Binding of N⁸-Acetylspermidine Analogues to Histone Deacetylase 10 Reveals Molecular Strategies for Blocking Polyamine Deacetylation

Abstract

Eukaryotic histone deacetylase 10 (HDAC10) is a Zn²⁺-dependent hydrolase that exhibits catalytic specificity for the hydrolysis of the polyamine N⁸-acetylspermidine. The recently determined crystal structure of HDAC10 from Danio rerio (zebrafish) reveals a narrow active site cleft and a negatively charged “gatekeeper” (E274) that favors the binding of the slender cationic substrate. Since HDAC10 expression is upregulated in advanced-stage neuroblastoma and induces autophagy, the selective inhibition of HDAC10 suppresses the autophagic response and renders cancer cells more susceptible to cytotoxic chemotherapeutic drugs. Here, we describe X-ray crystal structures of zebrafish HDAC10 complexed with eight different analogues of N⁸-acetylspermidine. These analogues contain different Zn²⁺-binding groups, such as hydroxamate, thiolate, and the tetrahedral gem-diolate resulting from the addition of a Zn²⁺-bound water molecule to a ketone carbonyl group. Notably, the chemistry that accompanies the binding of ketonic substrate analogues is identical to the chemistry involved in the first step of catalysis, i.e., nucleophilic attack of a Zn²⁺-bound water molecule at the scissile carbonyl group of N⁸-acetylspermidine. The most potent inhibitor studied contains a thiolate Zn²⁺-binding group. These structures reveal interesting geometric changes in the metal coordination polyhedron that accommodate inhibitor binding. Additional interactions in the active site highlight features contributing to substrate specificity. These interactions are likely to contribute to inhibitor binding selectivity and will inform the future design of compounds selective for HDAC10 inhibition.

Graphical Abstract

INTRODUCTION

Ubiquitous in all forms of life, cationic polyamines such as putrescine, spermidine, and spermine are present at millimolar concentrations in cells where they are essential for growth and function in countless biological processes.^1–4 In eukaryotes, polyamines derive from amino acid catabolism (Figure 1a). For example, L-arginine hydrolysis yields L-ornithine, decarboxylation of which yields putrescine; modification of putrescine with one or two equivalents of S-adenosyl 3-(methylsulfanyl)propylamine yields spermidine or spermine, respectively. Polyamines are subject to additional, highly regulated pathways of chemical modification and catabolism to meet the needs of the cell, and dysregulation of these pathways can result in various disease pathologies. Thus, enzymes of polyamine metabolism are attractive as potential drug targets.^5–8

Figure 1.

(a) Eukaryotic polyamine metabolism. Abbreviations: AAP, 3-acetylaminopropanal; ADC, arginine decarboxylase; APAO, N¹-acetylpolyamine oxidase; HDAC10, N⁸-acetylspermidine deacetylase; MTA, S-methyl-5’-thioadenosine; ODC, ornithine decarboxylase; P/CAF, p300/CBP-associated factor; SAMP, S-adenosyl 3-(methylsulfanyl)propylamine; SMS, spermine synthetase; SMOX, spermine oxidase; SRM, spermidine synthase; SSAT, spermidine/spermine acetyltransferase. (b) Structure of wild-type zebrafish HDAC10 (PDB 6UIL). Polyamine deacetylase domain (PDAC, cyan); pseudo-deacetylase domain (ΨDAC, tan); P(E,A)CE motif (magenta); E274 (red stick figure side chain); Zn²⁺, gray sphere. A bound inhibitor (compound 1, stick figure) indicates the location of the active site. (c) Close-up view of the active site showing the sterically constricted molecular surface that confers specificity for slender polyamine substrates. Protein features are color-coded as in (b).

The acetylation of spermidine and spermine comprises a key chemical modification in eukaryotic polyamine metabolism (Figure 1a). For instance, spermidine is selectively acetylated in the nucleus to yield N⁸-acetylspermidine,⁹ which is then exported to the cytosol. There, deacetylation yields spermidine,^10,11 which can be utilized for spermine biosynthesis. Alternatively, cytosolic spermidine can be acetylated to yield N¹-acetylspermidine,¹² which then undergoes oxidative deamination to yield putrescine and N-acetyl-3-aminopropanal.¹³

Although cytosolic spermidine deacetylase activity was first observed in 1978, the enzyme responsible for this activity was not identified at the time.^10,11 Notably, the substrate specificity of this activity was very strict, since the deacetylation of N¹-acetylspermidine or N¹-acetylspermine was not observed. Later studies showed that selective inhibition of cytosolic deacetylation activity in HeLa cells did not influence acetylated histone levels but did increase N⁸-acetylspermidine levels, so polyamine deacetylase activity is clearly distinct from histone deacetylase activity.¹⁴

We recently discovered that histone deacetylase 10 (HDAC10) enzymes from Homo sapiens (human) and Danio rerio (zebrafish) exhibit highly specific deacetylase activity toward N⁸-acetylspermidine, but not toward N¹-acetylspermidine, acetylspermine, or acetyllysine-containing peptide substrates in vitro.¹⁵ Thus, it is likely that HDAC10 is the cytosolic N⁸-acetylspermidine deacetylase in vivo. This catalytic function is especially intriguing in view of the discovery that HDAC10 induces autophagy, a cellular survival mechanism activated in response to cellular damage.¹⁶ Specifically, HDAC10 protects malignant cells from Doxorubicin, a cytotoxic drug used in cancer chemotherapy. The upregulation of HDAC10 signals a robust autophagic response and serves as a biomarker for poor outcome in advanced-stage neuroblastoma. However, inhibition of HDAC10 suppresses autophagy and renders cancer cells more susceptible to Doxorubicin. Therefore, clinical co-administration of a selective HDAC10 inhibitor along with Doxorubicin may improve the efficacy of cancer chemotherapy.

Our recent X-ray crystal structure of zebrafish HDAC10 revealed a “butterfly”-like architecture in which the catalytically active deacetylase domain is paired with a smaller pseudo-deacetylase domain with identical topology but no known catalytic function (Figure 1b).¹⁵ The active site in the deacetylase domain contains the catalytic Zn²⁺ ion as well as Y307, which assists the Zn²⁺ ion in polarizing the scissile carbonyl of the substrate for nucleophilic attack by a Zn²⁺-bound solvent molecule; tandem histidine residues H136 and H137 are believed to serve general base-general acid functions. The active site cleft of HDAC10 is sterically constricted by a 3₁₀ helix containing the P(E,A)CE motif, and an active site glutamate, E274, confers specificity for slender cationic substrates (Figure 1c). Both features are unique to HDAC10 orthologs.

In separate studies, we reported the design and synthesis of polyamine analogues (Figure 2) that inhibit the prokaryotic polyamine deacetylase, acetylpolyamine amidohydrolase from Mycoplana ramosa (mrAPAH), along with X-ray crystal structures of mrAPAH–inhibitor complexes.^17,18 Here, we present X-ray crystal structures of these inhibitors in their complexes with zebrafish HDAC10 as the first step in understanding structure-affinity relationships in the active site of a eukaryotic polyamine deacetylase. Ultimately, we expect that these structural data will inform the design and development of clinically useful HDAC10-selective inhibitors.

Figure 2.

Polyamine analogues studied in complexes with HDAC10. 1, 7-[(3-aminopropyl)amino]-1,1,1-trifluoroheptan-2-one, IC₅₀ = 80 ± 10 nM; 2, 7-[(3-aminopropyl)amino]heptan-2-one, IC₅₀ = 2.3 ± 0.2 μM; 3, 7-[(3-aminopropyl)amino]-1-methoxyheptan-2-one, IC₅₀ = 35.0 ± 0.1 μM; 4, 5-[(3-aminopropyl)amino]pentylboronic acid, IC₅₀ = 400 ± 70 μM; 5, 6-[(3-aminopropyl)amino]-N-hydroxyhexanamide, IC₅₀ = 120 ± 20 nM; 6, 5-[(3-aminopropyl)amino]pentane-1-thiol, IC₅₀ = 370 ± 20 nM; 7, S-{5-[(3-aminopropyl)amino]pentyl} thioacetate, IC₅₀ = 30 ± 10 nM; 8, 7-{[(3-aminopropyl)amino]-2-oxoheptyl} thioacetate, IC₅₀ = 180 ± 20 nM. All IC50 curves are shown in Figure S1.

MATERIALS AND METHODS

Reagents.

All chemicals, buffers, and general reagents were purchased from Sigma-Aldrich and used without further purification unless otherwise indicated. Polyamine deacetylase inhibitors 1–8 (Figure 2) were prepared as previously described.^17,18

Protein expression and purification.

Wild-type HDAC10 and Y307F HDAC10 from D. rerio were prepared and purified as previously described.¹⁵ Briefly, cells containing each plasmid were grown in 70 mL of 2x YT medium supplemented with 50 mg/L of kanamycin. Cultures were grown overnight at 37 °C – inoculation of growth medium was initiated by adding 5 mL of saturated culture to 1 L of 2x YT in the presence of 50 mg/L kanamycin and grown at 37 °C with shaking at 250 rpm. When OD₆₀₀ reached 0.8, shaking was stopped, the temperature was lowered to 16 °C, and the medium was cooled for 30 min. Expression was induced with the addition of 250 μM ZnSO₄ (Fisher Scientific) and 200 μM isopropyl-β-D-1-thiogalactopyranoside (IPTG; GoldBio), and cell cultures were grown for an additional 18 hours at 16 °C with shaking at 250 rpm. Cells were then centrifuged at 5,422 g, the supernatant was poured off, and cell pellets were stored at −80 °C until further use.

Immediately prior to purification, pellets were thawed in a water bath at 4 °C for no longer than 2 hours. Lysis buffer [50 mM HEPES (pH 7.5), 300 mM KCl, 10% glycerol, 2 mM TCEP, 10 μM ZnSO₄, 30 mM imidazole, 0.4 mg/mL lysozyme (MPbiomedicals), 2.8 units/mL benzonase nuclease (Sigma), and protease inhibitor tablets (Roche Applied Science)] was added to the cell pellet in a 2:1 ratio (v:w) of lysis buffer:cell pellet. The suspension was stirred at 4 °C for 45 minutes and then lysed by sonication. The lysate was centrifuged at 41,657 g at 4°C for 1 hour and then loaded onto a 5 mL HisTrap (GE) column pre-equilibrated with loading buffer [50 mM HEPES (pH 7.5), 300 mM KCl, 10% glycerol, 2 mM TCEP, 10 μM ZnSO₄, and 30 mM imidazole]. Protein was eluted with elution buffer [50 mM HEPES (pH 7.5), 300 mM KCl, 10% glycerol, 2 mM TCEP, 10 μM ZnSO₄, and 500 mM imidazole]. All fractions under the A₂₈₀ peak were pooled and digested using TEV protease overnight and extensively dialyzed into loading buffer.

After overnight digestion, the solution was loaded onto a tandem MBPTrap (GE)-HisTrap column pre-equilibrated with loading buffer; the flow-through was collected, pooled, concentrated, and loaded onto a HiLoad Superdex 200 column equilibrated with size exclusion buffer [50 mM HEPES (pH 7.5), 300 mM KCl, 5% glycerol, and 1 mM TCEP]. Protein fractions were identified by SDS-PAGE, concentrated to approximately 10 mg/mL, flash cooled, and stored at −80 °C.

Inhibitory potencies.

Inhibitors were evaluated using the acetylpolyamine hydrolysis assay described by Hai and colleagues.¹⁵ Substrate concentrations as well as reaction durations were optimized prior to affinity determinations. Briefly, 125–250 nM HDAC10 in assay buffer [20 mM HEPES (pH 7.5), 100 mM NaCl, 5 mM KCl, 1 mM MgCl₂] was incubated with 100 pM −10 mM inhibitor for 10 minutes. Substrate N-acetylputrescine (800 μM) was added; after 45–60 minutes, the reaction was quenched by the addition of 10 μL of 1.6 M sodium bicarbonate and 40 μL of 12.5 mM dansyl chloride and heating at 50 °C for 1 h. The dansylated product was identified and quantified using liquid chromatography-mass spectrometry to determine enzyme activity in the presence and absence of inhibitor. Data were fit to a logistic nonlinear regression model,

$$Y=Y_{max}+\frac{Y_{min}-Y_{max}}{1+{10}^{I{C}_{50}-[I]}}$$

where Y = enzyme activity, Y_min is the minimum asymptote, Y_max is the maximum asymptote, [I] is inhibitor concentration, and IC₅₀ is the inflection point of the curve.

Crystallization of enzyme-inhibitor complexes.

Individual protein solutions containing HDAC10 and each inhibitor shown in Figure 2 were prepared by adding 5 mM inhibitor to a protein solution comprised of 10 mg/mL HDAC10 in size exclusion buffer and equilibrated on ice for 1 h. Trypsin was then added (1:1000 trypsin:HDAC10 molar ratio). After digestion at room temperature for 1 h, the protein solution was filtered using 0.22-μm centrifuge filters.

All enzyme-inhibitor complexes were crystallized by the sitting drop vapor diffusion method at 4 °C. Typically, a 100-nL drop of freshly filtered protein solution was added to a 100-nL drop of precipitant buffer and equilibrated against 80 μL of precipitant solution in the well reservoir of a 96-well crystallization plate using a Mosquito crystallization robot (TTP Labtech). Crystallization of the Y307F HDAC10–6 complex was achieved using 0.2 M KH₂PO₄, 20% PEG 3350, and 4% v/v polypropylene glycol as precipitant. Crystallization of the HDAC10 complexes with 1 and 8 was achieved using 0.2 M KH₂PO₄, 20% PEG 3350, and 0.5% (w/v) n-dodecyl-β-D-maltoside as precipitant. Crystallization of the HDAC10–7 complex was achieved using 0.2 M (NH)₄H₂PO₄, 20% PEG 3350, and 4% (v/v) 1-propanol as precipitant. Crystallization of the HDAC10 complexes with 2 and 3 was achieved using 0.1 M KH₂PO₄, 0.1 M K₂HPO₄, and 20% PEG 3350 as precipitant; micro-seed crystals of the HDAC10–1 complex were also added to the crystallization drop. Crystallization of the HDAC10–4 complex was achieved using 0.15 M KH₂PO₄, 0.05 M K₂HPO₄, and 20% PEG 3350 as precipitant; micro-seed crystals of the HDAC10–1 complex were also added to the crystallization drop. Crystallization of the HDAC10-5 complex was achieved using 0.2 M (NH)₄H₂PO₄, 20% PEG 3350, and 0.1 M sodium citrate tribasic dihydrate as precipitant; micro-seed crystals of the HDAC10-1 complex were also added to the crystallization drop.

Crystal structure determinations.

X-ray diffraction data for the Y307F HDAC10–6 complex were collected on NE-CAT beamline 24-ID-E at the Advanced Photon Source, Argonne National Laboratory (Argonne, IL). X-ray diffraction data for the HDAC10 complexes with 1, 7, and 8 were collected at the Stanford Synchrotron Radiation Lightsource, beamline 9–2. X-ray diffraction data for the HDAC10 complexes with 2–5 were collected at the National Synchrotron Light Source II (NSLSII), Brookhaven National Laboratory (Upton, NY). Diffraction data from crystals of the Y307F HDAC10–6 complex, as well as HDAC10 complexes with 1, 4, 7, and 8 were integrated using iMosflm¹⁹ and scaled using Aimless in the CCP4 program suite.^20,21 Diffraction data from HDAC10 complexes with 2, 3, and 5 were reduced and integrated using X-ray Detector Software (XDS)²² as implemented on NSLSII servers. All structures were solved by molecular replacement using the crystal structure of the Y307F HDAC10–1 complex (5TD7)¹⁵ with ligands and water molecules removed as a search model with the program Phaser.²³ Model building was performed using COOT and crystallographic refinement, including simulated annealing refinement to minimize phase bias, was achieved using Phenix.^24,25 During the refinement of each enzyme-inhibitor complex, the inhibitor was fit to the electron density map in the final stages of refinement. A limited number of water molecules were also added in the final stage of each refinement; given the moderate resolutions of these structures, only water molecules characterized by strong electron density peaks in chemically reasonable positions were modeled. Some polypeptide segments were omitted from each final model due to poor or absent electron density; these segments belonged exclusively to the pseudo-deacetylase domain and are listed in Table S1. In general, the pseudo-deacetylase domain exhibited greater disorder/flexibility compared with the catalytically active polyamine deacetylase domain, and this difference was manifest in higher average B-factors for the pseudo-deacetylase domain (Table S1). MolProbity²⁶ was used to assess and validate all final refined structures. All data collection and refinement statistics are recorded in Table 1.

Table 1.

Crystallographic data collection and refinement statistics.

Complex	HDAC10–1	HDAC10–2	HDAC10–3	HDAC10–4	HDAC10–5	Y307F HDAC10–6	HDAC10–7	HDAC10–8
Space group	P31 2 1	P31 2 1	P31 2 1	P31 2 1	P31 2 1	P31 2 1	P31 2 1	P31 2 1
Unit cell dimensionsa
a,b,c (Å)	81.0 81.0 242.8	80.6 80.6 246.2	80.6 80.6 247.6	80.6 80.6 246.2	80.5 80.5 245.9	80.8 80.8 244.3	80.8 80.8 242.6	80.9 80.9 241.5
Rmergeb	0.109 (1.384)	0.120 (0.758)	0.154 (0.860)	0.348 (1.818)	0.147 (0.664)	0.132 (1.510)	0.164 (1.630)	0.168 (0.890)
Rpimc	0.056 (0.704)	0.057 (0.363)	0.073 (0.421)	0.180 (0.974)	0.070 (0.312)	0.064 (0.725)	0.084 (0.853)	0.085 (0.448)
CCl/2d	0.981 (0.589)	0.998 (0.847)	0.993 (0.806)	0.953 (0.502)	0.995 (0.855)	0.995 (0.639)	0.998 (0.512)	0.990 (0.842)
Redundancy	8.0 (9.1)	9.9 (9.9)	9.9 (9.4)	9.2 (9.5)	10.1 (10.2)	9.7 (10.0)	8.8 (8.7)	8.8 (9.2)
Completeness (%)	99.8 (100)	99.8 (97.7)	99.7 (96.7)	100 (100)	99.7 (97.1)	100 (100)	99.9 (99.9)	99.8 (100)
I/σ	14.9 (1.5)	14.5 (2.5)	10.5 (2.2)	8.5 (3.3)	10.3 (2.4)	11.5 (2.1)	10.4 (1.5)	7.2 (1.9)
Refinement
Resolution (Å)	49.50–2.85 (2.95–2.85)	28.74–2.70 (2.77–2.70)	28.84–2.68 (2.78–2.68)	69.76–2.80 (2.90–2.80)	28.72–2.53 (2.60–2.53)	60.71–2.65 (2.75–2.65)	40.43–2.90 (3.00–2.90)	40.44–2.75 (2.85–2.75)
No. reflections	22377 (2209)	26215 (2554)	27000 (2553)	23533 (2305)	31691 (3068)	27778 (2710)	21155 (2088)	24629 (2416)
Rwork/Rfreee	0.244/0.305 (0.358/0.367)	0.189/0.228 (0.276/0.377)	0.195/0.235 (0.265/0.344)	0.197/0.254 (0.258/0.328)	0.204/0.246 (0.279/0.320)	0.204/0.257 (0.283/0.357)	0.204/0.249 (0.329/0.389)	0.223/0.276 (0.342/0.421)
No. atomsf
Protein	4565	4878	4966	4834	4723	4685	4700	4756
Ligand	30	37	34	32	27	29	24	27
Solvent	19	51	55	39	70	25	18	14
Average B factors (Å2)
Protein	79	51	55	40	51	67	73	73
Ligand	63	58	61	44	45	66	58	56
Solvent	64	39	45	25	42	53	54	54
R.m.s. deviations
Bond lengths (Å)	0.003	0.005	0.065	0.002	0.003	0.005	0.005	0.004
Bond angles (°)	0.7	1.0	2.0	0.6	0.5	0.7	0.8	0.7
Ramachandran plotg
Favored	92.7	96.2	95.9	95.7	96.7	96.2	94.4	94.6
Allowed	6.8	3.7	4.0	4.3	3.1	3.7	5.6	4.9
Outliers	0.5	0.2	0.2	0.0	0.2	0.2	0.0	0.5
PDB code	6UIL	6UFN	6UFO	6UHU	6UHV	6UII	6UIJ	6UIM

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

^bR_merge = Σ_hklΣ_i|I_i,hkl − 〈I〉_hkl|/Σ_hklΣ_iI_i,hkl, where 〈I〉_hkl is the average intensity calculated for reflection hkl from replicate measurements.

^cR_p.i.m.= (Σ_hkl(1/(N-1))^1/2Σ_i|I_i,hkl − 〈I〉_hkl|)/Σ_hklΣ_i I_i,hkl, where 〈I〉_hkl is the average intensity calculated for reflection hkl from replicate measurements and N is the number of reflections.

^dPearson correlation coefficient between random half-datasets.

^eR_work = Σ∥F_o| − |F_c∥/Σ|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

Inhibitory potencies.

The IC₅₀ values for compounds 1–8 against zebrafish HDAC10 were determined using the fixed-point N-acetylputrescine hydrolysis assay.¹⁵ The IC₅₀ value of each inhibitor is recorded in the Figure 2 legend. Individual IC₅₀ plots are found in Figure S1 of the Supporting Information.

Binding modes of reactive substrate analogues 1–4.

The crystal structure of the wild-type HDAC10–1 complex was determined at 2.85 Å resolution and is very similar to the previously reported structure of the Y307F HDAC10–1 complex determined at 2.85 Å resolution (root-mean-square (rms) deviation = 0.30 Å for 522 Cα atoms).¹⁵ The trifluoromethylketone moiety binds to the catalytic Zn²⁺ ion as a gem-diolate in bidentate fashion, with Zn²⁺---O1 and Zn²⁺---O2 coordinate bond lengths of 2.0 Å and 2.2 Å, respectively (Figure 3). The Zn²⁺-bound oxyanion O1 accepts a hydrogen bond from Y307; Zn²⁺-bound hydroxyl group O2 forms hydrogen bonds with tandem histidine residues H136 and H137, believed to serve general base-general acid functions in catalysis.^15,27 This structure is the first to confirm the proposed role of Y307 in stabilizing the tetrahedral intermediate and its flanking transition states in catalysis, since the previously reported structure was that of the Y307F variant.¹⁵

Figure 3.

Polder omit map of 1 bound in active site of HDAC10 (contoured at 5.5σ). Atomic color codes are as follows: C = cyan (HDAC10) or orange (inhibitor 1), N = dark blue, O = red, S = yellow, F = lime green, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

It is likely that inhibitor 1 exists predominantly as the tetrahedral gem-diol in aqueous solution, based on studies of related trifluoromethylketone enzyme inhibitors first reported by Gelb and collegues.²⁸ The binding of one of these inhibitors with an oxyanion in the active site of chymotrypsin was definitively determined by NMR spectroscopy.²⁹ Accordingly, it is likely that the O1 hydroxyl group of 1, stabilized by a 2.0 Å coordination interaction with Zn²⁺, a hydrogen bond with Y307, and the electron-withdrawing inductive effects of the trifluoromethyl group, similarly binds as the oxyanionic gem-diolate in the active site of HDAC10. Notably, the first crystal structure determination of a Zn²⁺-metalloenzyme complexed with a trifluoromethylketone inhibitor similarly revealed the binding of the gem-diol form of the inhibitor; here, too, the gem-diol is likely stabilized as the oxyanionic gem-diolate.³⁰

Interestingly, the binding of 1 causes the D174-Zn²⁺ coordination interaction to shift from monodentate to bidentate, thereby increasing the coordination number to 6. This shift in metal coordination causes a slight increase in crowding around the Zn²⁺ ion, resulting in a slight lengthening of Zn²⁺ coordination interactions.

Additional hydrogen bond interactions are observed in the HDAC10–1 complex. One of the trifluoromethyl C–F groups accepts a hydrogen bond from Y307. Additionally, the secondary amino group (N4) donates a hydrogen bond to the side chain of D94. Finally, the primary amino group (N1) is within hydrogen bonding distance to E274; however, strong electron density is weak or absent beyond the N4 atom, likely reflecting some degree of molecular disorder. Thus, the inhibitor N1---E274 hydrogen bond interaction does not firmly anchor this end of the inhibitor.

The crystal structure of the wild-type HDAC10–2 complex was determined at 2.70 Å resolution. Inhibitor 2 is nearly isosteric with inhibitor 1, except that inhibitor 2 contains an unactivated methylketone in contrast with the highly electrophilic trifluoromethylketone moiety of 1. Surprisingly, however, inhibitor 2 nevertheless binds as a hydrated gem-diol(ate) in the active site of HDAC10, with Zn²⁺---O1 and Zn²⁺---O2 coordinate bond lengths of 2.0 Å and 2.3 Å, respectively (Figure 4). Since unactivated ketones such as that of 2 exist as less than 0.2% gem-diol hydrate in aqueous solution,³¹ it is likely that inhibitor 2 undergoes nucleophilic attack by the Zn²⁺-bound water molecule upon binding in the enzyme active site. Alternatively, it remains a possibility that the pre-formed gem-diol hydrate from solution is preferentially selected for binding, in which case the inhibitory potency of the actual inhibiting species would be approximately 500-fold better than that indicated by the IC₅₀ of 2.3 μM.

Figure 4.

Polder omit map of 2 bound in active site of HDAC10 (contoured at 5.5σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 2), N = dark blue, O = red, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

As observed for the binding of 1, the Zn²⁺-bound gem-diol(ate) of 2 is stabilized by hydrogen bond interactions. Here, too, the O1 hydroxyl group may be ionized to form a Zn²⁺-bound oxyanion also stabilized by a hydrogen bond with Y307. Hydroxyl group O2 hydrogen bonds with tandem histidine residues H136 and H137. Similar gem-diol binding modes are observed for unactivated aldehyde and ketone substrate analogues bound to the Zn²⁺ ion in the active site of carboxypeptidase A,^32–35 so the carbonyl hydration reaction is a chemical function shared by the greater family of Zn²⁺ metallohydrolases.

The secondary amino group of 2 (N4) donates a hydrogen bond to the carboxylate side chain of D94 and also engages in a cation-π interaction with the aromatic side chain of W205 (Figure 4). The primary amino group of 2 does not engage in any direct hydrogen bonds with protein residues.

The crystal structure of the wild-type HDAC10–3 complex was determined at 2.68 Å resolution. Inhibitor 3 contains an unactivated ketone like that of 2; similarly, inhibitor 3 binds as a hydrated gem-diol(ate) with Zn²⁺---O1 and Zn²⁺---O2 coordinate bond lengths of 2.0 Å and 2.2 Å, respectively (Figure 5). The O1 hydroxyl group of 3 may be ionized as the Zn²⁺-bound oxyanion, also stabilized by a hydrogen bond with Y307. Hydroxyl group O2 forms hydrogen bonds with H136 and H137.

Figure 5.

Polder omit map of 3 bound in active site of HDAC10 (contoured at 5.5σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 3), N = dark blue, O = red, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

The methoxy moiety of inhibitor 3 introduces additional steric bulk that is accommodated deep in the HDAC10 active site. Indeed, the lack of electron density corresponding to this group suggests that it is disordered. Hence, it is modeled into a chemically reasonable conformation in the final model of the enzyme-inhibitor complex (Figure 5). Similar to the binding interactions observed in the HDAC10–2 complex, inhibitor 3 donates a hydrogen bond from N4 to the carboxylate side chain of D94. The positively charged amino group is also stabilized by a cation-π interaction with W205. The primary amine of 3 (N1) may form a hydrogen bond with D94 as well, but because the electron density is broken and weak after N4, this likely represents only one of many possible conformations.

The crystal structure of the wild-type HDAC10–4 complex was determined at 2.80 Å resolution. Inhibitor 4 contains a trigonal planar boronic acid moiety that undergoes nucleophilic attack, presumably by the Zn²⁺-bound water molecule, to yield a tetrahedral boronate anion (Figure 6). Boronate hydroxyl groups O1 and O2 coordinate to Zn²⁺ with Zn²⁺---O coordinate bond lengths of 1.9 Å and 2.4 Å, respectively. Additionally, O1 forms a hydrogen bond with Y307, and O2 forms hydrogen bonds with H136 and H137. The binding mode of the tetrahedral boronate anion is similar to that of the tetrahedral gem-diol(ate) forms of inhibitors 1–3, which similarly undergo reaction with the Zn²⁺-bound water molecule to bind as tetrahedral transition state analogues. Similar chemistry has been observed for boronic acid binding to the manganese metalloenzyme arginase.^36–38

Figure 6.

Polder omit map of 4 bound in active site of HDAC10 (contoured at 5.0σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 4), N = dark blue, O = red, B = pink, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

The N4 secondary amine of inhibitor 4 forms two electrostatic interactions with the side chain of D94 and E274 and is further stabilized by a cation-π interaction with W205. The N1 primary amine is oriented toward the 3₁₀ helix that sterically constrains the active site (uniquely conserved among HDAC10 orthologs as P(E,A)CE), and also forms a hydrogen bond with E274. Both of these structural elements contribute to the polyamine substrate specificity of HDAC10.¹⁵

Binding mode of hydroxamate analogue 5.

The crystal structure of the wild-type HDAC10–5 complex was determined at 2.53 Å resolution. Inhibitor 5 contains a hydroxamic acid metal-binding group, which ionizes to the anionic hydroxamate and chelates the catalytic Zn²⁺ ion upon binding in the active site with C=O---Zn²⁺ and N–O⁻---Zn²⁺ coordinate bond lengths of 2.0 Å each (Figure 7). The hydroxamate carbonyl also accepts a hydrogen bond from Y307. The hydroxamate NH group donates a hydrogen bond to H137, and the Zn²⁺-bound hydroxamate N–O⁻ group accepts a hydrogen bond from H136. To accommodate these hydrogen bond interactions, the side chain of H136 must be the positively charged imidazolium cation and the side chain of H137 must be the neutral imidazole.

Figure 7.

Polder omit map of 5 bound in active site of HDAC10 (contoured at 5.5σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 5), N = dark blue, O = red, Zn²⁺ = gray sphere. Hydrogen bonds are shown as black dashed lines and metal coordination interactions are shown as black solid lines.

The orientation of the hydroxamate bound to the Zn²⁺ ion shifts the N4 secondary amine of inhibitor 5 further out of the active site cavity compared with previously described inhibitors. This shift allows N4 to donate a bifurcated hydrogen bond to the carboxylate side chain of E274.

Binding modes of thio-analogues 6–8.

The crystal structure of the Y307F HDAC10–6 complex was determined at 2.65 Å resolution. Inhibitor 6 contains a thiol metal-binding group, which ionizes to the anionic thiolate upon coordination to the catalytic Zn²⁺ ion to form a tetrahedral complex (Figure 8). The Zn²⁺---S coordinate bond length of 2.3 Å and the C–S⁻---Zn²⁺ angle of 107° are nearly ideal, but the C–C–S⁻---Zn²⁺ torsion angle of 52° deviates, from ideal values tabulated for cysteine thiolate–metal ion coordination.³⁹ The side chain of H136 donates a hydrogen bond to the Zn²⁺-bound thiolate anion. The secondary amine N4 of inhibitor 6 makes electrostatic interactions with both D94 and E274, but is not sufficiently close to consider these hydrogen bond interactions. Additionally, N4 is stabilized by a cation-π interaction with W205. Electron density for the primary amine N1 of inhibitor 6 is weaker, likely due to disorder.

Figure 8.

Polder omit map of 6 bound in active site of HDAC10 (contoured at 5.5σ). Atomic color codes are as follows: C = light blue (HDAC10) or orange (inhibitor 6), N = dark 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.

The crystal structure of the wild-type HDAC10–7 complex was determined at 2.90 Å resolution. Inhibitor 7 contains a thioester moiety, but the electron density map of the enzyme-inhibitor complex reveals that the thioester has undergone hydrolysis to yield a Zn²⁺-bound thiolate anion, i.e., the inhibitor has been chemically converted into inhibitor 6 and binds in generally similar fashion with tetrahedral coordination geometry (Figure 9). The Zn²⁺---S coordinate bond length is 2.2 Å, an ideal value for thiolate-metal coordination; however, the C–S⁻---Zn²⁺ angle of 124° and the C–C–S⁻---Zn²⁺ torsion angle of 7° deviate from ideal values.³⁹ The Zn²⁺-bound thiolate anion also accepts long hydrogen bonds from H136 and Y307.

Figure 9.

Polder omit map of 7 bound in active site of HDAC10 (contoured at 9.0σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 7), N = dark 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.

The secondary amine N4 of inhibitor 7 does not make any direct hydrogen bonds with the carboxylate side chains of D94 or E274, but it does appear to be well-positioned to make a cation-π interaction with W205. The N1 primary amine of 7 is oriented toward E274, but is not within hydrogen bonding distance.

It is interesting to note differences between the HDAC10–7 complex and the Y307F HDAC10–6 complex, even though these complexes involve identical inhibitors due to the hydrolysis of 7. These differences may arise from the Y307F substitution in the active site. For example, the hydrogen bond between the Zn²⁺-bound thiolate and H136 lengthens from 3.1 Å to 3.3 Å in the wild-type HDAC10–7 complex, possibly to optimize the additional interaction with the now-present phenolic hydroxyl group of Y307.

The crystal structure of the wild-type HDAC10–8 complex was determined at 2.75 Å resolution. Inhibitor 8 contains an unactivated ketone as well as a thioester moiety, but the electron density map of the enzyme-inhibitor complex reveals that the thioester has undergone hydrolysis to yield a Zn²⁺-bound thiolate anion with a metal coordination distance of 2.2 Å (Figure 10). Metal coordination geometry is 4-coordinate tetrahedral. While the Zn²⁺---S distance is nearly ideal, the C–S⁻---Zn²⁺ angle of 126° and the C–C–S⁻---Zn²⁺ torsion angle of 6° deviate from ideal values.³⁹ The unactivated ketone remains with an intact, trigonal planar carbonyl group; its oxygen atom is too far from the Zn²⁺ ion (3.0 Å) to make a coordination interaction. However, the ketone carbonyl oxygen accepts a hydrogen bond from Y307. This interaction seems to influence the position of the Zn²⁺-bound thiolate, which accepts a hydrogen bond from H136.

Figure 10.

Polder omit map of 8 bound in active site of HDAC10 (contoured at 7.0σ). Atomic color codes are as follows: C = cyan (HDAC10) or tan (inhibitor 8), N = dark 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.

Additional hydrogen bond interactions occur between the N4 secondary amine of 8 and the carboxylate side chain of E274. It is possible that the increased length of this inhibitor, due to the insertion of the unactivated ketone after the thioester moiety, facilitates the hydrogen bond interaction with E274.

DISCUSSION

The crystal structures described herein represent the first definitive study of inhibitor binding in the active site of HDAC10. Significantly, many of these structures illuminate key molecular features of substrate binding and catalysis. Especially intriguing are reactive substrate analogues, such as the activated trifluoromethylketone 1, unactivated methylketone 2, and unactivated methoxymethylketone 3, each of which binds as an analogue of the tetrahedral intermediate and its flanking transition states for the hydrolysis of N⁸-acetylspermidine. With IC₅₀ = 80 nM, trifluoromethylketone 1 is the best inhibitor among the ketonic substrate analogues and the second-best inhibitor of all compounds studied. Regardless of the reactivity of the carbonyl group in each ketonic substrate analogue, the tetrahedral gem diol(ate) is preferentially bound in the enzyme active site. Since the trifluoromethylketone gem-diol is present in aqueous solution at 97% based on measurements with trifluoroacetone,⁴⁰ it is possible that the pre-formed tetrahedral gem-diol binds directly from solution and displaces the Zn²⁺-bound water molecule of the resting enzyme. However, with the gem-diol being in facile equilibrium with the intact ketone carbonyl group, it is also possible that the ketone binds and undergoes nucleophilic attack by the Zn²⁺-bound water molecule. This latter possibility is more likely for the binding of unactivated ketones 2 and 3, since they exist to less than 0.2% as the gem-diol in aqueous solution based on measurements with acetone.³¹ While the measured inhibitory potency of trifluoromethylketone 1 is at least 29-fold greater than that of methylketone 2 or methoxymethylketone 3, the IC₅₀ values for the actual inhibitory gem-diol forms of 2 and 3 may be approximately 500-fold better and hence comparable to, if not better than, the IC₅₀ value of the gem-diol form of 1.

Typically, boronic acid inhibitors of other metalloenzymes such as arginase bind as tetrahedral boronate anions to mimic hydrolytic transition states,^36–38 in much the same way that ketone or aldehyde carbonyl groups undergo hydration reactions upon binding to yield tetrahedral gem-diol(ates) (Figures 4 and 5).^32–35 In binding to HDAC10, boronic acid inhibitor 4 undergoes nucleophilic attack by Zn²⁺-bound water, thereby mimicking the first step in the hydrolysis of substrate N⁸-acetylspermidine. Curiously, however, the boronic acid moiety of 4 does not always undergo nucleophilic attack to form a tetrahedral borate anion in the active sites of other deacetylases. For example, inhibitor 4 binds as the intact, trigonal planar boronic acid in its complex with acetylpolyamine amidohydrolase from the deep earth halophile Marinobacter subterrani.⁴¹

The relatively modest inhibitory potencies of 4 against these deacetylases contrasts with the submicromolar inhibitory potency of boronic acid inhibitors against HDAC3, but not other isozymes.⁴² Incorporation of a boronic acid Zn²⁺-binding group into bicyclic tetrapeptide HDAC inhibitors yields poor activity, in contrast with bicyclic tetrapeptides derivatized with a methoxymethylketone Zn²⁺-binding group.⁴³ Use of the boronic acid moiety as a zinc-binding group in deacetylase inhibitors may thus serve as a route toward achieving isozyme selectivity.

Interestingly, the structure of the HDAC10–8 complex reveals that the unactivated ketone moiety remains trigonal planar and does not react with a Zn²⁺-bound solvent molecule to yield a gem-diol. Instead, the Zn²⁺ ion is coordinated by the thiolate group liberated by thioester hydrolysis. This could indicate that the thiolate moiety is preferred as a Zn²⁺-binding group over the gem-diol(ate) moiety. It is intriguing to speculate that the enzyme may have catalyzed the thioester hydrolysis reaction with inhibitors 7 and 8 to generate the inhibitory thiol in situ. Thioester hydrolysis similarly activates inhibition of HDAC isozymes by the marine natural product Largazole, which binds to HDAC8 through a thiolate-zinc coordination interaction with nearly perfect geometry.^44,45

Of particular note, the most potent inhibitory activity is measured for thioester 7 (IC₅₀ = 30 nM), which undergoes hydrolysis to yield a Zn²⁺-bound thiolate anion (Figure 9). The IC₅₀ reported here, however, is likely only an upper limit to the inhibitory potency due to the concentrations of enzyme required to perform the assay. This result suggests that a thiol zinc-binding group is superior to the other zinc-binding groups represented in the array of inhibitors shown in Figure 2. Since the inhibitory form of 7 is identical to inhibitor 6 (IC₅₀ = 370 nM), what accounts for the 12-fold difference in potencies between these two inhibitors? Mass spectrometry reveals that the free thiol group of 6 is susceptible to oxidation and begins to forms a disulfide in aqueous solution, thereby diminishing the concentration of the inhibiting species and accounting for the higher IC₅₀ value. In contrast, compound 7 is essentially a prodrug, stable in aqueous solution as determined by mass spectrometry, but upon incubation with the enzyme undergoes hydrolysis to liberate the free thiol. Thus, the free thiol generated from 7 is not subject to oxidation for a long period of time prior to binding to the enzyme.

Given the array of different Zn²⁺-binding groups in this series of enzyme-inhibitor complexes, it is instructive to compare overall metal ion coordination geometries. For example, consider thiolate Zn²⁺-binding groups that consistently bind with tetrahedral coordination geometry. Thiolate-Zn²⁺ coordination displaces the Zn²⁺-bound solvent molecule expected to be present in the resting enzyme (based on analogy with the related class IIb enzyme HDAC6⁴⁶). In contrast, chelation of the catalytic Zn²⁺ ion by the ionized hydroxamate moiety of inhibitor 5, the gem-diol(ate) moieties of inhibitors 1–3, or the tetrahedral boronate anion of 4 yields a 5- or 6-coordinate metal ion (Figure 11a).

Figure 11.

(a) Superposition of HDAC10 complexes with inhibitors 1–4. (b) Superposition of HDAC10 complexes with inhibitors 2 and 5. (c) Superposition of HDAC10 complexes with inhibitor 6 and the free thiolate forms of inhibitor 7 and 8.

Closer analysis of Zn²⁺ coordination in the complexes with inhibitors 1–4 reveals that inhibitors 2 and 4 bind to Zn²⁺ similarly, resulting in mixed square pyramidal-trigonal bipyramidal complexes, while inhibitors 1 and 4 form 6-coordinate complexes. A 6-coordinate metal ion results when coordination by D174 shifts from monodentate to bidentate. Notably, inhibitors 1, 2, and 4 all bind to Zn²⁺ in similar positions, whereas inhibitor 3 is slightly shifted. This shift is likely due to steric interactions between the methoxy group of 3 and the bottom of the active site pocket.

The Zn²⁺ coordination geometry is different when the metal binding group is a hydroxamate (Figure 11b). Although inhibitors 2 and 5 both form 5-coordinate Zn²⁺ complexes, the metal coordination geometry in the HDAC10–5 complex is slightly more square pyramidal. In contrast, each Zn²⁺ complex involving thio-analogues 6–8 adopts nearly identical 4-coordinate tetrahedral geometry (Figure 11c). Identification of the geometry of each metal binding group is important for understanding inhibitor structure-affinity relationships, which in turn will inform the design of second-generation HDAC10 inhibitors.

Additional structural determinants of affinity in the active site of HDAC10 include the negatively charged carboxylate side chain of E274, which serves as a “gatekeeper” to ensure catalytic specificity for the binding of the cationic substrate N⁸-acetylspermidine.¹⁵ Substitution of E274 with leucine, as it appears in most other HDAC isozymes, attenuates polyamine deacetylase activity and increases lysine deacetylase activity.¹⁵ In the array of crystal structures presented herein, E274 is close to the secondary amino group of N⁸-acetylspermidine, but it is not always within hydrogen bonding distance. Indeed, in the structure of the HDAC10–4 complex, E274 forms a hydrogen bond with the primary amino group of the inhibitor. Thus, the structural basis of polyamine substrate selectivity conferred by E274 may derive more from the general electrostatic effect of this residue rather than a specific hydrogen bond with the substrate. When hydrogen bonding does occur between E274 and substrate analogues, it can occur with the secondary amino group (complexes with 5 and 8) or the primary amino group (complexes with 1 and 4). However, the primary amino group is typically characterized by weak or missing electron density, so its apparent disorder indicates that is not firmly anchored by any particular interaction.

Also present in the active site of zebrafish HDAC10 is D94, and this residue forms hydrogen bonds with the primary amino groups of most inhibitors studied. However, HDAC10 sequence alignments show that D94 of zebrafish HDAC10 aligns with an alanine residue in the human enzyme, so interactions observed with D94 cannot be targeted for inhibition of the human enzyme. Moreover, D94A HDAC10 exhibits near-normal catalytic activity,¹⁵ so D94 does not make catalytically important interactions with substrate N⁸-acetylspermidine.

CONCLUSIONS

Zebrafish HDAC10 serves as a readily studied surrogate of human HDAC10, particularly with regard to the X-ray crystallographic study of enzyme-inhibitor complexes. Structure-affinity relationships presented herein indicate that thiolate and trifluoromethylketone gem-diolate Zn²⁺-binding groups are most optimal for metal ion coordination as incorporated into the N⁸-acetylspermidine analogues shown in Figure 2. Additionally, these functional groups also engage in hydrogen bond interactions with catalytic residues H136, H137, and/or Y307. Thus, both metal ion coordination and hydrogen bond interactions with the zinc-binding groups are key to the design of tight-binding inhibitors.

Of particular interest are the ketonic substrate analogues of N⁸-acetylspermidine, which undergo the chemistry of hydration in binding to the catalytic Zn²⁺ ion regardless of whether the ketone carbonyl is activated by a trifluoromethyl group. The binding of these analogues requires identical chemistry to that required for catalysis, i.e., the activation of a Zn²⁺-bound water molecule for nucleophilic attack at a carbonyl group. The resulting tetrahedral gem-diol(ate) is a nearly perfect mimic of the tetrahedral transition state and its flanking intermediates in catalysis; thus, the binding of the gem-diol(ate) reflects the preferential binding of the transition state in the HDAC10 active site. It follows that Zn²⁺, H136, H137, and Y307 participate in transition state stabilization, which is a principal feature of catalysis in HDAC10 and other metal-dependent deacetylases.

Finally, since the negatively charged side chain of E274 and the 3₁₀ helix that constrains the active site (the P(E,A)CE motif) are unique to HDAC10 orthologues, these residues can be targeted for intermolecular interactions that will confer HDAC10 selectivity in future inhibitor designs. Indeed, recent inhibitor designs targeting interactions with E274 have yielded nanomolar potencies with up to 43-fold selectivity for inhibition of HDAC10.⁴⁷ Our future studies will probe the importance of these interactions as well as others in the active site as designs of isozyme-selective HDAC10 inhibitors are pursued.