Novel spirocyclic tranylcypromine derivatives as lysine-specific demethylase 1 (LSD1) inhibitors

Herein we describe the design, synthesis, and biological evaluation of a novel series of tranylcypromine-based LSD1 inhibitors via conformational restriction using spiro ring systems. A simple, direct spirocyclic analog of tranylcypromine (compounds 8a and 8b) was shown to be a 28- to 129-fold more potent inhibitor of LSD1 enzyme compared to tranylcypromine. Further incorporation of various substituted benzyl groups to the amino group resulted in a suite of 2′,3′-dihydrospiro[cyclopropane-1,1′-inden]-2-amines that are potent LSD1 inhibitors with excellent selectivity profiles (e.g.14a, 15b, 16a, 19a and 20b) against closely related enzymes such as MAO-A, MAO-B, and LSD2.


Introduction
Epigenetic modications of DNA or histone tails, such as methylation and acetylation, provide regulatory mechanisms to inuence gene expression. This dynamic process allows for the chromatin uncoiling or compaction, thereby altering the ability of the transcription machineries to access the DNA. Lysine-specic demethylase 1 (LSD1) is now recognized as the rst true histone demethylase, originally discovered by the group of Shi Yang in 2004. 1 LSD1 belongs to the avin adenine dinucleotide (FAD)dependent amine oxidase superfamily and specically demethylates mono-and dimethylated histone H3 lysine 4 (H3K4me and H3K4me2). 1,2 The catalytic cycle is thought to involve the methylated lysine being rstly oxidized by FAD co-factor to form an imine intermediate, followed by hydrolysis to generate formaldehyde and the demethylated lysine. FADH 2 , the reduced product of FAD, is then oxidized by O 2 to form FAD and H 2 O 2 . LSD1 is oen present in transcriptional co-repressor complexes such as RE1-silencing transcriptional factor corepressor 1 (CoREST), which can stabilize and recruit LSD1 aer binding to chromatin. 3 Alteration of the LSD1 substrate specicity occurs upon association with the androgen receptor, namely from H3K4 to H3K9. 4 LSD1 is also known for its ability to demethylate non-histone proteins, such as p53, thus inhibiting the apoptosis process mediated by p53. 5 In addition to LSD1, the LSD family has a homologue known as LSD2 which was rst indentied in 2009. 6,7 Both homologues contain the N-terminal SWIRM (Swi3p, Rsc8p and Moira) domain and the C-terminal amine oxidase domain. 8 However, while LSD1 contains the Tower domain which forms a binding site with CoREST, the LSD2 contains a CW-type zinc nger domain with unknown function. 6 There are numerous studies reporting that LSD1 is upregulated in various tumor tissue cells and tissues including retinoblastoma, 9 non-small cell lung cancer, 10 prostate cancer, 11 breast cancer, 12,13 and colon cancer. 14 LSD1 is also overexpressed in MLL-rearranged leukemia as a key regulator to promote the oncogenic potential of MLL-AF9 leukemia stem cell. 15 MLL gene translocation is a major biomarker in acute leukemia which is typically associated with a poor prognosis. The results of RNAi-mediated knockdown or pharmacological inhibition of LSD1 seem to suggest that this enzyme induces reexpression of epigenetically silenced tumor suppressor genes and regulates p53 transcriptional activity or down-regulation of several leukemic-related genes. [15][16][17] Overall, LSD1 represents a promising target for the treatment of leukemia as well as other solid tumors.
LSD1 exhibits homology with monoamine oxidases (MAOs) A and B with 17.6% identity. 18 Classic MAO inhibitors such as trans-phenylcyclopropylamine (1, tPCPA), phenelzine (2), and pargyline (3) were found to show LSD1 inhibitory activity by forming covalent adduct with FAD (Fig. 1). The (1R, 2S)-isomer of 1 obtained by chiral resolution of tPCPA reacts with FAD in the active site to produce N-(5) adduct A, while the (1S, 2R)-isomer generates N-(5) adduct B (Fig. 2). 18,19 A number of tPCPAcontaining inhibitors were developed with the aim to improve the potency at LSD1 as well as the selectivity over MAOs. The catalytic site for LSD1 is larger compared to MAOs, 20 with substitutions at the phenyl ring and amino nitrogen atom generally increasing the LSD1 inhibitory activity. Compound 4 is a potent LSD1 inhibitor with an IC 50 value of 98 nM that has been demonstrated to inhibit colony forming capacity of MLL-AF9 leukemia cells. 15 Compound 5 was reported to be a brain-permeable LSD1 inhibitor that can block memory consolidation in a mouse model. 21 Other classes of reversible LSD1 inhibitors have also been reported, including peptides, polyamine analogues, pyrimidinethioureas, 3-(piperidin-4-ylmethoxy)pyridine and 3,5-diamino-1,2,4-triazoles. 22-27 N1-[(1R, 2S)-2-phenylcyclopropyl]cyclohexane-1,4-diamine dihydrochlorideORY-1001 (6, ORY1001, RG6016, Oryzon Genomics) is an irreversible LSD1 inhibitor that was granted an orphan drug status for the treatment of acute myeloid leukemia (AML) in phase IIA clinical trial. 28 GlaxoSmithKline developed a potent LSD1 inhibitor, compound 7 (GSK2879552), which is currently in the phase I trials for treatment of relapsed/ refractory small cell lung cancer and AML. More recently, Oryzon Genomics announced tPCPA analogue ORY2001, a dual LSD1/ MAO-B inhibitor, that was advanced in phase I studies for Alzheimer's disease in 2016. Other compounds currently in phase I/ II clinical trials for oncology include INCB059872, IMG-7289 and CC-90011. 29 Although there are numerous medicinal chemistry campaigns on tranylcypromine-related analogues, 30 thus far there has been a limited report on the incorporation of spirocycle into the system. 31,32 Spiro containing system has been increasingly utilized in medicinal chemistry as it introduces not only diverse orientation but also structural novelty. 33 Within this context, the spirocycle constraint may inuence the potency and selectivity for LSD1 inhibition over other homologous enzymes, such as LSD2, MAO-A, and MAO-B. In the present study, we report a series of 2 0 ,3 0 -dihydrospiro[cyclopropane-1,1 0inden]-2-amine analogues and their potencies at inhibiting LSD1 and related enzymes (Fig. 3).
Following their synthesis, compounds 8a and 8b were tested for their inhibitory activities against the puried LSD1 recombinant. As shown in Table 1, compounds 8a and 8b drastically improved the inhibitory activity against LSD1 by 129and 28-fold compared to the control compound tPCPA (1), with the trans-isomer 8a being 5-fold more potent than the cis-isomer 8b. Furthermore, both isomers exhibited very high selectivity of >600-and >120-fold against MAO-A, but only modest selectivity for MAO-B. The observed trend suggested that the conformational restriction imposed by the spirocycle was benecial for the inhibitory activity against LSD1 and selectivity over MAO-A.
Encouraged by the preliminary in vitro results of the spirocycles 8a and 8b, we further explored structural extensions on the amino group (13a-25a, 13b-25b). As outlined in Scheme 2, reductive amination of the amines 8a and 8b with various substituted benzaldehydes or pyridine aldehydes using NaBH 4 gave the nal products 13a-24a and 13b-25b. Alkylation of 8a or 8b with 2-chloro-1-morpholinoethan-1-one in the presence of sodium hydride in anhydrous dimethylformamide (DMF) provided the amides 25a and 25b. Their IC 50 values for the  In particular, a 2-methoxybenzyl (14a, 14b) or 2-uorobenzyl substituents (15a, 15b) on the amino group enhanced the inhibitory potencies to single digit nanomolar levels, indicating that both electron donating and withdrawing groups are tolerated at the 2-position of the aromatic ring. Within the disubstituted benzyl analogues (16a-19a, 16b-19b), all the trans-isomers were found to be more potent inhibitors of LSD1 compared to their corresponding cis-isomers. Among the 2,5disubstituted benzyls analogues 16a-17a and 16b-17b, while the trans-isomers are almost equipotent at LSD1, the additional 5-uoro substitution gave slightly more potent compounds than the 5-bromo substitution (17b vs. 16b) for the cis-isomers, indicating potential steric issues at this region. Moving the uoro substituent from the 5-to the 4-position of the benzyl group did not elicit signicant changes in the IC 50 values (17a, 17b vs. 18a, 18b). The 3,4-dimethoxy analogues (19a, 19b) were almost as equipotent as the 2-methoxy analogue (14a, 14b). For the trisubstituted benzyl analogue 20, the cis-isomer (20b) was found to be 3-fold more potent than the trans-isomer (20a). Changing a benzyl to a 3-pyridylmethyl group (21a-23a, 21b-23b) resulted in essentially equipotent isomers, with the exception of the 2methoxy derivative, where the trans-isomer 22a (IC 50 ¼ 32 nM) was 3-fold less potent than the cis-isomer 22b (IC 50 ¼ 10 nM). The unsubstituted 3-pyridylmethyl analogues 21a and 21b were found to be 10-fold and 39-fold more potent than 8a and 8b, respectively. The 2-methoxy analogues 22a and 22b exhibited similar inhibitory activity against LSD1 compared with 21a and 21b. However, the 2-uoro derivatives 23a and 23b were approximately 4-fold less potent than the unsubstituted derivatives 21a and 21b (IC 50 ¼ 64 and 70 nM). Keeping the uoro group intact, a pyridyl walk from 3-to the 2-position (24a, 24b vs. 23a, 23b) resulted in at least a 4-fold reduction in their LSD1 inhibitory potencies. Swapping the aromatic group to an aliphatic morpholinecarbonyl (25a, 25b) showed a further signicant reduction in the LSD1 inhibitory activities, with IC 50 values of 1.4 and 22 mM, respectively. In terms of selectivity for LSD1 inhibition over other related enzymes LSD2, MAO-A and MAO-B, all of the synthesized target compounds were the most potent at inhibiting LSD1 with selectivity index varying between low (e.g. IC 50 values of 13b and 18b for LSD1 vs. MAO-A) and excellent (e.g. IC 50 values of 14a, 15b, 16a, 19a and 20b for LSD1 vs. MAO-A, MAO-B and LSD2).
The subsequent round of structural activity relationship (SAR) investigations involved structural extensions on the aromatic ring of the dihydroindene. As shown in Scheme 3, the commercially available 5-bromo-2,3-dihydro-1H-inden-1-one (26) were used as the starting material and subjected to the conditions as previously outlined in Scheme 2. The isomers 27a and 27b were separated by ash chromatography and the relative stereochemistry assigned based on a combination of their COSY and NOESY studies. Suzuki coupling of 27a with aromatic boronic acid in the presence of Pd(PPh 3 ) 4 at 80 C in degassed 1 M Na 2 CO 3 (aq)/DMF, followed by deprotection of the Boc group afforded the desired compounds 28a and 29a. Compounds 28b and 29b were synthesized following identical procedures with 27b as the starting material. The nal compounds 30a-32a and 30b were readily accessed via reductive amination of the corresponding precursors with various substituted benzaldehydes and NaBH 4 as the reducing agent. Their inhibitory activities at LSD1 and related enzymes are shown in Table 2.
As demonstrated in Table 2, an unsubstituted benzene ring extension at the aromatic group of the dihydroindene gave opposite effects depending on the stereochemistry of the cyclopropanamine. For the trans-isomers, a very slight increase in potency was observed (28a vs. 8a), whereas an opposite trend was seen for the cis-isomers (28b vs. 8b), with 28b having an IC 50 value of 2.7 mM against LSD1. A 4-uoro substitution on the newly added benzene ring resulted in approximately 3-fold reduction in the LSD1 inhibitory potencies (29a, 29b vs. 28a, 28b). The remaining synthesized compounds 30a, 30b, 31a and 32a all showed IC 50 values in the single digit micromolar range, indicating a potentially undesired steric effect at this region. As the most potent LSD1 inhibitor arising from this biphenyl series, compound 28a was further tested against other enzymes MAO-A, MAO-B and LSD2, with IC 50 values of 0.46 mM, 0.17 mM and 94 mM, respectively.

Conclusion
In summary, a novel series of 2 0 ,3 0 -dihydrospiro[cyclopropane-1,1 0 -inden]-2-amine analogues as potent LSD1 inhibitors have been developed. In line with the conformational restriction conferred by the spirocycle to the tranylcypromine system, compounds 8a and 8b displayed signicant improvements in their LSD1 inhibitory potencies and high selectivities for MAO-A. In order to further explore the SAR, structural modications on the amino group and the benzene ring of the dihydroindene were found to give more potent compounds. Specically, the addition of substituted benzyl moieties to the amino group (13a-25a, 13b-25b) showed single digit nanomolar potencies against LSD1 and excellent selectivity proles against the homologous MAO enzymes and LSD2. While most of the transisomers were generally found to be more potent than the corresponding cis-isomers, there were some exceptions in the SAR data. The additional aromatic moiety at the benzene ring of the dihydroindene (28a-32a, 28b-30b) resulted in a signicant drop of the LSD1 inhibitory potencies, indicating an unfavorable steric clash at this region. Overall, these studies warrant the further development of spirocyclic tranylcypromine derivatives as potent and selective LSD1 inhibitors.

Chemistry
Starting materials, reagents, and solvents were purchased from commercial suppliers and used without further purication, unless otherwise stated. Anhydrous tetrahedrofuran (THF) and CH 2 Cl 2 were obtained by distillation over sodium wire or CaH 2 , respectively. All non-aqueous reactions were run under a nitrogen atmosphere with exclusion of moisture from reagents, and all reaction vessels were oven-dried. The progress of reactions was monitored by TLC on SiO 2 . Spots were visualized by their quenching of the uorescence of an indicator admixed to the SiO 2 layer, or by dipping into phosphomolybdic acid ethanol solution followed by heating. SiO 2 for ash chromatography was of 200-300 mesh particle size, and an EtOAc/ PE mixture or gradient was used unless stated otherwise. 1 H NMR spectra were recorded at a spectrometer frequency of 400 MHz, 13 C NMR spectra at 101 MHz. Chemical shis are reported in d (ppm) using the d 0 signal of tetramethylsilane (TMS) as internal standard. High resolution mass spectra were performed using a Bruker ESI-TOF high-resolution mass spectrometer and Waters Micromass Q-TOF micro Synapt high denition mass spectrometer.

MAO-A and MAO-B inhibition assay
MAO-A and MAO-B enzymes were purchased from Sigma-Aldrich and the inhibition assay was tested with Promega MAO-Glo™ Assay kit according to the manufacturer's protocol. The luminescent signals were detected using Envision (Perki-nElmer). The test was carried out in triplicate. And the IC 50 data was calculated using the soware GraphPad Prism 5, and chosen the equation "sigmoidal dose-response (variable slope)" for curve tting.

Conflicts of interest
There are no conicts of interest to declare.