Discovery of Reversible Inhibitors of KDM1A Efficacious in Acute Myeloid Leukemia Models

Abstract

Lysine-specific demethylase 1 (LSD1 or KDM1A) is a FAD-dependent enzyme that acts as a transcription corepressor or coactivator by regulating the methylation status of histone H3 lysines K4 and K9, respectively. KDM1A represents an attractive target for cancer therapy. While, in the past, the main medicinal chemistry strategy toward KDM1A inhibition was based on the optimization of ligands that irreversibly bind the FAD cofactor within the enzyme catalytic site, we and others have also identified reversible inhibitors. Herein we reported the discovery of 5-imidazolylthieno[3,2-b]pyrroles, a new series of KDM1A inhibitors endowed with picomolar inhibitory potency, active in cells and efficacious after oral administration in murine leukemia models.

Histone lysine demethylases are a class of enzymes involved in the epigenetic control of gene expression by regulating, in concert with their counterpart histone lysine methylases, the equilibrium of histone lysine methylation state.^1,2

Several diseases, including cancer, can be related to the aberrant regulation of transcription that is induced by an unbalanced histone lysine methylation. Consequently, many of the enzymes involved in histone lysine methylation/demethylation are being studied as potential drug targets.¹

The lysine specific demethylase 1 (LSD1/KDM1A), which catalyzes the demethylation of lysine residues of histones H3K4me1/2 and H3K9me1/2, has emerged as a target of particular interest over the past years for its key role in tumor cell growth and survival. Several lines of evidence, which include overexpression in solid and hematological malignancies^3,4 as well as its correlation in certain tumor types with poor prognosis,⁵ indicate KDM1A as a promising anticancer target.

KDM1A is a flavin adenine dinucleotide (FAD) dependent histone demethylase with high structural homology with monoamine oxidase A (MAOA). This brought several groups to investigate the activity of MAO inhibitors and to optimize their potency toward KDM1A inhibition. In particular, a plethora of tranylcypromine (TCPA) based derivatives have been studied as mechanism-based, irreversible inhibitor of KDM1A. This strategy led to the development of several clinical candidates (IMG-7289, GSK 2879552, ORY1001/RG6016, and INCB059872)⁶ that share the 2-phenylcyclopropanamine key pharmacophore, which irreversibly binds to the flavin adenine dinucleotide (FAD) moiety within the KDM1A catalytic cavity.

The interest in finding reversible KDM1A inhibitors has also recently grown, to investigate if any advantage can be offered by the efficacy/safety profile of a noncovalent inhibitor. Several potent inhibitors, often endowed with high in vitro potency and cellular activities, have been disclosed,⁷⁻¹¹ and two of them demonstrated in vivo efficacy.^12,13 Ultimately a clinical candidate has been identified by Celgene: CC-90011.⁶ We also contributed to the field by reporting the identification of thieno[3,2-b]pyrrole-5-carboxamides as reversible inhibitors of KDM1A.^14,15 To gain insight into the role of the amide moiety of these derivatives in the affinity for KDM1A, we have explored its replacement with different residues. We here report on the identification and early expansion of a series of 5-imidazolylthieno[3,2-b]pyrroles as a new chemical class of reversible inhibitors.

This work led to the discovery of KDM1A inhibitors endowed with in vitro picomolar potency accompanied by a remarkable anticlonogenic effect on MLL-AF9 human leukemia cells. Eventually, the representative compound 15b demonstrated significant efficacy in two different murine leukemia models after oral administration at well-tolerated doses.

We initially evaluated a series of simple analogs bearing classical replacements of the benzamido moiety present in 1(14) (compounds 2–7, Table 1). Among them, only the phenylimidazole 7 showed a promising activity; thus we focused on the optimization of the imidazole series. The SAR study on compound 7 started with a preliminary exploration at position 4 of the imidazole ring. As the initial replacement of the phenyl ring in 7 with an ethyl residue (compound 9, Table 1) led to a significantly improved potency, we examined the effect of other small alkyl groups obtaining a set of compounds more potent than 9 (compounds 10–13, Table 1). For the next optimization step we postulated that these compounds could interact with KDM1A maintaining the same orientation of the previous carboxamido series. According to this hypothesis, we introduced substituents on the nitrogen of the imidazole ring following the SAR and structural information already available from our previous work.^14,15

Table 1. KDM1A Inhibitory Activity of 4-Methylthieno[3,2-b]pyrrole Derivatives 1–13^a.

^aThe synthesis of compounds 2–13 are reported in Schemes S1 and S2 in Suppoting Information.

^bData are expressed as the mean of at least four determinations with standard deviation.

To this end several chains characterized by the presence of a terminal secondary cyclic amine were linked to the imidazole ring through different spacers. This functionalization, carried out on our first lead compound 9 together with the most potent derivatives 10 and 13, strongly increased the inhibitory potency leading to compounds with IC₅₀ in the picomolar range (Table 2). The corresponding 1,5′ disubstituted derivatives, isolated as minor isomers, consistently showed 10-fold weaker activity (complete data in the Supporting Information, Table S1) and were not further evaluated. Noteworthy, the lack of the terminal amine, as well as the removal of the phenoxyethyl linker, led to a drop in potency by 1 order of magnitude (compounds 17 and 18 respectively in Table 2).

Table 2. KDM1A Inhibitory Activity of N-Substituted Imidazole Derivatives 14a–f, 15a–f, 16a–f, 17, 18^a.

^aAll the compounds have been prepared according Scheme S2 (Supporting Information).

^bData are expressed as the mean of at least four determinations with standard deviation.

In the course of this work the crystallographic structures of several inhibitors in complex with KDM1A were also solved and confirmed the initial working hypothesis.

Analogous to the previous carboxamido derivatives, this novel series also adopts a packed “U shaped”¹⁵ conformation within the catalytic site of KDM1A, where key interactions are (i) hydrophobic contacts between the thienopyrrole moiety and Val333, Thr335, Leu659, Tyr761 residues; (ii) π–π parallel displaced contact between the thienopyrrole and flavin rings; (iii) a salt-bridge interaction between Asp555 residue and the positively charged amino tail; (iv) the alkyl substituents on the imidazole ring nicely protruding into a subpocket defined by the Phe538, Leu659, Lys661, and Trp695 side chains (Figure 1). In particular, the distance between rings in the π–π stacking between the thienopyrrole of the inhibitor and the flavin ring of the FAD is 3.9 Å, while the one between the heterocyclic and the aryl ring of the imidazole side chain in the inhibitor is 4.4 Å (measured with the “distances to rings” tool available in PyMOL version 2.1.0).

Figure 1.

(A) X-ray crystal structure of compound 15f (orange sticks) in complex with KDM1A (white cartoon and sticks). Highlighted with cyan sticks is the FAD cofactor (Table S1 in Suppoting Information, PDB code 6TE1). 15f adopts the “U shaped” conformation within the catalytic site of KDM1A. The π–π parallel displaced contact between the FAD and thienopyrrole rings is highlighted with a black dashed line. (B) Image of the subpocket hosting the cyclobutyl substituent on the imidazole ring. KDM1A is depicted as gray surface, 15f as orange sticks, and FAD cofactor as cyan sticks. A second molecule of inhibitor 15f is also found in the crystal structure, binding in a less conventional way, juxtaposed to the described inhibitor, with its positively charged amino tail stacking to the one of the conventional inhibitor, without any contact with the FAD cofactor (not shown).

All the compounds reported in Table 2 were then tested on human leukemia cell line THP-1 to investigate their ability to engage and inhibit KDM1A in cell. Specifically, we decided to follow the induction of CD86 expression, a myeloid differentiation marker, as a surrogate cellular biomarker for pharmacological inhibition of the histone lysine-specific demethylase 1A.¹⁵ THP-1 cells were exposed to the compounds for 48 h at the concentration of 5 and 50 nM, and expression of CD86 was followed by flow cytometry. DDP-38003 (compound 15¹⁶), used as positive control, induced an increase of CD86 expression of 5.1 ± 1.3 and 41.2 ± 2.8 at doses of 5 and 50 nM, respectively.

As expected for KDM1A inhibitors, the compounds strongly and dose dependently induced an increase of leukemia cells expressing CD86 and, consequently, induced an increase of differentiating cells (Table 3).

Table 3. Target Modulation.

	CD86 fold increasea			CD86 fold increasea
compd	5 nM	50 nM	compd	5 nM	50 nM
14a	4.7 ± 0.1	27.2 ± 4.3	15d	2.7 ± 0.1	25.8 ± 0.9
14b	6.9 ± 0.1	39.1 ± 14	15e	7.3 ± 1.5	28.1 ± 1.4
14c	2.4 ± 0.1	27.8 ± 3.7	15f	17.9 ± 7.8	57.7 ± 2.5
14d	1.5 ± 0.3	22.0 ± 0.3	16a	29.0 ± 2.0	73.7 ± 8.9
14e	1.5 ± 0.6	19.6 ± 1.8	16b	4.8 ± 0. 5	28.8 ± 7.8
14f	10.9 ± 1.2	51.9 ± 0.6	16c	1.7 ± 0.4	24.5 ± 3.1
15a	22.5 ± 2.1	50.3 ± 2.4	16d	2.6 ± 0.03	22.6 ± 2.4
15b	19.1 ± 0.1	56.8 ± 3.2	16e	6.6 ± 0.1	22.7 ± 2.03
15c	4.4 ± 0.3	33.5 ± 3.4	16f	3.8 ± 0. 4	29.7 ± 2.9

^aData are expressed as fold increase compared to the vehicle (DMSO) and reported as the mean value of at least two biological replicates, measured after 48 h of treatment, with standard deviation.

The most potent compounds (14b, 14f, 15a, 15b, 15f, and 16a) were then characterized in terms of biochemical selectivity by assessing their activity on MAOA and MAOB enzymes, which are strongly structurally correlated to KDM1A. Strikingly, all these inhibitors showed a remarkable selectivity toward KDM1A vs MAOs with an average selectivity fold higher than 100 (Table S3, Supporting Information).

On the basis of the relevance of KDM1A in sustaining the oncogenic potential of MLL-AF9 leukemia stem cell,¹⁷ we evaluated the therapeutic potential of selected inhibitors by determining their ability to inhibit the colony forming capacity of human MLL-AF9 leukemia cell line (THP-1). DDP-38003 was used as positive control determining an inhibition of 62% at the dose of 500 nM.

As reported in Table 4, the molecules tested at the doses of 50 and 500 nM showed a dose dependent reduction of colony forming potential of human leukemia THP-1 cells. Several of the most potent compounds showed an in vitro overall favorable metabolic stability in both human and mouse microsomes (Table S4, Supporting Information). Among them 15b was chosen as an early representative compound for determination of pharmacokinetic properties.

Table 4. Cellular Data.

compd	colony formation, % inhib (50 nM)a	colony formation, % inhib (500 nM)a
14b	33.3 ± 10.7	61.4 ± 1.9
14f	27.6 ± 9.0	72.3 ± 4.5
15a	40.9 ± 5.3	74.2 ± 5.4
15b	38.8 ± 15.0	72.3 ± 8.6
15f	33b	73.3 ± 1.5
16a	46.6 ± 1.3	64.2 ± 6.0

^aColonies were counted after 13 days. Percentage of inhibition is referenced to the vehicle (DMSO) treated cells, and the data are reported as the mean value of at least two biological replicates with standard deviation.

^bSingle determination.

As summarized in Table 5, the pharmacokinetic profile of 15b turned out to be compatible with oral testing in an in vivo efficacy experiment.

Table 5. Pharmacokinetic Parameters of 15b after Single iv and os Administration in CD-1 Mice^a.

iv, 5 mg/kg		os, 15 mg/kg
Cmax (μM)	2.54 ± 0.48	Cmax (μM)	0.32 ± 0.04
AUC0–∞ (μM h)	4.2 ± 0.5	AUC0–∞ (μM h)	2.03 ± 0.5
CL (mL min–1 kg–1)	37.5 ± 7.4	tmax (min)	25 ± 9
Vss (L/kg)	10.2 ± 2.7	F (%)	17
t1/2 (h)	6 ± 1.4

^a15b was administrated to CD-1 mice in single intravenous (iv) or oral (os) doses of 5 or 15 mg/kg, respectively. The compound was dissolved in 5% glucose solution containing 10% Tween 80 for the iv dose or in 5% glucose solution containing 40% PEG for the oral dose. Data are reported as the mean value plus and minus standard deviation (n = 3). C_max: maximal concentration, AUC_0–∞: area under the curve. CL: clearance. V_ss: volume of distribution at steady state. t_1/2: half-life. t_max: time to reach maximal concentration. F: oral bioavailability.

Following, the antitumor activity of 15b was assessed in two leukemia models representative of acute promyelocytic leukemia driven (APL) by PML/RAR translocation and by acute myeloid leukemia (AML) driven by MLL-AF9 translocation.

On the basis of the results of the preliminary PK experiment and of the potency in the cellular assays, 15b was initially tested in the APL model at the doses of 15, 30, and 60 mg/kg for 5 days per week for 2 weeks. The administration started once leukemia cells were detected in recipient mice (around 10 days after leukemia blast injection). The survival of mice of the different experimental groups was analyzed and represented by a Kaplan–Meier survival plot. No significant body weight differences were observed during the treatment with the exception of the dose of 60 mg/kg (data not shown). At this dose the treatment was halted at the end of the first week as mice started to seriously lose body weight. Considering the safe doses of 15 and 30 mg/kg, as reported in Figure 2A, in the APL model, the compound significantly and dose dependently prolonged the survival of the mice of 20 and 55% at the doses of 15 and 30 mg/kg respectively.

Figure 2.

In vivo studies of compound 15b. (A) In vivo efficacy experiment in an established murine promyelocytic leukemia model of 15b. Shown are Kaplan–Meier survival curves of leukemic mice treated with compound 15b and its respective vehicle. Treatment started once blast cells are detected in the recipients’ peripheral blood (10 days after cell injection). 15b was orally administered at doses of 15 and 30 mg/kg for 5 days per week for 2 weeks: **p < 0.001. (B) In vivo efficacy experiment in an established murine MLL-AF9 leukemia model of 15b. Shown are Kaplan–Meier survival curves of leukemic mice treated with compound 15b and its respective vehicle. Treatment started once blast cells are detected in the recipients’ peripheral blood (10 days after cell injection). 15b was orally administered at doses of 15 and 30 mg/kg for 5 days per week for 2 weeks: **p < 0.001.

Next, 15b was also tested in MLL-AF9 model at the doses of 15 and 30 mg/kg following same protocol and analysis reported for APL model. No significant body weight differences were observed during the treatment (data not shown).

As reported in Figure 2B, also in the MLL-AF9 model, administration of the inhibitor prolonged significantly and dose dependently the survival of the mice (29% and 70% at the doses of 15 and 30 mg/kg, respectively). Remarkably the observed in vivo efficacy of 15b, obtained at safe doses, could be suggestive of its preferential activity on tumor cells with respect to normal cells.

In conclusion, the class of 5-imidazolylthieno[3,2-b]pyrroles represents a series of novel, potent, and selective noncovalent KDM1A inhibitors. The initial SAR exploration around these compounds led to inhibitors endowed with picomolar potency in vitro, which demonstrated engagement of KDM1A in THP-1 human leukemia cells and remarkable inhibition of their clonogenic potential. Finally, the in vivo efficacy demonstrated by 15b in two different leukemia models is similar to that previously observed with covalent inhibitors¹⁶ and strongly supports the further development of this class toward the identification of novel oral anticancer agents for the treatment of acute myeloid leukemia.

Glossary

Abbreviations

LSD1

lysine-specific demethylase

FAD

flavin adenine dinucleotide

MAO

monoamine oxidase

TCPA

tranylcypromine

DMSO

dimethyl sulfoxide

APL

acute promyelocytic leukemia

AML

acute myeloid leukemia

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.9b00604.

• Details of synthetic procedures, analytical data, and biological studies (PDF)

Author Present Address

^# A.R., A.C., S.B., M.R.C., R.D.Z., G.F., R.F., P.T., M.V., S.V., M.V., C.M.: IFOM, Fondazione Istituto FIRC di Oncologia Molecolare, Via Adamello 16, 20139 Milano, Italy.

Author Present Address

^∇ L.M.: Nuevolution A/S, Amgen Research Copenhagen, Rønnegade 8, 2100 Copenhagen, Denmark.

Author Present Address

^○ L.S.: S-IN Soluzioni Informatiche Srl, Via Ferrari 14, I-36100 Vicenza, Italy.

Author Contributions

^◆ A.R., A.C., and P.V. contributed equally to this work. All authors have given approval to the final version of the manuscript.

This work was supported by Rasna Therapeutics Inc.

The authors declare no competing financial interest.