X-ray Crystallographic Snapshots of Substrate Binding in the Active Site of Histone Deacetylase 10

Abstract

Histone deacetylase 10 (HDAC10) is a zinc-dependent polyamine deacetylase enriched in the cytosol of eukaryotic cells. The active site of HDAC10 contains catalytic residues conserved in other HDAC isozymes that function as lysine deacetylases: Y307 assists the zinc ion in polarizing the substrate carbonyl for nucleophilic attack, and the H136-H137 dyad serves general base-acid functions. As an inducer of autophagy, HDAC10 is an attractive target for the design of selective inhibitors that may be useful in cancer chemotherapy. Since detailed structural information regarding the catalytic mechanism of HDAC10 may inform new approaches to inhibitor design, we now report X-ray crystal structures of HDAC10 in which reaction intermediates with substrates N⁸-acetylspermidine and N-acetylputrescine are trapped in the active site. The Y307F substitution prevents activation of the substrate carbonyl for nucleophilic attack by the zinc-bound water molecule, thereby enabling crystallographic isolation of intact enzyme-substrate complexes. The H137A substitution removes the catalytically obligatory general acid, thereby enabling crystallographic isolation of oxyanionic tetrahedral intermediates. Finally, the acetate complex with the wild-type enzyme represents a product complex after dissociation of the polyamine coproduct. Taken together, these structures provide snapshots of the reaction coordinate of acetylpolyamine hydrolysis and are consistent with a mechanism in which tandem histidine residues H136 and H137 serve as general base and general acid catalysts, respectively. The function of the histidine dyad in the HDAC10 mechanism appears to be similar to that in HDAC6, but not HDAC8 in which both functions are served by the second histidine of the tandem pair.

Graphical Abstract

INTRODUCTION

Cytosolic polyamine deacetylase activity was first observed more than 40 years ago in rat liver extracts^1,2 and later found to exhibit specificity for the hydrolysis of N⁸-acetylspermidine but not N¹-acetylspermidine.³ Although the identity of this eukaryotic polyamine deacetylase (PDAC) was not immediately determined, it was clearly distinct from the lysine deacetylases that catalyze histone deacetylation (HDACs). In particular, selective inhibition of PDAC activity increased N⁸-acetylspermidine levels but did not influence histone acetylation levels.⁴ Lombardi and colleagues suggested that one of the class IIb HDACs, either HDAC6 or HDAC10, might be responsible for eukaryotic PDAC activity based on their sequence similarity with a structurally characterized bacterial PDAC as well as their localization in the cell cytosol.⁵ Most recently, Hai and colleagues established that HDAC10 enzymes from Danio rerio (zebrafish) and Homo sapiens (human) are indeed PDACs and not lysine deacetylases.⁶

The crystal structure of HDAC10 from D. rerio⁶ reveals two domains (Figure 1a), each of which adopts the α/β arginase-deacetylase fold, that assemble with an overall architecture similar to that of HDAC6.^7,8 However, while both domains of HDAC6 are catalytically active, only the PDAC domain of HDAC10 is catalytically active. The PDAC domain contains key catalytic elements: tandem histidine residues H136 and H137 serve general base-general acid functions, and Y307 assists the Zn²⁺ ion in polarizing the substrate carbonyl for nucleophilic attack by a Zn²⁺-bound water molecule. The PDAC domain of HDAC10 additionally contains two unique structural features conserved among HDAC10 enzymes from different species: a 3₁₀ helix containing the consensus sequence P²³(E,A)CE²⁶ (the “PEACE” motif) that sterically constricts the active site, and a gatekeeper residue (E274) that provides a complementary electrostatic environment for the cationic substrate (Figure 1a).⁶ These structural features favor the binding of long, slender polyamines such as N⁸-acetylspermidine (Figure 1a,b). The catalytic mechanism proposed for N⁸-acetylspermidine hydrolysis is summarized in Figure 1c.

Figure 1.

(a) The structure of HDAC10 consists of a polyamine deacetylase domain (PDAC, gray) and a catalytically inactive pseudodeacetylase domain (ΨDAC, pale green). The substrate N⁸-acetylspermidine is bound in the PDAC active site (CPK model, C = orange, N = blue, O = red). The catalytic Zn²⁺ ion is a black sphere. The “PEACE” motif 3₁₀-helix (P²³(E,A)CE²⁶, magenta) sterically constricts the active site, and E274 (red ball-and-stick figure) provides electrostatic complementarity to cationic polyamine substrates. (b) Acetylpolyamine substrates of wild-type human and zebrafish HDAC10 enzymes.⁶ (c) Proposed mechanism of N⁸-acetylspermidine hydrolysis catalyzed by HDAC10 based on the structural studies reported herein.

The unique structural features of the HDAC10 active site can be exploited in the design and development of selective inhibitors. HDAC10 inhibition is emerging as a strategy for cancer chemotherapy because inhibition blocks autophagy, thereby increasing the efficacy of cytotoxic drugs.⁹ The selectivity of inhibitor binding to HDAC10 over other HDAC isozymes can be enhanced by targeting interactions with the unique 3₁₀ helix and E274 in the HDAC10 active site.^10,11 For example, consider Tubastatin A,¹² originally designed as a selective inhibitor of HDAC6 but subsequently discovered to be more selective for inhibition of HDAC10.¹⁰ The 3₁₀ helix clamps the inhibitor in place while E274 makes an electrostatic interaction with the positively charged amino group of the inhibitor capping group.¹³ Tubastatin A analogues that optimize these interactions exhibit even better HDAC10 selectivity.^10,13

To further inform the design of HDAC10 inhibitors, detailed structural chemistry underlying the catalytic mechanism of acetylpolyamine hydrolysis is desirable. Here, we employ X-ray crystallography to study the binding of intact substrates, tetrahedral intermediates, and product acetate in the active sites of wild-type and variant forms of “humanized” D. rerio (zebrafish) HDAC10. Specifically, we employ variants of zebrafish HDAC10 in which the A24E and D94A substitutions make the active site more similar to that of human HDAC10 (the catalytic domains of zebrafish and human HDAC10 are related by 63%/79% sequence identity/similarity⁶).¹³ Both Zn²⁺ and Y307F are required to polarize the amide carbonyl for nucleophilic attack, so the loss of the hydroxyl group in the Y307F variant enables cocrystallization of precatalytic Michaelis complexes. The loss of the catalytic general acid in the H137A variant enables cocrystallization with substrates that react to form tetrahedral intermediates that cannot collapse in the absence of a proton donor. Taken together, these structures provide unprecedented snapshots of catalysis along the reaction coordinate of acetylpolyamine hydrolysis from start to finish, highlighting key interactions responsible for substrate specificity and transition state stabilization.

MATERIALS AND METHODS

Crystallization of HDAC10–substrate complexes.

“Humanized” HDAC10 (HDAC10) was prepared and purified as previously described.¹³ Individual solutions containing 10 mg/mL wild-type HDAC10, H137A HDAC10, or Y307F HDAC10 in buffer [50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; pH 7.5), 300 mM KCl, 5% glycerol, and 1 mM tris(2-carboxyethyl)phosphine (TCEP)] were allowed to incubate with 10 mM substrate on ice for 1 h. Trypsin was added to each of the HDAC10–substrate solutions at a 1:1000 (trypsin:HDAC10) molar ratio and allowed to digest for 1 h at room temperature. Each HDAC10–substrate solution was filtered using 0.22 μm centrifuge filters prior to crystallization.

Each HDAC10–substrate complex was crystallized by the sitting drop vapor diffusion method at 4 °C. Briefly, a 100 nL drop of the HDAC10–substrate solution was dispensed into a 100 nL drop of precipitant buffer on a 96-well crystallization plate using a Mosquito crystallization robot (TTP Labtech). The drop was equilibrated against 80 μL of precipitant buffer in the well reservoir and crystals formed in approximately one day. Microseed crystals of the HDAC10–Tubastatin A complex¹³ were added to every crystallization drop prior to sealing the 96-well tray. The general approach to crystallization involved screening conditions similar to those employed for crystallization of the HDAC10–Tubastatin A complex,¹³ with systematic variation of phosphate monobasic/dibasic salt concentrations, counterions (Na⁺ or K⁺), and additives contained in the Hampton Additive Screen HT.

Crystallization of the Y307F HDAC10–N⁸-acetylspermidine complex was achieved using 0.125 M Na₂HPO₄, 0.075 M NaH₂PO₄, and 20% (w/v) PEG3350 as the precipitant buffer. Crystallization of the Y307F HDAC10–N-acetylputrescine complex was achieved using 0.100 M NaH₂PO₄, 0.100 M Na₂HPO₄, and 20% (w/v) PEG3350 as the precipitant buffer. Crystallization of the H137A HDAC10–N⁸-acetylspermidine complex was achieved using 0.113 M NaH₂PO₄, 0.068 M Na₂HPO₄, 18% (w/v) PEG3350, and 3% (w/v) 1,5-diaminopentane dihydrochloride as the precipitant buffer. Crystallization of the H137A HDAC10–N-acetylputrescine complex was achieved using 0.113 M NaH₂PO₄, 0.068 M Na₂HPO₄, 18% (w/v) PEG3350, and 4% (v/v) polypropylene glycol P400 as the precipitant buffer. Crystallization of the HDAC10–acetate complex was achieved using 0.100 M KH₂PO₄, 0.100 M K₂HPO₄, and 20% (w/v) PEG3350 as the precipitant buffer.

X-ray diffraction data for all HDAC10 complexes reported here were collected on NE-CAT beamline 24-ID-C at the Advanced Photon Source of Argonne National Laboratory (Argonne, Il). Diffraction data were integrated and using iMosflm and scaled with Aimless in the CCP4 program suite.^14–16 All of the initial electron density maps were phased by molecular replacement using the Y307F HDAC10–trifluoroketone inhibitor complex (PDB 5TD7)⁶ using Phaser.^17,18 Model building was performed using COOT¹⁹ and refinement was achieved using PHENIX²⁰ in an iterative process that yielded each final model. For each complex, the substrate was built into electron density during the final stages of refinement. Each final model lacks certain segments of the polypeptide chain due to poor or missing electron density. Some of these sections of the polypeptide chain are located in the pseudodeacetylase domain, which is characterized as more dynamic than the polyamine deacetylase (PDAC) domain as evidenced by higher average B factors (Table S1). Additionally, a portion of the linker region between the PDAC and the pseudodeacetylase domain is missing from the model which is likely due to the trypsin digest required for successful crystallization. MolProbity²¹ was used to assess the quality of the final model. All data collection and refinement statistics are listed in Table 1.

Table 1.

Data collection and refinement statistics

HDAC10 Complex	Y307F HDAC10–N8-AcSpd	Y307F HDAC10–N-AcPut	H137A HDAC10–N8-AcSpd	H137A HDAC10–N-AcPut	HDAC10-Acetate
Space group	P3121	P3121	P3121	P3121	P3121
a,b,c (Å)	80.5, 80.5, 249.4	80.6, 80.6, 249.7	80.5, 80.5, 245.3	80.3, 80.3, 246.0	80.5, 80.5, 245.5
Rmergeb	0.217 (1.136)	0.112 (1.026)	0.113 (1.140)	0.125 (1.317)	0.138 (1.326)
Rpimc	0.084 (0.437)	0.039 (0.356)	0.039 (0.393)	0.042 (0.474)	0.048 (0.469)
CC1/2d	0.985 (0.617)	0.998 (0.684)	0.997 (0.604)	0.997 (0.656)	0.997 (0.575)
Redundancy	7.2 (7.4)	9.1 (9.2)	9.4 (9.1)	9.5 (8.5)	9.1 (8.9)
Completeness (%)	99.8 (99.9)	100 (100)	99.4 (99.5)	100 (100)	100 (100)
I/σ	5.5 (2.1)	11.8 (2.2)	10.5 (1.9)	11.1 (2.1)	11.0 (1.9)
Refinement
Resolution (Å)	69.72–2.10 (2.16–2.10)	69.77–2.10 (2.16–2.10)	69.71–2.00 (2.05–2.00)	69.50–2.05 (2.11–2.05)	69.70–2.15 (2.21–2.15)
No. reflections	55699 (5526)	55068 (5481)	62900 (6197)	58731 (5784)	51222 (5030)
Rwork/Rfreee	0.202/0.244 (0.292/0.334)	0.191/0.217 (0.278/0.315)	0.175/0.212 (0.252/0.294)	0.176/0.218 (0.249/0.314)	0.185/0.218 (0.270/0.310)
Number of Atoms
Protein	4858	4867	4930	4888	4878
Ligand	32	35	55	76	30
Solvent	203	256	322	329	280
Average B factor (Å2)
Protein	39	42	40	39	42
Ligand	41	46	48	48	43
Solvent	38	43	45	44	43
Root-Mean-Square Deviation
Bond lengths (Å)	0.008	0.007	0.007	0.007	0.007
Bond angles (°)	0.9	0.9	0.8	0.9	0.9
Ramachandran Plot (%)
Favored	95.58	97.28	96.82	96.38	96.50
Allowed	4.42	2.56	3.03	3.62	3.34
Outliers	0.00	0.16	0.16	0.00	0.16
PDB Entry	7KUQ	7KUR	7KUS	7KUT	7KUV

^aValues in parentheses refer to the highest-resolution shell of data.

^bR_merge = ∑_h∑_i|I_i,_h − ⟨I⟩_h|/∑_h∑_iI_i,h, where ⟨I⟩_h is the average intensity calculated for reflection h from i replicate measurements.

^cR_p.i.m. = (∑_h(1/(N-1))^1/2∑_i|I_i,h −⟨I⟩_h|)/∑_h∑_i I_i,_h, where N is the number of reflections and ⟨I⟩_h is the average intensity calculated for reflection h from replicate measurements.

^dPearson correlation coefficient between random half-datasets.

^eR_work = ∑||F_o| − |F_c||/∑|F_o| for reflections contained in the working set. |F_o| and |F_c| are the observed and calculated structure factor amplitudes, respectively. R_free is calculated using the same expression for reflections contained in the test set held aside during refinement.

^fPer asymmetric unit.

^gCalculated with MolProbity.

RESULTS

The “humanized” form of D. rerio HDAC10 is superior to the wild-type enzyme for cocrystallization with substrates and inhibitors, generally yielding higher quality crystals that diffract to higher resolution. As previously reported for wild-type zebrafish HDAC10, trypsin was added to protein solutions to facilitate crystallization. Proteolysis by trypsin likely accounts for approximately 30 residues missing from the linker between the deacetylase and pseudodeacetylase domains. Below, we present structures of this HDAC10 construct complexed with the intact substrates N⁸-acetylspermidine and N-acetylputrescine, the tetrahedral intermediates resulting from nucleophilic attack of a Zn²⁺-bound water molecule at the scissile carbonyl of these substrates, and the co-product acetate.

Substrate complexes.

The binding modes of N⁸-acetylspermidine and N-acetylputrescine are essentially identical in the active site of Y307F HDAC10 as observed in the crystal structures of complexes determined at 2.10 Å resolution (Figure 2, Figure S1). The catalytic Zn²⁺ ion is coordinated by protein residues D174, H176, and D267, the substrate carbonyl group, and a water molecule with distorted square pyramidal geometry (H176 is at the apical position). The O---Zn²⁺ distances for the substrate carbonyl oxygen and water molecule are 2.3 Å and 2.2 Å for the N⁸-acetylspermidine complex, and 2.2 Å and 2.4 Å for the N-acetylputrescine complex, respectively. The Zn²⁺-bound water molecule forms hydrogen bonds with H136 and H137. Ordinarily, both Zn²⁺ coordination and a hydrogen bond with Y307 are required to activate the amide carbonyl group for nucleophilic attack by the Zn²⁺-bound water molecule. Unable to achieve the latter interaction owing to the Y307F substitution, the Zn²⁺-bound substrate is insufficiently activated for reaction and is therefore trapped as the precatalytic Michaelis complex. Since the position of the Zn²⁺-bound water molecule lies outside of the Bürgi-Dunitz trajectory²² of the scissile carbonyl group, a slight conformational adjustment is likely necessary to enable an ideal approach for O–C bond formation. Such an adjustment might be reflected by the shape of the electron density for the scissile carbonyl group, which is somewhat bulbous and may accordingly allow for slight rotation of the acetyl group (Figure 2).

Figure 2. Enzyme-substrate complexes.

(a) Stereoview of Polder omit maps of N⁸-acetylspermidine (black mesh, contoured at 5.5σ) and the Zn²⁺-bound water molecule (green mesh, contoured at 8.0σ) in the active site of Y307F HDAC10. Atoms are color-coded as follows: C = white (HDAC10) or orange (substrate), N = blue, O = red, S = yellow, Zn²⁺ = large gray sphere, water molecules = small red spheres. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines. (b) Stereoview of Polder omit maps of N-acetylputrescine (black mesh, contoured at 7.5σ) and the Zn²⁺-bound water molecule (green mesh, contoured at 9.5σ) in the active site of Y307F HDAC10. Atoms are color-coded as in (a). Alternative orientations of these maps can be found in Figure S1.

Additional residues in the HDAC10 active site govern specificity for the binding of polyamine substrates. As previously mentioned, the PEACE motif helix is unique to HDAC10 and sterically constricts the active site, thereby conferring specificity for the binding of slender polyamine substrates. Additionally, the amide NH groups of N⁸-acetylspermidine and N-acetylputrescine each donate a hydrogen bond (within experimental error) to the backbone carbonyl of G145. This interaction contributes to the molecular recognition of the substrate amide group and helps to orient it for nucleophilic attack.

The primary ammonium cations of both N-acetylputrescine and N⁸-acetylspermidine each donate a hydrogen bond to E24. Interestingly, despite the different lengths of these two polyamine substrates, the side chain of E24 is sufficiently flexible to undergo a conformational change to form this hydrogen bonded salt link. This residue is unique to human HDAC10 and was introduced through the A24E substitution to “humanize” the active site of D. rerio HDAC10. Interestingly, gatekeeper residue E274 does not form a direct hydrogen bond with either substrate. Instead, the secondary ammonium cation of N⁸-acetylspermidine and the primary ammonium cation of N-acetylputrescine interact with E274 across two bridging hydrogen bonded water molecules (Figure 2, Figure S1).

Tetrahedral intermediate complexes.

Crystal structures determined at 2.00 Å and 2.05 Å resolution, respectively, reveal essentially identical binding modes for N⁸-acetylspermidine and N-acetylputrescine in the active site of H137A HDAC10: the scissile amide carbonyl group of each substrate has undergone nucleophilic attack to yield a tetrahedral adduct with its oxyanion coordinating to the catalytic Zn²⁺ ion (O---Zn²⁺ distances = 2.0 Å and 2.1 Å, respectively) (Figure 3, Figure S2). The Zn²⁺ ion is tetracoordinate with distorted tetrahedral geometry. The former Zn²⁺-bound water molecule becomes the hydroxyl group of the tetrahedral intermediate, moving away from inner-sphere metal coordination with Zn²⁺–O distances of 2.7 Å and 2.6 Å, respectively, in complexes with N⁸-acetylspermidine and N-acetylputrescine. The hydroxyl group retains the hydrogen bond with H136 as observed for the native Zn²⁺-bound water molecule (Figure 2).

Figure 3. Enzyme-tetrahedral intermediate complexes.

(a) Stereoview of a Polder omit map of the tetrahedral intermediate of N⁸-acetylspermidine (black mesh, contoured at 5.5σ) in the active site of H137A HDAC10. Atoms are color-coded as follows: C = white (HDAC10) or orange (substrate), N = blue, O = red, S = yellow, Zn²⁺ = large gray sphere, water molecules = small red spheres. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines. (b) Stereoview of a Polder omit map of the tetrahedral intermediate of N-acetylputrescine (black mesh, contoured at 7.5σ) in the active site of H137A HDAC10. Atoms are color-coded as in (a). Alternative orientations of these maps can be found in Figure S2.

It is curious that the high-energy tetrahedral intermediate of amide hydrolysis is stabilized in the HDAC10 active site in these crystalline complexes. Such an intermediate is metastable and is not typically one that can be isolated under standard laboratory conditions (although there is an exception for a twisted adamantyl amide²³). What features account for the stabilization of the hydrolytic tetrahedral intermediate in the active site of H137A HDAC10? Since this variant retains Y307, the substrate carbonyl group is fully activated by hydrogen bonding as well as Zn²⁺ coordination. This facilitates nucleophilic attack by the Zn²⁺-bound water molecule, assisted by general base H136. By Hammond’s Postulate,²⁴ we know that the resulting tetrahedral intermediate is structurally and energetically close to its flanking transition states; moreover, Pauling^25,26 noted that enzymes function by binding the transition state more tightly than substrate or product. Thus, with sufficient activation of the substrate carbonyl, the HDAC10 active site is set up for preferential binding and stabilization of the high-energy tetrahedral intermediate. Lacking general acid catalyst H137, however, the tetrahedral intermediate cannot proceed further along the hydrolytic reaction coordinate. Since crystals are harvested, stored, and subject to X-ray diffraction data collection at liquid nitrogen temperatures, the tetrahedral intermediate is literally frozen in place with its oxyanion stabilized by Zn²⁺ coordination and hydrogen bonding to Y307.

As observed for the binding of intact substrates, the NH group of each polyamine tetrahedral intermediate donates a hydrogen bond to the backbone carbonyl group of G145. In these structures, the side chain carboxylate group of E24 does not make direct hydrogen bond interactions with substrates. However, E24 interacts with the primary amino group of the N-acetylputrescine tetrahedral intermediate through a bridging water molecule. The primary amino group of N8-acetylspermidine is not definitively positioned due to weak or missing electron density, suggesting disorder for the amino end of the longer substrate. Additionally, gatekeeper residue E274 forms water-mediated hydrogen bonds with the secondary amino group of N⁸-acetylspermidine and the corresponding primary amino group of N-acetylputrescine. In each structure, the indole side chain of W205 shifts approximately 0.5 Å with respect to the Y307F HDAC10 structures; this conformational change may be due to the void resulting from the H137A substitution.

Product Complex.

The final step of catalysis is collapse of the tetrahedral intermediate to yield products acetate and spermidine or putrescine. The 2.15 Å-resolution structure of the wild-type HDAC10–acetate complex therefore represents a product complex from which spermidine has already dissociated. The acetate carboxylate group coordinates to the Zn²⁺ ion asymmetrically with Zn²⁺–O1 and Zn²⁺–O2 distances of 2.3 and 2.5 Å, respectively (Figure 4). The acetate carboxylate is additionally stabilized by hydrogen bond interactions: O1 accepts a hydrogen bond from Y307 (O---O distance = 2.5 Å), and O2 accepts hydrogen bonds from H136 and H137 (O2---N distances = 2.5 Å and 2.9 Å, respectively).

Figure 4. Enzyme-product complex.

Polder omit map of the acetate anion (contoured at 7.0σ) bound in the active site of wild-type zebrafish HDAC10. Atoms are color-coded as follows: C = white (HDAC10) or orange (acetate), N = blue, O = red, S = yellow, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

DISCUSSION

Chemical mechanism of acetylpolyamine deacetylation.

The substitution of catalytic residues in the active site of HDAC10 enables the X-ray crystallographic observation of key structures along the reaction coordinate of amide hydrolysis. Different snapshots of catalysis are revealed depending on the residue substituted, so these crystal structures yield powerful inferences regarding the specific function of each mutated residue. Crystalline enzymes are typically catalytically active, e.g., as demonstrated for the zinc hydrolase carboxypeptidase A,²⁷ so crystalline enzyme mutants can be employed in general approaches to trap reaction intermediates at the liquid nitrogen temperatures of X-ray diffraction data collection.

Early structural studies of the prototypical zinc hydrolases thermolysin and carboxypeptidase A indicated that the scissile peptide bond must be polarized for catalysis not only by the catalytic Zn²⁺ ion, but also (or instead) by an additional residue capable of donating a hydrogen bond to the peptide carbonyl group such as a histidine or arginine residue, respectively.^28,29 This is a general catalytic strategy for zinc hydrolases, including the HDACs: the class I, IIb, and IV enzymes contain a catalytic tyrosine residue, and the class IIa enzymes contain a catalytic histidine residue that serves this function.

In HDAC10, this catalytic tyrosine, Y307, assists the Zn²⁺ ion in polarizing the substrate carbonyl and also stabilizes the developing negative charge of the transition state enroute to the oxyanionic tetrahedral intermediate. The loss of this hydrogen bond in Y307F HDAC10 complexed with two different substrates, N⁸-acetylspermidine and N-acetylputrescine, results in the crystallographic entrapment of precatalytic Michaelis complexes. The amide carbonyl group of each intact substrate coordinates to the catalytic Zn²⁺ ion and the amide NH group of each donates a hydrogen bond to the backbone carbonyl of G145. The use of tyrosine variants of the lysine deacetylases HDAC8 and HDAC6 similarly enables trapping of precatalytic Michaelis complexes with tetrapeptide assay substrates, thus verifying the critical function of the catalytic tyrosine in the HDAC active site for activation of the scissile amide group and transition state stabilization.^7,30

At first glance, it is perhaps surprising that the hydrolytic tetrahedral intermediate is sufficiently stable in the active site of H137A HDAC10 for cocrystallization and X-ray structure determination. This intermediate is a high-energy species that, absent the enzyme active site, would not typically be isolable in aqueous solution. However, since the amino leaving group cannot leave this tetrahedral intermediate without being protonated, and since general acid H137 has been deleted by mutagenesis, the high-energy tetrahedral intermediate cannot proceed along the reaction coordinate of hydrolysis. While the reverse reaction leading to the intact enzyme-substrate complex could of course occur, the preferential binding of a transition state-like structure results in preferential stabilization of the tetrahedral intermediate.

Crystallographic results for HDAC10–acetylpolyamine substrate complexes reported here parallel crystallographic results for HDAC6–acetyllysine peptide substrate complexes:⁷ substitution of the catalytic tyrosine with phenylalanine yields an intact acetyllysine amide group coordinated to zinc in a precatalytic Michaelis complex, and substitution of the second histidine in the tandem pair yields a tetrahedral intermediate with its oxyanion stabilized by zinc coordination and a hydrogen bond with the catalytic tyrosine. These crystallographic snapshots of HDAC10 and HDAC6 are consistent with a mechanism in which the first histidine in the tandem pair serves as a general base in promoting the attack of the zinc-bound water molecule at the substrate carbonyl group, and the second histidine serves as a general acid to facilitate departure of the leaving amino group and collapse of the tetrahedral intermediate (Figure 1). Lacking a general base, the zinc-bound water molecule might not be sufficiently nucleophilic to attack the amide carbonyl, so the second histidine in the tandem pair is unlikely to serve as a general base in HDAC10 and HDAC6.

However, enzymological studies of HDAC8 indicate that the second histidine residue of the tandem pair serves both general base and general acid functions in this iszoyme.^31,32 Consistent with this finding, the X-ray crystal structure of the H143A HDAC8–substrate complex reveals the structure of an intact substrate bound in the precatalytic Michaelis complex and not as a tetrahedral intermediate.³³ Lacking the general base-general acid residue, the zinc-bound water molecule is not sufficiently reactive to attack the amide carbonyl. What accounts for this apparent difference in the function of the tandem histidine pair of HDAC8 compared with that of HDAC6 or HDAC10?

The answer to this question appears to be rooted in the hydrogen bond interactions of the tandem histidine residues in each deacetylase. Comparison of the active sites of these three deacetylases reveals that the first histidine of the tandem pair in each enzyme donates a hydrogen bond to a strictly conserved aspartate residue (Figure 5). A carboxylate-histidine interaction would enhance the basicity of the histidine residue due to charge-charge stabilization of the imidazolium form, raising the histidine pKa by about 1.5 units.^34,35 However, the second histidine of the tandem pair hydrogen bonds with a different residue in each deacetylase: in HDAC8 this residue is aspartate, but in HDAC6 and HDAC10 this residue is asparagine and glutamine, respectively. A hydrogen bond with the neutral carboxamide group of asparagine or glutamine will elevate the pKa of histidine by only half as much in comparison with a hydrogen bond to a negatively charged carboxylate,³⁶ so the second histidine residue in the tandem pair of HDAC6 and HDAC10 is not as basic as that of HDAC8. Thus, a working hypothesis is that the second histidine in HDAC6 and HDAC10 is not well-suited to serving as a general base in comparison with the second histidine of HDAC8. Accordingly, separate histidine residues in HDAC6 and HDAC10 would serve the separate functions of general base and general acid, whereas a single histidine residue in HDAC8 would be capable of serving both functions. This hypothesis unites all enzymological and structural data gathered to date for these three deacetylases.

Figure 5.

Stereoview showing the superposition of intact substrate complexes with Y745F HDAC6 (PDB 5EFK, cyan), Y306F HDAC8 (PDB 2V5W, yellow), and Y307F HDAC10 (current study, grey) active sites. Each substrate is shown with its scissile carbonyl coordinating to the catalytic Zn²⁺ ion (grey sphere); the Zn²⁺-bound water molecule is omitted for clarity. A strictly conserved aspartate hydrogen bonds with the first histidine of the tandem histidine pair in each enzyme, whereas the hydrogen bond partner of the second histidine of this pair is more variable: in HDAC6 it is asparagine, in HDAC8 it is aspartate, and in HDAC10 it is glutamine.

Ketone hydration reactions.

In general, hydrolytic enzymes exhibit a preference for binding tetrahedral transition state-like structures with non-hydrolyzable substrate analogues. For example, substrate analogues in which an aldehyde or unactivated ketone group is substituted for the scissile peptide linkage preferentially bind as high-energy tetrahedral gem-diol(ate)s in the active site of carboxypeptidase A.^37–39 Cyclic peptide inhibitors bearing epoxyketone side chains in which the ketone carbonyl is isosteric with the amide carbonyl of acetyllysine bind in the active sites of the lysine deacetylases HDAC6 and HDAC8 as gem-diol(ate)s.^7,40 Most recently, we observed that simple ketone analogues of N⁸-acetylspermidine bind as gem-diol(ate)s in the active site of HDAC10.⁴¹

While it is not surprising that activated, highly electrophilic trifluoromethylketone substrate analogues bind in the active sites of carboxypeptidase A, HDAC4, HDAC6, and HDAC10 as gem-diol(ate)s,^6,7,41–43 i.e., the form of the inhibitor that is ca. 99% predominant in aqueous solution, it is surprising to observe the binding of unactivated ketone substrate analogues as gem-diol(ate)s. Acetone is a representative example of an unactivated ketone and is in equilibrium with less than 0.2% of its gem-diol form in aqueous solution, indicating that the gem-diol is at least 3.8 kcal/mol less stable than the ketone.⁴⁴ If the pKa of a gem-diol hydroxyl group is ~16, then the gem-diol oxyanion would be ~12 kcal/mol less stable than the neutral species at pH 7. The total stabilization free energy of ~16 kcal/mol for the binding and stabilization of a gem-diolate – or an oxyanionic tetrahedral intermediate and its flanking transition states in catalysis – reflects the impressive power that a zinc hydrolase brings to bear on inhibitor binding and catalysis.

Structural basis of substrate specificity.

HDAC10 is unique among the greater family of HDAC enzymes in terms of its substrate specificity for acetylpolyamines, and even more so in terms of its specificity for N⁸-acetylspermidine but not N¹-acetylspermidine. Steady-state kinetic assays of human and zebrafish HDAC10 with various acetylpolyamine substrates suggest that the precise location of secondary and/or primary ammonium cations in the substrate dictates substrate specificity.⁶ As mentioned previously, the PEACE motif helix and gatekeeper E274 play key roles in governing substrate specificity for acetylpolyamines based on mutagenesis studies,⁶ but the direct visualization of substrate binding in the HDAC10 active site has not been achieved until now. The crystal structures reported herein reveal that slender acetylpolyamine substrates fit perfectly in the constricted active site. The gatekeeper residue, E274, as well as E24 in the humanized PEACE motif helix, establish a negative electrostatic potential in the active site that is complementary to cationic acetylpolyamine substrates. Gatekeeper E274 is unique to HDAC10 and is a critical determinant of polyamine substrate specificity,⁶ so it is remarkable to see that E274 dictates specificity through water-mediated and not direct hydrogen bonds.

Curiously, the deacetylation activity of HDAC10 against N¹-acetylspermidine is negligible in comparison with that measured with N⁸-acetylspermidine.⁶ Analysis of the crystal structures of complexes with N⁸-acetylspermidine (Figures 2a and 3a) suggests that gatekeeper E274 is responsible for substrate discrimination through its two hydrogen bonded water molecules, which form hydrogen bonds with the secondary ammonium group of N⁸-acetylspermidine with a pincer-like geometry (Figure 6). Presuming that both N⁸-acetylspermidine and N¹-acetylspermidine would bind in a similar extended conformation due to the steric constrictions of the active site, the secondary amino group of N¹-acetylspermidine would be oriented away from E274 and would be unable to form these hydrogen bonds (Figure 6). Thus, the geometry of hydrogen bonding with the E274-water pair favors N⁸-acetylspermidine binding and disfavors N¹-acetylspermidine binding.

Figure 6.

(a) Crystal structure of N⁸-acetylspermidine bound in the active site of Y307F HDAC10 (this orientation is rotated by approximately 90º compared with the orientation shown in Figure 2a). E274 is anchored in place by a hydrogen bond with Zn²⁺ ligand H176 and also accepts hydrogen bonds from two water molecules (red spheres), which in turn accept hydrogen bonds from each proton of the secondary ammonium group of the substrate with a pincer-like geometry. (b) Model of N¹-acetylspermidine bound in the active site of Y307F HDAC10 based on the structure in (a); switching the position of the secondary ammonium group and the adjacent methylene group converts N⁸-acetylspermidine into N¹-acetylspermidine. The protons of the secondary ammonium group of N¹-acetylspermidine are oriented away from E274 and its hydrogen bonded water molecules, preventing hydrogen bond formation. Thus, the binding of N¹-acetylspermidine is disfavored relative to N⁸-acetylspermidine.

CONCLUSIONS

The X-ray crystal structures of HDAC10–substrate complexes conclusively reveal the structural basis of polyamine substrate specificity. Previous mutagenesis experiments indicated that the carboxylate side chain of E274 was critical for molecular recognition of N⁸-acetylspermidine and N-acetylputrescine.⁶ Here, structures of enzyme-substrate complexes reveal that E274 forms hydrogen bonds with water molecules that, in turn, form hydrogen bonds with the secondary amino group of N⁸-acetylspermidine and the primary amino group of N-acetylputrescine. In other words, E274 exerts its effect not by forming a direct salt-linked hydrogen bond with the substrate, but instead by forming water-mediated hydrogen bonds. Additionally, the model of the HDAC10–N¹-acetylspermidine complex shows that the corresponding secondary amino group of N¹-acetylspermidine would be oriented away from E274 and its hydrogen bonded water molecules (Figure 6), consistent with the dramatic specificity for hydrolysis of N⁸-acetylspermidine compared to N¹-acetylspermidine (Figure 1b).^3,6

In addition to the steric constriction afforded by the PEACE helix, which allows for the binding of only long, slender substrates, E24 in the “humanized” form of this helix engages in salt-linked hydrogen bonds with substrate amino groups. This feature, too, contributes to catalytic specificity for long, slender, cationic polyamine substrates.

It is remarkable that the Zn²⁺-bound oxyanionic tetrahedral intermediate can be stabilized in an enzyme active site and trapped in the crystal for X-ray crystal structure determination (Figure 3). This result with the polyamine deacetylase HDAC10, and the similar result previously reported with the lysine deacetylase HDAC6,⁷ show that the enzyme active site preferentially binds a tetrahedral transition state-like structure rather than a planar substrate-like structure. This chemical behavior directly demonstrates the principle that enzymes catalyze chemical reactions in large part by preferentially stabilizing their transition states.^25,26

Finally, the structures reported herein and comparison with the structure of the H143A HDAC8–substrate complex³⁴ suggest that the catalytic function of the tandem histidine pair in the HDAC active site differs in the details of general base-general acid catalysis in the chemical mechanism of amide hydrolysis. Future enzymological studies will allow us to probe these differences in detail.