Structure, Mechanism, and Inhibition of the Zinc-Dependent Histone Deacetylases

Abstract

Zinc-dependent histone deacetylases (HDACs) regulate the biological function of histone and non-histone proteins through the hydrolysis of acetyllysine side chains to yield free lysine and acetate. Certain HDAC isozymes exhibit alternative catalytic activities, such as polyamine deacetylase or lysine fatty acid deacylase activity. To date, crystal structures have been reported for class I HDACs (1, 2, 3, and 8), class IIa HDACs (4 and 7), and class IIb HDACs (6 and 10). Conserved active site residues mediate the chemistry of substrate activation and hydrolysis in these isozymes through a metal-activated water molecule assisted by general base-general acid catalysis. Upregulated HDAC activity is observed in cancer and neurodegenerative disease, and four HDAC inhibitors are currently approved for use in cancer chemotherapy. Crystal structures of HDAC-inhibitor complexes guide the design of new inhibitors with high affinity and selectivity for specific HDAC isozymes implicated in human disease.

Graphical Abstract

Introduction

The reversible acetylation of lysine side chains on the surfaces of enzymes and other proteins comprises a vital regulatory strategy in myriad cellular processes such as transcription, the cell cycle, and metabolism [1–4]. At present, the acetylome encompasses nearly 40,000 unique protein acetylation sites as classified in the PhosphoSitePlus database [5]. Accordingly, acetylation rivals phosphorylation as a ubiquitous chemical modification in the regulation of protein function [6].

The chemistry of lysine acetylation requires “writers” – acetyl transferases that utilize acetyl-CoA as a co-substrate [7, 8], and “erasers” – deacetylases that catalyze the hydrolysis of acetyllysine to yield lysine and acetate [9]. With their identification rooted in Allfrey’s pioneering discovery of histone acetylation [10], lysine deacetylases are historically referred to as histone deacetylases (HDACs). However, this name belies the wider function of these enzymes in processing both histone and non-histone protein substrates. Moreover, certain HDACs are not lysine deacetylases, but instead are lysine fatty-acid deacylases [•11, •12]; others process non-protein substrates through their function as polyamine deacetylases [••13].

The HDACs are particularly important as targets for therapeutic intervention in the treatment of cancer, neurodegenerative disease, and other disorders [14–16]. To date, four different inhibitors of metal-dependent HDACs are approved in the US for cancer chemotherapy, and each inhibitor targets the active site Zn²⁺ ion. Vorinostat, Belinostat, and Panobinostat contain hydroxamic acid moieties that chelate Zn²⁺, and Romidepsin is a cyclic depsipeptide that, in the reduced form, bears a pendant thiol group that coordinates to Zn²⁺. These inhibitors are relatively non-selective, i.e., they do not exhibit significant preference for the inhibition of one particular HDAC isozyme or another, so off-target side effects can result from their use [14]. Accordingly, the development of isozyme selectivity in HDAC inhibitors is a priority for current drug design efforts.

Arginase-deacetylase fold

Metal-dependent HDACs adopt an α/β fold first observed in the binuclear manganese metalloenzyme arginase [17] and subsequently observed in the histone deacetylase-like protein from Aquifex aeolicus [18]. This structural and evolutionary relationship was unexpected, since these two enzymes share insignificant amino acid sequence identity. Even so, the Zn²⁺ binding site of the HDACs is conserved as the Mn²⁺_B site of the arginases [19, 20]. Thus, metal binding stoichiometry and selectivity in the arginase-deacetylase family diverged from a common metalloenzyme ancestor.

Recent phylogenetic analysis indicates 12 clades in the arginase-deacetylase family (Figure 1) [••13]. The arginase family [21] also includes agmatinases, ureohydrolases, formiminoglutamases, and even pseudo-arginases that lack metal binding sites, such as the arginase-like protein of unknown function in Trypanosoma brucei [22]. In addition to the HDAC isozymes, the deacetylase family also includes bacterial polyamine deacetylases, acetoin utilization proteins, and pseudo-deacetylases of unknown chemical function [23]. The closest relationship between the arginases and deacetylases depicted in Figure 1 is found between human agmatinase and the pseudo-deacetylase domain of guinea pig HDAC10, which share 19% amino acid sequence identity.

Figure 1: Arginase-deacetylase family of proteins.

Unrooted phylogenetic tree indicates 12 clades: arginases, pseudo-arginases (ΨARG), formiminoglutamases (FIGase) and ureohydrolases, yeast Hos3 homologues, bacterial acetylpolyamine amidohydrolases (APAH), bacterial histone deacetylase-like amidohydrolases (HDAH), class II HDACs, class I HDACs, bacterial acetoin utilization proteins (AcuC), class IV HDACs, uncharacterized protein family UPF0489, and pseudo-deacetylases (ΨDAC). Individual proteins, species, and UniProt accession numbers are listed in Supplementary Table 3 of ref. [••13]. Reprinted from ref. [••13] (Creative Commons Attribution 4.0 International License).

Metal-dependent HDAC isozymes

Since the discovery of the first mammalian HDAC [24], 11 metal-dependent HDACs have been identified and classified through phylogenetic analysis [25]: the class I isozymes HDAC1, HDAC2, HDAC3, and HDAC8; the class IIa isozymes HDAC4, HDAC5, HDAC7, and HDAC9; the class IIb isozymes HDAC6 and HDAC10; and the sole class IV isozyme, HDAC11. Parenthetically, class III HDACs are better known as sirtuins, but these are structurally and mechanistically distinct NAD⁺-dependent enzymes [26, 27]. X-ray crystal structures of metal-dependent HDACs reveal conservation of the α/β arginase-deacetylase fold: HDAC1 [28, ••29], HDAC2 [30], HDAC3 [••31], HDAC4 [32], HDAC6 [••33, ••34], HDAC7 [35], [HDAC8 [36, 37], and HDAC10 [••13]. Crystal structures of HDAC isozymes illuminate critical molecular features of biological function and inhibition.

To date, crystal structures have not yet been reported for HDAC5, HDAC9, or HDAC11. Among these isozymes, HDAC11 is particularly intriguing due to its role as an immunoregulator that interacts with the promoter of the gene encoding interleukin 10 (IL-10) [38]. By blocking the expression of IL-10, HDAC11 activates immune function; loss of HDAC11 leads to immune tolerance, but it is not clear how the catalytic activity of HDAC11 is involved in this process. Possibly, HDAC11 is involved in chromatin remodeling as it associates with the IL-10 promoter. Deletion of HDAC11 or treatment with small molecule inhibitors increases the immunosuppressive function of Foxp3+ T-regulatory cells, suggesting that HDAC11 may be a viable target for immunotherapy [39].

Curiously, however, HDAC11 is not a lysine deacetylase, but instead is a robust lysine fatty acid deacylase: steady-state kinetics indicate optimal catalytic efficiency for the hydrolysis of C₁₀ decanoyllysine, C₁₂ dodecanoyllysine, and C₁₄ myristoyllysine [•11, •12]. Thus, the immunomodulatory function of HDAC11 may be related to its catalytic activity as a fatty acid deacylase. Homology modeling studies suggest that the long fatty acid is accommodated in an internal pocket also shared with HDAC8 [•11], consistent with the weak fatty acid deacylase activity of HDAC8 [40]. The X-ray crystal structure determination of HDAC11 will be required to fully understand these unusual structure-function relationships.

Mechanism of amide bond hydrolysis

A catalytic mechanism for the hydrolysis of acetyllysine by a metal-dependent deacetylase was first proposed with the crystal structure determination of a bacterial histone deacetylase-like protein [18]. Subsequent studies refined this mechanistic proposal, based in large part on structural and enzymological studies of HDAC8 (Figure 2). In the precatalytic Michaelis complex, the scissile carbonyl group of acetyllysine coordinates to Zn²⁺ and accepts a hydrogen bond from Y306. A zinc-carbonyl interaction in the Michaelis complex is consistent with the metal-dependent variation of K_M when Fe²⁺ and Co²⁺ are substituted for Zn²⁺ [41], as well as the crystal structures of intact substrates bound to the Y306F and H143A HDAC8 variants [42, 43]. Y306 may undergo induced-fit conformational changes between “out” and “in” orientations to accommodate substrate binding, as indicated by computational and experimental studies of HDAC8 and related deacetylases [32, 44–47].

Figure 2: Proposed mechanism of acetyllysine hydrolysis catalyzed by HDAC8.

Enzymological measurements indicate that H143 serves as a single general base-general acid, while H142 remains in the positively charged imidazolium form and serves as an electrostatic catalyst. However, hydrogen bond differences for the tandem histidine pair in other HDAC isozymes may allow for dual general base-general acid function. For example, in HDAC6, H573 may serve as a general base, and H574 may serve as a general acid [••33].

Both metal coordination and hydrogen bond interactions are required to fully activate the acetyllysine carbonyl group for nucleophilic attack by a Zn²⁺-bound water molecule. Enzymological studies of wild-type HDAC8 and site-specific variants indicate that H143 serves as a general base to assist Zn²⁺ in activating the nucleophilic water molecule, the pKa of which is lowered by metal coordination; subsequently, H143 serves as a general acid to protonate the leaving amino group, thereby enabling the collapse of the tetrahedral intermediate to yield lysine and acetate [48, 49]. The imidazolium side chain of H142 serves as an electrostatic catalyst, remaining protonated throughout the catalytic cycle. Both H142 and H143 hydrogen bond with the Zn²⁺-bound water molecule, thereby ensuring its optimal orientation for nucleophilic attack. Product acetate may depart through a “backdoor” tunnel in the dissociation of the enzyme-product complex [50]. This tunnel may also accommodate the fatty acid product in the weak fatty acid deacylase activity measured for HDAC8 [40].

Binding interactions of the tetrahedral intermediate and its flanking transition states have been visualized through the study of reactive substrate analogues that, upon binding in active site, undergo a chemical reaction that mimics the first step of catalysis. Specifically, consider the cyclic tetrapeptide Trapoxin A (Figure 3a), the irreversible inhibitor of class I HDACs used in the isolation and identification of human HDAC1 [24]. Trapoxin A contains an amino acid bearing an unusual α,β-epoxyketone amino acid, (2S,9S)-2-amino-8-oxo-9,10-epoxydecanoic acid (L-Aoe). The ketone carbonyl group of the L-Aoe side chain is isosteric with the amide carbonyl group of acetyl-L-lysine, the HDAC substrate. The crystal structure of the HDAC8-Trapoxin A complex shows that the ketone carbonyl undergoes nucleophilic attack to yield a tetrahedral gem-diolate that mimics the tetrahedral intermediate in catalysis [••51]. The presumed oxyanion of the gem-diolate coordinates to Zn²⁺ and accepts a hydrogen bond from Y306, and the hydroxyl group of the gem-diolate forms hydrogen bonds with H142 and H143 (Figure 3a).

Figure 3: HDAC complexes with cyclic tetrapeptide inhibitors.

(a) Structure of the HDAC8-Trapoxin A complex determined at 1.24 Å resolution (PDB 5VI6). The α,β-epoxyketone side chain of L-Aoe binds as a gem-diolate, thereby mimicking the tetrahedral intermediate and its flanking transition states in catalysis. The flanking NH groups of L-Aoe donate hydrogen bonds to D101, a key determinant of isozyme-substrate recognition. Reproduced from ref [••51]. Copyright 2017 American Chemical Society. (b) Structure of the HDAC6-HC Toxin complex determined at 1.73 Å resolution (PDB 5EFJ). Here, too, the α,β-epoxyketone side chain binds as a gem-diolate. The backbone NH group of L-Aoe donates a hydrogen bond to S531, a key determinant of isozyme-substrate recognition. Reprinted from ref [••33].

HC Toxin is a cyclic tetrapeptide inhibitor similarly containing an L-Aoe residue like that of Trapoxin A, and the crystal structure of the HDAC6-HC Toxin complex similarly reveals the binding of the gem-diolate form of the α,β-epoxyketone side chain (Figure 3b) [••33]. Intriguingly, cocrystallization of H574A HDAC6 with an acetyllysine-containing substrate yields an electron density map showing tetrahedral instead of planar electron density for the scissile amide moiety, suggesting that the actual tetrahedral intermediate is trapped in the active site [••33]. Since the tetrahedral intermediate cannot collapse without a proton donor to the leaving amino group, the loss of general acid H574 results in the dead-end formation of the tetrahedral intermediate in much the same way as the tetrahedral gem-diolate of HC Toxin is formed and stabilized in the HDAC6 active site. Crystal structures of HDAC4 [32], HDAC6 [••33], and HDAC10 [••13] complexed with α-trifluoromethylketones similarly reveal the binding of the gem-diolate forms of these inhibitors in deacetylase active sites.

As indicated in Figure 2, key catalytic residues in the active sites of most HDACs are the tandem histidine dyad, a catalytic tyrosine, and the Zn²⁺ ion. However, in class IIa HDACs (4, 5, 7, and 9), the catalytic tyrosine is substituted with a histidine, and deacetylase activity is significantly compromised as a result [52]. Even so, robust lysine deacetylase activity of HDAC4 and HDAC7 can be restored by histidine-to-tyrosine substitutions [32, 35, 52]. The biological significance of this activity difference is unclear. The active sites of most zinc hydrolases contain a hydrogen bond donor in addition to the Zn²⁺ ion that helps polarize the scissile amide carbonyl, as first noted for the prototypical zinc hydrolases thermolysin and carboxypeptidase A [53, 54]. Possibly, the additional histidine in the active sites of class IIa HDACs confers an alternative substrate specificity that is yet to be discovered [52].

Substrate specificity

In cells, HDAC-substrate pairs can be identified using Stable Isotope Labeling of Amino Acids in Cell Culture (SILAC) proteomics coupled with the administration of a selective HDAC inhibitor [55], or the application of a genetically-encoded photocrosslinker by unnatural amino acid expression [56]. Acetylomics approaches have also been used to assess the biological effects of isozyme-selective HDAC inhibitors [57].

For enzymological studies, peptide-based assays have been developed to measure HDAC activity, allowing for analysis of substrate sequence-activity relationships. One such assay utilizes peptides of varying composition and length bearing a C-terminal acetyl-L-lysine (Aly) followed by an aminomethylcoumarin (AMC) moiety [58]. This assay exploits the fluorescence shift of the AMC moiety upon liberation from the peptide by trypsin, which only occurs after the Aly residue has been deacetylated. However, this assay only allows for variation of the peptide sequence on the amino side of the Aly residue, i.e., in the −1 position and beyond. Alternatively, peptide substrates bearing an internal Aly residue can be incubated with HDAC; after quenching, the newly liberated lysine side chain in the peptide product can be covalently modified with dansyl chloride, and the peptide product then identified and quantified using liquid chromatography-mass spectrometry [••33].

X-ray crystal structures of HDAC isozymes complexed with substrates or substrate analogues reveal the molecular basis of substrate recognition. As lysine deacetylases, HDAC6 and HDAC8 catalyze the hydrolysis of acetyllysine residues contained in polypeptide and protein substrates. The active site of each enzyme consists of a concave molecular surface surrounding a narrow cleft that accommodates the scissile acetyllysine side chain [••33, 36, 37]. Crystal structures of H143A HDAC8 and Y306F HDAC8 complexed with the p53-derived assay substrate Ac-Arg-His-Aly-Aly-AMC show that the backbone NH groups of the scissile acetyllysine residue and the AMC moiety at the +1 position donate hydrogen bonds to the carboxylate side chain of D101 (Figure 4a) [42, 43]. Corresponding hydrogen bonds are observed between HDAC8 and the NH groups flanking the α,β-epoxyketone residue of Trapoxin A (Figure 3a) [••51]. Dual hydrogen bond interactions with D101 suggest that a proline residue at the +1 position of the substrate cannot be accommodated by HDAC8, and this has been confirmed experimentally [••33]. No other hydrogen bonds are observed between the concave active site surface of HDAC8 and main chain atoms for residues in the −3 to −1 positions of the bound substrate; the guanidinium group of the arginine residue at the −3 position forms hydrogen bonds with the backbone carbonyl of I94 or the carboxylate group of E148, or makes no hydrogen bonds at all in several crystal structures [42, 43]. The relative dearth of enzyme-substrate hydrogen bond specificity ensures broad specificity for flanking amino acid sequences in protein substrates. Although no structural information is available with regard to peptide substrate binding at the +2 position or beyond, there is enzymological evidence for an exosite that recognizes the basic sequence KRHR at the +4 to +7 positions of a substrate derived from the N-terminal tail of the histone H4 protein [59].

Figure 4: Substrate binding to HDACs.

(a) Superposition of substrate complexes with Y306F HDAC8 (blue, PDB 2V5W) and H143A HDAC8 (grey, PDB 3EWF) reveals that the scissile carbonyl of acetyllysine is activated by Zn²⁺ coordination and a hydrogen bond with Y306. The flanking NH groups of the acetyllysine substrate donate hydrogen bonds to D101. (b) Superposition of substrate complexes with Y745F HDAC6 (light green, PDB 5EFK) and H574A HDAC6 (dark green, PDB 5EFN) similarly reveals activation of the scissile substrate carbonyl by Zn²⁺ coordination and a hydrogen bond with Y745. Moreover, nucleophilic attack of Zn²⁺-bound water at the substrate carbonyl group occurs in H574A HDAC6 to result in a tetrahedral intermediate. The backbone NH group of acetyllysine donates a hydrogen bond to S531. (c) The trifluoromethylketone analogue of N⁸-acetylspermidine binds as a tetrahedral gem-diolate that mimics the tetrahedral intermediate in the active site of HDAC10 (PDB 5TD7). A 3₁₀ helix unique to this isozyme (magenta) sterically constricts the active site and confers specificity for acetylpolyamine substrates.

In contrast with HDAC8, crystal structures of HDAC6 complexed with two different assay substrates, the α-tubulin-derived peptide Ac-Ser-Asp-Aly-AMC and the histone H4-derived peptide Ac-Arg-Gly-Aly-AMC, show that only a single hydrogen bond is made between the backbone NH group of the scissile acetyllysine residue and S531 (Figure 4b); the corresponding hydrogen bond is similarly observed between S531 and the NH group of the α,β-epoxyketone residue of HC Toxin [••33]. Accordingly, HDAC6 substrates can accommodate proline residues at the +1 position. In the HDAC6-substrate complexes, peptide segments preceding the scissile acetyllysine residues are disordered and hence not fully modeled in the refined crystal structures [••33]. However, the backbone carbonyl of the aspartate or glycine residue at the −1 positions of these substrates hydrogen bonds with N530. Here, too, minimal hydrogen bonding with main chain or side chain atoms ensures broad specificity for flanking amino acid sequences in protein substrates.

In contrast with other HDACs of known structure, the crystal structure of HDAC10 reveals an active site cleft sterically constricted by a 3₁₀-helix containing the highly conserved P(E,A)CE motif (Figure 4c) [••13, 60]. Consequently, peptide substrates cannot bind easily in the active site of HDAC10 due to the long, narrow active site cleft. Instead, the active site of HDAC10 readily accommodates long, slender polyamine substrates, exhibiting optimal catalytic efficiency for the deacetylation of N⁸-acetylspermidine. A conserved gatekeeper in the HDAC10 active site, E274, ensures electrostatic recognition of the positively charged secondary amino group of N⁸-acetylspermidine. Conserved glutamate and aspartate residues across the protein surface additionally contribute to the molecular recognition of the cationic substrate [••13]. HDAC10 appears to be the long-sought cytosolic N⁸-acetylspermidine deacetylase for which cellular activity was discovered more than 40 years ago [61, 62].

Regulation by corepressors

Several metal-dependent HDACs function in multi-protein regulatory complexes. For example, crystal structures of HDAC1 and HDAC3 complexed with corepressor proteins that activate catalysis reveal significant protein-protein interactions that sandwich inositol tetraphosphate at the interface [28, ••29, ••31]. Corepressor binding buttresses the L1 loop flanking the deacetylase active site, which reduces enzyme surface dynamics [••31]. Additionally, superposition of the structures of HDAC1 and HDAC3-corepressor complexes with the structure of an HDAC6-inhibitor complex (which lacks a corepressor) suggests that corepressor binding may slightly constrict the active site [•63]. Inositol tetraphosphate interacts with a conserved arginine side chain on the L6 loop, which forms one wall of the active site [••29]. Mutation of this arginine in the HDAC3-SMRT corepressor complex results in a dramatic loss of enzyme activity [28], suggesting that this interaction is vital for activation of these class I HDACs.

Molecular dynamics simulations of HDAC3 reveal conformational mobility for the catalytic tyrosine in the active site, which adopts the “out” conformation in the absence of corepressor and the catalytically-required “in” conformation upon the binding of corepressor [45]. Similar conformational mobility has been observed for this residue in molecular dynamics studies of HDAC8 [46, 47] and in X-ray crystallographic studies of a bacterial polyamine deacetylase [44]. Conformational mobility of the catalytic tyrosine in HDAC active sites is facilitated by its placement in a conserved glycine-rich loop [47]. Conformational control of the catalytic tyrosine in the class I HDAC active site by deacetylase-activating domains of corepressor proteins represents a simple molecular strategy for the regulation of deacetylase activity in multi-protein complexes.

Notably, HDAC8 is the only class I enzyme that does not require a corepressor for optimal activity due to structural differences in the L1 and L6 loops that allow the enzyme to adopt an active conformation without steric persuasion from a corepressor protein. Even so, active site loops in HDAC8 can adopt alternative conformations as they accommodate derivatives of the cyclic tetrapeptide inhibitor Largazole [64]; the flexibility of these loops may contribute to the significant entropic cost of inhibitor binding [65].

Inhibitor binding and isozyme selectivity

Given that HDACs are validated targets for therapeutic intervention in the treatment of cancer and other diseases, X-ray crystal structures provide valuable guidance for drug design. A key feature of an HDAC inhibitor is the functional group that coordinates to the active site Zn²⁺ ion. The hydroxamate group is most commonly utilized and is found in three out of four of the currently-approved HDAC-targeted drugs (Figure 5a) [14–16]. However, the hydroxamate group can be mutagenic due to degradation via the Lossen rearrangement, which generates a highly reactive isocyanate capable of alkylating cellular components (Figure 5b) [66]. This chemical reactivity can limit the utility of hydroxamate-containing drugs.

Figure 5: HDAC inhibitors.

(a) Hydroxamate-based HDAC inhibitors currently approved for cancer chemotherapy. (b) The hydroxamate moiety is susceptible to degradation through the Lossen rearrangement, which generates an electrophilic isocyanate. (c) Romidepsin is currently approved for cancer chemotherapy and is activated by reduction of the disulfide linkage. The marine natural product Largazole is structurally similar to Romidepsin and is activated by thioester hydrolysis. (d) The crystal structure of the HDAC8-Largazole complex shows that the ionized thiolate side chain coordinates to Zn²⁺ and accepts a hydrogen bond from Y306 (PDB 3RQD).

Macrocyclic peptides and depsipeptides comprise a fascinating class of HDAC inhibitors. These natural products contain pendant side chains that mimic the scissile acetyllysine side chain of an HDAC substrate. Romidepsin is a depsipeptide drug that, upon reduction of an intramolecular disulfide linkage, yields a side chain bearing a thiol group that coordinates to the active site Zn²⁺ ion. While the crystal structure of an HDAC-Romidepsin complex has not yet been reported, the crystal structures of HDAC8 complexed with Largazole and its synthetic analogues, which are structurally similar to Romidepsin, have been reported (Figure 5c,d) [64, 67].

As previously discussed, the macrocyclic tetrapeptide inhibitors HC Toxin and Trapoxin A have been studied in complexes with HDAC6 and HDAC8, respectively (Figure 3) [••33, ••51]. Hydrogen bond interactions between the NH groups flanking the L-Aoe residue are identical to those observed in structures of enzyme-substrate complexes [••33, 43] and highlight differences that contribute to inhibitor selectivity for HDAC6 over class I HDACs such as HDAC8. Specifically, D101 of HDAC8 accepts hydrogen bonds from both NH groups flanking the L-Aoe residue [••51], whereas S531 of HDAC6 accepts a hydrogen bond only from the NH group of LAoe (Figure 3) [••33]. Crystal structures of selective inhibitors bound to HDAC6 similarly reveal direct or water-mediated hydrogen bond interactions with S531, so this residue is a key determinant of isozyme selectivity [•63, 68, •69]. Additionally, calorimetric measurements indicate that inhibitor binding to HDAC6 is entropically favorable, in contrast with inhibitor binding to HDAC8 [65].

Other functional groups targeting Zn²⁺ coordination in HDAC active sites include α-trifluoromethylketones, which bind as gem-diolates in the active sites of HDAC4 [32], HDAC6 [••33], and HDAC10 (Figure 4c) [••13], and benzamides that chelate the Zn²⁺ ion of HDAC2 through the anilide nitrogen and amide carbonyl groups [30]. Most recently, structural studies of HDAC-mercaptoacetamide complexes reveal functional mimicry of hydroxamate binding to HDAC6 [70] but not HDAC8 [71]. Clearly, opportunities abound for new HDAC inhibitor designs incorporating alternative metal-binding groups, such as those exemplified in a recent review on bioinorganic fragment-based drug discovery [•72].

Concluding remarks

Recent years have witnessed rapidly increasing interest in the structural and chemical biology of the metal-dependent HDAC isozymes, particularly in view of their value as therapeutic targets for cancer, neurodegenerative disorders, and other diseases. At present, only the structures of HDAC5, HDAC9, and HDAC11 have yet to be determined. Chemical and biological studies reveal that HDACs catalyze the deacetylation of histone proteins and more, both in the nucleus and the cytoplasm of eukaryotic cells. Moreover, HDAC11 is a deacylase and not a deacetylase, with specificity for substrates bearing fatty acid acyl-lysine residues [•11, •12]. The preferred substrate of HDAC10 in vitro is not acetyllysine, but instead N^8-acetylspermidine [••13]. Thus, metal-dependent HDACs exhibit an impressively broad range of specificity for small molecule and protein substrates in different cellular compartments.

Crystal structures of HDAC-inhibitor complexes reveal the molecular features underlying substrate specificity and inhibitor selectivity. Isozyme selectivity is desirable to minimize side effects resulting from the inhibition of off-target HDAC isozymes. For example, recent structural studies of HDAC6 show that the most selective inhibitors contain capping groups that make favorable binding interactions in an active site pocket formed by the L1 loop, as well as variable Zn²⁺ coordination denticity [•63, •69]. Moreover, selective inhibitors can be used for imaging HDAC in the human brain using positron emission tomography, which exemplifies the forefront of HDAC studies [•73]. Clearly, HDAC structural and chemical biology will continue to provide exciting new opportunities at the interface of chemistry, biology, and medicine for many years to come.

Highlights.

• Zn²⁺-histone deacetylases (HDACs) belong to the arginase-deacetylase superfamily

• HDACs hydrolyze acetyllysine, fatty acid acyllysine, and acetylpolyamine substrates

• HDACs are drug targets for the treatment of cancer and neurodegenerative disease

• Crystal structures of HDACs guide current drug design efforts