Discovery of a novel class of non-ATP site DFG-out state p38 inhibitors utilizing computationally assisted virtual fragment-based drug design (vFBDD)

Article abstract of DOI:10.1016/j.bmcl.2011.09.078

Discovery of a new class of DFG-out p38α kinase inhibitors with no hinge interaction is described. A computationally assisted, virtual fragment-based drug design (vFBDD) platform was utilized to identify novel non-aromatic fragments which make productive hydrogen bond interactions with Arg 70 on the αC-helix. Molecules incorporating these fragments were found to be potent inhibitors of p38 kinase. X-ray co-crystal structures confirmed the predicted binding modes. A lead compound was identified as a potent (p38α IC₅₀ = 22 nM) and highly selective (≥150-fold against 150 kinase panel) DFG-out p38 kinase inhibitor.

Full text of DOI:10.1016/j.bmcl.2011.09.078

Bioorganic & Medicinal Chemistry Letters 21 (2011) 7155–7165
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Discovery of a novel class of non-ATP site DFG-out state p38 inhibitors
utilizing computationally assisted virtual fragment-based drug design (vFBDD) q
⇑
Kristofer Moffett , Zenon Konteatis, Duyan Nguyen, Rupa Shetty, Jennifer Ludington, Ted Fujimoto,
Kyoung-Jin Lee, Xiaomei Chai, Haridasan Namboodiri, Michael Karpusas, Bruce Dorsey, Frank Guarnieri,
Marina Bukhtiyarova, Eric Springman, Enrique Michelotti
Ansaris, Four Valley Square, 512 East Township Line Road, Blue Bell, PA 19422, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Discovery of a new class of DFG-out p38a kinase inhibitors with no hinge interaction is described. A com-
Received 8 August 2011
Revised 15 September 2011
Accepted 19 September 2011
Available online 28 September 2011
putationally assisted, virtual fragment-based drug design (vFBDD) platform was utilized to identify novel
non-aromatic fragments which make productive hydrogen bond interactions with Arg 70 on the C-helix.
Molecules incorporating these fragments were found to be potent inhibitors of p38 kinase. X-ray co-crys-
a
tal structures confirmed the predicted binding modes. A lead compound was identified as a potent (p38
a
IC50 = 22 nM) and highly selective (ꢀ150-fold against 150 kinase panel) DFG-out p38 kinase inhibitor.
Keywords:
p38 Kinase
Ó 2011 Elsevier Ltd. All rights reserved.
Virtual fragment-based drug design
DFG-out conformation
Over-expression of pro-inflammatory cytokines such as tumor
necrosis factor-a (TNF-a), interleukin-1b (IL-1b), and interleukin-6
demonstrated analgesic efficacy in acute post-surgical dental pain
in a Phase II clinical trial11 or are currently being evaluated in the
clinic for efficacy in acute inflammatory pain, myelodysplastic
syndromes, neuropathic pain, COPD, RA, cardiovascular disease
and depression.10
(IL-6) have been associated with inflammatory diseases such as
rheumatoid arthritis (RA), inflammatory pain and chronic obstruc-
tive pulmonary disease (COPD).1 Protein therapeutics etanercept,
infliximab, and adalimumab, which target these cytokines, are con-
sidered to be a standard treatment for RA.2 However, the limitations
of protein therapeutics (such as cost and parenteral administration)
have stimulated significant research aimed at developing orally bio-
available, small molecule inhibitors of cytokines.
The mitogen activated protein kinase (MAPK) p38 plays a central
role in the regulation of the cytokines TNF-
expression and activation of p38 isoforms in synovial tissue ex-
tracted from RA patients implicated p38 as a potentially relevant
target for therapeutic intervention.4–6 This prompted many research
a
and IL-1b.3 Studies on
a
groups to focus on discovering and developing small molecule p38
a
inhibitors as promising anti-inflammatory drugs.7,8 While recent
clinical data indicate that p38 inhibition alone may not be sufficient
for effective treatment of RA,9,10 several p38 inhibitors have either
For the treatment of chronic diseases, like RA, kinase inhibitors
must be very selective to avoid unwanted side effects that can be
caused by off-target kinase inhibition.7 While selective p38 kinase
inhibitors have recently been reported,12 the early type I and type
II p38 inhibitors, such as SB203580 (1)13 and BIRB-796 (2)14 were
also active against other kinases. It was suggested that this lack
of selectivity may have contributed to the failure to progress these
compounds in the clinic. These early inhibitors share the key fea-
ture of occupying a significant portion of the ATP binding site,
making one or more key hydrogen bonds within the highly
conserved hinge region.
Abbreviations: TNF-a, tumor necrosis factor alpha; IL, interleuken; RA, rheuma-
toid arthritis; MAPK, mitogen activated protein kinase; ATP, adenosine triphos-
phate; COPD, chronic obstructive pulmonary disease; FBDD, fragment-based drug
design; LPS, lipopolysaccharide; PBMC, peripheral blood mononuclear cells; FFP,
fresh-frozen plasma; GCMC, grand canonical Monte Carlo.
q
PDB codes for compounds in the text co-crystallized with p38
a are as follows:
13a, 3P5K; 13b, 3P78; 17a, 3P7B; 27, 3P7C; 34a, 3P7A; 37b, 3P79.
⇑
Corresponding author.
E-mail address: kmistry@earthlink.net (K. Moffett).
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.09.078

7156
K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
Figure 1. Cartoon of 2 in the p38 DFG-out conformation outlining the various regions (in blue) and important residues. The design strategy involved removing interactions
with the hinge, finding replacement fragments for the tolyl pocket, and maintaining interactions with Glu 71, Asp 168, and the selectivity pocket.
Table 1
The B-values of selected fragments found in the DFG-out conformation tolyl pocket of
p38 kinase
Entry
1
Fragment
B-value
À8.7
2
3
À11.9
À11.6
4
5
À10.7
À12.6
6
7
À15.9
À15.2
We describe here our efforts to discover highly selective, orally
bioavailable p38 kinase inhibitors. Our reasoning was that target-
ing the less conserved, inactive DFG-out conformation of the ki-
nase15 while avoiding the conserved hinge region would give a
better chance at developing selective inhibitors. Recently, two
groups reported success with similar strategies.16 In the original
work leading to the development of 2, compounds having no hinge
interaction were reported to be potent against p38 kinase; how-
ever, it was thought necessary to add hinge binding elements to
improve, at least in part, the physio-chemical properties of the
Figure 2. Computationally predicted binding modes of 1,1-dioxothiomorpholine in
the DFG-out conformation of p38: (a) output from the calculations gives three
clusters of poses for the dioxothiomorpholine mimicking the molecular dynamics of
the fragment, in three distinct modes. Mode 1 (pink/gray) interacts with Lys 53.
Mode 2 (yellow/dark gray) interacts with Arg 70 with one H-bond. Mode 3 (purple/
light gray) interacts with Arg 67 or both Arg 67 and Arg 70; (b) compound 2 (light
blue/white) is overlaid with one representative pose from each cluster shown in
Fig. 2a, and shows the overlap of the tolyl moiety with binding mode 1.

K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
7157
Scheme 1. Synthesis of compounds containing thiophene core. Reagents and conditions: (a) KOH, 1:1 MeOH:H2O, 80 °C, >95%; (b) 2 M phosgene in toluene, H2O, r.t. >90%; (c)
ArNH2, THF, 70 °C, 75–95%; (d) ArNCO, THF, 35–95%; (e) cyclic amine, EDCI, HOBt, DCM and/or DMF, 39–100%; (f) TrocCl, NaOH, 1:2 H2O:EtOAc, 5–15 °C, 17–76%; (g) aniline,
DMSO, DIEA, 80 °C, 18 h, 13–60%; (h) ArNCO, toluene or THF, 80 °C, 18 h, 10–58%.
The strategy we used to develop novel p38 kinase inhibitors
having no hinge interaction is outlined in Figure 1. The urea inter-
actions with Glu 71 and the backbone NH of Asp 168, as well as the
t-butyl group in the DFG-out pocket were all interactions that
would be maintained. Therefore, we focused the computational ef-
fort on evaluating fragments in the selectivity pocket and the tolyl
pocket. In order to maximize the drug-like character, any newly
designed compounds needed to reduce the hydrophobicity of com-
pounds containing both naphthyl and tolyl moieties. Finding a
fragment that had enough hydrophobic character to interact with
the tolyl pocket and polarity or non-aromatic character to increase
drug-like properties had high priority.
We have previously described our fragment-based computa-
tional design approach,17,18 which uses a grand canonical Monte
Carlo (GCMC) simulation to generate ensembles of small organic
Scheme 2. Reduction of amide to give compounds with a methylene linker.
compounds.14,15 The challenge for us was to demonstrate that
selective, potent and druggable compounds could be made without
significant interaction with the hinge region.
Scheme 3. Synthesis of compound having fragment directly linked to thiophene corre. Reagents and conditions: (a) (i) HCl, NaNO2, H2O, 0 °C, (ii) CuBr2, 100 °C, 35–60%; (b)
dioxothiomorpholine, Cs2CO3, Pd2(dba)3, xantphos, 1,4-dioxane, 85 °C, 18 h, 50%; (c) NaOH, 3:1 MeOH:H2O, 80 °C, 2 h, quant.; (d) (i) TEA, DPPA, 100 °C, 2 h, (ii) ArNH2, 100 °C,
3 h, 25%.

7158
K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
Scheme 4. Synthesis of two-atom linker compounds using pyrazole core. Reagents and conditions: (a) DIEA, toluene, reflux; (b) ArNCO, toluene, reflux, 47–59% (two steps);
(c) NaOH, 2:1:1 THF:MeOH:H2O, r.t.; (d) amine, EDCI, HOBt, DCM, r.t., 22–29% (two steps).
Scheme 5. Synthesis of one-atom linker compounds with pyrazole core. Reagents and conditions: (a) isocyanate, toluene, reflux, 50–58%; (b) phosgene, DIEA, DCM, 0 °C to
r.t.; (c) DIEA, 1,4-dioxane, 70 °C, 15–21 h, 17–23%.
Table 2
Evaluation of dioxothiomorpholine fragment, thiomorpholine derivatives and linker strategies against p38 kinasea
R
W
X
Y
Z
Ar
Compound
p38a IC50 (lM)
S
S
CH
CH
C
C
CO
CO
1-Naphthyl-
2-Naphthyl-
13a
13b
0.076
0.11
S
S
S
S
CH
CH
CH
CH
S
S
S
S
S
S
S
S
N
N
N
N
C
C
C
C
C
C
C
C
C
C
C
C
N
N
N
N
CO
4-ClPh-
13c
29a
29b
29c
14a
14b
14c
30a
30b
30c
34a
34b
37a
37b
41a
41b
0.25
0.22
0.9
CH2
CH2
CH2
CO
CO
CO
CH2
CH2
CH2
-
1-Naphthyl-
2-Naphthyl-
4-ClPh-
1-Naphthyl-
2-Naphthyl-
4-ClPh-
1-Naphthyl-
2-Naphthyl-
4-ClPh-
1-Naphthyl-
2-Naphthyl-
1-Naphthyl-
2-Naphthyl-
1-Naphthyl-
2-Naphthyl-
1.7
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
0.13
0.12
0.24
1.2
5.7
16
0.52
2
6.4
2.3
3.5
-
CH2CO
CH2CO
CO
CO
10
S
S
CH
CH
C
C
CO
CO
1-Naphthyl-
1-Naphthyl-
15
16
>10
2.0
a
The p38 kinase IC50 values were determined by measuring the incorporation of 33P from -[33P] ATP into the GST-ATF-2 protein substrate as described in the Experimental
c
(see Supplementary data). IC50 Values were calculated by non-linear regression with sigmoidal dose-response equation in GraphPad PrismÒ using data generated in two
independent dose response experiments.
fragments, calculate the free energy of binding of these fragments
to a protein, and assemble these fragments into potential drugs.
These ensembles, or clusters of poses, for each fragment and for
each molecule built by the GCMC method represents the dynamic
molecular motion of the ligand instead of a single static pose. We
applied this virtual fragment-based drug design (vFBDD) approach

K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
7159
Figure 3. X-ray co-crystal structures of DFG-out compounds in p38
a with overlays of representative poses from the predicted binding modes 1 and/or 2 of the
dioxothiomorpholine fragment (distances shown are measured from atom centers): (a) compound 13a (yellow/white) showing overlap with a representative pose from
predicted binding mode 1 (pink/dark gray); (b) compound 13b (light blue/white) showing overlap with a representative pose from binding mode 2 (yellow/dark gray).
Comparison to Fig. 3a shows how the 2-naphthyl substitution has extended the dioxothiomorpholine out of binding mode 1; (c) compound 34a (orange/white) showing the
dioxothiomorpholine moiety occupying space midway between binding modes 1 (pink/dark gray) and 2 (yellow/gray); (d) compound 37b (light blue/white) showing the
dioxothiomorpholine moiety in a position above binding mode 2 (yellow/dark gray).
to evaluate fragments and their preferred interactions with the p38
protein.
The interactions of eight hundred fragments with the 1kv2
structure of p38 kinase were simulated, and computed free ener-
gies were obtained.19 Fragments containing sp3 hybridized nitro-
gens were simulated in the non-protonated state. For each
fragment, a distribution or cluster of poses of equivalent energy
was found at various sites of interest on the protein surface. The
energy values for each cluster were given as a B-value, which is a
dimensionless number directly related to the free energy of bind-
ing.17 In the selectivity pocket, multiple fragments (mostly aro-
matic) were found with favorable energies. Some of these
fragments were selected to be used in new compounds as naphtha-
lene replacements. A variety of fragments showed up in the tolyl
pocket including aromatic, aliphatic, polar and mixed character
fragments.
From these results, the tolyl pocket fragments went through an
evaluation process. Fragments with unfavorable B-values were dis-
carded, leaving about fifty having favorable B-values compared to
the tolyl fragment. Fragments with B-values within ꢁ10 units were
Figure 4. X-ray co-crystal structure of 13a (yellow/white) with overlay of 2 (light
blue/dark grey) showing the two ring methylenes in the hydrophobic tolyl pocket.

7160
K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
Table 3
Evaluation of dioxothiomorpholine replacementsa
Assay (IC50 and EC50
values in M)
l
p38
PBMC:
a
:
0.076
2.5
0.018
2.5
0.11
1.3
0.025
>3
a
p38
PBMC:
a
:
0.11
2.7
0.012
2.8
0.018
2.5
0.012
0.27
b
c
p38
PBMC:
a
:
0.25
2.1
0.03
0.59
0.083
1.6
0.086
2.8
a
The p38 kinase IC50 values were determined by measuring the incorporation of 33P from -[33P] ATP into the GST-ATF-2 protein substrate. Cellular assay PBMC IC50 values
c
were determined by measuring LPS-induced TNF-a production in peripheral blood monocyte cells (PBMC) as described in the Experimental (see Supplementary data). IC50
values were calculated by non-linear regression with sigmoidal dose-response equation in GraphPad PrismÒ using data generated in two independent dose response
experiments.
considered similar. Table 1 shows a subset of these fragments.
Entries 1–4 were included in the initial round of selection while
entries 5–7 were evaluated in a second iterative round after data
on the first compounds were collected.
Overlays of the fragment distributions with 2 showed that
modes 1 and 2 could be accessed by either a direct bond to a 5-
membered ring core or through one or two atom linkers
(Fig. 2b). Binding mode 3 may also be linked to a core by a two
atom linker but the computationally minimized small molecule
structure showed that it would only cover the edge of the diox-
othiomorpholine distribution. Such compounds would not be fully
representative of the predicted mode and may not be as active as
compounds that put the fragment in the center of the predicted
distribution.
We chose to utilize both a thiophene urea core and a pyrazole
urea core as the scaffolds to assess the fragments suggested by
our computational methodology. Both cores have the advantage
of having been incorporated into active p38 compounds15,20 and
having the synthetic feasibility to make the targeted compounds.
The thiophene core was used to make compounds with the frag-
ment directly bonded and one-atom linked to the core while the
pyrazole core was used to prepare compounds with the fragment
attached via one- and two-atom linkages.
Further selections were made based on the criteria we had set:
(1) preferably non-aromatic; (2) some hydrophobic character; (3)
polar functionality; (4) improved drug-like properties. Aromatic
fragments, such as chlorobenzene (entry 2, Table 1), were dis-
carded as we wanted to reduce some of the aromatic character in
any new compounds. Fragments having polarity, such as carbonyl
or sulfoxide, and hydrophobocity, such as the ring methylenes seen
in structures like entries 3–7 in Table 1 met our criteria. While the
polarity and saturated ring systems would reduce aromaticity and
improve drug-like properties, enough hydrophobicity would re-
main in the methylenes to interact with the hydrophobic nature
of the tolyl pocket.
An evaluation based on synthetic feasibility was used to select
the final fragments. In order to quickly gather data from an initial
set of compounds, we wanted fragments that could easily be incor-
porated into compounds with core structures already validated to
have activity towards p38 kinase. These core structures would in-
clude 5-membered ring heterocycles similar to what is found in
compound 2. For this reason we felt that compounds such as pyr-
rolidonone (entry 3, Table 1) would not be suitable since the chem-
istry required to attach it to the core systems we had in mind
would be more elaborate and time consuming. Fragments with
hetero-atoms in appropriate positions, like those seen in entries
4–7 in Table 1, could more easily be incorporated into proof-of-
concept compounds.
Starting from the thiophene aminoesters 3 and 4,21 the car-
bonyl-linked target compounds were prepared via four separate
routes (Scheme 1). Anilines with sufficient nucleophilicity utilized
oxazinedione intermediates 5 and 6 while less reactive anilines
were incorporated via an isocyanate using amino acid 7. Amide for-
mation with the desired amine containing fragments allowed us to
rapidly arrive at the target compounds (13–28). Some compounds
were made via a parallel synthesis methodology utilizing interme-
diates 10–12. The methylene linked compounds 29a–c and 30a–c
were made from the amides 13a–c and 14a–c by reduction using
borane (Scheme 2). Direct-linked compounds 34a and 34b were
synthesized via a key palladium mediated coupling (Scheme 3). It
should be noted that the conversion of 4 to the bromide (analogous
to 31) was accomplished but all attempts to add dioxothiomorph-
oline using various conditions failed to give the desired product
(analog of 32).
The first fragment identified from the simulation and selection
process was 1,1-dioxothiomorpholine (entry 4, Table 1). The com-
putational simulation output shows the fragment as a cluster of
poses in various regions of the protein. Each cluster is identified
as a potential binding mode for that fragment to the protein
(Fig. 2). Three binding modes for this fragment were found
(Fig. 2a): the first is dominated by an H-bond interaction with
Lys 53 (mode 1); the second has a single H-bond interaction with
Arg 70 (mode 2); the third has two H-bond interactions with Arg
67 or with both Arg 67 and Arg 70 (mode 3).
Using a pyrazole core more easily allowed for synthesis of com-
pounds with a two atom linker to the fragment. Cyclization of 4,4-
dimethyl-3-oxopentanenitrile with ethyl 2-hydrazinylacetate
hydrochloride (Scheme 4) gave the amino pyrazole core (35) which

K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
7161
Table 4
subsequently allowed for the formation of the desired products
37a–b.
Effects of modifications in selectivity pocket with compounds containing diazepanone
fragmenta
One atom linked pyrazole compounds were made by first react-
ing aminopyrazole 38 with 1- or 2-isocyanatonaphthalene in reflux-
ing toluene to give products 39a–b with the desired regiochemistry
(Scheme 5). Coupling pyrazole ureas 39a–b with dioxothiomorpho-
line carbamoyl chloride 40 gave the desired products 41a–b.
Initial evaluation of inhibitors containing the 1,1-diox-
othiomorpholine moiety was done using both thiophene and pyr-
azole cores with naphthalene and 4-chlorobenzene occupying the
selectivity pocket. Compounds with a carbonyl linker to either of
the thiophene cores (13a–c and 14a–c) showed good potency in
a p38 kinase assay with IC50 values at or below 250 nM (Table 2).
There was no significant difference in activity between the 2-
and 3-aminothiophenes. Replacement of the carbonyl with a meth-
ylene significantly reduced activity (compare 13a–c with 29a–c
and 14a–c with 30a–c). Compounds with the dioxothiomorpholine
directly connected to the core (34a–b) also showed decreased po-
tency. The carbonyl linker may have the proper balance between
flexibility (compared to directly bonded compounds) and rigidity
(compared to the methylene-linked compounds) to provide an
Ar
Assay (IC50 and EC50 values in lM)
p38
PBMC:
a
:
0.006
1.3
17d
p38
PBMC:
a
:
0.042
1.6
17e
17f
p38
PBMC:
a
:
0.01
0.25
p38
PBMC:
a
:
0.013
4.8
17g
p38
PBMC:
a
:
0.022
0.23
17h
17i
p38
PBMC:
a
:
0.08
7
p38
PBMC:
a
:
0.13
5.8
17j
p38
PBMC:
a
:
0.002
0.23
17k
a
The p38 kinase IC50 values were determined by measuring the incorporation of
33P from -[33P] ATP into the GST-ATF-2 protein substrate. Cellular assay PBMC IC50
values were determined by measuring LPS-induced TNF- production in peripheral
c
a
blood monocyte cells (PBMC) as described in the Experimental (see Supplementary
data). IC50 values were calculated by non-linear regression with sigmoidal dose-
response equation in GraphPad PrismÒ using data generated in two independent
dose response experiments.
optimal conformation. The dioxothiomorpholine oxygens are
important for compound binding as demonstrated by the de-
creased activity for the compound containing thiomorpholine
(compare 13a with 16). Interestingly, compound 15, containing
the mono-oxythiomorpholine, also had reduced activity. Com-
pounds with a naphthyl moiety had increased activity over the
4-chlorophenyl (compare 13a–b and 14a–b with 13c and 14c).
This could be due to better space filling in the hydrophobic
selectivity pocket by the larger naphthalene moiety. The trend is
exaggerated when the compounds are less active (see 29a–c and
30a–c).
The carbonyl linker on the pyrazole core was about 50-fold less
active (compare 13a–b with 41a–b). This could be due to either a
difference in the electronics of the two classes of molecules affect-
ing their interaction with the protein, a difference in how each core
presents the dioxothiomorpholine moiety to the protein, or a
combination of the two. Addition of the two atom linker to the
pyrazole core showed no enhancement of activity (compare 41a–
b with 37a–b).
Figure 5. X-ray co-crystal structure of 17a (light blue/white) in p38 kinase: (a)
overlay with the co-crystal structure of 13a (pink/dark gray); (b) overlay with the
co-crystal structure of 13b (yellow/dark gray). Compound 17a has the diazepanone
is oriented with the carbonyl pointing towards Arg 70. Also shown is the
intramolecular hydrogen bond in 17a and the hydrogen bond between 17a and
Arg 70 (distances are measured from atom centers).

7162
K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
The predicted binding modes for the dioxothiomorpholine frag-
having a two atom linker, places the dioxothiomorpholine close
to binding mode 2 allowing for a hydrogen bond to Arg 70
(Fig. 3d). However, this linker pushes the moiety slightly up out
of the computed optimal distribution closer to the C-helix and
side-chain carbons of Arg 67 thus increasing unfavorable interac-
tions with the protein, reflected in the poor activity of 37b.
Overlaying the structures of 13a and 2 shows how the naphthyl
and core 5-membered ring systems overlap nicely, as would be ex-
pected (Fig. 4). Two methylenes from the dioxothiomorpholine
ring are positioned to interact with the hydrophobic region occu-
pied by the tolyl moiety of 2.
Having established an optimal linking strategy, diox-
othiomorpholine isosteres were evaluated with the expectation
that they would interact with Arg 70 through a hydrogen bond,
based on the simulation results. Compounds containing the isoster-
es showed an improvement in p38 kinase potency of 6–9-fold (dia-
zepanones 17a–c), 3–6-fold (dioxothiodiazepane 18a–c) and 3–10
fold (ketopiperazines 19a–c) compared to the dioxothiomorpholine
compounds (Table 3). The cell potency of these compounds, which
ment were confirmed by solving the X-ray co-crystal structures for
compounds containing the various linkers between the core and
dioxothiomorpholine (Fig. 3). Compounds with a carbonyl linker
were found in two binding modes. Compound 13a, with the one-
atom carbonyl linker, shows the dioxothiomorpholine placed in
the binding mode 1 closest to Lys 53 (Fig. 3a) making a hydrogen
bond. This is facilitated by the carbonyl being rotated (ꢁ90°) away
from the plane of the urea. Comparing this to a representative pose
from the predicted binding mode 1 shows good overlap between
the dioxothiomorpholine moiety and the fragment. The angle of
the dioxothiomorpholine moiety in the compound mimics very
closely the computationally predicted orientation. Compound
13b, differing from 13a by the 2-naphthyl substitution, places
the dioxothiomorpholine in binding mode 2 (Fig. 3b). The diox-
othiomorpholine moiety again shows good overlap with the pre-
dicted binding mode 2. The altered substitution point on the
naphthalene helps to push the rest of the molecule closer to Arg
70. The urea NH closest to the ring now has an intramolecular
hydrogen bond to the carbonyl linker which is helping to push
the dioxothiomorpholine even further out.
was assessed by measuring LPS-stimulated TNF-
a production in
isolated human peripheral blood mononuclear cells (PBMC), re-
mained relatively unchanged, with the exception of 19b which
showed an interesting 10-fold improvement of IC50 in the PBMC
assay.
The X-ray co-crystal structure of p38 with compound 17a
(Fig. 5) showed the carbonyl linker is in the plane of the urea, mak-
ing an intramolecular hydrogen bond with the urea NH. This is in
Compound 34a, a direct-connected compound, has the diox-
othiomorpholine placed between binding modes 1 and 2 (Fig. 3c).
The lower binding affinity of 34a to p38 kinase compared to 13a
and 13b would be expected with the dioxothiomorpholine moiety
occupying an area between the two predicted higher affinity
regions of binding modes 1 and 2. Compound 37b, the pyrazole
Table 5
Effect of substitution on the diazepanone amide nitrogena
Compound No.
R
X
Assay (IC50 and EC50 values in
PBMC
lM)
p38a
FFP–PBMC
20
21
22
H
H
H
0.014
0.14
3
—
4.2
0.83
3.8
0.013
2.8
9.7
2.2
23
24
H
H
0.031
0.019
0.39
0.16
25
26
H
H
0.029
0.039
>10
—
0.52
6.7
27
28
H
F
0.022
0.073
0.39
0.17
0.86
1.2
a
The p38 kinase IC50 values were determined by measuring the incorporation of 33P from
c
-[33P] ATP into the GST-ATF-2 protein substrate. Cellular assay PBMC and FFP
IC50 values were determined by measuring LPS-induced TNF- production in peripheral blood monocyte cells alone (PBMS assay) or in the presence of fresh frozen plasma
a
(FFP–PBMC assay) as described in the Experimental (see Supplementary data). IC50 values were calculated by non-linear regression with sigmoidal dose-response equation in
GraphPad PrismÒ using data generated in two independent dose response experiments.

K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
7163
125
100
75
a
-4
-4.5
-5
TNF
IL1
* p < 0.05
-5.5
-6
*
50
*
*
-6.5
-7
p38d
25
*
p38g
0
-7.5
-8
p38a
p38a
Ansaris
p38b
-8.5
Figure 7. Inhibition of TNF-a and IL-1b for compound 27 in mouse-LPS model. The
control is dexamethasone.
Kinase
-4
-4.5
-5
b
-5.5
-6
-6.5
-7
p38b
-7.5
-8
p38a
p38a
Ansaris
-8.5
Kinase
Figure 6. Broad kinase selectivity screening data for compounds 2 and 27 against a
panel of 150 kinases using the HotSpot™ filtration binding 33P kinase assay
(Reaction Biology, Malvern, PA). In the graph log IC50 values (M) for kinases in the
panel are plotted against an individual kinase in the screen listed in alphabetic
order. In-house kinase data for 2 and 27 are plotted for comparison (labeled as
‘p38a Ansaris’). The full list of kinases with corresponding IC50s is included in
Supplementary data. (a) compound 2 shows activity towards a number of kinases
with values comparable to its p38a IC50; (b) compound 27 is >150-fold less potent
towards any other kinase compared to p38a (except p38b).
Table 6
Selected data for compound 27
Hepatocyte (human) t1/2
CaCo2 permiability (Â10À6 cm/s)
Cyp3A4 IC50
hERG Screen
Oral PK (rat)
340 min
Pab: 0.39; Pba: 5.3
0.55
69% Bound @ 1
33%F (at 4 mg/kg)
lM
lM
contrast to what was seen with 13a (Fig. 5a) but is similar to 13b
(Fig. 5b). The carbonyl in the diazepanone ring of 17a is closer to
Arg 70 than the oxygens of the dioxothiomorpholine of 13b. This
allows the carbonyl to form a hydrogen bond with Arg 70. Also,
the diazepanone methylenes are in a hydrophobic pocket formed
by the side chain carbons of Glu71 and Arg 67. These interactions
may account for the improved kinase potency over 13b.
Compounds containing the diazepanone fragment were further
optimized due to their consistently good kinase potency. We eval-
uated the effect of modifications in the selectivity pocket (Table
4). Replacement fragments were picked, based on results from
simulations in the selectivity pocket, to mimic the space filling abil-
ity of naphthalene, including phenyl substituted with halogens and
polar groups in various configurations. These changes were an
Figure 8. X-ray co-crystal structure of compound 27 (light blue/white) in p38
kinase. (a) overlay of 17a (yellow/dark gray); (b) overlay of 2 (pink/dark gray)
Placement of the dimethylaminoethyl moiety in 27 is an approximation as no
electron density beyond the first methylene was seen. H-bond distance is measured
from atom centers.
attempt to improve cell permeability/activity, which would be
reflected in improved PBMC potency. Compounds containing 2,3-
dichlorophenyl (17f), benzodioxole (17h) and 2,3-dichloro-4-fluo-
rophenyl (17k) showed 3–10-fold improvement in cell potency
while maintaining kinase potency.

7164
K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
We chose to optimize around 17f in order to further improve
Predicted binding modes of the fragments were confirmed by
solving X-ray co-crystal structures. Compounds containing the
diazepanone fragment, which make a hydrogen bond to Arg 70
and have hydrophobic interactions in the tolyl pocket, result
in excellent selectivity while maintaining potency. Compound
27 demonstrates low nanomolar potency in p38 kinase assay
cell potency and pharmacokinetic properties. The co-crystal struc-
ture of 17a showed that the NH of the diazepanone is oriented to-
ward an open region of the protein and therefore could be
functionalized to modify compound properties including solubil-
ity. Most of the functionality incorporated at this position had good
potency towards p38 kinase, similar to 17f, in the enzymatic assay
(Table 5). Most compounds showed sub-micromolar activity in the
PBMC assay. However, many compounds lost potency when tested
in a cell assay conducted in the presence of human plasma (FFP–
PBMC assay). Encouragingly, compound 27 had FFP–PBMC IC50 of
860 nM. Interestingly, when the dimethylaminoethyl group was
added to 17k (a compound with better kinase and comparable
PBMC activity to 17f) the resulting compound 28 was found to
be 3-fold less potent toward the kinase compared to 27 even
though cell potency was comparable.
(IC50 = 0.022
PBMC EC50 = 0.86
kinase panel. Compound 27 showed oral bioavailability and
lM), good cell potency (PBMC EC50 = 0.34
lM; FFP–
lM) and high selectivity (>150-fold) in a 150
demonstrated suppression of TNF-
efficacy model.
a and IL-1b in an animal
Acknowledgments
The authors wish to thank Syngene International Pvt. Ltd for
synthesis of derivatized diazepanone intermediates, Robert Schiks-
nis for assistance with NMR studies, Brandon Campbell for per-
forming preparative SFC purification and SFC-MS studies, Tim
Allison for uploading crystal structures to the PDB, William Moore,
Bin Liu and Martha Kelly for scientific contributions and Janet
Gaboury valuable help in preparing this manuscript.
Compounds 2 (dual DFG-out/ATP site binder) and 27 (non-
hinge binder) were evaluated for broad kinase selectivity against
a panel of 150 kinases (Fig. 6).22 Both compounds showed similar
activity toward p38
IC50 = 0.020 M). These data compare well with data obtained from
the in-house p38 kinase assay (2: IC50 = 0.02 M; 27:
IC50 = 0.022 M). The only other kinase that was inhibited by com-
pound 27 within 150-fold range of p38 IC50 was another isoform
of p38 kinase – p38b (IC50 = 0.05 M; Fig. 6b). In sharp contrast,
compound 2 inhibited 29 kinases out of 150 with IC50 values
a in this assay (2: IC50 = 0.023 lM; 27:
l
a
l
l
Supplementary data
a
l
Supplementary data (data table containing results from the
broad kinase screen and experimental details for synthesis of com-
pounds) associated with this article can be found, in the online ver-
sion, at doi:10.1016/j.bmcl.2011.09.078.
63.0
60.1
l
l
M including 12 kinases being inhibited with IC50 values
M (Fig. 6a).
Compound 27 was evaluated in various in vitro, in vivo and effi-
cacy studies (Table 6). It showed good stability in human hepato-
cytes (340 min) and good oral bioavailability in rat (33%F at
4 mg/Kg). Compound 27 also showed the potential for efflux in
the CaCo2 assay, sub-micromolar activity towards Cyp3A4 and
high binding in the hERG screen. When administered orally in an
LPS-challenged mouse model, compound 27 showed a dose-depen-
References and notes
1. (a) Choy, E. H. S.; Panayi, G. S. New Eng. J. Med. 2001, 344, 907; (b) Dinarello, C.
A. Curr. Opin. Immunol. 1991, 3, 941.
2. (a) Jarvis, B.; Faulds, D. Drugs 1999, 57, 945; (b) Seymour, H. E.; Worsley, A.;
Smith, J. M.; Thomas, S. H. L. Br. J. Clin. Pharmacol. 2001, 51, 201; (c) Rutgeerts, P.
J. Aliment. Pharmacol. Ther. 1999, 13, 9.
3. Westra, J.; Limburg, P. C. Mini-Rev. Med. Chem. 2006, 6, 867.
4. Hale, K. K.; Trollinger, D.; Rihanek, M.; Mantley, C. L. J. Immunol. 1999, 162,
4246.
5. Korb, A.; Tohidast-Akrad, M.; Cetin, E.; Axmann, R.; Smolen, J.; Schett, G.
Arthritis Rheum. 2006, 54, 2745.
6. Schett, G.; Tohidast-Akrad, M.; Smolen, J. S.; Schmid, B. J.; Steiner, C. W.; Bitzan,
P.; Zenz, P.; Redlich, K.; Xu, Q.; Steiner, G. Arthritis Rheum. 2000, 43, 2501.
7. Goldstein, D. M.; Gabriel, T. Curr. Top. Med. Chem. 2005, 5, 1017.
8. Pettus, L. H.; Wurtz, R. P. Curr. Top. Med. Chem. 2008, 8, 1452.
9. Genovese, M. C. Arthritis Rheum. 2009, 60, 317.
10. Goldstein, D. M.; Kuglstatter, A.; Lou, Y.; Soth, M. J. J. Med. Chem. 2010, 53, 2345.
11. Tong, S. E.; Daniels, S. E.; Montano, T.; Chang, S.; Desjardins, P. Am. Soc. Clin.
Pharmcol. Ther. 2004, 75, 1.
12. Fabian, M. A.; Biggs, W. H., III; Treiber, D. K.; Atteridge, C. E.; Azimioara, M. D.;
Benedetti, M. G.; Carter, T. A.; Ciceri, P.; Edeen, P. T.; Floyd, M.; Ford, J. M.;
Galvin, M.; Gerlach, J. L.; Grotzfeld, R. M.; Herrgard, S.; Insko, D. E.; Insko, M. A.;
Lai, A. G.; Lelias, J. M.; Mehta, S. A.; Milanov, Z. V.; Velasco, A. M.; Wodicka, L.
M.; Patel, H. K.; Zarrinkar, P. P.; Lockhart, D. J. Nat. Biotechnol. 2005, 23, 329.
13. Cuenda, A.; Rouse, J.; Doza, Y. N.; Meier, R.; Cohen, P.; Gallagher, T. F.; Young, P.
R.; Lee, J. C. FEBS Lett. 1995, 364, 229.
14. Pargellis, C.; Tong, L.; Churchill, L.; Cirillo, P. F.; Gilmore, T.; Graham, A. G.;
Grob, P. M.; Hickey, E. R.; Moss, N.; Pav, S.; Regan, J. Nat. Struct. Biol. 2002, 9,
268.
dant inhibitory effect towards TNF-a and IL-1b release (Fig. 7).
The co-crystal structure of 27 with p38 shows many similarities
to the co-crystal of 17a (Fig. 8a). The dichlorophenyl moiety of 27
occupies similar space in the selectivity pocket as the naphthyl
group of 17a. The hydrogen bond between the carbonyl on the dia-
zepanone with Arg 70 is maintained, and is closer by 0.55 Å com-
pared to 17a. Electron density for the dimethylaminoethyl
moiety beyond the first methylene was not seen in the co-crystal
structure. This would indicate that the dimethylaminoethyl moiety
did not have any direct interactions with the protein but it does ap-
pear to influence the conformation of the diazepanone ring, rela-
tive to the unsubstituted ring, without significantly impacting
kinase potency. Therefore it appears to act as a solubilizing group,
with the effect of enhancing the overall properties of the molecule.
A comparison of 27 and 2 (Fig. 8b) shows significant overlapping of
the core 5-membered rings, the t-butyl and selectivity pockets. Be-
cause of the interaction with Arg 70, the diazepanone of 27 only
has hydrophobic contacts with the protein along one side of the
ring whereas the tolyl moiety of 2 presents an entire face to this
hydrophobic region.
15. Regan, J.; Breitfelder, S.; Cirillo, P.; Gilmore, T.; Graham, A. G.; Hickley, E.; Klaus,
B.; Madwed, J.; Moriak, M.; Moss, N.; Pargelis, C.; Pav, S.; Proto, A.; Swinamer,
A.; Tong, L.; Torcellini, C. J. Med. Chem. 2002, 45, 2994.
16. (a) Liu, H.; Kuhn, C.; Feru, F.; Jacques, S. L.; Deshmukh, G. D.; Ye, P.; Rennie, G.
R.; Johnson, T.; Kazmirski, S.; Low, S.; Coli, R.; Ding, Y.-H.; Cheng, A. C.; Tecle,
H.; English, J. M.; Stanton, R.; Wub, J. C. Bioorg. Med. Chem. Lett. 2010, 20, 4885;
(b) Ahn, Y. M.; Clare, M.; Ensinger, C. L.; Hood, M. M.; Lord, J. W.; Lu, W.-P.;
Miller, D. F.; Patt, W. C.; Smith, B. D.; Vogeti, L.; Kaufman, M. D.; Petillo, P. A.;
Wise, S. C.; Abendroth, J.; Chun, L.; Clark, R.; Feese, M.; Kim, H.; Stewart, L.;
Flynn, D. L. Bioorg. Med. Chem. Lett. 2010, 20, 5793.
17. Clark, M.; Guarnieri, F.; Shkurko, I.; Wiseman, J. J. Chem. Inf. Model. 2006, 46,
231.
18. Clark, M.; Meshkat, S.; Talbot, G. T.; Carnevali, P.; Wiseman, J. S. J. Chem. Inf.
Model. 2009, 49, 1901.
We used computational/virtual fragment-based drug design
software and methods to design a novel p38 kinase inhibitor with
outstanding selectivity. Using this approach we discovered a
number of fragments containing both aliphatic and polar func-
tionalities that occupied the tolyl pocket of the DFG-out confor-
mation of p38a kinase. The previously unrecognized 1,1-
dioxothiomorpholine fragment along with the isosteres diazepa-
none, dioxothiodiazepane and ketopiperazine were identified
and incorporated into compounds that showed good potency for
p38 kinase without needing the interaction with the hinge region.
19. For further details of the methodology as it was applied here, see the
Supplementary data for a brief description.

K. Moffett et al. / Bioorg. Med. Chem. Lett. 21 (2011) 7155–7165
7165
20. (a) Dumas, J.; Hatoum-Mokdad, H.; Sibley, R.; Riedl, B.; Scott, W. J.; Monahan,
M. K.; Lowinger, T. B.; Brennan, C.; Natero, R.; Turner, T.; Johnson, J. S.;
Schoenleber, R.; Bhargava, A.; Wilhelm, S. M.; Housley, T. J.; Ranges, G. E.;
Shrikhande, A. Bioorg. Med. Chem. Lett. 2000, 10, 2051; (b) Redman, A. M.;
Johnson, J. S.; Dally, R.; Swartz, S.; Wild, H.; Paulsen, H.; Caringal, Y.; Gunn, D.;
Renick, J.; Osterhout, M.; Kingery-Wood, J.; Smith, R. A.; Lee, W.; Dumas, J.;
Wilhelm, S. M.; Housley, T. J.; Bhargava, A.; Ranges, G. E.; Shrikhande, A.;
Young, D.; Bombara, M.; Scott, W. J. Bioorg. Med. Chem. Lett. 2001, 11, 9.
21. Ranges, G.; Scott, W.; Bombara, M.; Rauner, D.; Redman, A.; Smith, R.; Paulsen,
H.; Chen, J.; Gunn, D.; Renick, J. PCT Int. Appl. WO 98/52558, 1998.
22. A table identifying the kinases tested and the resulting IC50 values can be found
as Supplementary data.