Structure-based design, synthesis and crystallization of 2-arylquinazolines as lipid pocket ligands of p38α MAPK

Article abstract of DOI:10.1371/journal.pone.0184627

In protein kinase research, identifying and addressing small molecule binding sites other than the highly conserved ATP-pocket are of intense interest because this line of investigation extends our understanding of kinase function beyond the catalytic phosphotransfer. Such alternative binding sites may be involved in altering the activation state through subtle conformational changes, control cellular enzyme localization, or in mediating and disrupting protein-protein interactions. Small organic molecules that target these less conserved regions might serve as tools for chemical biology research and to probe alternative strategies in targeting protein kinases in disease settings. Here, we present the structure-based design and synthesis of a focused library of 2-arylquinazoline derivatives to target the lipophilic C-terminal binding pocket in p38α MAPK, for which a clear biological function has yet to be identified. The interactions of the ligands with p38α MAPK was analyzed by SPR measurements and validated by protein X-ray crystallography.

Full text of DOI:10.1371/journal.pone.0184627

RESEARCH ARTICLE
Structure-based design, synthesis and
crystallization of 2-arylquinazolines as lipid
pocket ligands of p38α MAPK
Mike Bu¨hrmann, Bianca M. Wiedemann, Matthias P. Mu¨ller, Julia Hardick, Maria Ecke,
Daniel Rauh*
Faculty of Chemistry and Chemical Biology, TU Dortmund University, Dortmund, Germany
* daniel.rauh@tu-dortmund.de
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
Abstract
In protein kinase research, identifying and addressing small molecule binding sites other
than the highly conserved ATP-pocket are of intense interest because this line of investiga-
tion extends our understanding of kinase function beyond the catalytic phosphotransfer.
Such alternative binding sites may be involved in altering the activation state through subtle
conformational changes, control cellular enzyme localization, or in mediating and disrupting
protein-protein interactions. Small organic molecules that target these less conserved
regions might serve as tools for chemical biology research and to probe alternative strate-
gies in targeting protein kinases in disease settings. Here, we present the structure-based
design and synthesis of a focused library of 2-arylquinazoline derivatives to target the lipo-
philic C-terminal binding pocket in p38α MAPK, for which a clear biological function has yet
to be identified. The interactions of the ligands with p38α MAPK was analyzed by SPR mea-
surements and validated by protein X-ray crystallography.
OPEN ACCESS
Citation: Bu¨hrmann M, Wiedemann BM, Mu¨ller
MP, Hardick J, Ecke M, Rauh D (2017) Structure-
based design, synthesis and crystallization of 2-
arylquinazolines as lipid pocket ligands of p38α
MAPK. PLoS ONE 12(9): e0184627. https://doi.
org/10.1371/journal.pone.0184627
Editor: Kjetil Tasken, University of Oslo, NORWAY
Received: June 14, 2017
Accepted: August 28, 2017
Published: September 11, 2017
Copyright: © 2017 Bu¨hrmann et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Introduction
Protein kinases have been a frequent topic in medicinal chemistry and drug development due
to their key function as mediating components in signal transduction, regulating cellular path-
ways on a molecular level, thereby playing a crucial role in the emergence of several diseases.
The conventional approach towards the treatment of kinase-related diseases has involved the
administration of ATP-competitive inhibitors which potently occupy and thereby block the
enzyme’s active site where the phosphotransfer from ATP to target substrates takes place [1,
2]. However, development of specifically selective inhibitors for a certain targeted kinase
within the related members of this enzyme family remains a major hurdle in drug research
[3, 4].
Data Availability Statement: Crystal structures
were deposited to the RCSB Protein Data Bank
(wwPDB) and can be found under the
corresponding entry IDs: 9c (5N63), 9g (5N64), 9h
(5N65), 9j (5N66), 9l (5N67) and 9m (5N68).
Funding: This work was co-funded by the German
Federal Ministry for Education and Research
(NGFNPlus and e:Med) (Grant No. BMBF
Successful strategies to gain improved selectivity within the kinome have revolved around
employing unique structural features of individual kinases, such as covalent modification of
cysteines [5, 6] or identifying and targeting alternative binding pockets distant from the active
site [7]. Alternative bindings sites far from the ATP-pocket can directly regulate kinase affinity
01GS08104, 01ZX1303C) and by the Deutsche
Forschungsgemeinschaft (DFG). https://www.
bmbf.de/en/index.html; http://www.dfg.de/. DR
thanks the German federal state North Rhine
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
1 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Westphalia (NRW) and the European Union
(European Regional Development Fund: Investing
In Your Future) (EFRE-800400).
and can potentially be addressed by small molecules which alter the kinase activity in a dual
manner, via both inhibition and activation [8, 9].
In addition to advantages in the development of selective kinase modulators, these binding
sites can aid in distinguishing so called non-catalytic functions, those processes triggered by
protein-protein (or protein-target) interactions, where kinases serve as scaffolds, e.g., for
the formation of multi-enzyme-complexes [10]. In this way, these remote sites can serve as
(allosteric) effectors of target molecules, directly affecting the location or the activity state of
interaction partners, or can influence cell proliferation, differentiation, and apoptosis. An
increasing number of those scaffolding functions of kinases are gradually being discovered
and their functions may even exceed the significance of the solely catalytic properties [10–12].
Accordingly, the identification and exploration of non-conserved regions may provide
insights into the putative unknown functions of protein kinases beyond catalysis and allosteric
regulation. Thus, elucidation of unique structural features that modulate protein kinases
through employing alternative binding pockets and investigations of the design of correspond-
ing small molecules as alternatives to ATP-competitive ligands has moved to the forefront of
kinase inhibitor research and kinase biology. Against this background, we recently identified a
novel class of p38α MAPK (mitogen-activated protein kinase) binders addressing a C-terminal
lipophilic binding pocket (LP) accessible for small molecules located at several angstroms dis-
tance from the enzymes active site. The discovered 2-arylquinazolines bind to the LP and their
corresponding co-crystal structures revealed a very distinct binding mode of these lipid pocket
ligands (LiPoLis) to p38α (Fig 1B). We considered that these ligands may serve as interesting
starting points to study the yet unexplored functions of this binding site in p38α [13]. Several
small molecules have been described to address this pocket (Fig 1A) and they can be classified
into detergent-like molecules, β-octyl-D-glucopyranoside (BOG) [14] and phosphatidylinosi-
tol ether lipid analogues [15], and small organic compounds, 1 (4-[3-(4-fluorophenyl)-1H-pyr-
azol-4-yl]pyridine) [16] and 2 (4-(trifluoromethyl)-3-(3-(trifluoromethyl)phenyl)-1H-
pyrazolo[3,4-b]pyridine-6(7H)-one) (Fig 1A) [17]. Overall, the majority of the published com-
pounds that address the LP have exhibited a rather low affinity towards p38α [17, 18].
The LP consists of the two α-helices 1L14 and 2L14, the αEF/αF loop and a deep lipophilic
sub-pocket that is mostly decorated with hydrophobic amino acid side chains (Fig 1B). Based
on published co-crystal structures and biochemical assay data, there are indications that some
of the already described LiPoLis alter the kinase conformation in a way that the enzymatic
activity could be directly influenced by ligand binding, although the corresponding data was
measured at high concentrations [15–17]. Previous publications also speculated that there
might be a regulatory fine-tuning as a result of conformational changes within the LP when
addressed by ligands [15, 18]. Thus, the biological role of the LP is still not fully understood
and may well serve some yet unknown function [19].
Competing interests: The authors have declared
that no competing interests exist.
To gain a more detailed insight into the role of the p38α LP, we undertook SAR studies
based on our previously found lead structure 3 and the known binding mode. Here, we present
a structure-based design, synthesis, and validation by surface plasmon resonance (SPR) analy-
sis and protein X-ray crystallography of novel LiPoLis that target the LP in p38α MAPK.
Materials and methods
Reagents and materials
Unless otherwise noted, all reagents and solvents were purchased from Acros, Alfa Aesar,
Apollo Scientific, Fluka, Merck or Sigma-Aldrich and used without further purification. Dry
solvents were purchased as anhydrous reagents from commercial suppliers. 1H and 13C spectra
were recorded on a Bruker Avance DRX 400 and a Bruker Avance DRX 500 spectrometer at
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
2 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 1. Binding modes of active site inhibitors and LiPoLis in p38α MAPK. (A) Superposed kinase domains of p38α
MAPK (cyan) in complex with active site inhibitor BIRB-796 (yellow) (PDB: 1KV2) and the quinazoline-based LiPoLi 3
(green) (PDB: 4DLJ). (B) Detailed binding mode of 3 (green) in the LP of p38α MAPK (cyan), highlighting key structural
elements and main interactions formed between the protein and the ligand. (C) Chemical structure of 3 with systematic
numbering of the quinazoline scaffold and highlighted moieties selected for derivatization.
https://doi.org/10.1371/journal.pone.0184627.g001
400 MHz or 500 MHz and 101 MHz or 125 MHz, respectively. Chemical shifts are reported in
δ (ppm) as s (singlet), d (doublet), dd (doublet of doublet), t (triplet), q (quartet), m (multiplet)
and bs (broad singlet) and are referenced to the residual solvent signal: DMSO-d6 (2.50) or
CDCl3 (7.26) for 1H and DMSO-d6 (39.52) or CDCl3 (77.16) for 13C. Compound identity was
further confirmed by LC-MS analysis on LCQ Advantage MAX (1200 series, Agilent) with
Eclipse XDB-C18-column (5 μM, 150 × 1.6 mm, Phenomenex). High resolution electrospray
ionization mass spectra (ESI-FTMS) were recorded on a Thermo LTQ Orbitrap (high resolu-
tion mass spectrometer from Thermo Electron) coupled to an "Accela" HPLC system supplied
with a "Hypersil GOLD" column (Thermo Electron). Analytical TLC was carried out on
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
3 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Merck 60 F245 aluminium-backed silica gel plates. Compounds were purified by column chro-
matography using silica gel from Baker (40–70 μm particle size) or VWR Prolabo (Normasil
60, 40–63 μm particle size). Flash column chromatography was done using a Biotage Isolera
One system with Biotage SNAP and SNAP Ultra columns, respectively, and monitored by UV
at 254 and 360 nm. Preparative HPLC was conducted on a Varian HPLC system (Pro Star 215)
with a VP 250/21 Nucleosil C18 PPN column from Macherey-Nagel and monitored by UV at
254 nm. All presented compounds were analyzed by HPLC to determine and ensure a purity
of ꢀ 95%.
Synthesis and analytics
Preparation of 7-nitro-2-phenylquinazolin-4-ol (7a). To a solution of 5a (5 g, 27.5
mmol) and benzamidine hydrochloride (3 g, 24.9 mmol) in 2-methoxyethanol (150 mL) was
added AcOH (3 mL) and the reaction mixture was stirred at 130˚C for 18 h. The warm solu-
tion was subsequently filtered and the residue washed with cold MeOH to obtain 7a as a pale
brown solid (1.8 g, 24%). 1H NMR (400 MHz, DMSO-d6) δ 12.93 (s, 1H), 8.43 (d, J = 2.1 Hz,
1H), 8.36 (d, 1H), 8.24 (d, J = 2.2 Hz, 1H), 8.24–8.20 (m, 2H), 7.65 (t, J = 7.3 Hz, 1H), 7.59
(t, J = 7.4 Hz, 2H); 13C NMR (126 MHz, DMSO-d6) δ 161.3, 154.5, 151.3, 149.1, 132.1, 131.9,
128.6, 128.1, 128.0, 125.3, 122.4, 112.0; HRMS (ESI-MS): Calculated for C14H10O3N3 [M+H]+:
268.07167; found: 268.07159.
Preparation of 8-nitro-2-phenylquinazolin-4-ol (7b). Anthranilic acid 5b (500 mg, 2.8
mmol) was placed in a microwave reaction vial and benzoic anhydride was added (1.55 g, 6.9
mmol). Prior to adding formamide (1.1 mL, 27.5 mmol) benzoic acid was melted at 45˚C to
homogenize the reaction mixture. After 10 min heating at 200˚C in a microwave reactor, the
solution was poured into ice-water and stirred for 10 min. Subsequent filtration yielded the
product 7b (268 mg, 37%). 1H NMR (500 MHz, DMSO-d6) δ 12.93 (s, 1H), 8.38 (dd, J = 8.0,
1.4 Hz, 1H), 8.31 (dd, J = 7.8, 1.4 Hz, 1H), 8.18–8.13 (m, 2H), 7.67–7.61 (m, 2H), 7.60–7.54 (m,
2H); 13C NMR (101 MHz, DMSO-d6) δ 160.97, 154.65, 146.68, 140.55, 132.23, 131.95, 129.72,
128.77, 128.24, 128.16, 126.02, 122.54.
Common procedure for the generation of 2-arylquinazoline scaffolds (7c-
e)
Anthranilic amide 6 (1 eq), NaHSO3 (1.2 eq) and a corresponding aldehyde building block (1
eq) were dissolved in DMAc and the solution stirred for 30 min. pTSA (0.1 eq) was added and
the mixture was stirred at 155˚C for 18 h. After concentration of the reaction mixture in vacuo
and separation between saturated NaHCO3 and DCM, the crude product was purified by flash
column chromatography to isolate the desired 2-arylquinazoline.
Preparation of 1-(4-(4-(4-hydroxy-7-nitroquinazolin-2-yl)phenyl)piperazin-1-yl)ethan-
1-one (7c). According to the common procedure, amide 6 (1 g, 5.5 mmol), NaHSO3 (690
mg, 6.6 mmol), aldehyde (1.3 g, 5.6 mmol) and pTSA (100 mg, 0.5 mmol) in DMAc (50 mL)
were converted to a crude reaction mixture that was purified by flash column chromatography
(0% ! 2% MeOH/DCM) to give 390 mg (18%) of the described target compound. 1H NMR
(500 MHz, DMSO-d6) δ 12.58 (s, 1H), 8.34 (d, J = 2.1 Hz, 1H), 8.30 (d, J = 8.7 Hz, 1H), 8.16 (d,
J = 9.0 Hz, 2H), 8.12 (dd, J = 8.7, 2.2 Hz, 1H), 7.06 (d, J = 9.0 Hz, 2H), 3.67–3.54 (m, 4H), 3.43–
3.38 (m, 2H), 3.34–3.29 (m, 2H), 2.05 (s, 3H); 13C NMR (126 MHz, DMSO-d6) δ 168.4, 154.1,
152.8, 151.3, 129.3, 128.1, 120.5, 119.0, 113.7, 46.8, 46.5, 45.0, 40.4, 21.2; HRMS (ESI-MS): Cal-
culated for C20H20O4N5 [M+H]+: 393.15098; found: 393.15035.
Preparation of 2-(4-morpholinophenyl)-7-nitroquinazolin-4-ol (7d). Following the
common procedure, amide 6 (2.2 g, 12.1 mmol), NaHSO3 (1.5 g, 14.4 mmol), aldehyde (2.3 g,
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
4 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
12.1 mmol) and pTSA (230 mg, 1.2 mmol) in DMAc (50 mL) were converted to a crude reac-
tion mixture that was purified by flash column chromatography (0% ! 2% MeOH/DCM) to
give 430 mg (10%) of the described target compound. 1H NMR (600 MHz, DMSO-d6) δ 12.60
(s, 1H), 8.33 (s, 1H), 8.29 (dd, J = 8.7, 1.5 Hz, 1H), 8.17–8.09 (m, 3H), 7.05 (d, J = 7.5 Hz, 2H),
3.73 (s, 4H), 3.28 (s, 4H); 13C NMR (101 MHz, DMSO-d6) δ 153.4, 151.4, 129.3, 128.2, 120.8,
119.1, 115.9, 115.7, 113.6, 65.9, 47.0; HRMS (ESI-MS): Calculated for C18H17O4N4 [M+H]+:
353.12443; found: 333.12444.
Preparation of 2-(4-(methylthio)phenyl)-7-nitroquinazolin-4-ol (7e). According to the
common procedure, amide 6 (400 mg, 2.2 mmol), NaHSO3 (270 mg, 2.6 mmol), aldehyde
(420 mg, 2.2 mmol) and pTSA (40 mg, 0.2 mmol) in DMAc (5 mL) were converted and the
crude product was subsequently used without further purification (235 mg, 34%). 1H NMR
(400 MHz, DMSO-d6) δ 12.85 (s, 1H), 8.40 (d, J = 2.0 Hz, 1H), 8.34 (d, J = 8.7 Hz, 1H), 8.21 (d,
J = 2.2 Hz, 1H), 8.19 (d, J = 2.8 Hz, 1H), 8.16 (s, 1H), 7.42 (d, J = 8.5 Hz, 2H), 2.56 (s, 3H); 13
C
NMR (101 MHz, DMSO-d6) δ 161.41, 153.99, 151.37, 149.28, 146.85, 144.03, 128.36, 128.19,
127.91, 125.25, 125.10, 122.30, 119.85, 14.04; HRMS (ESI-MS): Calculated for C15H12O3N3S
[M+H]+: 314.05939; found: 314.05934.
Common procedure for the preparation of 4-amino-7-/
8-nitroquinazolines (8a-d)
Substituted quinazolines 8a-d were synthesized following a common procedure. Quinazoline
7a (1 eq) was dissolved in SOCl2 and catalytic amounts of DMF. The reaction mixture was stir-
red for 4 h at 80˚C before the solvent was evaporated under reduced pressure. The residue was
dissolved in DCM/iPrOH (3:2) before the corresponding amine (1.5 eq) and DIPEA (2 eq)
were added. After stirring overnight at room temperature, extraction with saturated NaHCO3/
EtOAc, concentration in vacuo and silica gel column chromatography yielded the described
target compounds.
Preparation of N-benzyl-7-nitro-2-phenylquinazolin-4-amine (8a). According to the
general procedure, quinazoline 7a (200 mg, 0.75 mmol) was converted to 8a and the crude
product was purified by recrystallization from MeOH to yield 211 mg (79%) of the described
compound. 1H NMR (500 MHz, DMSO-d6) δ 9.32 (t, J = 5.7 Hz, 1H), 8.56 (d, J = 9.0 Hz, 1H),
8.48–8.42 (m, 3H), 8.22 (dd, J = 9.0, 2.4 Hz, 1H), 7.55–7.42 (m, 5H), 7.39–7.28 (m, 2H), 7.25
(t, J = 7.3 Hz, 1H), 4.94 (d, J = 5.7 Hz, 2H); 13C NMR (126 MHz, DMSO-d6) δ 161.2, 159.3,
150.1, 139.1, 137.7, 131.0, 128.3, 128.3, 128.1, 127.5, 125.2, 122.8, 118.5, 117.4, 44.1; HRMS
(ESI-MS): Calculated for C21H15N4O2 [M+H]+: 357.13460; found: 357.13446.
Preparation of N-(4-fluorophenyl)-7-nitro-2-phenylquinazolin-4-amine (8b). Follow-
ing the general procedure, quinazoline 7a (240 mg, 0.89 mmol) was converted to 8b and the
crude product was purified by recrystallization from MeOH to yield 213 mg (66%) of the
described compound. 1H NMR (400 MHz, DMSO-d6) δ 10.22 (s, 1H), 8.71 (d, J = 9.0 Hz, 1H),
8.45 (d, J = 2.3 Hz, 1H), 8.42–8.36 (m, 2H), 8.24 (dd, J = 9.0, 2.3 Hz, 1H), 7.94–7.83 (m, 2H),
7.54–7.46 (m, 3H), 7.29 (t, J = 8.9 Hz, 2H); 13C NMR (101 MHz, DMSO-d6) δ 161.05, 158.42
(d, J = 240.7 Hz), 157.75, 150.64, 150.05, 137.82, 136.64, 130.71, 128.43, 128.11, 125.71, 124.65
(d, J = 8.0 Hz), 122.68, 118.72, 118.53, 115.06 (d, J = 22.2 Hz); HRMS (ESI-MS): Calculated for
C20H14N4O2F [M+H]+: 361.10953; found: 361.10963.
Preparation of N-(4-fluorobenzyl)-7-nitro-2-phenylquinazolin-4-amine (8c). Follow-
ing the general procedure, quinazoline 7a (200 mg, 0.75 mmol) was converted to 8c and the
crude product was purified by silica gel column chromatography (5 ! 12% EtOAc/PE) to
yield 265 mg (95%) of the described compound. 1H NMR (400 MHz, DMSO-d6) δ 9.33 (t,
J = 5.7 Hz, 1H), 8.54 (d, J = 9.0 Hz, 1H), 8.46 (m, 3H), 8.22 (dd, J = 9.0, 2.3 Hz, 1H), 7.60–7.45
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
5 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
(m, 5H), 7.16 (t, J = 8.9 Hz, 2H), 4.91 (d, J = 5.6 Hz, 2H); 13C NMR (101 MHz, DMSO-d6) δ
161.28 (d, J = 242.4 Hz), 161.13, 159.28, 150.14, 137.68, 135.33, 130.79, 129.51 (d, J = 8.1 Hz),
128.40, 128.14, 125.24, 122.89, 118.65, 117.45, 115.14 (d, J = 21.3 Hz), 43.42; HRMS (ESI-MS):
Calculated for C21H16N4O2F [M+H]+: 375.12518; found: 375.12529.
Preparation of N-(3-fluorophenyl)-7-nitro-2-phenylquinazolin-4-amine (8d). Follow-
ing the general procedure, quinazoline 7a (240 mg, 0.89 mmol) was converted to 8d and the
crude product was purified by recrystallization from MeOH and precipitation in acetone to
yield 233 mg (72%) of the described compound. 1H NMR (500 MHz, DMSO-d6) δ 10.27 (s,
1H), 8.78 (d, J = 9.1 Hz, 1H), 8.53 (d, J = 1.4 Hz, 1H), 8.45–8.38 (m, 2H), 8.34–8.24 (m, 1H),
7.96 (d, J = 11.8 Hz, 1H), 7.76 (d, J = 8.0 Hz, 1H), 7.58–7.23 (m, 4H), 7.10–7.02 (m, 1H); 13
C
NMR (126 MHz, DMSO-d6) δ 161.95 (d, J = 241.4 Hz), 160.85, 157.63, 150.54, 150.31, 140.51
(d, J = 11.1 Hz), 137.34, 131.00, 130.08 (d, J = 9.5 Hz), 128.55, 128.06, 125.58, 123.09, 119.10,
117.91, 117.55, 110.57 (d, J = 21.0 Hz), 109.02 (d, J = 26.2 Hz); HRMS (ESI-MS): Calculated for
C20H14N4O2F [M+H]+: 361.10953; found: 361.11130.
Common procedure for the preparation of 4-amino-7-/
8-nitroquinazolines (8e-n)
Substituted quinazolines 8e-n were synthesized following a common procedure. Quinazolines
7a-e (1 eq) and HCCP (1 eq) were dissolved in DIPEA (5 eq) and MeCN. The reaction mixture
was stirred for 60 min at rt before the corresponding amine (6 eq) was added. After further
stirring for 18 h at rt, the crude products was purified by silica gel column chromatography to
yield the described target compounds.
Preparation of N-(3,4-difluorophenyl)-7-nitro-2-phenylquinazolin-4-amine (8e).
According to the general procedure, quinazoline 7a (150 mg, 0.56 mmol) was converted and
the crude product purified by silica gel column chromatography (1% ! 20% EtOAc/PE) to
the substituted aminoquinazoline 8e (114 mg, 56%). 1H NMR (500 MHz, DMSO-d6) δ 10.25
(s, 1H), 8.72 (d, J = 9.1 Hz, 1H), 8.49 (d, J = 2.1 Hz, 1H), 8.37 (m, 2H), 8.28 (dd, J = 9.0 Hz, 2.2
Hz,1H), 8.11 (ddd, J = 13.2 Hz, 7.5 Hz, 2.5 Hz, 1H), 7.71 (d, J = 9.0 Hz, 1H), 7.53 (m, 4H); 13
NMR (126 MHz, DMSO-d6) δ 160.7, 157.5, 150.3, 150.2, 137.1, 135.7, 135.6, 131.0, 128.5,
C
128.0, 125.4, 122.9, 119.0, 118.7, 117.3, 117.2, 117.0, 111.4 (d, J = 21.5 Hz); HRMS (ESI-MS):
Calculated for C20H13N4O2F2 [M+H]+: 379.10011; found: 379.1013.
Preparation of N-(3,4-difluorobenzyl)-7-nitro-2-phenylquinazolin-4-amine (8f).
According to the general procedure, quinazoline 7a (150 mg, 0.56 mmol) was converted and
the crude product purified by silica gel column chromatography (2% ! 20% EtOAc/PE) to
the substituted aminoquinazoline 8f (141 mg, 64%). 1H NMR (500 MHz, DMSO-d6) δ 9.30
(t, J = 5.7 Hz, 1H), 8.53 (d, J = 9.0, 1H), 8.46 (m, 3H), 8.23 (dd, J = 9.0 Hz, 2.3 Hz, 1H), 7.52
(m, 4H), 7.40 (dt, J = 10.7 Hz, 8.4 Hz, 1H), 7.33 (m, 1H), 4.90 (d, J = 5.6 Hz, 2H); 13C NMR
(126 MHz, DMSO-d6) δ 161.0, 159.3, 150.1, 150.1, 148.2, 137.6, 137.0, 130.7, 128.3, 128.1,
125.2, 124.1, 122.8, 118.6, 117.4, 117.2, 116.5, 116.4, 43.2; HRMS (ESI-MS): Calculated for
C21H15N4O2F2 [M+H]+: 393.11576; found: 393.11448.
Preparation of 7-nitro-2-phenyl-N-(thiophen-2-ylmethyl)quinazolin-4-amine (8g).
According to the general procedure, quinazoline 7a (150 mg, 0.56 mmol) was converted and
the crude product purified by silica gel column chromatography (1% ! 20% EtOAc/PE) to
the substituted aminoquinazoline 8g (144 mg, 71%). 1H NMR (400 MHz, DMSO-d6) δ 9.52
(s, 1H), 8.58 (d, J = 3.5 Hz, 2H), 8.52 (d, J = 8.9 Hz, 2H), 8.24 (dd, J = 8.7 Hz, 2.0 Hz, 1H), 7.56
(s, 3H), 7.38 (d, J = 4.8 Hz, 1H), 7.20 (d, J = 2.7 Hz, 1H), 6.99 (m, 1H), 5.09 (d, J = 5.1 Hz, 2H);
13C NMR (101 MHz, DMSO-d6) δ 161.7, 159.8, 151.1, 142.4, 131.8, 129.3, 129.2, 127.4, 127.2,
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
6 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
126.4, 126.1, 123.5, 119.7, 118.2, 40.0; HRMS (ESI-MS): Calculated for C19H15N4O2S [M+H]+:
363.09102; found: 363.09115.
Preparation of 7-nitro-2-phenyl-N-(2-(thiophen-2-yl)ethyl)quinazolin-4-amine (8h).
According to the general procedure, quinazoline 7a (150 mg, 0.56 mmol) was converted and
the crude product purified by silica gel column chromatography (10% ! 20% EtOAc/PE) to
the substituted aminoquinazoline 8h (137 mg, 65%). 1H NMR (500 MHz, DMSO-d6) δ 8.92
(s, 1H), 8.52 (dd, J = 6.6 Hz, 3.1 Hz, 2H), 8.48 (t, J = 5.3 Hz, 2H), 8.22 (dd, J = 8.9 Hz, 2.4 Hz,
1H), 7.53 (m, 3H), 7.34 (dd, J = 4.9 Hz, 1.3 Hz, 1H), 6.97 (dd, J = 8.3 Hz, 3.3 Hz, 2H), 3.94 (dd,
J = 12.8 Hz, 7.0 Hz, 2H), 3.28 (s, 2H); 13C NMR (126 MHz, DMSO-d6) δ 161.1, 159.3, 150.1,
141.4, 139.2, 137.7, 130.7, 128.3, 128.1, 127.0, 125.3, 125.1, 124.2, 122.8, 118.5, 117.4, 42.6, 28.3;
HRMS (ESI-MS): Calculated for C20H17N4O2S [M+H]+: 377.10667; found: 377.10703.
Preparation of 4-(2-(4-fluorophenyl)-4,5-dihydro-1H-imidazol-1-yl)-7-nitro-2-phenyl-
quinazoline (8i). According to the general procedure, quinazoline 7a (150 mg, 0.56 mmol)
was converted and the crude product purified by silica gel column chromatography (20%
EtOAc/PE ! 100% EtOAc) to the substituted aminoquinazoline 8i (137 mg, 65%). In devia-
tion from the common protocol K2CO3 (5eq, 323 mg, 2.34 mmol) was used as a base. 1H
NMR (500 MHz, DMSO-d6) δ 8.66 (m, 1H), 8.54 (m, 1H), 8.31 (m, 1H), 7.72 (m, 4H), 7.44
(m, 1H), 7.32 (m, 2H), 7.23 (m, 2H), 4.49 (t, J = 8.4 Hz, 2H), 4.12 (t, J = 8.4 Hz, 2H); 13C NMR
(126 MHz, DMSO-d6) δ 162.3, 161.6, 160.3, 153.0, 151.4, 137.2, 132.1, 131.0, 130.9, 129.8,
129.2, 128.9, 128.7, 124.1, 120.3, 120.2, 116.1 (d, J = 22.0 Hz), 56.4, 54.1; HRMS (ESI-MS): Cal-
culated for C20H17N4O2S [M+H]+: 414.13608; found: 414.13580.
Preparation of N-((4-(cyclopropylmethyl)furan-2-yl)methyl)-7-nitro-2-phenylquinazo-
lin-4-amine (8j). According to the general procedure, quinazoline 7a (180 mg, 0.67 mmol)
was converted and the crude product purified by silica gel column chromatography (25%
EtOAc/PE ! 50% EtOAc) to the substituted aminoquinazoline 8j as a hydrochloride salt (233
mg, 87%).
1H NMR (600 MHz, DMSO-d6) δ 8.59 (d, J = 9.0 Hz, 1H), 8.56–8.52 (m, 3H), 8.27 (dd,
J = 8.9, 2.0 Hz, 1H), 7.62–7.54 (m, 4H), 6.33 (d, J = 3.0 Hz, 1H), 6.05 (d, J = 3.1 Hz, 1H), 4.88
(d, J = 5.4 Hz, 2H), 2.60 (d, J = 7.1 Hz, 2H), 2.39–2.37 (m, 1H), 1.83–1.55 (m, 4H); 13C NMR
(151 MHz, DMSO-d6) δ 159.2, 154.6, 150.3, 145.8, 132.0, 128.7, 128.5, 128.3, 128.1, 125.5,
120.1, 117.3, 111.1, 108.4, 106.4, 106.0, 45.5, 31.9, 27.0, 25.3; HRMS (ESI-MS): Calculated for
C23H22N4O3Cl [M+H]+: 437.13749 and 439.13454; found: 437.13815 and 439.13526.
Preparation of 8-nitro-N-phenethyl-2-phenylquinazolin-4-amine (8k). According to
the general procedure, quinazoline 7b (268 mg, 1.00 mmol) was converted and the crude prod-
uct purified by silica gel column chromatography (30% EtOAc/PE) to the substituted amino-
quinazoline 8k (324 mg, 87%). 1H NMR (500 MHz, DMSO-d6) δ 8.86 (t, J = 5.4 Hz, 1H), 8.49
(d, J = 8.3 Hz, 1H), 8.46–8.42 (m, 2H), 8.24 (d, J = 7.6 Hz, 1H), 7.60 (t, J = 7.9 Hz, 1H), 7.55–
7.51 (m, 3H), 7.35–7.29 (m, 4H), 7.24–7.18 (m, 1H), 3.93 (dd, J = 14.1, 6.3 Hz, 2H), 3.08 (m,
2H); 13C NMR (101 MHz, DMSO-d6) δ 161.0, 159.2, 147.0, 141.9, 139.4, 137.7, 131.0, 128.8,
128.5, 128.5, 128.2, 126.7, 126.5, 126.2, 124.2, 114.9, 42.6, 34.4; HRMS (ESI-MS): Calculated
for C22H19N4O2 [M+H]+: 371.15025; found: 371.15018.
Preparation of 1-(4-(4-(7-nitro-4-(phenethylamino)quinazolin-2-yl)phenyl)piperazin-
1-yl)ethan-1-one (8l). According to the general procedure, quinazoline 3c (366 mg, 0.74
mmol) was converted and the crude product purified by flash chromatography (0% ! 10%
MeOH/DCM) to the substituted aminoquinazoline 8l (329 mg, 69%). 1H NMR (500 MHz,
DMSO-d6) δ 8.73 (t, J = 5.3 Hz, 1H), 8.47–8.33 (m, 4H), 8.12 (dd, J = 8.9, 2.3 Hz, 1H), 7.32 (d,
J = 4.4 Hz, 4H), 7.27–7.17 (m, 1H), 7.07 (d, J = 9.0 Hz, 2H), 3.88 (dd, J = 14.3, 6.1 Hz, 2H), 3.61
(s, 4H), 3.34 (dd, J = 10.1, 5.3 Hz, 2H), 3.30–3.27 (m, 2H), 3.10–3.02 (m, 2H), 2.06 (s, 3H); 13
C
NMR (126 MHz, DMSO-d6) δ 168.3, 161.3, 159.0, 152.3, 150.8, 150.3, 150.0, 139.4, 129.4,
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
7 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
128.7,128.4, 127.8, 126.1, 125.0, 122.4, 117.5,117.2, 114.2, 47.4, 47.1, 45.2, 42.5, 40.5, 34.3, 21.1;
HRMS (ESI-MS): Calculated for C28H29N6O3 [M+H]+: 497.22957; found: 497.22855.
Preparation of 2-(4-morpholinophenyl)-7-nitro-N-phenethylquinazolin-4-amine
(8m). According to the general procedure, quinazoline 7d (490 mg, 1.22 mmol) was con-
verted and the crude product purified by flash chromatography (0% ! 10% MeOH/DCM) to
the substituted aminoquinazoline 8m (268 mg, 48%). 1H NMR (400 MHz, DMSO-d6) δ 8.89–
8.71 (m, J = 5.0 Hz, 1H), 8.52–8.35 (m, 4H), 8.15 (dd, J = 8.9, 2.3 Hz, 1H), 7.33 (d, J = 4.3 Hz,
4H), 7.27–7.16 (m, J = 8.6, 4.3 Hz, 1H), 7.08 (d, J = 8.9 Hz, 2H), 3.96–3.84 (m, J = 13.0, 6.7 Hz,
2H), 3.77 (t, J = 4.0 Hz, 4H), 3.30–3.21 (m, J = 4.3 Hz, 4H), 3.06 (m, 2H); 13C NMR (126 MHz,
DMSO-d6) δ 162.2, 158.5, 153.7, 150.9, 140.4, 130.3, 130.2, 129.8, 129.4, 129.1, 128.8, 127.0,
125.9, 123.3, 114.7, 66.9, 48.4, 43.4, 35.2; HRMS (ESI-MS): Calculated for C26H26N5O3
[M+H]+: 456.20302; found 456.20230.
Preparation of 2-(4-(methylsulfinyl)phenyl)-7-nitro-N-phenethylquinazolin-4-amine
(18). Following the general procedure, quinazoline 7e (400 mg, 1.27 mmol) was converted
and the crude product 8n was subsequently used without any further purification. To a stirred
solution of nitroquinazoline 8n (500 mg, 1.20 mmol) in DCM (10 mL) mCPBA (800 mg, 4.80
mmol) was added in small portions at 0˚C. The reaction mixture was allowed to warm up to rt
and was stirred for 3 h. After quenching with saturated NaHCO3:NaS2O3 (1:1) and extraction,
the combined organic layers were concentrated under reduced pressure. The described com-
pound 18 was obtained after precipitation in cold H2O (478 mg, 87%). 1H NMR (500 MHz,
DMSO-d6) δ 8.96 (t, J = 5.4 Hz, 1H), 8.68 (d, J = 8.5 Hz, 2H), 8.50–8.45 (m, 2H), 8.23 (dd,
J = 8.9, 2.4 Hz, 1H), 8.09 (d, J = 8.5 Hz, 2H), 7.36–7.29 (m, 4H), 7.24–7.19 (m, 1H), 3.91 (dd,
J = 14.1, 6.3 Hz, 2H), 3.06 (t, J = 7.5 Hz, 2H); 13C NMR (126 MHz, DMSO-d6) δ 159.7, 159.4,
150.1, 149.8, 142.4, 142.3, 139.3, 128.7, 128.4, 127.1, 126.2, 125.2, 122.9, 119.1, 117.6, 43.5, 42.6,
34.3; HRMS (ESI-MS): Calculated for C23H21N4O3S [M+H]+: 433.13289; found 433.13260.
Common procedure for the generation of 7-/8-aminoquinazolines 9a-m
The corresponding nitroquinazoline was reduced with 10% Pd/C (0.2 mol-%) and ammonium
formate (6 eq) in EtOH for 2 h at 80˚C. The hot reaction mixture was filtered through Celite
and concentrated under reduced pressure. Sulphur-containing nitroquinazolines were alterna-
tively reduced using Fe (5 eq) and NH4Cl (8 eq) in MeOH:H2O (4:1) for 3–4 h at 80˚C. The
crude reaction mixture was partitioned between NaHCO3 and DCM and the combined
organic layers were concentrated in vacuo.
Preparation of N4-benzyl-2-phenylquinazoline-4,7-diamine (9a). According to the
common procedure, quinazoline 8a (100 mg, 0.30 mmol) was converted and the crude prod-
uct was purified by silica gel column chromatography (1% ! 10% MeOH/DCM) to yield the
amine 9a (90 mg, 93%). 1H NMR (500 MHz, DMSO-d6) δ 8.51 (s, J = 40.7 Hz, 1H), 8.40–8.33
(m, 2H), 7.98 (d, J = 12.2 Hz, 1H), 7.44 (m, 5H), 7.31 (t, J = 7.5 Hz, 2H), 7.22 (t, J = 7.3 Hz,
1H), 6.80 (dd, J = 8.8, 1.9 Hz, 1H), 6.78–6.75 (m, J = 1.9 Hz, 1H), 5.94 (s, 2H), 4.87 (d, J = 5.7
Hz, 2H); 13C NMR (126 MHz, DMSO-d6) δ 159.0, 158.8, 152.9, 140.3, 129.8, 128.2, 128.1,
127.7, 127.4, 126.6, 123.7, 116.0, 105.6, 104.3, 43.6; HRMS (ESI-MS): Calculated for C21H19N4
[M+H]+: 327.16042; found 327.16056.
Preparation of N-(4-fluorophenyl)-7-nitro-2-phenylquinazolin-4-amine (9b). Follow-
ing the common procedure, quinazoline 8b (513 mg, 1.42 mmol) was converted and the crude
product was purified by silica gel column chromatography (2% MeOH/DCM) to yield the
amine 9b (264 mg, 56%). 1H NMR (300 MHz, DMSO-d6) δ 9.48 (s, 1H), 8.43–8.29 (m, 2H),
8.19 (d, J = 9.0 Hz, 1H), 8.01–7.82 (m, 2H), 7.55–7.39 (m, 3H), 7.35–7.19 (m, 2H), 6.90 (dd,
J = 8.9, 2.1 Hz, 1H), 6.81 (d, J = 2.1 Hz, 1H), 6.06 (s, 2H); 13C NMR (75 MHz, DMSO-d6)
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
8 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
δ 158.8, 158.0 (d, J = 239.6 Hz), 157.2, 153.2, 152.5, 138.7, 136.3, 129.9, 128.3, 127.7, 124.0,
123.6 (d, J = 7.8 Hz), 116.4, 115.0 (d, J = 22.1 Hz), 106.0, 104.6; HRMS (ESI-MS): Calculated
for C20H16N4F [M+H]+: 331.13535; found 331.13574.
Preparation of N4-(4-fluorobenzyl)-2-phenylquinazoline-4,7-diamine (9c). According
to the common procedure, quinazoline 8c (280 mg, 0.75 mmol) was converted and the crude
product was purified by silica gel column chromatography (0.5% ! 2% MeOH/DCM) to yield
the amine 9c (157 mg, 61%). 1H NMR (300 MHz, DMSO-d6) δ 8.46 (m, 1H), 8.42–8.34 (m,
2H), 7.95 (d, J = 8.8 Hz, 1H), 7.59–7.29 (m, 5H), 7.14 (t, J = 8.9 Hz, 2H), 6.79 (dd, J = 8.8, 2.2
Hz, 1H), 6.74 (d, J = 2.1 Hz, 1H), 5.93 (s, 2H), 4.83 (d, J = 5.7 Hz, 2H); 13C NMR (75 MHz,
DMSO-d6) δ 161.1 (d, J = 242.0 Hz), 159.0, 158.9, 152.9, 138.8, 136.6, 129.8, 129.3 (d, J = 8.1
Hz), 128.1, 127.7, 123.7, 115.7, 115.0 (d, J = 21.2 Hz), 105.9, 104.5, 42.9; HRMS (ESI-MS):
Calculated for C21H18N4F [M+H]+: 345.15100; found 345.15117.
Preparation of N4-(3-fluorophenyl)-2-phenylquinazoline-4,7-diamine (9d). Following
the common procedure, quinazoline 8d (68 mg, 0.19 mmol) was converted and the crude
product was purified by silica gel column chromatography (1% ! 5% MeOH/DCM) to yield
the amine 9d (60 mg, 96%). 1H NMR (300 MHz, DMSO-d6) δ 9.67 (s, 1H), 8.45–8.32 (m, 2H),
8.24 (d, J = 9.0 Hz, 1H), 8.01 (d, J = 12.2 Hz, 1H), 7.74 (d, J = 8.4 Hz, 1H), 7.60–7.47 (m, 3H),
7.46–7.39 (m, 1H), 6.92 (td, J = 8.7, 2.2 Hz, 2H), 6.85 (d, J = 2.0 Hz, 1H), 6.19 (s, 2H); 13
C
NMR (75 MHz, DMSO-d6) δ 162.1 (d, J = 240.3 Hz), 158.6, 157.1, 153.5, 141.8 (d, J = 10.8 Hz),
130.1 (d, J = 18.66 Hz), 129.9, 128.4, 127.8, 124.2, 117.1, 116.7, 109.1 (d, J = 19.4 Hz), 108.1 (d,
J = 25.4 Hz), 104.6; HRMS (ESI-MS): Calculated for C20H16N4F [M+H]+: 331.13535; found
331.13580.
Preparation of N4-(3,4-difluorophenyl)-2-phenylquinazoline-4,7-diamine (9e).
According to the common procedure, quinazoline 8e (59 mg, 0.16 mmol) was converted and
the crude product was purified by silica gel column chromatography (0% ! 1% MeOH/
DCM) to yield the amine 9e (36 mg, 66%). 1H NMR (400.1 MHz, DMSO-d6) δ 9.57 (s, 1H),
8.38 (d, J = 7.8 Hz, 2H), 8.20 (d, J = 8.8 Hz, 2H), 7.71 (m, 1H), 7.48 (m, 4H), 6.92 (dd, J = 8.9
Hz, 1H), 6.83 (s, 1H), 6.11 (s, 2H); 13C NMR (100.6 MHz, DMSO-d6) δ 159.6, 157.7, 154.1,
163.6, 139.6, 138.1, 138.0, 130.8, 129.2, 128.5, 124.7, 118.3, 117.9, 117.8, 117.4, 111.1 (d, J = 21.6
Hz), 107.0, 105.5; HRMS (ESI-MS): Calculated for C20H15N4F2 [M+H]+: 349.12593; found
349.12594.
Preparation of N4-(3,4-difluorobenzyl)-2-phenylquinazoline-4,7-diamine (9f). Follow-
ing the common procedure, quinazoline 8f (53 mg, 0.13 mmol) was converted and the crude
product was purified by silica gel column chromatography (0% ! 10% MeOH/DCM) to yield
the amine 9f (35 mg, 72%). 1H NMR (500 MHz, DMSO-d6) δ 8.49 (s, 1H), 8.36 (dd, J = 6.6 Hz,
3.0 Hz, 2H), 8.14 (s, 1H), 7.95 (d, J = 8.8 Hz, 1H), 7.45 (m, 4H), 7.37 (dd, J = 19.2 Hz, 8.5 Hz,
1H), 7.28 (m, 1H), 6.82 (dd, J = 8.8 Hz, 1.7 Hz, 1H), 6.77 (s, 1H), 5.94 (s, 2H), 4.82 (d, J = 5.7
Hz, 2H); 13C NMR (101 MHz, DMSO-d6) δ 163.2, 158.9, 158.8, 153.0, 138.5, 138.4, 129.9,
128.2, 127.8, 124.0, 123.8, 117.4, 117.2, 116.4, 116.2, 115.9, 105.6, 104.3, 42.8; HRMS (ESI-MS):
Calculated for C21H17N4F2 [M+H]+: 363.14158; found: 363.14154.
Preparation of 2-phenyl-N4-(thiophen-2-ylmethyl)quinazoline-4,7-diamine (9g).
According to the common procedure, quinazoline 8g (59 mg, 0.16 mmol) was converted and
the crude product was purified by silica gel column chromatography (1% ! 10% MeOH/
DCM) to yield the amine 9g (45 mg, 83%). 1H NMR (500 MHz, DMSO-d6) δ 8.97 (s, 1H), 8.47
(dd, J = 6.6 Hz, 3.0 Hz, 2H), 7.98 (d, J = 9.4 Hz, 1H), 7.53 (m, 3H), 7.35 (dd, J = 5.1 Hz, 0.9 Hz,
1H), 7.13 (d, J = 2.9 Hz, 1H), 6.96 (dd, J = 5.0 Hz, 3.5 Hz, 1H), 6.83 (d, J = 6.6 Hz, 2H), 6.19 (s,
2H), 5.02 (d, J = 5.8 Hz, 2H); 13C NMR (101 MHz, DMSO-d6) δ 159.4, 158.9, 154.6, 143.3,
131.7, 129.2, 127.5, 127.3, 126.2, 125.1, 117.0, 104.3, 49.45; HRMS (ESI-MS): Calculated for
C19H17N4S [M+H]+: 333.11684; found: 333.11689.
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
9 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Preparation of 2-phenyl-N4-(2-(thiophen-2-yl)ethyl)quinazoline-4,7-diamine (9h).
Following the common procedure, quinazoline 8h (60 mg, 0.16 mmol) was converted and the
crude product was purified by silica gel column chromatography (0% ! 2% MeOH/DCM) to
yield the amine 9h (46 mg, 84%). 1H NMR (500 MHz, DMSO-d6) δ 9.76 (s, 1H), 8.33 (m, 2H),
8.17 (d, J = 9.1 Hz, 1H), 7.72 (t, J = 7.3 Hz, 1H), 7.66 (t, J = 7.5 Hz, 2H), 7.34 (dd, J = 5.0 Hz,
1.2 Hz, 1H), 7.03 (d, J = 2.1 Hz, 1H), 6.95 (m, 5H), 3.97 (dd, J = 12.9 Hz, 7.0 Hz, 2H), 3.27
(t, J = 7.2, 2H); 13C NMR (126 MHz, DMSO-d6) δ 158.8, 156.4, 155.3, 140.8, 133.0, 128.8,
128.7, 127.0, 125.5, 125.4, 124.3, 117.0, 101.2, 42.8; HRMS (ESI-MS): Calculated for
C20H19N4S [M+H]+: 347.13249; found: 347.13256.
Preparation of 4-(2-(4-fluorophenyl)-4,5-dihydro-1H-imidazol-1-yl)-2-phenylquinazo-
lin-7-amine (9i). According to the common procedure, quinazoline 8i (80 mg, 0.20 mmol)
was converted and the crude product was purified by silica gel column chromatography
(0% ! 1% MeOH/DCM) to yield the amine 9i (61 mg, 81%). 1H NMR (500 MHz, DMSO-d6)
δ 7.91 (m, 1H), 7.72 (d, J = 7.3 Hz, 2H), 7.61 (dd, J = 8.6 Hz, 5.6 Hz, 2H), 7.35 (t, J = 7.3 Hz,
1H), 7.27 (dd, J = 14.3 Hz, 6.6 Hz, 2H), 7.17 (t, J = 8.8 Hz, 2H), 6.99 (dd, J = 9.0 Hz, 2.1 Hz,
1H), 6.84 (d, J = 2.1 Hz, 1H), 6.30 (s, 2H), 4.27 (t, J = 8.7 Hz, 2H), 4.04 (t, J = 8.7 Hz, 2H); 13
NMR (126 MHz, DMSO-d6) δ 163.6, 161.7, 160.7, 157.8, 154.5, 153.8, 137.8, 129.9, 129.7,
129.7, 127.9, 127.3, 126.2, 117.7, 115.1, 115.0, 114.9, 114.8, 107.9, 105.1, 55.1, 53.3; HRMS
(ESI-MS): Calculated for C23H19N5F [M+H]+: 384.16190; found: 384.16173.
C
Preparation of N4-((4-(cyclopropylmethyl)furan-2-yl)methyl)-2-phenylquinazoline-
4,7-diamine (9j). Following the common procedure, quinazoline 8j (135 mg, 0.31 mmol)
was converted and the crude product was purified by silica gel column chromatography
(0.5% ! 5% MeOH/DCM) to yield the amine 9j as a hydrochloride salt (99 mg, 87%). 1H
NMR (600 MHz, DMSO-d6) δ 8.46 (dd, J = 17.5, 6.8 Hz, 2H), 7.95 (dd, J = 25.3, 8.7 Hz, 1H),
7.58–7.39 (m, 3H), 6.77 (d, J = 8.7 Hz, 1H), 6.74 (s, 1H), 6.22 (d, J = 2.6 Hz, 1H), 6.01 (d, J = 3.0
Hz, 1H), 5.87 (br, 1H), 4.76 (d, J = 5.0 Hz, 2H), 3.63–3.57 (m, 2H), 2.59 (d, J = 7.1 Hz, 2H),
2.40–2.37 (m, 1H), 1.79–1.62 (m, 4H); 13C NMR (151 MHz, DMSO-d6) δ 159.3, 154.6, 151.9,
139.8, 130.1, 128.6, 128.3, 124.2, 116.2, 108.0, 106.3, 45.5, 40.5, 31.9, 27.1, 25.3; HRMS (ESI-
MS): Calculated for C23H24N4OCl [M+H]+: 407.16332 and 409.16037; found: 407.16399 and
409.16100.
Preparation of N4-phenethyl-2-phenylquinazoline-4,8-diamine (9k). Following the
common procedure, quinazoline 8k (370 mg, 1.00 mmol) was converted and the crude prod-
uct was purified by silica gel column chromatography (25% ! 35% EtOAc/PE) to yield the
amine 9k (271 mg, 80%). 1H NMR (500 MHz, DMSO-d6) δ 8.59 (d, J = 8.2 Hz, 2H), 8.11
(t, J = 5.2 Hz, 1H), 7.55–7.42 (m, 3H), 7.35–7.30 (m, 4H), 7.29 (d, J = 8.4 Hz, 1H), 7.24–7.18
(m, 1H), 7.16 (t, J = 7.9 Hz, 1H), 6.90 (d, J = 7.6 Hz, 1H), 5.88 (s, 2H), 3.87 (dd, J = 13.7, 6.7 Hz,
2H), 3.06 (t, J = 7.5 Hz, 2H); 13C NMR (126 MHz, DMSO-d6) δ 159.5, 156.4, 144.8, 139.7,
138.9, 138.4, 129.6, 128.7, 128.4, 128.1, 127.7, 126.0, 125.8, 113.8, 112.0, 108.0, 42.4, 34.7;
HRMS (ESI-MS): Calculated for C22H21N4 [M+H]+: 341.17607; found: 341.17608.
Preparation of 1-(4-(4-(7-amino-4-(phenethylamino)quinazolin-2-yl)phenyl)piperazin-
1-yl)ethan-1-one (9l). According to the common procedure, quinazoline 8l (268 mg, 0.54
mmol) was converted and the crude product was purified by flash chromatography (2% !
10% MeOH/DCM) to yield the amine 9l (147 mg, 58%). 1H NMR (500 MHz, DMSO-d6) δ
13.26 (s, 1H), 9.60 (t, J = 5.4 Hz, 1H), 8.32 (t, J = 15.3 Hz, 2H), 8.14 (d, J = 9.1 Hz, 1H), 7.30
(d, J = 4.4 Hz, 4H), 7.23–7.17 (m, 1H), 7.11 (d, J = 9.1 Hz, 2H), 7.05 (d, J = 1.9 Hz, 1H), 6.86
(dd, J = 9.0, 2.0 Hz, 1H), 6.79 (s, 1H), 3.91 (dt, J = 13.9, 6.4 Hz, 2H), 3.64–3.58 (m, 4H), 3.47
(t, J = 4.9 Hz, 2H), 3.40 (s, 2H), 3.03 (t, J = 7.5 Hz, 2H), 2.06 (s, 3H); 13C NMR (126 MHz,
DMSO-d6) δ 168.4, 158.5, 155.7, 155.1, 153.5, 141.2, 139.0, 130.4, 128.7, 128.4, 126.2, 125.3,
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
10 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
119.2, 116.0, 113.4, 100.9, 97.4, 46.5, 46.2, 44.9, 42.6, 40.3, 34.8, 21.2; HRMS (ESI-MS): Calcu-
lated for C28H31N6O [M+H]+: 467.25539; found: 467.25443.
Preparation of 2-(4-morpholinophenyl)-N4-phenethylquinazoline-4,7-diamine (9m).
Following the common procedure, quinazoline 8m (21 mg, 0.05 mmol) was converted and the
crude product was purified by flash chromatography (2% ! 10% MeOH/DCM) to yield the
amine 9m (13 mg, 66%). 1H NMR (400 MHz, CDCl3) δ 8.46 (d, J = 8.9 Hz, 2H), 7.27 (d, J = 9.7
Hz, 6H), 6.98 (dd, J = 6.7, 4.2 Hz, 3H), 6.67 (dd, J = 8.7, 2.2 Hz, 1H), 5.93 (s, 1H), 4.11 (br, 2H),
3.97 (dd, J = 12.9, 6.9 Hz, 2H), 3.92–3.86 (m, 4H), 3.31–3.24 (m, 4H), 3.06 (t, J = 7.1 Hz, 2H);
13C NMR (101 MHz, CDCl3) δ 159.0, 152.8, 150.5, 143.5, 139.6, 129.8, 129.1, 128.8, 126.6,
122.4, 115.4, 114.6, 77.4, 67.0, 48.7, 42.5, 35.7; HRMS (ESI-MS): Calculated for C26H28N5O
[M+H]+: 426.22884; found: 426.22857.
Preparation of 2-(4-(methylsulfinyl)phenyl)-N4-phenethylquinazoline-4,7-diamine
(19). According to the common procedure, quinazoline 18 (478 mg, 1.11 mmol) was con-
verted and the crude product was purified by flash chromatography (0% ! 6% MeOH/DCM)
to yield the amine 19 (272 mg, 61%). 1H NMR (500 MHz, CDCl3) δ 8.73 (d, J = 8.4 Hz, 2H),
8.03 (d, J = 8.4 Hz, 2H), 7.41–7.30 (m, 4H), 7.28 (d, J = 7.7 Hz, 3H), 7.03 (d, J = 2.0 Hz, 1H),
6.79 (dd, J = 8.7, 2.0 Hz, 1H), 5.61 (s, 1H), 4.17 (s, 2H), 3.07 (s, 3H), 3.06 (s, 2H); 13C NMR
(126 MHz, CDCl3) δ 159.4, 159.2, 152.3, 150.7, 144.8, 141.1, 139.3, 129.3, 129.0, 128.9, 127.3,
126.8, 122.2, 116.7, 109.4, 106.7, 44.8, 42.5, 35.7; HRMS (ESI-MS): Calculated for C23H23N4OS
[M+H]+: 403.15871; found: 403.15755.
Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-7-yl)propionamide (10a).
To a stirred solution of 3 (40 mg, 0.12 mmol) in DIPEA (40 μL, 0.24 mmol) and THF (5 mL)
at 0˚C propionyl chloride (12.3 μL in 1 mL THF, 0.14 mmol) was added dropwise. The reac-
tion mixture was allowed to warm up to rt and was stirred for another 2 h. After extraction
with saturated NaHCO3, concentration of the combined organic layers in vacuo and silica gel
column chromatography (0% ! 1% MeOH/DCM) 10a (31 mg, 66%) was obtained. 1H NMR
(500 MHz, DMSO-d6) δ 10.19 (s, 1H), 8.50 (dd, J = 8.0, 1.6 Hz, 2H), 8.29 (t, J = 5.4 Hz, 1H),
8.13 (dd, J = 5.4, 3.2 Hz, 2H), 7.58 (dd, J = 8.9, 2.0 Hz, 1H), 7.54–7.45 (m, 3H), 7.36–7.29 (m,
4H), 7.25–7.18 (m, 1H), 3.86 (dt, J = 14.5, 6.0 Hz, 2H), 3.06 (m, 2H), 2.41 (q, J = 7.5 Hz, 2H),
1.13 (t, J = 7.5 Hz, 3H); 13C NMR (126 MHz, DMSO-d6) δ 172.6, 159.6, 159.1, 151.0, 142.9,
139.7, 138.8, 129.9, 128.7, 128.4, 128.2, 127.8, 126.1, 123.2, 117.7, 114.9, 109.5, 42.3, 34.7, 29.7,
9.5; HRMS (ESI-MS): Calculated for C25H25N4O [M+H]+: 397.20229; found: 397.20133.
Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-7-yl)propane-1-sulfon-
amide (10b). To a stirred solution of 3 (75 mg, 0.22 mmol) in pyridine (19.6 μL, 0.24 mmol)
and DCM (5 mL) at 0˚C propane-1-sulfonyl chloride (27.2 μL in 1 mL DCM, 0.24 mmol) was
added dropwise. The reaction mixture was then heated to 50˚C and was stirred for another 6
h. After quenching with 6 N NaOH and extraction, the combined organic layers were concen-
trated in vacuo. Silica gel column chromatography (0.5% MeOH/DCM) yielded 10b (68 mg,
80%). 1H NMR (500 MHz, DMSO-d6) δ 10.31 (s, 1H), 8.49 (dd, J = 7.7, 1.8 Hz, 2H), 8.36 (s,
1H), 8.16 (d, J = 8.9 Hz, 1H), 7.58–7.43 (m, 4H), 7.34–7.31 (m, 4H), 7.29 (dd, J = 8.9, 2.1 Hz,
1H), 7.25–7.18 (m, 1H), 3.87 (dd, J = 14.3, 6.1 Hz, 2H), 3.22 (m, 2H), 3.05 (m, 2H), 1.79–1.67
(m, 2H), 0.95 (t, J = 7.5 Hz, 3H); 13C NMR (101 MHz, DMSO-d6) δ 160.0, 159.3, 151.1, 142.4,
139.7, 138.6, 130.2, 128.8, 128.5, 128.3, 127.9, 126.2, 124.2, 117.0, 113.7, 109.7, 52.7, 42.4, 34.7,
17.0, 12.5; HRMS (ESI-MS): Calculated for C25H27N4O2S [M+H]+: 447.18492; found:
447.18460.
Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-7-yl)benzenesulfonamide
(10c). To a stirred solution of 3 (75 mg, 0.22 mmol) in pyridine (19.6 μL, 0.24 mmol) and
DCM (5 mL) at 0˚C benzenesulfonyl chloride (31 μL in 1 mL DCM, 0.24 mmol) was added
dropwise. The reaction mixture was then heated to 50˚C and was stirred for another 6 h. After
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
11 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
quenching with 6 N NaOH and extraction, the combined organic layers were concentrated in
vacuo. Silica gel column chromatography (30% ! 35% EtOAc/PE) yielded 10c (32 mg, 69%).
1H NMR (500 MHz, DMSO-d6) δ 10.87 (s, 1H), 8.47–8.41 (m, 2H), 8.08 (d, J = 8.9 Hz, 1H),
7.89–7.85 (m, 2H), 7.65–7.54 (m, 3H), 7.53–7.45 (m, 3H), 7.41 (d, J = 1.8 Hz, 1H), 7.35–7.27
(m, 4H), 7.26–7.17 (m, 2H), 5.74 (s, 1H), 3.83 (dd, J = 14.5, 6.1 Hz, 2H), 3.01 (m, 2H); 13
C
NMR (101 MHz, DMSO-d6) δ 159.9, 159.2, 154.5, 147.2, 141.7, 139.6, 139.3, 133.3, 130.3,
129.5, 128.7, 128.5, 128.3, 127.9, 126.7, 126.2, 124.1, 117.4, 114.6, 110.0, 42.4, 34.7; HRMS
(ESI-MS): Calculated for C28H25N4O2S [M+H]+: 481.16927; found: 481.16906.
Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-8-yl)propane-1-sulfon-
amide (10d). To a stirred solution of 9k (60 mg, 0.18 mmol) in pyridine (2 drops) and DCM
(5 mL) at 0˚C propane-1-sulfonyl chloride (23.4 μL in 1 mL DCM, 0.19 mmol) was added
dropwise. The reaction mixture was then heated to 50˚C and was stirred for another 6 h. After
quenching with 6 N NaOH and extraction, the combined organic layers were concentrated in
vacuo. Silica gel column chromatography (10% ! 20% EtOAc/PE) yielded 10b (65 mg, 83%).
1H NMR (500 MHz, DMSO-d6) δ 9.24 (s, 1H), 8.63 (dd, J = 7.8, 1.5 Hz, 2H), 8.57 (t, J = 5.4 Hz,
1H), 7.97 (d, J = 8.3 Hz, 1H), 7.76 (d, J = 7.6 Hz, 1H), 7.58–7.48 (m, 3H), 7.45 (t, J = 8.0 Hz,
1H), 7.36–7.28 (m, 4H), 7.24–7.17 (m, 1H), 3.90 (dd, J = 14.3, 6.2 Hz, 2H), 3.23–3.15 (m, 2H),
3.10–3.04 (m, 2H), 1.82–1.71 (m, 2H), 0.85 (t, J = 7.4 Hz, 2H); 13C NMR (126 MHz, DMSO-
d6) δ 159.5, 158.8, 141.4, 139.5, 138.2, 133.5, 130.3, 128.7, 128.4, 128.2, 128.2, 126.1, 125.0,
121.0, 117.6, 113.8, 53.0, 42.5, 34.5, 16.8, 12.5; HRMS (ESI-MS): Calculated for C25H27N4O2S
[M+H]+: 447.18492; found: 447.18435.
Preparation of N-(4-(phenethylamino)-2-phenylquinazolin-8-yl)benzenesulfonamide
(10e). To a stirred solution of 3 (60 mg, 0.18 mmol) in pyridine (2 drops) and DCM (5 mL)
at 0˚C benzenesulfonyl chloride (26.7 μL in 1 mL DCM, 0.19 mmol) was added dropwise. The
reaction mixture was then heated to 50˚C and was stirred for another 6 h. After quenching
with 6 N NaOH and extraction, the combined organic layers were concentrated in vacuo. Silica
gel column chromatography (10% ! 20% EtOAc/PE) yielded 10c (41 mg, 78%). 1H NMR
(300 MHz, DMSO-d6) δ 9.87 (s, 1H), 8.60 (m, 2H), 8.50 (t, J = 5.4 Hz, 1H), 7.95–7.84 (m, 3H),
7.71 (d, J = 7.1 Hz, 1H), 7.60–7.46 (m, 4H), 7.47–7.34 (m, 3H), 7.34–7.24 (m, 4H), 7.24–7.15
(m, 1H), 3.84 (dd, J = 14.3, 6.1 Hz, 2H), 3.02 (m, 2H); 13C NMR (75 MHz, DMSO-d6) δ 159.4,
158.7, 141.6, 139.6, 138.2, 133.0, 132.8, 130.4, 129.1, 128.7, 128.5, 128.4, 128.2, 126.8, 126.2,
124.8, 121.3, 118.0, 113.7, 42.5, 34.5; HRMS (ESI-MS): Calculated for C28H25N4O2S [M+H]+:
481.16927; found: 481.16884.
Preparation of N-(furan-2-ylmethyl)acetamide (12). To a solution of furan-2-ylmetha-
namine 11 (920 μL, 10.4 mmol) in DCM (50 mL) Ac2O (2.9 mL, 31 mmol) and TEA (2.86 g,
20.6 mmol) were added and the reaction mixture was stirred overnight at rt. The solvent was
removed under reduced pressure and after silica gel column chromatography (5% ! 50%
EtOAc/PE) 12 (1.26 g, 88%) was obtained. 1H NMR (300 MHz, DMSO-d6) δ 8.30 (s, 1H), 7.56
(d, J = 1.0 Hz, 1H), 6.38 (dd, J = 3.1 Hz, 1.9 Hz, 1H), 6.22 (d, J = 2.8 Hz, 1H) 4.23 (d, J = 5.7 Hz,
2H), 1.83 (s, 3H); 13C NMR (126 MHz, DMSO-d6) δ 169.0, 152.4, 142.0, 110.4, 106.7, 35.4,
22.4; HRMS (ESI-MS): Calculated for C7H10NO2 [M+H]+: 140.07061; found: 140.07070.
Preparation of N-((4-(cyclopropanecarbonyl)furan-2-yl)methyl)acetamide (13). AlCl3
(2.88 g, 21.57 mmol) and cyclopropanecarbonyl chloride (850 μL, 9.35 mmol) were dissolved
in DCM (15 mL) at 0˚C and the mixture was stirred at rt for 30 min. Furan 12 (1 g, 7.19
mmol) was added the reaction was kept stirring at rt overnight. H2O and EtOAc were added
and after extraction the combined organic layers were concentrated in vacuo. Silica gel column
chromatography (1% ! 2% MeOH/DCM) yielded 13 (861 mg, 58%). 1H NMR (500 MHz,
DMSO-d6) δ 8.44 (s, 1H), 7.51 (d, J = 3.5 Hz, 1H), 6.48 (d, J = 3.5 Hz, 1H), 4.32 (d, J = 5.7 Hz,
2H), 2.64 (tt, J = 7.2, 5.4 Hz, 1H), 1.86 (s, 3H), 1.07–0.88 (m, 4H); 13C NMR (126 MHz,
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
12 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
DMSO-d6) δ 187.4, 169.3, 157.8, 151.4, 119.5, 109.4, 35.7, 22.4, 16.6, 10.4; HRMS (ESI-MS):
Calculated for C11H14NO3 [M+H]+: 208.09682, found: 208.09768.
Preparation of N-((4-(cyclopropylmethyl)furan-2-yl)methyl)acetamide (14). A solu-
tion of 13 (200 mg, 0.97 mmol) in concentrated HCl (5 mL) and stirred overnight at rt. The
reduced crude product (591 mg, 3.58 mmol) was used without any further purification. After
dissolving in TFA (11 mL), TES (1.374 mL, 8.6 mmol) was added dropwise and the resulting
mixture was stirred at rt overnight. H2O and cold saturated NaHCO3 were added to the reac-
tion that was subsequently thoroughly extracted with EtOAc. The combined organic layers
were concentrated in vacuo and silica gel column chromatography (1% ! 5% MeOH/DCM)
yielded the target compound 14 as a hydrochloride salt (620 mg, 92%). 1H NMR (600 MHz,
DMSO-d6) δ 8.28 (s, 3H), 6.41 (d, J = 3.1 Hz, 1H), 6.14 (d, J = 3.1 Hz, 1H), 3.66 (t, J = 6.4 Hz,
2H), 3.17 (s, 1H), 2.63 (t, J = 7.3 Hz, 2H), 1.73 (m, 5H); 13C NMR (151 MHz, DMSO-d6) δ
156.1, 145.8, 110.9, 106.3, 45.0, 35.2, 31.4, 26.5, 24.8; HRMS (ESI-MS): Calculated for
C9H15NOCl [M+H]+: 188.08367 and 190.08072; found: 188.08411 and 190.08146.
Preparation of 2-(4-fluorophenyl)-4,5-dihydro-1H-imidazole (16). 4-Fluorobenzalde-
hyde 15 (1.28 mL, 12.1 mmol) was dissolved in tBuOH (20 mL) and ethane-1,2-diamine (1.1
mL, 15.72 mmol) was added. After 30 min of stirring at rt, K2CO3 (2.5 g, 18.14 mmol) and I2
(9.2 g, 36.28 mL) were added and the resulting mixture was heated to 70˚C and stirred for
another 3.5 h. The reaction was quenched with saturated Na2SO3 and extracted with saturated
NaHCO3 and EtOAc. The combined organic layers were concentrated under reduced pressure
and silica gel column chromatography (4% ! 10% MeOH/DCM ! 6% MeOH/DCM + 1%
NH3(aq)) yielded 16 (1.1 g, 54%). 1H NMR (500 MHz, DMSO-d6) δ 7.92–7.85 (m, 2H), 7.48
(br, 1H), 7.28 (t, J = 8.9 Hz, 2H), 3.63 (s, 2H, 2H); 13C NMR (126 MHz, DMSO-d6) δ 163.4 (d,
J = 247.7 Hz), 162.7, 129. 6 (d, J = 8.8 Hz, 2C), 126.4 (d, J = 2.8 Hz), 115.2 (d, J = 21.7 Hz, 2C),
49.2; HRMS (ESI-MS): Calculated for C9H10N2F [M+H]+: 165.08225; found: 165.08209.
Preparation of 4-(4-acetylpiperazin-1-yl)benzaldehyde (17a). 4-Fluorobenzaldehyde
15 (619 μL, 5.85 mmol), 1-acetylpiperazine (500 μL, 3.90 mmol) and Na2CO3 (620 mg, 5.85
mmol) were dissolved in H2O (25 mL) and the stirred at 100˚C overnight. After extraction
with DCM, the combined organic layers were concentrated under reduced pressure and silica
gel column chromatography (3% MeOH/DCM) yielded 17a (942 mg, 58%). 1H NMR (500
MHz, DMSO-d6) δ 9.73 (s, 1H), 7.73 (d, J = 8.9 Hz, 2H), 7.04 (d, J = 8.9 Hz, 2H), 3.65–3.54 (m,
4H), 3.50–3.44 (m, 2H), 3.42–3.36 (m, 2H), 2.04 (s, 3H); 13C NMR (126 MHz, DMSO-d6) δ
190.2, 168.42, 154.3, 131.4, 126.4, 113.2, 46.3, 46.0, 44.9, 40.3, 21.1; HRMS (ESI-MS): Calcu-
lated for C13H17N2O2 [M+H]+: 233.12845; found: 233.12871.
Preparation of 4-morpholinobenzaldehyde (17b). 4-Fluorobenzaldehyde 15 (1.8 mL,
17.24 mmol), 1-acetylpiperazine (1 mL, 11.49 mmol) and Na2CO3 (1.83 g, 17.24 mmol) were
dissolved in H2O (40 mL) and the stirred at 100˚C overnight. After extraction with DCM, the
combined organic layers were concentrated under reduced pressure and silica gel column
chromatography (20% EtOAc/PE) yielded 17b (2.36 g, 55%). 1H NMR (400 MHz, DMSO-d6)
δ 9.97 (s, 1H), 9.73 (s, 1H), 8.11–7.89 (m, 2H), 7.72 (d, J = 8.9 Hz, 2H), 7.43 (q, J = 8.4 Hz, 2H),
7.03 (d, J = 8.9 Hz, 2H), 3.72 (m, 4H), 3.32 (m, 4H); 13C NMR (101 MHz, DMSO-d6) δ 191.0,
154.9, 131.7, 126.7, 118.3, 65.8, 46.6; HRMS (ESI-MS): Calculated for C11H14NO2 [M+H]+:
192.10191; found: 192.10186.
Preparation of 3-(3-(trifluoromethyl)phenyl)-1H-pyrazol-5-amine (21). 3-oxo-3-(3-
(trifluoromethyl)phenyl)propanenitrile 20 (1.2 g, 5.63 mmol) was dissolved in EtOH (10 mL)
and hydrazine (141 μL, 8.80 mmol) was added dropwise. The reaction mixture was stirred
for 1 h at rt and then overnight at reflux. After cooling down to rt, addition of H2O (twice the
reaction volume) led to the precipitation of 21 (630 mg, 50%) as a white solid that was subse-
quently filtered off and collected. In bigger scales occurring impurities could be removed by an
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
13 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
additional work up with NaHCO3/10% MeOH/DCM. 1H NMR (500 MHz, DMSO-d6) δ 11.89
(d, J = 220.0 Hz, 1H), 7.98 (s, J = 9.9 Hz, 1H), 7.94 (s, 1H), 7.59 (s, 2H), 5.85 (d, J = 67.9 Hz,
1H), 4.88 (d, J = 218.2 Hz, 2H).
Preparation of 4-(trifluoromethyl)-3-(3-(trifluoromethyl)phenyl)-1,7-dihydro-6H-pyr-
azolo[3,4-b]pyridin-6-one (2). Ethyl 4,4,4-trifluoro-3-oxobutanoate (139 μL, 0.95 mmol)
and 21 (180 mg, 0.79 mmol) were dissolved in AcOH (5 mL) and stirred overnight under
refluxing conditions. After cooling to rt, ice-water was added to the stirring reaction mixture
and led to precipitation of crude product that was subsequently filtered off. Silica gel column
chromatography (60% ! 100% EtOAc/PE) and washing of the concentrated collected frac-
tions with hexane yielded pure 2 (145 mg, 53%). 1H NMR (500 MHz, DMSO-d6) δ 13.82 (s,
1H), 12.41 (s, 1H), 7.88 (s, 1H), 7.84–7.77 (m, 2H), 7.76–7.69 (m, J = 7.2 Hz, 1H), 6.61 (s, 1H);
HRMS (ESI-MS): Calculated for C14H8N3OF6 [M+H]+: 348.05661; found: 348.05715.
Preparation of 4-cyclopropyl-3-(3-(trifluoromethyl)phenyl)-1,7-dihydro-6H-pyrazolo
[3,4-b]pyridin-6-one (22). Ethyl 3-cyclopropyl-3-oxopropanoate (131 μL, 0.85 mmol) and
21 (160 mg, 0.70 mmol) were dissolved in AcOH (5 mL) and stirred overnight under refluxing
conditions. After cooling to rt, ice-water was added to the stirring reaction mixture and led to
precipitation of crude product that was subsequently filtered off and washed with hexane to
yield 22 (113 mg, 51%). 1H NMR (500 MHz, DMSO-d6) δ 12.36 (s, 1H), 8.28 (d, J = 5.3 Hz,
2H), 7.77 (d, J = 7.8 Hz, 1H), 7.71 (t, J = 8.0 Hz, 1H), 6.71 (s, 1H), 5.46 (s, 1H), 1.96–1.91 (m,
1H), 1.14–1.06 (m, 2H), 1.03–0.96 (m, 2H); 13C NMR (126 MHz, DMSO-d6) δ 156.3, 156.1,
151.2, 142.8, 133.6, 130.1, 129.9, 129.5, 125.2, 122.2, 90.5, 86.0, 13.0, 9.2; HRMS (ESI-MS): Cal-
culated for C16H13N3OF3 [M+H]+: 320.10052; found: 320.10100.
Preparation of 4-methyl-3-(3-(trifluoromethyl)phenyl)-1,7-dihydro-6H-pyrazolo[3,4-
b]pyridin-6-one (23). Ethyl 3-oxobutanoate (110 μL, 0.85 mmol) and 21 (160 mg, 0.70
mmol) were dissolved in AcOH (5 mL) and stirred overnight under refluxing conditions.
After cooling to rt, ice-water was added to the stirring reaction mixture and led to precipitation
of crude product that was subsequently filtered off and washed with hexane to yield 23 (123
mg, 60%). 1H NMR (500 MHz, DMSO-d6) δ 12.44 (s, 1H), 8.29 (s, 2H), 7.80–7.74 (m, 1H),
7.74–7.68 (m, 1H), 6.75 (s, 1H), 5.63 (s, 1H), 2.32 (s, 3H); 13C NMR (126 MHz, DMSO-d6) δ
156.0, 151.3, 150.5, 142.9, 133.6, 130.1, 129.9, 125.3, 122.2, 95.4, 85.9, 18.6; HRMS (ESI-MS):
Calculated for C14H10N3OF3Na [M+Na]+: 316.06682; found: 316.06720.
Protein expression, purification and crystallization
The expression and purification of inactive, non-phosphorylated p38α wt MAPK was done as
previously reported [20]. Briefly, an N-terminal His6-p38α wt construct was transformed in to
E. coli BL21(DE3) and expressed overnight at 18˚C. The protein was purified by Ni2+-NTA-
affinity chromatography, followed by anion exchange and size exclusion chromatography after
removal of the His-tag by proteolytic cleavage. For SPR experiments the corresponding protein
batch did not undergo His-tag cleavage. The pure protein was subsequently concentrated to
10–30 mg/mL, aliquoted, flash frozen in liquid N2 and stored at -80˚C.
Various LiPoLis were co-crystallized with p38α wt using conditions similar to those as
described previously [14]. Briefly, protein-ligand complexes were prepared by mixing 40 μL
p38α wt (10 mg/mL) with 1 μL compound (50 mM in DMSO) and incubated for 60 min on
ice. The samples were centrifuged at 13,000 rpm for 10 min to remove excess ligand. Crystals
were grown in 24-well crystallization plates (EasyXtal Tool, Quagen, Hilden, Germany) using
the hanging drop vapor diffusion method and by mixing 1.5 μL protein-ligand solution with
0.5 μL reservoir (100 mM MES pH 5.6–6.2, 20–30% PEG4000 and 50 mM BOG). In some
cases, crystals were obtained when BOG was absent in the reservoir solution and when 125 μM
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
14 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
BIRB-769 were used instead of BOG, respectively. The crystals were protected using 25%
PEG400 before they were flash frozen in liquid N2. Diffraction data of the p38α-ligand com-
plexes were collected at the PX10SA beam line of the Swiss Light Source (PSI, Villigen, Swit-
zerland) using wavelengths close to 1 Å. The datasets were integrated with XDS [21] and
scaled using XSCALE [21]. The complex structures were solved by molecular replacement
with PHASER [22] using the published p38α structure of 4 (PDB: 4DLI) [13] as template. Mol-
ecules in the asymmetric unit were manually modified using the program COOT [23]. Inhibi-
tor topology files were generated using the Dundee PRODRG server [24]. Final refinement
was done employing the PDB-redo server [25] and PHENIX [26]. Refined structures were vali-
dated by Ramachandran plot analysis with RAMPAGE [27]. Data collection, structure refine-
ment statistics, PDB-entry codes and the Ramachandran plot results are shown in S2 Table.
Electron densities used for the omit maps were generated via a simulated annealing refinement
with ligand and water molecule occupancy set to”0”. PyMOL [28] was used to produce
Figs 1–4 and S5 Fig.
Fig 2. Compound design. (A) Design of LiPoLis based on the alignment of the crystal structures of 3 (green) and 1 (yellow) in
complex with p38α (PDB-codes: 4DLJ and 3HVC). Overlay of 1 and modeled structures of (B) 9c (cyan), (C) 9j (white) and (D)
9i (magenta) showcasing the proposed binding modes.
https://doi.org/10.1371/journal.pone.0184627.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
15 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 3. a Synthesis and identity of 2-arylquinazolines. a Reagents and conditions: (a) method A: benzamidine
hydrochloride hydrate, AcOH, 2-methoxyethanol, 130˚C, 18 h, 18–32%; method B: benzoic anhydride, formamide,
200˚C, 5 min, MW, 31–37%; (b) method C: aldehyde, NaHSO3, pTSA, DMAc, 155˚C, 18 h, 10–34%; (c) method A: 1)
SOCl2, DMF, 80˚C, 4 h, 2) amine, DIPEA, DCM/iPrOH (3:2), rt, 18 h, 66–95%; method B: 1) HCCP, DIPEA, MeCN, rt, 1
h, 2) amine, rt, 18 h, 68–87%; (d) method A: 10% Pd/C, ammonium formate, EtOH, 80˚C, 1–3 h, 56–96%; method B: Fe,
NH4Cl, MeOH:H2O (4:1), 80˚C, 3–6 h, 81–87%; (e) acyl chloride, DIPEA, DCM, 0˚C to rt or 50˚C, 1–6 h, 69–80%.
https://doi.org/10.1371/journal.pone.0184627.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
16 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 4. a Synthesis of amine building block 14. a Reagents and conditions: (a) Ac2O, DCM, rt, 18 h, 88%; (b) AlCl3,
cyclopropanecarbonyl chloride, DCM, rt, 30 min, 58%; (c) 1) HCl (conc.), rt, 18 h, 2) TFA, TES, rt, 18 h, 92%.
https://doi.org/10.1371/journal.pone.0184627.g004
Surface plasmon resonance
For the kinetic measurements a SPR-2/4 from Sierra Sensors (Hamburg, Germany) was used.
All experiments were performed with a constant flow rate of 25 μL/min of running buffer (1x
PBS, 150 mM NaCl, pH 7.4, 0.001% Triton™ X-100, 3% DMSO). At first, a trisNTA sensor was
made by standard amine coupling of amino-trisNTA [29] (100 mM) to a commercially avail-
able HCA sensor (Sierra Sensors) using a 1:1-cocktail of EDC/NHS (0.4 M/50 mM) for activa-
tion and ethanol amine (1 M, pH 8.5) for blocking of unreacted surfaces (reactions outside the
device; no flow). Then, the sensor was mounted to the SPR-2/4 and 50 μL of 200 μM Ni(II)
acetate were injected over a reference and an active spot on the sensor surface, followed by
subsequent injection of 50 μL inactive His6-p38α wt (2 mg/mL in running buffer) over the
active spot and 50 μL His6-peptide (MBL) (2 mg/mL in running buffer) over the reference
spot. After 30 min of baseline stabilisation due to dissociation of the His-tagged kinase from
the trisNTA surface an immobilization level of ca. 3000–4000 RU was reached, with a remain-
ing, negligible drift of ca. 2 RU/min.
A concentration series of LiPoLis and SB203580 (200 nM—6 μM, 1–30 μM and 5–500 nM,
respectively) along with blank injections for referencing, were injected over 2 and 5 min,
respectively, to both, the reference and the active spot (association), followed by washing with
running buffer for 1 min and 4 min, respectively (dissociation). No surface regeneration was
needed after compound injection. For solvent calibration 15 μL injections of 4.4–5.6% DMSO
in running buffer were run. Finally, multiple injections of 100 μL 0.5 M EDTA regenerated the
trisNTA sensor completely. Raw data was processed und globally fitted with Analyser R2
v0.2.3.8 (Sierra Sensors).
Thermal shift assay
Melting curves of p38α wt were measured at two compound concentrations (50, 10 μM). Com-
pounds in 10 mM DMSO stock solutions were diluted to 2.5 mM and 0.5 mM with DMSO,
respectively. From each of the dilution 0.8 μL were added to a well of a 96-well plate (TW-MT,
Biozym), except water and reference (DMSO) wells. A protein stock solution (30 mg/mL p38α
wt) was diluted to 1 μM in sample buffer (10 mM HEPES, 150 mM NaCl, 5x dye, pH 7.5),
already containing the dye SYPRO1 orange (1000x diluted 5000x stock in DMSO, Sigma) and
kept protected from light. To each well of the measurement plate were added 39 μL of this pro-
tein solution and the samples were appropriately mixed. After the plates were sealed with opti-
cal foil and spun down (200 g, rt, 1 min), they were subsequently placed in a LightCycler1 480
II (Roche). Experiment was carried out at 492 nm excitation and 610 nm emission wavelength
and the temperature ramping from 25˚C to 95˚C with a rate of 1˚C/min. ΔT was determined
from the difference to DMSO control samples.
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
17 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Results and discussion
Structure-based design and synthesis of a focused 4-amino-
2-arylquinazoline compound library
Starting from the crystal structures of 3 and 4 in complex with p38α (PDB-codes: 4DLI and
4DLJ) and some knowledge of the underlying binding mode, we employed a rational approach
for the design and synthesis of new 4-amino-2-arylquinazolines. As a major design aspect, we
made no changes to the main interactions contributing to ligand binding, namely the parallel-
displaced π-π stacking between the quinazoline core and the aromatic side chain of Trp197
and a direct hydrogen bond between the anilinic amine and the Asp294 carboxylic acid func-
tion. The deep sub-pocket decorated with lipophilic amino acid residues is occupied by a
hydrophobic moiety, in this case a phenethyl substituent in the 4-position. This site was
thought to be suitable for derivatization to potentially increase compound affinity towards the
kinase, since it has been shown that small molecules with differently sized moieties at this posi-
tion could bind to the LP, with 3 being the most demanding LiPoLi observed thus far [13].
Since previous studies showed that 3 was poorly soluble in aqueous media, we selected the
2-phenyl ring pointing outside the LP towards the solvent for modification with solubilizing
groups (Fig 1B). Furthermore, the 7-position of the scaffold should be easily derivatized while
maintaining the NH-group as a hydrogen bond donor. Taken together, these observations led
us to choose the lipophilic moiety in the 4-position, the solvent-exposed 2-phenyl group as
well as the 7-amine as putative sites of modification for the preparation of a diverse quinazo-
line-based compound library (Fig 1C).
The alignment of LiPoLis 1 and 3 (PDB-codes: 3HVC and 4DLJ) followed by structural
analysis prompted us to our initial concepts of derivatization of the quinazoline core scaffold
(Fig 2A). In our first design approach, we increased the size of the lipophilic group in the
4-position by introducing fluoro- and difluorophenyl substituents (9b-f) (Fig 2B). These
hydrophobic elements might fill out the sub-pocket and thereby form additional favorable
interactions, e.g., halogen bonds to backbone carbonyls [30]. Furthermore, the binding of fluo-
rinated compounds to receptors may be entropically favored due to the liberation of water by
desolvation, particularly at hydrophobic moieties [31]. Substitution of the phenyl ring with a
thiophene (Tph) moiety and variation of the linker length to the scaffold (9g,h) emerged as
interesting alternatives since thiophene is a bioisostere with respect to the phenyl group and
features different chemical and electronic properties [32]. Using a shorter methylene linker
may provide sufficient space for expansion of the five-membered ring, which was the basis of
the design of 9j (Fig 2C). Aliphatic moieties like the present cyclopropyl group in building
block 14 that was used for the synthesis of 9j can make a greater contribution to the hydropho-
bic interactions than aromatic substituents, and thus these aliphatic structures may be advanta-
geous for improving ligand affinity [33]. As an alternative strategy, we also designed a hybrid
compound (9i), which retained the scaffold of 3 while the 4-position was derivatized with a
fragment of the known LiPoLi 1 (Fig 2D).
Furthermore, the solvent exposed 2-phenyl ring and the amine in the 7-position were cho-
sen as derivatization sites for the introduction of solubilizing groups, such as acetylpiperazine
(9l), morpholine (9m), methylsulfinyl (19), amides and sulfonamides (10a-e). Also, 8-amino
derivative 9k was designed to investigate any impact on the binding affinity, since new hydro-
gen bonds to the amino acid chains could be formed. The structural basis for this assumption
is the relatively high density of polar amino acid side chains in this region of the LP, such as
Asp294 being directly involved in the binding of 3.
We developed a common synthetic route which successfully led to the proposed target
compounds (Fig 3). The quinazoline cores 7a-d were built up in the reaction by use of
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
18 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 5. a Synthesis of amine building block 16. a Reagents and conditions: (a) ethane-1,2-diamine, K2CO3,
I2, tBuOH, 70˚C, 3.5 h, 54%.
https://doi.org/10.1371/journal.pone.0184627.g005
anthranilic acids 5a and 5b with benzamide and benzoic anhydride moieties, respectively [34,
35], or by using anthranilic amide 6 and aldehyde building blocks [36, 37]. Substitution in the
4-position with an amine took place after activation of the hydroxyl group with hexachlorocy-
clotriphosphazene (HCCP) [38], yielding the nitro compound precursors 8a-n, which were
subsequently reduced to the corresponding amines 9a-m. These were feasible substrates to
undergo nucleophilic substitution with either carboxylic acids or sulfonyl chlorides to yield
compounds 10a-e.
Some LiPoLis required the synthesis of building blocks to be used for coupling to the scaf-
fold (14, 16) or as components for condensation to the quinazoline core (17a,b). Starting with
the protection of 11 via acetylation and subsequent Friedel-Crafts acylation of the furan ring
12, the final amine 14 was generated by reduction and simultaneous deprotection of interme-
diate compound 13 under acidic conditions (Fig 4). Dihydroimidazole 16 was prepared by
condensation of 4-fluorobenzaldehyde 15 with ethylene diamine (Fig 5).
To decorate the quinazolines 9l,m with corresponding solubilizing groups, the aldehydes
17a and 17b were synthesized by substituting 4-fluorobenzaldehyde 15 with acetylpiperazine
and morpholine, respectively (Fig 6). For the generation of 19 starting from 7a, the intermedi-
ate 8n was oxidized to the sulfinyl compound 18 that was subsequently reduced under mild
conditions to give the final amine (Fig 7). Following these procedures, we synthesized a library
of over 30 compounds.
As a potential control ligand, we also synthesized 2 [17], starting with the condensation of
nitrile component 20 and hydrazine [39]. The resulting pyrazolamine 21 was converted with
4,4,4-trifluoro-3-oxobutanoate to yield the final product 2. Analogues 22 and 23 were designed
to explore possible chemical space with a putative alternative LiPoLi scaffold and were easily
generated by using the corresponding oxobutanoates (Fig 8). To subsequently validate our
design concept and to confirm that the newly developed LiPoLis would address the LP of the
p38α MAPK, we undertook SPR experiments.
Fig 6. a Synthesis of aldehyde building blocks 17a and 17b. a Reagents and conditions: (a) amine,
Na2CO3, H2O, reflux, 18 h, 55–58%.
https://doi.org/10.1371/journal.pone.0184627.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
19 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 7. a Synthesis of 19. a Reagents and conditions: (a) mCPBA, DCM, rt, 3 h, 87%.
https://doi.org/10.1371/journal.pone.0184627.g007
Surface plasmon resonance
For SPR measurements, the His6-p38α wt was immobilized on a trisNTA [29] sensor. The
system was tested with the active site inhibitor SB203580 that showed reproducible results
(kon = 1.43 Á 106 (M Á s)-1, koff = 2.43 Á 10−3 s-1, KD = 1.7 nM) in good agreement with the litera-
ture [34, 40] (S1A Fig). Thus, the chosen assay was subsequently used for the characterisation
of the synthesized library of quinazoline-based LiPoLis.
Injection of LiPoLi samples gave a concentration dependent response for some of the syn-
thesized derivatives, indicating a specific binding event to the p38α kinase domain (S1B–S1D
Fig) since other LiPoLis did not show any increasing signal upon compound injection (S1F
Fig). Hence, this renders the separation between actual ligands and non-binding molecules to
the immobilized protein and the identification of tolerated structural modifications to the
LiPoLis by SPR possible. Concerning the ligands that didn’t show any response, all nitro deriv-
atives 8a-m as well as 7- and 8-substituted LiPoLis 10a-e can be considered not to bind to
p38α under the chosen experimental conditions. Notably, also reference compound 2 and its
derivatives 22 and 23 did not show any response in the SPR experiments with increasing ana-
lyte concentrations.
Those ligands that were identified to positively bind to p38α MAPK in these studies all
exhibited a similar shape of the detected sensorgram as the initial lead compound 3. Starting
at baseline level, sensorgrams showed an increasing response during analyte injection in a
concentration-dependent manner, which indicates a positive binding event to the immobi-
lized protein. Unspecific binding to the sensor surface of the reference channel could be
Fig 8. a Synthesis of 2, 22 and 23. a Reagents and conditions: (a) hydrazine, EtOH, reflux, 21 h, 50%; (b)
oxobutanoate, AcOH reflux, 19 h, 51–60%.
https://doi.org/10.1371/journal.pone.0184627.g008
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
20 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
essentially excluded, since corresponding sensorgrams did not represent any signal indicat-
ing unwanted interactions with the reference surface blocked with the His6-peptide. At
higher concentrations the tested analytes did steadily bind to the active sensor surface with-
out reaching an equilibrium, indicating accumulating effects to the kinase. This behaviour
complicates the determination of kinetic parameters in terms of association as a 1:1-Lang-
muir fit model did not properly represent the binding event. However, the observed effect
was reversible, as baseline level was recovered after switching from compound injection to
running buffer flow.
Compounds that showed a characteristic response in the SPR measurements allow conclu-
sions regarding the tolerated chemical space concerning compound modifications that can be
made to the 2-arylquinazoline scaffold without impairing the ability to address the LP. The
exemplified results for the characterisation of 3, 9c and 9m showcase the commonly observed
shape of sensorgrams of the tested LiPoLis, that were reflected by slow association and signifi-
cantly fast dissociation (koff = 0.04–0.06 s-1) (S1B–S1D Fig), typically leading to weak binding
affinities. Substituting the phenethyl moiety in 4-position with bioisosteres in form of thio-
phenes 9g and 9h or replacement by fluorophenyl residues (9b-d) gave comparable and repro-
ducible sensorgrams. On note, derivatives 9e,f carrying the difluorophenyl moiety appeared to
lose any affinity to the protein compared to the other 4-aminoquinazolines. In direct compari-
son to 3, the solubilizing group-bearing derivatives 9l,m showed a significantly faster associa-
tion before reaching the equilibrium state at lower ligand concentrations (S1D Fig). Thus,
these results demonstrate, that even sterically more demanding moieties are tolerated at the
phenyl ring in the 2-postion. In one case, the binding signal of 9j detected from the active
channel was apparently masked by non-specific binding occurring at both, the active surface
and the blocked reference surface, reflected in pseudo-irreversible binding after double-
referencing (S1E Fig).
In summary, we could establish a robust SPR assay system that was set up using the known
inhibitor SB203580 and generated reliable and reproducible data for the characterisation of
the presented ligands. Weak affinity of the analytes towards the target kinase was generally
observed in SPR and orthogonal assays, e.g., thermal shift assays (S3 Fig). To further character-
ise the designed LiPolis regarding the exact binding mode when bound to p38α MAPK co-
crystallization experiments were conducted.
Crystallography
For the crystallization, we pursued different strategies that resulted in different co-crystallized
structures of p38α MAPK in complex with LiPoLi quinazolines. Crystal growth varied depend-
ing on the experimental protocol followed and compounds used. When previously published
conditions were used in the presence of BOG [14], crystals and the corresponding protein
structures were usually obtained that accommodated the detergent in the LP. BOG is known
to bind to the LP and the concentration used under these conditions may compete with the
LiPoLi quinazolines for the occupation of the LP. Therefore, we also conducted crystallization
trials in the absence of BOG. A third crystallization method focused on using the active site
inhibitor BIRB-769 to stabilize the kinase and thereby facilitate crystal growth and eventually
binding of the LiPoLis. Generally, crystal growth took place spontaneously or within 21 days,
yielding needle-shaped crystals when BOG or BIRB-796 were present. More spherical-shaped
crystals were obtained when the protein was exclusively incubated with the LiPoLis (S4 Fig).
Protein crystals showing the spherical morphology usually were occupied with bound LiPoLis
to the LP and could thereby serve as an indicator for successful co-crystallization (a summary
of these findings can be found in S1 Table).
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
21 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
When crystallization trials were set up with ligands bearing an alternative substituent other
than an amine in the 7-position, immediate precipitation of the protein was observed. Thus,
no crystals could be obtained for nitro compounds 8a-n and the derivatives 9k, 10a-e as well
as 2 and its analogues. This outcome indicated that sterically demanding substituents in the
7-position might interfere with protein binding, resulting in the loss of the previously observed
H bond between the amine and Asp294.
We started our trials of the 4-amino derivatives following the direct co-crystallization pro-
tocol and the complex of 9c was one of the first structures that was successfully solved (Fig
9A). The generally proposed binding mode and key interactions were still maintained, includ-
ing the phenethyl moiety binding deeply into the lipophilic sub-pocket, the π-π-interaction
between Trp197 and the core quinazoline scaffold, and a direct hydrogen bond between the
amine in the 7-position and Asp294 as well as a hydrogen bonding network with surrounding
water molecules and amino acid side chains and the protein backbone, e.g., including Ser251
and Lys249, and partly stabilized by contacts involving crystal symmetry mates, particularly
Glu336 (S5 Fig). Furthermore, we obtained crystals for 9g and 9h that were conceived as bioi-
sosteres of 3 by substituting the 4-residue with the corresponding thiophene moieties (Fig 9B
and 9C). Thus, we demonstrated that the hydrophobic sub-pocket of the LP was capable of
harboring that five-membered heterocycle while maintaining the general binding mode
described above. These findings were the starting point for the design of 9j as an “extended
derivative" of 9g, which was achieved by adding another methylene cyclopropyl to the furan
Fig 9. Crystal structures of LiPoLis in complex with p38α. Diagrams of the experimental electron densities of (A) 9c (yellow), (B) 9g (green), (C) 9h
(cyan), (D) 9j (white), (E) 9l (magenta) and (F) 9m (orange). At resolutions ranging from 1.85 to 2.4 Å; 2mFo-DFc map (grey) contoured at 1.0σ. Water
molecules are shown as red spheres. Hydrogen-bond interactions of the ligands with the protein and water molecules are illustrated by yellow dotted
lines. All LiPoLis bind to the LP flanked by helices 1L14, 2L14 and the αEF/αF loop.
https://doi.org/10.1371/journal.pone.0184627.g009
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
22 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
ring. We were gratified to observe that the co-crystallization of this LiPoLi succeeded in pres-
ence of BIRB-796. It is noteworthy that the binding mode of 9j within the LP was not signifi-
cantly different from the previously mentioned key interactions (Fig 9D). The cyclopropyl
moiety of 9j was able to access the deeply buried regions of the LP. Furthermore, the cyclopro-
pyl group did not lead to any perturbation of the protein structure based on the arrangement
of adjacent amino acid side chains, as compared to the other LiPoLi bound structures. The dif-
ference electron density map was not defined at the position where the cyclopropyl substituent
was situated, indicating a significant flexibility and the ability to twist within the lipophilic
sub-pocket of the LP.
Interestingly, some compounds only crystallized in the absence of BOG in the reservoir
solution, and these samples also exhibited a new crystal morphology as observed for 9l and 9m
(S4 Fig). The co-crystal structures for both compounds were successfully solved and confirmed
our design approach as they revealed the same binding geometry as the initial lead compound
3 (Fig 9E and 9F). Compounds 9l and particularly 9m were designed to improve the com-
pound’s solubility by introducing a solubilizing group at the phenyl ring in the para-position.
The 2-phenyl substituent was orientated towards the solvent, similar to 3, although no interac-
tions with the surrounding water molecules were found. Notably, the network of water mole-
cules surrounding the LP was best resolved in the p38α-9m complex crystal structure (Fig 9F).
Performing a simulated annealing refinement, mFo-DFc omit maps were generated for all co-
crystallized LiPoLis and show defined electron density that could be clearly assigned to the cor-
responding ligands and demonstrated a consistent mode of binding (S6 Fig). Only for 9j, no
defined density of the cyclopropyl moiety within the LP was observed, likely due to conforma-
tional flexibility, and 9h only showed partial density of the quinazoline scaffold, potentially
caused by lower occupancy at the chosen experimental conditions. When we compared our
new complex crystal structures with p38α-1 and p38α-3, we commonly observed subtle con-
formational changes of secondary structure elements shaping the LP. Interestingly, a more sig-
nificant displacement of helix 2L14 was found in the p38α-9h complex structure, resulting in
an almost complete opening of the lower section of the LP (Fig 10). This underlines a pro-
nounced flexibility of the entire binding site rendering it putatively addressable by even more
complex synthetic ligands or yet to be identified biological binding partners.
Conclusions
Due to their central role in cell signaling pathways, protein kinases are prominent targets in
drug research and development. Several diseases are caused by direct dysregulation of the cor-
responding kinase or their mediators and interaction partners, implying that not only the
integrity of enzymatic mechanisms, but also of non-catalytic functions, often referred to as
scaffolding functions [10], are mandatory to preserve the sensitive and well-regulated pro-
cesses within the cell. The successful targeting of scaffolding functions requires detailed knowl-
edge of the participating interaction partners as well as the mechanisms of communication,
which include the important structural elements such as binding epitopes and their ligands as
well as their biological roles. Here, we set out to target the previously identified LP in p38α
MAPK. By applying structure-guided derivatization of 3, we introduced structural variations
in the hydrophobic moiety in the 4-position and the functionalization of the 2-phenyl ring and
the 7-/8-positions that led to a focused library of over 30 compounds. An SPR assay system
was set up to get insight into the kinetics of LiPoLi binding towards p38α MAPK. The interac-
tions of the tested ligands and the immobilized kinase was typically described by a slow associ-
ation phase and saturation followed by fast dissociation. Most of the LiPoLis that showed a
characteristic response in the SPR experiments were also successfully co-crystallized with p38α
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
23 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Fig 10. Conformational changes of the LP upon LiPoLi binding. (A) Alignment of the p38α-9c (cyan) and p38α-9h (yellow) complex crystal
structures. Displacement of helix 2L14 and minor rearrangement of loop αEF/αF and helix 1L14 (trajectories are shown as red dots); (B) Opening of the
LP (red arrows) when 9c is bound compared to the closed LP in presence of 9h.
https://doi.org/10.1371/journal.pone.0184627.g010
MAPK. A series of six complex crystal structures showed that the designed LiPoLis indeed tar-
get the LP of p38α MAPK and validated our design approaches. The complex structure of 9j
could be solved although in the SPR experiments only unspecific binding to the sensor surface
was detected.
In summary, we identified substitution patterns of the LiPoLi scaffold to be crucial for the
opening of the LP underlining the flexible nature of this binding site. The characterization of
the ligand-binding event by SPR indicated, however, that these LiPoLis are most likely not
suitable to serve as functional probes given their weak binding affinities. Anyhow, the results
presented here will encourage further compound modifications to focus particularly on the
2-phenyl ring and alternative substitutions to generate more potent LiPoLis to finally dissect
the functional role of the lipid pocket in p38α.
Supporting information
S1 Fig. Representative sensorgrams of SB20350 and LiPoLis of p38α MAPK. Time-depen-
dent changes in resonance units (RU) were detected during the injection of (A) SB203580,
(B) 3, (C) 9h, (D) 9l, (E) 9j and (F) 2 at various concentrations to a sensor surface carrying
immobilized His6-p38α. Global 1:1-Langmuir binding fits are shown as black lines.
(TIF)
S2 Fig. Representative sensorgrams of all LiPoLi amine derivatives. Time-dependent
changes in resonance units (RU) were detected during the injection of LiPoLis in concentra-
tions ranging from 1 nM—30 μM to a sensor surface carrying immobilized His6-p38α. Global
1:1-Langmuir and multi-phasic binding fits, respectively, are shown as black lines. LiPoLi
nitro derivatives showed no response and are therefore not shown.
(TIF)
S3 Fig. Thermal shift assay. A) p38α MAPK melting curves and their first derivatives in pres-
ence of active site inhibitors and a selection of LiPoLis. Thermal shifts ΔT (˚C) were calculated
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
24 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
from substraction of melting point in presence of DMSO from measured melting points Tm
(˚C) in presence of compound. B) p38α MAPK melting curves and their first derivatives for all
presented LiPoLis.
(TIF)
S4 Fig. Protein crystals of p38α MAPK. Crystals grown in presence of A) 9g (needles), B) 9l
and C) 9m (cubic). Crystals were grown at 20˚C using 100 mM MES pH 5.6–6.2, 20–30%
PEG4000 and 50 mM BOG as reservoir solution.
(TIF)
S5 Fig. Water-mediated contact to crystal symmetry mate Glu336. Exemplified for p38α in
complex with 9c; symmetry mate is shown in grey, numbers represent H bond distances in Å
(PDB: 5N63).
(TIF)
S6 Fig. Omit maps of the crystallized LiPoLis. Performing a simulated annealing refinement,
mFo-DFc omit maps (green, contoured at 2.5σ) were calculated for A) 9c, B) 9g, D) 9h, E) 9j,
E) 9l and F) 9m, as well as for surrounding water molecules. 2Fo-Fc maps for the ligands,
waters and key residues Trp197 and Asp294 were contoured at 1.0σ (blue). Maps indicate par-
tial occupancy for 9h due to multiple molecules bound to the protein and conformational flex-
ibility of the cyclopropyl moiety in 9j.
(TIF)
S1 Table. Crystallographic statistics of p38α MAPK in complex with LiPoLi derivatives.
Statistics for co-crystals with LiPoLis 9c, 9g, 9h, 9j, 9l and 9m (PDBs: 5N63, 5N64, 5N65,
5N66, 5N67 and 5N68). Values in parenthesis refer to the highest resolution shell.
(TIF)
S1 File. NMR spectra.
(DOCX)
S2 File. Overview of collected data regarding LiPoLi characterization.
(DOCX)
S3 File. Molecular formula strings.
(CSV)
Author Contributions
Conceptualization: Daniel Rauh.
Data curation: Mike Bu¨hrmann, Matthias P. Mu¨ller.
Formal analysis: Mike Bu¨hrmann, Bianca M. Wiedemann, Julia Hardick, Maria Ecke.
Funding acquisition: Daniel Rauh.
Investigation: Mike Bu¨hrmann, Bianca M. Wiedemann, Julia Hardick.
Methodology: Mike Bu¨hrmann.
Project administration: Daniel Rauh.
Resources: Daniel Rauh.
Validation: Mike Bu¨hrmann.
Visualization: Mike Bu¨hrmann.
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
25 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
Writing – original draft: Mike Bu¨hrmann.
Writing – review & editing: Matthias P. Mu¨ller, Daniel Rauh.
References
1. Cohen P, Frame S. The renaissance of GSK3. Nat Rev Mol Cell Biol. 2001; 2(10):769–76. https://doi.
org/10.1038/35096075 PMID: 11584304
2. Hidaka H, Inagaki M, Kawamoto S, Sasaki Y. Isoquinolinesulfonamides, Novel and Potent Inhibitors of
Cyclic-Nucleotide Dependent Protein-Kinase and Protein Kinase-C. Biochemistry. 1984; 23(21):5036–
41. PMID: 6238627
3. Knight ZA, Shokat KM. Features of selective kinase inhibitors. Chem Biol. 2005; 12(6):621–37. https://
doi.org/10.1016/j.chembiol.2005.04.011 PMID: 15975507
4. Sawyers CL. Opportunities and challenges in the development of kinase inhibitor therapy for cancer.
Genes Dev. 2003; 17(24):2998–3010. https://doi.org/10.1101/gad.1152403 PMID: 14701871
5. Weisner J, Gontla R, van der Westhuizen L, Oeck S, Ketzer J, Janning P, et al. Covalent-Allosteric
Kinase Inhibitors. Angew Chem Int Ed Engl. 2015; 54(35):10313–6. https://doi.org/10.1002/anie.
201502142 PMID: 26110718
6. Engel J, Richters A, Getlik M, Tomassi S, Keul M, Termathe M, et al. Targeting Drug Resistance in
EGFR with Covalent Inhibitors: A Structure-Based Design Approach. J Med Chem. 2015; 58(17):6844–
63. https://doi.org/10.1021/acs.jmedchem.5b01082 PMID: 26275028
7. Cox KJ, Shomin CD, Ghosh I. Tinkering outside the kinase ATP box: allosteric (type IV) and bivalent
(type V) inhibitors of protein kinases. Future Med Chem. 2011; 3(1):29–43. https://doi.org/10.4155/fmc.
10.272 PMID: 21428824
8. Adrian FJ, Ding Q, Sim T, Velentza A, Sloan C, Liu Y, et al. Allosteric inhibitors of Bcr-abl-dependent cell
proliferation. Nat Chem Biol. 2006; 2(2):95–102. https://doi.org/10.1038/nchembio760 PMID: 16415863
9. Yang J, Campobasso N, Biju MP, Fisher K, Pan XQ, Cottom J, et al. Discovery and characterization of
a cell-permeable, small-molecule c-Abl kinase activator that binds to the myristoyl binding site. Chem
Biol. 2011; 18(2):177–86. https://doi.org/10.1016/j.chembiol.2010.12.013 PMID: 21338916
10. Rauch J, Volinsky N, Romano D, Kolch W. The secret life of kinases: functions beyond catalysis. Cell
Commun Signaling. 2011; 9(1):23.
11. Giet R, Prigent C. The non-catalytic domain of the Xenopus laevis auroraA kinase localises the protein
to the centrosome. J Cell Sci. 2001; 114(11):2095–104.
12. Garcia-Martinez JM, Calcabrini A, Gonzalez L, Martin-Forero E, Agullo-Ortuno MT, Simon V, et al. A
non-catalytic function of the Src family tyrosine kinases controls prolactin-induced Jak2 signaling. Cell
Signal. 2010; 22(3):415–26. https://doi.org/10.1016/j.cellsig.2009.10.013 PMID: 19892015
13. Getlik M, Simard JR, Termathe M, Gru¨tter C, Rabiller M, van Otterlo WA, et al. Fluorophore labeled
kinase detects ligands that bind within the MAPK insert of p38alpha kinase. PLoS One. 2012; 7(7):
e39713. https://doi.org/10.1371/journal.pone.0039713 PMID: 22768308
14. Bukhtiyarova M, Northrop K, Chai XM, Casper D, Karpusas M, Springman E. Improved expression,
purification, and crystallization of p38 alpha MAP kinase. Protein Expression and Purification. 2004; 37
(1):154–61. https://doi.org/10.1016/j.pep.2004.05.017 PMID: 15294293
15. Tzarum N, Eisenberg-Domovich Y, Gills JJ, Dennis PA, Livnah O. Lipid Molecules Induce p38 alpha
Activation via a Novel Molecular Switch. J Mol Biol. 2012; 424(5):339–53. https://doi.org/10.1016/j.jmb.
2012.10.007 PMID: 23079240
16. Perry JJP, Harris RM, Moiani D, Olson AJ, Tainer JA. p38 alpha MAP Kinase C-Terminal Domain Bind-
ing Pocket Characterized by Crystallographic and Computational Analyses. J Mol Biol. 2009; 391(1):1–
11. https://doi.org/10.1016/j.jmb.2009.06.005 PMID: 19501598
17. Comess KM, Sun CH, Abad-Zapatero C, Goedken ER, Gum RJ, Borhani DW, et al. Discovery and
Characterization of Non-ATP Site Inhibitors of the Mitogen Activated Protein (MAP) Kinases. ACS
Chem Biol. 2011; 6(3):234–44. https://doi.org/10.1021/cb1002619 PMID: 21090814
18. Diskin R, Engelberg D, Livnah O. A novel lipid binding site formed by the MAP kinase insert in p38
alpha. J Mol Biol. 2008; 375(1):70–9. https://doi.org/10.1016/j.jmb.2007.09.002 PMID: 17999933
19. Bu¨hrmann M, Hardick J, Weisner J, Quambusch L, Rauh D. Covalent Lipid Pocket Ligands Targeting
p38alpha MAPK Mutants. Angewandte Chemie. 2017. https://doi.org/10.1002/anie.201706345 PMID:
28834017
20. Simard JR, Gru¨tter C, Pawar V, Aust B, Wolf A, Rabiller M, et al. High-Throughput Screening To Identify
Inhibitors Which Stabilize Inactive Kinase Conformations in p38 alpha. J Am Chem Soc. 2009; 131
(51):18478–88. https://doi.org/10.1021/ja907795q PMID: 19950957
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
26 / 27

Structure-based design of lipid pocket ligands of p38α MAPK
21. Kabsch W. Automatic Processing of Rotation Diffraction Data from Crystals of Initially Unknown Sym-
metry and Cell Constants. Journal of Applied Crystallography. 1993; 26:795–800.
22. Read RJ. Pushing the boundaries of molecular replacement with maximum likelihood. Acta Crystallo-
graphica Section D-Biological Crystallography. 2001; 57:1373–82.
23. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallographica Section
D-Biological Crystallography. 2004; 60:2126–32.
24. Schu¨ttelkopf AW, van Aalten DMF. PRODRG: a tool for high-throughput crystallography of protein-
ligand complexes. Acta Crystallographica Section D-Biological Crystallography. 2004; 60:1355–63.
25. Joosten RP, Long F, Murshudov GN, Perrakis A. The PDB_REDO server for macromolecular structure
model optimization. IUCrJ. 2014; 1(Pt 4):213–20. https://doi.org/10.1107/S2052252514009324 PMID:
25075342
26. Adams PD, Afonine PV, Bunkoczi G, Chen VB, Davis IW, Echols N, et al. PHENIX: a comprehensive
Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;
66(Pt 2):213–21. https://doi.org/10.1107/S0907444909052925 PMID: 20124702
27. Lovell SC, Davis IW, Adrendall WB, de Bakker PIW, Word JM, Prisant MG, et al. Structure validation by
C alpha geometry: phi,psi and C beta deviation. Proteins-Structure Function and Genetics. 2003; 50
(3):437–50.
28. DeLano WL. The PyMOL Molecular Graphics System 2002. http://www.pymol.org.
29. Lata S, Reichel A, Brock R, Tampe R, Piehler J. High-affinity adaptors for switchable recognition of histi-
dine-tagged proteins. J Am Chem Soc. 2005; 127(29):10205–15. https://doi.org/10.1021/ja050690c
PMID: 16028931
30. Wilcken R, Liu XR, Zimmermann MO, Rutherford TJ, Fersht AR, Joerger AC, et al. Halogen-Enriched
Fragment Libraries as Leads for Drug Rescue of Mutant p53. J Am Chem Soc. 2012; 134(15):6810–8.
https://doi.org/10.1021/ja301056a PMID: 22439615
31. Biffinger JC, Kim HW, DiMagno SG. The polar hydrophobicity of fluorinated compounds. Chembio-
chem. 2004; 5(5):622–7. https://doi.org/10.1002/cbic.200300910 PMID: 15122633
32. Meanwell NA. Synopsis of Some Recent Tactical Application of Bioisosteres in Drug Design. J Med
Chem. 2011; 54(8):2529–91. https://doi.org/10.1021/jm1013693 PMID: 21413808
33. Karplus PA. Hydrophobicity regained. Protein Sci. 1997; 6(6):1302–7. https://doi.org/10.1002/pro.
5560060618 PMID: 9194190
34. Simard JR, Getlik M, Gru¨tter C, Schneider R, Wulfert S, Rauh D. Fluorophore Labeling of the Glycine-
Rich Loop as a Method of Identifying Inhibitors That Bind to Active and Inactive Kinase Conformations.
J Am Chem Soc. 2010; 132(12):4152–60. https://doi.org/10.1021/ja908083e PMID: 20201574
35. Nouira I, Kostakis IK, Dubouilh C, Chosson E, Iannelli M, Besson T. Decomposition of formamide assis-
ted by microwaves, a tool for synthesis of nitrogen-containing heterocycles. Tetrahedron Lett. 2008;
49(49):7033–6.
36. Hour MJ, Huang LJ, Kuo SC, Xia Y, Bastow K, Nakanishi Y, et al. 6-Alkylamino- and 2,3-dihydro-3’-
methoxy-2-phenyl-4-quinazolinones and related compounds: their synthesis, cytotoxicity, and inhibition
of tubulin polymerization. J Med Chem. 2000; 43(23):4479–87. PMID: 11087572
37. Cheng R, Guo T, Zhang-Negrerie Z, Du Y, Zhao K. One-Pot Synthesis of Quinazolinones from Anthra-
nilamides and Aldehydes via p-Toluenesulfonic Acid Catalyzed Cyclocondensation and Phenyliodine
Diacetate Mediated Oxidative Dehydrogenation. Synthesis. 2013; 45(21):8.
38. Shen ZL, He XF, Dai JL, Mo WM, Hu BX, Sun N, et al. An efficient HCCP-mediated direct amination of
quinazolin-4(3H)-ones. Tetrahedron. 2011; 67(9):1665–72.
39. Moe ST, Thompson AB, Smith GM, Fredenburg RA, Stein RL, Jacobson AR. Botulinum neurotoxin
serotype A inhibitors: small-molecule mercaptoacetamide analogs. Bioorg Med Chem. 2009; 17
(8):3072–9. https://doi.org/10.1016/j.bmc.2009.03.013 PMID: 19329331
40. Zhang WX, Wang RX, Wisniewski D, Marcy AI, LoGrasso P, Lisnock JM, et al. Time-resolved Forster
resonance energy transfer assays for the binding of nucleotide and protein substrates to p38 alpha pro-
tein kinase. Anal Biochem. 2005; 343(1):76–83. https://doi.org/10.1016/j.ab.2005.05.011 PMID:
15979553
PLOS ONE | https://doi.org/10.1371/journal.pone.0184627 September 11, 2017
27 / 27