Binding to an Unusual Inactive Kinase Conformation by Highly Selective Inhibitors of Inositol-Requiring Enzyme 1α Kinase-Endoribonuclease

Article abstract of DOI:10.1021/acs.jmedchem.8b01721

A series of imidazo[1,2-b]pyridazin-8-amine kinase inhibitors were discovered to allosterically inhibit the endoribonuclease function of the dual kinase-endoribonuclease inositol-requiring enzyme 1α (IRE1α), a key component of the unfolded protein response in mammalian cells and a potential drug target in multiple human diseases. Inhibitor optimization gave compounds with high kinome selectivity that prevented endoplasmic reticulum stress-induced IRE1α oligomerization and phosphorylation, and inhibited endoribonuclease activity in human cells. X-ray crystallography showed the inhibitors to bind to a previously unreported and unusually disordered conformation of the IRE1α kinase domain that would be incompatible with back-to-back dimerization of the IRE1α protein and activation of the endoribonuclease function. These findings increase the repertoire of known IRE1α protein conformations and can guide the discovery of highly selective ligands for the IRE1α kinase site that allosterically inhibit the endoribonuclease.

Full text of DOI:10.1021/acs.jmedchem.8b01721

Subscriber access provided by UNIV OF TEXAS DALLAS
Article
Binding to an unusual inactive kinase conformation by highly selective
inhibitors of inositol-requiring enzyme 1# kinase-endoribonuclease
Giampiero Colombano, John J Caldwell, Thomas Peter Matthews, Chitra Bhatia, Amar
Joshi, Tatiana McHardy, Ngai Yi Mok, Yvette Newbatt, Lisa Pickard, Jade Strover, Somaieh
Hedayat, Michael Ian Walton, Stephanie Myers, Alan Morenc Jones, Harry Saville, Craig
McAndrew, Rosemary Burke, Suzanne Eccles, Faith Davies, Richard Bayliss, and Ian Collins
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01721 • Publication Date (Web): 19 Feb 2019
Downloaded from http://pubs.acs.org on February 20, 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 69
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

Journal of Medicinal Chemistry
Page 2 of 69
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
Binding to an unusual inactive kinase conformation
by highly selective inhibitors of inositol-requiring
enzyme 1 kinase-endoribonuclease
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
1
1
3
3
Giampiero Colombano , John J. Caldwell , Thomas P. Matthews , Chitra Bhatia Amar Joshi ,
,
1
1
1
1
2
Tatiana McHardy , Ngai Yi Mok , Yvette Newbatt , Lisa Pickard , Jade Strover , Somaieh
1
1
1
1
1
Hedayat , Michael I. Walton , Stephanie M. Myers , Alan M. Jones , Harry Saville , Craig
1
1
1
2,†
3,4
McAndrew , Rosemary Burke , Suzanne A. Eccles , Faith E. Davies , Richard Bayliss , Ian
1
,
Collins *
1
. Cancer Research UK Cancer Therapeutics Unit and 2. Division of Molecular Pathology, The
Institute of Cancer Research, London, SW7 3RP, U.K.
3
. Department of Molecular and Cell Biology, University of Leicester, LE1 7RH, U.K.
4
. School of Molecular and Cellular Biology, Faculty of Biological Sciences, University of
Leeds, LS2 9JT, U.K.
ACS Paragon Plus Environment
1

Page 3 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Abstract
A series of imidazo[1,2-b]pyridazin-8-amine kinase inhibitors was discovered to allosterically
inhibit the endoribonuclease function of the dual kinase-endoribonuclease inositol-requiring
enzyme 1 (IRE1, a key component of the unfolded protein response in mammalian cells and
a potential drug target in multiple human diseases. Inhibitor optimization gave compounds with
high kinome selectivity that prevented endoplasmic reticulum stress-induced IRE1
oligomerization and phosphorylation, and inhibited endoribonuclease activity in human cells. X-
ray crystallography showed the inhibitors to bind to a previously unreported and unusually
disordered conformation of the IRE1 kinase domain that would be incompatible with back-to-
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
back dimerization of the IRE1 protein and activation of the endoribonuclease function. These
findings increase the repertoire of known IRE1 protein conformations and can guide the
discovery of highly selective ligands for the IRE1 kinase site that allosterically inhibit the
endoribonuclease.
Keywords
IRE1; kinase; endoribonuclease; allosteric inhibitor; protein dimerization
ACS Paragon Plus Environment
2

Journal of Medicinal Chemistry
Page 4 of 69
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
The dual kinase-endoribonuclease inositol requiring enzyme 1 (IRE1; ERN1) is a central
and conserved component of the unfolded protein response (UPR) activated by eukaryotic cells
in reaction to endoplasmic reticulum (ER) stress, caused by an excess of misfolded proteins in
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
,2
the ER lumen . Activation of the UPR reduces protein translation and increases protein folding
capacity. If this initial response does not alleviate ER stress, prolonged activation of the UPR
signaling network leads to apoptosis. Thus the balance of downstream anti- and pro-apoptotic
signaling pathways activated by the UPR determines cell fate in the presence of acute or chronic
proteotoxic stress.
IRE1 is an ER-membrane resident protein with serine/threonine kinase and endoribonuclease
3
(
RNase) functions contained in its cytosolic domain . The accumulation of unfolded or
misfolded proteins in the ER lumen triggers oligomerization of IRE1 and kinase
autophosphorylation, resulting in activation of the RNase4,5,6. The RNase specifically cleaves the
mRNA coding for the unspliced transcription factor X-box protein 1 (XBP1u) to remove a 26
base pair intron, and subsequent ligation produces the mRNA of the active, spliced form
7
(
XBP1s) of the transcription factor . XBP1s-dependent transcription includes upregulation of
8
ER-resident molecular chaperones required to relieve aberrant protein folding . In contrast, non-
specific cleavage of other mRNAs localized at the ER membrane by IRE1 RNase (termed
9
regulated IRE1-dependent decay, RIDD) is proposed to exert an apoptotic effect . Additionally,
the interaction of TRAF2 with oligomerized phospho-IRE1 elicits signaling through the JNK/c-
10
JUN axis, independent of IRE1 RNase activity . In mammalian cells, IRE1 functions with
the transcriptional repression and regulation of apoptosis effected by the other UPR elements,
1
,2
ATF6 and PERK, to alleviate ER stress or to direct cell death .
ACS Paragon Plus Environment
3

Page 5 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Dysregulation of IRE1-XBP1s signaling is implicated in a number of human diseases,
including cancer, diabetes, lipidemia, inflammatory disease and neurodegeneration1,11. Signaling
through IRE1-XBP1s has been proposed as important for the survival of myeloma cells, as well
as breast and prostate cancers12,13,14. Recently, in an ovarian cancer in vivo model, XBP1s
function has been shown to suppress proper presentation of tumor-associated antigens by tumor-
infiltrating dendritic cells, thus determining the degree of T-cell anti-tumor immune response,
and suggesting that modulation of XBP1 splicing in normal tissue may be an alternative
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
5
approach for an anticancer therapy .
There is a clear need for selective chemical tools to explore the consequences of IRE1
inhibition by different mechanisms in cancer and non-transformed cells, and to study the IRE1-
XBP1s pathway in other human diseases. Given the diverse nature of IRE1 signaling outputs,
inhibitors targeting the kinase or RNase functions, or both, may have different consequences for
cell fate, which may further vary according to cell context. Reversible covalent binders targeting
the RNase active site directly have been reported from a salicylaldehyde scaffold16,17,18, and have
been shown to inhibit cancer cell growth18,19,20. The pharmacology of inhibitors binding at the
IRE1 kinase site is complicated by the possibility of allosteric inhibition or activation of the
RNase function depending on the conformation of the kinase domain and oligomerization state
21
that is stabilized . Allosteric inhibition of the RNase function has been associated with Type II
2
1
kinase site ligands that bind the DFG-out conformation of the kinase or other compounds that
displace the C-helix of the kinase from the active conformation22,23,24. While a series of highly
selective, kinase site binding, allosteric inhibitors of IRE1 RNase have shown no direct
2
3
cytotoxic effect on a range of cancer cell lines , compounds with this profile remain of interest
25
for other therapeutic approaches that target non-malignant cells . In contrast to Type II and C-
ACS Paragon Plus Environment
4

Journal of Medicinal Chemistry
Page 6 of 69
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
helix displacing scaffolds, reported Type I IRE1 kinase site ligands promote dimerization and
activate the RNase function6,21. The identification of new activators and inhibitors of IRE1
RNase, and understanding the allosteric mechanisms linking the kinase and RNase sites, are
therefore of continued importance for the development of new therapeutics.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
We previously described the Type I imidazo[1,2-b]pyridazine kinase inhibitor 1 (Figure 1) that
6
allosterically activates IRE1 endoribonuclease in vitro . While screening chemical scaffolds
based on 1 we unexpectedly discovered the related Type I imidazo[1,2-b]pyridazin-8-amine 2
(Figure 1A) to be an inhibitor of both the kinase and RNase functions of IRE1. In this study we
show how 2 and more potent analogues bind to a previously undescribed inactive conformation
of monomeric IRE1 involving major rearrangements of key secondary structure in the kinase
domain, including disordering of the C-helix and disruption of the hydrophobic spine of the
protein, which together confer high IRE1 kinase binding selectivity and inhibition of both
kinase and RNase functions. We confirm that these allosteric modulators disrupt all downstream
signaling through IRE1 in human cells, thus validating targeting of this unusual binding mode
as a new way to generate selective and cell active inhibitors of IRE1.
ACS Paragon Plus Environment
5

Page 7 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
N
A
NH
O
NH2
O
Cl
N
N
HN
N
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N 1
8
N 1
NH
16
2
IRE1 Kinase inhibitor
IRE1 RNase activator
IRE1 Kinase inhibitor
IRE1 RNase inhibitor
B
N
N
NH2
H
N
H
N
CF3
N
O
KIRA321
Cl
H
N
S
N
O
O
O
NH2
N
N
N
H
Compound 1623
Figure 1: A) Structures of the IRE1 RNase activator 1 and the structurally related RNase
inhibitor 2; B) Structures of selected published IRE1 kinase-RNase inhibitors.
Synthetic Chemistry
We established short synthetic routes to prepare analogues of 2 using the differential reactivity
of the poly-halogenated imidazo[1,2-b]pyridazine 3 (Scheme 1). Selective displacement of the 8-
bromo substituent introduced the (cyclopropylmethyl)amine and was followed by Suzuki
reactions of the 3-bromo group of 4 to give compounds 5-12. The 6-chloro substituent of 3 and 4
might be expected to undergo competing nucleophilic substitution or palladium insertion but
ACS Paragon Plus Environment
6

Journal of Medicinal Chemistry
Page 8 of 69
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
26
proved notably stable under the conditions studied . To replace the indazol-5-yl substituent of 2
with 2-(N-phenyl)amino-1H-benzo[d]imidazole, the intermediate 13 was prepared from 5-
bromo-2-chloro-1H-benzo[d]imidazole by displacement of the 2-chloro group with aniline and
N-protection, before palladium-catalyzed installation of the boronate (Scheme 1). Compound 13
was obtained and used as a mixture of N-Boc 2-aminobenzimidazole regioisomers, with
complete N-deprotection observed in the subsequent Suzuki reaction. The 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
(
methylamino)benzimidazole 14 did not require N-protection for successful reaction. Other N-
methylated 2-aminobenzimidazoles 16 and 18 were synthesized from the nitrophenyl boronates
and 17, respectively, by nitro group reduction and condensation with
1
5
2
7
isothiocyanatobenzene .
a
Scheme 1 : Syntheses of compounds 2 – 18.
ACS Paragon Plus Environment
7

Page 9 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Br
Br
R1
Cl
N
Cl
N
Cl
N
Cl
N
N
i, ii
N
iii
N
iv
N
NH2
N
N
N
Br
Br
NH
NH
3
4
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
Br
N
NH
HN
Ph
O
B
N
NH
N
X
Ph
O
2
Y
N
N
NMe
9
HN
MeN
Ph
Ph
Ph
v
Cl
N
N
1
3 X = Y = NBoc
5
NH
vi, vii
HN
N
10
1
4 X = NH, Y = NMe
HN
6
7
NMe
O
O
NH
B
viii
O
11
12
HN
N
B
Y
X
Ph
NH
N
O
MeN
NH
Y
X
8
15 X = NHMe, Y = NH2
16 X = NMe, Y = N
18 X = N, Y = NMe
17 X = NH , Y = NHMe
2
a
Reagents and conditions: i) 2-Chloroacetaldehyde, EtOH, 50°C, 21 h, 78%; ii) NBS, CH Cl,
3
0
°C to rt, 16 h, 84%; iii) Cyclopropylmethanamine, THF, rt, 3 h, 93%; iv) Heteroaryl boronate,
Pd(OAc) , 1,1'-bis(di-tert-butylphosphino)ferrocene, 2M Na CO , dioxane, 120-135°C, 15-22 h,
-51%; v) a) PhNH , 180°C (microwave), 1 h; b) (Boc) O, DMAP, THF, rt, 16 h; c)
2 2
Bis(pinacolato)diboron, KOAc, Pd(dppf)Cl .CH Cl , 1,4-dioxane, 100°C, 16 h, 53% over 3
steps; vi) PhNHMe, 180°C (microwave), 1 h, 86%; vii) Bis(pinacolato)diboron, KOAc,
Pd(dppf)Cl .CH Cl , 1,4-dioxane, 100°C, 16 h, 36%; viii) Isothiocyanatobenzene, EDC, THF,
2
2
3
5
2
2
2
2
2
2
5
0°C, 5 h, 17-45%.
The synthesis was adapted to vary the 8-amino substituent of the imidazo[1,2-b]pyridazine by
reacting a range of amines with substrate 3 to give 19-21 and 27-29 (Scheme 2). For the
unsubstituted 1,4-diaminocyclohexanes 22 and 23, N-protection of the terminal amine, to ensure
selectivity, and higher temperatures were required in the nucleophilic substitution step. The more
elaborate 1,4-diaminocyclohexane substituents incorporated into 24-26 were made prior to
reaction with 3. Fluorine substituents were introduced into the 2-(N-phenyl)amino-1H-
ACS Paragon Plus Environment
8

Journal of Medicinal Chemistry
Page 10 of 69
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
benzo[d]imidazole group starting from the appropriate fluorinated 4-bromobenzene-1,2-diamines
via condensation with isothiocyanatobenzene, and the route was telescoped to generate
compounds 30-32.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
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
Scheme 2 : Syntheses of compounds 19 – 32.
ACS Paragon Plus Environment
9

Page 11 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Ph
HN
N
NH
NH
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
Cl
N
Cl
N
i, ii
N
N
i, ii
3
3
N
N
R2
R2
R2
HN
HN
19
27
O
H N
2
NH2
F
2
0
1
28
29
NMe
Br
NMe
2
HN
HN
x
ii, iii, iv
3
22
Ph
HN
N
NHBoc
NH2
NH
v, vi
23
N
7
4
F
NH2
NH2
O
6
Cl
N
N
vii, ii
vii, ii
3
24
25
HN
HN
N
HN
NMe
N
H
3
3
0 7-F
31 6-F
2 4-F
N
Cl
O
O viii, ix, ii
3
26
HN
N
O
O
O
N
N
H
ACS Paragon Plus Environment
1
0

Journal of Medicinal Chemistry
Page 12 of 69
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
aReagents and conditions: i) Amine, THF, rt, 16 h, 60-97%; ii) Heteroaryl boronate, Pd(OAc)2,
,1'-bis(di-tert-butylphosphino)ferrocene, 2M Na CO , 1,4-dioxane, 130°C, 16 h, 4-55%; iii)
tert-Butyl ((1r,4r)-4-aminocyclohexyl)carbamate, THF, 50°C, 16 h, quantitative; iv) HCl, 1,4-
1
2
3
dioxane, rt, 2 h, 99%; v) a) LiAlH , THF, 0°C to reflux, 4 h; b) Ethyl 1,3-dioxoisoindoline-2-
4
carboxylate, Et N, CH Cl , rt, 15 h; c) 3, Et N, 1,4-dioxane-DMSO, 100°C, 22 h, 52% over 3
3
2
2
3
steps; vi) 5-(4,4,5,5-Tetramethyl-1,3,2-dioxaborolan-2-yl)-1H-indazole, Pd(OAc) , 1,1'-bis(di-
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
tert-butylphosphino)-ferrocene, 2M Na CO , 1,4-dioxane, 130°C, 21 h, then NH NH .H O,
2
3
2
2
2
1
00°C, 48 h, 34%; vii) Amine, THF, 50°C, 16 h, 29-85%; viii) a) tert-Butyl ((1r,4r)-4-
aminocyclohexyl)carbamate, Et N, CH Cl , rt, 20 h; b) HCl-1,4-dioxane, MeOH, rt, 2 h, 37%
3
2
2
over 2 steps; ix) 3, Et N, 1,4-dioxane-DMSO, 50°C, 23 h, 58%; x) a) Isothiocyanatobenzene,
3
EDC, THF, 50 °C, 5 h; b) (Boc) O, DMAP, THF, rt, 16 h c) Bis(pinacolato)diboron, KOAc,
2
Pd(dppf)Cl .CH Cl , 1,4-dioxane, 135°C, 24 h; d) 3-bromo-6-chloro-N-((1-methylpiperidin-4-
2
2
2
yl)methyl)imidazo[1,2-b]pyridazin-8-amine,
Pd(OAc)2,
1,1'-bis(di-tert-butylphosphino)-
ferrocene, 2M Na CO , 1,4-dioxane, 135°C, 16 h, 2-5% over 4 steps.
2
3
Results
Compound 2 (Figure 1A, Table 1) was discovered through screening of Type I kinase
6
scaffolds related to the IRE1 kinase inhibitor-RNase activator 1 . Unexpectedly, we determined
that while inhibiting the kinase activity with similar potency to 1, compound 2 was an inhibitor
of the RNase function in vitro, suggesting a different allosteric effect on the RNase compared to
2
1
typical Type I IRE1 kinase inhibitors . The small size of 2 compared to reported IRE1
kinase-RNase inhibitors from Type II or Type I-extended scaffolds (e.g. KIRA3 and Compound
1
6, respectively, Figure 1B) was also attractive as a potential starting point for inhibitor
optimization21,23,24. We hypothesized that the orientation of 2 within the IRE1 ATP site would
6
be similar to that determined for 1 (PDB 4z7h) . We anticipated that the key hinge-binding
hydrogen bond from N-1 of the imidazo[1,2-b]pyridazine would be maintained and that a new
hydrogen bond could be formed to the hinge region from the 8-amino group, replacing the
CH…O interaction seen for 1. In this orientation, the indazole substituent would be directed
toward the C-helix, taking the place of the benzamide moiety in 1, while the cyclopropylmethyl
group would be directed out towards solvent. The imidazo[1,2-b]pyridazin-8-amine hinge-
ACS Paragon Plus Environment
1
1

Page 13 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
binding group has recently been shown to adopt such a binding mode to the kinase, monopolar
spindle 1 (MPS1)28,29
.
We used in vitro biochemical assays to measure compound binding and effects on IRE1
autophosphorylation and endoribonuclease activities, to investigate the structure-activity
relationships around 2 and test the proposed binding model. Recombinant IRE1 protein
corresponding to the cytoplasmic kinase-endoribonuclease extended (KEN) domain
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
(G547IRE1) was produced separately in phosphorylated and dephosphorylated states, as
3
0
confirmed by intact protein mass spectrometry (Supplementary Figure S1) . We used our
previously described DELFIA format to determine the inhibition of recombinant IRE1 S274
3
1
autophosphorylation . To measure compound binding to IRE1 we developed a LanthaScreen
Eu Kinase binding FRET assay which monitored the displacement of an ATP-site tracer. To
determine the allosteric effect of the kinase inhibitors on IRE1 RNase activity, we used a
FRET-based assay to measure the cleavage of a short stem-loop RNA sequence derived from the
endogenous cleavage site in XBP1u mRNA (Supplementary Figure S2)21,30
.
We initially investigated the role of the indazol-5-yl substituent in 2 (Table 1). The availability
and correct positioning of hydrogen bond donor/acceptor functionality was critical, with neither
the N-methyl (5) nor regioisomeric (6) analogues showing activity. Only weak inhibition of
IRE1 autophosphorylation in the sensitive DELFIA format was observed for the indol-5-yl
derivative (7). The benzimidazol-5-yl analogue (8) also inhibited IRE1 autophosphorylation but
showed significantly weaker ATP-site binding in the LanthaScreen assay and no RNase
inhibition. For the RNase assay, this may in part reflect the reduced potency of 8 compared to 2.
The weak potency for ATP-site binding assay was unexpected, but based on the promising
ACS Paragon Plus Environment
1
2

Journal of Medicinal Chemistry
Page 14 of 69
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
inhibition of IRE1 autophosphorylation we pursued further substitution of the benzimidazol-5-
yl scaffold.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
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 1. In vitro binding and inhibition of IRE1 kinase and RNase activities by 2, 5-12.
No.
IRE1 S724
IRE1 ATP site binding
IRE1 RNase activity
autophosphorylation
b
c
IC (M)
IC (M)
5
0
50
a
IC (M)
5
0
16
2
0.218 (±0.15)6
0.303 (±0.15)
>10
1.18 (±0.17)
1.60 (±0.91)
>10
Activator EC 0.143
6
5
0
8.90 (±0.21)
n.d.d
5
6
>10
>10
n.d.
7
4.06 (±0.74)
0.89 (±0.14)
0.81 (±0.4)
>10
>10
>10
46% @10 M e
0.84 (±0.42)
>10
8
>10
9
2.67 (±1.89)
n.d.
10
11
>10
>10
n.d.
12
2.7 (±1.1)
>10
n.d.
a
Inhibition of recombinant G547 IRE1 KEN domain pS274 autophosphorylation measured
b
in DELFIA format, mean (±SD) for n ≥ 3; Inhibition of ATP site LanthaScreen tracer binding
to recombinant dephosphorylated G547 IRE1 KEN domain, mean (±SD) for n ≥ 3; Inhibition
c
of G547 IRE1-dependent cleavage of a FRET-labelled stem-loop RNA containing the XBP1
d
e
cleavage site, mean (±SD) for n ≥ 3; n.d. = not determined; ±4%, n=4.
We assumed that the indazol-5-yl substituent of 2 potentially interacted with the same region
of the IRE1 active site as the benzamide of 1, in particular with residues such as D711 from the
DFG-motif or K599. Compound 1 also interacted with a water molecule that was in contact with
6
E612 from the C-helix . Therefore, we investigated 2-(N-phenyl)amino substitution of the
ACS Paragon Plus Environment
1
3

Page 15 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
benzimidazole 8. The 2-(N-phenyl)amino-benzimidazole motif is a known isostere for the diaryl
urea motif more commonly found in Type II kinase inhibitors, as it is capable of mimicking the
hydrogen bonding pattern made by Type II urea inhibitors in the inactive DFG-out conformation
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
32
of the kinase domain . The addition of aryl urea groups or their isosteres to convert Type I
3
3
kinase inhibitor scaffolds to Type II molecules is a well-established approach . Gratifyingly, we
observed that the 2-(N-phenyl)amino-benzimidazole 9 had similar kinase site binding affinity to
2
, and improved on the inhibition of IRE1 RNase activity. Consistent with a role for the 2-(N-
phenyl)aminoimidazole functionality in forming multiple hydrogen bonds, N-methylation at any
position reduced inhibitory activity (see 10, 11 and 12).
With two variations of the imidazo[1,2-b]pyridazin-8-amine core identified that led to RNase
inhibition, the effects of varying the 8-amino group on potency and physicochemical properties
were studied. The aqueous buffer solubilities of 2 (KSol = 0.8 M) and 9 (KSol = 2.9 M) were
low, consistent with their measured lipophilicities (2, LogD7.4 = 3.4; 9, LogD7.4 = 4.1). We
therefore sought to introduce polar or ionizable groups to improve solubility, and hypothesized
that the vector defined by the cyclopropylmethyl group should be directed toward solvent.
Importantly, studies leading to the IRE1 inhibitor Compound 16 (Figure 1B) found that
appending cyclic amine or cyclohexylamine substituents directed out of the ATP-binding pocket
2
3
was essential for achieving high potency. Our presumed binding mode of the core scaffolds 2
and 9 oriented the 8-amino substituent in a similar direction. We investigated a range of
substituents, including cyclic amines analogous to those described in the development of
Compound 16, with selected examples highlighted in Table 2.
ACS Paragon Plus Environment
1
4

Journal of Medicinal Chemistry
Page 16 of 69
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. In vitro binding and inhibition of IRE1 kinase and RNase activities by 19-32 and
published compound KIRA3.
No.
IRE1 S724
autophosphorylation
IRE1 ATP
site binding
IRE1 RNase
activity
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
b
c
IC (M)
IC (M)
IC (M)
5
0
50
50
1
2
2
2
2
2
2
2
2
2
2
9
0
1
2
3
4
5
6
7
8
9
0.19 (±0.11)
0.25 (±0.07)
0.48 (±0.19)
0.77 (±0.13)d
0.13e
0.16e
0.83 (±0.19)
0.87 (±0.24)
0.11 (±0.04)
>10
0.47 (±0.09)
0.32 (±0.09)
>10
0.072 (±0.011)
n.d.f
0.033e
0.15 (±0.05)
0.19 (±0.01)d
0.24 (±0.09)
0.20 (±0.05)
0.13 (±0.06)
0.13 (±0.03)
0.15 (±0.02)
0.16 (±0.06)
0.16 (±0.05)
0.14 (±0.02)
0.11 (±0.01)
0.122 (±0.026)
0.087 (±0.022) 0.027 (±0.007)
2.0 (±0.94)
0.67 (±0.23)
0.39 (±0.11)
0.33 (±0.04)
0.27 (±0.17)
0.48 (±0.05)
0.40 (±0.16)
0.51e
0.18e
0.09 (±0.05)
0.10 (±0.05)
0.08 (±0.05)
0.17 (±0.09)
0.45 (±0.10)
30
31
32
KIRA3 0.29 (±0.06)
a
Inhibition of recombinant G547 IRE1 KEN domain pS274 autophosphorylation measured
b
in DELFIA format, mean (±SD) for n ≥ 3; Inhibition of ATP site LanthaScreen tracer binding
to recombinant dephosphorylated G547 IRE1 KEN, mean (±SD) for n ≥ 3; Inhibition of G547
c
IRE1-dependent cleavage of a FRET-labelled stem-loop RNA containing the XBP1 cleavage
d
e
f
site, mean (±SD) for n ≥ 3; n = 2; single determination; n.d. = not determined.
Introduction of polar groups generally maintained the dual kinase-RNase inhibitory profile of
the compounds while enhancing potency. Thus, replacement of the cyclopropylmethyl
ACS Paragon Plus Environment
1
5

Page 17 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
substituent by tetrahydropyran-4-yl resulted in sub-micromolar inhibitors of IRE1 RNase
function, 19 and 27. Aqueous solubility remained low (19, KSol = 2.2 M; 27, KSol = 2.4 M),
but was increased in the basic N-methylpiperidin-4-yl derivatives 20 (KSol = 58 M, LogD7.4
.8) and 28 (KSol = 41 M, LogD7.4 = 2.7) which also showed improvements in RNase
inhibition.
=
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
For the 3-(indazol-5-yl) substituted analogues, extension of the piperidine substituent had a
negligible effect on compound affinity or potency (see 21). However, the trans-1,4-
diaminocyclohexane analogue 22 showed enhanced binding and RNase inhibition. Notably, N-8-
methylation to give 23 led to loss of activity, consistent with the assumed kinase hinge-binding
role of the 8-amino substituent. Further substitution of the trans-1,4-diaminocyclohexane group
was tolerated, exemplified by 24, 25 and 26, and led to potent in vitro inhibitors of IRE1 RNase
activity with acceptable aqueous solubility (e.g. 26, KSol = 80 M, LogD = 1.8).
7
.4
For the 3-(2-(N-phenyl)aminobenzimidazol-5-yl) substituted compounds, the (N-
methylpiperidin-4-yl)methyl analogue 29 provided an enhancement in binding and RNase
inhibitory potency without the need for further substitution. We examined fluorination of the
benzimidazole group but found only marginal (<2-fold) gains in affinity and potency (see 30, 31,
32). The most potent RNase inhibitor 31 from this scaffold showed moderate solubility (KSol =
3
8 M), reflecting the high lipophilicity (LogD = 3.7).
7
.4
Table 3: Cellular activities of selected IRE1 inhibitors
No.
XBP1-Luciferase fusion mRNA
splicing in HEK293 cells
Cytotoxicity in HEK293 reporter cells
b
EC (M)
5
0
a
IC (M)
5
0
2
4.2 (±1.7)
>49
ACS Paragon Plus Environment
1
6

Journal of Medicinal Chemistry
Page 18 of 69
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
2
2
2
3
2
6
9
1
2.9 (±0.87)
1.12 (±0.24)
0.89 (±0.1)
0.71 (±0.33)
28 (±6.9)
>49
12c
9.0 (±1.6)
>49
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
KIRA3 0.76c
a
Inhibition of tunicamycin-induced IRE1-dependent splicing of an XBP1u-luciferase mRNA
b
reporter stably expressed in HEK293 cells, mean (±SD) for n≥2; Cytotoxicity in HEK293 cells
stably expressing an XBP1u-luciferase mRNA reporter, measured using the Alamar Blue format,
mean (±SD) for n≥2; Single determination.
c
Cellular inhibition of IRE1 RNase function was measured for selected compounds using a
bioluminescent reporter assay in HEK293 cells stably expressing an XBP1-luciferase fusion
mRNA to quantify XBP1 mRNA splicing (Table 3). In the absence of ER stress, translation of
the XBP1u mRNA terminates before the luciferase sequence, while exogenous induction of ER
stress leads to IRE1-dependent XBP1 mRNA splicing and results in tandem translation of the
XP1s-luciferase reporter protein. The imidazo[1,2-b]pyridazine 2 moderately inhibited
tunicamycin-induced mRNA splicing. Enhancement of the in vitro potency and aqueous
solubility of the scaffold, to give 22 and 26, was accompanied by increased cellular activity.
Likewise, the most potent and soluble 3-(2-(N-phenyl)aminobenzimidazol-5-yl) substituted
analogues 29 and 31 gave sub-micromolar inhibitors of IRE1 activity in cells. The optimization
of in vitro potency in this series involved addition of basic groups, which may have impaired cell
permeability and countered gains in potency for some of the more basic molecules. The
cytotoxicity of compounds in the cell line was monitored to preclude interference in the assay,
and no compound showed high cytotoxicity in the HEK293 cell line over the duration of the
assay.
Compounds 2, 26 and 31 were profiled for broad kinase selectivity, as determined by
inhibition of probe binding to recombinant human protein and lipid kinase domains at a
ACS Paragon Plus Environment
1
7

Page 19 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
concentration of 1 M (KINOMEscan, Eurofins; Figure 2 and Supplementary Table S1).
Compound 2 gave >70% inhibition of 26/455 wild-type and mutant kinases tested. A distinct
selectivity pattern was seen, with specific CMGC (CLK1, CLK2, CLK4, DYRK1A, DYRK1B,
HIPK4) and TK (FLT3 wild-type and mutants, KIT wild type and mutants, PDGFRB) sub-
family members showing particularly high inhibition (>90%). The more potent derivative 26
maintained a very similar selectivity pattern, with additional kinases in the CMGC (DYRK2,
HIPK1, HIPK2, HIPK3) and TK (PDGFRA) sub-families related to those inhibited by 2 showing
potent (>90%) inhibition. Overall, 54/455 wild-type and mutant kinases tested showed >70%
inhibition, consistent with the increased potency of 26. In contrast, the 2-(N-
phenyl)aminobenzimidazole 31 had a different selectivity pattern and very high selectivity with
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
>
70% inhibition of only 4/455 kinases. While two of the CMGC kinases inhibited by 2 and 26
(CLK1 and CLK4) were inhibited to a lesser extent by 31, the most potently inhibited kinase was
another CGMC family member PCTK2 (97% inhibition at 1 M test concentration). Interaction
with wild-type and mutant TK sub-family kinases was minimal for 31. The inhibitors 26 and 31
were, respectively, 9- and 100-fold selective for IRE1 over the IRE1 isoform, which has 80%
34
sequence identity in the kinase domain , in parallel LanthaScreen binding assays
(Supplementary Figure S3).
ACS Paragon Plus Environment
1
8

Journal of Medicinal Chemistry
Page 20 of 69
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
2
26
31
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
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: Binding profiles of compounds 2, 26 and 31 tested at 1 M concentrations of
inhibitors against 455 wild type and mutant human protein and lipid kinases (KINOMEscan,
Eurofins) (See Supplementary Table S1). Kinases where probe binding was inhibited by >90% at
1 M test compound concentration are listed.
To better understand the origins of the potency and high selectivity of the inhibitors, we
determined the X-ray structure of 2 bound to the cytosolic kinase-endoribonuclease (KEN)
domain of IRE1 (PDB 6hx1; Figure 3A, B). The complex 2-IRE1 crystallized as a monomer
in the unit cell. The kinase domain adopted an unusual conformation in which the C-helix was
remarkably disordered, such that electron density was not modelled for most of this important
secondary structural component. The DFG motif occupied an intermediate position between the
ACS Paragon Plus Environment
1
9

Page 21 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
canonical DFG-in and DFG-out conformations, which we term ‘DFG-up’ for convenience. Thus
F712 was displaced from the ‘in’ position that characterizes the active conformation of the
regulatory hydrophobic spine of the kinase. A related intermediate conformation of the DFG
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
motif has been observed in complexes of ligands bound to mutated Aurora A kinases35,36. In the
current structure, the residue Y628, also part of the hydrophobic spine in IRE1, was directed
down towards the DFG motif, forming hydrogen bonds to D711 (2.7 Å) and the bound ligand.
This contrasted with the upward orientation of Y628 seen when the regulatory spine is intact and
6
the IRE1 kinase domain is in the active conformation . The DFG-up, Y628-down conformation
of 2-IRE1 was accompanied by an unusual folding of the activation loop in an anti-parallel
double -strand across the face of the ATP site, forming a region of -sheet with the P-loop that
enclosed the bound ligand. The residue H723 in the activation loop appeared to anchor this
folding, with direct hydrogen bonding to E651 (2.8 Å) in the kinase domain C-terminus and a
water-mediated hydrogen bond observed between H723 and L577 in the P-loop. Folding of the
activation loop across the active site is an auto-inhibitory mechanism for some kinases. For
example, structures of imidazo[1,2-b]pyridazines and other ligands bound to the kinase MPS1
show a single strand of the activation loop folded across the active site parallel to the P-loop,
2
9
potentially blocking the interaction with substrate peptides . However, in these cases the kinase
adopts a DFG-in conformation on ligand binding and the C-helix is preserved.
ACS Paragon Plus Environment
2
0

Journal of Medicinal Chemistry
Page 22 of 69
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
A
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
C
D
E
F
G
H
ACS Paragon Plus Environment
2
1

Page 23 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Figure 3: Characterization of IRE1 inhibitor binding modes. A) Part of the crystal structure
(PDB 6hx1) of 2 (green sticks) bound to IRE1 KEN domain (grey cartoon) showing the
conformations of the DFG motif (cyan), activation loop segment (pale cyan), Y628 (blue), K599
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
(grey) and the observed part of the C-helix (red); B) Binding of 2 (green sticks) in the IRE1
ATP site (grey sticks) with key residues labelled. Hydrogen bonds are shown as black dotted
lines and selected water molecules as red spheres; C) Detail of the interactions of 2 (green sticks)
with the DFG motif and Y628 in the IRE1 ATP site (grey cartoon) with key residues labelled.
Hydrogen bonds are shown as black dotted lines and selected water molecules as red spheres; D)
Comparison of 2 (purple), 1 (blue; PDB 4z7h), Compound 16 (light blue; PDB 4u6r) and
GSK2850163 (red; PDB 4yz9) binding to IRE1 KEN domain. Structures are overlaid on the
hinge region (E643-L644-C645; lines) and the C-helix (L602-S619; cartoon) is shown in the
three structures in which it is present; E) Schematic overlay showing the relationship of the
DFG-out, DFG-up and DFG-in conformations and corresponding orientations of the observed
activation loop segments (L714-S729) for GSK2850163-IRE1 (red), 2-IRE1 (purple),
Compound 16-IRE1 (light blue) and 1-IRE1 (blue); F) Comparison of components of the N-
terminal kinase domain dimerization interface for the DFG-in, active conformation of 1-IRE1
(blue & blue-white protomers) and DFG-up, inactive conformation of 2-IRE1 (purple).
Structures are aligned on the kinase C-lobe (shown) and endonuclease C-terminal domain (not
shown). Bound ligands are rendered as sticks; G) Model of 9 (green sticks) bound to IRE1 (grey
cartoon) based on the conformation observed for 2-IRE1. Predicted hydrogen bonds are shown
as black dotted lines; H) Detail showing the movement of Y628 as positioned in 2-IRE1 (grey
cartoon) required to accommodate the modelled binding of 9 (green sticks).
ACS Paragon Plus Environment
2
2

Journal of Medicinal Chemistry
Page 24 of 69
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
Within the unusual kinase ATP-site conformation, the ligand 2 bound at the hinge region, with
the imidazo[1,2-b]pyridazine ring sandwiched between L714 in the activation loop and V586 in
the 2-strand, C-terminal to the P-loop (Figure 3B). Canonical hydrogen bonds were made from
the imidazo[1,2-b]pyridazine N1 and 8-NH to the C645 NH (3.1 Å) and C=O (3.0 Å)
functionalities, respectively. This closely resembled the binding modes described for 8-amino-
imidazo[1,2-b]pyridazines in other kinases28,29,37,38. An additional hydrogen bond (2.8 Å) was
observed between the indazole NH and Y628, presumably contributing to stabilizing the tyrosine
‘down’ conformation. The indazole occupied a hydrophobic pocket formed by the gatekeeper
residue I642, F712 of the DFG-motif, F725 from the activation loop and the catalytic lysine
K599. The 6-chloro substituent of 2 packed closely against the -sheet surface formed by L577
and G578 of the P-loop and H723 and S724 of the activation loop. The cyclopropylmethyl side
chain of 2 reached the edge of the ATP-site and was oriented to provide a vector out into solvent.
The unusual binding mode seen for 2-IRE1 was independently reproduced with a close
analogue, 33 (PDB 6hv0), in which the cyclopropylmethyl group was replaced by an isopropyl
chain (Supplementary Figure S4), supporting the likely biological relevance of the structure.
Furthermore, the observed binding mode was consistent with the structure-activity relationships
described above for the indazole-substituted 8-amino-imidazo[1,2-b]pyridazines; specifically the
requirement for a hydrogen bond donor corresponding to the indazole N(1)-H (Table 1) and the
tolerance of a wide range of substitution of the 8-amino group on the imidazo[1,2-b]pyridazine
core (Table 2). Modelling of the more potent compound 26 into the conformation of IRE1
observed in the crystal structures reproduced this binding mode, with the suggestion of an
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
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
3

Page 25 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
additional hydrogen bond from the cyclohexylamine to E651 on the surface of the protein
(Supplementary Figure S5).
The unusual DFG-up inactive conformation of the kinase domain of 2- and 33-IRE1 differed
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
considerably from other active and inactive kinase conformations reported for ligand-bound
structures of IRE1 (Figure 3C, D)4,6,22,23. We previously described the IRE1 kinase inhibitor-
RNase activator 1 that stabilized a classical DFG-in, active conformation of the kinase with the
activation loop extended away from the active site, mimicking IRE1 autophosphorylation and
6
activating the RNase function through competent dimerization of the endonuclease domain . In
contrast, a large sulfonamide inhibitor (named Compound 16), although it also bound to a DFG-
in conformation of the kinase, was found to cause a shift of the C-helix that was incompatible
2
3
with the formation of the RNase domain dimer . A Type III inhibitor of IRE1, GSK2850163,
has been described that occupied a pocket adjacent to the hinge region of the kinase and
promoted a DFG-out kinase conformation, with the activation loop directed across the active site
2
2
and the C-helix further shifted and rotated away from the active conformation . In the latter
two cases, although the active kinase conformation was disrupted the C-helix remained intact.
For 2 and 33 our data showed a striking disruption of the C-helix, as well as a distinctive
refolding of the activation loop around the bound inhibitor. This conformation is incompatible
with the back-to-back dimerization required to form an active endoribonuclease domain dimer
(Figure 3E). The dimerization interface in the kinase domain is formed by the loops (-1)-(0)
6
and (2)-(3) from each monomer . Compared to the active IRE1 dimer structure represented
by 1-IRE1, the binding of 2 and 33 shifted the position of the kinase domain relative to the
endonuclease domain such that dimerization at the kinase domain would lead to steric clashes of
the endoribonuclease domains.
ACS Paragon Plus Environment
2
4

Journal of Medicinal Chemistry
Page 26 of 69
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
We were unable to obtain a crystal structure of a 2-aminobenzimidazole-substituted analogue
bound to IRE1. Modelling of 9, the closest analogue to 2, into the structure of 2-IRE1 showed
that alignment of the imidazo[1,2-b]pyridazine cores of the two inhibitors resulted in a clash
between the 2-aminobenzimidazole substituent and residue Y628. When Y628 was allowed to
move from the ‘down’ orientation observed in 2-IRE1 to the ‘up’ conformation observed in other
IRE1 structures, this minimal change resulted in a satisfactory binding mode for the
prototypical 2-(N-phenyl)amino-benzimidazole 9 (Figure 3F). The model retained binding of the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
imidazo[1,2-b]pyridazine at the hinge region and proposed a 180 rotation of the 2-
aminobenzimidazole relative to the indazole substituent of 2, such that a new hydrogen bond was
formed between the 2-amino group and D711 of the DFG motif. The terminal phenyl group
occupied the space vacated by Y628, and packed against this residue now in its upward
conformation. The introduction of a hydrogen bond to the DFG motif was consistent with
precedent for other 2-aminobenzimidazole kinase inhibitors, where this functionality is often an
isostere for urea-based inhibitors that bind to DFG-out conformations, as we had originally
speculated in developing the series. Moreover the modelled hydrogen bond to D711 was
consistent with the observed structure-activity relationships, where methylation of the 2-amino
group or imidazole nitrogens reduced activity (Table 1). The combination of multiple kinase
conformational disruptions accompanying binding of the indazole and aminobenzimidazole
scaffolds can account for the high IRE1 kinase selectivity of 26 and 31 (Figure 2). For 26, a
discrete set of kinases in the CAMK and TK sub-families were also inhibited. For this ligand we
cannot exclude the possibility that an alternative binding mode where the indazole contributes
the donor-acceptor hinge binding motif may be accessible in some kinases39,40. Interestingly,
while the majority of residues forming the binding site of 2 and 33 to IRE1 are identical in the
ACS Paragon Plus Environment
2
5

Page 27 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
sequence of the isoform IRE1, the histidine H723 in IRE1 is replaced by a cysteine residue
3
4
(C671) in IRE (Supplementary Figure S3B). Since hydrogen bonding of the imidazole side-
chain of H723 to E651 appeared important for the unusual activation loop fold in the structures 2
and 33 bound to IRE1 we speculate that the absence of this residue may in part be involved in
the observed selectivity of the compounds for IRE1 over IRE1. The other major local
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


difference (A646 in the hinge corresponds to R594 in IRE1 ) involves a side chain directed
away from the binding site.
To confirm the functional consequences of the unusual binding modes of the inhibitors, we
investigated the effects of 26 and 31 on ER stress-induced IRE1 signaling outputs in human
cells. The effects of the compounds on IRE1 oligomerization were determined in HEK293 cells
expressing doxycycline-inducible GFP-labelled IRE1 that forms quantifiable fluorescent foci in
the ER membrane under stress conditions41 (Figure 4A). GFP-IRE1 foci produced upon
treatment with tunicamycin were dose-dependently reduced by 26 (IC = 1.88 ±0.09 M) and
5
0
3
1 (IC = 0.74 ±0.03 M), indicating inhibition of IRE1 oligomer formation. No stimulation of
50
GFP-IRE1 foci formation was observed for either compound on its own. Thus the compounds
are full inhibitors of oligomerization in cells and do not show partial activation of IRE1 through
occupation of the nucleotide-binding site.
Compounds 26 and 31 inhibited both tunicamycin- and thapsigargin-induced IRE1-
dependent splicing of XBP1-luciferase fusion mRNA in HEK293 cells (Table 3 and
Supplementary Figure 6) with equivalent potencies (IC 0.68 – 1.63 M). In parallel, inhibition
5
0
of tunicamycin-induced production of endogenous XBP1s mRNA was demonstrated in H929
myeloma cells using RT-qPCR at similar concentrations (Supplementary Figure S7). The
ACS Paragon Plus Environment
2
6

Journal of Medicinal Chemistry
Page 28 of 69
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
expression of the spliced transcription factor XBP1s following ER stress was measured by
immunofluorescent staining in H929 myeloma cells (Figure 4C). Tunicamycin-induced
expression of XBP1s protein was inhibited by 26 and 31 with similar potencies to the inhibition
of IRE1 oligomerization and RNase activities in cells. The expected downstream effect on
XBP1s-dependent transcription in H929 cells was confirmed by monitoring production of the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
mRNA coding for DNAJB9, an ER-resident molecular chaperone regulated by XBP1s (Figure
4
D). Inhibition of IRE1 autophosphorylation in H929 cells was also observed on treatment with
6 and 31 (Figure 4B and Supplementary Figure 8), although full blockade appeared to require
2
higher concentrations than the oligomerization and RNase functions. This is consistent with the
proposed model for IRE1 activation, where conformationally productive oligomerization is the
4
critical step in activating the endoribonuclease , while autophosphorylation contributes to
6
stabilizing the active oligomers and XBP1-independent signaling to JNK through the binding of
10
6
4
TRAF2 . Previous studies by ourselves and others indicate that the trans-autophosphorylation
of IRE1 proceeds through a face-to-face encounter of IRE1 protomers distinct from the back-
to-back arrangement essential for activation of the RNase function.
ACS Paragon Plus Environment
2
7

Page 29 of 69
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
A
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
B
C
D
ACS Paragon Plus Environment
2
8