Discovery of Asciminib (ABL001), an Allosteric Inhibitor of the Tyrosine Kinase Activity of BCR-ABL1

Article abstract of DOI:10.1021/acs.jmedchem.8b01040

Chronic myelogenous leukemia (CML) arises from the constitutive activity of the BCR-ABL1 oncoprotein. Tyrosine kinase inhibitors (TKIs) that target the ATP-binding site have transformed CML into a chronic manageable disease. However, some patients develop drug resistance due to ATP-site mutations impeding drug binding. We describe the discovery of asciminib (ABL001), the first allosteric BCR-ABL1 inhibitor to reach the clinic. Asciminib binds to the myristate pocket of BCR-ABL1 and maintains activity against TKI-resistant ATP-site mutations. Although resistance can emerge due to myristate-site mutations, these are sensitive to ATP-competitive inhibitors so that combinations of asciminib with ATP-competitive TKIs suppress the emergence of resistance. Fragment-based screening using NMR and X-ray yielded ligands for the myristate pocket. An NMR-based conformational assay guided the transformation of these inactive ligands into ABL1 inhibitors. Further structure-based optimization for potency, physicochemical, pharmacokinetic, and drug-like properties, culminated in asciminib, which is currently undergoing clinical studies in CML patients.

Full text of DOI:10.1021/acs.jmedchem.8b01040

Subscriber access provided by Karolinska Institutet, University Library
Drug Annotation
Discovery of asciminib (ABL001), an allosteric
inhibitor of the tyrosine kinase activity of BCR-ABL1
Joseph Schoepfer, Wolfgang Jahnke, Giuliano Berellini, Silvia Buonamici, Simona Cotesta, Sandra
W Cowan-Jacob, Stephanie Dodd, Peter Drueckes, Doriano Fabbro, Tobias Gabriel, Jean-
Marc Groell, Robert M. Grotzfeld, Asm Quamrul Hassan, Christelle Henry, Varsha Iyer, Darryl
Jones, Franco Lombardo, Alice Loo, Paul William Manley, Xavier Pelle, Gabriele Rummel, Bahaa
Salem, Markus Warmuth, Andrew Wylie, Thomas Zoller, Andreas L. Marzinzik, and Pascal Furet
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01040 • Publication Date (Web): 23 Aug 2018
Downloaded from http://pubs.acs.org on August 23, 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 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Discovery of asciminib (ABL001), an allosteric inhibitor
of the tyrosine kinase activity of BCRꢀABL1
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
#,+
#,+
ǂ
ǂ
Joseph Schoepfer , Wolfgang Jahnke , Giuliano Berellini , Silvia Buonamici , Simona
#
#
##
#
ǂ
Cotesta , Sandra W. CowanꢀJacob , Stephanie Dodd , Peter Drueckes , Doriano Fabbro , Tobias
#
#
#
ǂ
#
Gabriel , JeanꢀMarc Groell , Robert M. Grotzfeld , A. Quamrul Hassan , Chrystèle Henry ,
ǂ
#
ǂ
##
#
#
Varsha Iyer , Darryl Jones , Franco Lombardo , Alice Loo , Paul W. Manley , Xavier Pellé ,
#
#
ǂ
ǂ
#
Gabriele Rummel , Bahaa Salem , Markus Warmuth , Andrew A. Wylie , Thomas Zoller ,
*,#,+
#,+
Andreas L. Marzinzik , Pascal Furet
#
#
Novartis Institutes for BioMedical Research, Novartis Campus, CHꢀ4056 Basel, Switzerland
# Novartis Institutes for BioMedical Research, 250 Mass Ave, Cambridge, MA 02139 USA
ǂ New address
+ These authors contributed equally
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 56
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
ABSTRACT
Chronic myelogenous leukemia (CML) arises from the constitutive activity of the BCRꢀABL1
oncoprotein. Tyrosine kinase inhibitors (TKIs) which target the ATPꢀbinding site have
transformed CML into a chronic manageable disease. However, some patients develop drug
resistance due to ATPꢀsite mutations impeding drug binding. We describe the discovery of
asciminib (ABL001), the first allosteric BCRꢀABL1 inhibitor to reach the clinic. Asciminib
binds to the myristate pocket of BCRꢀABL1 and maintains activity against TKIꢀresistant ATPꢀ
site mutations. Although resistance can emerge due to myristateꢀsite mutations, these are
sensitive to ATPꢀcompetitive inhibitors, so that combinations of asciminib with ATPꢀ
competitive TKIs suppress the emergence of resistance. Fragmentꢀbased screening using NMR
and Xꢀray yielded ligands for the myristate pocket. An NMRꢀbased conformational assay guided
the transformation of these inactive ligands into ABL1 inhibitors. Further structureꢀbased
optimization for potency, physicochemical, pharmacokinetic and drugꢀlike properties,
culminated in asciminib, which is currently undergoing clinical studies in CML patients.
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
2

Page 3 of 56
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
Figure 1: (A) Structures of FDAꢀapproved BCRꢀABL1 inhibitors, type I and type II ATP
1, 2
competitive inhibitors
(B) Structures of compounds employed for Xꢀray crystallographic
3
4
5
studies of assembled inactive state of ABL1 (C) GNFꢀ2 and GNFꢀ5, activities on the short
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 56
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
(
kinase domain) and long (kinase and myristate binding domain) ABL1 constructs (radiometric
filter binding assays) and cell proliferation data in Ba/F3 BCRꢀABL1 (p210) dependent cells
6
(ATPLite™).
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
4

Page 5 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Introduction
Chronic myelogenous leukemia (CML) and some cases of adult acute lymphoblastic leukemia
ALL) arise from a reciprocal translocation between chromosomes 9 and 22 in hematological
(
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
progenitor cells, leading to the fusion of the BCR and ABL1 genes on the Philadelphia
7
chromosome (Ph). Transcription and the resulting expression of the BCRꢀABL1 fusion protein,
which has a constitutively active ABL1 kinase domain, leads to the aberrant activation of
numerous signaling pathways which result in the dysregulated differentiation, growth and
8
survival of leukemic cells. Inhibiting the kinase activity of BCRꢀABL1 by targeting its ATPꢀ
binding site with drugs such as imatinib (Gleevec®), dasatinib (Sprycel®), nilotinib (Tasigna®)
and bosutinib (Bosulif®) greatly reduces leukemic tumor burden and increases the overall
survival rate of CML patients, thereby validating ABL1 as a target for therapeutic intervention
9
ꢀ12
for this disease.
With such targeted tyrosine kinase inhibitors (TKIs), the majority of chronic
phase CML patients continuously treated according to current guidelines have a normal life
1
3
expectancy. However, despite the success of these drugs in transforming the prognosis of a
fatal malignancy, some patients suffer from either poor initial response, loss of response or
1
4, 15
tolerability issues.
Loss of response is frequently associated with drug resistance, which can
mut
emerge when clones harboring mutations in the BCRꢀABL1 kinase (BCRꢀABL1 ) that hinder
1
6, 17
TKI binding, expand to further drive the disease.
Of these mutations the one harboring a
T315I
substitution of the threonine T315 to isoleucine (BCRꢀABL1
) is only inhibited by ponatinib
18
(
Iclusig®), the dosing of which is limited by adverse events. Many of the adverse events
recorded for TKIs which lead to tolerability issues in CML patients result from offꢀtarget
activities and, consequently a drug that specifically targeted ABL kinase activity would probably
have fewer side effects.
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 56
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
All of the TKIs currently approved for the treatment of CML target the catalytic ATP binding
site of ABL1, either in a typeꢀI or typeꢀII binding mode where they bind to either the
1
, 2
catalytically active or catalytically inactive conformation of the SH1 domain.
Although
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
imatinib and nilotinib are more selective than ponatinib and the typeꢀI TKIs, bosutinib and
19,
dasatinib, all of the drugs have distinct profiles leading to the inhibition of many other kinases.
2
0
Moreover, they tend to be susceptible to resistance mutations, a notorious one being the T315I
2
1
mutation involving the gatekeeper residue in the ATP site, which is only inhibited by ponatinib.
3, 22
In 2003, the research groups of Kuriyan and SupertiꢀFurga
reported that a myristoyl group
2
3
is involved in the autoregulation of ABL1 and the closely related ABL2 kinase: By binding to
an internal “myristate binding site / pocket” in the Cꢀlobe of the ABL1 kinase domain, the
myristoyl group that is covalently attached to the Nꢀterminus of ABL1 induces an assembled
inactive state in which the SH3 and SH2 domains dock against the kinase domain to reduce its
conformational flexibility (Figure 2). This mechanism of natural autoinhibition seen in ABL1 is
lost upon fusion with the fragment of BCR which replaces the Nꢀterminal cap region of ABL1
3
that contains the myristoylation site, thus leading to the constitutive activation of BCRꢀABL1.
This assembled inactive state differs from the typical kinase inactive DFGꢀout conformation
where ATP site directed typeꢀII inhibitors stabilize the inactive conformation of the kinase
24
domain by directly binding to the DFGꢀout loop conformation. Conceptually, a drug that
inhibited ABL1 kinase by interacting with this myristoyl site would be more selective than ATPꢀ
competitive TKIs, would maintain activity against resistance mutations and would not be subject
to adverse events resulting from offꢀtarget kinase inhibition, and hence could be well tolerated.
As an approach to identify novel BCRꢀABL1 inhibitors, Adrian and colleagues employed a
differential cytotoxicity screen using BCRꢀABL1 transformed and parental mouse hematopoietic
ACS Paragon Plus Environment
6

Page 7 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
3
2D cells. Whereas parental 32D cells require interleukin 3 (IL3) as a growth factor for
proliferation, the expression of BCRꢀABL1 allows the cells to grow in the absence of IL3 and
2
5
instead depend upon BCRꢀABL1 for proliferation. From a library of heterocyclic compounds,
an Nꢀphenylpyrimidinꢀ4ꢀamine derivative showed remarkable selectivity and subsequent
optimization to improve cellular activity afforded GNFꢀ2, which exhibited potent and selective
antiproliferative activity on BCRꢀABL1ꢀdependent cells (Figure 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
SAR around GNFꢀ2 indicated that a paraꢀCF Oꢀ group and the freeꢀNH of the CF OC H NHꢀ
3
3
6
4
pharmacophore were important for activity. Based on mutagenesis studies where amino acid
substitutions within the myristoyl binding pocket modulated the activity of GNFꢀ2, it was
hypothesized that GNFꢀ2 bound to myristoyl binding pocket of ABL1. Subsequently this
5
hypothesis was validated using NMR and Xꢀray crystallography, which demonstrated that GNFꢀ
2
bound to the myristate binding site and therefore inhibited ABL1 kinase activity allosterically.
In the latter study an analogue with improved pharmacokinetic properties, GNFꢀ5 (Figure 1),
demonstrated efficacy in vivo. Although attempts to exploit this chemotype as a lead expanded
the diversity of allosteric BCRꢀABL1 inhibitors, no clinical development candidates were
2
6
identified.
Following a strategy to elaborate myristate binding site inhibitors as potential drugs for the
treatment of CML, we identified asciminib (ABL001, compound 1), a potent inhibitor of BCRꢀ
2
7
ABL1ꢀdriven cell proliferation, which is currently undergoing clinical evaluation. Compound 1
has been shown to mimic the role of the myristoylated Nꢀterminus of ABL1 by occupying its
vacant binding site, thereby restoring the negative regulation of BCRꢀABL1 kinase activity. We
report here the main aspects of the discovery of compound 1, which include identification of a
weak fragment screening hit employing an NMRꢀbased conformational assay, and subsequent
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 56
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
optimization for potency and drugꢀlike properties, guided by Xꢀray crystallography and
molecular modeling (Figure 3). The in vivo evaluation of compound 1 and analogs in BCRꢀ
ABL1ꢀdependent tumor xenograft models, the full in vitro profile of compound 1, and its
pharmacokinetics in preclinical species are discussed.
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. (A) The assembled inactive state of ABL1 kinase as seen in the minimal autoinhibitory
construct, comprising the SH3, SH2 and SH1 domains, bound to PD166326 (Figure 1; a typeꢀI
3
tyrosine kinase inhibitor) and myristic acid (1opl). (B) The structure of the ABL1 SH1 domain
2
8
in complex with imatinib (2hyy). Note the two different conformations of helix I, which is bent
in the assembled inactive state (panel A, orange), but linear and partially disordered in the
absence of autoinhibition (panel B, light blue).
ACS Paragon Plus Environment
8

Page 9 of 56
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
Figure 3. From fragment to clinical candidate: Medicinal chemistry progression from fragments
2
7
to fragmentꢀderived hit 6, to lead 7 and finally clinical candidate, asciminib (ABL001, 1),
showing some NMRꢀderived K and GI values from the Ba/F3 cellular proliferation assay
d
50
6
(ATPLite™).
Fragmentꢀbased lead discovery
Attempts to discover novel ligands for the myristate pocket of ABL1 kinase started with a
fragment screen using NMR spectroscopy. Since only allosteric inhibitors of ABL1 were of
interest, the ATPꢀsite was blocked with imatinib and the ABL1ꢀimatinib complex was screened
using NMR T1ρ and water LOGSY ligand experiments. By screening approximately 500
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 56
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
29
diverse, qualityꢀcontrolled and soluble fragments , 30 hits were identified. The dissociation
constants of the hits were determined by NMR by either direct titration, measuring chemical shift
1
changes in a oneꢀdimensional H spectrum of the ABL1ꢀimatinib complex, or by competition
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
with an CF Oꢀ containing tool compound derived from GNFꢀ2, which had a K of (7.4 ± 1.0)
3
d
µM, as determined by direct NMR titration. Both NMRꢀbased methods for K determination
d
were compared for a subset of hits, and the results were in full agreement. Figure 3 shows the
progression from screening hits to followꢀup compounds, with NMRꢀderived K values.
d
Screening hit 2 was remarkable in terms of its affinity (K = 6 µM) and low molecular weight
d
(
185 Da), corresponding to a ligand efficiency of 0.6. SARꢀbyꢀarchive analysis showed clear
trends for 2, with replacement of Cl with Me (K = 12 µM), CF (K = 14 µM) or F (K = 31
d
3
d
d
µM) resulting in a loss in potency, whereas a replacement with Br gave an increase in potency
K = 1 µM).
(
d
None of the fragment hits or initial followꢀup compounds were active in a biochemical ABL1
kinase assay, which monitored substrate phosphorylation by capillary electrophoresis. This was
not surprising since at the time, the assay employed only the kinase domain of ABL1, and lacked
the minimal autoinhibitory construct (Figure 2A) capable of forming an assembled inactive state
induced by allosteric inhibitors. More surprisingly, none of them were active at concentrations <
5
0 uM in cellular assays. The cellular assays comprised of proliferation or autophosphorylation
assays in Ba/F3 BCRꢀABL1 dependent cells (viability determined by luminescent ATP
6
detection, ATPLite™ ) which were established as being sensitive to ATPꢀcompetitive BCRꢀ
ABL1 inhibitors and with BCRꢀABL1 inhibition being highly correlated with antiproliferative
1
9
activity. Although fragment hits are generally not expected to be active in cellular assays, our
ACS Paragon Plus Environment
10

Page 11 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
hits and initial followꢀup compounds exhibited low micromolar K values, and comprised of
d
several chemotypes, such that cellular activity was expected at a concentration of 50 µM.
A selection of the fragment hits were crystallized as ternary complexes with the ABL1ꢀ
imatinib complex. The Xꢀray crystal structures of these complexes confirmed that the hits bound
to the myristate pocket and revealed why they were not functional in cellular assays, thus
providing a path forward towards functional inhibitors: the structures, such as that structure
involving fragment hit 2 (Figure 3) showed that the binding mode of the fragment was
incompatible with bending of helixꢀI, which is a conformational change necessary for 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
3
0
formation of the assembled inactive state (Figure 2). In fact, 2 as well as other fragment hits
would sterically clash with helixꢀI in the bent conformation. Therefore, hits such as 2 are not
allosteric inhibitors of BCRꢀABL1, but would actually activate ABL1 by interfering with its
3
0, 31
autoinhibition.
Figure 4. (A) Crystal structure of ABL1 kinase domain in complex with imatinib and fragment
hit 2. (B) Structural detail, showing the position of 2 in the myristate pocket of ABL1 kinase.
Superimposed from the structure of ABL1 kinase in complex with myristic acid (1opl) are the
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 56
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
myristic acid ligand (orange), as well as helixꢀI (orange), in order to show the exit vector of 2,
and the steric clash of helixꢀI (side chain of isoleucine I521 displayed) with the ester group of 2
(PDB code 3MS9, 1.80 Å).
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
Based on the crystal structures, a strategy was developed to design analogs of 2 (Figure 4)
which would bind without sterically clashing with the bent conformation of helixꢀI, and expected
to be functional inhibitors. Removal of the methyl ester moiety would avoid the steric clash, and
substituting the aromatic amine with an aliphatic carbon chain such as in the natural myristate
ligand should restore the potency lost upon removal of the ester. The resulting compound 3
retained the affinity of the original fragment 2 (K = 4 µM) but was still inactive in cellular
d
assays.
In another approach to design allosteric inhibitors that bind solely to the bent conformation of
helixꢀI, we performed similarity and pharmacophore searches based on the crystal structures of
multiple crystallized fragments, and docked the resulting hits into the bent helixꢀI conformation
of the myristate pocket. Compounds such as 4 were identified, which bound with moderate
affinity (K = 6 µM), but still showed no cellular activity.
d
We then questioned whether our assumption of the ability of followꢀup fragments such as 3
and compound 4 to bend helixꢀI was correct. At this stage, the crystal structures only included
the kinase domain of ABL1, and in this system helixꢀI was not visible. We therefore used NMR
15
spectroscopy and developed an assay using a selectively Nꢀvalineꢀlabeled ABL1 kinase domain
3
0
to monitor the conformation of helixꢀI in solution. This was used to monitor valine V525 on
helixꢀI, taking advantage of the fact that V525 is in a flexibly disordered region when helixꢀI is
linear, but gets fully ordered when the helix is bent. Since the lineꢀwidth and intensity of an
NMR signal depends on the local correlation time and thus on the local flexibility, V525 gives
ACS Paragon Plus Environment
12

Page 13 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
15
1
rise to a strong and sharp signal in N, HꢀHSQC when helixꢀI is linear, while it gives rise to a
peak with average intensity when helixꢀI is bent. This is illustrated in Figure 5, with the
unliganded state as reference for the linear state of helixꢀI (Figure 5B), and the myristateꢀbound
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
state as reference for the fully bent helix (Figure 5C). Similar to the myristoylated peptide, GNFꢀ
30
2
also leads to a fully bent helixꢀI (Figure 5D).
Figure 5. The NMRꢀbased conformational assay allows the conformation of the Cꢀterminal helix
32
I to be monitored, particularly distinguishing the linear (light blue) and bent (orange) states.
The principle of the assay is depicted in panel (A), which shows a superposition of two ABL1
structures, one with linear helixꢀI (light blue) and one with bent helixꢀI (orange). Panels (BꢀF)
show results of the HSQC NMR conformational assay with the myristate pocket in various
states: with unliganded myristate pocket (B), with the myristate pocket complexed with a
2
2
4
myristoylated peptide (C), or in complex with GNFꢀ2 (D). Panel (E) shows that compound 4
does not bend helixꢀI, although it was designed to do so. However, optimized compound 5
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 56
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
wt
induces bending of helixꢀI (F) and consequently displays BCRꢀABL1 inhibition in cellular
assays.
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
To our surprise, followꢀup compounds 3 and 4, although designed to bind to the bent
conformation of helixꢀI, were shown by the conformation assay not to bind to this conformation
(
Figure 5E). This explains why those compounds, although binding to the myristate pocket of
ABL1, cannot induce the assembled inactive state and are therefore not BCRꢀABL1 inhibitors.
In GNFꢀ2, the CF Oꢀ group attached at the paraꢀposition of the aniline was found to be essential
3
for inhibitory activity. Our fragment hits lacked a CF Oꢀ group in this position. We reasoned that
3
the introduction of a CF Oꢀ group could induce helixꢀI bending and impart inhibitory activity.
3
Indeed, the resulting compound 5, having a K of 10 µM, showed a bent helix I in the
d
conformation assay (Figure 5F). Furthermore, this profile translated into a GI value of 8 µM in
5
0
wt
6
BCRꢀABL1 Ba/F3 cells (ATPLite™). Removal of the hydroxy group and the chlorine atom
from the benzamide core and replacement of the Nꢀmethylꢀpiperazine by a morpholine ring led
wt
to compound 6 having K = 2 µM and GI = 2 µM in the same BCRꢀABL1 Ba/F3 derived cell
d
50
line. Thus, with the help of the NMR conformational assay, we understood the molecular
requirements for allosteric inhibition of BCRꢀABL1, and were able to rationalize the design and
synthesis of allosteric BCRꢀABL1 inhibitors. Compound 6 provided the basis for further
optimization.
While the NMR assay was used to generate early SAR, the development of cellꢀfree
biochemical assays capable of quantifying inhibitory effects of myristate ligands was critical to
6
optimize activity. Thus, we developed a radiometric filter binding assay and later a microfluidic
capillary electrophoresis (Caliper) assay with the minimum autoinhibitory construct (comprising
ACS Paragon Plus Environment
14

Page 15 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the SH3ꢀSH2ꢀkinase domain) which could form the assembled inactive state. It proved important
to not add any surfactants to these assays since, due to their structural similarity with myristate,
they could block the myristate binding site and prevent the detection of myristate ligands. Using
the radiometric filter binding assay without detergent, the inhibitory potencies of myristate
ligands could be measured, as for example with compound 6, having an IC value of 0.55 µM in
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
5
0
the Caliper assay (Table 1).
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 56
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
C
E
B
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
D
F
Figure 6. (A) Crystal structure of ABL1ꢀimatinib complex bound to compound 6 (view of the
myristate binding pocket; PDB code 6HD6, 2.30 Å). Hꢀbonds are indicated by black dotted lines.
The orthogonal polar interaction between one of the fluorine atoms of the CF Oꢀ group and the
3
ACS Paragon Plus Environment
16

Page 17 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
carbon atom of the backbone carbonyl of leucine L359 is indicated by a red dotted line. The
same applies for the other figures. (B) The two possible orientations of the morpholine ring of 6
in the crystal structure. (C) Binding model of designed compound 8 in the myristate pocket. (D)
Binding model of designed compound 7 in the myristate pocket. (E) Crystal structure of ABL1ꢀ
imatinib complex bound to compound 7 (PDB code 6HD4, 2.03 Å). (F) Crystal structure of
ABL1ꢀimatinib complex bound to compound 1 (PDB code 5MO4, 2.17 Å).
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
The crystal structure of compound 6 (Figure 6A) shows the ligand bound in the myristate
pocket of the ABL1ꢀimatinib complex. Notably, the Oꢀ group of 6 binds deeply into the narrow
lipophilic channel, with space only available for substitution in the paraꢀposition of the aromatic
5
ring, as seen in crystal structure of GNFꢀ2 bound to the ABL1ꢀimatinib complex. There is a
favorable polar interaction between one of the fluorine atoms and the carbon atom of the
backbone carbonyl group of leucine L359. While the carbonyl of the amide bond in compound 6
does not make any interaction with the protein, the NH is involved in waterꢀmediated Hꢀbonding
interactions to the backbone carbonyls of alanine A452 and glutamic acid E481. The crystal
structure of 6 shows very good electron density of the aromatic core, but the 3ꢀ
morpholinomethyl is disordered and shows only weak interactions within the pocket, most of it
protruding into the solvent. The electron density is best rationalized by two different binding
modes of the morpholine ring where in one binding mode the morpholine is pointing into the
solvent, in another binding mode the morpholine is flipped by almost 180°, binding along a
lipophilic cleft at the exit of the myristate channel (Figure 6B).
The existence of two different binding modes revealed by the coꢀcrystal structure pointed to
some opportunities to increase the affinity of 6 by better filling the lipophilic cleft revealed by
the alternative conformation of the morpholine. We also noticed that an accessible hydrophobic
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 56
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
surface, formed by the side chain of leucine L529 on the other side of the myristate channel
opening, was not leveraged by the inhibitor. To exploit the latter observation analogues were
designed having an aryl or heteroaryl group directly attached at the phenyl positionꢀ3 in place of
the basic morpholine. In modeling experiments, such groups could form favorable hydrophobic
contacts with the side chain of leucine L529 as illustrated in Figure 6C. A small set of biaryls
designed according to this concept, in which the 3ꢀmorpholinomethyl substituent of 6 was
replaced by lipophilic, neutral or weakly basic, unsubstituted diazines, substantially improved
potency. For example compound 8, having a pyrimidine attached to the metaꢀposition of the
benzamide had an IC value of 18 nM in the Caliper assay. Compound 8 was a milestone 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
5
0
the project since it not only had remarkable biochemical potency, but also inhibited the
wt
proliferation of LucꢀBa/F3 BCRꢀABL1 cells (GI = 150 nM; in addition to being transfected to
5
0
express wildꢀtype BCRꢀABL1, these cells were engineered to express a puromycinꢀluciferase
fusion cassette to enable the direct detection of cell numbers using the BriteLite™ proliferation
T315I
assay) and LucꢀBa/F3 BCRꢀABL1
cells (GI = 1.8 µM; Table 1). In vitro profiling of 8
50
revealed low solubility (HTꢀSol FaSSIF = 0.01 mM, Table 1), and PK studies in mice indicated
high clearance (75 mL/min/kg) and negligible exposure after oral administration (Table 2). To
impart in vivo activity, our strategy focused on increasing the potency by picking up further
interactions and optimizing the physicochemical properties of compound 8.
To improve solubility we employed three approaches (Table 1): i) reducing clogP by
2
incorporating nitrogen at various positions in ringꢀB, ii) introducing an R substituent ringꢀB
1
1
ortho to R , with the aim to push aromatic residues in R outꢀofꢀplane to avoid intermolecular πꢀπ
2
stacking, iii) adding a solubilizing group to R . By incorporating a nitrogen in ringꢀB, the logP of
3
3
8 decreased from 4.3 to 2.7 in 9 and solubility (highꢀthroughput kinetic solubility) increased
ACS Paragon Plus Environment
18

Page 19 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
from 0.01 mM to 0.026 mM. Overall, compound 9 showed the best balance of solubility,
biochemical potency (IC50 0.024 µM) and antiproliferative activity against Luc-Ba/F3 BCRꢀ
WT
ABL1 cells (GI 0.078 µM). Compound 9 also showed low µM activity against Luc-Ba/F3
5
0
T315I
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
BCRꢀABL1
(GI50 1.65 µM) and did not show hERG binding up to 30 µM in a highꢀ
throughput dofetilide binding screening assay (this research assay was not compliant with Good.
3
4
Laboratory Practice (GLP) regulations).
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 56
Table 1. From hit to lead: Comparison of effects of BCRꢀABL1 in biochemical assay (IC ) and in cellular proliferation assays (GI ) with BCRꢀ
5
0
50
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
wt
T315I
ABL1 and BCRꢀABL1
dependent LucꢀBa/F3 cells (data is mean of n ≥ 2, ± SD). Comparison of selected physicochemical and in vitro
pharmacological properties of compounds 6ꢀ14.
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
2
164ꢀ515
Cpd
R
X
ABL1
IC (ꢀM)
LucꢀBa/F3
BCRꢀABL1
LucꢀBa/F3
BCRꢀABL1
clogP /
logP
pKa
FASSIF
(mM)
HTꢀ
perm
calc FA
Rat liver
microsoꢀ
me CLꢀ
hERG
dof.
Binding
IC50
wt
T315I
50
GI (ꢀM)
GI (ꢀM)
50
50
1
(
mLmin
(%)
ꢀ1
kg )
(ꢀM)
6
7
ꢀ
0.55 ± 0.69
0.0023 ±
0.253; 0.341
2.93; > 10
3.6 / 4.2
2.0 / 3.0
6.2
3.3
0.499
0.59
100
56
147
3.7
N
0.0017 ±
0.0001
0.073 ± 0.006
40
9.6
0
.0009
8
9
H
H
CH
N
0.018 ± 0.004
0.024 ± 0.004
0.011 ± 0.001
0.019 ± 0.004
0.117 ± 0.065
0.078 ± 0.018
0.004 ± 0.001
0.020 ± 0.003
1.80 ± 0.44
1.65 ± 0.44
0.80 ± 0.09
1.82; 1.59
3.3 / 4.3
2.5 / 2.7
2.9 / 4.6
3.7 / >3.4
3.3
n/a
n/a
8.9
0.01
0.026
0.16
0.5
100
98
28
20
16
60
> 10
>30
1
0
1
OMe
CH
CH
100
99
1.1
1
0.009
1
2
3
NH(CH
)
2
OH
3
N
N
0.018 ± 0.008
0.004; 0.007
0.511; 0.543
0.396 ± 0.043
2.7 / 3.2
3.1 / n/a
3.7
7.4
0.013
0.151
36
98
60
34
1.5
1
0.007 ± 0.001 0.005 ± 0.0001
0.004 ± 0.001 0.004 ± 0.001
0.31
14
N
0.294 ± 0.066
2.0 / 3.5
3.7
>1
40
50
17
ACS Paragon Plus Environment
2
0

Page 21 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
For further optimization, the pyrimidine core of 9 was kept constant and variations of the
2
substituent R were explored by employing structureꢀguided design (e.g. targeting Hꢀbonding at
the mouth of the pocket to further enhance binding), and adding groups to further improve
2
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
solubility. It was hypothesized that a substituent R ortho to R might improve potency through
having an improved topological fit to the binding site, while also promoting a torsion angle
between ringꢀB and the pyrimidine and thereby increase solubility by impacting crystal packing.
Compound 10 incorporated a CH Oꢀgroup, substantially improved solubility from 0.01 mM to
3
0
.16 mM and this was further enhanced by linking a tertiary amine substituent as in compound
1
1 (0.5 mM), which also exhibited good cellular potency (Table 1). Pyridine derivatives with
2
appropriate R substituents had improved solubility without possessing a strongly basic group,
and compound 12 potently inhibited the growth of LucꢀBa/F3 cells dependent upon BCRꢀ
wt
T315I
ABL1 and BCRꢀABL1
with GI values of 5 and 527 nM, respectively. Incorporating basic
50
groups, such as in the methylpiperazine analog 13, substantially enhanced solubility, but
imparted submicromolar affinity to hERG.
In an attempt to eliminate the hERG signal, molecules were designed to maintain the polarity
of the solubilizing group without incorporating a basic center. Molecular modeling studies
suggested that a 3ꢀhydroxyꢀpyrrolidine group would make hydrophobic contacts with the side
chain of Leu529 and form an Hꢀbond with the sideꢀchain of Glu481, located at the entrance of
the cavity. However, we realized that if we wanted to keep the pyrimidine ring, this had to bind
on the other side of the cavity, when compared to compound 8. In fact, the presence of the
hydroxyꢀpyrrolidine group increased the pyridineꢀpyrimidine torsion angle resulting in a steric
clash with Leu529 in the binding orientation hypothesized for this ring in compound 8 (Figure
6C), To our satisfaction, molecular modeling indicated that the pyrimidine ring could be
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 56
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
accommodated in this new orientation, where it made favorable hydrophobic contacts with
threonine T453, methionine M456, proline P480 (the residues forming the hydrophobic cleft
revealed by the alternative conformation of the morpholine group of 6 in the Xꢀray structure) and
possibly a Hꢀbond with the sideꢀchain of arginine R351 as depicted in Figure 6D.
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
Indeed, adding a 3ꢀhydroxyꢀpyrrolidine at R provided another boost in antiproliferative
activity (3, GI 2 nM). Compound 7 also displayed reduced hERG binding (9.6 ꢀM) compared
5
0
to the methylpiperazine 13 (0.31 ꢀM), although this still translated into activity in nonꢀGLP
hERG Qpatch (IC 4.2 µM) and patch clamp (IC 4.5 µM) assays. On comparing both
5
0
50
enantiomers the Sꢀenantiomer, 7, exhibited marginally better biochemical and cellular potency
compared to 14, and with this the project reached a milestone by clearly demonstrating the
potential of the series to serve as a starting point towards a clinical candidate. The next section
documents the in vitro and in vivo profiles of lead compound 7, followed by its optimization
culminating in the discovery of the clinical candidate 1.
Lead Characterization
The crystal structure of compound 7 bound to the ABL1ꢀimatinib complex showed a wellꢀ
defined electron density for the entire complex, confirming that the pyrimidine ring occupied the
hydrophobic channel formed by threonine T453, methionine M456 and proline P480 of the
myristate binding pocket (Figure 6E). In addition, the pyrimidine participated in waterꢀmediated
Hꢀbonds with the backbone nitrogen of glutamate E481, of the carbonyl oxygen with the sideꢀ
chain hydroxyl group of the tyrosine Y454 and from the nitrogen atom of the central pyridine.
As anticipated, the hydroxyꢀpyrrolidine group made favorable van der Waals interactions with
ACS Paragon Plus Environment
22

Page 23 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the surface formed by the sideꢀchains of E481 and L529, but did not participate in polar
interactions with the protein. The hydroxyl group remained solvent exposed rather than forming
an Hꢀbond with E481 as hypothesized in the design.
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
Replacing the typical phenylaminopyrimidine TKI pharmacophore as present in GNFꢀ2, with a
Nꢀphenylnicotinamide scaffold largely abrogated the risk of offꢀtarget TKI activity, as
demonstrated by profile of 7 in the KINOMEscan® (DiscoveRx, San Diego, U.S.A.) active
35
ATPꢀsiteꢀdirected competition binding assay, which contained more than 440 kinases.
However, it should be noted that a limitation of this platform is that it is based upon recombinant
proteins containing just the catalytic portion of the SHꢀ1 kinase domain, e.g. ABL1(229ꢀ512)
compared to ABL1(64ꢀ515) used in the present publication, and therefore potential myristoyl
binding sites are not present.
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 56
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
Table 2. Comparison of pharmacokinetic data for compounds 1, 7, 8 and 18ꢀ20. Compounds
were administered to mice as solutions for intravenous and oral use, and/or respectively as
*supensions (amorphous material) for oral use. AUC and Cmax values are dose normalized.
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
Cpd
1
Route
dose; mg/kg)
i.v. (1)
p.o. (30)
i.v. (1)
AUC
(ꢁmol.h/L)
Cmax
(ꢁmol/L)
Tmax
(hr)
F
(%)
CL
(mL/min/kg)
t½
(h)
1.1
(
3.2
12
1.7
0.4
2.0
53
7
8
4.2
8.3
2
p.o. (10)
4.2
1.2
0.8
1
99
76
*
*
p.o. (10)
i.v. (1)
3.2
0.8
0.61
0.005
1.9
75
22
11
26
0.2
1.1
1.0
0.8
p.o. (10)
i.v. (1)
0.01
0.6
0.33
1.3
0.8
>100
73
18
19
20
p.o. (30)
i.v. (1)
2.5
3.3
p.o. (30)
i.v. (1)
2.4
0.6
1.7
1.4
p.o. (30)
2.8
0.7
2.0
>100
Although replacing the benzylꢀmorpholine moiety of 6 with a biaryl ring system markedly
reduced the intrinsic clearance in rat liver microsomes (Tables 1 and 2), it was necessary to
demonstrate that compounds had adequate pharmacokinetic properties in mice before proceeding
with efficacy studies in murine models of leukemia. The pharmacokinetic properties of 7
compared to 8 supported the potential of the former compound to serve as a high quality lead
(
Table 2). Thus, mice were treated with either 7 or 8 at 1 mg/kg i.v. and 10 mg/kg p.o.:
Following i.v. administration the clearance of compound 7 was substantially slower than that of
, resulting in a longer halfꢀlife. Most strikingly, when 7 was administered orally it afforded
8
excellent bioavailability and exposure, with the mean plasma level 7 hours after administration
ACS Paragon Plus Environment
24

Page 25 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of 10 mg/kg (C7hr 3.1 ± 1.0 µM) still remaining >600ꢀfold the GI value (4.8 ± 0.5 nM) in our
50
wt
LucꢀKCL22 BCRꢀABL1 cellꢀline engineered from cells derived from a patient with blast crisis
3
6
disease . Considering the mouse plasma protein binding of 96% and a maximal fraction
unbound of 100% in the cell proliferation assay (incubation medium contained 10% fetal calf
serum) the concentration remaining after 7 h would be at least >20ꢀfold over the GI value.
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
5
0
The overall profile of 7 translated into efficacy in a murine, subcutaneous xenograft model of
CML, using the aboveꢀmentioned cellꢀline, where once daily oral administration of 7 to tumorꢀ
bearing mice resulted in doseꢀdependent growth inhibition after 7 days. At a dose of 30 mg/kg
afforded 15% tumor growth compared to that of vehicle control (T/C 15%), with no loss in body
weight (Supplementary Figure 1).
36
From lead compound 7 to clinical candidate 1
Although modifications of the CF Oꢀgroup had been unsuccessful for GNFꢀ2 related
3
2
6
compounds, in order to further optimize the potency and in vivo efficacy of lead compound 7,
the effect of modifications to this group in our chemotype were examined. Analysis of the crystal
structures of compound 6 or 7 bound to the ABL1ꢀimatinib complex showed the CF Oꢀgroup to
3
3
7
sit in the deepest part of the myristate pocket, adopting its characteristic orthogonal orientation
with respect to the plane of the phenyl ring, and with one fluorine atom making a highly
directional polar interaction with the carbonyl carbon of leucine L359 deep in the pocket.
Molecular modeling indicated that a slightly larger group could be accommodated in this part of
the cavity leading to the idea of either replacing the oxygen atom by a sulfur, or replacing one of
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 56
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 fluorine atoms with chlorine. Ab initio calculations predicted that such modifications should
not alter the preference for the desired orthogonal conformation of the group, and even suggested
an increased stabilization in the case of the oxygenꢀsulfur exchange (Supplementary Figure 2).
Remarkably, compounds 15 and 16, which incorporated these modifications, provided an
additional order of magnitude gain in potency over compound 7 in biochemical and cellular
assays. To further optimize these extremely potent allosteric ABL1 kinase inhibitors the
introduction of an additional hydroxyl group in the pyrrolidine ring of 16, a variation designed to
further improve solubility (compound 17), weakened the cellular activity.
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
26

Page 27 of 56
Journal of Medicinal Chemistry
Table 3. Comparison of ABL1 / BCRꢀABL1 inhibition of compounds progressing from the lead pyrimidine series A to the pyrazoles B (lead
compound 7 to clinical candidate 1). ABL1 inhibition was assessed in the biochemical transphosphorylation caliper assay (IC ) and BCRꢀABL
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
5
0
wt
T315I
inhibition was assessed in cell proliferation assays (GI ) with either BCRꢀABL1 of BCRꢀABL1
activity and selected physicochemical parameters of compounds 1 and 15ꢀ20 for dofetilide RLB ( n = 1, * n ≥ 2) Qpatch and patch clamp assays.
transformed LucꢀBa/F3 cells (n ≥ 3). hERG
5
0
#
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
Luc-Ba/F3
hERG IC (µM) or % inh. @ (µM)
5
0
6
4ꢀ615 BCRꢀABL1 GI (nM)
5
0
ABL1
clogP / logP
pKa
Cpd
1
Structure
Dofetilide
RLB*
Patch
clamp
IC (nM)
5
0
Qpatch
wt
T315I
1
5 @ 30
B
3.2 / 3.9
4.0
>30
26
X = Cl
0.5 ± 0.1
1.0 ± 0.1
25 ± 3
ꢀ1 @ 10
R = R = H
2
3
X = F,
Y = S
2
.6 / 4.6
.4 / 3.7
3.5
3.8
3.3
10.5
3.6
1.5
1.8
n/a
5.0
n/a
1
1
1
5
6
7
A
0.4 ± 0.05
< 0.3
< 0.3
12.4 ± 3.0
12.8 ± 1.6
213 ± 45
X = Cl,
Y = O
2
A
A
0.4 ± 0.05
9.0 ± 1.4
X = Cl,
Y = O
#
1.9 / 3.3
2.8 / 3.9
>30
>30
0.5 ± 0.04
R = OH
1
6
4
6 @ 30
@ 10
B
4.0
3.8
>30, 23
>30
>30
>30
1
8
9
X = F
X = F
1.1 ± 0.1
1.6 ± 0.3
2.6 ± 0.5
2.7 ± 0.5
90 ± 16
94 ± 17
R = R = H
2
3
B
= Me,
R = H
3
.3 / 4.4
.5 / 5.1
1 @ 10
1
R
2
3
B
= H,
3
3.6
18.5, >30
>30
16 @ 10
2
0
R
2
X = Cl
0.7 ± 0.2
0.6 ± 0.1
11 ± 3
R = F
3
ACS Paragon Plus Environment
2
7

Journal of Medicinal Chemistry
Page 28 of 56
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
Lead compound 7, together with 15 and 16, showed good potency and acceptable
physicochemical properties (Table 1 and 3), although their hERG activity in our highꢀ
throughput, nonꢀGLP screening assays was of concern. Compound 7 inhibited dofetilide binding
3
4
+
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
to hERG with an IC of 9.6 µM in a radioligand binding assay (RLB) assay, inhibited K flux
5
0
38
through hERG with an IC of 4.2 µM in an automated patch clamp (QPatch) assay and with an
50
3
9
IC of 4.5 µM in a manual patch clamp assay (Table 3). Reduction of the potential cardiac
50
4
0
safety risk associated with hERG inhibition was considered a key priority. Although several ion
41
currents and channels are involved in the cardiac action potential, we will here only discuss the
deꢀrisking strategy for QT prolongation based on minimization of hERG inhibition, since it is
well established that preclinical studies measuring hERG current are predictive for QT
42, 43
prolongation in patients.
With the aim of reducing the hERG activity of compound 7,
without reducing potency, we synthesized analogs in which the pyrimidine ring was replaced by
other moieties. From the crystal structure of 7 (Figure 6E) we discerned that the pyrimidine was
in close proximity to the backbone carbonyl group of glutamic acid E481. This suggested that
this carbonyl group could be a target for Hꢀbonding which might be achieved by replacing the
pyrimidine with a fiveꢀmembered heterocycle bearing an Hꢀbond donor functionality. Docking
studies indicated that a pyrazole ring was suitable for this purpose and compounds 1 and 18ꢀ20
were synthesized to explore this concept (Tables 2 and 3).
Although the pyrazole derivatives (1 and 18ꢀ20) were slightly less potent as ABL1 kinase
inhibitors compared to their pyrimidine counterparts (3, 15 and 16), their hERG activity was
markedly reduced. This finding cannot simply be attributed to changes in lipophilicity and
4
4
ionization state , since replacing the pyrimidine with pyrazole only slightly increased logP
(compare the matched pair, 7 and 18) and all of the compounds are weakly basic and largely
ACS Paragon Plus Environment
28

Page 29 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
44
unionized at physiological pH. Although hERG activity could also be reduced in the
pyrimidine series, as in the dihydroxyꢀpyrrolidine derivative 17, this was associated with a
marked decrease in cellular potency and this approach was not pursued. Our strategy of replacing
pyrimidine with pyrazole would appear not to be a general strategy to reduce hERG activity,
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
45
since such replacement has been reported not to influence hERG activity.
The pyrazole derivatives combined good antiproliferative activity with strongly reduced hERG
inhibition and lower QT prolongation risk, and consequently these were further profiled.
Compounds 1 and 18ꢀ20 all had similar cellular potency (GI 0.6ꢀ2.7 nM in the LucꢀBa/F3
5
0
WT
46
BCRꢀABL1 cell line; Table 3) similar in vitro clearance and permeability in liver microsome
4
7
and Cacoꢀ2 assays (Supplementary Table 1) and similar PK profiles in mice (Table 2).
However, the compounds were differentiated in terms of efficacy in our KCLꢀ22 mouse
xenograft model: When administered by oral gavage at 3 mg/kg BID for 1 week, compounds 1
and 20 afforded 68ꢀ70% tumor growth inhibition, whereas compounds 18 and 19 showed no
efficacy at this dose and schedule (Figure 7A). The increased efficacy of 1 and 20 over 18 and 19
might possibly result from the higher tumor concentrations achieved at 7 h for 1 and 20 (C7hr
0
.16 ± 0.04 and0.20 ± 0.03 µM) compared to 18 and 19 (C7hr 0.08 ± 0.01 and 0.10 ± 0.02 µM);
though all compounds 1, 19ꢀ20 showed overall similar tumor exposures at 1 h (~C1hr 0.2 µM)
Supplementary Table 3). Consequently, compounds 1 and 20 were regarded as being viable
(
development candidates.
Formulation of compound 1
Compound 1 is weakly basic (pKa 4.0) and exhibits a typical pHꢀdependent solubility profile,
with high solubility at acidic pH and low thermodynamic solubility at neutral pH
ACS Paragon Plus Environment
29