Biochemical, structural, and biological evaluation of tranylcypromine derivatives as inhibitors of histone demethylases LSD1 and LSD2

Article abstract of DOI:10.1021/ja101557k

LSD1 and LSD2 histone demethylases are implicated in a number of physiological and pathological processes, ranging from tumorigenesis to herpes virus infection. A comprehensive structural, biochemical, and cellular study is presented here to probe the potential of these enzymes for epigenetic therapies. This approach employs tranylcypromine as a chemical scaffold for the design of novel demethylase inhibitors. This drug is a clinically validated antidepressant known to target monoamine oxidases A and B. These two flavoenzymes are structurally related to LSD1 and LSD2. Mechanistic and crystallographic studies of tranylcypromine inhibition reveal a lack of selectivity and differing covalent modifications of the FAD cofactor depending on the enantiomeric form. These findings are pharmacologically relevant, since tranylcypromine is currently administered as a racemic mixture. A large set of tranylcypromine analogues were synthesized and screened for inhibitory activities. We found that the common evolutionary origin of LSD and MAO enzymes, despite their unrelated functions and substrate specificities, is reflected in related ligand-binding properties. A few compounds with partial enzyme selectivity were identified. The biological activity of one of these new inhibitors was evaluated with a cellular model of acute promyelocytic leukemia chosen since its pathogenesis includes aberrant activities of several chromatin modifiers. Marked effects on cell differentiation and an unprecedented synergistic activity with antileukemia drugs were observed. These data demonstrate that these LSD1/2 inhibitors are of potential relevance for the treatment of promyelocytic leukemia and, more generally, as tools to alter chromatin state with promise of a block of tumor progression.

Full text of DOI:10.1021/ja101557k

Published on Web 04/23/2010
Biochemical, Structural, and Biological Evaluation of
Tranylcypromine Derivatives as Inhibitors of Histone
Demethylases LSD1 and LSD2
Claudia Binda,†,# Sergio Valente,‡,# Mauro Romanenghi,§ Simona Pilotto,†,‡
Roberto Cirilli,| Aristotele Karytinos,† Giuseppe Ciossani,† Oronza A. Botrugno,§
Federico Forneris,†,¶ Maria Tardugno,‡ Dale E. Edmondson,⊥ Saverio Minucci,*,§,∇
Andrea Mattevi,*,† and Antonello Mai*,‡
Department of Genetics and Microbiology, UniVersity of PaVia, Via Ferrata 1, 27100 PaVia,
Italy, Department of Drug Chemistry and Technologies, UniVersity “La Sapienza”, P.le A.
Moro 5, 00185 Roma, Italy, European Institute of Oncology, Via Adamello 16, 20139 Milan,
Italy, Italian National Institute of Health, Department of Therapeutic Research and Medicines
EValuation, Via Regina Elena 299, 00161 Roma, Italy, Department of Biochemistry, Emory
UniVersity, 1510 Clifton Road, Atlanta, Georgia 30322, and Department of Biomolecular
Sciences and Biotechnologies, UniVersity of Milan, Via Celoria, 26, 20133 Milan, Italy
Received February 23, 2010; E-mail: saverio.minucci@ifom-ieo-campus.it; antonello.mai@uniroma1.it;
andrea.mattevi@unipv.it
Abstract: LSD1 and LSD2 histone demethylases are implicated in a number of physiological and
pathological processes, ranging from tumorigenesis to herpes virus infection. A comprehensive structural,
biochemical, and cellular study is presented here to probe the potential of these enzymes for epigenetic
therapies. This approach employs tranylcypromine as a chemical scaffold for the design of novel demethylase
inhibitors. This drug is a clinically validated antidepressant known to target monoamine oxidases A and B.
These two flavoenzymes are structurally related to LSD1 and LSD2. Mechanistic and crystallographic studies
of tranylcypromine inhibition reveal a lack of selectivity and differing covalent modifications of the FAD
cofactor depending on the enantiomeric form. These findings are pharmacologically relevant, since
tranylcypromine is currently administered as a racemic mixture. A large set of tranylcypromine analogues
were synthesized and screened for inhibitory activities. We found that the common evolutionary origin of
LSD and MAO enzymes, despite their unrelated functions and substrate specificities, is reflected in related
ligand-binding properties. A few compounds with partial enzyme selectivity were identified. The biological
activity of one of these new inhibitors was evaluated with a cellular model of acute promyelocytic leukemia
chosen since its pathogenesis includes aberrant activities of several chromatin modifiers. Marked effects
on cell differentiation and an unprecedented synergistic activity with antileukemia drugs were observed.
These data demonstrate that these LSD1/2 inhibitors are of potential relevance for the treatment of
promyelocytic leukemia and, more generally, as tools to alter chromatin state with promise of a block of
tumor progression.
has revealed two types of enzymatic mechanisms.3 The iron-
dependent enzymes can demethylate lysine side chains in all
three methylation states. Conversely, the oxidative chemistry
that underlies the function of flavin-dependent histone dem-
ethylases makes it impossible for these enzymes to act on
trimethylated lysine and restricts their activity to mono- and
dimethylated substrates.4
Mammals contain two flavoenzyme demethylases: LSD1 and
LSD2 known also as KDM1A and KDM1B.5,6 LSD1 is typically
associated with the corepressor protein CoREST,7 and removes
methyl groups from mono- and dimethyl Lys4 of histone H3, a
gene activation mark. The enzyme is an interesting target for
Introduction
Alterations in the structural and functional states of chromatin
are involved in the pathogenesis of a variety of diseases.1
Histone lysine demethylases represent very attractive targets for
epigenetic drugs and are gaining increasing attention. A lysine
can be mono-, di-, and trimethylated. Each modification on the
same amino acid can exert different biological effects.2 Con-
sistently, the recent discovery of histone lysine demethylases
† University of Pavia.
# These authors contributed equally.
‡ University “La Sapienza”.
§ European Institute of Oncology.
| Italian National Institute of Health.
¶ Present address: Crystal and Structural Chemistry, Bijvoet Center for
Biomolecular Research, Utrecht University, The Netherlands.
⊥ Emory University.
(2) Bhaumik, S. R.; Smith, E.; Shilatifard, A. Nat. Struct. Mol. Biol. 2007,
14, 1008.
(3) Anand, R.; Marmorstein, R. J. Biol. Chem. 2007, 282, 35425.
(4) Forneris, F.; Binda, C.; Battaglioli, E.; Mattevi, A. Trends Biochem.
Sci. 2008, 33, 181.
∇ University of Milan.
(1) Best, J. D.; Carey, N. Drug DiscoVery Today 2009, 15, 65.
9
10.1021/ja101557k  2010 American Chemical Society
J. AM. CHEM. SOC. 2010, 132, 6827–6833 6827

A R T I C L E S
Binda et al.
expressed in Escherichia coli and purified as described.6 Human
recombinant LSD1/CoREST were expressed in E. coli as separate
proteins and co-purified following previously reported procedures.20
Enzymatic activities and inhibition assays with both demethylases
were carried out at pH 7.5-8.0 using methylated H3 peptides.6,20
Crystals of LSD1/CoREST inhibitor complexes were obtained by
cocrystallization under conditions identical to those previously
reported for other structural studies on these proteins.14,20 Diffrac-
tion data were measured at the European Synchrotron Radiation
Facility and the Swiss Light Source. Data processing and structure
refinement were carried out using standard procedures.20,21 Crystal-
lographic and refinement statistics together with the PDB accession
codes are reported in Table S5. Figures were prepared with Pymol
(www.pymol.org).
NB4 cells were treated at different concentrations of 14e (Table
1). Whole cell lysates were subjected to SDS-PAGE, and then
immunoblotted using antibodies against different histone modifica-
tions (from Abcam: H3K4me2, H3K9me2, H3; antiacetylated H4,
T2522). Analysis of NB4 cells and murine APL blasts growth and
differentiation was performed as described previously.23,24
Data Deposition. Coordinates have been deposited with the
Protein Data Bank. Accession codes are 2XAF, 2XAG, 2XAH,
2XAJ, 2XAQ, 2XAS (Table S5).
epigenetic drugs as suggested by its overexpression in solid
tumors,8 its role in various differentiation processes,9 its
involvement in herpes virus infection,10 and its association to
histone deacetylase 1, a validated drug-target. LSD2, like LSD1,
displays a strict specificity for mono- and dimethylated Lys4
of H3. However, the biology of LSD2 is proposed to differ from
that of LSD1 since LSD2 does not bind CoREST and has not
been found so-far in any LSD1-containing protein complex.6,11,12
LSD1 and LSD2 share a similar catalytic domain (45%
sequence identity) that is structurally homologous with the amine
oxidases, a class of flavin-dependent enzymes that act on
biogenic amines.4,5 Among these proteins, human monoamine
oxidases (MAOs) A and B have been the subject of more than
50 years of research that has led to the development of a
multitude of inhibitors including antidepressive and antiparkin-
son drugs.13 Their similarity in the catalytic and structural
properties prompted the investigation of antiMAO drugs as
potential LSD1 inhibitors.14 It was found that tranylcypromine
(Table 1), a MAO inhibitor used as antidepressive drug, is able
to inhibit LSD1.15-18 On this basis, we sought to design
compounds that would be more selective for demethylases using
tranylcypromine as the lead scaffold. Here, we report the
synthesis of a series (more than 40) of new tranylcypromine
analogues and a biochemical and biological evaluation of their
inhibitory properties with human LSD1, mouse LSD2, human
MAO A, and human MAO B (see Supporting Information). The
results demonstrate that many of these compounds are effective
LSD1 and LSD2 inhibitors, and most importantly, the prototype
14e (Table 1) exhibits synergistic activities with antileukemia
drugs.
Results
Enantioselectivity of the Inhibition. Tranylcypromine is a
racemic mixture of (()-trans-2-phenylcyclopropyl-1-amine ·HCl
(tPCPA) that covalently inhibits the LSD and MAO enzymes.
The first question we addressed in our study was the difference,
if any, between the two tPCPA enantiomers with respect to their
inhibition activities. (+)- and (-)-tPCPA were synthesized and
their absolute configuration determined (see Supporting Infor-
mation). Biochemical analysis showed that the inversion of the
configuration has a marginal effect on inhibition of the two LSD
enzymes, whereas it is significant for inhibition of MAO B.25
In addition, the crystallographic analysis of LSD1/CoREST
inhibitor complexes highlighted a surprising feature: the two
enantiomers differ both in the binding orientation and nature
of the covalent adduct with the flavin (Figure 1-3; Scheme 1).
This feature was confirmed using the para-brominated deriva-
tives whose electron-rich substituent provided enhanced clarity
of the electron density maps (Figure 1). (-)-tPCPA engages
its carbonyl carbon in a covalent bond with the flavin N5 atom
and positions its phenyl ring in the core of the substrate-binding
pocket, stacking above the flavin ring. In contrast, the phenyl
ring of the (+)-tPCPA binds in a lateral niche of the substrate-
binding pocket and points away from the flavin ring (Figure
1-3). Furthermore, the covalent bond with the flavin N5 atom
involves the phenyl-substituted carbon of the inhibitor rather
than the carbonyl carbon as observed for the (-)-enantiomer
(Scheme 1).
Materials and Methods
Syntheses of all tested compounds are described in the Supporting
Information. Human recombinant MAO A and MAO B were
expressed in Pichia pastoris and purified as published.19 Inhibition
assays and Ki values were measured using kynuramine (MAO A)
and benzylamine (MAO B) substrates at pH 7.5 according to
published procedures (Table 1).19 Mouse recombinant LSD2 was
(5) Shi, Y.; Lan, F.; Matson, C.; Mulligan, P.; Whetstine, J. R.; Cole,
P. A.; Casero, R. A.; Shi, Y. Cell 2004, 119, 941.
(6) Karytinos, A.; Forneris, F.; Profumo, A.; Ciossani, G.; Battaglioli,
E.; Binda, C.; Mattevi, A. J. Biol. Chem. 2009, 284, 17775.
(7) Lee, M. G.; Wynder, C.; Cooch, N.; Shiekhattar, R. Nature 2005, 437,
432.
(8) Schulte, J. H.; Lim, S.; Schramm, A.; Friedrichs, N.; Koster, J.;
Versteeg, R.; Ora, I.; Pajtler, K.; Klein-Hitpass, L.; Kuhfittig-Kulle,
S.; Metzger, E.; Schu¨le, R.; Eggert, A.; Buettner, R.; Kirfel, J. Cancer
Res. 2009, 69, 2065.
(9) Hu, X.; Li, X.; Valverde, K.; Fu, X.; Noguchi, C.; Qiu, Y.; Huang, S.
Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 10141.
(10) Gu, H.; Roizman, B. J. Virol. 2009, 83, 4376.
(11) Ciccone, D. N.; Su, H.; Hevi, S.; Gay, F.; Lei, H.; Bajko, J.; Xu, G.;
Li, E.; Chen, T. Nature 2009, 461, 415.
To further analyze these different binding modes, we com-
pared our crystal structures with those previously determined
(12) Yang, Z.; Jiang, J.; Stewart, D. M.; Qi, S.; Yamane, K.; Li, J.; Zhang,
Y.; Wong, J. Cell. Res. 2010, 20, 276.
(13) Edmondson, D. E.; Binda, C.; Wang, J.; Upadhyay, A. K.; Mattevi,
A. Biochemistry 2009, 48, 4220.
(14) Yang, M.; Culhane, J. C.; Szewczuk, L. M.; Gocke, C. B.; Brautigam,
C. A.; Tomchick, D. R.; Machius, M.; Cole, P. A.; Yu, H. Nat. Struct.
Mol. Biol. 2007, 14, 535.
(20) Forneris, F.; Binda, C.; Adamo, A.; Battaglioli, E.; Mattevi, A. J. Biol.
Chem. 2007, 282, 20070.
(21) Collaborative Computational Project, Number 4. Acta Crystallogr.
1994, D50, 760.
(15) Lee, M. G.; Wynder, C.; Schmidt, D, M.; McCafferty, D. G.;
Shiekhattar, R. Chem. Biol. 2006, 13, 563.
(22) Ronzoni, S.; Faretta, M.; Ballarini, M.; Pelicci, P.; Minucci, S.
Cytometry, Part A 2005, 66, 52.
(16) Schmidt, D. M.; McCafferty, D. G. Biochemistry 2007, 46, 4408.
(17) Yang, M.; Culhane, J. C.; Szewczuk, L. M.; Jalili, P.; Ball, H. L.;
Machius, M.; Cole, P. A.; Yu, H. Biochemistry 2007, 46, 8058.
(18) Mimasu, S.; Sengoku, T.; Fukuzawa, S.; Umehara, T.; Yokoyama, S.
Biochem. Biophys. Res. Commun. 2008, 366, 15.
(23) Minucci, S.; et al. Blood 2002, 100, 2989.
(24) Di Croce, L.; Raker, V. A.; Corsaro, M.; Fazi, F.; Fanelli, M.; Faretta,
M.; Fuks, F.; Lo Coco, F.; Kouzarides, T.; Nervi, C.; Minucci, S.;
Pelicci, P. G. Science 2002, 295, 1079.
(19) Binda, C.; Li, M.; Hubalek, F.; Restelli, N.; Edmondson, D. E.; Mattevi,
A. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 9750.
(25) Zirkle, C. L.; Kaiser, C.; Tedeschi, D. H.; Tedeschi, R. E.; Burger, A.
J. Med. Pharm. Chem. 1962, 91, 1265.
9
6828 J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010

Histone Demethylase Inhibition
A R T I C L E S
Table 1. Inhibition of Selected Tranylcypromine Derivatives against LSD1, LSD2, and Monoamine Oxidases
a Enzymatic activities were measured at 25 °C using the peroxidase-coupled assay (see Materials and Methods). Errors in the determination of Ki are
within 30% of their values; nd, not determined. The Ki values were determined by steady-state competition experiments. The slow rate of irreversible
inhibition allowed these experiments to be performed by normal steady-state approaches. b LSD1 activities were assayed in 50 mM Hepes/NaOH pH 7.5
using a histone H3 peptide monomethylated at Lys4 as substrate. LSD2 activities were measured in 50 mM Hepes/NaOH pH 8.0 with the substrate
histone H3 peptide dimethylated at Lys4. c MAO A and MAO B assays were performed in 50 mM Hepes/NaOH pH 7.5, 0.5% (v/v) reduced Triton
X-100 by using kynuramine and benzylamine as substrates, respectively. d (+)-tPCPA·HCl and (+)-Br-tPCPA·HCl have 1S,2R absolute configuration,
and (-)-tPCPA·HCl and (-)-Br-tPCPA·HCl have 1R,2S absolute configuration. e (+)-cPCPA·HCl and (+)-Br-cPCPA·HCl have 1S,2S absolute
configuration, and (-)-cPCPA·HCl and (-)-Br-cPCPA·HCl have 1R,2R absolute configuration. f No detectable inhibition effect at the maximum tested
concentrations 133 µM, 267 µM, and 67 µM for 14e, 15, and 14l, respectively. Higher concentrations produced turbidity.
of LSD1 in complex with racemic tranylcypromine.17,18 In these
complexes, the inhibitor was found to bind exactly with the
same binding mode as observed with (-)-tPCPA indicating that,
when probed with a racemic mixture, LSD1 selectively binds
the (-)-enantiomer. This is consistent with (-)-tPCPA being a
slightly stronger LSD1 inhibitor than the opposite enantiomer
(Table 1). The differences in the binding modes of the two
stereoisomers become even more remarkable when compared
with the case of human MAO B. In this enzyme, (+)- and (-)-
tPCPAs adopt identical binding modes: they both form covalent
adducts with the flavin C4a (rather than N5 as found in LSD1)
atom and their phenyl moieties bind in positions that closely
resemble that observed for (+)-tPCPA in LSD119 (Figures 1
and 2B; to be published elsewhere). In summary, tPCPAs are
confirmed to be nonselective inhibitors, but key aspects of
binding differ among enzymes and inhibitor enantiomers: (i)
the cyclopropyl atom forming the covalent linkage, (ii) the flavin
ring position that is involved in the covalent bond with the
inhibitor, and (iii) the position of the inhibitor aromatic ring
with respect to the flavin.
9
J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010 6829

A R T I C L E S
Binda et al.
Figure 1. Crystallographic data on inhibitor binding to human LSD1/
CoREST. The pictures show the weighted 2Fo - Fc maps (1.2 σ contour
level, 3.0 Å resolution) for the FAD cofactor and p-bromo-(-)-tPCPA (upper
left panel), (+)-tPCPA (upper right panel), and 14e (bottom). The electron
densities clearly indicate the presence of a flavin-inhibitor covalent adduct.
The orientation of the two phenyl rings of 14e was further confirmed by
the structural analysis of a brominated derivative. Carbons are in gray,
nitrogens in blue, oxygens in red, and bromines in dark red.
Cis versus Trans Diastereoselectivity of Inhibition. The above-
described findings raised the question of the effect of changing
the configuration of only one chiral center, that is, the inhibitory
efficacy of the cis isomers of 2-phenylcyclopropyl-1-amine ·HCl
(cPCPA). Both the (+) and (-)-enantiomers of cPCPA are
covalent inhibitors of LSD1, LSD2, and MAO B, although they
exhibit weaker affinities than the corresponding trans diaster-
eomers (Table 1). The crystal structure of LSD1/CoREST with
(+)-Br-cPCPA shows an inhibitor binding mode which is
identical to that of the (-)-enantiomer of tPCPA in terms of
both orientation of the phenyl ring and covalent linkage with
the flavin N5 atom (Figure 1). These findings fully support the
notion that 2-phenylcyclopropyl-1-amines are nonspecific in-
hibitors and predict that they are likely to inhibit other flavin-
dependent amine oxidases.
Figure 2. Three-dimensional structure of (A) 14e and (B) (+)-tPCPA
binding. The pictures show the FAD (yellow), the inhibitor (pink carbons),
and the surrounding side chains (gray carbons). For Ala809, also the
backbone atoms are shown. Nitrogens are in blue, oxygens in red, and
bromines in dark red.
Toward Selective Inhibitors. We sought to design and
evaluate new molecules with improved inhibitory properties
against LSD1/2. Our strategy was to exploit the vastly different
architectures of the substrate binding sites in the amine oxidase
enzymes.17,19,20,26 LSD1 has an open cleft that hosts the H3
N-terminal tail residues through a network of very specific
interactions (Figures S8 and S9). MAO A and MAO B feature
internal cavities that are gated by surface loops. On the basis
of these structural observations, we synthesized tPCPA deriva-
tives with increasing larger substituents that contained a mix
of hydrophobic and hydrophilic groups.
Virtually all synthesized compounds inhibit either LSD1 and
LSD2 though with different efficacies. The most potent com-
pounds were 13a, 13b, and 14e with Ki values of 1-2 µM for
LSD1/CoREST and of 20-60 µM for LSD2 which are at least
100-fold better than tPCPA (Table 1). Structural analysis of their
complexes with LSD1/CoREST confirmed that they form a
covalent bond with the flavin N5 atom (Figures 1 and 3; Scheme
1). In all cases, the structure analysis revealed that the
phenylpropyl group is in a conformation identical to that ob-
served for (-)-tPCPA (Figure 1 and Figure S10). This binding
mode enables the bulky inhibitor chain to extend from the flavin
in an orientation that is roughly orthogonal to the cofactor ring.
In particular, one of the aromatic rings of the “branched” 14e
partly protrudes out of the binding cleft (Figure S9). In essence,
these inhibitors act as a sort of plug that fills the substrate-
binding cleft of LSD1 and is “anchored” to the flavin ring
through a covalent linkage.
Having identified relatively strong LSD1/CoREST inhibitors,
we probed them for their inhibitory power against human MAO
A and MAO B. First of all, the branched 14e and 14l inhibitors
did not inhibit MAO B. This is consistent with the MAO B
three-dimensional structure that shows two adjacent flat cavities
that would be too small to accommodate these bulky mol-
ecules.19 Furthermore, smaller linear compounds such as 13a
(26) Son, S. Y.; Ma, J.; Kondou, Y.; Yoshimura, M.; Yamashita, E.;
Tsukihara, T. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 5739.
9
6830 J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010

Histone Demethylase Inhibition
A R T I C L E S
cavity which is gated by a flexible loop.26,28 This property
apparently enables the enzyme to bind relatively large molecules
such as 14e and its cis diastereomer 15 (Table 1). These findings
emphasize the adaptability of the MAO A active site as
compared to that of MAO B, which is consistent with chemical
labeling studies.29
Biological Evaluation. The good selectivity (and water
solubility) of 14e identified this compound as a logical choice
for an initial evaluation of the biological activities of the
synthesized molecules. We chose a cellular model of acute
promyelocytic leukemia (APL) since the pathogenesis of this
disease involves the aberrant activities of several chromatin
modifiers (histone deacetylases and methyltransferases). These
abnormalities are associated with a block of the normal
hematopoietic differentiation at the promyelocytic stage. Phar-
macological concentrations of retinoic acid, acting on the fusion
protein PML-RAR (originated by the t15;17 chromosomal
translocation that is the trigger of the disease), release the
differentiation block and lead to clinical remission of the
patients.30
Initially, we checked whether 14e modulated its target in the
APL-derived NB4 cell line. We treated the cells with various
concentrations of the compound for various times. After a short
(6 h) treatment, we observed a selective, dose-dependent
increase in H3-Lys4 dimethylation (Figure 4A). At later time
points (12 h), the increase in Lys4 dimethylation showed a less
evident dose-dependence that could be explained by the
irreversible mechanism of action of the inhibitor. We could not
measure significant changes in the levels of H3-dimethylLys9,
whereas after 24 h, we observed an increase in histone H4
acetylation (Figure S11), in line with the postulated cross-talks
between LSD1 and deacetylases.4 Next, we measured the effect
of 14e on cell growth. Singly, the compound did not affect the
growth of NB4 cells up to 7 days of treatment, even at a
concentration of 2 µM, which was sufficient to achieve the
highest target modulation. However, LSD1 inhibition exhibited
a striking synergism with retinoic acid growth inhibitory effect:
at low retinoic acid concentrations (10 nM) that affect only
mildly cell growth, co-treatment with 14e led to a >10-fold
stronger effect, with a reduction in cell number even stronger
than that observed with high retinoic acid concentrations given
alone (1 µM, Figure 4B). In part, this phenomenon is due to an
enhanced apoptosis rate (Figure S12). The effect on cell growth
led us to hypothesize that 14e could also affect differentiation
of NB4 cells. We therefore measured differentiation by appear-
ance of surface markers (such as CD11b), and by morphological
features. 14e alone led to a significant enhancement of CD11b
expression (from 10% to 40% of CD11b positive cells) with
limited morphological changes. It also potentiated the dif-
ferentiating effect of retinoic acid at all concentrations (Figure
S13), greatly enhancing CD11b expression at low retinoic acid
concentrations (Figure 4C). To verify that the effect was not
limited to a cell line, we took advantage of the availability of
primary murine APL blasts derived from an animal model.23
14e exhibited a strong effect on the growth of leukemic blasts
in a semisolid medium, even when given alone (Figure 4D).
This correlated with enhanced differentiation as gathered from
Figure 3. Schematic diagram of the inhibitor binding sites; (A) 14e, and
(B) (+)-tPCPA. H-bonds are shown with dashed lines. Consistent with the
data of Yang et al.17 on racemic tPCPA, the electron densities for 14e (Figure
1), (-)-tPCPA, and 13b do not rule out the possibility that at least part of
the bound inhibitor molecules forms a cyclic adduct that in addition to the
depicted bond between flavin N5 and the inhibitor carbonyl carbon contains
also a covalent linkage between the flavin C4a and the phenyl-substituted
carbon of the inhibitor.
Scheme 1
and 13b displayed good Ki values against MAO B but did not
induce spectral perturbations of the flavin, indicating that they
are noncovalent inhibitors. This may reflect a binding mode that
does not position the cyclopropyl unit in the proper orientation
with respect to the flavin to promote formation of a covalent
linkage.27 The data on MAO B inhibition contrast with those
measured on MAO A. All tested compounds were effective
covalent inhibitors of this enzyme with measured Ki values in
the 1-50 µM range. MAO A has a large 650 Å3 active site
(28) De Colibus, L.; Li, M.; Binda, C.; Lustig, A.; Edmondson, D. E.;
Mattevi, A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 12684.
(29) Hubalek, F.; Pohl, J.; Edmondson, D. E. J. Biol. Chem. 2003, 278,
28612.
(27) Culhane, J. C.; Wang, D.; Yen, P. M.; Cole, P. A. J. Am. Chem. Soc.
2010, 132 (9), 3164.
(30) Botrugno, O. A.; Santoro, F.; Minucci, S. Cancer. Lett. 2009, 280,
134.
9
J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010 6831

A R T I C L E S
Binda et al.
Figure 4. Biological evaluation of 14e. (A) 14e modulates histone methylation in whole cells. NB4 cells were treated for 6 h with increasing concentrations
of 14e (0.25-2 µM) or vehicle (NT), and then whole cell lysates were analyzed by Western blot against the indicated histone post-translational modifications
(H3K4me2, dimethylation level of Lys4 of histone H3; H3K9me2, dimethylation level of Lys9 of histone H3; AcH4, lysine acetylation levels of histone
H4). Total H3 served as loading control. (B) 14e synergizes with retinoic acid (RA) in inhibiting cell growth. NB4 cells were treated with increasing
concentrations of retinoic acid (10 nM, 100 nM, and 1 µM), in the absence or in the presence of 14e (2 µM). At the indicated time points, cells were counted
by Trypan blue exclusion. NT, untreated cells (vehicle only). A representative experiment is shown. (C) 14e enhances expression of the differentiation
marker CD11b in NB4 cells. Cells were treated as in panel B. After 48 h, the CD11b marker was analyzed by FACS. (D) 14e inhibits cells growth of murine
APL blasts. Primary leukemic cells (20 000 cells) from spleen were seeded in triplicate in semisolid medium (MethoCult 3434) in the presence of increasing
concentrations of 14e (0.25 and 2 µM) or vehicle (NT). After 7 days, colonies (C.F.U.) were counted. (E) 14e induces differentiation of murine APL blasts.
APL blasts were seeded in semisolid medium as in panel D. At the end of the assay, colonies were pooled, and cells, cyto-spun on glass slides, were stained
(May Grunwald-Giemsa). Percentage of differentiated cells is indicated.
morphological features of the single cells (Figure 4E) and
colonies (that show a spread pattern as opposed to the compact
structure of colonies derived from APL blasts; Figure S14) as
well as induction of genes associated with differentiation (Figure
S15).
cyclopropylamines and normal substrates is that the former can
react to generate stable covalent adducts.
Independently from the exact mechanism, it is clear that
inhibition occurs with cyclopropyl ring-opening. An insightful
finding is that (+)- and (-)-tPCPA generate positionally distinct
inhibitor-flavin linkages, possibly via differing mechanistic
pathways (Scheme 1). This feature contrasts with MAO B, in
which the two enantiomers exhibit the same orientation and
identical C4a adduct.19 A tentative explanation for the N5 versus
C4a difference between LSD1 and MAO B is that the MAO B
adduct initially occurs as an N5 adduct that undergoes a
migration to the C4a atom as found in model flavin reactions.32
Taken together, these data indicate that, although the ( and
cis/trans stereoisomers are unspecific LSD and MAO inhibitors,
the effect of chirality must not be neglected.
Another crucial aspect revealed by these studies is that most
of the investigated molecules inhibit both MAO and LSD
enzymes. It appears that the common evolutionary origin and
folding topology in these otherwise functionally divergent
enzymes manifest themselves in similar inhibition and binding
properties. However, the “branched” set of compounds does
show some selectivity. 14e and 14l are unable to inhibit MAO
Discussion
Our investigation of 2-phenylcyclopropylamines has led to
the discovery of novel compounds with improved inhibitory
activities against LSD1/2, unraveling several key features about
their inhibition mechanism. The prevailing view in the current
literature is that tPCPA reacts by transferring one electron to
the flavin, followed by recombination of the radicals resulting
from homolytic cyclopropane ring cleavage to form the covalent
adduct (Scheme 1).31 In the course of our studies, we have
observed that inhibition of both MAO B and LSD1 generates
hydrogen peroxide with a H2O2/protein molar ratio of about
0.3-0.7. Thus, the reaction is partly uncoupled: occasionally,
the flavin is reoxidized and/or the oxidized tPCPA is released
before formation of the covalent linkage can occur. This lends
support to the idea that 2-phenylcyclopropylamines behave like
canonical amine oxidase substrates that are oxidized by the
enzyme, resulting in H2O2 generation. The difference between
(32) Walker, W. H.; Hemmerich, P.; Massey, V. HelV. Chim. Acta 1967,
50, 2269.
(31) Silverman, R. B. J. Biol. Chem. 1983, 258, 14766.
9
6832 J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010

Histone Demethylase Inhibition
A R T I C L E S
B, representing per se a valuable property because there is only
a handful of known inhibitors that are so effective in the MAO
A versus MAO B selectivity.13 A similarly bulky tPCPA
derivative recently described exhibits similar properties.33 The
inability of the “branched” compounds to discriminate between
LSD1 and MAO A likely reflects the absence of specific
directional inhibitor-protein interactions (i.e., H-bonds) as
indicated by the crystal structure of 14e bound to LSD1/
CoREST. Nevertheless, molecules such as 14e represent a step
forward the development of selective inhibitors. 14e is not only
an effective inhibitor of LSD1 in Vitro, but it also exhibits
relevant biological activities in cellular models. The compound
strongly enhances the efficacy of retinoic acid on growth and
differentiation of acute promyelocytic leukemia cells, including
primary murine APL blasts. This result demonstrates that LSD1
inhibitors have potential to both alter differentiation via
alterations of the chromatin functional state and act synergisti-
cally with drugs that block and/or reduce tumor progression.
Abbreviations: LSD, lysine-specific histone demethylase;
MAO, monoamine oxidase; tPCPA, trans-2-phenylcyclopropyl-
1-amine as hydrochloride salt; cPCPA, cis-2-phenylcyclopropyl-
1-amine as hydrochloride salt; APL, acute promyelocytic leu-
kemia.
Acknowledgment. The support by the Italian Ministry of
Science (PRIN08 and FIRB), Ministry of Health, Fondazione Roma,
Fondazione Cariplo, EU (Epitron), and the Italian Association for
Cancer Research is gratefully acknowledged.
Supporting Information Available: Statistics about crystal-
lographic procedures, experimental details about the biological
evaluation of the inhibitors, detailed description of the inhibitor
syntheses, and complete reference 23. This material is available
free of charge via the Internet at http://pubs.acs.org.
(33) Ueda, R.; Suzuki, T.; Mino, K.; Tsumoto, H.; Nakagawa, H.;
Hasegawa, M.; Sasaki, R.; Mizukami, T.; Miyata, N. J. Am. Chem.
Soc. 2009, 131, 17536.
JA101557K
9
J. AM. CHEM. SOC. VOL. 132, NO. 19, 2010 6833