Thieno[3,2-b]pyrrole-5-carboxamides as New Reversible Inhibitors of Histone Lysine Demethylase KDM1A/LSD1. Part 1: High-Throughput Screening and Preliminary Exploration

Article abstract of DOI:10.1021/acs.jmedchem.6b01018

Lysine specific demethylase 1 KDM1A (LSD1) regulates histone methylation and it is increasingly recognized as a potential therapeutic target in oncology. We report on a high-throughput screening campaign performed on KDM1A/CoREST, using a time-resolved fluorescence resonance energy transfer (TR-FRET) technology, to identify reversible inhibitors. The screening led to 115 hits for which we determined biochemical IC₅₀, thus identifying four chemical series. After data analysis, we have prioritized the chemical series of N-phenyl-4H-thieno[3, 2-b]pyrrole-5-carboxamide for which we obtained X-ray structures of the most potent hit (compound 19, IC₅₀ = 2.9 μM) in complex with the enzyme. Initial expansion of this chemical class, both modifying core structure and decorating benzamide moiety, was directed toward the definition of the moieties responsible for the interaction with the enzyme. Preliminary optimization led to compound 90, which inhibited the enzyme with a submicromolar IC₅₀ (0.162 μM), capable of inhibiting the target in cells.

Full text of DOI:10.1021/acs.jmedchem.6b01018

Subscriber access provided by University of Florida | Smathers Libraries
Article
Thieno[3,2-b]pyrrole-5-carboxamides as New Reversible
Inhibitors of Histone Lysine Demethylase KDM1A/LSD1. Part
: High Throughput Screening and Preliminary Exploration
1
Luca Sartori, Ciro Mercurio, Federica Amigoni, Anna Cappa, Giovanni Fagá, Raimondo
Fattori, Elena Legnaghi, Giuseppe Ciossani, Andrea Mattevi, Giuseppe Meroni, Loris
Moretti, Valentina Cecatiello, Sebastiano Pasqualato, Alessia Romussi, Florian Thaler, Paolo
Trifiró, Manuela Villa, Stefania Vultaggio, Oronza A. Botrugno, Paola Dessanti, Saverio
Minucci, Elisa Zagarrí, Daniele Carettoni, Lucia Iuzzolino, Mario Varasi, and Paola Vianello
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.6b01018 • Publication Date (Web): 10 Feb 2017
Downloaded from http://pubs.acs.org on February 11, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Experimental Oncology
Zagarrí, Elisa; Fondazione FADOI
Carettoni, Daniele; Axxam, Screening Technologies
Iuzzolino, Lucia; Axxam, Screening Technologies
Varasi, Mario; IFOM - the FIRC Institute of Molecular Oncology Foundation,
Experimental Therapeutics Unit
Vianello, Paola; IFOM - the FIRC Institute of Molecular Oncology
Foundation, Experimental Therapeutics Unit
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Thieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamides as New
Reversible Inhibitors of Histone Lysine
Demethylase KDM1A/LSD1. Part 1: High
Throughput Screening and Preliminary
Exploration
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
,†,§
†,#,±
†
†,#
†,§
Luca Sartori,* Ciro Mercurio,
Federica Amigoni, Anna Cappa, Giovanni Fagá,
†
,§
†
‡
‡
Raimondo Fattori, Elena Legnaghi, Giuseppe Ciossani, Andrea Mattevi, Giuseppe
†,§
†¥
Θ,
∥
Meroni,
Loris Moretti,
Valentina Cecatiello,
Sebastiano Pasqualato,∥ Alessia
†
,#
†,§
†,§
†,§
†,#
Romussi, Florian Thaler, Paolo Trifiró, Manuela Villa, Stefania Vultaggio, Oronza
†
†
†,
†
⊥
A. Botrugno, Paola Dessanti, Saverio Minucci, Elisa Zagarrí, Daniele Carettoni,^ Lucia
†
,§
,†,#
Iuzzolino,^ Mario Varasi, Paola Vianello*
Department of Experimental Oncology, Academic Drug Discovery, European Institute of
Oncology, Via Adamello 16, 20139 Milan, Italy
Department of Biology and Biotechnology, University of Pavia, Via Ferrata 1, 27100
Pavia, Italy
ꢀ Crystallography Unit, Department of Experimental Oncology, European Institute of
†
‡
∥
Oncology, Via Adamello 16, 20139 Milan, Italy
Θ IFOM, Fondazione Istituto FIRC di Oncologia Molecolare, Via Adamello 16, 20139
Milano, Italy
⊥ꢀDepartment of Biosciences, University of Milan, Via Celoria, 26, 20133 Milan, Italy
^Axxam, via A. Meucci 3, 20091 Bresso (MI), Italy
ACS Paragon Plus Environment

Page 3 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
± Genextra Group, DAC s.r.l., Via Adamello 16, 20139 Milan, Italy
KEYWORDS: epigenetics; lysineꢀspecific demethylaseꢀ1; KDM1A; LSD1; TRꢀFRET; HTS;
Reversible inhibitors.
ABSTRACT
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Lysine specific demethylase 1 KDM1A (LSD1) is one regulator of histone methylation and it
is increasingly recognized as a potential therapeutic target in oncology. We report on a highꢀ
throughput screening campaign performed on KDM1A/CoREST, using a time resolved
fluorescence resonance energy transfer (TRꢀFRET) technology, to identify reversible
inhibitors. The screening led to 115 hits for which we determined biochemical IC , thus
50
identifying 4 chemical series. After data analysis, we have prioritized the chemical series of
Nꢀphenylꢀ4Hꢀthieno[3, 2ꢀb]pyrroleꢀ5ꢀcarboxamide for which we obtained Xꢀray structures of
the most potent hit (compound 19, IC50 = 2.9µM) in complex with the enzyme. Initial
expansion of this chemical class, both modifying core structure and decorating benzamide
moiety, was directed towards the definition of the moieties responsible for the interaction
with the enzyme. Preliminary optimization brought to compound 90 which inhibited the
enzyme with a submicromolar IC (0.162µM), capable to inhibit the target in cells.
50
INTRODUCTION
Histone Lysine demethylases are responsible for postꢀtranslational modifications of the
1
,2
histone tails, representing some of the most important partners in eukaryotic chromatin
3
regulation. Among them KDM1A (LSD1), homologous to polyamine oxidase (PAO) and
4
monoamine oxidase (MAO) A and B, catalyses monoꢀ and dimethyl histone 3 lysine 4
4
(H3K4me1, H3K4me2) and H3K9 demethylation. High expression of KDM1A has been
5
6
7
reported in acute myeloid leukemia , lymphoid neoplasm , prostate cancer, ERꢀnegative
8
9
breast cancer and neuroblastoma. Moreover its high expression correlates with poor
7
8, 10
prognosis in several tumor types, such as breast cancer, prostate cancer,
non small cell
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
1
12
lung cancer and hepatocellular carcinoma. A role of KDM1A into conferring stem cellꢀ
13
like characteristics in breast cancer has also been proposed.
From the discovery of tranylcypromine (1 in Figure 1) as an irreversible inhibitor of
1
4,15
16ꢀ20
KDM1A,
a number of analogues bearing the cyclopropyl moiety has been described
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
and three clinical candidates are currently under investigation, two of them being disclosed,
2
1
22,23
ORY1001/RGꢀ6016 as compound 6 and GSK2879552
as compound 7, whereas
24
INCB059872 from Incyte is still structurally unknown. We have also contributed to the
25
field with the recent identification of a potent oral selective inhibitor [(1S,2R)ꢀ3, Fig.1] able
to significantly prolong survival time in a murine promyelocytic leukemia model.
Several examples of reversible KDM1A inhibitors have been reported in the literature (Fig.
1
6ꢀ20, 25ꢀ28
1),
the development of which appears to be in the preꢀclinical stage. A highꢀ
throughput screening (HTS) using Timeꢀresolved fluorescence energy transfer assay (TRꢀ
27
FRET) on KDM1A has also been described, although no inhibitors were reported. Here we
describe the identification of four chemical series that are active on KDM1A through the
screening of a representative subset of our chemical collection (ca. 34000 commercially
available compounds) and the process used to select the thieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamide
series for further development. A very preliminary exploration allowed us to obtain
submicromolar inhibitors. A subsequent extensive medicinal chemistry effort has led to a
further optimization of the series, up to the identification of highly selective nanomolar
inhibitors, endowed with high potency in cellular assays. This in depth exploration is the
28
subject of an accompanying paper.
Insert Figure 1
RESULTS AND DISCUSSION
The Chemical Collection
ACS Paragon Plus Environment

Page 5 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The starting point for any HTS activity is a proper compound collection. To perform HTS for
our projects, we built a structurally diverse collection, starting from five millions of
commercially available compounds, filtering out molecules on the basis of their physicoꢀ
2
9,30
chemical properties, removal of undesired features,
diversity and chemotype
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
representativeness and we obtained a subset of 196,000 compounds, which were purchased.
This library was designed to represent each selected chemotype (scaffoldꢀbased) with a
sufficient number of close analogues to obtain preliminary SARs for each hit identified
during HTS. Although the chemical diversity is limited, the hit identification process is
reinforced with the reduction of false positive rate. The analogues would help in deciding if
positive (or negative) results are true values simply by looking at the whole activity range
across the cluster. At the same time the higher the number of cluster members (close
analogues), the lower the “chemical diversity”, independently from chemical diversity space
definition.
31
In order to maximize signal over noise ratio , and to minimize the total number of assayed
compounds still maintaining high number of confirmed hits, the collection has been divided
into 5 subsets. Four of them, containing 8,500 chemical diversity representatives each, for a
total of 34,000, have been selected to be assayed in the first step of HTS. Close analogues of
hits are pooled from the fifth set (Full Diversity Set), represented by the remaining 162,000
compounds. These analogues are used during the hit expansion step. All compounds are
plated into 384 well plates, and are grouped by similarity, thus generating a sort of
“
compound cartridges” ready for screening and allowing hit expansions with a minimal
effort.
Quantitative HighꢀThroughput Screening
Step1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The screening was conducted on 34000 compounds, the representative subset of our chemical
collection obtained from commercially available generic libraries, at the fixed concentration
of 10 ꢀM (see Figure 2). The KDM1A inhibiting activity was determined using a TRꢀ
FRET(time resolved fluorescence resonance energy transfer) assay. The assay was performed
at room temperature (RT) in 384 well plates in a final volume of 30 µL of assay buffer (Tris
HCl 50 mM pH 8.8, NaCl 50 mM, DTT 1 mM, Tweenꢀ20 0.01%, proteaseꢀfree BSA 0.01%
w/v) at the enzyme (KDM1A/CoREST) and substrate (histone H3K4 monomethylated)
concentrations of 0.079 nM and 51 nM respectively. After 60 min at RT, the reaction was
stopped with tranylcypromine at the concentration of 300 ꢀΜ. The primary screening
campaign was completed in 5 days and the assays provided an overall Robust z prime factor
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(RZ’) of 0.85, demonstrating the robustness and the quality of the obtained data. The
threshold for hit selection was set at 15% of activity, which corresponded to mean plus three
times the standard deviation computed on neutral controls (maximal signal in absence of
inhibitor). Compounds with high autoꢀfluorescence and therefore interfering with the assay
were discarded. Applying these criteria, 320 out of 34000 compounds were identified as hits,
corresponding to 0.95% hit rate. The 320 primary hits were then tested at 30 ꢁM, 10 ꢁM and
3
ꢁM with interꢀplate triplicates. 113 out of 320 retested compounds were confirmed with a
confirmation rate of 35 % and with an overall RZ’ of 0.7. This result was consistent with the
threshold applied in primary screening for hit selection. Indeed, the majority of compounds
that did not confirm their activity were close to the primary screening threshold, whereas
most of the compounds showing high activity in primary screening, belonged to the group of
confirmed hits. Using these compounds as templates, substructure and similarity searches
32
against the fifth set of the collection was performed with ad hoc developed Pipeline Pilot
protocols. The protocols performed iteratively substructure and similarity searches (See
Experimental Section for details) for each of the reconfirmed hits and then collected and
ACS Paragon Plus Environment

Page 7 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
merged all the results, thus producing a series of plates giving us the opportunity to screen an
enriched set of compounds similar to those obtained at 10 ꢁM.
Step 2
19 mother plates were selected from the Full Diversity set to be tested in triplicate at 10 ꢁM
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
to identify hits from analogous compounds present in the library. An overall RZ’ of 0.81 was
obtained and 70 compounds among the 6080 tested were found to be active according to the
same threshold of the primary screening, thus producing a hit rate of 1.15% based on all
compounds and of 8% based on close analogs.
Step 3
Finally, a dose response experiment was conducted on 115 selected compounds. The
compounds were tested in doseꢀresponse at 8 concentrations ranging from 30 ꢀM to 0.1 ꢀM
with intraꢀplate quadruplicate data points. The overall RZ’ was of 0.76 and 73 active
compounds were identified. Among them 27 compounds showed an IC50 higher than 30
micromolar, 32 compounds showed an IC50 between 10 and 30 micromolar and 14
compounds showed and IC50 lower than 10 micromolar. Overall four chemical series and 14
singletons were identified (Table 1).
Insert Figure 2
HTS Results
The four main chemical series identified are reported in Table 1. They were evaluated on the
basis of novelty, preliminary SAR evidence, chemical feasibility and Structure Based Drug
Design (SBDD) amenability. Our interest was thus focused on compounds 19 and 21.
Compound 19 (and its furo analogue) showed an interesting preliminary SAR having 20
analogues with IC50 ranging from 10 to 30 micromolar. Compound 21, although belonging to
a much smaller cluster, has a low molecular weight, comparable to a fragment, with a single
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
digit micromolar IC50, and was considered for further investigation, taking into account its
high potential for derivatization.
Insert Table 1
The chemical series prioritized, thieno(furo)pyrroles 19 and spiro derivatives 21, were
selected for further studies to better define their biochemical mechanism of action and to
verify their amenability for Xꢀray crystallography.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Orthogonal assay, reversibility and selectivity
Compounds 19, 21 and tranylcypromine 1 were assessed with an orthogonal KDM1A assay
based on Horseradish Peroxidase (HRP) assay to reduce the probability to pursue false
positives and to obtain an independent confirmation of their histone demethylases activity. As
reported in Table 2, they inhibited KDM1A/CoREST complex also in the HRP assay
confirming to be genuine KDM1A inhibitors. Different experimental assay conditions in
terms of enzyme, substrate concentrations and incubation time used in the two assays justify
the differences between the TRꢀFRET and HRP reported IC s. Similar differences were
5
0
observed between HTS and reconfirmation data.
Insert Table 2
Since HRP assay is a continuous assay well suited to perform enzymatic inhibition kinetic
analysis, it was used to verify reversibility and biochemical mechanism of action of
compounds 19 and 21. The reversibility was assayed by measuring the recovery of enzymatic
activity after a rapid and large dilution of the enzymeꢀinhibitor complex in presence of the
substrate. Consistently with its irreversible mechanism of action, incubation of 1 resulted in
complete inactivation of the enzyme, which was not recovered after dilution with the assay
buffer containing the substrate (Figure 3A). On the contrary, when KDM1A/CoREST was
incubated with 19 and 21, the recovery of enzymatic activity after dilution was almost
complete, confirming for both compounds the reversibility of inhibition.
ACS Paragon Plus Environment

Page 9 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
KDM1A/CoREST enzymatic kinetic was assessed at different concentrations of substrate and
inhibitor. Initial velocities were fitted to the equations for competitive, nonꢀcompetitive and
uncompetitive inhibition, using SigmaPlot (Systat Software, San Jose, CA). As reported in
Figure 3BꢀC, LineweaverꢀBurk plot indicates a non competitive mechanism for 19 and a
competitive mechanism for 21. In particular 19 showed to be a non competitive type inhibitor
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3
3
(
α=1.1) with same affinity for free KDM1A/CoREST complex and the KDM1A/CoREST
enzyme substrate complex, with a Ki of 43 ꢀM, whilst 21 showed to be a competitive
inhibitor, with a Ki of 48.4 ꢀM.
Insert Figure 3
Finally, we investigated enzyme selectivity of 19 and 21 against KDM1B, MAOA and
MAOB using tranylcypromine 1 as reference. Both compounds showed high selectivity in the
secondary assay, as illustrated in Table 3.
Insert Table 3
Structural analysis
The Nꢀ[3ꢀ(methoxymethyl)phenyl]ꢀ4ꢀmethylꢀthieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamide (19) was
successfully crystallized in complex with KDM1A/CoREST construct, see Materials and
Methods for details, and the Xꢀray diffraction quality allowed us to solve the complex at a
resolution of 3.2Å, adequate for structural analysis (see figure 4A and Table 1S in Supporting
Information). We also attempted to produce KDM1A/CoREST crystal in complex with
compound 21 but all the trials failed. To our knowledge 19 is the first small molecule in close
contact with FAD, reversibly inhibiting KDM1A, for which the experimentally determined
structure of the complex has been reported.
Key interactions
The ligand is located within the catalytic site of the KDM1A enzyme, as seen in Figure 4A,
with a relatively planar conformation due the electronic delocalization between the 2 ring
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
systems bridged by the amide group. The inhibitor interaction with the target is mainly
hydrophobic with no visible polar contacts. Its core, the thienopyrrole moiety, is caged
between residues Val333, Thr335, Leu663, Tyr765 and Thr814. The flavin ring represents
the back wall of such cage and the geometry of its contact with the bicyclic system is
compatible with an aromaticꢀaromatic interaction. The second ring system, an aryl group, is
packed between residues Val333, Phe542 and Trp699. The exit vector of the meta substituent
on the second ring system points toward Gln358, suggesting a possible interaction region.
Stability of the complex and its interactions have been evaluated using Molecular Dynamics
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(MD) technique, propaedeutic to structure base drug design setup (see next paragraph).
Molecular details of the nonꢀcompetitive inhibition mechanism for 19 are revealed by the
comparison of this crystallographic structure with the one of KDM1A in complex with
34
histone H3 peptide (2V1D) depicted in figure 4B. The superposition of the two structures
shows a very partial overlap of the thienopyrrole molecule with the atoms of the peptide, only
the oxygen of the methoxymethyl superimpose with the histone, in a non effective manner. In
particular, the ligand is proximal to the region where the dimethylated lysine interacts with
35
the target to undergo the demethylation reaction. Further comparison of the two KDM1A
complexes (2V1D vs internal) highlights subtle differences of bindingꢀsite residue
conformations. The two binding sites are quite identical with the exception of Trp699, which
is present with different rotamers. The different conformation of this amino acid is due to the
induced fit effect of thienopyrrole ligand binding to the target.
Insert Figure 4
Structure Based Drug Design
To support the development of this class of inhibitors, a specific molecular docking protocol
was developed to better reproduce the binding mode of the thienopyrrole chemotype. A
procedure was developed by combining docking programs, clustering techniques and energy
ACS Paragon Plus Environment

Page 11 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
estimates (see Experimental Section for details). The KDM1A protein conformations used for
the docking procedures were those obtained from internal Xꢀray trials, where the
conformation of Trp699 has shifted with respect to its reported position. This effect is
induced by the binding of the ligand and in particular is due to the phenyl ring position
(Figure 4B). The use of these protein structures and the docking protocols gave satisfactory
results to reproduce the Xꢀray crystallographic conformation of 19 and produced reasonable
poses.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The conformational space of the ligand inside the target was explored with more than one
36
37
38
methodology: Autodock , Plants and Glide produced multiple docking poses. They were
39
clustered with the ACIAP (Autonomusꢀhirerarchical clustering), a method based on their
distances as RMSD between conformations. Finally, one representative pose per family was
40
submitted to BEAR (Binding estimation after refinement) evaluation.
To identify binding site characteristics, such as hydrophobicꢀinteraction pattern and hydrogen
bonding propensity, we chose to explore the target protein with a gridꢀbased approach.
3
6
Autogrid software was used to build different types of energetics maps applying various
atomic probes and the visualization of grid points below a certain threshold allowed us to
scan the binding site for favorable interaction regions. Of all obtained maps, the hydrophobic
pattern was quite informative and, associated to the electronegative areas, gave useful
suggestions for the development of the thienopyrrole series. As seen in Figure 5 the ligand is
buried into the hydrophobic region.
Insert Figure 5
The analysis of the electrostatics maps reveals two negatively charged regions, one in the
upward direction with respect to the thienopyrrole, where the residues Asp379 and Glu383
are found, and one downward, corresponding to residues Asp559, Asp560 and Glu563. In
particular, the former corresponds to the site where the guanidine group of the Arg8 of the
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
histone substrate interacts with the enzyme, see Figure 5B. Thus, we decided to grow the
chemical class elongating the substituent in position meta of the anilide in search of a
favorable interaction regions along the upward direction. In addition, the distance between
the benzene carbon in para position and the electronegative region is approximately 12
Angstroms. Ultimately these modeling suggestions prompted us to elongate the dimethylene
chain of compound 19 into a chain ending with a positively charged group.
Preliminary SAR from HTS
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Preliminary SAR indications referred to compound 19 can be obtained directly from the HTS
data (Table 4): thienopyrrole bicyclic ring is favoured to the furopyrrole analogue (19 vs 25,
22 vs 29); meta substitution on phenyl ring results in more potent compounds than the ortho
analogues (24 vs 31, 22 vs 33); amido nitrogen substitution with methyl is detrimental for
activity (38 and 40); electron donor groups are preferred to halogens or to electron
withdrawing groups.
Insert Table 4
Explorative SAR
A small chemical exploration to further investigate the series represented by compound 21
has been performed, but none of the synthesized compound reported any inhibitory activity
(data not shown) and the explorative activity concluded. For thienopyrroles derivatives
instead we initiated a more comprehensive SAR study.
We approached the modification on four areas (see Figure 6):
A: Different heterocycles to replace thienopyrrole;
Small substituents on the thienopyrrole;
B: Different chains replacement of the Nꢀmethyl group of the thienopyrrole;
C: Isosters of the amido function;
D: Variations of the meta substituent of the phenyl group.
ACS Paragon Plus Environment

Page 13 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Insert Figure 6
As demonstrated by Xꢀray structure, A and B regions are involved in binding/interactions
with flavin adenine dinucleotide (FAD), thus they have been explored together.
Region D is involved in induced fitting of Trp699 and carries the exit vector toward putative
interactions with negatively charged region.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Region C has been explored as linker replacement.
Synthetic Chemistry
The appropriate heterocyclic aldehydes (102a−l) were converted into the azido derivatives
(103aꢀl), by reaction with ethyl 2ꢀazidoacetate and potassium ethylate in ethanol. Subsequent
cyclization occurred by refluxing in xylene, providing 104a−l, which reacted with methyl
iodide and sodium hydride in DMF to give 105a−l. Ester function could either be directly
transformed into amide by reaction with aniline and LiHMDS in THF, or hydrolyzed to the
acids as in 106c in LiOH, which in turn reacted with aniline previous activation to acyl
chloride with thionyl chloride (see Scheme 1).
Insert Scheme 1
Halogenation of thienopyrrole 23 was performed according to Scheme 2. Chlorination with
4
1
Nꢀchlorosuccinimide as reported went exclusively in position 6 on the pyrrole ring
affording 6ꢀchloroꢀ4ꢀmethylꢀNꢀphenylꢀthieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamide (56) with 26%
4
2
yield. Bromination with Nꢀbromo succinimide has been reported on a similar substrate , but
in our case it gave an inseparable mixture of the major 2ꢀbromo derivative 44 and the 6ꢀ
bromo analogue 57. Compound 44 was alternatively prepared according to Scheme 1,
starting from 5ꢀbromothiopheneꢀ2ꢀcarbaldehyde (102i).
NIS gave a mixture of 6ꢀIodo (58), 2ꢀIodo (59) and 2,6ꢀdiIodo (60). Fluorination of the
thiophene in position 2 (61) was achieved with Nꢀflurobenzenesulfonimide (NFSI) using
41
LDA as a base. Attempts with NFSI and butyllithium failed to give the desired product.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Insert Scheme 2
The synthesis of 6ꢀmethoxymethyl and 6ꢀmethyl thienopyrroles (62 and 63 respectively) and
the synthesis of 2ꢀ and 6ꢀ(dimethylaminomethyl) thienopyrroles (64 and 65 respectively) are
reported in Supporting Information.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Thienopyrrole 104a was either treated with alkyl halide, benzyl chloride, or phenylsulphonyl
chloride, in presence of NaH in DMF, to give the intermediates 105m−q, or was treated with
phenylboronic acid and Cu(OAc)
2
in pyridine, to give 105p. Subsequent hydrolysis to
106m−r, conversion into the corresponding acyl chloride by treatment with thionyl chloride,
and reaction with the required amine, gave final amides 66ꢀ91.
Insert Scheme 3
2
Compounds 92ꢀ94 and the anilines ArNH in Scheme 3, when not commercially available, in
particular those used to generate 90 and 91, were prepared according to the procedures
reported in Supporting Info.
A: 1_ Variation of the Heterocycle
In order to verify if the interaction with FAD in the binding pocket could be mimicked by
other heterocycles, we replaced the bicycle ring with several other hetero aromatic systems.
As illustrated in Table 5, only the regioisomer 45 resulted almost equipotent to 23,
furopyrrole 28 was more than three times less potent, whereas pyrrolothiazoles (46 and 47),
pyrroloimidazoles (48 and 50), pyrrolopyrrole (49), pyrrolopyrazole (51), thienothiophene,
thienoimidazole (52, 93) or benzothiophene completely abolished the inhibitory activity.
Insert Table 5
A: 2_ Ring Substitution
We approached a limited exploration on all the accessible positions of the bicyclic system
(Table 6). None of the substitutions demonstrated the opportunity to decorate the core
scaffold. Among the residues selected to substitute the pyrrolic nitrogen, only ethyl in
ACS Paragon Plus Environment

Page 15 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
compound 66 improved potency of almost three times when compared to the methyl
analogue 23, whilst the other chains were all detrimental for the activity (see 2ꢀaminoethyl,
phenyl, phenylsulfonyl and benzyl in compounds 68ꢀ71). The reason for this inactivity can be
explained by steric hindrance of substituents in position 4 of thienopyrrole moiety: anything
larger than ethyl group in fact goes toward residues Leu663, Tyr765, as suggested in
comments to Fig 6A. Analogously, the proximity of the enzyme wall to the ligand does not
allow any atom replacing hydrogens in position 2, 3 and 6 of the thienopyrrole, neither small
alkyls, nor halogens of various type (44, 53−56, 58ꢀ65).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Insert Table 6
C: Amide replacement
In search of new functional groups substituting the amide spacer, we made some isosteric
replacements (ester, reverse amide, oxime) without any activity improvement (data not
shown).
D:1_Phenyl replacement
The phenyl group was replaced by several nitrogen containing six member rings, as
illustrated in Table 7, such as pyridine (72ꢀ74) or pirimidine (75ꢀ76) but none of them
resulted to be more potent than the parent compound.
Insert Table 7
D: 2_Benzamide meta substituted
From the data obtained, we concluded that the benzamido moiety is the only portion allowing
substitution, without substantially abolishing the inhibitory activity. Keeping in mind that 19
resulted among the most interesting derivatives, we investigated several replacements of the
methoxy methyl in meta position (see Table 8).
Halogens, methoxy groups, tertiary amines (22−24, 26, 77, 78), did not influence much the
activity, being the chlorine substituted (26) the most active compound, whereas primary
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
amides and bromine analogue were highly detrimental as in 79 and 80 (see Table 8a).
Variation of methoxymethyl brought to the methylthiomethyl derivative 32 and
morpholinomethyl 82 being slightly more potent, removal of the terminal methyl as in 85
worsen the activity of four fold, whereas phenoxymethyl 86 caused a complete abolishment
of inhibitory potency, which was partially recovered with a pꢀOH substitution 84 (see Table
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8b). The introduction of a pyridine (Table 8c) linked to the phenyl by a ꢀCH
2
Oꢀ group (as in
8
8) or by ꢀOCH ꢀ (as in 87) brought to a slight improvement of the activity, whereas the 4ꢀ
2
pyridinol derivative (89) resulted less active.
When observing the whole series of derivatives, a tendency in potency improvement in
presence of longer chains carrying a distal basic nitrogen emerges. In support of this
hypothesis, we observed a clear improvement with derivatives 90 and 91: in both derivatives
pꢀpiperidinoxyphenoxy substituent was able to increase the potency down to a
submicromolar range (90: IC50=0.162µM, 91: IC50=0.442µM, Table 8d). Docking
experiments of 90 have been performed using internal produced protein structure, but the
presence of two negatively charged regions allowed dual pose orientations. Although the
uniqueness of the pose could not be established, in both cases the activity improvement is
well justified by the interactions of the piperidine moiety with one aspartate residue (Asp555
or Asp375). In Figure 7 is reported one of the accessible orientations.
Insert Table 8
Insert Figure 7
To further characterize this chemical class, we analyzed the selectivity profile of the two
most potent compounds: 90 and 91 using 6 as positive control. As reported for the initial hit
19, they demonstrated a high selectivity for KDM1A.
Insert Table 9
ACS Paragon Plus Environment

Page 17 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Compounds 90 and 91 were evaluated in terms of KDM1A inhibitory activity in acute
myeloid leukemia cell line THPꢀ1. This cell line is characterized by the MLLꢀAF9
translocation, and was selected in consideration of the experimental evidences demonstrating
the relevance of KDM1A in sustaining the oncogenic potential of MLLꢀAF9 leukemia stem
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4
3
cells.
The absence of evident effect on global level of histone H3K4 methylation after genetic
44
45
ablation of KDM1A, both in cancer cells and hematopoietic cells, let us to deprioritize
this experimental read out for cellular KDM1A inhibition experiments.
We evaluated instead the compound effects on the expression level of two differentiation
markers: CD14 and CD11B (modulated by KDM1A down regulation as reported in
Supporting Information), considering that KDM1A inhibition induces differentiation of
5
human AML cells . As reported in table 9, 90 and 91, as well as 6, induced an increase of
mRNA expression of CD14 and CD11B selected genes, similarly to KDM1A downꢀ
regulation by short hairpin RNA(shRNA). These results demonstrate the compounds ability
to block KDM1A activity in cells and to induce differentiation of THPꢀ1 cells. To
additionally characterize their effects on leukemia cells, cell cycle analysis and assessment of
apoptosis induction were performed in THPꢀ1 cells. As reported in Supporting Information
(Figure S2) treatment for 72 hours at the dose of 1 ꢀM for 90 and 91 and at the dose of 0.5
ꢀ
M for 6 did not induce apoptosis and cell death consistentely with the expected cytostatic
25,43
proꢀdifferentiation mechanism, due to KDM1A inhibition in AML cells
. Finally, the
anticlonogenic potential of 90 was determined in the human MLLꢀAF9 leukemia cell line
THPꢀ1. As reported in the table 10, at dose of 5 ꢀM, 90 was able to inhibit the colony
formation in THPꢀ1 cells. Moreover, 90 ꢀ as well as 6 ꢀ induces differentiation in THPꢀ1 cells
as demonstrated by expression increases of CD86 in the cells ꢀ recovered at the end of
semisolid culture experiment.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
CONCLUSIONS
We run an HTS (TRꢀFRET assay) on 34000 compounds, in search of reversible inhibitors of
the demethylase KDM1A, identifying 4 chemical classes inhibiting the enzyme at a low
micromolar concentration. We prioritized thieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamides and 3,8ꢀ
diazaspiro[4.5]decanes, demonstrating the biochemical mechanism of action for each hit
selected in both the series.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The binding mode of a representative compound of thienopyrrole series was experimentally
determined through Xꢀray crystallography and represents the first structural knowledge of
KDM1A reversible inhibitors. With this notion we expanded the class modifying the
scaffold, the amido linker and the aryl on the benzamide moiety. We discovered that residues
bearing a terminal basic moiety located at about 18
Å away from thienopyrrole centroid
drastically improved the activity. Further SAR exploration of the chemical series is reported
28
in part 2, in which we approached the ambiguity of the binding poses.
EXPERIMENTAL SECTION
HTS Assay Conditions (KDM1A Enzyme Inhibition)
To identify small molecule reversible inhibitor of KDM1A enzyme, TRꢀFRET technology
was used to follow KDM1A enzymatic activity. In our assay a biotinylated H3K4me1
peptide substrate underwent demethylation and the demethylated products were then captured
by the Euꢀlabeled antibody (EuꢀAb) and U Light Streptavidin (U LightꢀSA), thus bringing the
Eu donor and U Light acceptor dye molecules into close proximity.
Upon irradiation at 340 nm, the energy from the Eu donor is transferred to the U Light
acceptor dye which, in turn, generates light at 665 nm. The intensity of the light emission is
proportional to the level of biotinylated reaction product. Once identified the best
configuration of the assay in 384 format, optimizing the reaction and the detection conditions,
assay adaptation to the automated screeningꢀcompatible conditions was performed, including
ACS Paragon Plus Environment

Page 19 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the analysis of intraꢀday variability and the spatial uniformity assessment. The overall RZ’
mean for the day 1 and day 2 plates respectively of 0.79 and 0.89, indicated the robustness of
the identified conditions and protocol. Inter and intraꢀplate variability analysis identified no
issues, the variability observed being below the threshold of 20%. RZ’ is defined as:
RZ’ =1ꢀ [3RSD (NC) + 3RSD (IC)]/ (|<NC> ꢀ <IC>|)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
RSD = robust standard deviation;
<
>= median of the signal values for the indicate control wells of a plate (NC for neutral
control, IC for Inhibitor control).
The KDM1A inhibiting activity was determined using a TRꢀFRET assay (time resolved
®
fluorescence resonance energy transfer, Lance Ultra Demethylase technology, Perkin Elmer,
Waltham, MA, USA), which comprises a Europium chelate donor dye (TRF0404, Perkin
TM
Elmer, Waltham, MA, USA) together with ULight (TR0102, Perkin Elmer, Waltham, MA,
USA), a small molecular weight acceptor dye with a redꢀshifted fluorescent emission, and a
biotinylated 21 aminoacids histone H3ꢀderived monomethylated peptide (H3K4me)
[
Lys(Me1)4]ꢀHistone H3 (1ꢀ21)ꢀGGK(biotin), (64355, Anaspec, Fremont, CA, USA) as
substrate. The intensity of the light emission is proportional to the level of biotinylated
reaction product. The complex of human recombinant KDM1A/CoREST protein was
46 34
produced in E. coli as separate proteins and coꢀpurified as previously described.
Demethylase Assay conditions: the enzymatic assay was performed in 384 well white plates
TM
(
6007290 OptiPlate , Perkin Elmer, Waltham, MA, USA) in a final volume of 30 µL of
assay buffer (Tris HCl 50 mM pH 8.8, NaCl 50 mM, DTT 1 mM, Tweenꢀ20 0.01%, proteaseꢀ
free BSA 0.01% w/v) as follows: 10 µL of 237 nM KDM1A/CoREST protein were first
incubated 10 minutes at RT with 10 µL of 3 fold concentrated compounds (preꢀdiluted in
assay buffer at 1.5% DMSO). The reaction was then started adding with 10 µL of 150 nM
histone H3K4 monomethylated substrate. After 60 min at RT, the reaction was stopped with
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
10 µL of 1.2 mM tranylcypromine (P8511ꢀ1G, SigmaꢀAldrich, St. Louis, MO 63103).
Detection step conditions: 10 µL of detection Mix containing 1 nM Euꢀantibody and 4 nM
UꢀLightꢀStreptavidin in 1X Lance Detection Buffer (TRF0404, TR0102, CR97100, Perkin
Elmer, Waltham, MA, USA) were dispensed on 40 µL stopped reaction. The resulting
mixture was incubated in the dark for 1 h at RT. Then, TRꢀFRET signal was read by a
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Plus
fluorimeter (PHERAstar , BMG LABTECH GmbH, Germany) (Excitation 337 nm,
Emission 665 nm and 620 nm, delay time 100 µs, window time 200 µs). Compound activity
and IC50 values in dose response experiments were calculated on TRꢀFRET signal at 665 nm
®
using Genedata Screener 10 (Genedata, Basel, Switzerland).
Routine Assay Conditions (KDM1A Enzyme Inhibition)
Demethylase Assay conditions: 0.25 nM KDM1A/CoREST protein and compound in 100%
DMSO were added in a final volume of 48 ꢀL assay buffer (Tris HCl 50 mM pH 8.8, NaCl
50 mM, DTT 1 mM, Tweenꢀ20 0.01%) to each well of a 96 well half area flat bottom white
plate (3693 Costar, SigmaꢀAldrich, St. Louis, M, USA).
Demethylase reaction was started by the addition of 50nM histone H3K4 monomethylated.
After 20 min at RT, 300 µM tranylcypromine (P8511ꢀ1G, SigmaꢀAldrich, St. Louis, MO
63103) was added to stop the reaction.
Detection step conditions: 10 µL of the assay mixture was transferred from the original plate
TM
into a 384 well white plate (6007290 OptiPlate , Perkin Elmer, Waltham, MA, USA) and
10 µL of the detection Mix containing 2 nM Euꢀantibody and 10 nM UꢀLightꢀStreptavidin in
1X Lance Detection Buffer (TRF0404, TR0102, CR97100, Perkin Elmer, Waltham, MA,
USA). The resulting mixture was incubated in the dark for 1 h at RT. Then, TRꢀFRET signal
was read by a fluorimeter (Infinite® F200, Tecan, Männedorf, Swirzerland) (Excitation
320nm, Emission 665 nm and 620 nm, delay time 50µs, window time 100µs).
Reversibility and Michaelis–Menten Analysis
ACS Paragon Plus Environment

Page 21 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Biochemical characterization of reversibility and
K
i
was performed using the
KDM1A/CoREST Assay Horseradish Peroxidase. The complex of human recombinant
KDM1A/CoREST protein was produced in Escherichia coli as separate proteins and coꢀ
purified. The experiments were performed in 96 well half area white plates (cat. 3693,
Corning, Corning, NY) using a monoꢀmethylated H3ꢀK4 peptide containing 21 amino acids
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(custom synthesis done by Thermo Scientific) as substrate and in a in 40 ꢁl volume of 50 mM
TRISꢀHCl, pH 8.0 and 0.05 mg/ml BSA buffer. The peptide purity was >95% as checked by
analytical highꢀpressure liquid chromatography and mass spectrometry. The demethylase
activity was estimated under aerobic conditions and at RT by measuring the release of H O
2
2
®
produced during the catalytic process by the Amplex UltraRed detection system coupled
with horseradish peroxidase (HRP). Briefly, 20 nM of KDM1A/CoREST complex was
incubated at RT for 15 min in the absence and/or the presence of various concentrations of
®
the inhibitors, 50 ꢁM Amplex UltraRed (Life Technologies) and 0.023 ꢁM HRP (Sigma) in
50 mM Tris–HCl pH 8.0 and 0.05 mg/ml BSA. The inhibitors were tested twice in duplicates
at each concentration. Tranylcypromine (Sigma) was used as control. After preincubation of
the enzyme with the inhibitor, the reaction was initiated by addition of 4.5 ꢁΜ of monoꢀ
®
®
methylated H3ꢀK4 peptide. The conversion of the Amplex Ultra Red reagent to Amplex
UltroxRed was monitored by fluorescence (excitation at 510 nm, emission at 595 nm) for
12 min and by using a microplate reader (Infinite 200, Tecan Group, Switzerland). Arbitrary
units were used to measure the level of H O produced in the absence and/or in the presence
2
2
of inhibition. The maximum demethylase activity of KDM1A/CoREST was obtained in the
absence of inhibitors and corrected for background fluorescence in the absence of
KDM1A/CoREST. The IC50 was calculated using GraphPad Prism version 4.0 (GraphPad
Software, San Diego, CA).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Reversibility was determined using jump dilution. Specifically KDM1A/CoRest complex
was incubated at a concentration 100 fold over the concentration required for the enzymatic
activity with saturating concentrations of the inhibitor (10 fold of its biochemical IC50). After
15 minutes of incubation, a 100ꢀfold dilution into reaction buffer containing the H3K4me
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
substrate was then performed to initiate the reaction. The reaction progress curves were
followed and compared to a positive control (KDM1A/CoREST enzyme complex without
inhibitor). The irreversible inhibitor 1 was used as a control of irreversible inhibition.
Michealis Menten analysis
The K of compound 19 and 21 was determined using a Michaelis–Menten kinetic analysis
i
across multiple concentrations of both compounds. Initial velocities of biochemical assay
were then plotted toward substrate enzyme concentrations and fitted to the equations for
competitive, noncompetitive and uncompetitive inhibition, using SigmaPlot (Systat Software,
San Jose, CA)
BioluminescentꢀCoupled Assay for Monoamine oxidases (MAOꢀGlo Assay)
The MAO Glo Assay from Promega (cat. V1402, Promega, Madison, WI) was used to
measure the effect of inhibitors on MAO A and MAO B activity. Human recombinant MAO
A and MAO B from SIGMA (M7316 and M7441) have been used for the screening The
assay was performed at RT in 50 µL (25 µL reaction solution + 25 µL detection reagent) in
96 well half area white plates (cat. 3693, Corning, Corning, NY). Luminescence was
measured after 20 min incubation in the dark using a microplate reader (Infinite F200, Tecan
Group, Switzerland) with an integration time of 0.25 s per well. 16 ng/ꢀL MAO A or
20ng/ꢀL MAO B were incubated with different inhibitor concentrations (from 0.004 µM to
100 µM) for 15 min at RT in Promega MAO Buffer or Promega MAO B Buffer (MAO Glo
Assay kit, catalogue number V1402, Promega, Madison, WI). After 30 min of incubation the
reaction was stopped with the Promega detection reagent. All compounds were tested twice
ACS Paragon Plus Environment

Page 23 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
and IC50 values were calculated using GraphPad Prism version 4.0 (GraphPad Software, San
Diego, CA).
Chemoinformatics
To perform all chemoinformatics activities, including substructure searches, similarity
searches, clustering and to control all docking activities Pipeline Pilot (vesion 9.1) protocols
have been used; Clustering (as implemented in Pipeline Pilot component) and Similarity
Searches (similarity threshold between 0.85 and 9.0) have been performed using Tanimoto
distance and FCEP6 descriptors as implemented in Pipeline Pilot (BIOVIA).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
29
Pipeline Pilot has been used also to verify absence of PAINS inhibitor features in the
synthesized compounds, by means of SubStructureSearches (SSS) extracted from the PAINS
paper.
Cellular Assay
Cell type
Human leukemia THPꢀ1 cells wer obtained from the Deutsche Sammlung von
Mikroorganismen und Zellkulturen and resulted to be mycoplasma free both by PCR and
Mycoalert (Lonza kit). Cells were grown in RPMI supplemented with 10% fetal bovine
serum, 2 mM lꢀglutamine, and antibiotics and maintained in a humidified tissue culture
2
incubator at 37 °C in 5% CO .
Retroviral Constructs and Production of Retroviruses
shRNA constructs were prepared in the MSCVꢀbased pLMP retroviral vector. The hairpins
used in the study were #12 and #13 (scramble ShRNAs) and 1A#3 and 1A#5 (shRNAs
against KDM1A):
#12, AGTACGCGAAGAATACTATCGA
#
13, AGTACGTTCAGAATATCATCGAT
A#3, AAGTGATACTGTGCTTGTCCAC
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1A#5, ATCTCAGAAGATGAGTATTATT
Supernatants from transfected Phoenix packaging cells were collected at 48 and 72 h after
transfection and immediately used for infections.
Retroviral Transduction of THPꢀ1 cells
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
For THPꢀ1 cells infection, cells were diluted in retroviral supernatants added with 8 ꢁg/mL
6
polybrene and seeded at 10 cells/well in 24ꢀwell plates. Spin infection was performed for 2
consecutive days (2 infections per day) at 1800 rpm for 45 min. Puromycin selection
(3ꢁg/mL) started 24 h after the last cycle of infection. Assays were carried out 8 days after
beginning of infection.
Immunoblots and Antibodies
Whole cell extracts were obtained by lysis in SDS buffer (50 mM TrisꢀHCl, 10% glycerol,
2% SDS). Proteins were separated by SDS–PAGE, blotted onto PVDF membrane and probed
with the indicated antibodies. The antibodies used in the study were antiꢀLSD1(KDM1A)
(no. 2139, Cell Signaling), antiꢀKDM1B (Sigma HPA031269), and antiꢀHDAC1 (ab7028),
Transcriptional assay
Human acute myeloid leukemia THPꢀ1 were grown in RPMI supplemented with 10% foetal
bovine serum, 2 mM lꢀglutamine, and antibiotics and maintained in a humidified tissue
culture incubator at 37 °C in 5% CO . Cells were treated at the fixed dose of 1 ꢀM or with
2
vehicle (DMSO). After 24 h the cells were collected for RNA analysis. Total RNA was
purified using RNeasy Mini Kit (Qiagen, Valencia, CA), quantified and reverse transcribed.
mRNA levels were measured by quantitative RTꢀPCR (Fast SYBR Green Master mix,
Applied Biosystems Foster City, CA) using specific primers and normalized against TBP
mRNA. Results are presented as fold induction relative to vehicle treated cells (DMSO).
Primers used in this study were:
CD14, GTTCGGAAGACTTATCGACCAꢀ ATCGTCCAGCTCACAAGGTT
ACS Paragon Plus Environment

Page 25 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
CD11B, AACCCCTGGTTCACCTCCTꢀ CATGACATAAGGTCAAGGCTGT
TBP, GCTGGCCCATAGTGATCTTT – CTTCACACGCCAAGAAACAGT.
Assessment of apoptosis by Annexin V and propidiun iodide double staining
Vehicle (DMSO) or drugꢀtreated cells were stained with Annexin V and propidiun iodide and
the percentages of apoptotic cells were determined by flow cytometry. Briefly, cells were
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
resuspended in annexinꢀV binding buffer (10mM Hepes pH 7.4, 150mM NaCl, 1mM MgCl
2
,
3.6mM CaCl2, 5mM KCl) containing 1/50 dilution of annexinꢀVꢀAPC (eBioscence and 3ul
of PI (Sigma) solution 50ug/ml in PBS. Cells were freshly analyzed using FACSCanto II
flow cytometers (BD Biosciences).
Cell Cycle Analysis
Cells were treated with vehicle (DMSO) or the reported inhibitors for 72 hours, then
harvested and fixed dropwise with 70% ethanol and kept at 4°C before analysis by flow
cytometry. Nucleic acids from fixed cells were stained with 1ml propidium iodide (50 ug
/mL) plus RNase A (Sigma) (250ug/mL), overnight at 4°C. DNA content was measured by
using FACSCanto II flow cytometers (BD Biosciences) analyzed using BD FACS Diva
Software and ModFit LT3.0.
Cell Studies (Clonogenic Assay)
For the clonogenic assay, THPꢀ1 cells were plated at a density of 250 cells/plate and cultured
with drugs at reported doses in MethoCult H4435 Enriched (StemCell Technologies,
Vancouver, BC) according with the manufacturer instructions for 13 days. After this time,
colonies were counted. Percentage of inhibition is referenced versus the vehicle (DMSO)
treated cells.
Crystallography
47
Crystals of KDM1A/CoREST were obtained as described. For structure determination of
enzymeꢀinhibitor complexes, crystals were soaked for 3ꢀ8 h in a solution consisting of 1.3 M
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
sodium/potassium tartrate, 100 mM Nꢀ(2ꢀacetamido)ꢀ2ꢀiminodiacetic acid pH 6.5, 12% (v/v)
glycerol, and 0.5e1 mM inhibitor. Diffraction data were measured at beamlines of the
European Synchrotron Radiation Facility (Grenoble, France) and the Swiss Light Source
(Villigen, Switzerland). Data processing and coordinate refinement were performed with
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
48
programs of the CCP4 package.
Statistical Analysis
49
SigmaPlot v11
Protein preparation for docking
50
The crystallographic coordinates of 19 were prepared as previously reported. In summary,
the protein was added of hydrogen atoms in Chimera software, optimizing the Hꢀbond
networking and minimizing their position with 1000 steps of Steepest Descent. Then with
Semiempirical Quantum Mechanics (SQM) first the hydrogen atoms were optimized till a
gradient norm of 50 and then the whole protein was fully optimized SQM for 50 steps. The
resulting structure was used to develop the inputs necessary for all modeling programs by
50
means of the procedure presented (BASH scripted) in the work.
Favorable interaction regions
Favorable interaction areas are derived from the exploration of proteins binding site with
various atoms as probes. This procedure maps the interaction energy between a probe, which
characterizes the type of interaction, and the target. Examples of the probes are aliphatic
carbon for the hydrophobic pattern or amine nitrogen for Hꢀbond donor, etc. The Autogrid
facility (from Autodock software) is used for this gridꢀbased energy evaluation. Thus, the
interaction is recorded placing the probe at each node of the grid in the binding site. Finally,
selecting the appropriate energy cutꢀoff and displaying only the grid's points with energy
below such value, the favorable interactions are visualized.
ACS Paragon Plus Environment

Page 27 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
5
1
The Autodock plugin for PyMOL software (The PyMOL Molecular Graphics System,
Version 1.7.4 Schrödinger, LLC.) reads these Autogrid files and generates the surface
representation on the graphical rendering of proteinꢀligand coordinates.
Pose production
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
To predict the putative binding pose of ligands interacting with the target a docking protocol
was used. In order to explore a larger portion of the conformational space for these
molecules, results from various docking programs were mixed together. The combination of
various methodologies has been previously observed to higher the chances of finding ligand
conformation close to the crystallographic structure and thus to enhance of docking
39
36
37
38
performances. In our study, 3 programs were used: Autodock , Plants and Glide with
50
the settings described in our previous work.
Pose clustering
The docking predictions were processed as one ensemble undergoing cluster analysis to
reduce dataset dimensionality and to identify binding conformational families. The
methodology used for this step is ACIAP, Autonomous hierarchical agglomerative Cluster
39
Analysis based Protocol. In brief, this method consists on the construction of a full
hierarchical tree of the predicted poses using the Root Mean Square Deviation (RMSD)
between ligand conformations as metric and the average linkage as criterion to determine the
distances. Finally, the clustering level to cut the tree is identified with the KGS penalty
function, which is based on the average spread of clusters at each tree level. The global
minimum of this function ensures the best compromise between the number of clusters and
the mean distances between clusters.
This methodology was implemented as automatic computational scripts by means of R
52
software . Once the procedure is performed the medoid conformation for each cluster is kept
as the representative for the family of binding. In addition, the number of cluster members is
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 78
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
also used to judge the quality of a binding mode. The more a solution is found the more
probable it is.
Pose scoring
40
In our protocol, the surviving poses are then optimized and scored by means of BEAR
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(binding estimate after refinement). With this method each docking solutions is minimized in
complex with the target for 2000 steps, then a 100 picoseconds Molecular Dynamics
simulation of sole ligand is run keeping frozen the protein. The resulting conformation of the
complex is reꢀminimized and scored (single point) with GBSA and PBSA. In addition to the
interaction energy estimation another value is used to judge the final pose: the RMSD value
of the ligand between the initial conformation and the one after the MD. The higher the
RMSD the less stable is the predicted conformation and thus the reliability of the binding
pose.
53
54
BEAR was implemented in house by means of Amber software package and Chimera .
Modeling protocols implementation
Pose production, clustering and scoring were implemented within the architecture described
50
32
in our recent work. The protocol was constructed in Pipeline Pilot with its usual
components plus the ‘CADDD components’ that we developed to use the modeling software
on a powerful GNU/Linux environment. This consists of PP sub protocols that invoke BASH
scripts via SSH calls. The scripts are a combination of command lines, to interface with the
wanted program, and scripting variables to automatize the process.
Chemistry
General Procedures.
Melting points were determined on a BUCHI Melting point Bꢀ540 and are uncorrected.
Reagents and solvents used, unless stated otherwise, were of commercially available reagent
grade quality and were used without further purification. Reaction solvents were used
ACS Paragon Plus Environment

Page 29 of 78
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
anhydrous, purchased in Sure/Seal bottles stored over molecular sieves. Silica gel flash
chromatography purifications were performed on Merck silica gel 60 (0.04ꢀ0.063 mm) or
Biotage Isolera One system using preꢀpacked columns from Grace and Biotage. Nuclear
1
magnetic resonance spectra ( H NMR) were recorded on a Agilent VNMRS 500 MHz
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
spectrometer at 300 K and are referenced in ppm (δ) relative to TMS. Coupling constants (J)
are expressed in hertz (Hz). HPLCꢀMS experiments were performed on an Acquity UPLC
apparatus, equipped with a diode array and a Micromass SQD single quadruple (Waters). The
compounds used for biological tests were monitored at 220 and 254 nM, with two different
methods and their purity was found to be at least 95%. Semiꢀpreparative HPLC was
performed on a Waters apparatus eluting the sample through a BEH C 18 column (5µm,
1
9mmx100mm) using decreasingly polar mixtures of water (containing 5% of MeCN and
.1% of trifluoroacetic acid) and MeCN (containing 0.1% of trifluoroacetic acid).
0
General procedure to prepare benzamido derivatives by direct amidation of esters
3 mL (3 mmol) of a 1 M solution of LiHMDS in THF was added dropwise at 0°C to a stirred
solution of 1 mmol of ester and 1.25 mmol of the required aniline. The mixture was allowed
4
to spontaneously warm to rt then was poured into sat. NH Cl and extracted with DCM. The
combined organic layers were dried over Na SO , filtered and evaporated and the oil residue
2
4
was purified by column chromatography (eluent: nꢀhexane/EtOAc, from 2% to 20% of
EtOAc) providing the desired compound.
Similarly prepared:
N-(3ꢀMethoxyphenyl)ꢀ4ꢀmethylꢀthieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamide (22) (from 105a and
1
3
ꢀmethoxyaniline): yield 40% (23 mg); pale brown solid; H NMR (DMSOꢀd6) δ (ppm): 9.96
(s, 1 H), 7.52 (d, J= 5.4 Hz, 1 H), 7.45ꢀ7.41 (m, 1 H), 7.34ꢀ7.30 (m, 2 H), 7.26ꢀ7.19 (m, 2 H),
+
6
.68ꢀ6.61 (m, 1 H), 4.01 (s, 3 H), 3.75 (s, 3 H); MS (ESI): m/z: 287 [M+H] .
4ꢀMethylꢀN-phenylꢀthieno[3,2ꢀb]pyrroleꢀ5ꢀcarboxamide (23) (from 105a and aniline):
ACS Paragon Plus Environment