Creation of a Novel Class of Potent and Selective MutT Homologue 1 (MTH1) Inhibitors Using Fragment-Based Screening and Structure-Based Drug Design

Article abstract of DOI:10.1021/acs.jmedchem.7b01884

Recent literature has both suggested and questioned MTH1 as a novel cancer target. BAY-707 was just published as a target validation small molecule probe for assessing the effects of pharmacological inhibition of MTH1 on tumor cell survival, both in vitro and in vivo.(1) In this report, we describe the medicinal chemistry program creating BAY-707, where fragment-based methods were used to develop a series of highly potent and selective MTH1 inhibitors. Using structure-based drug design and rational medicinal chemistry approaches, the potency was increased over 10,000 times from the fragment starting point while maintaining high ligand efficiency and drug-like properties.

Full text of DOI:10.1021/acs.jmedchem.7b01884

Subscriber access provided by - Access paid by the | UCSB Libraries
Article
Creation of a novel class of potent and selective MutT Homolog 1 (MTH1)
inhibitors using fragment-based screening and structure-based drug design
Fredrik Rahm, Jenny Viklund, Lionel Tresaugues, Manuel Ellermann, Anja Giese, Ulrika Ericsson,
Rickard Forsblom, Tobias Ginman, Judith Günther, Kenth Hallberg, Johan Lindström, Lars Boukharta
Persson, Camilla Silvander, Antoine Talagas, Laura Díaz-Sáez, Oleg Fedorov, Kilian V. M. Huber,
Ioanna Panagakou, Paulina Siejka, Matyas Gorjanacz, Marcus Bauser, and Martin Andersson
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01884 • Publication Date (Web): 27 Feb 2018
Downloaded from http://pubs.acs.org on February 28, 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.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Creation of a novel class of potent and selective
MutT Homolog 1 (MTH1) inhibitors using
fragmentꢀbased screening and structureꢀbased drug
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
#
§
#
#
†
†
Fredrik Rahm, Jenny Viklund*, Lionel Trésaugues, Manuel Ellermann, Anja Giese, Ulrika
#
#
#
†
#‡
Ericsson, Rickard Forsblom, Tobias Ginman, Judith Günther, Kenth Hallberg, Johan
#
#
#
#
∥,$
Lindström, Lars Boukharta Persson, Camilla Silvander, Antoine Talagas, Laura DíazꢀSáez,
Oleg Fedorov, ∥,$ Kilian V. M. Huber,∥,$ Ioanna Panagakou,∥,$ Paulina Siejka,∥,$ Mátyás
†
†
#
Gorjánácz, Marcus Bauser, and Martin Andersson
#
Sprint Bioscience AB, Novum, 141 57 Huddinge, Sweden
Bayer AG, Muellerstrasse 178, 133 53 Berlin, Germany
†
∥Structural Genomics Consortium, Nuffield Department of Medicine, University of Oxford,
Oxford, United Kingdom, OX3 7DQ
$Target Discovery Institute, Nuffield Department of Medicine, University of Oxford, Oxford,
United Kingdom, OX3 7FZ
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 84
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
KEYWORDS: MTH1, fragmentꢀbased drug discovery, fragment growing, structureꢀbased drug
discovery, target validation, BAYꢀ707
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 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
ABSTRACT
Recent literature has both suggested and questioned MTH1 as a novel cancer target. BAYꢀ707
was just published as a target validation small molecule probe for assessing the effects of
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
pharmacological inhibition of MTH1 on tumor cell survival, both in vitro and in vivo. In this
report, we describe the medicinal chemistry program creating BAYꢀ707, where fragmentꢀbased
methods were used to develop a series of highly potent and selective MTH1 inhibitors. Using
structureꢀbased drug design and rational medicinal chemistry approaches, the potency was
increased over 10.000 times from the fragment starting point, while maintaining high ligand
efficiency and drugꢀlike properties.
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 84
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
INTRODUCTION
One prominent feature of cancer is the increase in reactive oxygen species (ROS) in cancer cells
2
as compared to normal cells. ROS are generated during cell metabolism and are highly reactive
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
towards most biological macromolecules such as nucleic acids and proteins. Hence, cancer cells
have a higher dependency than normal cells on various ROSꢀrelated defense mechanisms, and
4
consequently, such pathways have been suggested as potential targets for cancer treatments.
The nucleotide pool is also subject to oxidative damage by ROS, with the main oxidation
products being 8ꢀoxoꢀ7,8ꢀdihydroꢀ2'ꢀdeoxyguanosine (8ꢀoxoꢀdG) and its corresponding
5
nucleotides. Oxidized nucleotide triphosphates (such as 8ꢀoxoꢀdGTP) are mutagenic and
6
cytotoxic when incorporated into DNA. To clear the nucleotide pool of such damaged
nucleotides, cells are equipped with enzymes, such as members of the NUDIX family
7
(NUcleoside DIphosphate linked to another moiety X). Recent studies have suggested a
dependency of cancer cells to a particular enzyme of this family; Nudix hydrolase 1 (NUDT1) ꢀ
8
also called MutT Homolog1 (MTH1), which is upregulated in several cancer forms. MTH1
catalyzes the hydrolysis of 8ꢀoxoꢀdGTP and scarcer oxidized nucleotides triphosphates to their
monophosphate counterparts, thus preventing their incorporation into DNA. The 8ꢀoxoꢀdGMP
9
product is further hydrolyzed to 8ꢀoxoꢀdG and excreted from the cell.
The first discoveries relating MTH1 to a role in cancer biology was published in 2011, in a study
1
0
that identified MTH1 as one of the genes induced by RAS overexpression. Given its role in
nucleotide pool sanitization and ROS stress defense, further studies by the same authors showed
11
that MTH1 knockdown inhibits RASꢀdriven tumor cell growth in vitro and in vivo. More data
was published during 2014 lending further support to the hypothesis that MTH1 has a role in
ACS Paragon Plus Environment
4

Page 5 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
prevention of DNA damage and cancer cell survival. It was shown that a phenotypic screening
12
hit, SCH51344 that inhibits the nonꢀadherent growth of cancer cells, is an MTH1 inhibitor. (S)ꢀ
Crizotinib was also identified in this screen as a more selective MTH1 inhibitor, and it was
shown that these compounds induce DNA damage and inhibit tumor cell growth both in vitro
and in vivo. In a third study, pharmacological target validation also looked promising, both 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
8
vitro and in vivo, using the novel MTH1 inhibitor TH588.
1
3,14
The first crystal structure of MTH1 became available in the PDB in 2011.
The global interest
in this potential cancer target is revealed by the large number of structures of complexes which
have been released since then. More than 20 complex structures with either nucleotide analogues
or small molecule inhibitors have been published: the first one being with 8ꢀoxoꢀGMP, the
15,16
product formed by hydrolysis of the main substrate, 8ꢀoxoꢀdGTP.
These ligands are all
8
,12,17
competitive binders,
and they share common binding motifs with the purine ring of the
substrate, including one or more hydrogenꢀbonds to the “key Asp119ꢀAsp120 recognition
1
6
element”. Even though the ligand interactions to these aspartates appear to be crucial, the
number and positions of the atoms involved in hydrogenꢀbonding can vary significantly (Figure
1
).
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 84
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. Representations of the key interactions of selected inhibitors of MTH1 (the
corresponding PDB ID is shown in parenthesis), with varying hydrogen bonding patterns to the
18
Asp119ꢀAsp120 recognition element (dotted lines): 8ꢀoxoꢀGMP (3ZR0) , (S)ꢀCrizotinib
1
2
8
17
(
4C9X) , TH588 (4N1U) , AZ Compound 19 (5ANV) .
1
7,19
In 2016, reports started to appear that questioned the role of MTH1 in cancer cell survival.
In
one study, proteomic profiling was used to conclude that the proteomeꢀwide fingerprints of cells
exposed to some of the pharmacological tools described in the 2014 studies did not resemble the
proteomic fingerprints of other DNAꢀdamage inducing agents or other MTH1 inhibitors,
2
0
indicating that the observed effects of these compounds might be due to offꢀtarget effects.
These opposing results compared to the previous reports from 2014 emphasize the importance of
2
1–24
using high quality, orthogonal probes for target validation.
Since none of the above publications from 2016 reported results from in vivo studies, the
hypothesis that MTH1 inhibition affects tumor growth in vivo could not be ruled out. To test this
suggestion, a novel MTH1 inhibitor (BAYꢀ707) was recently used to evaluate the effects of
pharmacological target inhibition in a broad array of model systems, both in vitro and in vivo,
ACS Paragon Plus Environment
6

Page 7 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
confirming the deꢀvalidation results from 2016 . This report describes the medicinal chemistry
program that created BAYꢀ707, using fragmentꢀbased drug discovery methods.
CHEMISTRY
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
Compounds described in this report are based on 7ꢀazaindole, 7ꢀazaindazole or 7ꢀ
azabenzimidazole core structures, substituted in the 4ꢀposition and, in the first, optionally in the
2
ꢀposition. The compounds were made by following a few general synthetic routes, as illustrated
by Method A and Method B outlined in Scheme 1 below. Depending on the nature of the
25
substituent in the 4ꢀposition, either the SuzukiꢀMiyaura crossꢀcoupling reaction, the Buchwaldꢀ
2
6,27
Hartwig
crossꢀcoupling reaction or direct nucleophilic aromatic substitution was used to
introduce this substituent. To enable further functionalization in the 2ꢀposition of the core
structure, suitable handles were introduced by standard deprotonation procedures followed by
reaction with an appropriate electrophile. Some compounds were synthesized by combining one
of these routes with additional standard functional group transformations. Full details regarding
the synthesis of all compounds can be found in the experimental section.
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 84
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
General method Aa
1. (ii) or (iii)
(i)
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
III
2
. (iv)
I
II
(v)
IV
General method Ba
(vi)
(vii)
VI
I
V
(iii)
VII
a
Scheme 1. Reagents and conditions: (i) LDA, EtOCOCl, 2ꢀMeTHF, ꢀ78 °C to rt; (ii) R1=aryl:
1
2 3
boronic acid reagent, Pd(amphos)Cl , K CO , MeCN:H O, 120 °C; (iii) R =ꢀNR R : cyclic
2
2
3
2
.
secondary amine, NMP, 160 °C; (iv) LiOH H O, THF:MeOH:H O, 50 °C; (v) (a) SOCl , 80 °C,
2
2
2
4
5
(
b) R R NH, THF, 0 °C to rt; (vi) LDA, 2ꢀMeTHF, CO , ꢀ78 °C to rt; (vii) (a) SOCl , 80 °C, (b)
2
2
EtNH , THF, 0 °C to rt.
2
ACS Paragon Plus Environment
8

Page 9 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
RESULTS AND DISCUSSION
Primary screening was performed using a fluorescence‐based thermal shift assay (TSA) to screen
a library of 723 fragments at a compound concentration of 1 mM. The TSA monitors ligand
effects on temperature‐dependent protein unfolding, where ligand binding would lead to a
positive T shift, as compared to the T of the apoꢀprotein control. Hits were determined based
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
m
m
on a change in T of MTH1 equal to or greater than three standard deviations of the mean of the
m
T for the reference. Consequently, fragments with a ꢁT value of more than +1.0°C in both
m
m
duplicates were defined as hits. Applying this threshold, 166 fragments were identified that
significantly increased the thermal melting point of MTH1, with ꢁT values ranging from 1.0 to
m
2
8
1
1.8°C. Such high hitꢀrate (23%) is usually indicative of a druggable active site.
Doseꢀresponse experiments were performed to verify the stabilization of MTH1 for the most
promising hits from the singleꢀpoint screening. The fragment concentrations were varied from
6
2.5 to 1000 µM and it was determined that the shift in ꢁT was doseꢀdependent for 48 of 49
m
tested initial hit fragments. One of the screening hits was a 7ꢀazaindole compound, fragment hit
, with a single point ꢁT of 3.9 °C at 1 mM (Figure 2). 1:1 binding was verified with ITC, with
1
m
a K of 49.5 ꢀΜ (Figure S1). The azaindole was neither the most potent nor the most ligand
d
efficient fragment among the verified hits. Still, it was selected to be explored considering its
tractability from several other important perspectives: the azaindole core was providing
synthetically reasonable attachment points to grow the fragment into interesting parts of the
protein. Also, we were intentionally avoiding mainly lipophilic fragments containing a basic
2
9
amine. And last, all fragments containing the same aspartateꢀbinding aminoꢀpyrimidine and
aminoꢀpyridine cores as the reference compounds TH588 and SꢀCrizotinib were deselected, to
ensure the creation of a novel and orthogonal series (Figure 1).
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 84
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 fragment hit 1 was coꢀcrystallized with human MTH1 protein and the 7ꢀazaindole core was
shown to make hydrogen bonds to both residues in the key recognition element mentioned
above: the acceptor is interacting with Asp119 and the donor is interacting with Asp120 (Figure
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
a). The 7ꢀazaindole core is sandwiched through piꢀstacking interactions between aromatic
moieties of both Phe72 and Trp117 (Figure. 2b and 2c). The ester motif of 1 occupies the same
area as the ribose part of the product, indicating the 4ꢀposition of the 7ꢀazaindole ring system as
an interesting starting point for growing the fragment hit into a larger molecule, aiming to
increase binding affinity. Analysis of the crystal structure of 1 in complex with MTH1 revealed
the existence of a minor pocket nearby Asp120, which we hypothesized could be targeted from
the 2ꢀposition of the 7ꢀazaindole core structure (pocket 1, Figure 2b).
ACS Paragon Plus Environment
10

Page 11 of 84
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
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 84
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
Figure 2. (a) The 2D structure of fragment hit 1. Vectors from the 2ꢀ and 4ꢀpositions illustrate
the attachment points for synthetic expansions presented in this work. (b) and (c) Overlay of
complex between MTH1 and 1 (green, PDB ID: 6F20) and complex between MTH1 and 8ꢀoxoꢀ
dGMP (cyan, PDB ID: 3ZR0). Sideꢀchains of Phe72, Trp117, Asp119 and Asp120 in the
structure of MTH1ꢀ1 complex are shown as sticks with their carbon atoms colored green. Water
molecules present in the inhibitor binding site of MTH1ꢀ1 complex are shown as red spheres. 8ꢀ
oxoꢀdGMP is displayed as thin sticks, with its carbon atoms colored cyan. The ester motif
occupies the same area as the ribose of 8ꢀoxoꢀdGMP, and we refer to this area as the ribose
pocket. Clipped molecular surface of MTH1 is shown in the MTH1ꢀCompound 1 complex
structure. Figure 2 (c) is rotated 90 degrees as compared to figure 2 (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
With the crystal structure of 1 in hand, our first approach for improving the fragment was to
investigate close analogues in our inꢀhouse compound collection. Based on the crystal structure,
docking studies were performed, and two compounds (2 and 3, Table 2) both with a substituent
in the 4ꢀposition of the 7ꢀazaindole core, were predicted to maintain the binding mode of 1. TSA
data showed an increase in melting temperature for both compounds, suggesting an increased
binding affinity as compared to compound 1 (Table 1). Encouraged by these data, compound 4,
with a more lipophilic group in the 4ꢀposition, was synthesized and tested together with
compounds 1, 2 and 3 in a biochemical MTH1 inhibition assay. All compounds showed an
increased potency as compared to 1 with at most a 5ꢀfold increase in potency for 4, while
3
0
maintaining a high ligand efficiency (LE) (Table 1).
Next, we set out to explore whether pocket 1 (Figure 2b) could fit a substituent in the 2ꢀposition
of the 7ꢀazaindole core. By visual inspection of the pocket, we reasoned that compound 5 was
ACS Paragon Plus Environment
12

Page 13 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the only representative from our inꢀhouse collection with a substituent (amide) that was small
enough to have a chance to fit into this minor pocket. It was encouraging to see that the amide in
the 2ꢀposition increased biochemical potency 20 times as compared to the matched pair
compound 3. In addition, compound 5 maintained a high LE, high solubility and even lower
lipophilicity (Table 1), indicating that exploring the 2ꢀposition would be an interesting option for
optimization.
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
13

Journal of Medicinal Chemistry
Page 14 of 84
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 1. Biophysical, biochemical and physicochemical data of fragment hit 1 and close
analogues
TSA
Potency
Physicochemical properties
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
in buffer
Compo
und
ꢁTm
(°C)
Biochemical
IC50 (ꢀM)
MW
(Da)
clogP
[b]
R1=
R2=
H
LE
(
mM)[c]
1
2
3.9
7.9
9.9
0.49
190.2
234.3
1.8
2.2
2
H
2.5
0.42
0.5
3
4
H
H
8.2
3.7
1.9
0.49
0.49
195.2
228.7
1.6
3.5
2
n.d. [a]
0.25
5
14.4
0.2
0.48
252.3
0.3
1
a) n.d. not determined
b) cLogP was calculated using the ChemAxon Partitioning Plugin v16.10.24
c) Solubility was estimated by visual inspection after dilution of the DMSO stock solution in buffer.
The compounds were diluted at least 1/100, to give a final DMSO concentration not higher than 1
%.
ACS Paragon Plus Environment
14

Page 15 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Encouraged by the large potency gain from the amide of compound 5, we synthesized a small set
of five additional amideꢀcontaining compounds to further investigate the 2ꢀposition of the 7ꢀ
azaindole core (6-10, Table 2). To generate matchedꢀpair compounds with a balanced
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
1
lipophilicity , mꢀchlorophenyl (as used in 4) was chosen as substituent in the 4ꢀposition.
The crystal structure of the ethyl amide analogue 7 was solved in complex with MTH1, to
understand the potency increase owing to the amide substituent in the 2ꢀposition. While the 7ꢀ
azaindole core remains in the same position as seen for the fragment hit 1, the amide makes two
additional hydrogen bonds to the protein; the donor is interacting with Asp120, and the carbonyl
acceptor is interacting with the backbone nitrogen of Gly34. In contrast to the crystal structure of
1, Asn33 has moved into a different conformation, and is now replaced by the amide carbonyl
(Figure 3). A similar interaction pattern, with a reversed amide substituent, was later published
1
7
by others (Figure 1, AZ Compound 19).
Figure 3. Crystal structure of 7, showing that the 7ꢀazaindole core remains in the same position
as the fragment hit 1. MTH1ꢀ1 and MTH1ꢀ7 complexes are colored green and violet
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 84
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
respectively. Residues interacting with the compounds, and compound themselves, are shown as
thin (complex with 1) or thick (complex with 7) sticks with their carbon atoms colored following
the same code as the secondary structure elements. Water molecules present in the bindingꢀsite
of MTH1ꢀ7 complex are shown as red spheres. Hydrogen bonds are displayed as orange dashes.
Clipped molecular surface of MTH1 is shown in the MTH1ꢀ7 complex structure (PDB ID:
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
6
F1X).
From a computational chemistry perspective, it is interesting to note that neither the binding
mode of 7, nor any of its analogues, could be predicted correctly via docking studies, using any
of the crystal structures available at that time (4C9X, 4C9W, 3ZR0, 3ZR1, 3Q93, 4N1T, 4N1U,
Compound 1). In all but one of these structures, the carbonyl of the flexible residue Asn33 is
occupying the same position as the carbonyl of the amide of 7. The only exception is 4CX9,
where Asn33 occupies a similar position as in the MTH1ꢀ7 complex. However, even when
modifying the protonation states of the 4C9X aspartates to accommodate the hydrogen bonding
pattern of 7, the binding mode of the amide could still not be predicted correctly. We believe this
is an interesting example of how computational selection methods of new compounds should be
complemented by visual inspection of the crystal structures, to take advantage of the intuition of
experienced medicinal chemists when doing structureꢀbased design. In retrospect, however, we
did try induced fit docking, allowing for the active site residues to be flexible, and with this more
timeꢀconsuming protocol, the correct amide binding mode could indeed be predicted correctly.
In combination with the crystal structure of 7, the small set of five amideꢀcontaining compounds
generated distinct knowledge for the structure activity relationships (SAR) in 2ꢀposition: the
importance of the hydrogen bond interaction with Asp120 is clearly seen in the severe drop in
potency for the tertiary amide of 8, lacking the donor. The equally severe drop in potency seen
ACS Paragon Plus Environment
16

Page 17 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
for the primary amide of 9, shows the importance of filling the remaining part of pocket 1 with a
small alkyl substituent, which is then surrounded by the hydrophobic residues Leu11, Leu20,
Met116 and Phe124. Thus, only secondary amides (6, 7, and 10) were well tolerated and resulted
in compounds with low nM potencies and maintained LE (Table 2).
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
To further explore pocket 1, and particularly to investigate whether more diverse types of
substituents other than amides were allowed in the 2ꢀposition, four additional matchedꢀpaired
compounds were synthesized (Table 2). Attempts to gain potency by either simply filling the
pocket with a methyl group (11) or by attempting to make polar interactions with the protein, as
predicted by docking studies, were unsuccessful. Both compounds 12 and 13 contain other types
of acceptors, aiming to mimic the amide carbonyl hydrogen bond to the backbone of Gly34.
However, both the nitrile and the oxazole rather reduced potency, as compared to the matched
pair compound 4, without a substituent in the 2ꢀposition. Compound 14, with a methylene
hydroxyl substituent, was predicted by the docking studies to be able to either interact either as
an acceptor to Gly34, or as a donor to Asp120. However, despite both of these hydrogenꢀbonding
possibilities, 14 was merely equally potent as the matched pair compound 4. To summarize to
optimization efforts in the 2ꢀposition, the secondary amide analogues 6, 7 and 10 were
considered the most promising compounds so far, with low nM biochemical potency and high
solubility.
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 84
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. Biochemical potency and physicochemical properties for variations in 2ꢀposition.
Potency
Physicochemical properties
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
Biochemical IC50
Solubility in
buffer (mM)
Compound
6
R2
LE
clogP
2.1
MW
(
µM)
0.012
0.020
2.7
0.54
285.7
0.25
7
8
9
0.5
2.5
2.4
299.8
299.8
0.25
0.25
0.36
5.8
0.41
0.49
1.9
2.5
271.7
311.8
0.1
0.1
10
11
0.010
> 10
5.6
3.5
3.1
2.6
242.7
253.7
295.7
0.05
0.25
0.05
1
1
2
3
0.42
0.44
>10
1
4
1.4
2.4
258.7
0.01
ACS Paragon Plus Environment
18

Page 19 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Since our main objective was the development of both potent and highly selective MTH1
inhibitors, we wanted to assess the kinome selectivity profile already at this early stage of the
program. We were especially concerned by the presence of the 7ꢀazaindole core in our series,
32
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
since this substructure is commonly seen as a hinge binding motif in kinase inhibitors.
Consequently, we assessed the kinome selectivity profile of 10, as a representative structure of
the azaindole series. Aiming for a selectivity window of at least 1000 times, compound 10 was
tested in single point at 1 ꢀM in a 97ꢀmember kinase panel [DiscoverX’s scanEDGE™ Kinase
Assay Panel]. It was encouraging to see that the compound did not hit any of the included
kinases at significant levels [Sꢀscore = 0, Supporting information].
After the 2ꢀposition had been explored, we decided to continue the optimization work using the
substituent in the 4ꢀposition, aiming to explore the ribose pocket further (Figure 2). A sitemap
analysis using the crystal structure of MTH1ꢀ7 proposed that this area would favor quite bulky,
lipophilic substituents, possibly in combination with a donor to interact with any of the backbone
carbonyls of Thr8, Phe27 or Gly34 (Figure S2, Supporting information). Consequently, we
aimed to introduce larger lipophilic substituents with high 3D character, still retaining rigidity, in
hope of improving potency by filling the ribose pocket more optimally. In addition, increasing
the fraction of sp3 carbons is a general strategy to improve vital drug properties, such as
33,34
metabolic stability, solubility and selectivity.
A new set of compounds (15 ꢀ 19) with cyclic amines in the 4ꢀposition were selected for
synthesis based on docking studies and availability of starting materials. We were intentionally
3
5
avoiding additional donors to maintain high permeability. Compound 16, with a 2ꢀ
phenylpyrrolidine in the 4ꢀposition, resulted in the highest increase of potency and LE (Table 3).
The structure of the complex between 16 and MTH1 showed a maintained binding mode of the
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 84
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
7
ꢀazaindole core, and the bulky phenyl tightly occupying the pocket between the lipophilic
residues Tyr7, Leu9 and Met81 (Figure 4). The most potent compounds 15 and 16, together with
the previous representative compound 10, were assessed for primary ADME properties. In
general, solubility was high, Cacoꢀ2 permeability promising whereas mouse microsomal stability
was low (Table 3).
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
20

Page 21 of 84
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 4. Overlay between the structure of MTH1ꢀ16 (white) and MTH1ꢀ1 (green). Compound
16 and compound 1 are displayed as thick and thin sticks respectively, with their carbon atoms
colored following the same code as the secondary structure elements. The sidechains of MTH1
residues forming the cavity surrounding the phenyl moiety of 16 are displayed as sticks. Clipped
molecular surface of MTH1 is shown in the MTH1ꢀ16 complex structure (PDB ID: 6F23)
To determine whether a parallel SAR with respect to potency could be established, a few
pyrrolidineꢀbased substituents in the 4ꢀposition were combined with a small number of
secondary amide substituents in the 2ꢀposition. Indeed, all combinations generated compounds
(20-25) that were highly potent, some of them even below the assay wall of 0.1 nM (Table 3).
When assessing Cacoꢀ2 permeability and metabolic stability, the most lipophilic representative
(20) was clearly the most permeable but also the least metabolically stable analogue. Comparing
to the more hydrophilic examples (21-24), it seemed reasonable to believe that microsomal
stability was favored by decreasing lipophilicity, which in turn seemed unfavorable for
36
permeability. This type of imbalance is commonly seen for small molecule drugs. Still, we
considered the compounds 24 and 25 as the most promising representatives in this series so far,
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 84
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
being highly potent, fairly stable in microsomes and with a medium Cacoꢀ2 permeability (Table
3
).
Considering the very high potency of our compounds, we decided to deliberately discard the
affinityꢀincreasing amideꢀsubstituent in the 2ꢀposition, aiming to investigate whether the
permeability and/or microsomal stability could be improved at the expense of potency. In the
design process, we were also bearing in mind that increasing the fraction sp3ꢀcarbons and
decreasing molecular weight and the number of aromatic rings may be favorable to improve drug
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
1,36
properties.
We hypothesized that exchanging the carbon in the 2ꢀposition of the 7ꢀazaindole
core for a nitrogen would lead to a small decrease in lipophilicity, which might be beneficial for
3
7
increasing metabolic stability. More importantly, this extra nitrogen would remove the CꢀC
double bond of this core, which we hypothesized could be significant for improving metabolic
38
stability. Thus, a set of the most promising 4ꢀposition substituents of earlier compounds and
some additional new variations, were transferred to the 7ꢀazaindazole core, creating compounds
26-30 (Table 3). Three of these showed biochemical potencies in the low nanomolar range,
which is equivalent to the original 7ꢀazaindole compounds, but as expected, neither of them were
more potent than the corresponding 7ꢀazaindole amide analogues. Permeability was very high for
all of these small compounds (MW < 350), with no signs of efflux (Table 3). However, the
metabolic stability was still low, and in fact all 7ꢀazaindazole compounds were less stable than
the 7ꢀazaindole amide analogues. Also, the matched paired compounds 16 and 27, with the 2ꢀ
phenylpyrollidine substituent combined with the 7ꢀazaindole and 7ꢀazaindazole cores
respectively, were equipotent and showed similar permeability, stability and solubility profiles.
The crystal structure of 29 showed a structural water within hydrogenꢀbonding distance from the
newly introduced 2ꢀposition nitrogen of the 7ꢀazaindazole core. Interestingly, this water is also at
ACS Paragon Plus Environment
22

Page 23 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
hydrogenꢀbonding distance from both interaction points (Gly34 nitrogen and Asp120) of the
earlier compounds with an amide substituent in the 2ꢀposition (Figure 5). While making
3
9,40
hydrogenꢀbonding interactions with a structural water in some cases may increase potency
, 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
our case, all matchedꢀpair compounds (9 and 26, 20 and 27, 22 and 28) clearly showed that the
amide interactions were more important for potency than the corresponding interactions made by
the structural water.
Figure 5. Overlay between the complex structure of MTH1ꢀ29 (yellow) and MTH1ꢀ1 (green).
Compound 29 and 1 are displayed as thick and thin sticks, respectively with their carbon atoms
colored following the same code as the secondary structure elements. Sidechain of MTH1
residues interacting with the compounds are shown as thin (1) and thick (29) sticks, respectively.
Water molecules present in the binding site of 29 are depicted as red spheres. Hydrogen bond
between MTH1, water molecules and 29 are displayed as orange dashes. Clipped molecular
surface of MTH1 is shown in the MTH1ꢀ29 complex structure (PDB ID: 6F22).
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 84
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
To assure a maintained high selectivity profile for the most promising representatives of our
chemical series, compound 24 was tested in the same 97ꢀmembered kinase panel as compound
10. Again, excellent selectivity was achieved [Sꢀscore = 0, Supporting information]. In addition,
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
4 was also tested in a 215ꢀmembered ATPꢀase panel, including the structurally related MTH1ꢀ
analogues NUDT2, NUDT3 and NUDT7. In summary, 24 was more than 1000 times selective
against all additional antiꢀtargets at 1 ꢀΜ [Supporting information].
Cellular onꢀtarget engagement was another important property we wanted to verify for our
MTH1ꢀinhibitors. The most promising amideꢀcontaining compounds 20, 24 and 25 were selected
for evaluation, together with two less potent compounds 16 and 28. These compounds were
selected both as representatives for the different subꢀseries and to cover a biochemical potency
8
range between 50 and 0.1 nM. TH588 was also included for comparisons. To produce a reliable
tool compound, we were aiming at creating compounds with a target engagement EC value that
50
was more potent than the reference compound TH588. The compounds were evaluated in K562
4
1
cells in doseꢀresponse, using the cellular thermal shift assay
(CETSA). There was a strong
correlation between biochemical potency and CETSA EC . The three most potent compounds
50
(
20, 24 and 25) had significantly higher CETSA EC values than the reference, and compounds
50
20 and 24 were 100 times more potent in both the biochemical and CETSA assays (Figure 6).
8,17
Similar EC values of TH588 have been reported previously.
50
ACS Paragon Plus Environment
24

Page 25 of 84
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 6. There was a strong correlation between biochemical potency and target engagement in
cells, using CETSA. Dotted line is showing y=xꢀ1.5, indicating a dropꢀoff from the biochemical
assay to the cellular assay of about 30 times. Three compounds (20, 24 and 25) had significantly
higher pEC in CETSA than the reference TH588.
5
0
During the design process, a computational model was iteratively developed for predicting the
potency of compounds in the 7ꢀazaindole and 7ꢀazaindazole series. The docking protocols
usually predicted the correct binding mode, as assessed by crystal structures. Initially,
substituents in the 4ꢀposition were prioritized for synthesis based only upon predicted binding
mode of the final compound, primarily selecting the ones that allowed the core to maintain the
same interaction mode to MTH1 as compound 1. However, the docking score could not predict
4
2
potency. Considering the importance of the hydrogenꢀbond between the pyridineꢀtype nitrogen
N7 of the 7ꢀazaindole ring and the Asp119 carboxylate, we combined the docking score and
predicted pKa of N7 into a small partialꢀleastꢀsquares (PLS) model. This model was later fineꢀ
tuned by including a simple heavy atom count descriptor. During the optimization work, this
model was continuously improved and used for selecting substituents in the 4ꢀposition., and the
last iteration resulted in a predictive PLS model (R2 0.87), covering a potency range of more
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 84
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
than five logꢀunits [Figure S3 and Table S2, Supporting information]. In a recent study, another
4
3
dockingꢀbased model was used to virtually optimize fragments for MTH1ꢀbinding.
Inspired by the importance of the pKa of N7 in the core structure, the high potency of the 7ꢀ
azaindazole core, and considering that a large portion of a ligand’s affinity to a protein usually
comes from electronic interactions with key residues, the core itself was also subjected to further
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
4
4,45
investigation.
Two matched pair variations of the azaindole and azaindazole cores,
introducing an extra nitrogen in 3ꢀ and 5ꢀposition respectively, were synthesized. They were
designed to investigate how these electronic variations would affect the hydrogenꢀbonding
strength to the aspartates. None of them resulted in an improved potency (Table S3, Supporting
information). To examine whether other nitrogen variations would be more favorable for affinity,
another type of predictive potency model was developed. We were inspired by a previously
published model, using density functional theory (DFT) calculations, where other types of core
4
6
structures are interacting with two key aspartates of a protease. Our model used the same
methodology, and was used to predict the potency of three additional, virtual variations of the
azaindole. However, since all three virtual compounds were suggested to be less potent than the
original azaindole and azaindazole cores, we decided not to synthesize any additional variations
(Figure S4, Supporting information).
To summarize the optimizationꢀwork so far, we decided to keep both the original 7ꢀazaindole
core and the secondary amide in the 2ꢀposition, since compounds such as 24 were combining the
highest potencies in both the biochemical and CETSA assays with high kinome and ATPase
selectivity, as well as showing the most promising permeability and stability profiles. In the next
optimization cycle, we were aiming to fineꢀtune the ADME properties, with respect to Cacoꢀ2
permeability and metabolic stability. Using a combination of the collected SAR knowledge from
ACS Paragon Plus Environment
26

Page 27 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
all previous compounds, the PLS potency model and the above mentioned medicinal chemistry
rules of thumb, four new compounds (31ꢀ34) were designed with both new and previously used
substituents in the 4ꢀposition. In these compounds, a combination of picomolar potency, high
solubility, high permeability and acceptable metabolic stability was finally realized (Table 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
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 84
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 3 Biochemical potency, physicochemical properties, metabolic stability and Cacoꢀ2
permeability data for variations in 2ꢀ and 4ꢀposition substituents on the azaindole and
azaindazole cores.
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
Mouse liver
microsomes
Physicochemical
properties
Potency
Caco-2 permeability
%
Metabo
lised
after 30
min
Papp A
Papp B
to A
(10
Solubil
Biochem
ical IC50
(µM)
mClint
(µL/mi
n/mg)
Effl
ux
ratio
to B
ity in
buffer
(mM)
ID
R1
A
R2
LE
ꢀ6
ꢀ6
clogP
MW
(10
cm/s)
cm/s)
1
0
C
C
C
10
0.49
0.53
0.57
115
high
244
83
100
97
18
18
1
2.5
2.3
3.7
311.8
251.3
263.3
0.1
1
1
5
6
H
H
0.04
0.005
39
69
35
46
0.9
0.7
0.25
0.25
17
C
C
H
0.2
0.7
0.48
0.4
3.3
3.0
249.3
279.3
0.01
0.25
1
8
H
H
19
20
C
C
0.3
0.5
2.1
2.7
236.3
334.4
0.25
0.1
<0.0001
0.56
130
49
85
52
41
4
63
31
1.6
8.3
2
1
C
<0.0001
0.55
1.4
335.4
0.5
ACS Paragon Plus Environment
28

Page 29 of 84
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2
2
C
0.0001
0.5
24
29
1
1
0.7
0.8
363.4
0.25
2
2
2
3
4
5
C
C
C
<0.0001
<0.0001
<0.0001
0.56
0.54
0.51
48
44
25
53
50
33
15
13
7
72
76
46
4.7
5.7
6.7
1.7
1.7
1.1
335.4
347.4
363.4
0.5
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
0.25
0.25
2
6
N
3
0.47
2.8
229.7
0.1
2
2
7
8
N
N
0.001
0.03
0.63
0.51
222
45
99
54
74
44
63
50
0.8
1.1
3.0
1.7
264.3
265.3
0.25
1
2
3
9
0
N
N
C
0.002
0.005
0.56
0.59
0.59
232
153
54
97
92
56
88
75
73
66
0.8
0.9
1,3
2.4
1.6
1.3
280.3
260.3
330.4
1
1
1
31
*
<0.0001
170
225
32
*
C
C
C
0.0008
<0.0001
0.0001
0.59
0.54
0.59
36
156
45
43
90
48
148
63
209
89
1,4
1,4
3,3
0.6
2.0
1.0
288.3
350.4
314.4
0.5
0.25
0.5
33
*
34
*
24
79
ACS Paragon Plus Environment
29