Introduction

Post-translational modifications of proteins play a crucial role in the modulation of their function by altering protein–protein and protein–nucleic acid interactions as well as their subcellular localization and activity. Specialized enzymes catalyze the addition and elimination of these modifications at specific sites. For example, kinases and phosphatases are responsible for the addition and removal of phosphate groups, respectively, on factors regulating signaling pathways or the cell cycle, whereas ubiquitin ligases are mainly involved in protein targeting to the proteasome. Histone acetyl transferases (HATs) and histone deacetylases (HDACs) regulate the balance of lysine acetylation on histones and non-histone proteins (Fig. 1; Spange et al. 2009). Lysine acetylation is deeply implicated in the control of highly regulated biological functions. Thus, alteration of the acetylation status is involved in the development of many diseases, including tumorigenesis. Therefore, the control of the acetylation status by modulating HDAC activities is a promising approach to anticancer therapy. Over the last 30 years, numerous synthetic and natural products, including a broad range of dietary compounds, have been identified as HDAC modulators. This review focuses on promising modulators of HDAC activity from natural sources and discusses their potential for anticancer therapy.

Fig. 1
figure 1

Acetylation/deacetylation of lysine residues. The reaction of acetylation on the ε-nitrogen of lysine residues is catalyzed by enzymes within the histone acetyl transferase (HAT) family, using acetyl-coenzyme A (acetyl-CoA) as donor of acetyl moiety. This is a reversible reaction catalyzed by enzymes within the histone deacetylase (HDAC) family

Full size image

The HDAC family: classification, roles and substrates

HDAC isoforms

HDACs are ubiquitous enzymes, initially identified as being responsible for the deacetylation of histones (Fig. 1), and are capable of targeting many non-histone proteins as substrates (Table 1). The 18 HDACs identified in mammals are divided into four classes according to both their sequence identity and catalytic activity.

Table 1 Presentation of the different HDAC isoforms
Full size table

Class I includes HDAC1, 2, 3 and 8, which are located mainly in the nucleus. HDAC3 forms an exception to this class because it possesses, in addition to the nuclear localization signal (NLS) present in this HDAC class, a nuclear export signal (NES), which suggests a possible cytoplasmic localization. The activity of these enzymes is modulated when they interact with chromatin proteins such as Sin3 (Switch intensive 3), NuRDs (nucleosome remodeling and deacetylating) and Co-REST (Co-repressor of REST; Zhang et al. 1999).

Class II is divided into two subclasses: subclass IIa includes HDAC4, 5, 7 and 9 and subclass IIb consists of HDAC6 and 10. HDAC9 is localized exclusively to the nucleus, whereas HDAC4, 5 and 7 contain an NLS and NES. These three enzymes are localized to both the nucleus and the cytoplasm. The catalytic activity of class IIa HDACs is very low and is hardly detectable in vitro (Fischle et al. 2002). In addition, it has been shown in cellulo that these enzymes are often complexed with other HDACs, suggesting a role different from that of a deacetylase. For example, HDAC4 allows the recruitment of the SMRT/N-CoR-HDAC3 complex to specific promoters, leading to transcriptional repression of targeted genes (Fischle et al. 2002). HDAC6 is mainly localized to the cytoplasm, where it is involved in the deacetylation of α-tubulin and consequently, regulation of microtubule-dependent cell motility. Moreover, this enzyme is able to target the chaperon protein HSP90, which stabilizes client proteins (Bali et al. 2005). This HDAC has two catalytic domains arranged in tandem and one functional HUB (HDAC6-, USP3- and Brap2-related zinc finger motif) domain absent in other HDACs. This domain is located in the C-terminal region and acts as a signal for ubiquitination of HDAC6, leading to its degradation. HDAC10 possesses, like other class IIa HDACs, both NLS and NES domains. Although the catalytic domain is located in the N-terminal region, a second putative domain is located in the C-terminal region. In addition, two putative retinoblastoma (Rb) protein-binding domains are present, which could explain the involvement of this enzyme in regulating the cell cycle. HDAC10 interacts with six other HDACs (HDAC1, 2, 3, 4, 5 and 7), suggesting a recruiting role rather than acting as a deacetylase (Tong et al. 2002). Class III, also called sirtuins for their analogy to the protein SIR2 (Silent Information Regulator 2) in yeast, contains seven members, SIRT1 to 7. SIRT1, 6 and 7 are localized to the nucleus. SIRT1 positively regulates transcription factors such as the nuclear factor kappa-B (NF-κB), p53 and forkhead box (FOX) proteins and plays a role in chromatin condensation (i.e., histone deacetylation). SIRT6 and 7 are involved in DNA repair and ribosomal RNA transcription, respectively. SIRT2 is mainly localized to the cytoplasm and affects α-tubulin acetylation status. SIRT3, 4 and 5 are mitochondrial proteins. SIRT3 positively regulates the activity of the acetyl-Coenzyme A synthetase 2, leading to the production of acetyl-coenzyme A (Hallows et al. 2006; Schwer et al. 2006), and deacetylates complex I of the respiratory chain involved in ATP production (Ahn et al. 2008). Thus, SIRT3 regulates energetic cell homeostasis. SIRT4 lacks deacetylase activity; however, it has an ADP-ribosylase activity. This ribosylation inactivates glutamate dehydrogenase, which in turn inhibits insulin secretion in pancreatic β-cells (Argmann and Auwerx 2006). Lastly, SIRT5 deacetylates and thus activates carbamoyl phosphate synthetase 1, which promotes ammonia detoxification (Nakagawa and Guarente 2009; Nakagawa et al. 2009).

Class IV comprises only HDAC11, which is localized in the nucleus and has a catalytic domain in the N-terminal region. This enzyme is structurally related to HDAC3 and 8 of class I, but also has similarities with class II; for these reasons, this new class of enzyme was created (Gao et al. 2002).

The reaction of deacetylation

Classes I, II and IV possess a zinc-dependent mechanism of action. Their catalytic domain in the N- or C-terminal region forms a tubular pocket with the zinc ion at the bottom. This ion is stabilized by two aspartate and one histidine residue and forms electrostatic bonds with a catalytic water molecule and the oxygen of the acetyl functional moiety of the substrate. Due to an electron transfer between the aspartate and histidine residues, the catalytic water molecule is activated, allowing a nucleophilic attack on the carbon of the acetyl moiety. This reaction leads to an oxyanion tetrahedral intermediate stabilized by a tyrosine residue in the catalytic site. Then, the nitrogen of the substrate lysine involved in the bond with the acetate accepts a proton from an aspartate-histidine relay. Hence, the carbon–nitrogen bond is broken and both the deacetylated substrate and the acetate are released (Finnin et al. 1999; Hodawadekar and Marmorstein 2007).

Class III relies on a nicotinamide adenine dinucleotide (NAD+)-dependent mechanism. However, based on crystallographic studies, two alternative mechanisms have been suggested. In the first model, the nicotinamide group is hydrolyzed, allowing the nucleophilic attack of the acetate’s oxygen to the 1′ carbon of the remaining ribose. In the second model, the nucleophilic attack of the acetate’s oxygen at the 1′ carbon of the NAD+-ribose leads to the release of the nicotinamide. In both cases, once the acetyl group is bonded to the 1′ carbon of the ribose, there is a proton transfer facilitated by a histidine in the catalytic site, which renders the 2′ oxygen nucleophilic. This oxygen will then attack the carbon of the acetate, and this reaction will result in the formation of a cyclic intermediate. The nitrogen of the substrate’s lysine then accepts a proton from a histidine. This transfer then breaks the connection between the acetate and the lysine. Reaction products are nicotinamide, deacetylated protein and 2-O′-acetyl-ADP-ribose. The latter will be recycled with the addition of nicotinamide and removal of the acetyl group (Hodawadekar and Marmorstein 2007; Hoff et al. 2006).

Cellular targets of acetylation

HDAC activity is involved, due to the large range of substrates, in many cellular processes (Table 1). Indeed, gene expression is modulated by the acetylation status of histones and transcription factors. Acetylation of histone lysine residues modulates histone-DNA and histone-protein interactions as well as chromatin remodeling. Thus, hyperacetylation of histones is associated with a relaxed state of chromatin and active gene expression (Minucci and Pelicci 2006). Moreover, acetylation of transcription factors affects their cellular localization. Indeed, NF-κB, as well as STAT1 (signal transducer and activator of transcription 1) and STAT3, is translocated into the nucleus following acetylation of specific lysine residues, where they activate the transcription of target genes. Interestingly, the transactivation function of other transcription factors, such as p53 and FOXO proteins, is also positively regulated by acetylation. In addition, the activity of the Rb protein is modulated by the presence of acetyl groups blocking its cyclin E-CDK (cyclin-dependent kinase) 2-dependent phosphorylation. This acetylation-dependent hypophosphorylation leads to cell cycle arrest. Moreover, the acetylation of factors involved in maturation, stability and translation of messenger RNA can have an impact on their function. Indeed, the acetylation of hnRPA1 (heterogeneous nuclear ribonucleoprotein A1) and PAP (poly-[A]-polymerase) modifies the interaction between these proteins with pre-mRNA. Protein stability can also be affected by acetylation of proteins destined to the proteasome. Indeed, the acetylation of p53, p73, Smad7 (mothers against decapentaplegic homolog 7) and c-myc prevents their ubiquitination and thus their degradation. In contrast, acetylation of HIF-1α (hypoxia-inducible factor-1α) facilitates its interaction with ubiquitin-ligase E3 and transit to the proteasome. Interestingly, cell mobility is dependent on α-tubulin and cortactin acetylation status. Indeed, HDAC6- and SIRT2-mediated deacetylation of α-tubulin promotes microtubule depolymerization and therefore increases microtubule dynamics and cell mobility (Hubbert et al. 2002). In addition, deacetylated cortactin promotes cell motility by interacting with F-actin, leading to actin polymerization (Aldana-Masangkay and Sakamoto 2011).

Finally, the microtubule network is necessary for an active clearance of cytotoxic misfolded protein through the aggresome pathway (Johnston et al. 1998). This alternative pathway of protein degradation via aggresome organelles protects cells against an overloading of the ubiquitin proteasome-system (UPS) due to an accumulation of misfolded proteins in cellular stress conditions. In this system, HDAC6 that possesses a ZnF-UBP (zinc-finger ubiquitin binding) domain binds polyubiquitinated misfolded proteins and delivers them to p150(Glued), a component of dynein/dynactin microtubule motor complex. Misfolded protein are then transported to the MTOC (microtubule organizing center), where aggresomes are formed leading to the activation of the autophagic clearance activation that allow degradation of misfolded proteins in a proteasome-independent manner (Aldana-Masangkay and Sakamoto 2011; Lee et al. 2010).

HDAC inhibitors in cancer therapy

Epigenetic alterations and cancer

Carcinogenesis is not exclusively due to mutations of the nucleotide sequence but also involves epigenetic alterations. Thus, according to the major role played by protein acetylation in gene expression, a disruption in the balance between HAT and HDAC activities may contribute to the transformation of a normal cell to a cancerous cell (Fig. 2). Indeed, aberrant hypoacetylation of histones and non-histone proteins, caused by a decrease in HAT activities, aberrant HDAC recruitment to the promoter regions of certain genes (e.g., tumor suppressor genes [TSGs]) and/or increase in HDAC expression leading to excessive gene silencing, is frequently observed in many forms of cancer. More precisely, a decrease in HAT activities could be due to either mutations of HAT isoforms, such as p300, or chromosomal translocations. The latter leads to the formation of fusion proteins such as MLL-CBP, MLL-p300, MOZ-CBP, MOZ-p300 or MOZ-TIF2 involved in aberrant gene expression. For example, in a normal state, MOZ controls RUNX1 expression, which is involved in cell differentiation, whereas the fusion protein MOZ-CBP induces the repression of this protein, thus promoting carcinogenesis (Bug and Ottmann 2010; Florean et al. 2011). Moreover, chromatin modifications induced by alterations in histone acetylation status of specific genes could be due to aberrant HDAC recruitment triggered by leukemia-associated fusion proteins, such as PLZF-RAR (promyelocytic leukemia zinc finger-retinoic acid receptor-α) or PML-RAR (promyelocytic leukemia-retinoic acid receptor-α) in promyelocytic leukemia. Indeed, the RAR is able to bind DNA with HDAC-containing complexes and repress transcription of target genes in the absence of a ligand. The presence of retinoic acid disrupts these inhibitory complexes to the benefit of transcriptional activator complexes, which contain HAT. The fusion-proteins PLZF-RAR and PML-RAR block this dissociation and recruit other enzymes involved in chromatin structure modifications, promoting a drastic inhibition of transcription (Florean et al. 2011; Minucci and Pelicci 2006). This mechanism is responsible for aberrant recruitment of HDACs to specific promoters involved in differentiation and impairs p53 functions, which play an important role in leukemogenesis. In addition to their involvement in the early stages of tumorigenesis and cancer development, aberrant cooperation between HATs and HDACs can influence tumor invasion and metastasis properties (Tai et al. 2007). Moreover, in some cancers, HDAC isoforms are either overexpressed, as with HDAC1, 2, 6, 8 and SIRT1, or underexpressed, as with HDAC4, 5 and SIRT4 (Bradbury et al. 2005). Finally, sirtuins are known to induce angiogenesis, which is critical for tumor growth and metastasis (Liu et al. 2009; Potente et al. 2007). In addition to histone acetylation, other histone modifications such as methylation are involved in chromatin structure and act together with DNA methylation on TSG silencing (Fig. 2; Florean et al. 2011; Hassan et al. 2007; Jones et al. 1998; Wade et al. 1999). Altogether, these mechanisms are implicated in apoptosis resistance, differentiation arrest, cell cycle perturbations leading to aberrant proliferation associated with tumorigenesis and tumor aggressiveness, including invasiveness and metastasis.

Fig. 2
figure 2

Effects of HDAC inhibitors in cancer cells. In healthy cells, promoters of TSGs are unmethylated and present an enrichment of active histone marks such as acetylation and methylation (e.g., H3K4, H3K79, H3R17), which are actively maintained by HATs and HMTs, respectively. Repressive histone marks (H3K9 and H3K27) are maintained unmethylated by specific HDMs. These chromatin marks promote the recruitment of co-activator complexes such as chromatin remodeling complex (e.g., SWI/SNF), transcription factors and RNA polymerase II, which in turn trigger transcription. In cancer cells, TSGs are transcriptionally silenced due to DNMT-mediated DNA hypermethylation, which allows the recruitment of MBDs that bind methyl CpG, which in turn promote the recruitment of repressive complexes (e.g., mSin3A or SMRT/N-CoR) and HDACs leading to histone hypoacetylation. Additionally, active histone methylation marks are removed by specific HDMs and replaced by repressive methylation marks (e.g., H3K9, H3K27, H3R2) by specific HMTs. These modifications promote chromatin condensation and block the recruitment of TF and RNA poly II complex. Consequently, inhibition of HDAC activities is responsible for the modulation of protein degradation, induction of cellular differentiation and apoptosis, and inhibition of telomerase expression and cell cycle progression. Bax Bcl-2-associated X protein, Bcl-2 B-cell lymphoma 2, CDK cyclin-dependent kinase, CDKN CDK inhibitor, DNMT DNA methyl transferase, HAT histone acetyl transferase, HDAC histone deacetylase, HDACi HDAC inhibitor, HIF hypoxia-inducible factor, HDM histone demethylase, HMT histone methyl transferase, hTERT human telomerase reverse transcriptase, MBD methyl binding domain protein, Pol II RNA polymerase II, RUNX runt-related transcription factor, Sin3 switch intensive 3, Smad7 mothers against decapentaplegic homolog 7, SMRT/N-CoR silencing mediator of retinoic acid and thyroid hormone receptors/nuclear hormone receptor co-repressor, SWI/SNF switch/sucrose non-Fermentable, TF transcription factor, TSG tumor suppressor gene

Full size image

Classification of HDAC inhibitors

Epigenetic mechanisms, in contrast to mutations, are potentially reversible and are thus, interesting therapeutic targets. With regard to the involvement of HDACs in cancer, HDAC inhibitors (HDACi) are currently under development for anticancer therapy. Most of these molecules have a binding site for catalytic zinc, a tail (linker) mimicking the side chain of lysine and a “cap” for obstructing the entrance to the active site. These inhibitors can be classified according to their structure into five categories: short-chain fatty acids, hydroxamates, cyclic peptides, benzamides and depsipeptide (Bieliauskas and Pflum 2008; Khan et al. 2008). Nevertheless, many molecules with a potential inhibitory activity against HDAC but with a chemical structure different than the first five have been identified. A significant number of these molecules were isolated from natural sources (Table 2). Among these molecules, pan-HDAC inhibitors and isoform- or class-specific inhibitors have been identified. Interestingly, HDAC inhibitors are used in anti-cancer therapy for their anti-proliferative, pro-apoptotic and anti-inflammatory activities. Moreover, HDACi can restore differentiation processes in cancer cells (Fig. 2).

Table 2 List of natural HDAC modulators
Full size table

HDAC inhibitors of natural and dietary origins

Within the chemical class of short-chain fatty acids, sodium butyrate is present in the gastrointestinal tract as a consequence of microbial fermentation of dietary fiber (Fig. 3). It has been shown that butyrate inhibits HDAC classes I, IIa and IV (Davie 2003). This compound leads to growth arrest, differentiation of leukemic cells and induces apoptosis following the deterioration of the anti-apoptotic protein Bcl-2 (Rosato et al. 2003; Schnekenburger et al. 2006). In addition, another short-chain fatty acid, propionate, was shown to inhibit HDAC activity (Davie 2003). Both compounds induce autophagy with the up-regulation of LAM-2 (lysosomal-associated membrane protein 2) expression and LC3 conversion (Tang et al. 2011). However, considering the low stability of this compound, its use is limited to in vitro studies.

Fig. 3
figure 3

Structure of HDAC modulators

Full size image

The hydroxamic acid trichostatin A (TSA) from the actinomycete Streptomyces hygroscopicus is a potent pan-HDACi. This molecule presents anti-proliferative properties, induces apoptosis through induction of both the pro-apoptotic protein Bax and caspase 3 and reduces telomerase activity by inhibiting hTERT (human telomerase reverse transcriptase) mRNA expression (Woo et al. 2007). Suberoyanilide hydroxamic acid (SAHA) is a synthetic derivative of TSA, which was approved by FDA (Food and Drug Administration) for the treatment of cutaneous T-cell lymphoma in 2006 (Wagner et al. 2010). Interestingly, cells treated by these two related molecules enter in autophagy to protect themself from the stress induced by these drugs (Shao et al. 2004; Zou et al. 2011). Moreover, regarding the involvement of HDAC6 in the formation of aggresomes, tubacin, a synthetic hydroxamic acid that inhibits specifically HDAC6, should be able to impair autophagy (Hideshima et al. 2005). Indeed, Lee et al. 2010 shown that knock-down of HDAC6 induces a defect of fusion of autophagosomes and lysosomes resulting in an accumulation of ubiquitinated proteins.

Natural cyclic peptides have also been demonstrated to act as HDACi. This is the case of FR235222, isolated from Acremonium sp., which induces an accumulation of cells in G1 phase accompanied by an increase in p21 expression and down-regulation of cyclin E in the human promyelocytic leukemia U937 cell line (Petrella et al. 2008).

The depsipeptide FK228 (or Romidepsin), isolated from Chromobacterium violaceum, which preferably inhibits class I HDACs, leads to cell growth inhibition and apoptosis induction in leukemic cells (Furumai et al. 2002; Rosato et al. 2003). Moreover, this compound leads to induction of the autophagy pathway with the involvement of AIF (apoptosis inducible factor) translocation from mitochondria to the nucleus (Watanabe et al. 2009). The FDA approved this compound in 2009 for the treatment of cutaneous T-cell lymphoma (Wagner et al. 2010).

Benzamides represent a chemical class wherein no promising natural compounds have yet been identified. However, some synthetic products belong to this class, such as MS-275 and MGCD0103. Such compounds are interesting due to their selectivity against class I HDACs compared to the pan-HDACi SAHA. These molecules are currently undergoing phase II clinical trials (Khan et al. 2008).

Other chemical classes that demonstrate interesting anticancer properties have also been investigated. Psammaplin A, isolated from the marine sponge Aplysinella rhax, demonstrates inhibitory activities against class I HDACs, accompanied by an increase in histone H3 acetylation (Kim et al. 2007).

Organosulfur compounds isolated from garlic, such as allyl mercaptan (AM) and diallyl disulfide (DADS), represent an interesting class of HDACi. Indeed, AM and DADS induce, following inhibition of HDACs (including HDAC8), an increase in the acetylation of core histones H3 and H4 in the promoter region of the CDKN1A (CDK inhibitor 1A) gene. Increased CDKN1A promoter acetylation induces p21 overexpression, which subsequently leads to cell cycle arrest and apoptosis in cancer cells (Druesne et al. 2004; Link et al. 2010; Nian et al. 2008).

Isothiocyanates such as phenethyl isothiocyanate (PEITC) from cruciferous vegetables (e.g., broccoli) show inhibitory activities against HDACs and induce an increase in histone acetylation, particularly at promoters of genes encoding p21 and Bax, which are involved in the cell cycle and apoptosis, respectively (Wang et al. 2008).

Indole-3-carbinol from broccoli and cauliflower is metabolized to 3,3′-diindolylmethane (DIM) and induces the proteasomal degradation of class I HDACs (1, 2, 3 and 8), thereby inhibiting the anti-apoptotic protein survivin. Moreover, it has been reported that this treatment induces the expression of the genes encoding p21 and p27, which are proteins involved in cell cycle arrest in G1 phase (Bhatnagar et al. 2009; Link et al. 2010).

Regarding class III HDACs, it seems therapeutically interesting to find activators of sirtuins. The activators resveratrol and quercetin are polyphenols isolated from grapes and citrus fruits, respectively. These molecules induce SIRT1 activity, which in turn deacetylates survivin and subsequently triggers apoptosis (Wang et al. 2008). However, it has been reported that inhibition of class III HDACs might favor tumor suppression. For example, isoflavones isolated from soybeans are capable of inhibiting HDACs and activate HATs. These modulations result in TSG activation (CDKN1A, CDKN2A and PTEN) and oncogene inhibition (hTERT; Link et al. 2010). Finally, dihydrocoumarin, a molecule isolated from Melitous officinalis that is widely used as a dietary supplement, is able to reverse heterochromatin-induced epigenetic silencing concomitantly with induction of a senescence- and aging-like phenotype associated with inhibition of SIRT1 activity, leading to an increase in p53 acetylation and apoptosis rate (Olaharski et al. 2005).

Conclusion and perspectives

Alterations in the balance of HAT/HDAC activities resulting in aberrant protein acetylation status are frequently observed in many forms of cancer. In particular, increased deacetylation is involved in epigenetically mediated TSG silencing associated with tumorigenesis. Therefore, HDACs represent a promising class of anticancer drug targets. Regarding HDACi, there are numerous molecules currently undergoing clinical trials, and some are already approved for anticancer therapy. HDACi have multiple biological effects (Fig. 2) that could probably be attributed to the large number of proteins post-transcriptionally modified by lysine acetylation. Consequently, it would be more appropriate to call HATs and HDACs lysine acetyltransferases (KATs) and lysine deacetyltransferases (KDACs), respectively. Furthermore, HDACi, from a chemosensitization point of view, seem particularly attractive when combined with either conventional chemotherapeutic drugs inducing apoptosis (e.g., doxorubicin or etoposide) and/or with other epigenetically active drugs, such as the DNA demethylating agent 5-aza-2′-deoxycytidine, to restore epigenetically mediated gene silencing (Chai et al. 2008; Schnekenburger et al. 2011; Yoon et al. 2010). Hence, it seems that HDACi offer promising alternatives for anticancer therapies. However, researchers are still actively looking for new HDACi from natural origins because they represent an almost inexhaustible source of new chemical structures that could be used for the design of future potent HDACi and the development of new class- or isoform-specific HDACi. Among natural HDACi, dietary compounds are of particular interest because of their potential for chemoprevention purposes. Nevertheless, to fully exploit this aspect, rigorous studies still need to be conducted in order to evaluate the potential as well as the proportion of specific dietary HDACi, compared to other dietary constituents, to reduce the occurrence of cancer. In addition, it is necessary to take in account the possible side effects of a diet enriched with naturally occurring epigenetic modulators. Nonetheless, besides technical challenges, the potential of dietary HDACi and, by extension, epigenetic modulators should be considered for future cancer chemoprevention.