Potent, Selective, and CNS-Penetrant Tetrasubstituted Cyclopropane Class IIa Histone Deacetylase (HDAC) Inhibitors

Abstract

Potent and selective class IIa HDAC tetrasubstituted cyclopropane hydroxamic acid inhibitors were identified with high oral bioavailability that exhibited good brain and muscle exposure. Compound 14 displayed suitable properties for assessment of the impact of class IIa HDAC catalytic site inhibition in preclinical disease models.

Inhibition of class IIa HDAC enzymes has been suggested as a therapeutic strategy for a number of indications, including Huntington’s disease (HD) and muscular atrophy. Class IIa HDACs are large proteins with multiple functions including transcription factor binding and N-acetyl lysine recognition.^1,2 Of most interest to our laboratory is the role of class IIa HDAC biology in HD, in particular the beneficial effect, which has been observed following HDAC4 genetic suppression.³⁻⁵ Replication of these effects in preclinical models of HD via occupancy of the class IIa HDAC catalytic domain would provide a rationale for small molecule therapy. Currently there are no marketed HDAC class IIa-selective inhibitors, whereas four pan-HDAC inhibitors, vorinostat (SAHA), romidepsin, belinostat, and panobinostat are on the market.

Class IIa-selective HDAC inhibitors would represent important tools for elucidating the therapeutic potential of this protein family. We recently reported the structure-based design of trisubstituted cyclopropane class IIa-selective HDAC inhibitors as potential therapeutics in HD.⁶ This improved selectivity was driven by exploiting a selectivity pocket (Figure 1, shown with compound 13) that is not present in the class I HDAC isoforms. This pocket is formed as a consequence of a tyrosine-histidine substitution.⁷

Figure 1.

X-ray structure of HDAC4 cd (class IIa HDAC) with trisubstituted cyclopropane 13 (4CBT)⁶ showing the presence of a selectivity pocket due to a Tyr-His substitution. Residue numbering from PDB files.

We now report the discovery of tetrasubstituted cyclopropane hydroxamic acid class IIa HDAC inhibitors, with additional substitution at C1 (Figure 1). These compounds exhibited improved pharmacokinetic profiles, and so may provide a further means for evaluating efficacy in preclinical in vivo HD disease models.

Determination of the biochemical and Jurkat E6.1 cell potencies of compounds against the HDAC isoforms has been described previously, employing artificial substrates Boc-Lys(TFA)-AMC (class IIa/HDAC8-specific) and Boc-Lys(Ac)-AMC (class I/IIb-specific).⁶⁻¹⁰ The majority of class IIa HDAC activity in the Jurkat E6.1 cell line is derived from HDAC4.⁶ Biochemical HDAC isoform selectivity was compared to the difference in activity in the cell assays between the Lys-Ac (class I/IIb-specific) and the Lys-TFA (class IIa/HDAC8-specific) substrates.

In the trisubstituted cyclopropane series reported previously, the most potent compounds in the biochemical and cell assays, albeit with a significant drop-off in cell potency, had comprised either pyrimidine or oxazole capping groups.⁶ Further improvements in biochemical potency and pharmacokinetics were targeted. The design of these molecules was inspired by the expectation that a fluorine atom alpha to the hydroxamic acids may increase their acidity. It has previously been proposed that fluorination of known HDAC inhibitors at the carbon bearing the hydroxamic acid moiety enhanced HDAC biochemical activity.¹¹ Analysis of the X-ray structures of the trisubstituted cyclopropanes, and docking studies for the new tetrasubstituted cores indicated that there was sufficient space for small substituents to be accommodated at the C1-cyclopropane position.

Trisubstituted cyclopropane hydroxamic acids previously profiled in mouse hepatocytes displayed high intrinsic clearance values.⁶ It was assumed that direct glucuronidation would be their principal route of clearance, presumably via O-glucuronidation of the hydroxamic acid, and that the contribution of oxidative metabolism to the overall clearance of these molecules would be small. Compounds with very short half-lives (<5 min) in mouse liver microsomes were not progressed. Substitution of the cyclopropane carbon atom bearing the hydroxamic acid moiety (C1), with either a methyl group or a halogen might limit, either via steric or electronic means, the glucuronidation process. Some of the previously reported trisubstituted cyclopropanes had demonstrated high passive permeability and low P-gp efflux in MDR1-MDCK monolayers. We anticipated that reduced clearance, while maintaining high passive permeability and low P-gp efflux, would lead to higher plasma and CNS exposures. The C1-substituent on the cyclopropane ring was selected to maintain the favorable CNS-compliant physicochemical properties of the compounds.

The effect of C1-substitution on potency was investigated initially with compounds comprising the 4-methyl pyrimidin-2-yl capping group (Table 1). Methyl and fluoro substitution at C1 (10 and 12) conferred an improvement in HDAC4 biochemical activity over trisubstituted cyclopropane 9, while the potency values for the chloro analogue 11 and compound 9 were similar. In the cellular HDAC class IIa assay, fluoro derivative 12 exhibited a significant potency enhancement over compounds 9, 10, and 11, with no improvement in cellular activity for methyl substituted 10versus9. All of the 1-fluoro derivatives (12, 14, 16, 18, 20, and 22) demonstrated a consistent positive effect on biochemical and cell activities when compared to the trisubstituted cyclopropanes (9, 13, 15, 17, 19, and 21).

Table 1. Impact of Cyclopropane Substitution Alpha to the Hydroxamic Acid upon Potency and in Vitro ADME Properties.

^aArithmetic mean and standard error of ≥3 measurements.

^bIntrinsic clearance (Cl_int) values of >257 mL/min/kg in mouse liver microsomes (MLM) and >52 mL/min/kg for human liver microsomes (HLM) indicate a rapid rate of oxidative metabolism. Under the assay conditions used, Cl_int values <65 mL/min/kg in MLM and <36 mL/min/kg in HLM indicate low rate of metabolism by CYP450 enzymes.

^cBased upon previous work in our laboratory, an effective efflux ratio (EER) value of >4 suggests that a compound is a substrate for P-gp, whereas values <2 suggest the compound is unlikely to be a P-gp substrate. For delivery to the CNS, a drug should ideally have an in vitro passive permeability >150 nm/s and should not be a good P-gp substrate (B to A/A to B ratio <2.5).¹²

^dPoor A to B permeability in both cell types. Low A to B mass balance.

The opposite enantiomer of 14, prepared via the synthetic route in Scheme 1, was 40-fold less potent in the HDAC4 biochemical and cell assays (compound 26, see Supporting Information).

Scheme 1. Synthesis of 1-Substituted Cyclopropanes and Alternative Route Towards Compound 14.

Reagents and conditions: (a) 12-crown-4, LiHMDS, DCM, −20 °C, 1 h; (b) LDA in THF, −78 °C, 30 min then MeI or CCl₄ or NFSI, 2 h; (c) (i) bis(pinacolato)diboron, KOAc, Pd(dppf)Cl₂, dioxane, 90 °C, 4 h; (ii) heteroaryl halide, Pd(dppf)Cl₂, CsF, dioxane, 100 °C, 17 h; (d) NH₂OH 50% aq, KOH, THF/MeOH (1:1), r.t., 16 h, chiral HPLC; (e) LDA in THF, −78 °C, 30 min then NFSI, 2 h, (rac)-5 separated by flash chromatography; (f) SFC chiral chromatography; (g) (i) bis(pinacolato)-diboron, KOAc, Pd(dppf)Cl₂, dioxane, 90 °C, 4 h; (ii) 2-chloro-5-fluoropyrimidine, Pd(dppf)Cl₂, CsF, dioxane, 100 °C, 17 h; (h) NH₂OH 50% aq, KOH, THF:MeOH (1:1), r.t., 16 h.

The tetrasubstituted cyclopropanes were assessed for in vitro ADME properties (Table 1). As an indication of oxidative metabolism, compounds 16 and 22 were more stable in mouse liver microsomes (Cl_int values of <65 and <67 mL/min/kg BW) than 14 (306 mL/min/kg BW). A mouse microsomal preparation incubated with compound 14 showed the presence of a metabolite that was assigned as the 5-hydroxypyrimidin-2-yl derivative, which may, at least in part, account for the increased microsomal turnover compared to 16 or 22. Compound 18 was moderately stable in mouse liver microsomes but showed poor solubility and poor passive permeability in MDCK monolayers, while 20 also exhibited low passive permeability.

In mouse hepatocytes, all compounds exhibited high intrinsic clearance values (data not shown).

C1-substitution of 9 with fluorine, chlorine, or a methyl group retained the low P-gp efflux (measured in MDR1-MDCK cell monolayers) characteristic of the parent trisubstituted cyclopropanes, with moderate to good passive permeability across wild type MDCK cells. The methyl analogue 10 also displayed some improvement in mouse microsomal stability.

A direct measure of compound binding to the target was determined by surface plasmon resonance (SPR) where the HDAC4 catalytic domain was immobilized via amine coupling to the sensor surface (see Supporting Information). At steady state, the K_D for compound 14 was determined to be 0.011 μM (Table 2), with a k_on of 196800/M·s and a k_off of 0.0044/s, i.e., moderate on and off rates. The corresponding des-fluoro compound 13 displayed lower affinity, consistent with the biochemical assay, with a K_D value of 0.118 μM, driven by larger k_off of 0.0186/s (the k_on value was 190270/M·s).

Table 2. Comparison of SPR K_D Values with Biochemical Activity Values for Compounds 13, 14, and 16.

entry	biochemical HDAC4 IC50 (μM)a	SPR KD (μM)a
13	0.05 ± 0.02	0.11 ± 0.01
14	0.01 ± 0.001	0.02 ± 0.01
16	0.03 ± 0.01	0.08 ± 0.06

^aArithmetic mean and standard error of ≥3 measurements.

An X-ray cocrystal structure of 14 with the HDAC4 cd (L728A mutant)⁶ was obtained confirming the absolute configuration and demonstrating the key interactions of the hydroxamate molecule with the protein binding site. The binding mode of 14 was similar to that of trisubstituted cyclopropane 13 (Figure 2B). Comparison with the class I HDACs (Figure 2A, with SAHA) shows the tyrosine residue of the class I HDACs projecting into the region of the selectivity pocket in the class IIa isoforms formed by the tyrosine-histidine switch.

Figure 2.

(A) X-ray structure of HDAC2 (class I HDAC) with SAHA (PDB code: 4LXZ).¹³ (B) Protein structure superposition of HDAC4 cd X-ray structures with compounds 13 and 14 (PDB codes: 4CBT and 5A2S, respectively). (C) Enlarged region of the HDAC4 active site with compound 14. Residue numbering from PDB files.

The previously reported trisubstituted cyclopropanes are selective for the class IIa versus class I and IIb HDAC isoforms. HDAC isoform biochemical selectivity was investigated with the Boc-Lys(TFA)-AMC and Boc-Lys(Ac)-AMC substrates (vide supra). The potency values against each of the HDAC isoforms for the trisubstituted cyclopropane 13 and its 1-fluoro derivative 14 are shown in Table 3.

Table 3. HDAC Isoform Biochemical Activity Data^a for Compounds 13 and 14.

	class IIa (cd)				class I				class IIb
entry	HDAC4	HDAC5	HDAC7	HDAC9	HDAC1	HDAC2	HDAC3b	HDAC8	HDAC6
13	0.05 ± 0.02	0.03 ± 0.01	0.12 ± 0.06	0.20 ± 0.09	21c	ND	9.7 ± 2.2	1.4 ± 0.2	1.9 ± 0.4
14	0.01 ± 0.001	0.01 ± 0.004	0.03 ± 0.01	0.06 ± 0.03	14 ± 4	>50	7.4 ± 2.1	0.28 ± 0.03	3.1 ± 0.4d

^aInhibition IC₅₀ values quoted in μM. Arithmetic mean and standard error of ≥3 measurements.

^bHDAC3-NCoR2.

^cLess than three independent replicates.

^dDetermined from a separate assay utilizing HDAC6 overexpression in a HEK cell line, due to a lack of availability of active HDAC6 protein. Lysates from these overexpressing cells demonstrated activity that was HDAC6 expression-dependent with only a small background signal due to endogenous HDAC activity.

Compound 14 maintained a good class IIa HDAC selectivity profile, demonstrating, for example, at least 100-fold selectivity over HDACs 1, 2, and 3.

Compounds 14 and 16 were progressed to PK studies on account of their high potency, MDCK data, and kinetic solubility. At the time of writing, compound 22 had not yet been advanced to further studies.

Trisubstituted cyclopropane 13 and fluoro cyclopropanes 14 and 16 were dosed intravenously and via oral gavage in fed male C57Bl/6 mice. Following an oral dose, all three compounds were rapidly absorbed (Figure 3).

Figure 3.

Mouse plasma and brain pharmacokinetic profiles for compounds 13, 14, and 16 post-oral dose (10 mg/kg).

Compounds 14 and 16, incorporating fluoro- or difluoromethoxy pyrimidine capping group substituents, respectively, demonstrated higher distribution to brain tissue versus the trisubstituted cyclopropane 13. All three compounds exhibited biphasic elimination profiles and high oral bioavailability (Table 4). In addition to high volumes of distribution, these compounds displayed high plasma clearance, as expected from the mouse hepatocyte data. Following oral dosing only, a broad secondary peak was observed for 14 and 16. This may be a result of slow dissolution of compound following gavage.

Table 4. Pharmacokinetic Parameters of 13, 14, and 16 Following Administration to Fed Male C57Bl/6 Mice.

		poa				ivb
entry	compartment	F (%)	AUCnorm (nM·h·kg/mg)	Cmax norm (nM·kg/mg)	tmax (h)	Clp (L/h/kg)	Vdss (L/kg)	t1/2 (h)
13	plasma	∼100	110c	97	0.5	11	4.1	0.3
	brain	ND	830	570	0.5	ND	ND	ND
	muscle	ND	330	320	0.5	ND	ND	ND
14	plasma	79	480	180	0.25	4.5	5.1	2.2
	brain	ND	1900	780	0.5	ND	ND	ND
	muscle	ND	640	290	0.5	ND	ND	ND
16	plasma	∼100	660	150	0.25	4.4	10	3
	brain	ND	580	120	1	ND	ND	ND
	muscle	ND	300	56	0.5	ND	ND	ND

^a13, 14, and 16 dosed at 10 mg/kg.

^b13, 14, and 16 dosed at 5, 1, and 2.5 mg/kg, respectively.

^cCalculated using AUC_last.

Compound 14 demonstrated a linear increase in exposure with dose in both plasma and brain matrices across 10, 30, and 100 mg/kg oral doses.

From HDAC4 SPR studies (vide supra) the dissociation half-life for compound 14 is approximately 160 s, considerably shorter than its in vivo half-life of ∼2 h. Binding kinetics are therefore unlikely to impact the duration of any pharmacodynamic effect of 14.

In the absence of known intracellular concentrations of compound in the brain, we considered that the best estimate of compound available to interact with the target is likely to be the unbound fraction of compound that may be determined either by microdialysis, CSF exposure, or correction of the total exposure from in vitro equilibrium dialysis.

Unbound brain concentrations of 14 were estimated by correcting total brain concentrations for the fraction unbound (0.41%) determined in mouse brain homogenate by equilibrium dialysis. CSF concentrations in mouse were estimated using the CSF to brain ratio determined in rat, assuming this remained constant across species. These gave very similar estimates of concentration over the time points (at 10 mg/kg po dose the unbound C_max in mouse was estimated to be at approximately the cell IC₅₀ value).

Compounds were synthesized according to Scheme 1. Cyclopropanation of the methyl cinnamate derived from 1 as previously described,⁶ followed by deprotonation of the racemic intermediate 2 using LDA (1.1–3 equiv) and quenching with an appropriate electrophile (MeI, CCl₄ or NFSI)¹⁴ afforded the tetrasubstituted cyclopropane. The desired isomer was isolated by flash chromatography. The heterocyclic capping groups were introduced by Suzuki coupling and the enantiopure hydroxamic acid products were obtained from chiral HPLC purification.

In order to access larger quantities of 1-fluoro cyclopropane 14, a modification of the general route was employed (Scheme 1, lower branch). Herein, the bromophenyl fluoro precursor (7) was obtained as the single desired stereoisomer. After the key electrophilic fluorination reaction of the cyclopropane ester (rac)-2, a 1:2 diastereomeric ratio of epimers (rac)-5 and (rac)-6, respectively, was achieved using THF and LDA (3 equiv) at −78 °C with NFSI (N-fluorobenzenesulfonimide), on a multigram scale. This resulted in a 25% isolated yield of (rac)-5 following separation by flash chromatography. The addition of LiCl, previously reported to increase the yields of 1-fluorination of cyclopropyl esters,¹⁵ was discovered to favor further the formation of the undesired diastereoisomer (rac)-6. The enantiomerically pure methyl ester 7 was isolated by chiral supercritical fluid chromatography (SFC). Subsequently the heterocyclic capping group was introduced by Suzuki coupling, followed by hydroxamic acid formation to deliver compound 14.

In conclusion, we have discovered potent and selective class IIa HDAC hydroxamic acid inhibitors comprising a tetrasubstituted cyclopropane scaffold. Compounds such as 14 displayed high oral bioavailability with good brain and muscle exposure. Such compounds exhibit suitable properties for assessment of the impact of class IIa HDAC catalytic site inhibition in preclinical HD disease models.

Glossary

ABBREVIATIONS

A

apical

B

basolateral

SAHA

suberoylanilide hydroxamic acid

cd

catalytic domain

ND

not determined

po

per os, oral administration

iv

intravenous route of administration

CSF

cerebrospinal fluid

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.5b00302.

• Additional compound data, synthesis procedures and characterization of compounds, quantification of compound binding to HDAC4 cd using SPR, X-ray crystallography data collection and refinement statistics, and rat pharmacokinetics (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.