Discovery of triazole aminopyrazines as a highly potent and selective series of PI3Kδ inhibitors

Article abstract of DOI:10.1016/j.bmcl.2016.11.004

A novel class of potent PI3Kδ inhibitors with >1000-fold selectivity against other class I PI3K isoforms is described. Optimization of the substituents on a triazole aminopyrazine scaffold, emerging from an in-house PI3Kα program, turned moderately selective PI3Kδ compounds into highly potent and selective PI3Kδ inhibitors. These efforts resulted in a series of aminopyrazines with PI3Kδ IC₅₀???1?nM in the enzyme assay, some of the most selective PI3Kδ inhibitors published to date, with a cell potency in a JeKo-cell assay of 20–120?nM.

Full text of DOI:10.1016/j.bmcl.2016.11.004

Accepted Manuscript
Discovery of triazole aminopyrazines as a highly potent and selective series of
PI3Kδ inhibitors
Ina Terstiege, Matthew Perry, Jens Petersen, Christian Tyrchan, Tor Svensson,
Helena Lindmark, Linda Öster
PII:
S0960-894X(16)31144-1
DOI:
http://dx.doi.org/10.1016/j.bmcl.2016.11.004
BMCL 24399
Reference:
Bioorganic & Medicinal Chemistry Letters
To appear in:
Received Date:
Revised Date:
Accepted Date:
9 September 2016
31 October 2016
1 November 2016
Please cite this article as: Terstiege, I., Perry, M., Petersen, J., Tyrchan, C., Svensson, T., Lindmark, H., Öster, L.,
Discovery of triazole aminopyrazines as a highly potent and selective series of PI3Kδ inhibitors, Bioorganic &
Medicinal Chemistry Letters (2016), doi: http://dx.doi.org/10.1016/j.bmcl.2016.11.004
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Graphical Abstract
Leave this area blank for abstract info.
Discovery of triazole aminopyrazines as a highly potent and selective series of
PI3Kδ inhibitors
Ina Terstiege, Matthew Perry, Jens Petersen, Christian Tyrchan,Tor Svensson, Helena Lindmark,
Linda Öster
Asp787
Trp760
7m
Tyr813

Bioorganic & Medicinal Chemistry Letters
Discovery of triazole aminopyrazines as a highly potent and selective series of PI3Kδ
inhibitors
Ina Terstiegea,*, Matthew Perrya, Jens Petersenb, Christian Tyrchana,Tor Svenssona, Helena Lindmark b,
Linda Österb
aRespiratory, Inflammation & Autoimmunity, Innovative Medicines and Early Development Biotech Unit, AstraZeneca, Pepparedsleden 1, Mölndal, 43183, SE
bDiscovery Sciences, Innovative Medicines and Early Development Biotech Unit, AstraZeneca, Pepparedsleden 1, Mölndal, 43183, SE
*Corresponding author.
E-mail address: Ina.Terstiege@astrazeneca.com (I. Terstiege)
ARTICLE INFO
ABSTRACT
Article history:
Received
Revised
Accepted
Available online
A novel class of potent PI3Kδ inhibitors with >1000-fold selectivity against other class I PI3K
isoforms is described. Optimisation of the substituents on a triazole aminopyrazine scaffold,
emerging from an in-house PI3K program, turned moderately selective PI3Kδ compounds into
highly potent and selective PI3Kδ inhibitors. These efforts resulted in a series of aminopyrazines
with PI3Kδ IC50 ≤1 nM in the enzyme assay, some of the most selective PI3Kδ inhibitors
published to date, with a cell potency in a JeKo-cell assay of 20 – 120 nM.
Keywords:
2009 Elsevier Ltd. All rights reserved.
Phosphoinositide 3-kinase
PI3Kδinhibitor
aminopyrazine
inflammation
The phosphoinositide 3-kinases (PI3Ks) are a family of lipid
kinases which play an important role in numerous intracellular
signaling pathways regulating cell growth, survival and
proliferation.1 The conversion of phosphatidyl-inositol (4,5)-
bisphosphate to phosphatidylinositol (3,4,5)-triphosphate by
PI3Ks results in the activation of AKT and other downstream
signaling events.2 Therefore these kinases are interesting targets
for the treatment of cancer, inflammatory and autoimmune
disorders.3
The PI3K family is divided into three classes (class I, II and
III). Efforts within the pharmaceutical industry have
predominantly been directed towards the class I PI3Ks. This class
is divided into two subgroups - class IA consisting of the PI3K,
PI3Kβ and PI3Kδ isoforms, whereas PI3K is the only known
class IB isoform. PI3K and β are ubiquitously expressed and
have been identified as potential drug targets for oncology
indications.3 In contrast, the PI3K and δ isoforms are mainly
found in the hemapoietic system.3 PI3Kδ is predominantly
expressed in leukocytes, and is central in their activation and
recruitment to sites of inflammation, making it an attractive
target for the treatment of inflammatory conditions.3-5
1
2
(GSK2292767)
(Roche)
Owing to the high degree of homology of the four class I
PI3K’s the development of isoform selective PI3K inhibitors by
rational design has proven to be challenging.3,6 Nevertheless,
several potent and highly selective inhibitors of PI3Kδ have
previously been described.2,6-11
3
4
(Idelalisib)
(AMG319)
Figure 1. Structures of selected PI3Kδ isoform selective inhibitors 1–4.

These can be divided into two classes, binding either to the
“closed” form of the protein, e.g. compound 1 and 2, or an
“open” form, like the 9H-purine-6-amines, e. g. 3 and 4 (Figure
1).2,9,10,12 The latter compounds induce a conformational change
of the ATP-binding pocket of the PI3K protein, resulting in the
formation of the so-called specificity pocket between Trp760
(PI3Kδ numbering) and Met752, which is attributed to the
increase in isoform selectivity.2,8,13-15
In this paper we describe the discovery of a novel series of
highly potent and selective PI3Kδ inhibitors emerging from a
series of aminopyrazine based PI3K inhibitors developed in
AstraZeneca (Figure 2).17,18 Optimization of these compounds,
guided by X-ray structural data, resulted in PI3Kδ inhibitors with
sub-nanomolar potency and more than a 1000-fold selectivity
over the other class I PI3K isoforms.19
5a-o are potent PI3Kδinhibitors and are more selective towards
PI3Kβ than PI3K or .19 For the amino pyrazine analogs 5a–f,
substitution on the methylene adjacent to the triazole (R2, R3 ≠ H,
Table 1) increased the selectivity over PI3K and β and, to a
lesser extent over  compared to the unsubstituted analogs (R2, R3
= H). This increase was due to loss of activity against PI3K, β
and compare 5a vs. 5b; 5c vs. 5d, e and f).
For one of these early selective compounds, morpholine-
analog 5d, approximately 20-fold selective for PI3Kδover , we
were able to obtain co-crystal structures with 5d bound both to
PI3K and δ.15 These structures revealed that the aminopyrazine
binds to the “closed” form of PI3Kδ and does not induce a
conformational shift in the ATP-binding pocket (Figure 3). The
central triazole-aminopyrazine scaffold adopts an almost planar
conformation in both isoforms. More interestingly, they show
two significantly different binding modes for the R1 substituent
of the triazole. In PI3Kδ the morpholine was positioned over
Trp760, the so-called tryptophan shelf, formed by the small
threonine residue (Thr750) lining the tryptophan.2,15 PI3Kδ is the
only isoform where this shelf is fully accessibleIn the  isoform
(as in PI3K and β) this corresponding position is blocked by a
larger charged residue (Lys802 for PI3K), resulting in a
significant rotation of the morpholine.
Figure 2. Triazole aminopyrazine scaffold.
For the morpholine analogs 5c–f neither substitution (R2, R3)
nor stereochemistry impacted PI3Kδ potency. However, for the
phenyl analogs 5g–k we observed a 6-fold increase in PI3Kδ
potency for cyclopropyl analog 5k combined with a 9-fold loss in
PI3Kβ potency, proving 5k to be the most potent and selective
PI3Kδ inhibitor in this subset of benzotriazole substituted
aminopyridines. Other aminopyrazines in Table 1, bearing
aliphatic substituents (5l–o), displayed only moderate PI3Kδ
potency and selectivity except for 5o, the benzylic analog of 5k
which showed similar potency but slightly lower isoform
selectivity.
The X-ray structure of the most potent compound 5k in
complex with PI3Kδ(Figure 4) showed an almost planar
conformation of the central triazole aminopyrazine. The phenyl-
substituent was directed towards the tryptophan shelf, probably
interacting with the protein via a hydrophobic interaction and not
an edge-to-face -stacking; indicated by the fact that varying the
electronics and the substitution pattern of this phenyl ring with
substituents in para and meta had remarkably little effect on
potency and selectivity (data not shown).22
In an internal program aimed to identify PI3K inhibitors a
small number of aminopyrazines with significant δ-activity and
modest selectivity versus PI3K (data not shown) were
discovered. After an initial exploration of the hits, we identified
the triazole aminopyrazine scaffold, methylated at the triazole
nitrogen distal to the pyrazine (Figure 2), as a promising lead
structure for the development of selective PI3Kδ inhibitors. Our
plan was to improve the δ-activity and isoform selectivity by
optimizing the R1 and R2 substituents of this scaffold.
Initially, we directed our attention towards exploration of the
R1 substituent. Based on modeling studies this residue binds to
the solvent exposed region of the active site.17,18
a)
b), c)
A
B
C
d)
e)
E
D
Scheme 1. (a) benzotriazole, neat, 140 °C, 4 h, 80%; (b) NBS, DMF, r.t., 1
h, 73%; (c) Zn(CN)2, Pd2(dba)3, Xantphos, TMEDA, DMF, 100 °C, 3 h,
71%; (d) methylhydrazine, EtOH, 50 °C, 16 h, 75%; (e) carboxylic acid,
TBTU, 4-methylmorpholine, NMP, r.t., overnight, chromatography, 5-55%;
A first set of analogs to the PI3K lead compounds was
synthesized according to Scheme 1: Reacting benzotriazole neat
with chloroamino pyrazine A resulted in formation of the
substituted amino pyrazine B.20 Subsequent bromination and
palladium-catalyzed cyanation yielded the nitrile C.17 Heating
intermediate C with methylhydrazine gave the imidohydrazide D,
which was converted to the triazole E by coupling with the
appropriate carboxylic acid, followed by ring-closure.21
Figure 3. Crystal structure of 5d in complex with mPI3Kδ(yellow)
(resolution 2.6 Å) and hPI3K (green) (resolution 2.8 Å), view from solvent
exposed side of the ATP-binding pocket towards hinge-region and affinity
pocket, surface of PI3Kδ is shown in grey.
Table 1 provides an overview of the most significant findings:
As the data of the enzyme inhibition assay indicates, compounds

Table 1. Potencya) (pIC50) of compounds 5a-o in enzyme inhibition assays.19
5a - o
5
a
R1
R2
R3
PI3Kδ
PI3K
PI3K
PI3Kβ
H
H
7.6
6.5
6.4
6.8
6.1
CH3
H
CH3
H
b
c
7.7
7.2
6.5
6.7
5.2
6.2
6.4
CH3
H
H
d
e
7.1
5.8
5.9
5.4
CH3
-CH2-
H
7.0
< 6.0
6.1
6.0
5.5
-CH2-
H
f
6.9
6.0
4.6
g
h
i
7.6
6.5
6.8
6.2
CH3
H
H
n.d. b)
7.4
n.d. b)
6.3
n.d.b)
6.3
n.d. b)
5.8
CH3
CH3
-CH2-
-CH2-
-CH2-
-CH2-
-CH2-
CH3
-CH2-
-CH2-
-CH2-
-CH2-
-CH2-
n.d. b)
8.5
n.d. b)
6.3
n.d. b)
6.6
n.d. b)
5.2
j
k
l
6.9
6.3
6.8
5.9
m
n
o
6.5
6.2
6.3
4.7
6.8
5.0
5.7
< 4.5
5.6
8.3
6.4
6.7
a) geometric mean of at least three independent experiments, b) not tested, low solubility
According to the calculated torsion profile (Diagram 1), the
observed conformation of 5k with a torsion angle of 93° between
the R1 substituent 1-phenylcyclopropyl and the methyltriazole is
lowest in energy compared to the profiles of the analogs 5ka and
5kb. The methyl-isomer 5ka displays only slightly higher energy
at this torsion angle whereas the hydrogen-substituted analog 5kb
has to pay a significant energy penalty to adopt this favorable
conformation. In the enzyme inhibition assay isomer 5k showed
similar PI3Kδ potency as 5ka but was 5-fold more potent against
PI3Kδ than 5kb. This observation supported the use of the
triazole aminopyrazine scaffold, methylated at the triazole

nitrogen distal to the pyrazine (Figure 2), as basis for further
exploration of the 1-phenylcyclopropyl analogs.
nitrogen heterocycle and NaH to yield the desired target
compound M. Alternatively, compound L could be coupled
with the appropriate boronic acid/ester or stannane under Pd-
catalysis for the synthesis of M.23
As shown in Table 2, the position of the heterocyclic
nitrogens had a marked effect on potency. Initially, we speculated
that one nitrogen of benzotriazole 5k (Y = N) could be involved
in an intramolecular H-bond with the aminopyrazine NH2.24
However, a torsion angle of ca. 30° of the bound compound 5k
(Figure 4) indicated that this intramolecular hydrogen bond was
probably not present in the protein-ligand complex.
Several groups have reported interactions of PI3Kδ inhibitors
with the tyrosine gatekeeper and the aspartate residues of the
DFG motif via a hydrogen bond network involving water.1,2,8,9
We postulated that such an interaction could explain the overall
higher potency of 5k (Z = N) compared to compound 6a (Z =
CH). The crystal water in the X-ray structure of 5k was within H-
bond distance of Tyr813 and Asp787 but was too remote to
interact with the inhibitor (distance 3.8 Å).
Figure 4. Crystal structure of 5k in complex with mPI3Kδ(resolution 2.8
Å), Trp760 highlighted in green, torsion angles between triazole
aminopyrazine scaffold and substituents annotated, binding pocket surface is
shown in grey.
With the potent triazole 5k as starting point we focused next
on the exploration of the right-hand side of our scaffold, the
substituent occupying the so-called affinity pocket (R2 in Figure
2). Although this pocket is highly homologous across all four
PI3K isoforms it has been demonstrated to play an important role
when optimizing compounds towards PI3Kδ selectivity.2,7,9
Diagram 1. Calculated torsion profile for 5k (in green), 5ka (in red) and
5kb (in blue) and PI3Kδ potency in enzyme inhibition assay.19
Scheme 2. (a) Zn(CN)2, Pd2(dba)3, Xantphos, TMEDA, 100 °C, 2 h, quant.;
(b) CH3NHNH2, EtOH, 14 hr, r.t., 33%; (c) 1-phenylcyclopropanecarboxylic
acid, SOCl2, then N-methylmorpholine, J, DMF, r.t. 2 h,120 °C, 1 h 77%; (d)
NBS, CH3CN, 0 °C to r.t., 4 h, 79%; (e) pyrazole/indazole, NaH, NMP, r.t. to
100 °C, 16 h or boronic acid/boronic ester/stannane, Pd(dppf)Cl2-DCM,
NaHCO3, dioxane, water, heat.23
Comparing compounds 6b and 6c revealed an even larger
potency difference, although both compounds could potentially
interact with a bridging water. In addition both compounds
should show an increased torsion angle between the
aminopyrazine and the affinity pocket substituent due to a steric
clash of the hydrogen substituent (Y = CH) and the pyrazine
NH2. Assuming that a torsion angle of ca. 30° between amino
pyrazine and affinity pocket substituent R2, as observed in the X-
ray of 5k (Figure 4), is relevant for binding, both compound 6b
and 6c should have similar potencies. We analyzed the
conformations of compounds 5k and 6a–c by QM.25 These
calculations showed a large energy penalty for 6b to adopt a
conformation in line with 5k, 6a and 6c (greater than 3 kcal/mol
for 6b, 0.02 kcal/mol for 5k and 6a, 0.18 kcal/mol for 6c).
Although these compounds were potent and selective PI3Kδ
inhibitors, none of the changes made in 6a–c resulted in a
significant improvement over compound 5k.
Initially, we looked at close analogs of the benzotriazole
substituted aminopyrazine 5k (Table 2). The compounds 6a–c
were synthesized according to Scheme 2: Bromo-
aminopyrazine H was converted into the nitrile I via
palladium-catalyzed cyanation. Treatment of
I
with
methylhydrazine yielded J. Formation of the methyl-triazole
K was conducted in a three step, one-pot fashion: formation
of 1-phenylcyclopropanecarbonyl chloride was followed by
reaction of the acid chloride with J and ring closure at
elevated temperature. Bromination of K with NBS gave
bromide L which could be treated with the appropriate

Table 2. Potencya) (pIC50) of compounds 5k, 6a–c in enzyme inhibition
first compound synthesized with low nanomolar potency for
PI3Kδ combined with a very high selectivity towards the other
isoforms (~500-fold against PI3K and  and ~3000-fold against
PI3Kβ). Compared to the leading compound, benzotriazole 5k,
the selectivity profile of the analog 7e represented a step change.
The reduction of logD by two log units (measured logD 4.3 and
2.1 for 5k and 7e respectively) resulted in a potency drop of 7e
against all isoforms except against PI3Kδ. The reason for this
unexpected result remains unclear. However, the H-bond network
of a crystal water between the inhibitor and Asp787, observed in
the X-ray structure of 7e (Figure 5a) but not of 5k, might be an
explanation. The reduced size of the pyrazole substituent of 7e,
compared to the larger benzotriazole in 5k, allows it to adopt a
favorable conformation for the formation of this H-bond, without
sterical clash with the pyrazine scaffold.
The non-symmetrical pyrazole analog 7f, on the other hand,
showed reduced potency against all isoforms. In comparison to
the close analog 7e, the potency of dimethyl pyrazole 7g gives
another indication of the importance of the torsion angle between
aminopyrazine and the affinity pocket substituent. The two
methyl substituents force the pyrazole and pyrazine rings into a
perpendicular conformation which cannot be accommodated in
the binding pocket, resulting in loss of activity. Overall,
substituents reaching into the affinity pocket are tolerated if they
don’t, as in 7g, prevent the compound to adopt the binding
conformation. Oxadiazole 7h, for example, bearing a large
lipophilic t-Bu group which reaches deeper into the affinity
pocket, was a selective sub-nanomolar PI3Kδ inhibitor and one
of the most potent compounds synthesized.
Compound 7i, the amino formyl substituted analog of 7e, was
designed to increase potential interactions with water in the
affinity pocket. The crystal structure of 7i (Figure 5b) confirmed
the presence of a conserved crystal water molecule within
hydrogen bond distance to a pyrazole nitrogen atom, as well as
an interaction with the amide functionality of 7i, which is within
hydrogen bond distance of Lys779 and Asp911 (2.7 Å, not
shown). However, this water is too remote for an interaction with
the protein and 7i shows similar potency and selectivity as the
parent compound. The close analog sulfonamide 7j, designed to
pick up an interaction with Lys779, displayed increased potency
against all isoforms compared to 7i at the expense of increased
lipophilicity. The crystal structure of 7j (Figure 5a) shows,
similarly to the structure of 7e and 7i, a hydrogen bond
interaction between a pyrazole nitrogen atom and a crystal water,
and in addition a hydrogen bond between the sulfone and
Lys779.
assays.19
5k, 6a-c
V
X
Y
Z
PI3Kδ
PI3Kβ
PI3K PI3K
C
N
N
N
5k
6a
6b
6c
8.5
6.3
5.6
5.1
6.4
6.6
5.4
4.9
6.7
5.2
C
C
N
N
N
C
N
C
C
C
N
N
7.8
7.3
8.7
>4.7
>4.5
5.4
a) geometric mean of at least three independent experiments
We then decided to explore a larger set of substituents in the
affinity pocket (Table 3): Close to the gatekeeper this pocket is
narrow and lipophilic, however polar interactions with e.g.
Lys779 are possible in the deeper end of the pocket. We
hypothesized that an interaction of the inhibitor with Tyr813 and
Asp787 mediated by crystal bound water is important to stabilize
the protein conformation.1,2,8,9 Thus, for our target library, we
selected small aromatic/heteroaromatic substituents with the
possibility to interact with the cluster of water molecules in order
to improve binding affinity. The compounds were synthesized
according to Scheme 2.23
The lack of potency for compound 7a showed that a
substituent positioned in the affinity pocket is necessary for
binding of the amino pyrazines, in contrast to what is observed
for PI3Kδ inhibitors binding to the “open” form of the protein
(e.g. compounds 3 and 4). Aminopyrazine 7b, with a nitrile-
substituent reaching into the affinity pocket, started to display
activity in the biochemical assays. The fluoro-pyridine
substituted analog 7c gained an additional log-unit in PI3Kδ
potency. The matched pair aminopyridine 7d showed further
improvement of potency compared to 7c. Compound 7e, with a
symmetrical pyrazole substituent in the affinity pocket, was the
Figure 5. a) Crystal structure of 7e (blue, resolution 2.3Å) and 7j (red, resolution 2.9 Å) in complex with mPI3Kδandcrystal water, surface rendering from the
PI3Kδ-7e complex shown in grey; b) Crystal structure of 7i (purple, resolution 2.6 Å) with crystal water, 7k (yellow, resolution 2.3 Å) with crystal water, 7l
(green, resolution 2.6 Å), and 7m (orange, resolution 2.8 Å) in complex with mPI3Kδsurface rendering from the PI3Kδ-7i complex shown in grey.

Table 3. Potencya) (pIC50) of compounds 7a-n in enzyme inhibition assays, measured logD, LLE.19
logD
(measured)
7
a
R
PI3Kδ
PI3Kβ
LLEb)
PI3K
PI3K
1.8
2.2
2.5
2.0
2.1
3.0
2.9
4.7
<4.5
<4.5
<4.5
4.3
4.9
6.0
6.2
4.2
b
c
d
e
f
6.5
7.4
8.0
8.3
7.2
<4.6
4.8
5.2
5.6
5.2
<4.5
5.0
<4.5
<4.5
<4.6
4.8
5.4
5.6
<4.6
<4.5
2.3
3.6
3.2
5.6
g
5.5
9.2
<4.7
6.6
<4.5
7.1
<4.5
6.0
h
2.2
2.8
2.8
5.9
6.3
5.9
i
j
8.1
9.1
8.7
5.4
6.2
6.0
5.4
6.4
6.3
<4.5
5.6
k
4.9
2.8
2.2
4.4
7.0
l
7.2
9.2
5.2
6.0
4.6
5.9
<4.5
5.0
m
2.2
6.9
n
9.1
5.8
5.4
4.7
a) geometric mean of at least three independent experiments, b) LLE = pIC50 (PI3Kδ) – logD (measured)
The primary amide 7k, a chemically more stable isomer of 7i,
is a more potent PI3Kδ-inhibitor with similar selectivity, despite
the lack of a nitrogen acceptor for a potential interaction with
bridging water. In the X-ray structure of aminopyrazine 7k, a
crystal water in a hydrogen bond network is observed, but this
time with Asp787 and Tyr 813 of the protein and not with the

inhibitor (Figure 5b). The distance of 4.2 Å between the water
and the amide function of 7k is too far for a hydrogen bond
interaction. Moving the amide substituent to the 3-position of the
pyrazole, as in regioisomer 7l, resulted in less favorable binding
interactions and therefore loss of potency and selectivity. The
crystal structure of 7l indicates that the amide is directly
interacting with Tyr 813 (Figure 5b).
Finally, we explored the amide substituted phenyl analogs 7m
and 7n. Owing to the different geometry of a six-membered ring
the amide substituent of 7m and 7n are directed into the affinity
pocket along a different vector. The interactions of these
inhibitors with the protein seem to be especially advantageous for
the PI3Kδ isoform resulting in sub-nanomolar -inhibitors with
remarkably selectivity against all isoforms. According to the
crystal structure for 7m the amide substituent binds in a position
between the two extremes observed earlier (7i and 7k vs.7l)
(Figure 5b) and is interacting with Lys 779. The fluorine
substituted analog 7n is equipotent with 7m against PI3Kδ but
displays even higher selectivity. This compound shows more
than 1000-fold selectivity against the other PI3K isoforms (4000-
fold against PI3K, 24000-fold against PI3Kβ and 1600-fold
against PI3K) and is one of the most selective PI3Kδ inhibitors
published to date.9-11
(Table 4).28 In addition, compounds 7h and 7j proved to be >100-
fold more potent than in a RAW264 cell assay, which was used
as a readout for PI3K selectivity.19,29-31 The sulfonamide 7j
shows, in contrast to the analog 7e, high clearance. This
metabolic instability might be caused by the dimethylamine
substituent. The potent and highly soluble compound 7m
displays, despite a low cell permeability and a high efflux, good
cell potency. By introduction of a F-substituent we were able to
increase permeability for analog 7n, probably due to an
intermolecular interaction between the primary amide and the F-
substituent. The reason for the reduced efflux of 7n remains
unclear. Surprisingly, the increased permeability did not result in
increased cell potency for 7n. Tested in vivo this compound
showed reasonable pharmacokinetics with good bioavailability
and moderate in vivo half-life. (Table 5).
Diagram 2. Kinase profile of 7n (Thermo Fisher, 1µM, 350 kinases).
Compound 7n was tested in a kinase panel (Thermo Fisher)
against 350 kinases at a concentration of 1M. Only five kinases
showed more than 50% inhibition of which three showed more
than 70% inhibition.26 This compound proved to be clean in a
secondary pharmacology panel of 117 targets with only three hits
with less than 1 M binding.28
When testing a selection of compounds in a JeKo 1-cell assay
the aminopyrazines showed activity in the nanomolar range
Table 4. Additional data for compounds 7e, h, j, m and o.19
PI3Kδ
JeKo
IC50
RAW
264
IC50
sol.
Hu
PPB
HLM
Clint
RH
Clint
hERG PSA
11µM
Inhib
Caco2
Efflux
(B to A) /
(A to B)
Caco2
A to B
(1E-6 cm/s)
a)
a)
IC50
7
e
R
(µM)
(µM)
(µM)
(µM) (% free) (µL/min/mg) (µL/min/1E6)
(%)
0.005
0.11
n.d.
37
39
<3
5
9
99
6.6
24
5
0.0006
0.0008
0.02
0.09
3.7 b)
3
3.1
20
12
29
105
0.42
h
j
> 10 c)
11
7.8
191
157
11
125
19
0.66
0.0006
0.0009
0.05
0.12
n.d.
n.d.
177
32
22
20
< 3
< 3
2
8
118
118
2.0
13
> 22
2.1
m
n
< 1
24
a)geometric mean of at least three independent experiments, b) n = 2, c) n = 1.

Table 5. Rat PK data for compound 7n.
Whittington, D. A.; Whoriskey, J.; Yu, G.; Zalameda, L.; Zhang, D.;
Metz, D. P. J. Med. Chem. 2015, 58, 480.
F (po) Vss (iv)
t1/2 (iv)
(h)
CL (iv)
(mL/min/kg)
12
9.
Down, K.; Amour, A.; Baldwin, I. R.; Cooper, A. W. J.; Deakin, A. M.;
Felton, L. M.; Guntrip, S. B.; Hardy, C.; Harrison, Z. A.; Jones, K. L.;
Jones, P.; Keeling, S. E.; Le, J.; Livia, S.; Lucas, F.; Lunniss, C. J.; Parr,
N. J.; Robinson, E.; Rowland, P.; Smith, S.; Thomas, D. A.; Vitulli, G.;
Washio, Y.; Hamblin, J. N. J. Med. Chem. 2015, 58, 7381.
(%)
41
(L/kg)
2.8
7n
3.4
10. Patel, L.; Chandrasekhar, J.; Evarts, J.; Haran, A. C.; Ip, C.; Kaplan, J.
A.; Kim, M.; Koditek, D.; Lad, L.; Lepist, E.-I.; McGrath, M. E.;
Novikov, N.; Perreault, S.; Puri, K. D.; Somoza, J. R.; Steiner, B. H.;
Stevens, K. L.; Therrien, J.; Treiberg, J.; Villaseñor, A. G.; Yeung, A.;
Phillips, G. J. Med. Chem. 2016, 59, 3532.
11. Shin, Y.; Suchomel, J.; Cardozo, M.; Duquette, J.; He, X.; Henne, K.;
Hu, Y.-L.; Kelly, R. C.; McCarter, J.; McGee, L. R.; Medina, J. C.;
Metz, D.; San Miguel, T.; Mohn, D.; Tran, T.; Vissinga, C.; Wong, S.;
Wannberg, S.; Whittington, D. A.; Whoriskey, J.; Yu, G.; Zalameda, L.;
Zhang, X.; Cushing, T. D. J. Med. Chem. 2016, 59, 431.
12. Fowler, K. W.; Huang, D.; Kesicki, E. A.; Ooi, H. C.; Oliver, A. R.;
Ruan, F.; Treiberg, J.; Quinazolinones as inhibitors of human
phosphatidylinositol 3-kinase delta. WO2005/113556, December 1,
2005.
13. Berndt, A.; Miller, S.; Williams, O.; Le, D. D.; Houseman, B. T.;
Pacold, J. I.; Gorrec, F.; Hon, W.-C.; Ren, P.; Liu, Y.; Rommel, C.;
Gaillard, P.; Ruckle, T.; Schwarz, M. K.; Shokat, K. M.; Shaw, J. P.;
Williams, R. L. Nat. Chem. Biol. 2010, 6, 117.
In conclusion, we have described how the triazole
aminopyrazine based inhibitors, originating from a series of
compounds targeting PI3K could be converted into a series
of highly selective and potent PI3Kδ inhibitors by careful
optimization of substituents supported by structural
information. Essential for achieving the high degree of
isoform selectivity was the optimization of the inhibitor
interaction with the tryptophan shelf as well as with the
affinity pocket in PI3Kδ. This program clearly demonstrates
the value of crystal structures to guide hypothesis generation
and aids to our SAR understanding, but also highlights the
remaining difficulties in predicting and optimizing potency
and selectivity in the affinity pocket of the class I PI3Ks, fully
homologous between the four isoforms.
14. Somoza, J. R.; Koditek, D.; Villaseñor, A. G.; Novikov, N.; Wong, M.
H.; Liclican, A.; Xing, W.; Lagpacan, L.; Wang, R.; Schultz, B. E.;
Papalia, G. A.; Samuel, D.; Lad, L.; McGrath, M. E. J. Biol. Chem.
2015, 290, 8439.
15. Amino acid numbering refers to mPI3Kδ if not stated otherwise, PI3K
numbering refers to hPI3K.
All X-ray complex structures have been deposited to the
protein data-bank with accession codes 5T23 (5d, hPI3K),
5T27 (5d, mPI3Kδ) , 5T28 (5k), 5T2B (7e), 5T2D (7j), 5T2G
(7i), 5T2I (7k), 5T2L (7l) and 5T2M (7m) respectively.
Determination of PI3K-5d crystal structure: Human PI3K
(residues 144-1102) with a c-terminal 6-His tag was expressed and
purified as previously described.16 Compound 5d was solubilised in
DMSO to 50 mM and added to the protein (at 12 mg/mL in 20 mM Tris
pH 7.2, 50 mM AmSO4, 1% betaine, 1% ethylene glycol, 0.02%
CHAPS and 2 mM TCEP) to a final concentration of 2 mM. The
mixture was left to incubate for 1 hour and then centrifuged at 13.000
rpm for 30 minutes. Hanging drops were set up with a one to one ratio
of protein solution and well solution containing 20% PEG3350, 175
mM AmSO4, 10 mM DTT and 100 mM Hepes (pH 7.4-7.9) at 20°C.
Several rounds of optimization and streak seeding improved the quality
and size of the crystals. Prior to flash freezing, the crystals were quickly
dipped in a cryo solution consisting of 15% glycerol, 20% PEG3350,
175 mM AmSO4, 10 mM DTT and 100 mM Hepes pH 7.5. X-ray
diffraction data were collected at beam line I04-1 at the Diamond Light
Source, UK. Determination of the mPI3Kδ crystal structures: The
expression and purification of mouse PI3Kδ followed the same
procedure as previously described.13 Compound 5d, 5k, 7e, 7i, 7j, 7k, 7l
and 7m were all solubilised to 50 mM in DMSO. In each case, 0.4 μL of
compound stock was mixed with 3 μL of 5% w/v n-dodecyl-β-D-
maltoside (DDM) from Hampton Research® prior to the addition of 25
μL mPI3Kδ (at 5 mg/mL in 20 mM Tris pH 7.2, 50 mM AmSO4, 1%
ethylene glycol, 1% betaine, 0.02% CHAPS and 5 mM DTT). The
mixture was left to incubate for 1 hour and then centrifuged at 13.000
rpm for 30 minutes. Hanging drops were set up with a one to one ratio
of protein solution and well solution around condition G6 of the
Morpheus screen (Molecular Dimensions®): 50% ethylene glycol/PEG
8000, 0.1 M carboxylic acids mix, 0.1 M Buffer system 2 pH 7.5. Streak
seeding improved the size and quality of the crystals. For data collection
crystals were flash frozen in liquid nitrogen without the addition of any
cryo protection. All data was collected at the European Synchrotron
Radiation Facility, France; compound 5d, 5k, 7e, 7i, 7k and 7l at beam
line ID29 and compound 7j and 7m at beam line ID23-1
Complex.structures of both PI3K and mPI3Kδ were determined by
Molecular replacement using the program “Molrep” using internal
structures as search models. Manual fitting and inspection of structures
were carried out in the program coot and refinement carried out using
the program Buster.
Acknowledgments
The authors would like to thank the AZ PIKC-team,
especially Vanessa Gonnot; the chemists at BioDuro and
Pharmaron for their synthetic contributions; Johan Carlsson
for providing the biochemical data and the Separation Science
and Analytical Team at AZ Gothenburg for purification of
final compounds and analytical support.
References and notes
1.
2.
Furet, P.; Guagnano, V.; Fairhurst, R. A.; Imbach-Weese, P.; Bruce, I.;
Knapp, M.; Fritsch, C.; Blasco, F.; Blanz, J.; Aichholz, R.; Hamon, J.;
Fabbro, D.; Caravatti, G. Bioorg. Med. Chem. Lett. 2013, 23, 3741.
Sutherlin, D.; Baker, S.; Bisconte, A.; Blaney, P. M.; Brown, A.; Chan,
B. K.; Chantry, D.; Castanedo, G.; DePledge, P.; Goldsmith, P.;
Goldstein, D. M.; Hancox, T.; Kaur, J.; Knowles, D.; Kondru, R.;
Lesnick, J.; Lucas, M. C.; Lewis, C.; Murray, J.; Nadin, A. J.;
Nonomiya, J.; Pang, J.; Pegg, N.; Price, S.; Reif, K.; Safina, B. S.;
Salphati, L.; Staben, S.; Seward, E. M.; Shuttleworth, S.; Sohal, S.;
Sweeney, Z.; Ultsch, M.; Waszkowycz, B.; Wei, B. Bioorg. Med. Chem.
Lett. 2012 22, 4296.
3.
4.
Ameriks, M. K.; Venable, J. D. Curr. Top. Med. Chem. 2009, 9, 738.
Yap, T.; Bjerke, L.; Clarke, P. A.; Workman, P. Curr. Opin. Pharmacol.
2015, 23, 98-107.
5.
Rowan, W. C.; Smith, J. L.; Affleck, K.; Amour, A. Biochem. Soc.
Trans. 2012, 40, 240.
6.
7.
Jeong, Y.; Kwon, D.; Hong, S. Future Med. Chem. 2014, 6, 737.
Safina, B. S.; Baker, S.; Baumgardner, M.; Blaney, P. M.; Chan, B. K.;
Chen, Y.-H.; Cartwright, M. W.; Castanedo, G.; Chabot, C.;
Cheguillaume, A. J.; Goldsmith, P.; Goldstein, D. M.; Goyal, B.;
Hancox, T.; Handa, R. K.; Iyer, P. S.; Kaur, J.; Kondru, R.; Kenny, J.
R.; Krintel, S. L.; Li, J.; Lesnick, J.; Lucas, M. C.; Lewis, C.; Mukadam,
S.; Murray, J.; Nadin, A. J.; Nonomiya, J.; Padilla, F.; Palmer, W. S.;
Pang, J.; Pegg, N.; Price, S.; Reif, K.; Salphati, L.; Savy, P. A.; Seward,
E. M.; Shuttleworth, S.; Sohal, S.; Sweeny, Z. K.; Tay, S.;
Tivitmahaisoon, P.; Waszkowycz, B.; Wei, B.; Yue, Q.; Zhang, C.;
Sutherlin, D. P. J. Med. Chem. 2012, 55, 5887.
16. Walker, E. H.; Pacold, M. E.; Perisic, O.; Stephens, L.; Hawkins, P. T.;
Wymann, M. P.; Williams, R. L. Mol. Cell. 2000, 6, 909.
17. Barlaam, B.; Cosulich, S.; Fitzek, M.; Green, S.; Harris, C. S.; Hudson,
K.; Lambert-van der Brempt, C.; Ouvry, G.; Page, K.; Ruston, L.;
Ward, L.; Delouvrié, B. Bioorg. Med. Chem. Lett. 2015, 25, 2679.
18. Barlaam, B.; Cosulich, S.; Delouvrié, B.; Ellston, R.; Fitzek, M.;
Germain, H.; Green, S.; Hancox, U.; Harris, C. S.; Hudson, K.;
Lambert-van der Brempt, C; Lebraud, H.; Magnien, F.; Lamorlette, M.;
Le Griffon, A.; Morgentin, R.; Ouvry, G.; Page, K.; Pasquet, G.;
8.
Cushing, T. D.; Hao, X.; Shin, Y.; Andrews, K.; Brown, M.; Cardozo,
M.; Chen, Y.; Duquette, J.; Fisher, B.; Gonzalez-Lopez de Turiso, F.;
He, X.; Henne, K. R.; Hu, Y.-L.; Hungate, R.; Johnson, M. G.; Kelly, R.
C.; Lucas, B.; McCarter, J. D.; McGee, L. R.; Medina, J. C.; San
Miguel, T.; Mohn, D.; Pattaropong, V.; Pettus, L. H.; Reichelt, A.;
Rzasa, R. M.; Seganish, J.; Tasker, A. S.; Wahl, R. C.; Wannberg, S.;

Polanska, U.; Ruston, L.; Saleh, T.; Vautier, M.; Ward, L. Bioorg. Med.
Chem. Lett. 2015, 25, 5155.
23. 6a, b, 7k, l were synthesized via treatment of L with N-heterocycle and
NaH; 6c was synthesized from L via Stille coupling with Pd(PPh3)4;7c –
g, j, m, n were synthesized via cross-coupling of L with the appropriate
boronic acid/ester; 7e,f were synthesized from L via cross-coupling
with the appropriate THP-protected boronic acid followed by
deprotection with HCl, 7i was synthesized from 7e by subsequent
treatment with phenyl chloroformiate and ammonia; 7h was synthesized
from L via the carboxylic acid in analogy to US2014/0206700 A1; the
Suzuki couplings were in some cases performed alternatively with
Pd(dbpf)Cl2 as catalyst, K2CO3 as base and acetonitrile or DME as
solvent.
19. Data is the geometric mean of at least three independent experiments,
Potency of the aminopyrazines was routinely measured using:
Enzymatic activity assay for rec human PI3K, β, δ and : The
activity of rec human PI3K ((aa144-1102)-6His) and PI3K, βδ (6-
His(p110– p85was determined by measuring the ADP level after
phosphorylation of DiC8-PIP2 using a commercially available ADP-
GloTM kit from Promega. The assay was carried out in white low
volume 384 well plates in a final volume of 14 µL at r.t. The assay
conditions contained the following: 50 mM Tris buffer pH 7.4, 2.1 mM
DTT, 3 mM MgCl2, 0.05% CHAPS, 20 µM ATP, 80 µM DiC8-PIP2 and
1.2 nM PI3Kβ, or0.6 nM PI3Kδ. Potential inhibitors were made up
in DMSO and then diluted in the assay to give a final concentration of
not exceeding 1% (v/v) DMSO. A 10-point half-log dilution series of
the inhibitors (highest concentration typically 0.1 μM for  and 33 μM
for ,  and ) was tested and the pIC50 determined using a 4-paramater
logistic equation in a non-linear curve fitting routine. Routinely,
inhibitors were pre-incubated with 3 μL of enzyme for 15 min prior to
the addition of 2 μL substrate mixture for a further 60 min enzyme
reaction. The phosphorylation was stop with the addition of 3 μL ADP-
GloTM reagent (stop solution) followed by a 40 min incubation. Prior to
detection 6 μL of ADP-GloTM Kinase Detection Reagent was added and
the plates were read in a micro plate reader using a Luminescence filter.
All additions were followed by a short centrifugation step.
24. Kenny, P. W.; Montanari, C. A.; Prokopcyk, I. M.; Ribeiro, J. F. R.;
Sortori, G. R. J. Med. Chem. 2016, 59, 4278.
25. Torsion profiles were calculated within Jaguar (Maestro 9.8) using
relaxed coordinate scan (10 degrees) with density functional theory
MO6-2x and the 6-31*+ basis set. For single point energy calculations
the same MO6-2x functional and 6-31*+ basis set was used
(Schrödinger Release 2014-2: Maestro, version 9.8, Schrödinger, LLC,
New York, NY, 2014).
26. 7n was tested against these kinases to give the following IC50 values:
DYRK1A (Dual specificity tyrosine-phosphorylation-regulated kinase
1A) IC50
= 0.14 µM, DYRK1B (Dual specificity tyrosine-
phosphorylation-regulated kinase 1B) IC50 = 0.33 µM, CSNK1G3
(casein kinase 1 gamma 3) IC50 = 0.46 µM; all three kinases were more
than 100-fold less inhibited than PI3Kδ.
27. Hits in the secondary pharmacology panel with IC50 < 1 µM: endothelin
converting enzyme 1, IC50 > 0.3 µM; 5-hydroxytryptamine receptor 1A,
IC50 = 0.6 µM; delta-2-opoid receptor, IC50 > 0.9 µM.
28. Meadows, S. A.; Kashishian, A.; Johnson, D.; Ulrich, R. G.; Miller, L.
L.; Lannutti, B. J. Blood 2011, 118, 3730.
29. Winkler, D. G.; Faia, K. L.; DiNitto, J. P.; Ali, J. A.; White, K. F.;
Brophy, E. E.; Pink, M. M.; Proctor, J. L.; Lussier, J.; Martin, C. M.;
Hoyt, J. G.; Tillotson, B.; Murphy, E. L.; Lim, A. R.; Thomas, B. D.;
MacDougall, J. R.; Ren, P.; Liu, Y.; Li, L. S.; Jessen, K. A.; Fritz, C. C.;
Dunbar, J. L.; Porter, J. R.; Rommel, C.; Palombella, V. J.; Changelian,
P. S.; Kutok, J. L. Chem. Biol. 2013, 20, 1364.
30. Vakkalanka, S. K. V. S.; Bhavar, P. K.; Viswanadha, S.; Babu, G.; Dual
selective PI3 delta and gamma kinase inhibitors. US2014/0364447,
December 11, 2014.
A
10-point half-log dilution series of the inhibitors (highest
concentration typically 0.1 μM for δ and 33 μM for  and ) was
tested and raw data output was analyzed in Screener software (Genedata
AG) and pIC50 values calculated using a four parameter logistic fit.
Variability of the enzymatic activity assays for a reference compound
(pIC50 (mean ± SD)): PI3Kδ 7.342 ± 0.093 (n = 161), PI3K 7.424 ±
0.139 (n = 97), PI3K 6.798 ± 0.081 (n = 136), PI3Kβ 7.204 ± 0.086 (n
= 127).
Inhibition of cellular activity in JeKo-1 cells: PI3Kδ inhibitors were
identified by measuring the phosphorylation of AKT (at Ser473) using
TR-FRET technology (Phospho-AKT (Ser473) kit, Cisbio) in JeKo-1
cells - Lymphoblast B cell line (ATCC). Cryo preserved cells were
thawed rapidly in RPMI 1640 medium (10% FBS, 2 mM Glutamine) at
37 ˚C. Cells were plated in white low volume 384 well plates at a
density of 1.25E6 cells/mL in assay buffer (RPMI 1640 medium, 2 mM
Glutamine). Potential inhibitors were made up in DMSO and then
diluted in the assay to give a final concentration of not exceeding 1%
(v/v) DMSO. Inhibitors were incubated with 8 μL of cells for 60 min at
30 °C prior stimulated with 2 µL Anti-IgM (250 ng/mL) in assay buffer.
After an incubation of 10 min at 37 °C the cells were lysed for 30 min at
r.t. on a plate shaker at 500 rpm. A 5 µL mix of d2/cryptate were added
to the plates and the FRET signal were read in a micro plate reader after
3-4 h at 620 nm & 665 nm.
31. Camps, M.; Rückle, T.; Hong, J.; Ardissone, V.; Rintelen, F.; Shaw, J.;
Ferrandi, C.; Chabert, C.; Gillieron, C.; Françon, B.; Martin, T.;
Gretner, D.; Perrin, D.; Leroy, D.; Vitte, P.-A.; Hirsch, E.; Wymann, M.
P.; Cirillo, R.; Schwarz, M. K.; Rommel, C. Nat. Med. 2005, 11, 936.
Variation of a reference compound in the JeKo-1 assay was (pIC50
(mean ± SD)): 7.217 ± 0.133 (n = 57).
Inhibition of cellular activity in RAW264 cells: PI3K inhibitors were
identified by measuring the phosphorylation of AKT (at Ser473) using
TR-FRET technology (Phospho-AKT (Ser473) kit, Cisbio) in RAW264
cells - mouse monocytic cell line (ATCC). Cryo preserved cells were
thawed rapidly in DMEM + Glutamax medium (10% FBS) at 37 ˚C.
Cells were plated in white low volume 384 well plates at a density of
2.5E6 cells/mL in assay buffer (DMEM + 1% FBS). Potential inhibitors
were made up in DMSO and then diluted in the assay to give a final
concentration of not exceeding 1% (v/v) DMSO. Inhibitors were
incubated with 8 μL of cells for 3 hours at 37 °C prior stimulated with 2
µL C5a (2 nM) in assay buffer. After an incubation of 3 min at 37 °C
the cells were lysed for 30 min at RT on a plate shaker at 500 rpm. A 5
µL mix of d2/cryptate were added to the plates and the FRET signal
were read in a micro plate reader after 3-4 h at 620 nm & 665 nm.
Variation of a reference compound in the RAW264 assay was (pIC50
(mean ± SD)): 6.986 ± 0.148 (n = 23).
20. Delouvrié, B.; Germain, H.; Harris, C. S.; Lamorlette, M.; Lebraud, H.;
Nguyen, H. T. H.; Noisier, A.; Ouvry, G. Tetrahedron Lett. 2012, 53,
5380; (safety see ref. 3 in ref. 20).
21. Synthesis of 5b analogous to Scheme 1: Cyclization to the
methyltriazole required heating of the open chain intermediate to 100
°C in AcOH; synthesis of 5j analogous to Scheme 2; synthesis of 5o
analogous to Scheme 1: C was treated with hydrazine hydrate instead of
methylhydrazine, methylation of the triazole was performed as the last
step with MeI and NaH in DMF.
22. A methoxy- or fluorine-substituent in ortho- instead of meta- or para-
position of the phenyl ring resulted in reduced PI3Kδ-potency, probably
due to a change of conformation.