Discovery of Reversible DNA Methyltransferase and Lysine Methyltransferase G9a Inhibitors with Antitumoral in Vivo Efficacy

Article abstract of DOI:10.1021/acs.jmedchem.7b01926

Using knowledge- and structure-based approaches, we designed and synthesized reversible chemical probes that simultaneously inhibit the activity of two epigenetic targets, histone 3 lysine 9 methyltransferase (G9a) and DNA methyltransferases (DNMT), at nanomolar ranges. Enzymatic competition assays confirmed our design strategy: substrate competitive inhibitors. Next, an initial exploration around our hit 11 was pursued to identify an adequate tool compound for in vivo testing. In vitro treatment of different hematological neoplasia cell lines led to the identification of molecules with clear antiproliferative efficacies (GI₅₀ values in the nanomolar range). On the basis of epigenetic functional cellular responses (levels of lysine 9 methylation and 5-methylcytosine), an acceptable therapeutic window (around 1 log unit) and a suitable pharmacokinetic profile, 12 was selected for in vivo proof-of-concept (Nat. Commun. 2017, 8, 15424). Herein, 12 achieved a significant in vivo efficacy: 70% overall tumor growth inhibition of a human acute myeloid leukemia (AML) xenograft in a mouse model.

Full text of DOI:10.1021/acs.jmedchem.7b01926

Subscriber access provided by - Access paid by the | UCSB Libraries
Article
Discovery of reversible DNA methyltransferase and lysine
methyltransferase G9a inhibitors with antitumoral in vivo efficacy
Obdulia Rabal, Edurne San José - Eneriz, Xabier Agirre, Juan A Sánchez-Arias, Amaia Vilas-Zornoza,
Ana Ugarte, Irene De Miguel, Estibaliz Miranda, Leire Garate, Mario Fraga, Pablo Santamarina, Raul
Fernandez Perez, Raquel Ordoñez, Elena Sáez, Sergio Roa, Maria Jose Garcia-Barchino, Jose
Angel Martinez-Climent, Yingying Liu, Wei Wu, Musheng Xu, Felipe Prosper, and Julen Oyarzabal
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01926 • Publication Date (Web): 28 Jun 2018
Downloaded from http://pubs.acs.org on June 29, 2018
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 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 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.
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 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of Navarra, Small Molecule Discovery
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
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 101
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
Discovery of reversible DNA methyltransferase and lysine
methyltransferase G9a inhibitors with antitumoral in vivo efficacy
1
,6
2,6
2,6
Obdulia Rabal, Edurne San JoséꢀEnériz, Xabier Agirre, Juan Antonio Sánchezꢀ
1
2
1
1
2
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
Arias, Amaia VilasꢀZornoza, Ana Ugarte, Irene de Miguel, Estíbaliz Miranda, Leire
2
3
3
3
2
Garate, Mario Fraga, Pablo Santamarina, Raul Fernandez Perez, Raquel Ordoñez,
1
2
2
2
Elena Sáez, Sergio Roa, María José GarcíaꢀBarchino, José Angel MartínezꢀCliment,
4
4
4
2,5,*
1,*
Yingying Liu, Wei Wu, Musheng Xu, Felipe Prosper
and Julen Oyarzabal
1
2
Small Molecule Discovery Platform, Molecular Therapeutics Program, Area de
HematoꢀOncología, IDISNA, Ciberonc, Center for Applied Medical Research (CIMA),
University of Navarra, Avenida Pio XII 55, Eꢀ31008 Pamplona, Spain
3
Nanomaterials and Nanotechnology Research Center (CINNꢀCSIC)ꢀUniversidad de
OviedoꢀPrincipado de Asturias, Avenida de la Vega, 4ꢀ6, 33940 El Entrego, Spain
4
WuXi Apptec (Tianjin) Co. Ltd., TEDA, No. 111 HuangHai Road, 4th Avenue, Tianjin
300456, PR China
5
Departmento de Hematología, Clinica Universidad de Navarra, University of Navarra,
Avenida Pio XII 36, Eꢀ31008 Pamplona, Spain
6
These authors equally contributed to this work.
ABSTRACT
Using knowledgeꢀ and structureꢀbased approaches, we designed and synthesized
reversible chemical probes that simultaneously inhibit the activity of two epigenetic
targets, histone 3 lysine 9 methyltransferase (G9a) and DNA methyltransferases
(DNMT), at nanomolar ranges. Enzymatic competition assays confirmed our design
strategy: substrate competitive inhibitors. Next, an initial exploration around our hit 11
1
/101
ACS Paragon Plus Environment

Page 3 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
was pursued to identify an adequate tool compound for in vivo testing. In vitro treatment
of different hematological neoplasia cell lines led to the identification of molecules with
clear antiꢀproliferative efficacies (GI50 values in the nanomolar range). Based on
epigenetic functional cellular responses (levels of lysine 9 methylation and 5ꢀ
methylcytosine), an acceptable therapeutic window (around 1 log unit) and a suitable
pharmacokinetic profile, 12 was selected for in vivo proofꢀofꢀconcept (ref 53). Herein,
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
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
12 achieved a significant in vivo efficacy: 70% overall tumor growth inhibition of a
human AML (Acute Myeloid Leukemia) xenograft in a mouse model.
INTRODUCTION
Epigenetic regulation of gene expression is controlled by DNA methylation,
nucleosome positioning, histone variant exchange and postꢀtranslational histone
1
,2
modifications. Epigenetic alterations contribute to many human diseases, including
cancer, inflammation, brain disorders, and metabolic and cardiovascular diseases. Thus,
the development of epigeneticꢀbased therapies has attracted considerable interest in the
scientific and pharmaceutical communities, especially after the approval of the first
therapies targeting DNA methyltransferases (DNMTs) and histone deacetylases
(HDACs) for treatment of hematological malignancies. Thus, among possible
epigenetic modifications, methylation has gained increased attention.
Methyltransferases, the methyl writers of the histone and DNA code, catalyze the
transfer of a methyl group from SꢀadenosylꢀLꢀmethionine (SAM or AdoMet) to its
corresponding substrate—histones, nonhistone proteins or DNA—yielding Sꢀadenosylꢀ
Lꢀhomocysteine (SAH or AdoHcy). Methyltransferases acting on lysine and arginine
2
/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 101
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
residues of histones and other proteins are referred to as Protein Methyltransferases
(PMTs), while those that methylate DNA are known as DNMTs.
G9a (also known as KMT1C (lysine methyltransferase 1C) or EHMT2 (euchromatic
histone methyltransferase 2)) is a histoneꢀlysine Nꢀmethyltransferase that catalyzes the
transfer of one or two methyl groups to the εꢀamino group of lysine 9 of histone H3
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
(H3K9me1 and H3K9me2), a hallmark associated with transcriptional gene silencing.
3
Other protein targets of G9 include the tumor suppressor p53, whose methylation leads
4
to its inactivation. G9a is upregulated in various cancers, and its overexpression has
5–7
been associated with poor prognosis and metastasis. There is mounting evidence that
8–11
the knockdown or pharmacological inhibition of G9a decreases cell growth,
delays
1
2
6,7
acute myeloid leukemia (AML) progression and blocks tumor metastasis. Beyond
13
14
cancer, G9a is involved in the maintenance of HIVꢀ1 latency, colitis, cognitive
15
disturbances (mental retardation, cocaine addiction, ageꢀrelated cognitive decline),
1
6
embryonic development and stem cell reprogramming, which has been used to
17
produce inducible pluripotent stem cells (iPSCs).
Compound 1 (BIXꢀ01294, Chart 1) was the first selective inhibitor of G9a and GLP
G9aꢀlike protein, its closely related protein, also known as EHMT1), discovered in a
(
18
screening campaign. As shown in Chart 1, subsequent optimizations of 1 led to the
identification of a series of compounds possessing the same quinazoline core. Thus,
inhibitor 2 (UNCꢀ0224), around two log units more potent than 1, was the first G9aꢀ
small molecule inhibitor coꢀcrystallized with G9a, demonstrating that these quinazolineꢀ
based inhibitors bind at the substrate peptide binding site and are not SAMꢀcompetitive
19
20
inhibitors. Chemical probe 3 (UNCꢀ0638) was the result of an extensive medicinal
19,21
chemistry effort
to obtain more potency against G9a and cellꢀpermeable quinazoline
derivatives. Finally, compound 4 (UNCꢀ0642), has an optimized pharmacokinetic (PK)
3
/101
ACS Paragon Plus Environment

Page 5 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2
2
profile for in vivo assays, and only very recently its efficacy in a PraderꢀWilli
23
syndrome mouse model has been proven. Only a few studies have been published
aiming to identify new alternative chemotypes of the quinazoline ring: the chemical
24
probe 5 (Aꢀ366) having a spiro[cyclobutaneꢀ1,3'ꢀindole] ring and, more recently, the
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
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
2
,4ꢀdiamino quinoline 6, which was found to be a potent G9a inhibitor after exhaustive
25
heterocycle replacement of the quinazoline ring of BIX derivatives. The in vivo
administration of compound 5 using osmotic miniꢀpump dosing in an AML flank
2
6
xenograft model resulted in moderate (45%) tumor growth inhibition. With the
2
0,27
exception of 3 (IC50 against DNMT1 in the micromolar range),
all of these
compounds are inactive against DNMT1 at concentrations higher than 10 ꢀM or even
5
0 ꢀM. Other G9a inhibitors with a pharmacophore different from 1 have been
28–30
31,32
reviewed
or were published very recently.
2
The involvement of DNA methylation in cancer is well acknowledged. DNMTs are
responsible for the methylation of the C5ꢀcytosine of DNA. The methylation of
promoter CpG islands leads to gene inactivation, as described for tumor suppressors and
DNA repair genes in different tumors. DNMTs are grouped into two families: DNMT1
and DNMT3A/DNMT3B, each family being predominantly involved in methylation
3
3
maintenance (DNMT1) or de novo methylation (DNMT3A/DNMT3B). These three
proteins are overexpressed in different cancer types or present mutations, as DNMT3A
34
in leukemia. Additionally, there is recent evidence that DNA methylation may play an
35
important role in various diseases from different therapeutic areas: memory function,
36
37
Alzheimer’s disease, amyotrophic lateral sclerosis, schizophrenia and bipolar
38
39
40
41
42
disorder, addiction, hypertension, atherosclerosis, type 2 diabetes, systemic
43
44
lupus erythematosus and fibrosis. As mentioned above, two nucleoside analogs, 7
(Azacytidine) and 8 (Decitabine) (Chart 2), have been approved by the U.S. FDA and
4
/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 101
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
EMA for treatment of hematological malignancies. These drugs incorporate into DNA
45
and inhibit DNMT1 through an irreversible covalent complex. These two drugs show
4
5
poor bioavailability, chemical instability and metabolic stability, leading to the search
for optimized nucleoside analogs, such as 9 (SGIꢀ110), currently being explored in
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
46
clinical trials. However, due to the toxic side effects associated with its mechanism of
action, there is increasing interest in finding potent nonꢀincorporating reversible DNMT
47,48
inhibitors. Among others (see refs
for comprehensive reviews), the quinolineꢀbased
10 (SGIꢀ1027) has demonstrated inhibitory activity against DNMT1, DNMT3a and
DNMT3b in biochemical assays and its potential to induce hypomethylation of CpG
49
islands of tumor suppressor genes. Interestingly, DNMT1 directly binds G9a and both
50
methyltransferases cooperate during cell division, promoting the transcriptional
5
1
52
silencing of target genes, including the reactivation of tumor suppressor genes.
53
54
Recently, we and others have shown that simultaneous blocking of G9a and DNMT1
exerts a synergistic effect on reducing the growth of cancer cells, either when
5
4
combining G9a knockdown with pharmacologic inhibition of DNMT or using specific
5
3
53
siRNAs against both targets or by inhibiting them with a combination of 5 and 8.
Based on these results, we postulated that smallꢀmolecule reversible dual inhibitors of
both methyltransferases might represent a novel approach in cancer therapeutics. Our
efforts toward the identification of such dual inhibitors commenced with compound 11
(Chart 3), a quinolineꢀbased matched pair of 3. Subsequent optimization led to the
discovery of the lead compound 12 (CMꢀ272), with a suitable PK profile for in vivo
administration and the ability to prolong the survival of AML, acute lymphoblastic
53
leukemia (ALL) and diffuse large Bꢀcell lymphoma (DLBCL) mice. Here, we report
the discovery and an initial exploration leading to the identification of our candidate for
in vivo proofꢀofꢀconcept compound 12.
5
/101
ACS Paragon Plus Environment

Page 7 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
N
N
N
HN
N
HN
N
HN
N
O
O
O
O
O
O
N
N
N
N
N
N
N
N
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
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
N
1
2
3
BIX-01294
UNC-0224
UNC-0638
IC50 G9a = 1200 nM
IC50 G9a = 15 nM
IC50 G9a = 55 nM
IC50 DNMT1 > 10000 nM
IC50 DNMT1 = 2010 nM
N
N
O
HN
NH2
HN
O
N
N
O
O
O
N
N
O
N
N
F
N
F
N
N
4
5
6
UNC-0642
A-366
IC50 G9a < 2.5 nM
IC50 DNMT1 > 50000 nM
IC50 G9a = 3.3 nM
IC50 DNMT1 > 50000 nM
IC50 G9a = 13 nM
IC50 DNMT1 > 50000 nM
Chart 1. Known G9a inhibitors: quinazolineꢀbased derivatives (compounds 1, 2, 3, 4)
and two novel alternative scaffolds of quinazoline: compounds 5 and 6. G9a and DNMT
IC50 values for 1 and 3 were determined internally (see footnote in Table 1).
2
0
Alternatively, reported G9a IC50 values for 1 are 180 nM (SAHHꢀcoupled assay) and
5
5
1900 nM (mass spectrometryꢀbased inhibition assay) and < 15 nM for 3 (SAHHꢀ
2
0
coupled assay). Concerning DNMT1, a percentage of inhibition of 35%±8% at 100
56
27
µM was reported for 1 and an IC50 value of 1287 nM was reported for 3, in both
cases using a radioactive methyl transfer assay. Discrepancies with initially reported
2
0
DNMT1 IC50 values for 3 (IC50 = 107,000 nM) have been attributed to a different
6
/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 101
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
2
7
19
22
24
25
assay format. The IC values for compounds 2 , 4 , 5 , and 6 have been published
50
previously.
NH2
N
NH2
N
NH2
N
H N
N
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
2
N
N
N
N
N
O
N
O
N
O
HO
HO
HO
O
N
NH
O
O
O
O
N
N
N
N
H
O H O H
OH
O
P
NH2
HN
N
O
O
–
O
O
O H
7
8
9
10
SGI-1027
IC50 G9a > 10000 nM
IC50 DNMT1 = 3550 nM
5
-Azacitidine
Decitabine
SGI-110
IC50 G9a > 10000 nM
DNMT1 = 300 nM
IC50 G9a > 10000 nM
DNMT1 = 30 nM
Chart 2. Known DNMT1 inhibitors: nucleoside (compounds 7, 8, 9) and nonꢀ
nucleoside (compound 10). G9a IC values for 7, 8 and 10 and DNMT1 IC values for
50
50
10 as determined internally (see footnote in Table 1). Alternatively, DNMT1 IC50
values of 12500 nM (Poly(dIꢀdC)) and 6000 nM (hemimethylated DNA) have been
reported for compound 10 using a radioactive methyl transfer assay with two different
49
57
substrates. For G9a, reported value for 10 is 59 ± 4 µM. DNMT1 values for 7 and 8
have been reported previously and correspond to the lower concentration at which DNA
45
hypomethylation is observed.
7
/101
ACS Paragon Plus Environment

Page 9 of 101
Journal of Medicinal Chemistry
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
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
11
12
CM-272
IC50 G9a = 0.7 nM
IC50 G9a = 8 nM
IC50 DNMT1 = 619 nM
IC50 DNMT1 = 382 nM
Chart 3. 11, quinoline matched pair of 3, and lead compound 12 (CMꢀ272). G9a and
DNMT1 IC50 values determined internally (see footnote in Table 1). Biochemical
inhibitory data of 12, and its close analogue (CMꢀ579), was confirmed by orthogonal
biochemical assays as well as by direct binding measurements to DNMT1 using
microscale thermophoresis (MST) and by cellular functional response (using
53
pyrosequencing).
RESULTS
Rational Design
We pursued knowledgeꢀ and structureꢀbased approaches to design novel G9a and
DNMT methyltransferase activity inhibitors; targeting both enzymes. For G9a, the
1
9
20
crystal structures of 2 (PDB entry 3K5K) and 3 (PDB entry 3RJW) revealed that
20
these quinazolineꢀbased inhibitors could bind to the histoneꢀbinding pocket of G9a. A
detailed analysis of the interactions of 3 with the substrateꢀbinding pocket of G9a
(Figure 1A) shows that i) the pyrrolidine side chain interacts with the lysine binding
channel (hydrogenꢀbond contact with the carbonyl of Leu1086 and cationꢀπ interaction
with the sideꢀchain of Tyr1154); ii) the –NH group at the 4ꢀposition of the quinazoline
8
/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 101
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
establishes a hydrogenꢀbond with Asp1083; iii) the basic nitrogen of the piperidine ring,
which is directed toward the solvent region, is in close proximity with Asp1078 (4.5 Å);
and iv) the N1 of the quinazoline ring, which is expected to be protonated at
20,55
physiological pH,
forms a salt bridge with Asp1088. As observed in Figure 1A, the
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
ring nitrogen atom (N3) of the quinazoline scaffold does not participate in any direct
key interaction with the protein, and no conserved water molecules appear proximal to
this residue as mediators in a waterꢀmediated hydrogenꢀbond contact. Thus, from the
viewpoint of G9a activity, the replacement of the quinazoline ring by a quinoline would
be well tolerated. Thus, we set out to synthesize and test the quinolineꢀmatched pair of
3
3
, compound 11, to validate our initial hypothesis for G9a activity. The observation that
displayed micromolar activity against DNMT1 (Chart 1) prompted us to use it as our
initial starting point for the proofꢀofꢀconcept of the dual G9aꢀDNMT inhibitor design.
56
Moreover, during the course of our investigations, Rotili disclosed a series of
quinazolineꢀbased compounds (based on chemical exploration around the G9a inhibitor
1) that inhibit DNMT3A at low micromolar levels without any significant inhibition of
DNMT1 and G9a.
For the DNMT1 structureꢀbased ligand design, different crystal structures of DNMT1 in
complex with SAH, SAM, sinefungin (a SAMꢀcompetitive inhibitor) and/or DNA are
publicly available. However, no crystal structures of any DNMT (including DNMT1,
DNMT3A and DNMT3B) complexed with small molecule reversible inhibitors such as
5
8–61
for example 10 are available, and only modeling studies have been published.
Thus,
the binding mode as well as the binding cavity (SAM or DNA pockets), for most of
these compounds, remain unknown. To guide the design of DNMT inhibitors, we
carried out docking studies using the productive DNMT1ꢀDNAꢀSAH complex structure
6
2
(
mouse DNMT1, mDNMT1, PDB entry 4DA4) containing a central hemimethylated
9
/101
ACS Paragon Plus Environment

Page 11 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
CpG site and lacking the autoinhibitory linker (mDNMT1: 699ꢀ733) and zincꢀfingerꢀ
like (CXXC motif, residues 646ꢀ692) domains at the Nꢀterminal region of the catalytic
domain (mDNMT1, 1140ꢀ1602, hDNMT1 1139ꢀ1616). This structure was chosen
because the catalytic domain of the active DNMT1 undergoes a strong conformational
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
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
62,63
change compared with the autoinhibited structure,
in which the DNA minor groove
is excluded from the catalytic site by the autoinhibitory linker. Thus, to model whether
compound 11 would bind to the SAM or the DNA pocket, a protein structure that can
accommodate both options was selected. Our docking studies of compound 11 showed
that the compound fits within the DNA binding pocket, superimposing well with the
hemimethylated CpG (Figure 1B) and the pyrrolidine ring occupying the cytidineꢀ
binding pocket, establishing a hydrogen bond with the carboxylate group of the catalytic
Glu1269 (Glu1266 in human DNMT1, hDNMT1) (Figure 1C). Additionally, the
oxygen directly bonded at the 7ꢀposition establishes a hydrogen bond with the
guanidine of Arg1315 (Arg1312 in hDNMT1), a guanidine of the conserved motif TRR
flanking the entrance of the cytidine. The NH of the 4ꢀposition establishes a hydrogen
bond with the side chain of Ser1233 (Ser1230 in hDNMT1), and the isopropylpiperidyl
group overlays with the guanine of the DNA chain exiting the catalytic pocket and
establishing hydrophobic contacts with Met1235 (Met1232 in hDNMT1) (Figure 1C).
The methoxy group at the 6ꢀposition orients towards the 3’ direction of the DNA chain.
Finally, the cyclohexyl ring at the 2ꢀposition lies in a cavity flanked by Thr1530
(Thr1528, hDNMT1), Gln1230 (Gln1227, hDNMT1), Tyr1243 (Tyr1240, hDNMT1)
and Arg1576 (Arg1574, hDNMT1), without making any explicit contact with any
residue. Comparing this active DNMT1 structure with the autoinhibited conformation of
63
DNMT1 (PDB entry 3PTA) that has both the CXXC and autoinhibitory domains, it
can be seen that the Cꢀterminal end of the CXXC linker (residues 692ꢀ700) lies
1
0/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 101
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
proximal to the region occupied by the cyclohexyl group (Figure S1). Thus, without any
further structural insight concerning the conformation adopted by the CXXC linker in
the active form, no clear conclusion can be stated regarding the potential ligand
interactions at this site. Similar conclusions were drawn when analyzing in detail a more
recently released crystal structure of hDNMT1 with a higher resolution that is also autoꢀ
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
64
inhibited (PDB entry 4WXX). Concerning the role of the central quinoline scaffold, it
seems to act as a mere scaffold, without establishing any key contact with the protein.
Thus, the replacement of the quinazoline of 3 by a quinoline 11 will predictably not
have a strong influence on the inhibitory activity against DNMT1. In contrast to G9a,
no apparent interaction between the nitrogen ring N1 and protein was observed, and no
negatively charged residue lies in its neighborhood.
Figure 1. (A) Complex of 3 with G9a (PDB accession code: 3RJW). (B) Docked
structure of compound 11 (orange) into DNMT1 (PDB accession code 4DA4) overlaid
1
1/101
ACS Paragon Plus Environment

Page 13 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
with DNA (pink). The cofactor SAH is shown in green. (C) Detailed analysis of the
predicted hydrogen bonds between compound 11 and DNMT1.
Scheme 1.
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
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
i)
iii)
iv)
13
16
1
4, R = NO2
5, R = NH2
ii)
1
v)
ii)
17
18
19
vi)
vii)
20
11
Conditions: i) 3ꢀpyrrolidinꢀ1ꢀylꢀpropanꢀ1ꢀol, PPh
3
, DEAD, THF, 0 ºC, followed by rt, 5 h; ii) Pd/C,
MeOH, H (1 atm), 15ꢀ25 ºC, 3ꢀ10 h; iii) malonic acid, POCl , rt, 4 h, followed by 90 ºC overnight; iv) 2ꢀ
2
3
(
cyclohexꢀ1ꢀenꢀ1ꢀyl)ꢀ4,4,5,5ꢀtetramethylꢀ1,3,2ꢀdioxaborolane, K
2 3 3 4 2
CO , Pd(PPh ) , 1,4ꢀdioxane/H O (5:1),
MW, 120 ºC, 1 h; v) tertꢀbutyl 4ꢀaminopiperidineꢀ1ꢀcarboxylate, Pd (dba) , BINAP, Cs CO , 1,4ꢀ
2
3
2
3
dioxane, 130 ºC, 36 h; vi) HCl/EtOAc (1.0 M), 16 ºC, 4 h; vii) acetone, AcOH, NaBH
15 h.
3
CN, THF, 50 ºC,
Compound 11 was synthesized as shown in Scheme 1 from commercially available 2ꢀ
methoxyꢀ5ꢀnitroꢀphenol (13). This compound was first transformed into amine (15) by
the Mitsunobu reaction and hydrogenation. Subsequently, 2,4ꢀdichloroquinoline (16)
3
was prepared by reaction with malonic acid in POCl , and intermediate 17 was isolated
after Suzuki coupling. Next, the desired substitution at position 4 was installed by
reaction with tertꢀbutyl 4ꢀaminopiperidineꢀ1ꢀcarboxylate obtaining intermediate 18,
which was converted into compound 19 by hydrogenation. Finally, deprotection of the
1
2/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 101
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
secondary amine under acidic conditions and reductive amination using acetone led us
to the desired compound 11.
Compound 11 was evaluated for G9a and DNMT1 inhibitory activity using the timeꢀ
resolved fluorescence resonance energy transfer (TRꢀFRET) assay. Intriguingly,
compound 11 was more potent than 3 against the two targets, with IC values of 0.7
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
5
0
nM against G9a (1.9 log units of increased potency) and 619 nM against DNMT1 (0.5
log units of increased potency) (Chart 3). Interestingly, in our hands, compound 3 was
also a better inhibitor of DNMT1 than compound 10 (Chart 2), a reversible DNMT
inhibitor commonly considered a reference compound for the DNMT inhibitor design.
Concerning G9a activity, the higher potency of the quinoline than that of the
quinazoline scaffold is in line with the results recently published for the quinolineꢀ
25
matched pair of compound 1 (BIXꢀ01294), compound 6 (Chart 1). In agreement with
the conclusions drawn by Srimongkolpithak et al, this improvement in G9a activity can
be attributed to the increased basicity of the N1 nitrogen that enhances the ionic
65
interaction with Asp1088, with the Pipeline Pilot predicted pKa values of 6.54
(compound 11) and 6.32 (compound 6) for the quinoline series compared with the
predicted pKa values of 4.69 (compound 3) and 4.48 (compound 1) for their respective
quinazolines. Compared with the experimentally determined pKa values of N1 for
2
5
compound 6 (pKa of 9.50) and compound 1 (pKa of 6.94), the pKa predictor
underestimates the pKa values, although the general trend is correctly predicted. Based
on the proposed binding mode of compound 11 into DNMT1, no explanation could be
stated for the improvement in the DNMT1 activity of compound 11 compared with that
of compound 3.
1
3/101
ACS Paragon Plus Environment

Page 15 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
We next determined the mechanism of action of compound 11 by carrying out
enzymatic competition studies varying the concentrations of SAM and the two
substrates: histone monomethylꢀH3K9 peptide for G9a and polyꢀdeoxy inosine polyꢀ
deoxy cytosine (pdIꢀpdC) DNA for DNMT1 (Figure 2). These competition experiments
confirmed that compound 11 is a substrateꢀcompetitive inhibitor of both targets, G9a
and DNMT1. In fact, the inhibitory potency of 11 is not affected by increasing
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
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
concentrations of the SAM cofactor (G9a IC ranging from 2 nM to 0.5 nM in Figure
50
2A and DNMT1 IC50 ranging from 1.4 µM to 2.4 µM in Figure 2C). However,
increasing the concentrations of the substrates, histone peptide (G9a, with IC50 values
spanning 0.7 nM to 29 nM, Figure 2B) and pdIꢀpdC (DNMT1, IC50 values from 1.4 µM
to > 10 µM, Figure 2D), have a clear impact on the binding of 11 to these two targets
(
pIC differences are greater than 1 log unit). Nonꢀcompetition with the SAM cofactor
50
is relevant to minimize offꢀtarget promiscuity. When compound 11 was assayed at a
concentration of 10 µM against a panel of 14 lysine and arginine histone
methyltransferases, only GLP, a closely related protein to G9a, was considerably
inhibited (IC50 of 180 nM, with some selectivity towards G9a); the rest of histone
methyltransferases were marginally inhibited, lower than 25% inhibition
(
Supplementary Table S1). For other DNMTs, only DNMT3A was significantly
inhibited by 11 at 10 µM (80%), while had no effect on DNMT3B (Supplementary
Table S1).
However, compound 11 had no antiꢀproliferative response against ALL CEMOꢀ1 (GI50
>
10 µM), AML MV4ꢀ11 (GI50 > 10 µM) and DLBCL OCIꢀLy3 and OCIꢀLy10 (0%
inhibition at 1 µM) cell lines. Interestingly, it showed no cytotoxicity in the healthy
hepatic cell line THLEꢀ2 (LC50 = 83.8 µM).
1
4/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 101
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
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
Figure 2. Competition experiments performed with compound 11 on G9a (A, B) and
DNMT1 (C, D) by varying the SAM concentrations (A, C), histone monomethylꢀH3K9
peptide (B) and pdIꢀpdC (C). Of note, the difference of 0.35 log units between the
determined DNMT1 IC50 value for this competition assay (1.4 µM) and the routinely
screening (619 nM, Chart 3), was exceptionally accepted for the purpose of the
competition assay (accepted difference in routinely screening is 0.3 log units, see
footnote in Table 1; requirement is met by 11).
In light of the high in vitro potency of 11 against G9a and DNMT1, we set out to
optimize the quinoline series with the first initial goal of enhancing DNMT1 activity
with potent antiꢀproliferative response against different hematological neoplasms and
with an adequate ADME profile to be administered in vivo (Chart 4). Given, that GLP is
1
5/101
ACS Paragon Plus Environment

Page 17 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a closely related homologue of G9a, compounds were routinely screened against G9a,
and only promising compounds were profiled against GLP.
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
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
1
1
SAR to improve DNMT1 potency and
achieve antiproliferative response.
SAR to improve ADME profile
Chart 4. SAR exploration strategy to optimize our initial hit compound 11.
SAR and antiproliferative response: Exploration of the 2-position (R1).
Initially, we decided to focus our SAR exploration at the 2ꢀposition because i)
modifications to the 2ꢀsubstituent of the quinazoline scaffold are well tolerated from the
1
9,21,66
viewpoint of G9a activity
, and ii) we hypothesized that it might have a great
impact on the permeability of the compounds to achieve a cellular response. For this
exploration, the 1ꢀmethylpiperidinꢀ4ylamino substituent was fixed at the 4ꢀposition as
the Nꢀcapping group because this fragment is directed toward the solventꢀexposed
region of both enzymes (predictably in the case of DNMT1), and it should not strongly
influence the in vitro potency of compounds while reducing their molecular weight and
easing synthetic accessibility. In fact, this fragment was found not to cause any potency
loss against G9a compared with the 1ꢀbenzylpiperidinꢀ4ꢀylamino group of compound 1
1
9
in the quinazoline series. Thus, compounds 23a-q and 31 (Table 1) with diverse
substituents at 2ꢀposition were designed and synthesized as shown in Schemes 2 and 3.
53
The synthesis of compound 12 has been previously reported. Compounds 23a-e and
1
6/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 101
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
2
3g-q were prepared from intermediate 16. First, the desired substitution at position 2
was installed by Suzuki coupling (17, 22d and 22g-r), BuchwaldꢀHartwig amination
22c) or reaction with NaOMe (22e). Next, the desired methylpiperidines were isolated
(
by different Pdꢀmediated coupling reactions. In the case of compound 23f, the
trifluoromethyl group at position 2 was installed through reaction of intermediate 15
with 4,4,4ꢀtrifluoroꢀ3ꢀoxobutanoate, and then compound 22f was prepared by heating in
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
POCl . Finally, compound 23f was synthesized by reaction with 1ꢀmethylpiperidinꢀ4ꢀ
3
amine under BuchwaldꢀHartwig amination conditions.
Scheme 2.
Cl
X
O
O
O
O
O
i) or ii)
iii), iv), v) or vi)
N
NH2
N
O
N
Y
N
N
R1
15
17, 22c-r
1
6, X = Y = Cl
2
1, X = OH, Y = CF3
N
HN
O
O
iv) or iv) and vii)
or viii) or ix)
N
N
R1
23a-q
N
R1
=
a)
b)
c)
S
N
N
d) -CH3
e) -OCH
3
f) -CF
3
g)
p)
h)
i)
N
O
N
NH k)
O
l)
m)
n)
o)
O
q)
O
r)
j)
N
N
O
N
H
Conditions: i) malonic acid, POCl
oxobutanoate, phenylpropanolamine, 120 ºC, 12 h; iii) corresponding boronic acid or ester, K
Na CO , Pd(PPh , 1,4ꢀdioxane or 1,4ꢀdioxane/H O (5:1 or 15:1), 110ꢀ120 ºC, MW or conventional
heating, 1ꢀ12 h; iv) 1ꢀmethylꢀpiperazine or 1ꢀmethylpiperidinꢀ4ꢀamine, Pd (dba) , Cs CO , BINAP, 1,4ꢀ
dioxane, 110ꢀ130 ºC, MW or conventional heating, 1ꢀ12 h; v) NaOMe, rt, overnight; vi) POCl , 110 ºC, 2
3
, rt, 4 h, followed by heating at 90 ºC overnight; ii) 4,4,4ꢀtrifluoroꢀ3ꢀ
2
CO or
3
2
3
3
)
4
2
2
3
2
3
3
2
h; vii) Pd/C, EtOH, H (1 atm), 25ꢀ35 ºC, 4ꢀ15 h; viii) 1ꢀmethylpiperidinꢀ4ꢀamine, biphenylꢀ2ꢀylꢀ
1
7/101
ACS Paragon Plus Environment

Page 19 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
dicyclohexylꢀphosphane, Pd
2
(dba)
3 3 4
, K PO , DME, 110 ºC, MW, 3 h; ix) 1ꢀmethylpiperidinꢀ4ꢀamine, tꢀ
BuOK, xantphos, Pd
2
(dba) , toluene, 130 ºC, MW or conventional heating, 3ꢀ10 h.
3
Scheme 3.
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
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
i)
iii)
iv)
24
27
28
2
5, R = CH CH
2 3
ii)
2
6, R = H
v)
vi)
vii)
29
30
31
Conditions: i) ethyl 3ꢀchloroꢀ3ꢀoxoꢀpropanoate, Et N, CH Cl , 0 ºC, followed by rt, 12 h; ii) LiOH·H O,
3
2
2
2
THF/MeOH/H
2
O (3:3:2), rt, 16 h; iii) POCl
3
, 90 ºC, 2 h; iv) K
2
CO
3
, Pd(PPh
3
)
4
, 4,4,5,5ꢀtetramethylꢀ2ꢀ(5ꢀ
methylꢀ2ꢀfuryl)ꢀ1,3,2ꢀdioxaborolane, 1,4ꢀdioxane, 110 ºC, 16 h; v) Cs CO , BINAP, Pd (dba) , 1ꢀ
2
3
2
3
methylpiperidinꢀ4ꢀamine, 1,4ꢀdioxane, 110 ºC, 16 h; vi) Pd/C, H
Cs CO , 1ꢀ(3ꢀchloropropyl)pyrrolidine, DMF, 110 ºC, 16 h.
2
(50 Psi), MeOH, 50 ºC, 16 h; vii)
2
3
On the other hand, compound 31 (Scheme 3) was prepared from commercially available
ꢀbenzyloxyꢀ4ꢀmethoxyꢀaniline (24), which was converted into carboxylic acid 26 by
reaction with ethyl 3ꢀchloroꢀ3ꢀoxoꢀpropanoate and ester hydrolysis. Next, 2,4ꢀ
dichloroquinoline 27 was achieved after heating in POCl . Subsequent reaction with
3
3
4,4,5,5ꢀtetramethylꢀ2ꢀ(5ꢀmethylꢀ2ꢀfuryl)ꢀ1,3,2ꢀdioxaborolane led to intermediate 28. As
described above, the desired substitution at position 4 of the quinoline was then
installed through BuchwaldꢀHartwig amination. Posterior hydrogenation afforded
intermediate 30, which was finally transformed into the desired compound 31 by
reaction with 1ꢀ(3ꢀchloropropyl)pyrrolidine in DMF.
1
8/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 101
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
According to our initial hypothesis, the replacement of the Nꢀcapping group at the 4ꢀ
position of compound 11 had a minor influence on the in vitro potency against both
targets (compound 23a, G9a and DNMT1 IC50 values of 2 and 497 nM, respectively).
From the viewpoint of G9a activity, in general, the cyclohexyl group of 23a can be
replaced by diverse substituents (small groups, alkyl rings, aromatic rings,
heteroaromatic rings and heterocycles), most of them resulting in low nanomolar G9a
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
inhibitors (IC < 50 nM). Analysis of the basicity of the N1 nitrogen can only partially
50
explain SAR results: decreased potency for electronꢀwithdrawing groups such as –CF3
(compound 23f, 637 nM). However, it does not explain the reduced activity of other
small electronꢀdonating groups such as the methoxy group of compound 23e (136 nM).
Here, lipophilic interactions of 6ꢀmembered rings with Val1096 (Figure 1A) seem to
confer G9a potency, as solely an explicit potential hydrogenꢀbond interaction was
predicted for compound 23m (between the NH of its methylpyrrole group and Asp1088,
Figure S2). Interestingly, this compound 23m is the most potent against G9a in Table 1
(IC50 = 0.5 nM). As for compound 11, a substrate competitive mechanism was observed
for 23m (Figure S3), indicating binding site conservation in both targets.
Considering DNMT1 activity, a clear SAR trend was not observed for the diverse
substituents in Table 1. Small substituents such as methyl (23d), methoxy (23e) and –
3
CF (23f) resulted in low micromolar (23d, 23e) or inactive (IC50 > 10 ꢀM, 23f)
compounds against DNMT1. Alkyl rings (23a, 23b) tended to be more potent (IC50
below or around 500 nM) than nonꢀaromatic heterocycles such as Nꢀmethylpiperazine
(
23c) and oxygenꢀcontaining rings (31, 23p, 23q) with DNMT1 inhibitory activity in
the low micromolar range ( M) (Table 1). Regarding aromatic and heteroaromatic
rings, the introduction of 5ꢀmembered heteroaromatic rings such as 5ꢀmethylꢀ2ꢀfuryl
∼
1 ꢀ
1
9/101
ACS Paragon Plus Environment

Page 21 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(
12), 5ꢀmethylꢀ2ꢀthienyl (23l) and 5ꢀmethylꢀ1Hꢀpyrrolꢀ2ꢀyl (23m) resulted in a slightly
increased DNMT1 inhibition compared with the cyclohexyl 23a (0.1ꢀ0.5 log units).
Analysis of the DNMT1 cavity around the 2ꢀposition of the quinoline core did not
reveal any clue to interpret the SAR, probably due to the missing CXXC domain.
However, comparative analysis of the electrostatic potential of these compounds
suggests a preference for negative potentials surrounding the 2ꢀposition of the
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
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
quinolines, as shown in Figure 3 for derivatives 12 (IC50 = 382 nM) and 23m (IC50
=
232 nM). However, as highlighted by compound 23b (IC = 494 nM), this is not a strict
5
0
mandatory requirement.
Figure 3. Electrostatic maps of potent DNMT1 inhibitors 23b, 12 and 23m. The blue
and red grids correspond to positive and negative potentials, respectively.
Concerning the antiꢀproliferative effect of the most biochemically potent compounds in
Table 1 against ALL cell lines (CEMOꢀ1 and primary culture LALꢀCUNꢀ2), DLBCL
(OCIꢀLy3 and OCIꢀLy10) and AML (MV4ꢀ11) cell lines, only compound 12 showed
GI values lower than 1 ꢀM against all five cell lines, indicating that 5ꢀsubstitutedꢀ2ꢀ
5
0
furyl groups at the 2ꢀposition are optimal for cellular activity. Other compounds with
both, similar in vitro inhibitory profile and moderate PAMPA permeability (10 < Pe <
2
0/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 101
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
6
7
3
0 nm/s) to that of 12, had either no effect on proliferation (23a) or were poorer
responders (23b, 23l). Even compound 23m, which displayed high activity in the in
vitro assay against G9a and DNMT1, had minor effect on cellular viability than 12. It
remains to be seen if differences in active transport may account for this difference. For
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
65
passive permeability, compounds with predicted LogD values at pH =7.4 ranging
between 1 and 3 (23a, 23b, 23g, 12, 23l), exhibited better permeability (moderate)
compared to more hydrophilic compounds (e.g. 23i, 23o, 23p, 23q), with Pe PAMPA
values less than 10 nm/s). Cytotoxicity of selected compounds in the healthy hepatic
cell line THLEꢀ2 allowed for a therapeutic window (absolute difference between pLC50
and pGI50) of around 0.8 – 1 log units against their corresponding most responsive
hematological neoplasia cell line (compounds 12, 23b, 23g, 23l and 23m), or even
higher (1.8 log units for compound 23k in OCIꢀLy10 cell line).
Table 1. Exploration of the 2-position
Cpd R1
G9a
DNMT1
PAMP
CEMO- LAL- OCI- OCI- MV4- THL
IC50
IC50
A
Pe
1
CUN- Ly3
2 GI50 GI50
Ly10 11
E-2
[c]
[c]
nM
nM
(nm/s)
GI50
GI50
GI50
LC50
[
d]
[e]
[e]
[e]
[e]
[e]
[f]
nM
nM
nM
nM
nM
nM
2
1/101
ACS Paragon Plus Environment

Page 23 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
[
a]
2
2
2
3a
3b
3c
2
497
494
945
19.8
20.9
20%
>1000 34800
0
6
1175
4290
7580
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
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
[
b]
b]
42
14%
[
2
2
3d
3e
6
929
14%
[a]
136
2290
1%
[
a]
2
2
3f
637
15
>10000
1050
24%
[
a]
3g
16.9
658
7%
717
420
2595
5620
2
3h
52
582
[
a]
[a]
[a
[a
23i
40
1410
1250
2.2
0%
16%
12%
17%
]
]
2
3j
142
5
3
12
8
382
978
230
12.9
2.2
218
664
409
455
386
269
1780
22300
4060
[
b]
23k
7
4560
555
54%
>400
0
[a
[a
2
3l
43
29.1
21%
42%
3095
]
]
22/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 101
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
[
a]
2
2
3m
3n
0.5
232
1650
12%
896
915
7680
1080
>10000
0.9
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
1
2
3
4
5
6
7
8
9
0
[
a]
[a]
[a
[a]
23o
44
27
969
10%
16%
16%
0%
]
[
g]
3
1
1370
[
a]
[a]
[a]
[a
2
2
3p
3q
12
17
957
3.7
2.2
8%
0%
6%
6%
0%
16%
]
[
a]
[a]
[a]
[a]
4420
20%
3%
[a]
[b]
Percentage of growth inhibition at 1 µM of the compound; percentage of growth inhibition at 10 µM
[c]
of the compound. For data in Tables 1ꢀ3 and Charts 1ꢀ3: all biochemical results are the average of at
least two independent replicates performed at different days. If absolute pIC50 difference was higher than
0
.3 log units, additional replicates were performed until satisfying the experimental error (by discarding
[
d]
individual results with values outside 2 MADs of the mean value). The PAMPA assay was performed
[e]
[f]
in triplicate.
Proliferation assays are the average of three replicates at different days.
THLEꢀ2
cytotoxicity results after 72 hours of incubation are the average of at least two independent experiments
performed at different days. If absolute pLC50 difference was higher than 1 log unit, additional replicates
were performed until satisfying the experimental error (by discarding individual results with values
[
g]
outside 3 MADs of the mean value). racemic mixture.
Subsequent SAR exploration of the 2ꢀposition was centered on studying the effects of
the replacement of the methyl fragment of the 5ꢀmethylꢀ2ꢀfuryl group of compound 12
23/101
ACS Paragon Plus Environment

Page 25 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
by diverse chemical moieties having representative potential ligandꢀreceptor interaction
68
patterns.
Thus, compounds 34a-f and 36 in Table 2 were synthesized according to Scheme 4
from compound 16. 4ꢀChloroquinolines 33a, 33f and 35 were directly prepared by
Suzuki coupling from this intermediate, while compounds 33b-e were afforded from
aldehyde 32. Next, the desired 4ꢀaminoquinolines 34a-f and 36 were isolated by
BuchwaldꢀHartwig amination.
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
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
Scheme 4.
i)
i)
i) or ii) or iii) or iv)
32
16
3
3a, R = CH CH
2 3
33, R = CH NHCH CH
3
2
2
i)
33c, R = CH N(CH )
2 3 2
3
3d, R = CH OH
2
33e, R = CN
3f, R = Ph
3
v) or vi)
35
v)
3
4a, R = CH CH
2 3
3
4b, R = CH NHCH CH
2
2
3
3
4c, R = CH N(CH )
2 3 2
36
3
4d, R = CH OH
2
3
4e, R = CN
34f, R = Ph
Conditions: i) corresponding boronic acid or ester, K
2
CO
3
or Na
2
CO
3
, Pd(PPh
)
3 4
, 1,4ꢀdioxane/H
2
O (5:1,
1
0:1 or 15:1), 80ꢀ120 ºC, MW or conventional heating, 1ꢀ12 h; ii) ethylamine or dimethyl amine
hydrochloride, MeOH, rt, 90 minutes, followed by NaBH CN, rt, 12 h; iii) NaBH , MeOH, rt, 2 h; iv)
phenylphosphonic dichloride, pyridine, NH OH·HCl, MeOH/CH Cl (1:4), rt, 15 h, then NaBH CN, rt,
12 h; v) Cs CO , BINAP, Pd (dba) , 1ꢀmethylpiperidinꢀ4ꢀamine, 1,4ꢀdioxane, 120ꢀ130 ºC, MW, 2ꢀ5 h; vi)
ꢀmethylpiperidinꢀ4ꢀamine, tꢀBuOK, xantphos, Pd (dba) , toluene, 130 ºC, MW, 2ꢀ3 h.
3
4
2
2
2
3
2
3
2
3
1
2
3
Table 2. Focused exploration on 5-substituted-2-furyl groups
2
4/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 101
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
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
Cp
d
R1
G9a
DNMT PAMP CEMO LAL-
OCI-
OCI-
Ly10
GI50
nM
IC50
1
IC50
A
Pe -1
CUN-2 Ly3
[
b]
[b]
nM
2
nM
234
618
(nm/s)
23.9
1.3
GI50
nM
56
GI50
nM
577
GI50
nM
95
3
3
4a
4b
64
[a]
[a]
[a]
[a]
48
0%
3%
0%
0%
34c
148
1270
[
a]
[a]
[a]
[a]
3
3
4d
4e
15
759
0.5
0%
19%
26%
1%
264
1600
3
3
4f
2720
3140
1050
1070
6
[a]
Percentage of growth inhibition at 1 µM of the compound. For compound 34a, GI50 and LC50
values in MV4ꢀ11 and THLEꢀ2 cell lines are 679 nM and 2320 nM, respectively.
2
5/101
ACS Paragon Plus Environment

Page 27 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Except for the 5ꢀethylꢀ2ꢀfuryl (compound 34a), the introduction of different polar
substituents with either hydrogen bond donors (compounds 34b, 34c) or hydrogen bond
acceptors (compound 34e) or both (compound 34d) had a detrimental impact on both
biochemical activities, emphasizing a stronger preference for hydrophobic groups.
Bulky aromatic substituents (34f, 36) were not well tolerated, especially for G9a,
probably caused by steric clashes with Val1096. Because no significant improvement
was observed, no further exploration was performed.
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
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
Compound 34a, with similar biochemical profile and slightly higher PAMPA
permeability than 12, showed higher antiꢀtumour activity against all cell lines, except
for MV4ꢀ11. Especially, for CEMOꢀ1, OCIꢀLy3 and OCIꢀLy10 lines (GI50 < 100 nM),
an interesting therapeutic window of 1.4 – 1.6 log units was achieved. In summary, as a
result of the exploration of the 2ꢀposition of the quinoline series, compounds 12 and 34a
were selected for further profiling (see below).
2
SAR and anti-proliferative response: Exploration of the 4-position (R )
Next, a small exploration of the Nꢀcapping group at the 4ꢀposition (cyclopropyl and
isopropyl groups) was carried out by holding constant the two optimal groups (5ꢀ
methylꢀ2ꢀfuryl and 5ꢀethylꢀ2ꢀfuryl) at the 2ꢀposition (Table 3). Our aim was twoꢀfold:
avoiding the potential Nꢀdemethylation in vivo of the 1ꢀmethylpiperidine ring and
decreasing the basicity of the nitrogen of this piperidine to decrease potential hERG
binding.
Compounds 38a and 39a-d were prepared as illustrated in Scheme 5. Starting from
previously described 4ꢀchloroquinolines 22r and 33a, BOCꢀprotected piperidines 37a-b
2
6/101
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 101
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
were isolated after BuchwaldꢀHartwig amination. Next, the BOC protecting group was
removed under acidic conditions, and the desired isopropyl or cyclopropyl substituents
were installed by reductive amination or cyclopropanation, respectively.
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
Scheme 5.
i)
ii)
3
7a, R1 = CH3
7b, R1 = CH CH3
2
2r, R1 = CH3
3
3
3a, R1 = CH CH
2
2
3
iii) or iv)
3
9a, R1 = CH , R2 = isopropyl
3
3
8a, R1 = CH3
8b, R1 = CH CH
39b, R1 = CH CH , R = isopropyl
2
3
2
3
2
3
39c, R1 = CH , R2 = cyclopropyl
3
3
9d, R1 = CH CH , R = cycloprpyl
2 3 2
Conditions: i) tertꢀbutyl 4ꢀaminopiperidineꢀ1ꢀcarboxylate, CS
2
CO
3
, BINAP, Pd
2
(dba)3, 1,4ꢀdioxane, 80ꢀ
130 ºC, 5ꢀ16 h; ii) HCl/EtOAc (1.0 M), 22ꢀ25 ºC, 2ꢀ3 h; iii) acetone, NaBH CN, AcOH, iꢀPrOH, 50 ºC,
3
3
15 h; iv) (1ꢀethoxycyclopropoxy)ꢀtrimethylꢀsilane, NaBH CN, AcOH, MeOH, 60 ºC, overnight.
Table 3. Exploration of the N-capping group at the 4-position
2
7/101
ACS Paragon Plus Environment

Page 29 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Cpd R1
R2
G9a
IC50
nM
DN
PAM CE
LAL-
OCI- OCI-
MV4 THL
MT1 PA
MO- CUN-2 Ly3
Ly10
GI50
nM
-11
E-2
IC50
Pe
1
GI50
GI50
GI50
nM
LC50
nM
nM
(nm/ GI50
nM
nM
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
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
s)
nM
1980 >5000
3
3
8a
9a
ꢀH
6
3
137
231
2.0
1270
898
>4000
792
8850
1340
25.8
38.3
247
649
971
39b
3
284
441
1720
3
3
9c
22
24
413
274
40.4
55.6
453
917
1380
187
94
1590 455
3506 703
9d
1560
236
For DNMT1 activity, the replacement of the methyl group by an isopropyl or
cyclopropyl group had a minor effect (IC50 between 200 and 500 nM, as for 12 and
34a). For G9a, the introduction of the cyclopropyl ring diminished G9a activity (up to 1
log unit in the case of compound 39d), emphasizing the impact of the basicity of the
alkylated nitrogen for G9a inhibitory activity, predictably decreased by the vinylic
69
character of the cyclopropyl moiety. Considering the trend in improved permeability
cyclopropyl > isopropyl > methyl), this drop in G9a potency might be one factor to
(
explain the reduced cellular potency against CEMOꢀ1, OCIꢀLy3 and MV4ꢀ11 cell lines
of compound 39d (low micromolar range) compared to 34a (low to middle nanomolar
range). Notably, compound 39c retained an acceptable growth inhibitory profile and
was even more potent against DLBCL cell lines than compound 12; then, 39c was
28/101
ACS Paragon Plus Environment