Structure-activity studies on N-Substituted tranylcypromine derivatives lead to selective inhibitors of lysine specific demethylase 1 (LSD1) and potent inducers of leukemic cell differentiation

Article abstract of DOI:10.1016/j.ejmech.2017.12.001

FAD-dependent lysine-specific demethylase 1 (LSD1) is overexpressed or deregulated in many cancers such as AML and prostate cancer and hence is a promising anticancer target with first inhibitors in clinical trials. Clinical candidates are N-substituted derivatives of the dual LSD1-/monoamine oxidase-inhibitor tranylcypromine (2-PCPA) with a basic amine function in the N-substituent. These derivatives are selective over monoamine oxidases. So far, only very limited information on structure-activity studies about this important class of LSD1 inhibitors is published in peer reviewed journals. Here, we show that N-substituted 2-PCPA derivatives without a basic function or even a polar group are still potent inhibitors of LSD1 in vitro and effectively inhibit colony formation of leukemic cells in culture. Yet, these lipophilic inhibitors also block the structurally related monoamine oxidases (MAO-A and MAO-B), which may be of interest for the treatment of neurodegenerative disorders, but this property is undesired for applications in cancer treatment. The introduction of a polar, non-basic function led to optimized structures that retain potent LSD1 inhibitors but exhibit selectivity over MAOs and are highly potent in the suppression of colony formation of cultured leukemic cells. Cellular target engagement is shown via a Cellular Thermal Shift Assay (CETSA) for LSD1.

Full text of DOI:10.1016/j.ejmech.2017.12.001

Accepted Manuscript
Structure-activity studies on N-Substituted tranylcypromine derivatives lead to
selective inhibitors of lysine specific demethylase 1 (LSD1) and potent inducers of
leukemic cell differentiation
Johannes Schulz-Fincke, Mirjam Hau, Jessica Barth, Dina Robaa, Dominica
Willmann, Andreas Kürner, Julian Haas, Gabriele Greve, Tinka Haydn, Simone Fulda,
Michael Lübbert, Steffen Lüdeke, Tobias Berg, Wolfgang Sippl, Roland Schüle,
Manfred Jung
PII:
S0223-5234(17)31007-3
10.1016/j.ejmech.2017.12.001
EJMECH 9979
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 25 October 2017
Revised Date: 28 November 2017
Accepted Date: 1 December 2017
Please cite this article as: J. Schulz-Fincke, M. Hau, J. Barth, D. Robaa, D. Willmann, A. Kürner, J.
Haas, G. Greve, T. Haydn, S. Fulda, M. Lübbert, S. Lüdeke, T. Berg, W. Sippl, R. Schüle, M. Jung,
Structure-activity studies on N-Substituted tranylcypromine derivatives lead to selective inhibitors of
lysine specific demethylase 1 (LSD1) and potent inducers of leukemic cell differentiation, European
Journal of Medicinal Chemistry (2018), doi: 10.1016/j.ejmech.2017.12.001.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Structure-Activity Studies on N-Substituted
Tranylcypromine Derivatives Lead to Selective
Inhibitors of Lysine Specific Demethylase 1 (LSD1)
and Potent Inducers of Leukemic Cell
Differentiation
†
, ‡, §
†
₸,‡
∥
⊥
Johannes Schulz-Fincke
,Mirjam Hau , Jessica Barth ,Dina Robaa , Dominica Willmann ,
†
, ‡, §
†
╩,#
⁞,₸,§
⁞,₸,§
Andreas Kürner
, Julian Haas , Gabriele Greve ,Tinka Haydn , Simone Fulda
,
╩,‡
†
₸,‡
∥
⊥
Michael Lübbert , Steffen Lüdeke , Tobias Berg , Wolfgang Sippl , Roland Schüle , and
†
*
Manfred Jung
†
Institute of Pharmaceutical Sciences, University of Freiburg (Germany)
‡
German Cancer Consortium (DKTK), Freiburg (Germany)
§
German Cancer Research Center (DKFZ), Heidelberg (Germany)
∥
Institute of Pharmacy, Martin-Luther-University of Halle-Wittenberg (Germany)
1

ACCEPTED MANUSCRIPT
⊥
Department of Urology/Women´s Hospital and Center for Clinical Research, University of
Freiburg Medical Center (Germany)
₸
German Cancer Consortium (DKTK), Frankfurt (Germany)
╩
Division of Hematology, Oncology and Stem Cell Transplantation, Department of Internal
Medicine, University of Freiburg Medical Center (Germany)
#
Faculty of Biology, University of Freiburg (Germany)
⁞
Institute for Experimental Cancer Research in Pediatrics, Goethe-University Frankfurt
Germany)
(
KEYWORDS: Epigenetics, KDM1A, AML, small molecule inhibitors, SAR, mechanism-based
inhibitors
Abbreviations: AML, acute myeloic leukemia; MAOs, Monoamine oxidases A and B; HRP,
Horseradish peroxidase; TLC, thin layer chromatography; vdW, van-der-Waals; CFU Colony
forming units; Copper(II) acetylacetonate, Cu(acac) ; Ethyl diazoacetate, EDA; DFT, density
2
functional theory.
2

ACCEPTED MANUSCRIPT
ABSTRACT
FAD-dependent lysine-specific demethylase 1 (LSD1) is overexpressed or deregulated in many
cancers such as AML and prostate cancer and hence is a promising anticancer target with first
inhibitors in clinical trials. Clinical candidates are N-substituted derivatives of the dual
LSD1-/monoamine oxidase-inhibitor tranylcypromine (2-PCPA) with a basic amine function in
the N-substituent. These derivatives are selective over monoamine oxidases. So far, only very
limited information on structure-activity studies about this important class of LSD1 inhibitors is
published in peer reviewed journals. Here, we show that N-substituted 2-PCPA derivatives
without a basic function or even a polar group are still potent inhibitors of LSD1 in vitro and
effectively inhibit colony formation of leukemic cells in culture. Yet, these lipophilic inhibitors
also block the structurally related monoamine oxidases (MAO-A and MAO-B) which may be of
interest for the treatment of neurodegenerative disorders, but this property is undesired for
applications in cancer treatment. The introduction of a polar, non-basic function led to optimized
structures that retain potent LSD1 inhibitors but exhibit selectivity over MAOs and are highly
potent in the suppression of colony formation of cultured leukemic cells. Cellular target
engagement is shown via a Cellular Thermal Shift Assay (CETSA) for LSD1.
3

ACCEPTED MANUSCRIPT
Introduction
LSD1 (KDM1A) is an epigenetic modifier that demethylates distinct mono- and dimethylated
lysines in histones, but also non-histone proteins, such as p53 via a FAD-dependent mechanism.
Methylation at H3K4 is generally associated with gene activation, as removal of this mark by
LSD1 leads to transcriptional repression. Upon interaction with the androgen receptor, LSD1
shifts the demethylation to H3K9me and promotes gene activation.[1] In the monoamine oxidase
2
reaction, FAD is reduced to FADH with concomitant conversion of the substrates N-methyl
group to an imine intermediate. This is hydrolyzed subsequently and decomposes to
formaldehyde and the demethylated lysine. Simultaneously the cofactor is reoxidized by
consuming molecular oxygen which forms hydrogen peroxide (Scheme 1 A).[2] During the last
decade, LSD1 has emerged as a key epigenetic regulator of various disease states and hence
represents a promising drug target. It regulates cellular signaling pathways in various cancer
entities such as breast cancer where it interacts in a deacetylase containing multiprotein complex
(NuRD).[3] In prostate cancer high levels of LSD1 are associated with cancer progression and
metastasis and hence LSD1 levels could be a useful biomarker.[4, 5] Harris et al described the
consequence of overexpression of LSD1 in MLL-AF9 leukemia stem cells and the use of potent
LSD1 inhibitors (1b) to induce the differentiation and decrease of the clonogenic potential.
Similar effects have been described for other hematopoietic lineages as well, when LSD1 was
inhibited by small molecules or a knockdown was performed using siRNA. Hematopoiesis is
regulated by the genes Hoxa9, Meis1 and Gfi1b among others, which are critically up-regulated
by LSD1.[6, 7] The enzyme´s activity affects key cellular events, like gene expression, metabolic
processes and maintenance of oncogenes. Besides cancer, LSD1 has also been proposed as a
target for the treatment of neuronal diseases.[8-11]
4

ACCEPTED MANUSCRIPT
Scheme 1. Catalytic mechanism of LSD1 and activation of 2-PCPA.
(A) Catalytic mechanism of KDM1A; (B) Activation of 2-PCPA and formation of
FAD-adduct. The putative FAD-8c-adduct is shown according to the known mechanism.
Due to the great therapeutic potential of LSD1 inhibition, quite a large number of inhibitors
has been described and the first two selective inhibitors (1c,1d)[12, 13] have already advanced to
clinical trials for the treatment of cancer and neurodegenerative diseases.[14, 15] The
prototypical LSD1 inhibitor is trans-2-PCPA (1a) which was approved for the treatment of
depression based on its covalent inhibition of MAOs. trans-2-PCPA is also in clinical trials for
AML treatment but derivatives show significantly increased potency and compounds with
significant selectivity over MAOs have been identified. Generally, LSD1 inhibitors can be
5

ACCEPTED MANUSCRIPT
distinguished between reversible and irreversible inhibitors of LSD1. The irreversible inhibitors
1
a-1j[11, 16-20] (Figure 1 A) are suicide inhibitors. Protein crystallography shows evidence for
the formation of a covalent adduct with the cofactor after an FAD-dependent activation process
Scheme 1 B). Upon activation, the cyclopropyl structure undergoes a ring opening and
(
subsequently forms a new bond with the cofactor at position N(5) and C(4a). Crystal structures
of FAD adducts with either (1R,2S) or (1S,2R) trans-2-PCPA, respectively, show that its
orientation depends on the absolute configuration of the inhibitor.[16, 21, 22] Examples for
reversible compounds 1k-1o[1, 23-26] are mentioned in Figure 1B.
6

ACCEPTED MANUSCRIPT
Figure 1. Structures of published KDM1A inhibitors. The compounds represent covalent (A)
and reversible (B) inhibitors. 2-PCPA-based inhibitors are characterized as racemates.
References for IC values are cited in the main text. IC of 1a has been determined according to
5
0
50
the described methods.
7

ACCEPTED MANUSCRIPT
Substitution with hydrophobic groups in the cyclopropyl ring of 2-PCPA at 1-position such as
alkyl and aryls increase the activity against LSD1, whereas esters and carboxylic acids in this
position decrease the inhibitory effect.[27, 28] To increase the potency and selectivity of 2-
PCPA, modifications both on the phenyl ring and the nitrogen have been performed (1e-g, 1i).
Detailed structure-activity studies are available for phenyl-ring substituted compounds.
Derivatization in meta and para position of the phenyl ring leads to IC50 values in lower
nanomolar range and potent cellular activity.[29, 30] There are patents and publications on
studies employing N-alkylated analogues as selective LSD1 inhibitors but the SAR is not
discussed profoundly and mostly a few potent derivatives are presented with emphasis on
biological effects.[31-33] Hence, we wanted to investigate the effect of N-substitution on
biological activity.
In the present study, we investigate the SAR of N-alkylated derivatives of the parent trans-2-
PCPA (1a, Figure 1) which is not well documented in the scientific literature. Firstly, we present
the introduction of a methyl group in the unsubstituted position of the cyclopropyl ring leading to
series 6 (Table 1). Subsequently, N-alkylation is the major focus of this study and resulted in a
series of highly potent LSD1 inhibitors, that have a disubstituted cyclopropane ring (series 7 and
8
, Table 2 and 3), with IC values in the lower nanomolar range. We provide an insight in how
50
substitutions have an effect on selectivity and show that selected compounds exhibit strong
antileukemic properties in different cellular models.
8

ACCEPTED MANUSCRIPT
Results and discussion.
Based on the mechanism of trans-2-PCPA activation, we reasoned that a methyl group at the
usually unsubstituted position of the cyclopropyl ring might stabilize the formed radical in alpha
position to the carbon which is substituted by the methyl group and could result in inhibitors with
superior potency compared to trans-2-PCPA or N-substituted derivatives without a methyl group
at this position.
For the synthesis of the cyclopropyl ring a copper-catalyzed cyclopropanation was used. The
stereochemistry of the reaction leads preferably to trans-oriented substituents at the cyclopropyl
ring.[34] In addition to the trans-selective cyclopropanation, the relative orientation of the third
substituent is controlled by the stereochemistry of the starting olefin (Scheme 2). Z-methyl
styrole leads to a cis-oriented phenyl and methyl group and accordingly, the E-congener places
the phenyl and the methyl group on different sides of the cyclopropane plane. We used
copper(II) acetylacetonate as a catalyst precursor instead of the catalytically active species Cu(I)
due to stability reasons. A wide range of metal catalysts such as Rh, Ru and Co are suitable to
initiate the intramolecular cyclopropanation through decomposition of the α-diazoester.[35] With
a good yield of 60-70 % in total we stopped the reaction after 4 hours and separated the cis and
trans isomers ethyl (1S*,2S*,3S*)-2-methyl-3-phenylcyclopropane-1-carboxylate (2a) and the
(1R*,2S*,3S*)-diastereomer (2b) by flash chromatography (see Scheme 2). The appropriate
styrene could be recycled and reused in further batches. We analyzed 2a and 2b by NMR to
determine the relative configuration (Figure S1).
Due to the nature of the cyclopropanation the relative orientation of the phenyl and the methyl
group is trans. We observed an H-H-coupling with J = 9.3 Hz for the protons next to the ester
cis
and the methyl group in 2a, implying that ester and phenyl group are both trans-oriented as well.
9

ACCEPTED MANUSCRIPT
The same coupling constant (Jcis = 9.3 Hz) has been noticed for the protons next to the ester and
phenyl moiety in 2b. These coupling constants of 2a and 2b corresponds to literature values.[36]
Upon mild hydrolysis in methanol, the corresponding carboxylic acids (3a and 3b) were heated
with diphenylphosphoryl azide and triethylamine in toluene to form a cyclopropyl carbonyl azide
which decomposes in a Curtius rearrangement when heated in solvent to 120 °C. This leads to a
cyclopropyl isocyanate which reacts with added anhydrous tert-ButOH to a Boc protected amine
(4a, 4b) in 50-70 % yield. Free unprotected amine is scavenged by di-tert-butyldicarbonate. The
mixture of 2c and 2d could not be separated as esters and carboxylic acids (3c and 3d) and was
used as a mixture for the Curtius rearrangement to get 4c in low yields but not its diastereomer.
Removal of the Boc protecting group using 2 M HCl in 1,4-dioxane afforded cyclopropylamines
5
a-c as hydrochlorides. The relative configuration of 5c was determined via NMR with a
coupling of J = 8.7 Hz for protons 1 and 3 and thus resulting in the structure of 5c and its
cis
precursors, as depicted in Table 1.
As it had been shown that trans-2-PCPA was more potent than the cis-isomer, we only
synthesized the trans-oriented target compounds.[22] N-alkylation was achieved by reductive
amination of the primary amine using aldehydes and sodium triacetoxyborohydride as the
reducing agents to yield the compounds in Table 1 and 2.
10

ACCEPTED MANUSCRIPT
Scheme 2. General route of the synthesis for di- and trisubstituted substrates. All compounds
were produced as racemates and the depicted stereochemistry reflects the relative orientation of
the groups, i.e. only one enantiomer is displayed.
Reagents and conditions: (a) Cu(acac) , EDA, DCM, 60 °C, 4 h; (b) KOH, MeOH, over night;
2
(
c) NEt , DPPA, Boc O, tert-butanol, 120 °C, 12 h; (d) 1,4-dioxane, 2 M HCl, over night; (e)
3 2
STAB, appropriate aldehyde, DCE, 2h.
For in vitro testing the standard Peroxidase assay was used employing a truncated dimethyl
histone peptide (H3K4aa1-20) as substrate and quantification of the product hydrogen peroxide
using 10-acetyl-3,7-dihydroxyphenoxazine. Unexpectedly, all trisubstituted cyclopropanes were
devoid of LSD1 inhibitory activity, even in the presence of N-substituents that lead to highly
potent analogues when appended to tranylcypromine. C-methylation completely abolished LSD1
inhibitory activity (e.g. 7v vs. 6c). The assay kit MAO-Glo was used for selectivity
determination against MAOs. Some of these derivatives were fairly potent inhibitors of
11

ACCEPTED MANUSCRIPT
monoamine oxidase A with IC50 values in the nanomolar range and showed moderate activity
against monoamine oxidase B (Table 1) and they could be used as negative controls for target
engagement and assay development with 2-PCPA based LSD1 inhibitors in biological studies.
Table 1. Inhibition values of trisubstituted 2-PCPA analogues.
IC50 [nM] or inhibition [%]
Compound
LSD1
MAO-A
MAO-B
1a
6.5·10³
3.4·10³
-
b
b
1
b
65.5 ± 3.5
n.i.
n.i.
a
b
b
5a
n.i.
n.i.
n.i.
a
b
b
5
b
n.i.
n.i.
n.i.
a
b
b
5c
n.i.
n.i.
n.i.
c
b
6a
n.i.
41.4·10³
14%
c
b
b
6
b
n.i.
32%
n.i.
b
6
c
n.i.
310 ± 66
10.5·10³
56.2·10³
a
6
d
n.i.
780 ± 140
b
6e
n.i.
237 ± 45
2.2·10³
12

ACCEPTED MANUSCRIPT
IC50 values were calculated from at least 8 data points and are the mean of at least two
independent determinations; (a) Percentage of inhibition at 100 µM; (b) Percentage of inhibition
at 10 µM; (c) unspecific, not dose-dependent low inhibition; (n.i.) no inhibition
As the sterical demand of the 3-methyl group seems to prevent LSD1 binding, we focused
subsequently on N-alkylation to achieve higher potency and selectivity against LSD1. There is
little precedence in peer reviewed journals on structure-activity studies of N-alkylation despite a
wealth of patent examples. To get an overview of the effect of non-basic compounds, we
synthesized some of the examples mentioned in the referred patents.[37] Based on the structure
of compounds used in clinical trials (1c, 1d) and on analyses in a series of reversible inhibitors
that are thought to have overlapping binding sites with the N-substituted derivatives, it was
postulated that a basic moiety (aliphatic amine or pyridine) in the substituent is required for
LSD1 inhibitors that are both potent and selective (e.g. 1b and 1d).[38]
To actually probe this hypothesis, we prepared a series of N-substituted tranylcypromines that
mostly bear lipophilic or weakly polar substituents, such as 7b or 7ad. N-substitution was again
achieved by reductive animation of trans-2-PCPA. We were pleased to see that potent LSD1
inhibitors can be identified among lipophilic N-substituted tranylcypromines.
In this series we synthesized single and multiple halide substitutions at different positions
which led to potent compounds in all cases without selectivity over MAOs, though. The similar
IC50 values support our conclusion that LSD1 does not have a general preference for ortho-
meta- or para- halide resp. methoxy-substituted compounds (Table 2). Those non-pyridine
analogues of 1b showed good potency but less selectivity regarding MAOs. An exception is the
nitro substitution as the IC50 of the ortho-substituted phenyl group in compound 7ac is 6-fold
higher than the one of its meta-isomer 7ad. We observed selectivity over MAOs by using the
pyridine moiety in 7b, which already improves the potency compared to trans-2-PCPA. The
13

ACCEPTED MANUSCRIPT
combination of methoxy group and pyridine had been published as a very potent and selective
compound (1b). With our methodology we determined an IC of 65.5 ± 3.5 nM for reference
5
0
inhibitor 1b (literature value 98 nM).
14

ACCEPTED MANUSCRIPT
Table 2. IC50 values of 2-PCPA derivatives with nonpolar substituents
IC50 [nM]
IC50 [nM]
MAO-A
Compound
Compound
LSD1
MAO-A
MAO-B
LSD1
MAO-B
a
a
a
7
a
321.1 ± 69.7
202.4 ± 4.4
5.7 ·10³
33%
7u
7v
103.9 ± 3.1
77.5 ± 16.3
62%
21%
a
a
7
b
n.i.
n.i.
205.5 ± 13.4
14.1 ·10³
a
7c
66.1 ± 0.1
9.7 ·10³
36%
(S,R)-7v
95.6 ± 4.8
-
-
a
a
7
d
157.9 ± 5.5
79%
n.i.
(R,S)-7v
94.9 ± 16.3
-
-
b
b
a
a
7e
100.2 ± 27.2
30%
n.i.
7w
202.8 ± 41.9
73%
11%
a
a
b
b
7
f
110.0 ± 2.4
95.9 ± 8.2
69%
n.i.
7x
7y
7z
122.0 ± 12.2
64.5 ± 36.6
276.5 ± 12.0
30%
17%
a
a
b
b
7g
47%
24%
69%
n.i.
b
b
b
b
7
h
112.1 ± 14.2
38%
n.i.
33%
12%
a
a
b
b
7
i
294.5 ± 30.1
43%
27%
7aa
817.1 ± 19.0
22%
27%
b
b
b
b
7
j
79.7 ± 18.3
117.5 ± 2.1
27%
n.i.
7ab
7ac
91.1 ± 20.4
12%
16%
a
a
b
b
7
k
20%
16%
643.5 ± 14.9
18%
15%
a
a
7
l
108.5 ± 2.1
114.1 ± 28.6
100.0 ± 18.4
159.0 ± 42.4
117 ± 0.1
69%
20%
7ad
99.5 ± 6.4
51.8 ± 8.8
111.7 ± 2.4
35.3 ± 3.5
154.5 ± 9.2
488.5 ± 36.1
1.4 ·10³
b
b
(S,R)-
7
m
27%
n.i.
-
-
-
-
7ad
b
b
(R,S)-
ad
7
n
43%
n.i.
7
b
b
a
a
7
o
p
q
27%
n.i.
7ae
7af
100%
13%
b
b
b
b
7
94%
79%
93%
52%
b
b
b
b
7
119.9 ± 28.6
60.0 ± 4.2
93%
82%
7ag
7ah
7ai
82.6 ± 5.0
66.5 ± 0.3
80.7 ± 2.7
100%
44%
1
2
60.5 ±
3.33
a
b
b
7
r
n.i.
60%
23%
a
a
b
b
7s
325 ± 34.8
48%
17%
100%
32%
b
b
a
a
7
t
417.5 ± 3.5
30%
n.i.
7aj
69.1 ± 1.0
54%
14%
15

ACCEPTED MANUSCRIPT
The IC50 values were calculated from at least 8 data points and are the mean value of at least
two independent determinations; (n.i.) no inhibition; (-) not determined; (a) Percentage of
inhibition at 10 µM; (b) Percentage of inhibition at 1 µM. All compounds are racemates and only
one enantiomer (relative configuration) is depicted (unless marked otherwise).
The inhibitor potency is significantly decreased by alkyl substituents, at least in para position
of the N-benzyl moiety (7z, 7aa). To elucidate how far the compounds can be extended in size,
we synthesized derivatives with larger and bulkier N-substituents, again with and without a
halide substitution. Both biphenyl analogues exhibit a nanomolar effect in vitro, but the
meta-substituted derivative has a twofold lower IC against LSD1. This was also observed for
50
benzyloxybenzyl analogues where the meta-isomer 7ah shows also a twofold higher affinity than
its para analogue 7af. Interestingly, a bromo substitution increased the potency in the para
aryloxy compound 7ag to the level of the meta-congener 7ai.
Furthermore, we wanted to clarify the contribution of the stereochemistry for the in vitro
potency. It has already been shown that the enantiomers of trans-2-PCPA slightly differ in their
IC50 values.[22] To elucidate whether this effect has also been amplified by activity enhancement
from N-alkylation, we investigated the biological activity of the single enantiomers of 7v and
7
ad. We separated the stereoisomers by the use of chiral HPLC on a semipreparative scale.
Indeed, the biological evaluation revealed that (1S,2R)-7ad is more potent than (1R,2S)-7ad
while the IC50 values for the two enantiomers (1S,2R)-7v and (1R,2S)-7v are similar to the
IC value of racemic 7v. The absolute configuration of (1S,2R)-7ad and (1R,2S)-7ad were
5
0
assigned by comparison of vibrational circular dichroism (VCD) spectra to spectra from quantum
chemical calculations – a well-established method for the determination of the absolute
configuration (Figure S2).[39] Different from electronic CD, which samples electronic
transitions in chiral molecules, VCD provides characteristic spectral patterns from 6N−3
16

ACCEPTED MANUSCRIPT
vibrational modes and does not depend on the existence of a chromophoric system. The high
content of structural information in these spectral patterns and the comparable high accuracy of
spectrum prediction (usually on DFT level) makes VCD a highly valuable tool for the
stereochemical analysis and is nowadays widely applied for drug molecules.[40]
7v and 7ad were also rather potent inhibitors of MAO-A but not MAO-B. Thus, while
N-benzylation is sufficient to increase LSD1 inhibitory potency substantially, it does not lead to
selectivity over MAOs. We still investigated the activity of compounds in cellular models that
have previously been shown to respond strongly to LSD1 inhibition. The colony-forming unit
assay (CFU assay) is enumerating hematopoietic progenitor cancer cells with colony formation
potential. Treatment with LSD inhibitors decreases the number of units and allows the
quantification of the effect of inhibitors on cancer entities.
First, we tested the unselective compounds 7v and 7ad for their antileukemic properties. They
showed an effect when THP1 cells were treated with nanomolar concentrations and decreased
the number of colony forming units by 80 % at 0.5 µM (Figure 2 A). In contrast the trisubstituted
analogues 6c and 6e did not show any influence in this cell line even at higher concentrations,
which is in accordance with an inactivity toward LSD1 (Figure 2 B). However, the reduction of
colony formation is not specific for only LSD1 inhibition which is why we showed target
engagement of LSD1 with the compounds 7v and 7ad in the CETSA (Fig. 6) in the same cell
line as was used in the colony forming unit assay.
17

ACCEPTED MANUSCRIPT
A
1
1
20
00
8
6
4
2
0
0
0
0
0
7
v
7ad
8
d
b
1
8c
0
.05
0.1
0.5
1.0
µM
B
1
1
20
00
8
6
4
2
0
0
0
0
0
6
6
e
c
0
.5
1.0
10.0
50.0
µM
Figure 2. Antileukemic effect on THP-1 cells of selected compounds; (A) N-benzylated
compounds reduce ability to create colony forming units, 1b is used as reference inhibitor at the
concentration of 0.1 and 0.5 µM; (B) no cellular effect of trisubstituted 2-PCPA analogues even
at high concentrations.
18

ACCEPTED MANUSCRIPT
It has previously been shown that LSD1 inhibition induces differentiation in murine and
mammalian leukemia models[6, 41, 42]. In order to evaluate the differentiation effects induced
by LSD1 inhibition in vitro, we used Hoxa9/Meis1-transformed murine progenitor cells that
elicit leukemia upon injection into irradiated recipient mice. Cells were treated with increasing
concentrations of 7v and 7ad and analyzed for morphological signs of granulocytic
differentiation as well as changes in surface marker expression. We observed differentiation
including characteristic changes in nuclear morphology (Figure 3 A). It is known that the
leukemic stem cell population in murine leukemia models resides in the c-Kit-positive
fraction.[43] We therefore used flow cytometric analysis to determine the reduction in the
fraction of c-Kit-positive cells upon treatment with the LSD1 inhibitors 7v and 7ad (Figure 3 B).
Both compounds induced differentiation effects even at 50 nM and led to a reduction in c-Kit-
positive cells suggesting a good cellular penetration and potency (Figure 3 C).
19

ACCEPTED MANUSCRIPT
20

ACCEPTED MANUSCRIPT
Figure 3. Inhibition of LSD1 by 7v and 7ad induces granulocytic differentiation and loss of
c-Kit-expressing immature cells in Hoxa9/Meis1-transformed hematopoietic progenitors. (A)
Representative images of cytospin preparations of Hoxa9/Meis1 cells after treatment with
1
00 nM of 7v or 7ad at indicated time points stained with Pappenheim stain, Scale bar: 50 µm;
B) Representative flow cytometric analysis for c-Kit and Sca1 in Hoxa9/Meis1-transformed
myeloid progenitors; (C) Quantitative analysis of c-Kit-positive cells in
(
Hoxa9/Meis1-transformed cells as analyzed by flow cytometry. Cells were treated with indicated
concentrations of LSD1 inhibitors 7v and 7ad for 96 h. The mean percentage ± SEM of
c-Kit-positive cells from 3 experiments is shown; * p≤0.05, ** p≤0.01
Given that strongly basic tertiary amines as well as the weakly basic pyridine in inhibitor 1b
give rise to compounds that are potent and selective, we reasoned that just a polar function might
be enough to increase selectivity over monoamine oxidases. Hence, we prepared a second series
of N-benzylated tranylcypromines with more polar substituents. For example the sulfonamide 8c
and the N-methylsulfonamide 8d have calculated logP-values of 1.0 resp. 2.0 as opposed to 2.5
in derivative 7v (predicted with Schrödinger suite 2013).[44]
The compounds 8a-e retained the strong inhibitory properties of the lipophilic series on LSD1
with IC50 values below 200 nM (Table 3). As already observed in the first series, the meta-
substituted compound is again more potent than ortho substituted derivatives. Different from the
first series, they also present significant selectivity over MAOs with compound 8c showing the
best selectivity. The compounds 8c and 8d also showed even more potent antileukemic
properties and reduced the count of CFUs by 80 % at only 50 nM (Figure 2 A).
21

ACCEPTED MANUSCRIPT
Table 3. IC50 values of tranylcypromine derivatives with polar substituents
IC50 [nM]
MAO-A
Selectivity
MAO-A/LSD1
Compound
LSD1
MAO-B
MAO-B/LSD1
1052
8a
91.0 ± 0.2
27.9 ·10³
24.7 ·10³
95.7 ·10³
307
127
a
8
b
194.6 ± 7.4
146.5 ± 0.1
187.3 ± 0.1
50%
>1282
>1700
>1337
a
8c
21.6 ·10³
17.1 ·10³
42%
147
91
a
8
d
n.i.
The IC values were calculated from at least 8 data points and are the mean value of at least
5
0
two independent determinations; (n.i.) no inhibition; (a) Percentage of inhibition at 250 µM; All
compounds are racemates and only one enantiomer is depicted.
To rationalize the observed selectivity profiles of the herein described N-substituted 2-PCPA
derivatives and to predict the interaction for LSD1, molecular docking was performed. As
already mentioned 2-PCPA derivatives bearing polar substituents at the N-benzyl group (8c and
8
d) show a preference for LSD1 over MAOs. Meanwhile, 3-methyl-2-PCPA derivatives
completely lack activity for LSD1, however, some still display considerable activity for MAO-A.
Previous studies as well as crystal structure analysis of LSD1-2-PCPA derivatives[21, 22, 28, 45]
demonstrate the formation of either N5, C4a or cyclic adducts with the cofactor FAD. In the case
of MAO-B, available crystal structures reveal the formation of the C4a-adduct.[46] Based on the
co-crystallized ligand structure of the C4a-adduct of the 1-ethyl-2-PCPA derivative in complex
with LSD1[28] and of 2-PCPA with MAO-B[46], similar C4a adducts of the herein reported
compounds were modelled as N-benzyl-imines and subsequently docked into the crystal
22

ACCEPTED MANUSCRIPT
structures of LSD1 (PDB ID 4UV8)[28] and MAO-A (PDB ID 2BXS)[47] using the program
Glide.[48]
3
3
In contrast to MAOs, LSD1 shows a wide active site cleft (~550 Å and ~400 Å in
3
MAOs[49], respectively, compared to ~1245 Å in LSD1[50]), which allows the binding of the
modelled C4a adducts in an extended conformation, as exemplified by the binding modes of 7v,
7
ah, 7r, and 8c (Figure 4A-D). The 2-phenyl moiety adopts the same position as observed in the
crystal structure of analogous 1-ethyl-2-PCPA (PDB ID 4UV8) and is embedded in a
hydrophobic region undergoing hydrophobic interactions with Phe538 and Val333. Meanwhile,
the N-substituent is hosted either in a sub-pocket lined by Asn540, Trp552, Val764, Tyr773,
Pro808, and Ala809 (7v Figure 4A), or in an adjacent sub-pocket formed by Thr335, Asp555,
Glu559, Phe560, and Tyr807 (7ah Figure 4B). The docking poses of derivatives bearing
hydrophobic and polar substituents, as exemplified by 7r and 8c respectively, are shown in
Figure 4C and 4D. While the biphenyl group of 7r is able to form van-der-Waals (vdW)
interactions with Phe560, the polar sulfonamide group of 8c undergoes polar interactions and
forms two H-bonds with the side chains of His812 and Asp555. This might explain why LSD1
can tolerate both polar and apolar moieties at the N-benzyl substituent. The obtained docking
results also demonstrate that an additional methyl group in 2-PCPA would cause a steric clash,
especially with Thr810 and, thus, disrupt the formation of the observed binding mode. This
might explain the lack of activity of the reported 2-methyl-3-phenylcyclopropan-1-amine
derivatives on LSD1.
23

ACCEPTED MANUSCRIPT
24

ACCEPTED MANUSCRIPT
Figure 4. Selected docking poses of the modelled C4a-adducts in LSD1. The surface of the
protein is depicted on the left side and removed on the right side for more clarity. The protein
residues are shown as white sticks. The adduct is displayed as cyan and yellow sticks, for the
ligand and the flavin ring of FAD, respectively. Water molecules are shown as red spheres; (A)
Docking pose of 7v C4a-adduct in LSD1; (B) Docking pose of 7ah C4a-adduct in LSD1; (C)
Docking pose of 7r C4a-adduct in LSD1; (D) Docking pose of 8c C4a-adduct in LSD1.
25

ACCEPTED MANUSCRIPT
MAO-A possesses a relatively narrow active site cleft, which can still host the C4a adducts of
described N-benzyl-PCPA derivatives, albeit in a U-shaped conformation (Figure 5). Again, the
2
-phenyl group is embedded in a hydrophobic region, where it undergoes vdW interactions with
Tyr69 and Phe352. In contrast to the binding mode predicted in LSD1, the substituent at the N-
benzyl group of the studied compounds is hosted in a notably hydrophobic region and
surrounded by the residues Ile180, Phe208, Val210, Cys323, Ile335, and Leu337. Hence, a polar
substituent at the benzyl group would be highly unfavorable, which explains the selectivity of the
sulfonamide derivatives 8c and 8d for LSD1 over MAO-A. The derived docking results also
supported the notion that MAO-A can host 2-Methyl-3-phenylcyclopropan-1-amine derivatives,
as exemplified by the docking pose of 6c in MAO-A (Figure 5B). The obtained docking pose
shows a highly similar binding mode as for the unsubstituted derivatives, with the 2-methyl
group placed in the vicinity of Tyr444.
26

ACCEPTED MANUSCRIPT
Figure 5. Selected docking poses of C4a-adducts in MAO-A. The surface of the protein is
depicted on the left side and removed on the right side for more clarity. The protein residues are
shown as white sticks. The adduct is displayed as cyan and yellow sticks, for the ligand and
flavin ring of FAD, respectively. Water molecules are shown as red spheres; (A) Docking pose
of 7ag C4a-adduct in MAO-A; (B) Docking pose of 6c C4a-adduct in MAO-A.
27

ACCEPTED MANUSCRIPT
As it has been shown that LSD1 inhibition does not generally lead to global hypermethylation
in leukemic cells[51, 52], we sought to demonstrate target engagement by alternative means
using a binding assay which is independent from putative downstream signaling effects. One
approach to measure target binding in a cellular setting is the Cellular Thermal Shift Assay
(CETSA). Treated cells with and without inhibitor are stressed thermally and binding to the
target might lead to a thermostabilization of the target in question.[53] In the assay only the
intended target protein is monitored and possible off-targets of analyzed compounds would stay
unnoticed. First, we realized a well-defined “melting curve” of LSD1 in THP-1 cells, using
western blotting as a means of quantification of LSD1 in the supernatant of lysates of cells after
thermal stress. Incubation of the cells with the positive control 1b (T = 50.8 ± 0.5 °C) and the
m
inhibitors 7v (T = 52.8 ± 2.9 °C) and 7ad (T = 53.2 ± 0.7 °C) showed a strong right shift of the
m
m
melting curve (DMSO control T = 46.0 ± 1.1 °C; mean and standard deviation calculated from
m
four independent experiments). The investigated negative controls 6c (T = 47.9 ± 0.4 °C) and 6e
m
(Tm = 45.6 ± 0.3 °C) did not show a strong shift of the melting curve. This strongly supports
LSD1 target engagement in these cells (Figure 6, S4-S9) but also the usefulness of the negative
controls 6c and 6e for target engagement studies of LSD1.
28