulated is characteristic of the effects that HDAC inhibitors have on these cells, e.g., inhibition of cell growth, as reflected in changes in the expression of the genes p21, TRPM-2, thymidylate synthetase, and CTP synthase, and possibly Hep27. The uniqueness of the regulation of these genes by HDAC inhibitors is confirmed by the lack of effects that structurally similar yet inactive analogues of these HDAC inhibitors have on the cells. One gene, Hep27, seems to be a unique gene of a nuclear protein regulated by HDAC inhibitors that is a member of the SCAD family and of which the function is still uncertain but may regulate proliferative responses in the nucleus.

We have analyzed the gene expression patterns of three different HDAC inhibitors in three different cell lines using microarray studies and identified a core set of 13 genes that are regulated by all of these HDAC inhibitors in all of the cell lines evaluated. Analysis of the expression profiles revealed a strong cell line dependence of gene changes induced by small molecule HDAC inhibitors. Expression profiling also distinguished hydroxamate and nonhydroxamate inhibitors, as well as potent HDAC inhibitors from structurally related inactive analogues. These studies clearly demonstrate the utility of expression profiling in identifying subtle differences in HDAC inhibitors that may affect both the efficacy and tolerability of these agents. The identification of a common set of genes modulated by multiple HDAC inhibitors also provides valuable information regarding possible clinical utility of these agents to enhance the therapeutic potential of conventional chemotherapeutics (as demonstrated for 5-fluorouracil based on the results presented herein; Ref. 45) or, based on the cell or tumor type-specific genes regulated by HDAC inhibitors, the ability to identify novel therapeutic targets in cancers.