Development of Potent and Selective Pyrazolopyrimidine IRAK4 Inhibitors

Article abstract of DOI:10.1021/acs.jmedchem.9b00439

A series of pyrazolopyrimidine inhibitors of IRAK4 were developed from a high-throughput screen (HTS). Modification of an HTS hit led to a series of bicyclic heterocycles with improved potency and kinase selectivity but lacking sufficient solubility to progress in vivo. Structure-based drug design, informed by cocrystal structures with the protein and small-molecule crystal structures, yielded a series of dihydrobenzofurans. This semisaturated bicycle provided superior druglike properties while maintaining excellent potency and selectivity. Improved physicochemical properties allowed for progression into in vivo experiments, where lead molecules exhibited low clearance and showed target-based inhibition of IRAK4 signaling in an inflammation-mediated PK/PD mouse model.

Full text of DOI:10.1021/acs.jmedchem.9b00439

Subscriber access provided by UNIV AUTONOMA DE COAHUILA UADEC
Article
Development of potent and selective pyrazolopyrimidine IRAK4 inhibitors
Marian C. Bryan, Joy Drobnick, Alberto Gobbi, Aleksandr Kolesnikov, Yongshen Chen, Naomi Rajapaksa,
Chudi O Ndubaku, Jianwen A. Feng, Willy Chang, Ross Francis, Christine Yu, Edna F. Choo, Kevin
DeMent, Yingqing Ran, Le An, Claire Emson, Zhiyu Huang, Swathi Sujathabhaskar, Hans Brightbill, Antonio
DiPasquale, Jonathan Maher, John Wai, Brent McKenzie, Patrick J. Lupardus, Ali Zarrin, and James R. Kiefer
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b00439 • Publication Date (Web): 13 May 2019
Downloaded from http://pubs.acs.org on May 13, 2019
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 67
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
Development of potent and selective pyrazolopyrimidine IRAK4 inhibitors
Marian C. Bryan,*,A Joy Drobnick, Alberto Gobbi, Aleksandr Kolesnikov, Yongsheng Chen, Naomi
A
A
A
B
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
A
A°
A∆
A#
A
A
Rajapaksa, Chudi Ndubaku, Jianwen Feng, Willy Chang, Ross Francis, Christine Yu, Edna F.
A
A≠
A
A
A□
A
Choo, Kevin DeMent, Yingqing Ran, Le An, Claire Emson, Zhiyu Huang, Swathi Sujatha-
A
A
A
A
B
A
Bhaskar, Hans Brightbill, Antonio DiPasquale, Jonathan Maher, John Wai, Brent S. McKenzie,
Patrick J. Lupardus,A● Ali A. Zarrin, James R. Kiefer
A
A
A
Genentech Inc., One DNA Way, South San Francisco, CA 94080
B
WuXi Apptec, 288 Fute Zhong Road, Waigaoqiao Free Trade Zone, Shanghai, 200131, P. R. China
KEYWORDS
Kinase, inhibitor, drug discovery, lupus, inflammation, interleukin-1 receptor activated kinase, IRAK,
structure based drug design
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 67
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: A series of pyrazolopyrimidine inhibitors of IRAK4 were developed from a high throughput
screen (HTS). Modification of an HTS hit led to a series of bicyclic heterocycles with improved potency
and kinase selectivity but lacking sufficient solubility to progress in vivo. Structure-based drug design,
informed by co-crystal structures with the protein and small molecule crystal structures yielded a series
of dihydrobenzofurans. This semi-saturated bicycle provided superior drug-like properties while
maintaining excellent potency and selectivity. Improved physicochemical properties allowed for
progression into in vivo experiments, where lead molecules exhibited low clearance and showed target-
based inhibition of IRAK4 signaling in an inflammation-mediated PK/PD mouse model.
0
1
2
3
4
5
6
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
INTRODUCTION
The innate immune system forms a first line of response to infection in mammals. This system is
required to sensitively distinguish between self and non-self antigens and immediately respond to
infections and injuries. Receptors, both on the cell surface and in endosomal compartments, are critical
sensors of innate immunity, recognizing pathogens and initiating signaling cascades to mount immune
1
responses. Two such examples are the IL-1 receptor (IL-1R) family and toll-like receptors (TLRs),
transmembrane pattern recognition receptors (PRRs) critical in the detection of pathogen-associated
molecular patterns (PAMPs) of foreign microbes. Together with the IL-1R family, TLRs enable cells to
recognize molecular structures present in pathogens and quickly respond to them by initiating and
amplifying the pro-inflammatory cytokine response.
While this system is critically important to the innate immune system, dysregulation of TLR signaling
in particular has been implicated in autoinflammatory diseases such as rheumatoid arthritis (RA) and
several others.2-3 In RA for example, multiple TLR family members are upregulated when compared to
healthy individuals. Stimulation of these receptors leads to production of the inflammatory cytokines such
ACS Paragon Plus Environment

Page 3 of 67
Journal of Medicinal Chemistry
1
2
3
4
as IL-6 and tissue-destructive matrix metaloproteins. TLR family members are also implicated in the
pathophysiology of systemic lupus erythematosus (SLE), another debilitating autoimmune disease. SLE
is characterized by chronic systemic inflammation of the joints and kidneys among other organs.
Ligand binding to either IL-1R or TLR family members leads to dimerization and recruitment of
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
adaptor molecules to a conserved motif on the cytoplasmic domain of the receptor called the Toll/IL-1R
5
(
TIR) domain. Binding then leads to recruitment of MyD88 and formation of the myddosome complex
which includes the interleukin-1 receptor activated kinase (IRAK) family.4-9 This family of kinases then
plays the crucial role of signal transduction propagating both the innate immune response.
Autophosphorylation of IRAK4, the first step in the pathway, leads to IRAK1 phosphorylation which in
turn leads to downstream signaling through multiple cascades, ultimately altering gene transcription and
initiation of the inflammatory response.
The IRAK kinase family is comprised of four members, IRAK1 and IRAK4 which perform
scaffolding functions in addition to being active kinases, and IRAK2 and IRAK3 (also called IRAKm),
9
which are catalytically inactive. Of the two family members with kinase activity, IRAK4 has been called
the “master IRAK” and is the most upstream kinase required for IL-1R and TLR signaling.8, 10
IRAK4’s role in the innate immune response has led to both interest in it as a target for therapeutic
intervention in autoinflammatory diseases as well as concerns of safety due to its role as a critical node.
While IRAK4-deficient patients show increased susceptibility to certain pyogenic infections, adults show
no increased risk de-risking IRAK4 as a therapeutic target to an extent and interest in targeting IRAK4
has grown. Several reports of small molecule kinase inhibitors targeting the IRAK4 kinase domain,
including those shown in Figure 1, have been recently published, including the clinical candidate PF-
0
6650833 (1), pyridine indazole 2, BMS-986126 (3) and aminopyridine 4.2, 9, 11 Lactam 1 was advanced
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 67
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
into Phase II in RA though results have not yet been disclosed (NCT02996500). This molecule and others
continue to undergo clinical evaluation.
0
1
2
3
4
5
6
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
O
O
F
F
NH2
H
N
N
N
N
N
NH
F
F
O
O
H
N
N
N
HO
N
N
O
Me
Me
NH
O
N
S
N
O
N
NH2
O
HN
O
N
N
NH
O
F
PF-06650833, 1
2
3
4
Figure 1. Structures of IRAK4 kinase inhibitors PF-06650833 (1), indazole 2, 1,6-naphthyridine 3 and
aminopyridine 4.
RESULTS AND DISCUSSION
From a high throughput screening campaign, we identified hits 5 and 6, differing only in piperidine
substitution, as starting points from which to design potent and selective IRAK4 inhibitors. Both of these
quinolines showed excellent biochemical potency against IRAK4 (Table 1) and reasonable LLE based on
cLogP.12-13 Both molecules possessed selectivity over IRAK1. While IRAK1 plays important scaffolding
roles in the cell, its kinase activity has been reported as unnecessary for homeostasis, and several studies
have highlighted the integral role of IRAK4 alone in Toll-like receptor (TLR)-mediated immune
responses.3, 14 As there is a high degree of sequence homology between the ATP binding sites of IRAK1
2
and IRAK4, we hypothesized that measured selectivity of greater than 200-fold against IRAK1 would
provide a useful initial kinase selectivity triage parameter. Evaluation of compounds 5 and 6 against a
broad kinase panel showed compound 5 inhibited ≥70% of the activity of 34 out of 236 kinases at 1 µM
inhibitor concentration, while compound 6 inhibited 22 out of 220 kinases tested.
ACS Paragon Plus Environment

Page 5 of 67
Journal of Medicinal Chemistry
1
2
3
Table 1. Profile of high throughput screening hits, quinolines 5 and 6.
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
N
N
N
N
O
O
NH
N
N
NH
N
N
N
N
NH2
OH
0
1
2
3
4
5
6
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
6
5
6
LLEcLogP13 / Calc’d pKaa
6.78 / 9.3
0.79 / 22
2 ± 0.71
6.54 / 7
3.1 / 1
b
LogD / K. Sol. (μM)
c
IRAK4 K (nM)
1.7 ± 0.73
(n=6)
i
(
n=183)
510 ± 110
n=172)
1 ± 2.5
n=133)
4.5 / 14 / 51
c
IRAK1 K (nM)
400 ± 110
(n=5)
i
(
4
c
NFkB Cell IC (nM)
35 ± 4.1 (n=2)
12 / 43 / 75
5
0
(
H/R/MLMd
a
Moka v.2.6.515 bKinetic Solubility: experimental procedures for kinetic solubility experiments can be
c
found in the Supporting Information; Data are the geometric mean of at least two independent
d
experiments ± standard deviation of the average for the number (n) of experiments conducted.; H/R/MLM
defined as predicted hepatic clearance in liver microsomes for human (H), rat (R) and mouse (M).
Reported in units of mL/min/kg.
Both compounds showed strong inhibition of NFkB reporter gene activity (41 and 35 nM, respectively,
Table 1) in a THP1-XBlue stable cell line. As a secondary readout, cell viability was assayed in this cell
line, and no inhibition of proliferation was seen at up to 20 µM concentrations for either compound. The
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 67
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
proliferation assay was also used to give a preliminary indication of kinase selectivity, as inhibition of
IRAK4 alone should not lead to target-related cytotoxicity due to the previously mentioned safety
rationale. While both compounds were potent in the biochemical and cellular assays, the scaffold
presented challenges that needed to be addressed, including kinase selectivity compatible with chronic
dosing and in vitro ADME 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
In order to understand their binding mode to enable SAR development, we obtained a 2.5Å resolution
x-ray co-crystal structure of 5 with fully phosphorylated IRAK4 (Figure 2A, Figure S2 in the Supporting
Information). Compound 5 binds to the hinge through the backbone of Met265 and Val263 via the amide
carbonyl and the C2 hydrogen of the pyrazolopyrimidine, referred to as the “hinge binding element”
1
6
(Figure 2B). This was of interest to the team as the pyrazolopyrimidine moiety is known to bind to hinge
residues in an alternate configuration. In an example from the JAK kinase family, the N1-nitrogen of the
pyrazolopyrimidine core makes a hydrogen bond with a backbone NH and the amide carbonyl forms a
1
7
hydrogen bond with a water molecule. This binding configuration has also been reported for other
16
IRAK4 inhibitors. The leading edge of this fused 5,6 ring system forms van der Waals contacts and
potentially -stacking interactions with the kinase “gatekeeper” residue Tyr262, an amino acid relatively
unique to the IRAK family among protein tyrosine kinases.18
The quinoline system, referred to as the “linker” in Figure 2B, stacks under Met192 and forms multiple
additional van der Waals interactions to residues of the P-loop and floor of the kinase binding site (Figure
2C). Finally, the methylamine substitution on the piperidine, occupying the ATP ribose pocket, forms a
hydrogen bond with the backbone carbonyl of Ala315 and proximal solvent molecules.
ACS Paragon Plus Environment

Page 7 of 67
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 2. Co-crystal structures of IRAK4 initial lead inhibitors. The 2.5Å resolution co-crystal structure
of 5 (A, PDB ID 6O9D) bound in the IRAK4 active site reveals three polarized functional groups forming
hydrogen bonds to the protein. Further annotation (B) shows the sub-regions together with key molecule
features and their interactions. The solvent accessible surface of the binding site (C) shows substantial van
der Waals contacts made between nonpolar (green mesh) protein residues and the inhibitor as well as focal
polar (pink mesh) contact points. Regions above and below the plane of the quinoline ring (dotted circles)
are potentially accessible through elaboration of the inhibitor linker. A 2.0Å resolution co-crystal
structure of 17 (D, PDB ID 6O94) shows the inhibitor benzimidazole forms an additional polar interaction
with the protein.
Analysis of the protein-inhibitor contacts in the co-crystal structure of 5 and comparison of them
to known IRAK4 small molecule inhibitors (SMI) revealed additional potential interactions and vectors
towards the solvent front that could be exploited for optimization of the linker region of the compound.2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 67
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
As such, we synthesized a set of 6,6- and 5,6- ring systems (Table 2). Truncation of the ribose pocket
group to methyl ether 7 led to a 30-fold loss in activity and a 5-fold loss in selectivity when compared to
6 though allowed for rapid development of linker SAR. Replacement or movement of the quinoline N to
yield naphthalene 8 and isomeric quinoline 9, respectively, lost significant potency when compared to
parent 7. The dramatic erosion in potency of 9 was presumably due to a repulsion between the N lone pair
and a backbone carbonyl.
0
1
2
3
4
5
6
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
Substitution at R1 was then undertaken with morpholine 11. The morpholine moiety was selected
2
due to its presence in a number of known IRAK SMI. This compound lost potency compared to parent
1
0. In order to understand if this was due to steric interference, we generated the smaller dimethylamino
analog (12) and modified the angle by introduction of a second nitrogen to the linker to give quinazoline
3. While removal of some of the morpholine bulk was able to regain some of the potency (26 nM for 12
1
versus 8 nM for 10), quinazoline 13 was equipotent to quinoline 11 and none of these changes to the 6,6-
biaryl systems were able to improve LLEcLogP compared to the initial HTS hit 6.
Table 2. 6,6 versus 5,6 linkers.
ACS Paragon Plus Environment

Page 9 of 67
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
IRAK4 Ki
NFkB
Cmpd
#
IRAK1
(nM)a
R1
Core
R2
(nM)a
Cell IC50
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
LLEcLogP13)
(nM)a
(
4
00 ±
N
N
1.7 ± 0.73
35 ± 4.1
(n=2)
O
N
6
7
-
NH
N
110
OH
(n=6) (6.5)
N
R2
(n=5)
2
,800 ±
600
N
N
O
52 ± 15 (n=2)
(5.6)
370 ± 71
(n=4)
-
N
O
NH
N
R2
(n=2)
N
N
O
b
8
9
-
-
N
O
610 (3.6)
ND
ND
ND
NH
R2
N
N
O
b
>30,000 (ND)
ND
ND
N
NH
N
O
R2
N
N
O
N
8b (5.0)
1
0
-
N
ND
NH
N
R2
1
,300 ±
620
N
N
290 ± 130
4,700 ±
0 (n=3)
O
N
N
11
12
NH
N
O
(n=3) (3.1)
R1
N
R2
(n=4)
5
00 ±
N
N
O
26 ± 9.8 (n=6)
3.8)
98 ± 24
(n=4)
N
N
NH
R2
N
230
(
R1
N
(n=6)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 67
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
5
,400 ± 3,000 ±
N
N
170 ± 300
O
N
N
1
3
NH
N
860
2,000
(n=4)
O
N
(
n=4) (3.8)
R1
R2
N
(n=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
N
N
O
O
H
b
14
-
N
O
790 (4.4)
ND
ND
NH
N
R2
330 ±
150
N
N
6.7 ± 0.8 (n=2) >5,700
6.4) (n=2)
H
N
N
1
5
-
N
NH
OH
(
N
R2
(n=4)
N
N
O
O
O
1.7 ± 0.6 (n=5) >5,500 250 ± 33
N
N
N
N
H
1
1
6
7
N
NH
O
O
O
N
OH
OH
OH
R1
(7.2)
(n=5)
(n=3)
N
R2
N
N
1.8 ± 0.9 (n=9) >6,100
7.1) (n=9)
97 ± 97
(n=7)
N
N
NH
N
N
N
R1
(
R2
N
N
95 ± 22 (n=4) >10,000 910 ± 50
18
N
NH
N
N
R1
(4.9)
(n=4)
(n=2)
R2
340 ±
94
N
N
O
O
1.2 ± 0.5 (n=6)
6.8)
60 ± 19
(n=6)
N
N
N
1
2
9
0
N
NH
O
O
O
N
OH
R1
(
R2
(
n=4)
0.5 ± 0.2 (n=6) 24 ± 5.9 8 ± 2.1
6.9) (n=5) (n=3)
N
N
N
NH
N
S
N
OH
R1
(
R2
ACS Paragon Plus Environment

Page 11 of 67
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
N
N
O
O
0.5 ± 0.2 (n=5) 95 ± 36 53 ± 20
N
N
2
2
1
2
N
NH
S
N
OH
R1
(6.3)
(n=5)
(n=3)
R2
190 ±
12
N
N
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
0.8 ± 1.0 (n=5)
7.1)
76 ± 18
(n=5)
N
N
NH
N
O
S
N
O
R1
(
R2
(n=4)
aData are the geometric mean of at least two independent experiments ± standard deviation of the average
b
for the number (n) of experiments conducted. ND = not determined. Assay conditions and experimental
datafor these molecules were reported in WO2012007375A1.19
At this point we shifted focus to optimize the linker to 5,6-ring systems. Indole 14 was analogously
to naphthalene 8 highlighting the same trend towards a requirement for a critical H-bond acceptor in the
quinoline linker. Unsubstituted benzimidazole 15 was 4x less potent than the quinolone 6 in the
biochemical assay with similar LLEcLogP but continued to experience a larger cell shift (45x vs 20x).
1
Appending a morpholine R substituent resulted in compound 16 that was equipotent to 6 in a biochemical
context with improved LLEcLogP but continues to show a sizeable shift in cellular potency. Capping the
NH of the benzimidazole with a methyl group (17) led to a compound that was both potent and selective
over IRAK1. This kinase selectivity was recapitulated when the molecule was assessed against a broader
kinase panel as shown in Table 3, where the molecule inhibited to ≥70% the activity of only 13 out of 220
kinases when measured at 1 µM (Supporting Information Table S1). Introduction of an additional
methylene (18) eroded both of these potency and selectivity gains, presumably due to steric clashes with
the protein backbone. Alternate 5,6-linker systems such as benzoxazole 19 and benzothiazole 20 either
maintained or improved potency relative to initial molecule 6 with morpholine 20 being 3-4x more potent
in the assays. Trimming the morpholine to a dimethylamine (21) retained enzymatic potency but
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 67
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
negatively impacted the cellular potency (52 nM versus 8 nM). In addition, while selectivity over IRAK1
was improved with compound 21 relative to 20, evaluation in broader panels revealed dimethylamine 21
inhibited 23 kinases (in addition to IRAK4) in a panel of 41 to ≥70% activity at 1 µM, while morpholine
0
1
2
3
4
5
6
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
0 inhibited only 15 additional kinases to the same degree against a much broader panel of 220 at the same
concentration (Supporting Information Table S1).
Analysis of the 2.0Å resolution co-crystal structure of benzimidazole 17 (Figure 2D, Figure S2 in
the Supporting Information) shows that the compound maintains the same key hinge interactions of 5 and
projects its morpholine towards solvent, forming additional van der Waals interactions with the protein.
The bis-morpholine inhibitor 22 lost cellular potency compared to 20 but importantly improved kinase
selectivity, reducing the number of off-target kinases with activity inhibited ≥70% to 9 out of 221 at 1 µM
including IRAK4 (Supporting Information Table S1).
With comparably promising biochemical and cellular potency, as well as reasonable kinase
selectivity, molecules 17, 20, and 22 were examined in in vitro metabolic stability assays. These analogs
were predicted to be stable-to-moderate in mouse and human based on liver microsomes (Table 3: 7.4 to
8
.5 mL/min/kg in HLM and 33 to 46.4 mL/min/kg in MLM) but differentiated when advanced to
hepatocytes. Benzimidazole 17 and benzothiazole 20 both showed heightened lability in human
hepatocytes with predicted clearances of >70% of liver blood flow (LBF)(15.3 and 17.8 mL/min/kg,
respectively) though 17 appeared stable in mouse hepatocytes at <30% LBF (MH = 20.7 mL/min/kg).
Only compound 22 remained relatively stable in human hepatocytes and showed improved predicted
stability in mouse hepatocytes (7.5 mL/min/kg and 20.7 mL/min/kg, Table 3). Compound 22 also
exhibited higher permeability when compared to 17 and 20 (12.4 versus 8.4 and 2.4 cm/s x10-6,
respectively) though benzimidazole 17 was less plasma protein bound in both species (Table 3).
ACS Paragon Plus Environment

Page 13 of 67
Journal of Medicinal Chemistry
1
2
3
4
5
When dosed intravenously in mice, all three compounds (17, 20 and 22) were stable, with clearance
below 30% of liver blood flow. Upon oral dosing, benzothiazole 20 was not detected (F below the level
of detection), while 17 and 22 showed only marginal bioavailability (F = 3% and 5%). While morpholine
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
2
2 initially showed sub-micromolar concentrations at Cmax and low bioavailability, this was improved
upon formulation as a nanosuspension (at low dose). Given its improved in vitro stability profile relative
to 17 and 20, 22 was advanced into further studies. Unfortunately, compound 22 showed limited increase
in exposure (AUC and Cmax) in dose-escalating mouse PK study as a result of dose limited absorption due
to low solubility, despite a high in vitro permeability measurement and low observed clearance. Analysis
of the material properties of 22 along with other members of the 5,6-series revealed crystalline material
2
0
with high melting points (>200°C), resulting in poor solubility across a range of pH, and several crystal
packing interactions for the fused biaryl linker leading to extensive -stacking interactions in the crystal
lattice (see small molecule crystal structures in Figure S4 in the Supporting Information). Despite many
changes probing exploration of the solvent front and ribose pocket of this and related biaryl systems, we
were not able to see measurable improvements in solubility. Therefore, we next designed molecules
incorporating a reduction in linker aromaticity and planarity to drive solubility and bioavailability
improvements.
Table 3. In vitro and in vivo PK of 17, 20 and 22.
Mouse
Mouse PO PK
IV PK
H/MLMa MDCK
CLp /
Cmpd
#
PPB
Kinase
Dose
Cmax
AUC∞
F
H/M
Papp
Vss
(H / M) countd
(mg/kg) (μM) (μM*h) (%)
CLhepb
(A-B)c
(L/kg)e
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 67
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
.4 / 33
5 / 21
8.5 / 42
8 / 47
13 /
1
2
7
0
2.4
8.4
88 / 95
23 / 0.4
18 / 0.4
5f
0.2
0.5
3%
1
220
16 /
220
5f
BLODh
98 / 99
0
1
2
3
4
5
6
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
5f
0.2
1.3
2.0
3.3
0.6
12
11
40
5%
25%
12%
6%
7.6 / 46
.5 / 21
10g
30g
2
2
12
95 / 99 9 / 221 17 / 0.9
7
1
00g
aH/MLM defined as predicted hepatic clearance in liver microsomes for human (H) and mouse (M).
b
Reported in units of mL/min/kg. H/M CLhep defined as predicted hepatic clearance in hepatocytes for
c
human (H) and mouse (M). Reported in units of mL/min/kg. A-B, apical-to-basolateral permeability as
−
6
measured in Madin–Darby canine kidney (MDCK) cells. Reported in units of ×10 cm/s; PPB = plasma
d
protein binding, % bound. Ratio of number of kinases inhibited at ≥70% against the total number of
e
kinases assessed at 1 μM. Cl = blood clearance. Reported in units of mL/min/kg. Compound dosed IV
p
f
(
1 mg/kg) as a solution in DMSO:Cremophor:Saline (30:10:60). Compound dosed po as a suspension in
g
h
MCT; Compound dosed po as a nanosuspension in MCT. BLOD = Below the Limit of Detection
Similar to other kinases, most of the active site of IRAK4 is a relatively flat, narrow pocket;
however, the solvent front adjacent to the hinge region (Figure 2C) broadens and can accommodate three-
3
dimensional ligand topology; thus, we sought to introduce more sp character into this region of the
molecules in order to improve solubility.21-22 Docking studies with semi-saturated systems suggested that
they would complement the protein shape, and their enhanced 3-dimensionality could disrupt the crystal
packing observed in the small molecule x-ray structures, while maintaining the key vectors to the ribose
ACS Paragon Plus Environment

Page 15 of 67
Journal of Medicinal Chemistry
1
2
3
pocket and solvent front regions. Additionally, for dihydrobenzofuran linkers, the docking pose suggested
that the oxygen atom would occupy a comparable position to the conserved hydrogen bond acceptor
functionality observed in the previous scaffolds, such as 17 (Figure 2D). This feature could maintain a
water mediated hydrogen bond with the protein, previously observed to enhance potency (e.g. the pyridine
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
N of 7 compared with the CH of 8, K = 52 nM and 613 nM, respectively, Table 2). The design also
i
provided flexibility to introduce modifications to the hinge and ribose pocket groups.
The first molecule synthesized in this series validated the hypothesis by demonstrating that a
dihydrobenzofuran linker provided biochemical and cellular potency within an order of magnitude of
previous leads 20 and 22 though with lower efficiency (LLEcLogP = 5.7 and 5.6 vs 6.9 and 7.1,
respectively). Both 23 and 24 also maintained or improved in vitro permeability. Although 23 remained
poorly soluble (kinetic sol. = 1 µM), modification of the ribose pocket group in 24 improved kinetic
solubility (kinetic sol. = 95 µM) while maintaining potency and improving permeability. Furthermore,
these molecules maintained excellent selectivity over IRAK1 (220x and 210x, respectively).
In order to swiftly assess kinase selectivity without requiring a full screen of >200 kinases, we
profiled off-target kinases that either routinely showed significant inhibition with our core scaffold, were
also involved in the cytokine production pathway which would complicate on-target efficacy evaluation,
or whose inhibition would lead to known safety liabilities. This group formed a core panel of 27 kinases
that could be rapidly screened. A kinase selectivity ratio (KS27) was developed for rank-ordering
molecules on selectivity, with a lower value correlating to inhibition of fewer off-target kinases inhibited
at ≥70% at 1μM. Evaluation of 23 and 24 along with the previous lead molecules provided a range of
KS27 ratios with initial quinoline 6 showing a KS27 = 0.52 reflecting its inhibition of 14 kinases out of
2
7 and 22 out of 220 when screened at 1 μM. Improvements in kinase selectivity for benzothiazoles 20
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 67
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
and 22 also translated to an improved KS27 (KS27 = 0.44 and 0.26, respectively). Gratifyingly, 23 and
4 demonstrated improved kinase selectivity with a KS27’s of 0.15 and 0.22, respectively.
2
Table 4. Initial efforts toward reducing planarity with semi-saturated 5,6-ring systems and modification
0
1
2
3
4
5
6
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
of the pyrazolopyrimidine hinge binding element.
N
N
N
N
OH
N
N
N
N
O
O
O
O
R
N
N
N
N
NH
N
NH
N
NH
N
NH
N
O
O
O
O
O
OH
O
O
23
24
25
R=
R=
26
27
28
29
OH
30
31
32
CN
O
OCF2H
CF2H
NH2
OH
NFkB
Cell
IRAK4 Ki
nM)a
K.
MDCK
Cmpd
#
KS27
Ratioc
(
Sol.b
P (A- H/MLMe
app
IC50
(LLEcLogP
)
(μM)
1
B)d
nM)a
(
5
.3 ± 2.1
(n=4) (5.7)
.5 ± 0.5
170 ± 38
(n=4)
2
2
3
4
0.15
0.22
--
14
16
--
5.3 / 81
9.5 / 74
13 / 26
3.9 / 44
15 / 58
1
38 ± 6.7
(n=4)
95
77
3.1
1
(n=5) (5.6)
>8,400
>20,000
(n=2)
25
(
n=2)
.2 ± 0.4
(n=4) (6.3)
4 ± 11
n=4) (4.4)
3
330 ± 34
(n=4)
2
2
6
7
0.15
--
14
0.6
3
>20,000
(n=4)
(
ACS Paragon Plus Environment

Page 17 of 67
Journal of Medicinal Chemistry
1
2
3
4
5
1
,200 ±
3
1 ± 11
2
8
1,000
2.5
--
--
10 / 56
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
(
n=4) (5.4)
(n=4)
>20,000
(n=2)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
5 ± 9.0
2
3
9
0
1
1
--
--
7.4 / 54
14 / 47
12 / 49
3.9 / 57
(
n=7) (4.0)
6.1 ± 0.4
84 ± 10
(n=4)
0.19
0.19
0.074
2.7
0.7
18
(
n=4) (5.9)
1
.1 ± 0.4
(n=4) (7.5)
.5 ± 0.9
n=8) (6.4)
56 ± 9.7
(n=4)
31
1
6
66 ± 5.2
(n=4)
3
2
15
(
aData are the geometric mean of at least two independent experiments ± standard deviation of the average
b
for the number (n) of experiments conducted. ND = not determined; Kinetic Solubility: experimental
c
procedures for kinetic solubility experiments in the Supporting Information; KS27 Ratio defined in
d
Supporting Information. A-B, apical-to-basolateral permeability as measured in Madin–Darby canine
−6
e
kidney (MDCK) cells. Reported in units of ×10 cm/s. H/MLM defined as predicted hepatic clearance
in liver microsomes for human (H) and mouse (M). Reported in units of mL/min/kg.
To this point, no changes had been undertaken on the pyrazolopyrimidine. However, we were
interested in the ability of pyrazolopyrimidine substitution to impact selectivity given the relative rarity of
the tyrosine gatekeeper. Modification of the pyrazolopyrimidine of 23, selected for synthetic feasibility
over hydroxymethyl piperidine 24, was then undertaken (Table 4). The only example of substitution at
C7, hydroxypyridine 25, lost measurable potency while C6 substitution was well tolerated in the
biochemical assay with hydroxyl-substituted pyrimidine 26 being equipotent to the parent pyrimidine 23
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 67
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
with the lowered lipophilicity reflected in the improved LLEcLogP. Moieties with increased lipophilicity at
4
R led to 6-fold loss in biochemical potency and lower lipophilic efficiency as seen with 27-29 despite
differences in steric bulk or hydrogen bond donating / accepting capabilities. This is particularly
interesting for 26 when compared to 28. Both have the potential for hydrogen bond donation but 28 loses
an order of magnitude in biochemical potency. Modifications with comparable or improved lipophilicity,
such as nitrile 30 and primary amide 31, maintained biochemical potency with superior cellular potency.
However, these groups led to poor permeability with low solubility and no improvement on KS27 (Table
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
). The one modification that appeared universally beneficial was the hydroxymethyl substituent in 32.
This molecule maintained biochemical potency and permeability while improving solubility, in vitro
stability and kinase selectivity (K. Sol. = 15 μM, H/MLM = <3.9 / 57 mL/min/kg, KS27 = 0.074, Table
4
).
While only modest improvements were obtained through modifying the hinge binding element,
we were interested in the ability of the ribose pocket to allow for polar moieties given the polar surface
area shown in Figure 2C and the interaction between the primary amine of 5 and the backbone carbonyl
of Ala315 (Figure 2B). Related chemotypes such as those in Figure 1 or disclosed elsewhere either had
aryl groups in this region or very small moieties. 2, 9, 11, 16 We envisioned that the ribose group could be
used to tailor the overall physicochemical properties and indeed significant improvements in multiple
parameters including solubility and stability were seen through alternate groups in the ribose pocket as
was established for 23 vs 24 (Table 5). Several saturated ring systems were tolerated by the ribose pocket
of the IRAK4 ATP binding site. Contraction of the morpholine to a pyrrolidine (33), modification to a
piperazine (34), substitution or oxidation of the hydroxymethylene (35 and 36) and expansion to an
azepane (37) all led to analogs within 4-fold biochemical potency to each other (K = 3.2 – 13 nM). None
i
of these modifications improved either IRAK4 or the cellular potency compared to 24, although all 6-
ACS Paragon Plus Environment

Page 19 of 67
Journal of Medicinal Chemistry
1
2
3
membered ring modifications gave improved in vitro metabolic stability with piperazine 34 showing
modest gains in solubility as well (Table 5). Appending hydrogen-bond donors via N-linked substitution
of 34 gave 38–40, with biochemical and cellular potency comparable to 24 . This is presumably due to
forming similar hydrogen bonding interactions as discussed with 5 earlier. However, potency gains seen
with primary amide 39 appeared to be at the cost of permeability and selectivity (Papp = 2.6, KS27 = 0.3).
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
Table 5. Diversification of the ribose pocket moiety.
N
N
O
N
NH
O
R
IRAK4 Ki
NFkB
K.
Cmpd
#
KS27
MDCK
(nM)a
Cell IC50 Sol.b
H/MLMe
R
Ratioc Papp(A-B)d
(
LLEcLogP
)
(nM)a (μM)
12 ± 1.6
140 ± 29
7.1
N
33
--
--
--
--
7.4 / 83
(
n=4) (4.3)
3 ± 1.5
(n=8) (5.6)
0 ± 2.3
(n=4)
1
56 ± 3.6
110
(n=4)
N
3
4
<3.9 / 26
NH
1
56 ± 5.3
1
N
35
0.48
--
4.4 / 45
OH
(n=4) (4.0)
(n=4)
CF3
3.2 ± 0.2
53 ± 7.3
7.1
(n=4)
N
3
3
6
7
--
--
--
--
3.9 / 61
16 / 84
CONH2
(n=6) (6.2)
.8 ± 0.6
5
130 ±
25
N
OH
(n=4) (4.7) 9.7 (n=4)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 67
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
.6 ± 0.6
97 ± 26
(n=4)
N
3
3
4
8
9
0
120
10
0.15
0.3
8.2
2.6
13
5.2 / 36
4.1 / 37
4.5 / 40
N
OH
(n=4) (6.1)
2.1 ± 0.5
(n=16)
2
5 ± 18
n=8)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N
CONH2
(
(6.9)
3.5 ± 0.9
60 ± 13
(n=4)
N
23
0.19
N
CONHMe
(
n=4) (6.2)
aData are the geometric mean of at least two independent experiments ± standard deviation of the
b
average for the number (n) of experiments conducted. ND = not determined. Kinetic Solubility:
c
experimental procedures for kinetic solubility experiments in the Supporting Information; KS27 Ratio
d
defined in Supporting Information. A-B, apical-to-basolateral permeability as measured in Madin–
−
6
e
Darby canine kidney (MDCK) cells. Reported in units of ×10 cm/s. H/MLM defined as predicted
hepatic clearance in liver microsomes for human (H) and mouse (M). Reported in units of mL/min/kg.
Having explored the SAR on the ribose pocket group and hinge binding element with the
dihydrobenzofuran linker, we sought to introduce further changes to the solvent front leading to molecules
with a combination of improved potency, PK properties, and selectivity. Replacing a methyl in the solvent
front with a hydroxymethyl (41 and 42) improved the solubility of the scaffold (Table 6). Resolution of
the enantiomers revealed 41 to be equipotent with 23 with maintained selectivity and improved mouse
metabolic stability, while 42 lost roughly 2-3x potency in the biochemical and cellular systems and
suffered from slightly reduced kinase selectivity.
Table 6. Modification of the dihydrobenzofuran linker substitution.
ACS Paragon Plus Environment

Page 21 of 67
Journal of Medicinal Chemistry
1
2
3
4
N
N
IRAK4 Ki
nM)a
(LLEcLogP
NFkB
MDCK
O
Cmpd
NH
N
N
K. Sol. KS27
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
(
Cell IC50
Papp
H/MLMe
5.3 / 81
6.4 / 54
4.2 / 68
3.9 / 29
5.1 / 22
O
(μM)b Ratioc
#
)
(nM)a
(A-B)d
0
1
2
3
4
5
6
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
.3 ± 0.9
(n=8)
(5.7)
O
1
1
3
1
2
70 ± 38
n=4)
2
3
1
2
3
4
1
0.15
0.15
0.19
0.15
0.11
14
15
27
21
13
(
HO
5.1 ± 1.2
(n=14)
(6.9)
O
70 ± 71
4
4
4
4
10
(n=13)
HO
15 ± 3
(n=14)
(6.4)
O
50 ± 79
10
(n=12)
7
.3 ± 2.4
HO
O
70 ± 92
(n=14)
(5.9)
150
130
(n=8)
N
7.4 ± 4.2
(n=10)
(5.9)
O
30 ± 92
n=4)
(
aData are the geometric mean of at least two independent experiments ± standard deviation of the average
b
for the number (n) of experiments conducted. Kinetic Solubility: experimental procedures for kinetic
c
d
solubility experiments in the Supporting Information; KS27 Ratio defined in Supporting Information. A-
B, apical-to-basolateral permeability as measured in Madin–Darby canine kidney (MDCK) cells. Reported
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 67
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
−6
e
in units of ×10 cm/s. H/MLM defined as predicted hepatic clearance in liver microsomes for human
(
H) and mouse (M). Reported in units of mL/min/kg.
Further disruption to the planarity of the molecule with introduction of the tertiary alcohol (43) or
0
1
2
3
4
5
6
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
-oxa-8-azaspiro[4.5]decane 44 greatly improved the solubility and predicted mouse metabolic stability,
either maintained or improved permeability, and maintained selectivity and potency. X-ray co-crystal
structures of 23 and 41 with IRAK4 confirmed that these substitutions along the solvent front added to
the van der Waals contacts to the protein without perturbing the binding mode of the inhibitors (Figure 3).
Additionally, the hydroxymethyl substituent of 41 formed hydrogen bonds to a network of water
molecules at the solvent front.
Figure 3. Co-crystal structures of dihydrobenzofuran IRAK4 inhibitors. (A, PDB ID 6O8U) A 1.8Å
resolution co-crystal structure of 23 bound in the IRAK4 active site revealed three hydrogen bonds formed
to the protein or solvent molecules and the protrusion of the gem-dimethyl moiety into the region of the
site unoccupied by inhibitor atoms in the co-crystal structure of 5 with IRAK4. (B, PDB ID 6O95) A
1.8Å resolution crystal structure of 41 revealed additional polar contacts made by the hydroxymethyl
substituent that further fills a previously unoccupied region, also adding 3-dimensional vectors to the
compound. (C) The solvent accessible surface for 41, colored as in Figure 2.
ACS Paragon Plus Environment

Page 23 of 67
Journal of Medicinal Chemistry
1
2
3
4
5
In order to understand if these changes to solubility and other properties improved in vivo
properties relative to the 5,6-fused heteroaryl systems, 41 was taken first into PK studies. Gratifyingly,
initial in vivo PK data showed an order of magnitude improvement in the Cmax, AUC and bioavailability
for mouse when compared to compound 22, with in vivo CL in line with the in vitro CL observed across
species (Table 7). Dose escalating PK studies in mice showed a less than dose proportional increase in
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
exposure between 10 and 100 mg/kg, though all doses were sufficient to cover IC for at least part of the
50
1
h time course for PK/PD (see Figure 4). Modification of formulation also led to improved bioavailability
and exposure as shown in Table 7. With this superior PK in hand, we advanced 41 into an interleukin-6
IL-6) PK/PD mouse model.
(
Table 7. PK profile of 41.
IV CLpd
/ Vss
LMa /
Species
PPB
(%)
Dose
Cmax AUC∞
F
t1/2
(h)
CLhepb
(L/kg)
(mg/kg) (μM) (μM*h) (%)
5g
2.1
5.4
7.2
4.8
9.7
10
2.3
10
59% 0.5
10g
-
-
1.3
1.3
0g
3
17
Mouse
56 / 51
86
54 / 1.1e
100g
15.1
98
19% 1.6
49% 3.3
22% 23.2
33% 42.0
42% 3.8
30% 1.7
1
3
3
00h
00g
00h
1g
63
14
340
3
Rat
8.1 / 10
8.8 / 7.8
79 6.6 / 2.3e
9.3 / 8f
0.4
0.2
Dog
66
1g
1.4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 67
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
Monkey
6.7 / 8.8
51
11 / 4.4e
1g
0.3
3.3
85% 3.0
Human <4.3 / <6.2 56
a
b
LM defined as predicted hepatic clearance in liver microsomes. Reported in units of mL/min/kg; CLhep
0
1
2
3
4
5
6
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
defined as predicted hepatic clearance in hepatocytes. Reported in units of mL/min/kg; Cl = plasma
p
e
clearance. Reported in units of mL/min/kg; PPB = plasma protein binding, % bound. Compound dosed
f
IV (1 mg/kg) as a solution in PEG400:EtOH:H O (35:5:60); Compound dosed IV (0.3 mg/kg) as a
2
g
h
solution in PEG400:EtOH:H O (60:5:35); Compound dosed po as a suspension in MCT; Compound
2
dosed po as a suspension in 10% captisol.
Attenuation of pro-inflammatory cytokines upon treatment with IRAK4 inhibitors in a PK/PD
model was of critical interest to the project as many efficacy models for autoimmune disease are both time
and material consuming. While both vertebrates and bacteria store critical genomic information in DNA
using the same bases, there are slight but immunologically important differences.23-24 Unmethylated CpG
dinucleotide (CpG) is one such motif found in bacterial and viral DNA but not mammalian DNA.25
Immunocompetent cells such as dendritic cells (DCs) and macrophages recognize this “non-self” motif
26
through TLR9, leading to pro-inflammatory cytokine release including TNF-α and IL-6 along with other
interleukins and interferons. As recognition and activation of TLR9 in the innate immune system may
contribute to the pathogenesis of SLE, inhibition of TLR9’s downstream signaling and cytokine
stimulation by IRAK4 inhibition has strong potential for therapeutic intervention.
In order to validate the hypothesis of IRAK4 involvement in cytokine release, mice were treated
with 41 1 h prior to CpG1826 stimulation. Blockade of TLR9 signaling via inhibition of IRAK4 should
lead to suppression of both IL-6 and TNFα, as they are downstream of TLR9. After 1 h, blood and serum
samples were collected for cytokine and PK analysis. Statistically significant inhibition of both IL-6
ACS Paragon Plus Environment

Page 25 of 67
Journal of Medicinal Chemistry
1
2
3
(69%) and TNFα (39%) induction were observed at a 10 mg/kg dose (Figure 4, A, B). Increasing the dose
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
yielded increasing exposure up to 100 mg/kg (Figure 4, C), with this top dose level resulting in 91%
inhibition of IL-6 induction and 89% inhibition of TNFα (Figure 4, A, 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
Figure 4. Cytokine inhibition upon treatment with 41 in a CpG-stimulated mouse PK/PD model. (A) IL-
6
response, (B) TNFα response (C) Plasma compound PK.
Model design: 5 mice per group were enrolled. Single dose of the compound or vehicle was given orally
at -1 hour; the stimulus CpG1826 or PBS was dosed intravenously at 0 hour; and whole blood was
collected at 1 hour post CpG1826 for plasma PK and serum cytokines (IL-6 and TNFa) measured by
MSD. Plasma compound PK is from blood samples collected from the same experiment of the cytokine
analysis as shown in A) and B). One-way ANOVA analysis comparing all groups to the stimulus group
(Group 2: vehicle + GpG1826) was performed. *: P<0.05; **: P<0.01; ***: P<0.001; ***: P<0.0001.
CHEMISTRY
A synthetic route amenable to diversifying multiple portions of the molecules and applicable to
both the 5,6 and 6,6-biaryl systems was derived. The 5,6 bi-aryl systems were prepared via a similar route,
allowing diversification of the ribose pocket group and installation of the pyrazolopyrimidine hinge
27
binder. Routes for the benzothiazole and benzoxazole cores have been previously disclosed. Synthesis
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 67
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
of benzimidazole linker C is shown below (Scheme 1). Addition of 4-chloro-5-nitrobenzene-1,2-diamine
45) to a nitrile containing the desired substitution (I1) promoted by ytterbium amides provides an efficient
(
2
8
one-step synthesis of benzimidazole core I2 via an N-arylamidine intermediate, which could be 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
manipulated with alkylation of the imidazole nitrogen to give intermediate I3. Extrapolating this to all the
5
,6-biaryl systems gave late-stage intermediate I4. Nucleophilic aromatic substitution with various
secondary amines proceeded in high yield to give intermediate I5. Reduction of the nitro group via
palladium-catalyzed hydrogenation with subsequent amide coupling then yielded final targets 15-22.
Routes to quinoline and naphthyridine cores to provide 5 through 13 as well as indole 14 have been
previously reported.19
Scheme 1. Synthesis of benzimidazole intermediates and final 5,6-biaryl molecules.
O
O
O
R1
N
R2
N+
H
N
N+
N+
H2N
H2N
O-
a
R1
N
R2
b
R1
N
R2
X
N
O-
N
+
O-
Cl
N
Cl
Cl
I1
45
I2
I3
X = NH,NR
N
N
N
O
N
O
O
NH2
N
N+
c,d
R1
N
R2
X
N
OH
R1
N
R2
X
N
O-
NH
N
R1
N
R2
X
N
R3
N
e
R3
N
Y
R4
I4
I5
15-22 R4
X = S, O , NH, NR
Y = Br, F, Cl
Reagents and conditions: (a) Ytterbium triflate, xylenes, 140ºC, 48h; (b) MeI/EtBr, Cs CO , DMF, 90
2
3
3
4
min; (c) NHR R , 100 ºC, 4h; (d) Pd/C, MeOH (e) HATU, DIPEA, DMF.
The semi-saturated dihydrobenzofuran core was obtained through multiple routes. The original
synthetic route provided a direct approach to initial lead 23 via late stage intermediate 49 as well as targets
2
5-32 (Scheme 2). Etherification of 3-chloro-4-nitrophenol 46 with 3-bromo-2-methylprop-1-ene
followed by nucleophilic aromatic substitution with morpholine provided 4-(5-((2-methylallyl)oxy)-2-
nitrophenyl)morpholine 47 in high yield over two steps. Claisen rearrangement-rearomatization under
ACS Paragon Plus Environment

Page 27 of 67
Journal of Medicinal Chemistry
1
2
3
microwave conditions provided a complex mixture including desired alcohol 48, regioisomer 50, cyclized
furan 51 and recovered starting material (27%). Gem-dimethyl dihydrobenzofuran 49 was then generated
from phenol 48 under acidic conditions. Hydrogenation of the nitro group followed by amide coupling
then gave furan 23 with poor overall conversion. The analogous route also provided hydroxymethyl
piperidine 24.
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
Scheme 2. Initial route for generating dihydrobenzofuran 23.
N
N
NO2
N
NO2
N
NO2
N
O
Br
(a),
b), HN
(c)
(d)
(e)
(f)
NO2
Cl
NH
N
N
O
HO
O
(
N
N
HO
O
O
O
O
O
O
O
46
47
48
49
OH
N
23
[1,3] Sig Rearrangement Products:
NO2
N
NO2
N
NO2
HO
HO
N
O
O
O
O
4
8
50
(16%)
51
(20%)
(12%)
Reagents and conditions: (a) K CO , MeCN, 55 ºC, quant.; (b) K CO , DMSO, 100 ºC, 30 min, 93%
2
3
2
3
yield; (c) μw, DMF, 220 ºC, 60 min; (d) 35% aqueous HCl in MeOH, 65% yield;(e) H , Pd/C, MeOH;
2
(f) PyAOP, 2,4,6-trimethylpyridine, DMF (45% yield).
Through further optimization, two synthetic routes were identified that allowed for scaffold
diversification. The first route (Scheme 3, Route A) involved epoxidation of intermediate 48 with mCPBA,
which was accompanied by cyclization to generate the hydroxymethyl dihydrobenzofuran, which could
then be elaborated to pyrazolopyrimidine IRAK4 inhibitors following the same transformations described
in Scheme 2 (compound 49 to 23). However, the epoxidation-cyclization sequence was low yielding. An
improved route to the dihydrobenzofurans allowed for increased flexibility beginning with bromide 52
(Scheme 3, Route B). Bromide 52 was converted to the corresponding Grignard reagent and reacted with
methyl 2-oxopropanoate to afford tertiary alcohol 53, which could be cyclized under basic conditions to
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 67
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
give acid 54. Nitration proceeded selectively at the C3 position to give intermediate 55, which, after
reduction using borane, provided the versatile hydroxymethyl dihydrobenzofuran intermediate I6. This
intermediate allowed for modification of the ribose pocket group as well as diversification of the hinge
binding element via nitro reduction and amide formation analogous to those shown in Schemes 1 and 2.
Scheme 3. Revised route to dihydrobenzofuran analogs.
0
1
2
3
4
5
6
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
Route A:
N
N
NO2
N
O
(a)-(b)
c),
NH
N
N
HO
N
N
(
O
O
O
HO
(e)
48
41-42
O
OH
N
Route B:
O
O
(d)
(f)
(g)
NO2
Cl
Br
F
O
NO2
Cl
HO
HO
HO
OH
Cl
O
F
Cl
O
O
O
Cl
52
53
54
55
I6
Reagents and conditions: (a) 1) mCPBA, DCM, 20% yield; (b) H , Pd/C; (c) PyAOP, 65% yield; (d)
2
Mg, cat. I , Et O, Δ then methyl 2-oxopropanoate; THF, -78 ºC to rt, 2 h, 51% yield (e) tBuOK, THF, 60
2
2
ºC, 12 h, 64% yield; (f) HNO , DCM, rt, 5 min, 54% yield; (g) BH3, THF, rt, 2 h, 32% yield.
3
highest resolution shell; n.d. parameter not determined.
Modification of 1-(bromomethyl)-4-chloro-2-fluorobenzene 52 and carboxylic acid 55 also allowed for
modification of the solvent front and linker regions. The synthetic details of additional intermediates
and characterization details are provided in the Experimental Section and Supporting Information.
CONCLUSIONS
In summary, we have identified a series of highly potent and selective IRAK4 inhibitors through a
structure-based-design approach. Initial molecules contained either 6,6 or 5,6-biaryl linker systems. While
the 5,6-biaryl ring systems such as benzimidazoles and benzothiazoles provided potent and stable
ACS Paragon Plus Environment

Page 29 of 67
Journal of Medicinal Chemistry
1
2
3
molecules in rodents, poor oral exposure was uniformly seen due to the formation of poorly soluble, highly
stable crystals. Reducing planarity while maintaining key hydrogen bond interactions was hypothesized
as a path forward to improve the physicochemical properties of inhibitors. This approach led to a new
dihydrobenzofuran linker which were optimized to provide 41. This was the first molecule from the
pyrazolopyrimidine class to show improved stability and oral exposure across species. Most gratifyingly,
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
4
1was also able to inhibit production of two cytokines that are critical to autoinflammatory response, IL-
and TNFα, after TLR induction. Given this precedence for scaffold hopping and the accompanying
6
strategy for generating inhibitors with improved exposure, further development is being pursued and will
be reported in due course.
EXPERIMENTAL SECTION
All chemicals were purchased from commercial suppliers and used as received. 1H NMR spectra were
recorded on Bruker Avance 400 or 500 spectrometers. Chemical shifts are expressed in δ ppm referenced
to an internal standard, tetramethylsilane (δ = 0 ppm). Abbreviations used in describing peak signals are
br = broad signal, s = singlet, d = doublet, dd = doublet of doublets, t = triplet, q = quartet, m = multiplet.
All final compounds were purified to have purity higher than 95% by reverse phase high-performance
liquid chromatography (HPLC), supercritical fluid chromatography (SFC) or normal phase silica gel flash
chromatography. The purity was assessed by reverse phase HPLC with an isocratic gradient of 5–95%
acetonitrile in water (with either acid or base modifier) and monitored by diode array ultraviolet detection
at 254 nm. Low-resolution mass spectra were recorded on a liquid chromatography–mass spectrometer in
electrospray positive (ES+) mode. HRMS experiments were performed on a Dionex LC Ultimate3000
coupled with a ThermoScientific Q Exactive orbitrap mass spectrometer using ESI as the ionization source
and a Phenomenex XB-C18, 1.7 mm, 50 mm × 2.1 mm column with a 0.7 mL/min flow rate at 40 °C for
ACS Paragon Plus Environment