This review focuses on the mechanisms of action of histone deacetylase (HDAC) inhibitors (HDACi), a group of recently discovered ‘targeted’ anticancer agents. There are 18 HDACs, which are generally divided into four classes, based on sequence homology to yeast counterparts. Classical HDACi such as the hydroxamic acid-based vorinostat (also known as SAHA and Zolinza) inhibits classes I, II and IV, but not the NAD+-dependent class III enzymes. In clinical trials, vorinostat has activity against hematologic and solid cancers at doses well tolerated by patients. In addition to histones, HDACs have many other protein substrates involved in regulation of gene expression, cell proliferation and cell death. Inhibition of HDACs causes accumulation of acetylated forms of these proteins, altering their function. Thus, HDACs are more properly called ‘lysine deacetylases.’ HDACi induces different phenotypes in various transformed cells, including growth arrest, activation of the extrinsic and/or intrinsic apoptotic pathways, autophagic cell death, reactive oxygen species (ROS)-induced cell death, mitotic cell death and senescence. In comparison, normal cells are relatively more resistant to HDACi-induced cell death. The plurality of mechanisms of HDACi-induced cell death reflects both the multiple substrates of HDACs and the heterogeneous patterns of molecular alterations present in different cancer cells.
Acetylation and deacetylation of histones play an important role in transcription regulation of eukaryotic cells (Lehrmann et al., 2002; Mai et al., 2005). The acetylation status of histones and non-histone proteins is determined by histone deacetylases (HDACs) and histone acetyl-transferases (HATs). HATs add acetyl groups to lysine residues, while HDACs remove the acetyl groups. In general, acetylation of histone promotes a more relaxed chromatin structure, allowing transcriptional activation. HDACs can act as transcription repressors, due to histone deacetylation, and consequently promote chromatin condensation. HDAC inhibitors (HDACi) selectively alter gene transcription, in part, by chromatin remodeling and by changes in the structure of proteins in transcription factor complexes (Gui et al., 2004). Further, the HDACs have many non-histone proteins substrates such as hormone receptors, chaperone proteins and cytoskeleton proteins, which regulate cell proliferation and cell death (Table 1). Thus, HDACi-induced transformed cell death involves transcription-dependent and transcription-independent mechanisms (Marks and Dokmanovic, 2005; Rosato and Grant, 2005; Bolden et al., 2006; Minucci and Pelicci, 2006).
In humans, 18 HDAC enzymes have been identified and classified, based on homology to yeast HDACs (Blander and Guarente, 2004; Bhalla, 2005; Marks and Dokmanovic, 2005). Class I HDACs include HDAC1, 2, 3 and 8, which are related to yeast RPD3 deacetylase and have high homology in their catalytic sites. Recent phylogenetic analyses suggest that this class can be divided into classes Ia (HDAC1 and -2), Ib (HDAC3) and Ic (HDAC8) (Gregoretti et al., 2004). Class II HDACs are related to yeast Hda1 (histone deacetylase 1) and include HDAC4, -5, -6, -7, -9 and -10 (Bhalla, 2005; Marks and Dokmanovic, 2005). This class is divided into class IIa, consisting of HDAC4, -5, -7 and -9, and class IIb, consisting of HDAC6 and -10, which contain two catalytic sites. All class I and II HDACs are zinc-dependent enzymes. Members of a third class, sirtuins, require NAD+ for their enzymatic activity (Blander and Guarente, 2004) (see review by E Verdin, in this issue). Among them, SIRT1 is orthologous to yeast silent information regulator 2. The enzymatic activity of class III HDACs is not inhibited by compounds such as vorinostat or trichostatin A (TSA), that inhibit class I and II HDACs. Class IV HDAC is represented by HDAC11, which, like yeast Hda 1 similar 3, has conserved residues in the catalytic core region shared by both class I and II enzymes (Gao et al., 2002).
HDACs are not redundant in function (Marks and Dokmanovic, 2005; Rosato and Grant, 2005; Bolden et al., 2006). Class I HDACs are primarily nuclear in localization and ubiquitously expressed, while class II HDACs can be primarily cytoplasmic and/or migrate between the cytoplasm and nucleus and are tissue-restricted in expression.
The structural details of the HDAC–HDACi interaction has been elucidated in studies of a histone deacetylase-like protein from an anerobic bacterium with TSA and vorinostat (Finnin et al., 1999). More recently, the crystal structure of HDAC8–hydroxamate interaction has been solved (Somoza et al., 2004; Vannini et al., 2004). These studies provide an insight into the mechanism of deacetylation of acetylated substrates. The hydroxamic acid moiety of the inhibitor directly interacts with the zinc ion at the base of the catalytic pocket.
This review focuses on the molecular mechanisms triggered by inhibitors of zinc-dependent HDACs in tumor cells that explain in part: (I) the effects of these compounds in inducing transformed cell death and (II) the relative resistance of normal and certain cancer cells to HDACi induced cell death. HDACi, for example, the hydroxamic acid-based vorinostat (SAHA, Zolinza), are promising drugs for cancer treatment (Richon et al., 1998; Marks and Breslow, 2007). Several HDACi are in phase I and II clinical trials, being tested in different tumor types, such as cutaneous T-cell lymphoma, acute myeloid leukemia, cervical cancer, etc (Bug et al., 2005; Chavez-Blanco et al., 2005; Kelly and Marks, 2005; Duvic and Zhang, 2006) (Table 2). Although considerable progress has been made in elucidating the role of HDACs and the effects of HDACi, these areas are still in early stages of discovery.
Recent phylogenetic analyses of bacterial HDACs suggest that all four HDAC classes preceded the evolution of histone proteins (Gregoretti et al., 2004). This suggests that the primary activity of HDACs may be directed against non-histone substrates. At least 50 non-histone proteins of known biological function have been identified, which may be acetylated and substrates of HDACs (Table 1) (Glozak et al., 2005; Marks and Dokmanovic, 2005; Rosato and Grant, 2005; Bolden et al., 2006; Minucci and Pelicci, 2006; Zhao et al., 2006). In addition, two recent proteomic studies identified many lysine-acetylated substrates (Iwabata et al., 2005; Kim et al., 2006). In view of all these findings, HDACs may be better called ‘N-epsilon-lysine deacetylase’. This designation would also distinguish them from the enzymes that catalyse other types of deacetylation in biological reactions, such as acylases that catalyse the deacetylation of a range of Nα-acetyl amino acids (Anders and Dekant, 1994).
Non-histone protein targets of HDACs include transcription factors, transcription regulators, signal transduction mediators, DNA repair enzymes, nuclear import regulators, chaperone proteins, structural proteins, inflammation mediators and viral proteins (Table 1). Acetylation can either increase or decrease the function or stability of the proteins, or protein–protein interaction (Glozak et al., 2005). These HDAC substrates are directly or indirectly involved in many biological processes, such as gene expression and regulation of pathways of proliferation, differentiation and cell death. These data suggest that HDACi could have multiple mechanisms of inducing cell growth arrest and cell death (Figure 1).
HDACi have been discovered with different structural characteristics, including hydroximates, cyclic peptides, aliphatic acids and benzamides (Table 2) (Miller et al., 2003; Yoshida et al., 2003; Marks and Breslow, 2007). Certain HDACi may selectively inhibit different HDACs. For example, MS-275 preferentially inhibits HDAC1 with IC50, at 0.3 μM, compared to HDAC3 with an IC50 of about 8 μM, and has little or no inhibitory effect against HDAC6 and HDAC8 (Hu et al., 2003). Two novel synthetic compounds, SK7041 and SK7068, preferentially target HDAC1 and 2 and exhibit growth inhibitory effects in human cancer cell lines and tumor xenograft models (Kim et al., 2003a). A small molecule, tubacin, selectively inhibits HDAC6 activity and causes an accumulation of acetylated α-tubulin, but does not affect acetylation of histones, and does not inhibit cell cycle progression (Haggarty et al., 2003). No other HDACi for a specific HDAC has been reported.
HDACi selectively alters gene expression
HDACi can affect transcription by inducing acetylation of histones, transcription factors and other proteins regulating transcription (Table 1) (Glozak et al., 2005; Marks and Dokmanovic, 2005; Bolden et al., 2006; Minucci and Pelicci, 2006). Early differential display experiments with lymphoid cell lines cultured with TSA showed that only 2% of 340 genes examined were altered in their expression, either increased or decreased, compared to untreated cells (Van Lint et al., 1996). Recent studies using cDNA arrays showed as many as 7–10% of genes were altered in their expression in cell lines of leukemia, multiple myeloma, and carcinomas of colon, bladder, kidney, prostate and breast, cultured for up to 48 h with butyrate, TSA, MS-275, vorinostat or FK228 (depsipeptide) using twofold change as the cut-off value (Chambers et al., 2003; Glaser et al., 2003; Mitsiades et al., 2004; Peart et al., 2005; Sasakawa et al., 2005). The time of culture, concentration and the HDACi used affect the number of genes detected with altered transcription. Short time points (Chambers et al., 2003; Mitsiades et al., 2004; Sasakawa et al., 2005) and low concentrations (Glaser et al., 2003) cause fewer changes in gene transcription, while the magnitude of change and number of genes altered increase to certain extent with the increase of time of culture and concentration of HDACi. Some changes in gene expression are probably direct effects of the HDACi on the gene promoter and other secondary and downstream effects. The patterns of alterations of gene expression are similar for different HDAC inhibitors, but show definite differences induced by different agents in various transformed cells (Glaser et al., 2003; Gray et al., 2004; Mitsiades et al., 2004; Peart et al., 2005; Sasakawa et al., 2005). In these several studies, it has been found that HDACi induce about as many genes as are repressed.
The cyclin-dependent kinase (CDK) inhibitor p21 (WAF1/CIP1) is one of the most common genes induced by HDACi (Archer et al., 1998; Richon et al., 2000; Sasakawa et al., 2002). HDACi-induced expression of p21 is independent of p53 and correlates with an increase in the acetylation of histones associated with the p21 promoter region (Richon et al., 2000; Gui et al., 2004). In ARP-1 cells, vorinostat caused specific modifications in the pattern of acetylation and methylation of lysines in histones H3 and H4 associated with the proximal promoter of the p21 gene (Gui et al., 2004). These changes did not occur in the histones associated with the promoter region of the expressed p27 (KIP1) or the silent epsilon globin gene in ARP-1 cells, and neither gene was altered in its expression by vorinostat. The protein complex associated with the proximal promoter region of the p21 gene contained HDAC1 and -2, Myc, BAF155, Brg-1, GCN5, P300 and Sp1. Vorinostat caused a marked decrease in HDAC1 and Myc, and recruitment of RNA polymerase II, with little detectable changes in HDAC2 or other proteins in the complex. The loss of HDAC1 from the complex was not associated with a decrease in this protein in the nuclear extract (Gui et al., 2004). These findings suggest that the selective alteration of transcription of a gene by HDACi may be determined by the specific composition and configuration of proteins in the transcription factor complex including the HDACs.
HDAC activity is required for transcriptional activation mediated by signal transducer and activator of transcription 5 (STAT5) (Rascle et al., 2003). Inhibiting HDAC activity can prevent expression of genes for which STAT5 is required, and result in repression of their expression. HDACi-induced transcriptional repression of androgen receptor (AR) results from the induction of a suppressor complex (Wang et al., 2004).
HDACi-induced antitumor pathways
HDACi can induce transformed cell growth arrest, terminal differentiation, cell death and/or inhibition of angiogenesis (Figure 1). Normal cells are relatively resistant to HDACi-induced cell death (Burgess et al., 2004; Insinga et al., 2005; Ungerstedt et al., 2005). The cell death pathways identified in mediating HDACi-induced transformed cell death include apoptosis (Rosato and Grant, 2005; Bolden et al., 2006; Minucci and Pelicci, 2006) by the intrinsic (Ruefli et al., 2001) and extrinsic pathways, mitotic catastrophe/cell death (Qiu et al., 2000; Dowling et al., 2005; Xu et al., 2005), autophagic cell death (Shao et al., 2004), senescence (Xu et al., 2005) and reactive oxygen species (ROS)-facilitated cell death (Rosato and Grant, 2005; Ungerstedt et al., 2005). The response to HDACi appears to depend, in part at least, on the nature of HDACi, concentration and time of exposure, and importantly, the cell context.
HDACi induces cell cycle arrest
HDACi induce cell cycle arrest in both normal and transformed cells (Marks and Dokmanovic, 2005; Ungerstedt et al., 2005). Low concentrations of HDACi predominantly induce G1 arrest, while high concentrations induce both G1 and G2/M arrests (Richon et al., 2000). G1 and G2 arrests are largely associated with induction of p21, which inhibits CDKs regulating G1 progression (CDK4/6) and G1/S transition (CDK2), the activity of proliferating cell nuclear antigen that is required for DNA replication (Vidal and Koff, 2000), and cdc2/CDK1 that regulates G2/M transition. Loss of p21 abolishes HDACi-induced G1 arrest (Archer et al., 1998; Rosato et al., 2001; Xu et al., 2005). G1 arrest was observed in cells without p21 (Hitomi et al., 2003). In this case, HDACi may induce other CDK inhibitors that cause cell cycle arrest. TSA induced G1 arrest in human colon HCT116 p21−/− cells associated with the induction of p15 (INK4b), which is an inhibitor of the cyclin D-dependent kinases (Hitomi et al., 2003).
p27, which inhibits CDK4- and CDK2-containing complexes (Vidal and Koff, 2000), was induced by vorinostat and/or TSA, in leukemia cells K562 and LAMA-84 (Nimmanapalli et al., 2003), and breast cancer cells MCF-7 and MDA-MB-231 (Huang and Pardee, 2000). In cells cultured with HDACi, the increase of the CDK inhibitors and the decrease of cyclins may act together to account for the reduced CDK activity, causing dephosphorylation of retinoblastoma protein (Rb), which blocks E2F activities in the transcription of genes for G1 progression and G1/S transition (Bolden et al., 2006). Transformed cells sensitive to HDACi-induced cell death are generally cell growth-arrested with increase of p21 expression (Huang and Pardee, 2000; Xu et al., 2006). HDACi can kill both proliferating and non-proliferating cells (Burgess et al., 2004). This is in contrast to the action of many chemotherapeutic drugs, which are effective only on proliferating cells.
In some cells, the G1 arrest is associated with terminal differentiation (Marks et al., 1996). The property of vorinostat to induce differentiation of transformed cells was discovered before identifying its inhibitory activity against HDACs (Marks and Breslow, 2007).
HDACi activates the extrinsic apoptotic pathways
The extrinsic pathway of apoptosis is initiated by the binding of death receptors, including Fas (Apo-1 or CD95), tumor necrosis factor (TNF) receptor-1 (TNFR-1), TNF-related apoptosis-inducing ligand (TRAIL or Apo2-L) receptors (DR-4 and -5), DR-3 (Apo3) and DR-6, to their ligands, such as FasL, TNF, TRAIL and TL1A (Apo3L), leading to activation of caspase-8 and caspase-10 (Ashkenazi, 2002). HDACi can upregulate the expression of both death receptors and their ligands, in vitro and in vivo, in transformed cells, but not in normal cells (Nakata et al., 2004; Insinga et al., 2005). Fas and FasL were induced in human neuroblastoma cells by M-carboxycinnamic acid bihydroxamide (Glick et al., 1999), nude mice xenograft of human osteosarcoma cells by FK228, and mouse model of APL by VPA (Insinga et al., 2005). TRAIL and its receptor DR-5 were induced in the mouse model of APL by VPA (Insinga et al., 2005), and Jurkat human T-cell lymphoblast leukemia cells and SKW6.4B lymphoblast cells by LAQ824 (Rosato et al., 2006). HDACi upregulated DR-5 in various transformed cells. Expression of TNF-α was upregulated by FK228 in HL-60 and K562 cells. c-FLIP, an inhibitor of the death receptor pathway, was downregulated by HDACi (Nakata et al., 2004; Peart et al., 2005; Rosato and Grant, 2005; Sutheesophon et al., 2005).
HDACi-induced cell death can be reduced by chimeras of DR5-Fc (anti-TRAIL) or Fas-Fc (anti-Fas), monoclonal antibodies against FasL or TRAIL and small inhibitor RNA (siRNA) of TRAIL or Fas (Nakata et al., 2004; Insinga et al., 2005; Rosato and Grant, 2005). Taken together, the data indicate that the extrinsic apoptotic pathway can account for HDACi-induced cell death in many transformed cells.
HDACi activates the intrinsic apoptotic pathways
The intrinsic apoptosis pathway is mediated by mitochondria, with the release of mitochondrial intermembrane proteins, such as cytochrome c, apoptosis inducing factor (AIF) and Smac, and the consequent activation of caspases. It is regulated, in part, by pro- and antiapoptotic proteins of bcl-2 family (Jiang and Wang, 2004). Activation of the intrinsic apoptotic pathway is a major pathway for HDACi to induce cell death. By mechanisms that are still not well understood, HDACi leads to release of cytochrome c from mitochondrial intermembrane space and activation caspase-9 (Marks and Dokmanovic, 2005; Bolden et al., 2006). Overexpression of Bcl-2 or Bcl-XL, which protect mitochondria, inhibits HDACi-induced apoptosis. Inhibition of Bcl-2 by a chemical inhibitor HA14-1 increases HDACi-induced cell death (Xu et al., 2006).
HDACi alter the factors that mediate or regulate the intrinsic apoptosis pathway. Bid cleavage, which can initiate the intrinsic pathway (Bolden et al., 2006), occurred before mitochondrial disruption in CEM cells cultured with vorinostat or oxamflatin (Ruefli et al., 2001; Peart et al., 2003). HDACi upregulate proapoptotic proteins of Bcl-2 family, such as Bim, Bmf, Bax, Bak and Bik (Zhang et al., 2004; Zhao et al., 2005; Xu et al., 2006). The mechanism of this effect is not understood. It has been shown that vorinostat and TSA increase Bim transcription by increasing the activity of E2F1 (Zhao et al., 2005). HDACi decrease antiapoptotic proteins of Bcl-2 family, such as Bcl-2, Bcl-XL, Bcl-w and Mcl-1 (Zhang et al., 2004; Rosato et al., 2006; Xu et al., 2006). HDACi decrease the inhibitor of apoptosis (IAP) XIAP by suppressing its transcription (Zhang et al., 2004; Rosato et al., 2006), and survivin by inducing protein degradation (Rosato et al., 2006). HDACi cause the release of the mitochondrial intermembrane proteins, cytochrome c, AIF and Smac (Ruefli et al., 2001; Rosato et al., 2006), and may also increase their levels (Xu et al., 2006). The effects of HDACi to increase the proapoptotic proteins and decrease the antiapoptotic proteins are cell context dependent. The basal level of these proteins vary dramatically in different tumor cells even of the same type of cancer, for example, prostate, as do the changes in these proteins induced by HDACi (Xu et al., 2006).
Using the Eu-myc mouse model of B-cell lymphoma, vorinostat was found to selectively induce lymphoma cell death (in vivo), which was independent of p53 and death receptor pathways (Lindemann et al., 2007). Sensititivity to the HDACi was dependent on expression of the BH3 – only proteins and BID and BIM.
HDACi induces mitotic cell death
HDACi can induce mitotic defects associated with aberrant acetylation of histones in heterochromatin and centromere domains. Newly replicated chromatin contains acetylated histones. In culture with TSA, histones in newly synthesized chromatin remain acetylated, and this disrupts the structure and function of the centromere and the pericentric heterochromatin, with loss of binding to heterochromatin binding proteins (Taddei et al., 2001; Cimini et al., 2003). Histone acetylation interferes with histone phosphorylation and disrupts the function of mitotic spindle checkpoint proteins, such as BubR1, hBUB1, CENP-F and CENP-E (Dowling et al., 2005; Robbins et al., 2005). As a result, the cells show a transient arrest at prometaphase, followed with aberrant mitosis such as missegregation and loss of chromosomes, resulting in cell death by either apoptosis or, mitotic cell death/catastrophe (Qiu et al., 2000; Cimini et al., 2003; Xu et al., 2005). HDACi-induced α-tubulin acetylation does not affect mitosis, although α-tubulin is a component of mitotic spindle that mediates mitosis. The HDAC6-specific inhibitor tubacin induced α-tubulin acetylation, but did not affect cell cycle progression (Haggarty et al., 2003). FK228 caused mitotic arrest, but did not inhibit tubulin deacetylase HDAC6.
HDACi induces autophagic cell death and senescence
HeLa cells with Apaf-1 knockout or Bcl-XL overexpression were induced to autophagic cell death with autophagic vacuoles in the cytoplasm, when cultured with vorinostat or butyrate (Shao et al., 2004). In colon carcinoma cells, senescence phenotype was observed in vorinostat-induced polyploidy cells associated with mitotic defects (Xu et al., 2005).
ROS, thioredoxin and Trx binding protein 2 in HDACi-induced cell death
Accumulation of ROS occurs in transformed cells cultured with HDACi, such as vorinostat, TSA, butyrate or MS-275 (Ruefli et al., 2001; Rosato et al., 2003; Ungerstedt et al., 2005; Xu et al., 2006). Accumulation of ROS may be important in HDACi-induced cell death. ROS accumulation occurs within 2 h of culture with HDACi, before disruption of mitochondria. Free radical scavengers such as N-acetylcysteine decrease HDACi-induced apoptosis (Ruefli et al., 2001; Rosato et al., 2003).
In many transformed cells, ROS-oxidation–reduction pathways are important mechanisms of HDACi-induced transformed cell death (Ungerstedt et al., 2005; Xu et al., 2006). Thioredoxin (Trx) acts as a hydrogen donor required for activation of many proteins, including ribonucleotide reductase that is essential for DNA synthesis, and transcription factors, for example, nuclear factor κB (NF-κB), and is an antioxidant scavenger of ROS (Lillig and Holmgren, 2007). HDACi upregulates the expression of Trx binding protein 2 (TBP2) (Butler et al., 2002; Xu et al., 2006), which binds and inhibits Trx activity (Nishiyama et al., 1999), and can cause downregulation of Trx in transformed but not normal cells (Butler et al., 2002; Ungerstedt et al., 2005). Trx is an inhibitor of apoptosis signal regulating kinase 1 (ASK1) (Saitoh et al., 1998). ASK1 promotes apoptosis by activation of SET1-JNK and MKK3/MKK6-p38 signaling cascades, and by enhancing the expression of proapoptotic protein Bim through a positive feedback on E2F1 activity (Tan et al., 2006). Inhibition of Trx by TBP2 activates ASK1. Further, HDACi increase the expression of ASK1. These effects can act together to promote apoptosis. ROS accumulation may also be a consequence of apoptosis. The pan-caspase inhibitor Z-VAD-fmk blocked MS-275-induced apoptosis as well as ROS accumulation in human chronic lymphocytic leukemia cells (Lucas et al., 2004).
Antitumor effects of HDAC6 inhibition
HDAC6 is an unique HDAC, localized in the cytoplasm, where it associates with non-histone substrates, such as HSP90 and α-tubulin. It has two catalytic domains and a ubiquitin binding domain, named BUZ domain (Zhang et al., 2006; Zou et al., 2006). Overexpression of HDAC6 leads to deacetylation of α-tubulin and increases cell motility (Hubbert et al., 2002; Haggarty et al., 2003). HDAC6 can not only bind both mono and poly-ubiquitinated proteins but also promote its own mono-ubiquitination. Specific inhibition of HDAC6 activity or its downregulation by siRNA increases α-tubulin and HSP90 acetylation, which reduces cellular motility, and induces HSP90 client proteins degradation, cell growth inhibition and cell death (Bali et al., 2005; Kovacs et al., 2005). Acetylated HSP90 cannot form stable complex with client proteins and its deacetylation by HDAC6 is required to regenerate functional HSP90 (Aoyagi and Archer, 2005). HSP90 acetylation is associated with loss of function, and its client proteins such as pro-survival and pro-proliferation proteins Akt, Bcr-Abl, c-Raf and ErbB2 can be poly-ubiquitinated and degraded via proteosome (Bali et al., 2005; Chen et al., 2005). Inhibition of HDAC6 by either specific or pan-HDACi can trigger different mechanisms of cell death. HDAC6 may also be involved in upregulating p21. Runx2 (Cbfa1, AML-3) is a transcription factor that can bind HDAC6, recruit it to the p21 promoter and repress its expression. Inhibition of HDAC6 activity could result in induction of p21 and, consequently, cell cycle arrest (Westendorf et al., 2002).
Activation of protein phosphatase 1
HDACi can disrupt phosphatase complexes. Recombinant HDAC6 binds directly to protein phosphatase 1 (PP1) catalytic subunit. TSA disrupts endogenous HDAC6–PP1 complexes (Brush et al., 2004). HDAC1 and 10 are also components of cellular phosphatase complexes. HDACi disrupt HDAC–PP1 complexes and activate PP1, which inactivates Akt by dephosphorylation (Chen et al., 2005). The interactions between HDACs and PP1 provide a mechanism by which HDACi can cause simultaneous changes in cellular protein phosphorylation and acetylation, which contribute to the antitumor activity of HDACi.
Disruption of the function of chaperonin HSP90
HDACi can cause accumulation of acetylated HSP90 through HDAC6 inhibition, with consequent inactivation of HSP90 (Bali et al., 2005). As indicated above, this chaperone protein is essential for the stability and function of many client proteins, including steroid hormone receptors and protein kinases, that are crucial for numerous cell signaling processes and cellular homeostasis (Solit and Rosen, 2006). Recent studies have demonstrated both a direct physical interaction between HDAC6 and HSP90, and HDAC6 as a regulator of HSP90 activity, through its deacetylation (Bali et al., 2005; Kovacs et al., 2005). Considering the number of HSP90 client proteins, many molecular alterations can be anticipated as a result of HSP90 inactivation through HDAC6 inhibition by HDACi, or its downregulation by HDAC6 siRNA. The effects of inactivation of HSP90 include non-functional glucocorticoid receptor with defects in ligand binding, nuclear translocation and transcription activation (Kovacs et al., 2005).
AR can be acetylated and become more transcriptionally active, suggesting that some HDACi could stimulate cell proliferation (Fu et al., 2003). However, HDACi inhibit HDAC6 and consequently HSP90, which hypothetically can overcome the AR activation (as a consequence of its acetylation) by inducing its degradation. Further, HDACi suppress AR transcription (Wang et al., 2004). These findings suggest that HDACi may be promising for the treatment of hormone-refractory prostate cancer.
In human leukemia K562 cells, the knockdown of HDAC6 by siRNA or culture with HDACi LAQ824 induces acetylation of HSP90 and α-tubulin, and inhibits the binding of HSP90 to ATP (Bali et al., 2005), disrupting HSP90 chaperone function with client proteins, inducing poly-ubiquitinalation and partial depletion of Bcr-Abl. Targeted inhibition of HDAC6 led to polyubiquitylation and depletion of other pro-growth and pro-survival HSP90 client proteins, including Akt. These studies indicate that inhibition of HDAC6 leads to disruption of the antiapoptotic Akt pathway, either by Akt dephosphorylation and/or by its degradation.
Disruption of the aggresome pathway
HDAC6 is a component of the aggresome, a cellular structure that constitutes the major site of degradation for misfolded protein aggregates, both non-ubiquitinated and ubiquitinated misfolded proteins (Kawaguchi et al., 2003). Direction of misfolded proteins to aggresomes is essential for cell survival, since these proteins are susceptible to forming cytotoxic aggregates that can interfere with normal cell function. Aggresome formation requires the microtubule network and the microtubule-associated motor, dynein (Johnston et al., 2002). HDAC6 can bind p150, a component of dynein motor complex, and act as a bridge between the dynein motors and the ubiquitination process, directing the poly-ubiquitinated proteins to aggresome. HDAC6 has a high affinity for ubiquitin molecule (due to the presence of the ZnF-UBP or BUZ domain) and is involved in the transport of poly-ubiquitinated proteins (Boyault et al., 2006). HDAC6 deacetylase activity is important for transport of misfolded poly-ubiquitinated proteins to the aggresome, and loss of HDAC6 function makes cells more sensitive to misfolded protein stress induced by protease inhibitor and, as a consequence, cell death (Kawaguchi et al., 2003).
HDACi inhibits angiogenesis
HDACi can block tumor angiogenesis by inhibition of hypoxia inducible factors (HIF) (Liang et al., 2006). HIF-1 and HIF-2 are transcription factors for angiogenic genes (Brown and Wilson, 2004). The oxygen level can control HIF activity through two mechanisms. First, under normoxic conditions, HIF-1α binds to von Hippel–Lindau protein (pVHL) and is degraded by the ubiquitination–proteasome system. Second, HIF activity depends on its transactivation potential (TAP), which is affected by the interaction with the coactivator p300/CBP among others. This complex can be disrupted by Factor Inhibiting HIF (FIH). Hypoxic conditions activate HIF through repression of the hydroxylases responsible for HIF degradation and loss of function. Hypoxia is a common event in tumors. Hypoxic conditions can induce transcription activation of HDAC1, -2 and -3 in transformed cells (Liang et al., 2006). These class I HDACs downregulate expression of p53 and pVHL, which reduces FIH expression and consequently activates HIF-1α and promotes angiogenesis (Liang et al., 2006).
TSA, vorinostat, FK228, butyrate and LAQ824 were found to repress angiogenesis in vitro and in vivo, and reduce expression of pro-angiogenesis factors, including HIF-1α and VEGF (Deroanne et al., 2002; Bolden et al., 2006; Liang et al., 2006). HDACi can induce HIF-1α degradation by acetylation at Lys532, leading to the interaction with, and ubiquitination by pVHL (Jeong et al., 2002). Further, HDACi can induce HIF-1α degradation in a VHL-independent mechanism (Kong et al., 2006). Class II HDACs, HDAC4 or HDAC6, physically associate with HIF-1α, and their selective inhibition by siRNA induced HIF-1α degradation (Qian et al., 2006). Moreover, HIF-1α binds to HSP90, and HDACi can disrupt HSP90 chaperone function, exposing HIF-1α to proteasomal degradation. These observations suggest that HDAC6 should be inhibited for an effective abolishment of HIF-1α function.
Another mechanism by which HDACi disrupt HIF-1α function is through repression of its TAP. Low doses of HDACi, which did not induce HIF-1α degradation, did repress HIF-1α TAP under both normoxic and hypoxic conditions (Fath et al., 2006). This repression occurs due to the targeting of HIF-1α/p300 complex by HDACi. HDACi could also decrease HIF-1α activation by inhibiting HDAC7 activity, since under hypoxic conditions, HDAC7 translocates to the nucleus, where it can interact with HIF-1α and increase its transcriptional activity (Kato et al., 2004).
HDACi inhibit angiogenesis by preventing endothelial cells from responding to the angiogenic stimulus generated by VEGF (Deroanne et al., 2002). TSA and vorinostat inhibit VEGF-induced expression of VEGF receptors and neuropilin-1, and induction of semaphoring III expression in endothelial cells. The anti-angiogenic effects of HDACi can contribute their antitumor activities. These observations support the use of combination therapies with VEGF inhibitors.
Combination of HDACi with other antitumor agents
The HDACi have shown synergistic or additive antitumor effects with a wide range of antitumor reagents, including chemotherapeutic drugs, new targeted therapeutic reagents and radiation, by various mechanisms, some unique for particular combinations (Rosato and Grant, 2004; Bhalla, 2005; Marks and Dokmanovic, 2005; Bolden et al., 2006).
HDACi have shown synergy with chemotherapeutic agents, such as antimetabolites 5-fluorodeoxyuridine and gemcitabine, antitubule agents docetaxel and epothilone B, topoisomerase (Topo) II inhibitors doxorubicin, epirubicin, VP-16 (etopside) and ellipticine (Munster et al., 2001; Kim et al., 2003b), and DNA crosslinking agent cisplatin (Kim et al., 2003b). The synergistic effects may depend on the sequence of drug administration. For example, prior treatment with HDACi induced chromatin decondensation and increased Topo IIβ/DNA cleavable complex formation, resulting in synergy of HDACi plus Topo II inhibitors (Marchion et al., 2005). The reverse order of administration of the drugs resulted in antagonistic effects, or had no more effect than each drug alone (Kim et al., 2003b). Further, pretreament with HDACi had more effect (four times) than the reverse (1.8 times) in the combination with cisplatin, although the reverse was not antagonistic (Kim et al., 2003b). HDACi have also been reported to have synergy with transcription modulator all-trans retinoid acids, vitamin D3 and its analogs, DNA demethylating agent 5-aza-2′deoxycytidine, abl kinase inhibitor imatinib (Gleevec, STI571) in both imatinib-sensitive and imatinib-resistant chronic myelogenous leukemia (CML) cells, HSP90 inhibitor 17-ally-amino-demethoxy geldanamycin, proteasome inhibitor bortezomib (PS-341), trastuzumab (herceptin), which is a monoclonal antibody against Her-2/neu (erbB2) receptor, and radiotherapy (Nimmanapalli et al., 2003; Marks and Dokmanovic, 2005; Bolden et al., 2006).
Upregulation of death receptors and/or reducing the inhibitory regulators of death receptor pathway by HDACi sensitize tumor cells to TRAIL (Bolden et al., 2006). HDACi also achieve synergy with TRAIL by simultaneous activation of the intrinsic and the extrinsic apoptotic pathways, without changing the expression of TRAIL receptors or the inhibitory protein c-FLIP (Rosato and Grant, 2004).
Many kinase inhibitors, including CDK inhibitor flavopyridol, phosphatidylinocitol 3 kinase inhibitor LY294002, FLT3 inhibitor, PKC412 and MEK1/2 inhibitor PD184352, suppress HDACi-mediated p21 induction, and potentiate the cell killing effect of HDACi. Blocking NF-κB activation by I Kappa B Alpha (IκBα) phosphorylation inhibitor Bay 11-7082 markedly increase HDACi-induced apoptosis (Almenara et al., 2002; Rosato and Grant, 2005).
Clinical development of HDACi
The studies with tumor bearing animals and clinical trials of HDACi have been extensively reviewed elsewhere (Rosato and Grant, 2004; Bhalla, 2005; Marks and Dokmanovic, 2005; Bolden et al., 2006; Marks and Breslow, 2007) (Table 2). Vorinostat is the first of the new HDACi to be approved by Food and Drug Administration for the clinical use in cancer patients, namely the treatment of cutaneous T-cell lymphoma (CTCL) (Duvic et al., 2007; Garber, 2007). In a phase II study with orally administered vorinostat on 33 previously treated patients with refractory cutaneous T-cell lymphoma, partial response were observed in eight patients (24.2%) and 14 of 31 evaluable patients (45.2%) had pruritis relief (Duvic et al., 2007).
At least 14 different HDACi are in some phase of clinical trials as monotherapy or in combination with retinoids, taxols, gemcitabine, radiation, etc, in patients with hematologic and solid tumors, including cancer of lung, breast, pancreas, renal and bladder, melanoma, glioblastoma, leukemias, lymphomas, multiple myeloma (see National Cancer Institute website for CTEP clinical trials, ctep.cancer.gov or clinicaltrials.gov, and website of companies developing HDACi; Table 2).
The resistance to HDACi
Development of resistance to HDACi is a major concern as with any new antitumor therapy. In preclinical studies, resistance to HDACi-induced transformed cell death was observed in human bladder carcinoma cells (T24) and prostate cancer cells (PC3) (Butler et al., 2000; Richon et al., 2000; Xu et al., 2006). Although vorinostat achieved 24.2% response rate in a phase II trial on CTCL, a considerable proportion of patients with CTCL did not respond well (Duvic et al., 2007). Resistance has been observed in clinical trials with other HDACi in different tumors. The basis of resistance to HDACi is not well understood. High levels of Bcl-2 (Pommier et al., 2004), Trx (Powis et al., 2000) and peroxiredoxins (Chung et al., 2001) have been associated with resistance of transformed cells to chemotherapy, and may play a role in the resistance to HDACi. Upregulation of Trx protects normal cells against HDACi-induced cell death (Ungerstedt et al., 2005). Overexpression of Bcl-2 blocks HDACi-induced transformed cell death (Mitsiades et al., 2003). Peroxiredoxins reduce ROS generation (Kang et al., 1998) and may protect transformed cells from HDACi-induced cell death, which is strongly associated with ROS production (Rosato and Grant, 2005). Resistance to FK228 has been associated with multiple drug resistance, upregulation of and efflux by P-gp (MDR1) (Glaser, 2006). The MDR1-mediated resistance does not affect vorinostat (Ruefli et al., 2002).
Conclusions and perspectives
HDACs have multiple substrates involved in many biological processes, including proliferation, differentiation, apoptosis and other forms of cell death. Indeed, the fact that HDACs have histone and multiple nonhistone protein substrates suggests these enzymes should be referred to as ‘lysine deacetylases’. HDACi can cause transformed cells to undergo growth arrest, differentiation and/or cell death. Normal cells are relatively resistant to HDACi. HDACi are selective in altering gene expression, which may reflect, in part, the proteins composing the transcription factor complex to which HDACs are recruited. Both altered gene expression and changes in non-histone proteins caused by HDACi-induced acetylation play a role in the antitumor activity of HDACi. This is reflected in the different inducer-activated antitumor pathways in transformed cells (Figure 1). The functions of HDACs are not redundant. Thus, a pan-HDAC inhibitor such as vorinostat may activate more antitumor pathways and have therapeutic advantages compared to HDAC isotype-specific inhibitors.
Almost all cancers have multiple defects in the expression and/or structure of proteins that regulate cell proliferation and death. Compared to other antitumor reagents, the plurality of action of HDACi potentially confers efficacy in a wide spectrum of cancers, which have heterogeneity and multiple defects, both among different types of cancer and within different individual tumors of the same type. The multiple defects in a cancer cell may be the reason for transformed cells being more sensitive than normal cells to HDACi. Thus, given the relatively rapid reversibility of vorinostat inhibition of HDACs, normal cells may be able to compensate for HDACi-induced changes more effectively than cancer cells.
HDACi have synergistic or additive antitumor effects with many other antitumor reagents – suggesting that combination of HDACi and other anticancer agents may be very attractive therapeutic strategies for using these agents. Complete understanding of the mechanisms underlying the resistance and sensitivity to HDACi has obvious therapeutic importance. Targeting resistant factors will enhance the antitumor efficacy of HDACi. Identifying markers that can predict response to HDACi is a high priority for expanding the efficacy of these novel anticancer agents.
The studies reported in this paper from the authors' laboratory have been supported, in part, by grants from the National Institute of Health (P30CA08748-41), Jack and Susan Rudin Foundation, David H Koch Foundation, and the Prostate Cancer Research Award, Experimental Therapeutics Center at Memorial Sloan-Kettering Cancer Center and the DeWitt Wallace Research Fund. MSKCC and Columbia University jointly hold patents on hydroxamic acid-based polar compounds, including vorinostat (SAHA), that were exclusively licensed to Aton Pharma Inc., a biotechnology company that was acquired by Merck Inc. in April 2004. PAM was a founder of Aton and has a financial interest in Merck's further development of vorinostat.