Addressing the Glycine-Rich Loop of Protein Kinases by a Multi-Facetted Interaction Network: Inhibition of PKA and a PKB Mimic

Article abstract of DOI:10.1002/chem.201503552

Protein kinases continue to be hot targets in drug discovery research, as they are involved in many essential cellular processes and their deregulation can lead to a variety of diseases. A series of 32 enantiomerically pure inhibitors was synthesized and tested towards protein kinase A (PKA) and protein kinase B mimic PKAB3 (PKA triple mutant). The ligands bind to the hinge region, ribose pocket, and glycine-rich loop at the ATP site. Biological assays showed high potency against PKA, with K_i values in the low nanomolar range. The investigation demonstrates the significance of targeting the often neglected glycine-rich loop for gaining high binding potency. X-ray co-crystal structures revealed a multi-facetted network of ligand-loop interactions for the tightest binders, involving orthogonal dipolar contacts, sulfur and other dispersive contacts, amide-π stacking, and H-bonding to organofluorine, besides efficient water replacement. The network was analyzed in a computational approach.

Full text of DOI:10.1002/chem.201503552

DOI: 10.1002/chem.201503552
Full Paper
&
Kinases
Addressing the Glycine-Rich Loop of Protein Kinases by a Multi-
Facetted Interaction Network: Inhibition of PKA and a PKB Mimic
Birgit S. Lauber,[a] Leo A. Hardegger,[a] Alam K. Asraful,[b] Bjarte A. Lund,[b] Oliver Dumele,[a]
Michael Harder,[a] Bernd Kuhn,[c] Richard A. Engh,*[b] and FranÅois Diederich*[a]
Dedicated to Professor Donald Hilvert on the occasion of his 60th birthday
Abstract: Protein kinases continue to be hot targets in drug
discovery research, as they are involved in many essential
cellular processes and their deregulation can lead to a variety
of diseases. A series of 32 enantiomerically pure inhibitors
was synthesized and tested towards protein kinase A (PKA)
and protein kinase B mimic PKAB3 (PKA triple mutant). The
ligands bind to the hinge region, ribose pocket, and glycine-
rich loop at the ATP site. Biological assays showed high po-
tency against PKA, with Ki values in the low nanomolar
range. The investigation demonstrates the significance of
targeting the often neglected glycine-rich loop for gaining
high binding potency. X-ray co-crystal structures revealed
a multi-facetted network of ligand–loop interactions for the
tightest binders, involving orthogonal dipolar contacts,
sulfur and other dispersive contacts, amide–p stacking, and
H-bonding to organofluorine, besides efficient water replace-
ment. The network was analyzed in a computational ap-
proach.
Introduction
than 500 when considering sequence alone, making up ap-
proximately 2% of human genes.[6]
Protein kinases, with their function to phosphorylate other pro-
teins as a core mechanism of cellular signal transduction, con-
stitute one of the most important enzyme classes. The “proto-
type” for the class, protein kinase A (PKA), was first described
in 1968 by Krebs and co-workers as a mediator for the second
messenger cyclic adenosine monophosphate (cAMP).[1] Because
its phosphotransferase activity occurs after dissociation into
a single and constitutively active catalytic domain, PKA is well
suited as the prototype for mechanistic studies of protein kin-
ases. Its kinetic properties and the crystal structure of its
apoenzyme provided detailed insight into the signaling path-
way, including the activation and deactivation mechanisms
and the constitution of the active site.[2–5] These data have
proven invaluable for understanding the mechanisms of the
other protein kinases in the human genome, numbering more
The large number of kinases, a high degree of sequence
conservation, and the competition with millimolar concentra-
tions of adenosine triphosphate (ATP) in cells make the kinases
challenging targets in drug discovery. However, HA1077 (Fasu-
dil) was approved as a ROCK1 inhibitor in Japan in 1995, and
imatinib (Glivec or Gleevec) was approved as an ABL kinase in-
hibitor in 2001, initiating an era of rapidly growing investment
into pharmaceutical research, now with some 30 drugs on the
market.[7]
Among the protein kinase signaling proteins and drug tar-
gets, protein kinase B (PKB, also known as Akt) has a central
role in regulatory pathways, including cell survival, metabolism,
motility, transcription, and cell-cycle progression. It is a down-
stream target in the phosphoinositide (PI) 3-kinase signaling
pathway, and it has been shown that cells with dysregulated
PKB-production, for example, knockout or over-expression,
may cause cardiac or skeletal muscle hypertrophy, tumors, and
a disordered insulin secretory pathway.[8–11] This highly con-
served enzyme belongs to the AGC subfamily (together with
PKA and several other subgroups) and consists of three iso-
forms, PKBa, PKBb, and PKBg (Akt1, Akt2, and Akt3, respective-
ly). The three isoforms are encoded by distinct genes but share
a highly conserved structure across species. The architecture of
the kinases consists of three functional domains, a C-terminal
regulatory domain with the hydrophobic motive (HM), the N-
terminal pleckstrin-homology domain (PH), and a central
kinase domain.
[a] Dr. B. S. Lauber, Dr. L. A. Hardegger, O. Dumele, Dr. M. Harder,
Prof. Dr. F. Diederich
Laboratorium für Organische Chemie, ETH Zürich
Vladimir-Prelog-Weg 3, 8093 Zürich (Switzerland)
Fax: (+41)44-632-11-09
E-mail: diederich@org.chem.ethz.ch
[b] A. K. Asraful, B. A. Lund, Prof. Dr. R. A. Engh
Department of Chemistry, University of Tromsø
Forskningsparken 3, Sykehusvegen 23, 9037 Tromsø (Norway)
E-mail: richard.engh@uit.no
[c] Dr. B. Kuhn
Small Molecule Research, Roche Innovation Center Basel
F. Hoffmann-La Roche, Grenzacherstrasse 124, 4070 Basel (Switzerland)
As mentioned above, PKA is one of the best studied kinases
and is closely related to PKB, sharing a sequence identity of
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.201503552.
Chem. Eur. J. 2016, 22, 211 – 221
211
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
45% in the catalytic domain, rising up to 80% in the active
pocket.[12] Although they share high sequence similarity in
their catalytic domains, PKA and PKB are involved in different
signaling pathways, making their feasibility as drug targets de-
pendent on the ability to achieve selectivity for their inhibition.
At the ATP binding sites of PKA (PKAa) and PKB (PKBa/
PKBg), as few as three amino acids need to be mutated in one
kinase to mimic the binding pocket of the other.[13] In the PKA
triple mutant enzyme PKAB3 used herein, two mutations
mimic the parent active pocket of the enzyme PKB (V123A and
L173M). A third mutation (Q181K) is required to prevent a rota-
mer conformation of Q181, enabled by the V123A mutation,
that can occlude the ATP site.[13] A fourth site of possible muta-
tion, V104T, may play a role especially for polar interaction net-
works that are not the focus of this work. It has been shown
by Engh and co-workers that PKAB3 works as a PKB surrogate,
features similar enzyme kinetics, and is generally more suitable
for enzyme crystallization than PKB.[13] Further, comparative
studies of ligand–enzyme interactions using point mutants
allow increased focus on the effects of individual interactions,
removing spurious indirect effects from enzyme differences
elsewhere.
moved for modeling and the enzyme kept rigid during energy
minimization. We designed ligands based on a potential gain
of selectivity for PKB over PKA and to investigate the ligand-
binding properties of the glycine-rich loop. Our initial design
targeted the L173M substitution, which was hoped to deliver
selectivity of PKB over PKA through sulfur–aromatic interac-
tions (Figure 1).[18]
The design of the new inhibitors followed the binding mode
of ATP in the hinge region, with two hydrogen bonds formed
between the quinazolinone (as an adenine substitute) and the
backbone NH of Ala123 and the backbone C=O of Glu121 (Fig-
ure 1a). In PKAB3, the quinazolinone ring was expected to be
sandwiched between the side chains of Val57 and Met173, en-
gaging in sulfur–aromatic interactions at optimal distances of
3.4–3.8 Š (Figure 1b).[18]
The ribose pocket was addressed with a protonated pyrroli-
dine ring, which was hoped to engage in ionic hydrogen
bonds to Glu127 and the backbone carbonyl of Glu170 (Fig-
ure 1c). The glycine-rich loop was targeted by incorporating
different halogenated benzoyl or carbothienyl substituents, as
shown for ligand (+)-1i in Figure 1c (see Table 1 for all ligands
1a–p). The organohalogens should interact with the glycine-
rich loop by orthogonal dipolar CÀX···C=O interactions.[19,20] Ac-
cording to the modeling studies, only the (3R,4S)-enantiomer
was expected to strongly bind into the active pockets of PKA
and PKAB3.
Herein, we report the structure-based design and synthesis
of a series of ligands for inhibition of PKA and PKAB3 and the
subsequent in vitro evaluation. We present a series of novel
PKA/PKAB3 selective inhibitors and an SAR analysis of their
binding to the glycine-rich loop supported by several X-ray co-
crystal structures. This motif, with the general sequence
GxGxxG, is adjacent to the phosphate groups of bound ATP
and also forms a hydrophobic “lid” over the adenine ring of
ATP and the typically heteroaromatic moieties of inhibitors
binding to the protein kinase hinge. This fact and the slightly
different sequence highlights the significant difference to the
so called Walker-motif (GxxxxGKT/S) of ATP binding pro-
teins.[14,15]
Table 1. Ki values [nm] of quinazolinone-based inhibitors determined by
the multi-component spectrophotometric Cook assay. The reported
values are the average of three repeated measurements. Inhibitor (+)-1e
decomposed over time and is marked with *. n.a: Ki value over 2000 nm
and/or inhibition did not show a characteristic inhibition curve. clogD
and clogP values were calculated with the ACD Labs program[a] at pH 7.4.
We aimed primarily to synthesize highly potent, low-molecu-
lar-weight inhibitors and evaluate their binding mode and in
particular their interactions with the glycine-rich loop. In con-
trast to the hinge region, the adenine pocket with the gate-
keeper residues, the ribose pocket, the DFG (Asp-Phe-Gly)
loop, and apolar back pockets, the glycine-rich loop has largely
not been addressed in protein kinase inhibitor design. The ex-
perimental Ki values, the SAR analysis, and several X-ray analy-
ses have been used to obtain structural information to explain
new potent binding motifs to the glycine-rich loop involving
a highly diverse set of weak intermolecular interactions. To
complete the analysis, a computational treatment of the coop-
erative interaction network was performed, revealing multiple
interactions and hotspot atoms.
Ar
X
clogP clogD Ki(PKA) [nm] Ki(PKAB3) [nm]
(+)
(À)
(+)
(À)
1a
1b
1c
1d
Ph
Ph
Ph
Ph
H
0.98
1.22
1.67
1.77
2.17
1.83
À0.43 30
À0.02 10
n.a
70
n.a
3-F
3-Cl
3-Br
3-I
3-CF3
3-OCF3 2.33
1600 80
1500
1800 n.a
400
500
n.a
400
n.a
0.04
0.55
0.95
0.46
1.14
0.23
0.54
0.54
1.00
0.67
1.02
À0.5
0.44
0.62
300 200
80
80
1e* Ph
200
740
n.a
n.a
50
70
20
20
80
370
1600
80
1 f
1g
1h
1i
1j
1k
1l
Ph
Ph
Ph
Ph
Ph
Ph
Ph
200 200
400 500
70
3
3
5
n.a
4-F
1.46
1.73
1.73
2.20
1.91
n.a
1200
1300
1900
n.a
n.a
n.a
1600
1500
1000
4-Cl
4-Br
4-I
4-CF3
4-OCF3 2.20
400
600
700
400
1100
2
12
1m Ph
1n
1o
1p
2-thienyl
2-thienyl 5-Cl
2-thienyl 5-Br
H
0.74
1.60
1.77
170 n.a
23
0.9
n.a
500
Results and Discussion
40
Inhibitor Design
[a] ACD/Structure Elucidator, Version 15.01, Advanced Chemistry Develop-
ment, Inc., Toronto, ON, Canada, www.acdlabs.com, 2015.
Molecular modeling was performed with the program
MOLOC[16] using the MAB force field. The X-ray co-crystal struc-
ture of PKAB3 (PDB ID: 2uw0)[17] was used with the ligand re-
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
212
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 1. Design of inhibitors for the triple mutant PKAB3. a) Schematic representation of the active site of PKAB3. Inhibitor (+)-1 (blue) is shown in the bind-
ing mode proposed by MOLOC.[16] Favorable interactions are highlighted with red dashed lines. The orthogonal C=O···C=O interaction to Gly50 is omitted for
clarity. b) Close-up view of the targeted sulfur–aromatic interactions between Met173 and the quinazolinone moiety of the ligand. c) Proposed binding mode
of ligand (+)-1i to PKAB3 (2.00 Š resolution, PDB ID: 2uw0) as predicted by MOLOC. The expected orthogonal C=O···C=O interaction to Gly50 is shown. Pic-
tures generated using PyMOL (The PyMOL Molecular Graphics System, Version 1.7.4 Schrçdinger, LLC). Distances are given in Š, and interactions are shown as
dashed lines. Color code: Cenzyme gray, Cligand green, N blue, O red, S yellow, Cl limon.
Synthesis
with LiAlH4 to 9, followed by oxidation with PCC, yielded alde-
hyde 5.[24]
The synthetic route is based on a Horner–Wadsworth–Emmons
reaction to generate activated (E)-olefins, followed by a 1,3-di-
polar cycloaddition with azomethine ylides to afford the cen-
tral pyrrolidine ring. The synthesis of protected, 3,4-disubstitut-
ed pyrrolidines by a similar route had been reported previous-
ly,[21] however the protocol had to be altered so that unpro-
tected variants could be accessed instead.[22] Furthermore, as
only the 3R,4S-enantiomer was expected to be active accord-
ing to the modeling, conditions for HPLC enantiomer separa-
tion by HPLC on a chiral stationary phase had to be identified.
In this route, esters 2a–p which were either commercially
available or produced from their corresponding acids (3b/c/d/
i/h/m) following well-established procedures,[23] were trans-
formed to b-ketophosphonates 4a–p with diethyl methyl-
phosphonate at À788C in good yields (Scheme 1).
b-Ketophosphonates 4a–p were coupled to aldehyde 5 to
afford alkenes 10a–p in moderate to good yields with (E/Z)
ratios between 97:3 and 83:17.[25,26] The use of KOtBu instead
of NaH as base led to increased (E) selectivities.[21] Purification
by HPLC afforded diastereoisomerically pure products, but
upon concentration, the alkenes showed limited stability as in-
1
dicated by LCMS chromatography and H NMR spectroscopy.
In a first step, the reaction conditions for the 1,3-dipolar cy-
cloaddition with azomethine ylides generated from glycine
and para-formaldehyde to afford unprotected pyrrolidines
were optimized. The protocol for the synthesis of protected
pyrrolidines (Bn or Me protecting groups) is well-established,[21]
but there is only one example present in the literature in
which 3,4-disubstituted unprotected pyrrolidines were ob-
tained through a 1,3-dipolar cycloaddition.[22] Trapping of the
water formed in the course of the reaction was crucial for the
formation of the unprotected pyrrolidines. For this purpose,
a layer of MgSO4 or Na2SO4 was installed on top of the flask
containing the reaction mixture at reflux. Yields for this key
Aldehyde 5 was synthesized starting from 2-aminotereph-
thalate 6 in a cyclization reaction to afford quinazolinone 7,
which was converted to the Me-protected quinazolinone 8 via
its chlorinated intermediate in two steps (Scheme 2). Reduction
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
213
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Scheme 1. Synthesis of b-ketophosphonates 4a–p. i) H2SO4, EtOH, 808C, 4–30 h, quant.; ii) diethyl methylphosphonate, nBuLi, THF, À78 to 228C, 1.5–6 h,
62–100%.
obtained by the X-ray co-crystal structures of these
inhibitors with PKA.
Biological affinities and X-ray co-crystal structures
Both enantiomers of ligands 1a–p were evaluated for
their in vitro affinities against PKA and PKAB3 via the
activity-based ATP-regenerative NADH (nicotinamide
adenine dinucleotide)-consuming Cook-assay (for de-
tails see the Supporting Information, Section S4).[29]
Scheme 2. Synthesis of aldehyde 5. i) H4NHCO2, HCO2H, 1008C, 19 h, 63%; ii) NH4HCO2,
formamide, 1408C, 7 h, 51%; iii) POCl3, DBU, toluene, 1208C, 2.5 h; iv) NaOMe, 0 to 228C,
1 h, 42%; v) LiAlH4, THF, 0 to 228C, 2 h; vi) PCC, CH2Cl2, 08C, 2 h, 75%. DBU=1,8-diazabi-
Assay interference tests were done to exclude any
undesired interactions of the inhibitors. Neither
adding ADP instead of ATP to mimic the effects of
ADP production by the protein kinase, with and with-
out inhibitor, or measuring the production of ADP by
spontaneous hydrolysis of ATP (with and without inhibitor)
showed any side effects.
cyclo[5.4.0]undec-7-ene; PCC=pyridinium chlorochromate.
step ranged from low to moderate (14–45%), most probably
due to the decreased stability of the olefins. In contrast, by
employing the same reaction conditions, para-formaldehyde,
glycine, and the dipolarophile trans-chalcone (1,3-diphenyl-
prop-2-en-one; 11) afforded, after Boc-protection, the corre-
sponding pyrrolidine (12) in 83% yield over two steps (for the
structures of 11 and 12, see the Supporting Information,
Scheme S1).
The velocity was measured at different inhibitor concentra-
tions, allowing the determination of Ki values for the different
ligands. Using the Cheng–Prusoff equation with the known KM
values of ATP, the Ki values were obtained as summarized in
Table 1. Herein, we focus on the Ki values of the (+)-1a–p
enantiomers, as they were significantly more active, strongly
supporting our modeling studies.
The olefins 10a–p were directly subjected to cyclization re-
actions with para-formaldehyde and glycine to afford the cor-
responding pyrrolidines, which were subsequently N-protected
(Boc2O) to afford the respective carbamates (Æ)-13a–p in mod-
erate overall yields (Scheme 3). The tert-butylcarbamate group
greatly facilitated the optical resolution of (Æ)-13a–p by chiral
HPLC to obtain (+)-13a–p and (À)-13a–p, respectively, in
enantiomerically pure form. Removal of the protecting group
by treatment with 2m aq. HCl in THF afforded the target com-
pounds (+)-1a–p and (À)-1a–p which were isolated as HCl
salts.[27] The absolute configuration of (+)-1c was assigned by
modeling the optical rotatory dispersion (ORD) of (3S,4R)-
(+)-1c, employing density functional theory methods. The cal-
culated [a]D value of +47.78 for chlorine derivative (+)-1c in
the four lowest-energy conformations is in acceptable agree-
The majority of the optically pure inhibitors displayed nano-
molar affinity (Ki values) against PKA, while being significantly
less active against PKAB3 (Table 1). We obtained X-ray co-crys-
tal structures of inhibitors (+)-1b,c,e,l,p with PKA and two
structures of (+)-1c,p with PKAB3. All X-ray co-crystal struc-
tures with PKA were loaded to the PDB database ((+)-1b (PDB
ID: 4UJB, 1.95 Š), (+)-1c (PDB ID: 4UJ1, 1.76 Š), (+)-1e (PDB
ID: 4UJ2, 2.02 Š), (+)-1l (PDB ID: 4UJ9, 1.87 Š) and (+)-1p
(PDB ID: 4UJA, 1.95 Š)), as well as with PKAB3 ((+)-1c (PDB ID:
4Z84, 1.55 Š), (+)-1p (PDB ID: 4Z83, 1.80 Š)). In addition to the
observed biological affinities, they form the basis of our analy-
sis and are discussed below.
Binding mode
20
ment with the measured [a]D of +77.58 (c=0.1 in MeOH)
(see the Supporting Information, Section S2.1).[28] Ultimate
All bound inhibitors observed in the X-ray co-crystal structures
of PKA superimpose well, and adopt a similar binding mode
proof of the absolute configurations of (+)-1b,c,e,l,p was later
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
214
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Scheme 3. 1,3-Dipolar cycloaddition to pyrrolidines 1a–p. i) NaH or tBuOK, THF, 0 to 228C, 5–29 h, 50–100%; ii) (H2CO)n, glycine, toluene, 1208C, 1–4 h;
iii) Boc2O, MeOH, 0 to 228C, 1.5–24 h, chiral HPLC, 14–45%; iv) 2m aq. HCl, THF, 0 to 228C, 1–2 d, 15–98%. Boc=tert-butyloxycarbonyl.
Selectivity of the inhibitors
within the hinge region and in the ribose pocket, in good
agreement with the modeling predictions. Their phenyl/thienyl
vectors reach into the glycine-rich loop region (Figure 2a). In
the X-ray co-crystal structure of (+)-1b bound to PKA (Fig-
ure 2b), the fluorinated inhibitor is anchored with two hydro-
gen bonds to the hinge region (d(Nligand···OGlu121)=
2.6 Š, d(Oligand···NVal123)=3.2 Š) and by an additional
hydrogen bond to the gatekeeper residue Thr183
(d(Nligand···OThr183)=2.8 Š). The flat quinazolinone
moiety is intercalated between the hydrophobic side
chains of Ala70 and Val57 on one side and Leu173 on
the other, at distances ranging from 3.2 to 4.2 Š. The
central, most probably protonated pyrrolidinium ring
is perfectly orientated to reach into the ribose pocket
where it forms an ionic hydrogen bond to the back-
bone C=O of Glu170 (d(Nligand···OGlu170)=3.0 Š) and
a salt-bridge-type hydrogen bond to the side chain
of Asp184 (d(Nligand···OAsp184)=2.7 Š) (Figure 2c). Addi-
tional longer contacts exist between the pyrrolidini-
um N-atom and the side chains of Glu127 and
Asn171. An orthogonal dipolar C=O···C=O interaction
of the benzoyl C=O group of the inhibitor to the
C=O of Gly50, at a distance of 3.0 Š and an angle
C···C=O of 968 (Figure 2b) completes the interaction
pattern observed in all X-ray co-crystal structures
which guides the different aromatic substituents into
The tested inhibitors were more active against PKA than
PKAB3 (Table 1). Comparison between two X-ray co-crystal
structures of the thienyl inhibitor (+)-1p with PKA and PKAB3,
Figure 2. a) Overlay of the X-ray co-crystal structures of inhibitors (+)-1b (1.95 Š resolu-
tion, PDB ID: 4UJB), (+)-1e (2.02 Š resolution, PDB ID: 4UJ2), (+)-1l (1.87 Š resolution,
the glycine-rich loop. Similar interactions and
a nearly identical alignment of the quinazolinone-pyr- PDB ID: 4UJ9), and (+)-1p (1.93 Š resolution, PDB ID: 4UJA) bound to PKA. Color code:
(1b)ligand/enzyme red, (1e)ligand/enzyme blue, (1l)ligand/enzyme green, (1p)ligand/enzyme magenta. The
rolidine moiety in all inhibitor series allowed a com-
four inhibitors show a nearly identical binding mode to the active site of PKA. b) Binding
parison of the PKA and PKAB3 binding modes and
mode of inhibitor (+)-1b. c) Detailed view of the binding mode in the ribose pocket. Dis-
a comprehensive SAR analysis of the interactions
with the glycine-rich loop.
tances are given in Š and interactions are shown as dashed lines. Color code: Cenzyme
gray, Cligand green, N blue, O red, S yellow, F light blue.
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
215
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
respectively, reveals the origin of this difference in binding af-
finity. Modeling with MOLOC had predicted that the ligand
(magenta in Figure 3) undergoes a salt-bridge-type hydrogen
bond to the carboxylate side chain of Glu127 (d(N···O)=2.9 Š).
In the X-ray co-crystal structures (Figure 3, green ligand in PKA
and blue ligand in PKAB3 shown with the protein environment
of PKAB3), the pyrrolidine ring of the inhibitor reaches deeper
into the ribose pocket to form an ionic hydrogen bond with
the backbone C=O of Glu170 (d(N···O)=2.8 Š (PKAB3) and
2.9 Š (PKA)) and a salt bridge with Asp184 (d(N···O)=2.7 Š
(PKAB3) and 2.5 Š (PKA)). As a consequence, the methyl group
of Met173 in PKAB3 is forced to an outwards orientation in an
energetically less favored gauche conformation. In the inward,
anti orientation (see Figure 1b) a steric clash with the C-atom
a to the pyrrolidine ring N-atom would occur, at a modeled,
strongly repulsive distance of d(H3CMet173···C(2)pyrrolidine)=2.3 Š.
As a result of this conformational flip, Met173 in PKAB3 inter-
acts less favorably with the quinazolinone ring of the inhibitor.
The deeper positioning of the pyrrolidine ring in the ribose
pocket, on the other hand, does not affect the favorable loca-
tion of the shorter Leu173 side chain in PKA. The outward
gauche conformation of Met173 was found in both X-ray co-
crystal structures obtained for PKAB3 with inhibitor (+)-1c and
(+)-1p. Overall, the flip of the methionine side chain to a less
favorable conformation, together with the reduced interactions
with the quinazolinone ring, might contribute to the lower af-
finity of the inhibitors towards PKAB3.
Interactions of meta-substituted phenyl rings with the
glycine-rich loop
The meta-substituted inhibitors are generally less active than
their para-substituted analogues and do not possess as many
favorable interactions with the glycine-rich loop (Figure 4). X-
ray co-crystal structures of inhibitors (+)-1b, (+)-1c, and even
(+)-1e with PKA were obtained. The iodide derivative (+)-1e
slowly decomposes at ambient conditions, leading to unreli-
able Ki values; yet an X-ray co-crystal structure could be
Figure 4. X-ray co-crystal structures of meta-aryl-substituted inhibitors
(+)-1b (1.95 Š resolution, PDB ID: 4UJB), (+)-1c (1.76 Š resolution, PDB ID:
4UJ1) and (+)-1e (2.02 Š resolution, PDB ID: 4UJ2) bound to PKA. a) Surface
(grey) of the active site with bound fluorinated inhibitor (+)-1b. b) Closer
look into the interactions of the fluorophenyl ring with the glycine-rich loop.
c) Binding mode of iodo-substituted inhibitor (+)-1e in the glycine-rich loop
region. d) Interactions of chloro-substituted inhibitor (+)-1c with the gly-
cine-rich loop, with the ligand-induced flip of the peptide bond Ser53-
Phe54. Distances are given in Š, and interactions are shown as dashed lines.
Color code: Cenzyme gray, O red, N blue, Cligand green, F light blue, I violet, Cl
limon. Water molecules are shown as red spheres.
Figure 3. Overlay of inhibitor (+)-1p modeled in PKAB3 (PDB ID: 2uw0; ma-
genta) and in the X-ray co-crystal structures of (+)-1p bound to PKA (green,
1.93 Š resolution, PDB ID: 4UJA) and PKAB3 (blue, 1.80 Š resolution, PDB ID:
4Z83). The protein surrounding shown is PKAB3 (PDB ID: 4Z83). Distances
for inhibitor (+)-1p bound to PKAB3 (PDB ID: 4Z83) and the modeled inhibi-
tor are given in Š, and interactions are shown as dashed lines. Color code:
Cenzyme gray, Cligand,XÀray green and blue, Cligand, modeled magenta, N blue, O red,
S yellow, Br dark red.
SAR analysis
solved.
With a plausible explanation for the differences in selectivity,
the inhibitors were analyzed for their binding with PKA. As
they all display the same binding pattern, a meaningful com-
parative analysis of their contacts with the glycine-rich loop
became possible. The following discussion proceeds from the
weaker ligands with a meta-substituted phenyl ring, to the
stronger ligands with para-substituted phenyl rings, and to the
thienyl derivatives.
Remarkably, all substituents in meta-position, with the ex-
ception of fluorine in (+)-1b, reduce binding affinity with re-
spect to the unsubstituted phenyl derivative (++)-1a (Table 1).
The glycine-rich loop is rather open on the side approached
by the aromatic ring (Figure 4a), and the meta-substituents
can orient in different ways. The halogenated aromatic ring of
(+)-1b and (+)-1c is sandwiched between the peptide bond
Thr51–Gly52 of the loop and the hydrophobic side chain of
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
216
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Lys72, which occupies a conserved position in all co-crystal
structures due to a salt bridge with the side chain of Glu91
(Figure 4a, d(N···O)=2.7 Š). The organofluorine in (+)-1b favor-
W34; additionally for steric reasons, it can geometrically
engage less well in attractive interactions with the loop.
The interactions of the F3C-phenyl ring of (+)-1l with the
loop are indeed extensive (Figure 5b). Besides the mentioned
sandwiching of the phenyl ring, which involves the parallel-
+
ably interacts with the NH3 terminus of Lys72 at a distance
d(F···N) of 2.9 Š (Figure 4b). A similar interaction is also seen
for organoiodine in (+)-1e (d(I···N) of 3.4 Š, Figure 4c); howev-
er, due to the reduced polarization of the CÀI bond, as com-
pared to CÀF, a gain in ligand binding strength is not ob-
served. The organochlorine in (++)-1c takes a different orienta-
tion, pointing inwards towards Gly55 to engage in an orthogo-
nal dipolar interaction (d(Cl···C=OGly55)=2.9 Š, a(Cl···C=OGly55)=
928), however without identifiable gain in binding strength.
This orientation of the chlorine apparently disturbs the geome-
try of the loop as seen by a complete flip of the peptide bond
between Ser53 and Phe54 (Figure 4d). Stacking interactions of
the chlorophenyl ring with the loop are nearly eliminated,
which presumably explains the lowest binding affinity in the
meta-halophenyl series.
A high degree of solvation is observed in the X-ray co-crystal
structures. Particularly one water molecule (W34, Figure 4b
and c), located on the aromatic face of Phe54 in the structures
of fluoride (+)-1b (Figure 4b) and iodide (+)-1e (Figure 4c),
needs to be considered in the energetics of protein–ligand
binding (see below). Its major interaction with the protein is
a short H-bond (d(O···N)=2.7 Š) to the backbone N–H of
Phe54; additionally OÀH···p H-bonding to the phenyl ring of
Phe54 may be assumed. It further interacts with one water
molecule in a larger water cluster. Compared to bulk water, the
solvation of W34 is clearly reduced, and therefore, it can be
most probably considered as enthalpically strained.[20,30]
Figure 5. X-ray co-crystal structure of compound (+)-1l bound to PKA
(1.78 Š resolution, PDB ID: 4UJ9). a) Structure of (+)-1l and a closer look
into the glycine-rich loop, where the water molecule W34 is replaced.
b) Contacts between the F3C-phenyl ring in the glycine-rich loop region. Dis-
tances are given in Š, and interactions are shown as dashed lines. Color
code: Cenzyme gray, Cligand green, O red, N blue, F light blue. Water molecules
are shown as red spheres.
slipped stacking of the ring with the peptide bond Thr51-
Gly52,[32] the CF3 substituent engages in multiple favorable
contacts. The complete set of interactions is shown in the Sup-
porting Information, Figure S2. The CF3 group undergoes three
orthogonal dipolar CÀF···C=O interactions with the loop back-
bone,[33] to the backbone C=O of Gly55 (d(F···C)=3.3 Š, a(F···C=
O)=938), of Phe54 (d(F···C)=3.6 Š, a(F···C=O)=1068), and
Gly52 (d(F···C)=3.6 Š, a(F···C=O)=1278) (Figure 5b). Additional-
ly, several favorable CÀF···HÀN contacts are observed, such as
with Lys72 (d(F···N)=3.7 Š) and Ser53 (d(F···N)=3.5 Š (see the
Supporting Information, Figure S2). The CF3 group further in-
teracts at short F···C distances around 3.1–3.2 Š with the HÀCa
bonds of loop residues Gly55 and Gly52, and maintains other
C-F···H-C contacts to the side chains of Phe54 and Leu74. Clear-
ly, the CF3 group is embedded into a truly fluorophilic environ-
ment.[19b,34]
Interactions of para-substituted phenyl rings with the
glycine-rich loop
The binding affinities of the para-substituted inhibitors, except
fluorine derivative (+)-1h, are greatly increased, by up to an
order of magnitude in Ki value, with respect to the unsubstitut-
ed phenyl derivative 1a. The co-crystal structure of (+)-1l
bound to PKA (Ki =2 nm) was solved and revealed an extended
interaction network[31] between the para-F3C-phenyl ring and
the loop. An obvious first advantage of the para-series, com-
pared to the meta-series, is the absence of different conforma-
tions of the substituted ring, which again is sandwiched be-
tween the loop (d(CPhe···CThr51)=3.9 Š) and the side chain of
Lys72 (d(CPhe···CLys72)=4.2 Š). The total spacing in the sandwich
(8.1 Š) is somewhat larger than needed for the intercalation of
an aromatic ring (6.8–7.2 Š), which is favorable to minimize
binding-induced losses in conformational entropy (“wiggling”)
of the phenyl substituent and the flexible loop (see Figure S1
in the Supporting Information).
The experimental finding that the water molecule W34, seen
in the co-crystal structures of the meta-series, had been dis-
placed by the para-CF3 group of (+)-1l led us to perform
a PDB search (see the Supporting Information, Section 3) to
obtain further evidence for the presence of a conserved water
molecule at this position. This analysis showed that the active
pocket of PKA is highly solvated and that solvation extends to
the glycine-rich loop. Most interestingly, the water molecule
W34 was found in eight of the thirteen analyzed structures
with a loop conformation similar to the one seen in our study.
Engh and co-workers had previously shown that the glycine-
rich loop is quite flexible[3] (see Supporting Information, Fig-
ure S1); therefore only a small number of structures were in-
corporated into the analysis. Still this investigation gave us fur-
The CF3 substituent displaces the previously mentioned
water molecule W34 into the bulk solvent (Figure 5a). The fa-
vorable energetics of this water replacement contribute to the
difference between the Ki value of the para-fluoro-inhibitor
(+)-1h (70 nm) and those of (+)-1i–l (2–3 nm) with bulkier
substituents (Table 1). Most probably, the F-atom is too small,
as compared to the larger substituents, to efficiently displace
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
217
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
ther evidence for the hypothesis, that the para-substituted in-
hibitors gain energy by replacing that water molecule.[35] As
the water molecule W34 seems to be energetically frustrat-
ed[20,30] (see above), the replacement is expected to be favor-
able. A water score analysis further supports this hypothesis. In
this analysis, we used a geometric scoring similar to the Rank
score developed by Kellogg and co-workers, by exploring the
coordination of structural water to neighboring water mole-
cules as well as to the protein.[36]
Dunitz and co-workers had previously analyzed the geometries
for nonbonded atomic contacts of divalent sulfur with electro-
philes and nucleophiles.[42]
The bromine is surrounded by the glycine-rich loop and un-
dergoes similarly favorable interactions to those discussed
before for the CF3 group of inhibitor (+)-1l (Figure 6b). Halo-
gen bonding was not observed as the interaction distances
and angles do not fulfill the rigorous geometric require-
ments.[43]
An additional contribution to binding affinity could originate
from the favorable dipolar alignment of the bromothienyl ring
and the peptide bond Gly52ÀThr51 in the loop, which under-
go parallel-shifted amide–p stacking at a distance of 3.9–4.1 Š.
Harder et al. reported a computational study which revealed
that the energetics of this stacking interaction is strongest if
strong dipoles are aligned in an antiparallel fashion.[44] The
best inhibitors (+)-1p and (+)-1l exhibit substantial local
dipole moments in their aromatic rings (3.6 and 2.4 Debye)
with favorable alignment angles of 1398 and 1638, respectively
(see Supporting Information, Figure S5).
Interactions of the thienyl inhibitors with the glycine-rich
loop
A sharp increase in binding affinity was observed upon adding
halides in ortho-position to the sulfur atom of the thiophene
substituent. While the parent thiophene ligand (+)-1n (Ki =
170 nm) is less potent than the phenyl compound (+)-1a (Ki =
30 nm), addition of a chlorine enhances binding to a Ki value
of 23 nm and bromination afforded the most potent inhibitor
of all three series, (+)-1p, with a Ki value of 0.9 nm, Figure 6).
Gratifyingly, a co-crystal structure of (+)-1p with PKA was ob-
tained. In this structure, some electron density is observed in
proximity of Phe54, which is attributed to a bromine atom, de-
rived from radiation damage, rather than the conservation of
a water molecule. The close proximity of the inhibitor to the
phenyl ring (4.6 Š) further supports this.[37]
The numerous weak intermolecular interactions of (+)-1p
with the glycine-rich loop are indeed remarkable and result in
the nearly sub-nanomolar inhibition (Ki value) of PKA. Although
the contribution of each isolated contact cannot be quantified
on its own, their simultaneous presence is important for the
high inhibitory activity of the bromothienyl ligand.
The very strong binding is again a result of an extensive in-
teraction network in the complex. The thiophene inhibitor
adopts the expected cis conformation, enforced by an intramo-
lecular chalcogen bonding interaction of one of the two s*-or-
bitals of the divalent sulfur atom with the lone pair of the C=O
oxygen atom.[18c,38–41] The conformational preorganization
through this intramolecular 1,4-S···O interaction[18c] (d(S···O)=
3.0 Š) directs the bromothienyl ring into the glycine-rich loop
(Figure 6). Here, the highly polarizable and apparently ideally
sized bromine atom of (+)-1p finds a perfect environment for
multiple dispersive and orthogonal dipolar interactions.
Computational analysis of the interaction network
To further decipher the impact of multiple interactions ob-
served in the binding of ligand (+)-1p to PKA, an enzyme-
ligand network interaction analysis, as developed by Kuhn
et al., was performed.[31] It specifically identifies networks of fa-
vorable enzyme–ligand interactions and categorizes close
enzyme–ligand contacts to distinct classes of interactions,
which allows an advanced analysis of binding modes and pro-
vides a rational for biological activity.
The thiophene sulfur atom undergoes additional interactions
A large number of favorable contacts are detected around
with neighboring backbone C=O groups in the loop (d(S···C= the carbonylbromothienyl moiety (Figure 7b; for the full net-
OThr51)=3.4 Š, d(S···C=OGly55)=4.1 Š and d(S···C=OArg56)=3.9 Š;
Figure 6a), which presumably are mostly dispersive.[18c,38–41]
work see the Supporting Information, Figure S3). In this analy-
sis, the ligand C=O interacts with the polarized Ca of Gly50 by
Figure 6. X-ray co-crystal structure of inhibitor (+)-1p bound to PKA (1.93 Š resolution, PDB ID: 4UJA). a) S···C=O contacts. b) Contacts between the Br atom
and the glycine-rich loop. The amide–p stacking distance is also shown. Distances are given in Š. Color code: Cenzyme gray, Cligand green, O red, N blue, S
yellow, Br dark red.
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
218
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
a weak hydrogen bond (d=3.0 Š). The thiophene moiety un-
dergoes several dispersive interactions with the glycine-rich
loop, and p–p stacking with the backbone amide moiety is de-
tected. Finally, the bromine atom is suitably embedded in an
environment with three close apolar contacts to the enzyme
(d(Br···enzyme)=3.3–3.6 Š). The more quantitative network
contribution analysis reveals six ligand atoms with remarkably
high scores (ꢀ1.0), suggesting that these may act as binding
hot-spots (Figure 7a). They comprise three atoms of the hinge-
binding quinazolinone heterocycle, the pyrrolidine nitrogen
atom in the ribose pocket, and the thienyl sulfur and bromine
atoms of the loop-binding substituent. According to the net-
work analysis, the interactions of the 5-bromothienyl moiety
with the glycine-rich loop clearly make an important contribu-
tion to the excellent ligand binding affinity of (+)-1p.
able water replacement by the ligand in the loop region con-
tributes to binding strength. The aromatic rings of the inhibi-
tors interact with the loop by undergoing amide–p stacking in-
teractions. The loop was best addressed with para-CF3-phenyl
and 5-bromothienyl rings which both feature in the co-crystal
structures a remarkable network of weak interactions, includ-
ing orthogonal dipolar contacts and dispersive interactions.
While the energetic contribution of each contact cannot be
quantified on its own, the ensemble of weak interactions with
the loop is key to high inhibitory activity. The interaction net-
work was analyzed computationally for the bromothienyl-sub-
stituted ligand, and both the sulfur and bromine atoms were
identified as interaction hotspots. In summary, this work
strongly validates binding to the glycine-rich ring as a promis-
ing strategy in the development of potent inhibitors of protein
kinases.
Conclusion
Experimental Section
We described design, synthesis, and biological evaluation of
a new family of enantiopure inhibitors for protein kinase A
(PKA) and its mutant PKAB3 as mimic of PKB. The ligands bind
to the ATP site of the proteins, with a quinazolinone ring inter-
acting with the hinge, a most probably protonated pyrrolidine
ring filling the ribose pocket, and substituted aromatic rings
(phenyl, thienyl) binding to the glycine-rich loop, the ATP-tri-
phosphate binding site. Depending on the nature and substi-
tution of these aromatic vectors, binding affinities down to Ki
values in the single-digit nanomolar range were measured. The
inhibitors exhibited a preference for binding to PKA over
PKAB3, the origin of which could be elucidated by X-ray co-
crystal structure determination. A total of seven co-crystal
structures revealed a conserved binding mode, which allowed
establishment of an SAR, with a focus on the interactions of
the ligands with the glycine-rich loop. This loop has not been
addressed in most protein kinase inhibitor design, but this
work shows that proper ligand interactions with the loop can
greatly enhance binding potency. The data suggest that the
glycine-rich loop is best addressed by an intricate network of
dispersive contacts and dipolar interactions. In addition, favor-
Materials and methods: For detailed descriptions of the chemical
analytical equipment, see the Supporting Information S5. This Sec-
tion provides general procedures for the pyrrolidine ring synthesis
and deprotection, and reports the preparation of inhibitor (+)-1j.
All other synthetic protocols, analytical data, and the co-crystal
structure analyses are included in the Supporting Information, Sec-
tion S5.
General procedure A (GP-A) for the phosphonate synthesis
A solution of diethyl methylphosphonate (1.5 equiv) in THF (10 or
20 mL) was treated at À788C dropwise with a 1.6m nBuLi solution
in hexane (1.5 equiv), stirred for 10 min, treated with a solution of
the halogenated ethyl-benzoate (2) (1 equiv) in THF (10 or 20 mL),
stirred for 2 h, and diluted with sat. aq. NH4Cl solution (200 mL).
The aqueous phase was extracted with EtOAc (3”200 mL), and the
combined organic phases were dried over MgSO4, filtered, and
evaporated.
General procedure B (GP-B) for the olefin synthesis
1) A suspension of 60wt% NaH on mineral oil (1 equiv) in THF
(1.5 mL) was treated with a solution of the corresponding phos-
phonate 4 (1 equiv) in THF (3.0 mL) at 08C. The mixture was stirred
at 08C for 10 min, treated with aldehyde 5 (1 equiv), stirred for
30 min at 08C, warmed to 228C, and stirred for additional 0.25–
Figure 7. Interaction network analysis of inhibitor (+)-1p bound to PKA (1.93 Š resolution, PDB ID: 4UJA). a) Atom-based score contributions in a blue to red
color scheme, with gray indicating no interactions. Binding hot spots (contribution score ꢀ 1) are shown in red with the highest scores for: Br 1.7, S 2.2,
Npyrrolidine 1.6, Nquinazolinone 1.2, Oquinazolinone 1.0, and Cquinazolinone 1.0 b) Illustration of the favorable interaction network around the Gly-rich loop. Color code: Cenzyme
gray, Cligand green, O red, N blue, S yellow, Br dark red. Interactions are shown as dashed lines. The color code for interactions is hydrogen bond: red, disper-
sive: yellow, ionic: pink, cation-dipole: magenta, p···p: orange.
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
219
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
22 h before being diluted with a sat. aq. NH4Cl solution (50 mL).
The aqueous phase was extracted with CH2Cl2 (3”50 mL). The
combined organic phases were dried over MgSO4, filtered, and
evaporated.
J=8.8 Hz, 1H; H-C(5)), 8.87 ppm (s, 1H; H-C(2)); 13C NMR (100 MHz,
CDCl3): d=54.52 (OMe), 117.48 (C(4a)), 124.35, 124.38, 125.73 (C(5)),
128.36 (CH=CH-C=O), 128.41 (C(4’)), 130.10 (2 C; C(2’)), 132.11 (2 C;
C(3’)), 136.52 (C(1’)), 139.64 (C(7)), 143.58 (CH=CH-C=O), 151.22
(C(8a)), 155.30 (C(2)), 166.98 (C(4)), 188.84 ppm (C=O); IR (ATR):
=2923 (w), 2853 (w), 1658 (m), 1619 (w), 1586 (m), 1574 (s), 1495
(m), 1455 (m), 1397 (w), 1380 (s), 1361 (s), 1322 (m), 1276 (m), 1209
(m), 1168 (m), 1096 (m), 1071 (m), 1025 (m), 993 (s), 977 (m), 906
(w), 875 (m), 862 (m), 813 (s), 784 (s), 764 (w), 745 (m), 732 (m), 682
2) A suspension of KOtBu (1 equiv) in THF (1.5 mL) was treated
with a solution of the corresponding phosphonate 4 (1 equiv) in
THF (3.5 mL) at 08C, stirred for 30 min, treated with aldehyde 5
(1 equiv), stirred for 20 min at 08C, for 5–22 h at 228C, and diluted
with sat. aq. NH4Cl (50 mL). The aqueous phase was extracted with
CH2Cl2 (3”50 mL). The combined organic phases were dried over
MgSO4, filtered, and evaporated.
(m), 673 (m), 629 cmÀ1 (m); HR-MS (ESI): m/z (%): 371.0211 (63,
+
[M+H]+, calcd for C18H1481BrN2O2
[M+H]+, calcd for C18H1479BrN2O2
:
371.0213), 369.0230 (67,
+
: 369.0233), 177.1271 (96),
General procedure C (GP-C) for the pyrrolidine synthesis
150.0367 (100).
A solution of alkene 10 (1 equiv) in toluene (6 mL) at 1008C or
1208C (with Na2SO4 in a frit above) was simultaneously treated
with para-formaldehyde (6 equiv) and glycine (2 equiv), stirred at
1208C for 2 h, treated with additional para-formaldehyde (6 equiv)
and glycine (2 equiv), stirred for 2 h at 1208C, cooled to 228C, and
the solvent was removed under reduced pressure. A solution of
the crude intermediate in MeOH (6 mL) was treated with Boc2O
(1.2 equiv), stirred at 228C overnight, and the solvent was removed
under reduced pressure.
(+)- and (À)-tert-Butyl (3,4-trans)-3-(4-bromobenzoyl)-4-(4-me-
thoxyquinazolin-7-yl)pyrrolidine-1-carboxylate ((+)-13 and (À)-
13): GP-C starting from alkene 10j (67 mg, 0.18 mmol). FC (SiO2;
MeOH/CH2Cl2 1:99
!
2.5:97.5) and chiral HPLC (Reprosil;
15 mLminÀ1
,
hexane/iPrOH/CHCl3 75:3:22) gave pyrrolidines
(+)-13j (18 mg, 19%) and (À)-13j (19 mg, 20%) as colorless waxes.
Data of (+)-13j: Rf =0.33 (SiO2; MeOH/CH2Cl2 1:19); tR (prep.
HPLC)=40 min (Reprosil; 15 mLminÀ1
,
hexane/iPrOH/CHCl3
1
75:3:22); [a]D20 = +23.2 (c=0.1 in CHCl3); H NMR (400 MHz, CDCl3;
approx. 1:1 mixture of diastereoisomers, assignment based on
HSQC and HMBC spectra of (+)-13j): d=1.51 (s, 9H; C(CH3)3), 3.60
(dd, J=10.5, 7.9 Hz, 1H; Ha-C(2 or 5)), 3.64–3.71 (m, 1H; Ha-C(2 or
5)), 3.96–4.26 (m, 4H; Hb-C(2), H-C(3), H-C(4), Hb-C(5)), 4.18 (s, 3H;
CH3), 7.49 (dd, J=8.5, 1.4 Hz, 1H; H-C(6’’)), 7.57, 7.74 (AA’MM’, J=
8.6 Hz, 4H; H-C(2’), H-C(3’)), 7.82 (br. s, 1H; H-C(8’’)), 8.11 (d, J=
8.5 Hz, 1H; H-C(5’’)), 8.80 ppm (s, 1H; H-C(2’’)); 13C NMR (100 MHz,
CDCl3; approx. 1:1 mixture of diastereoisomers): d=28.48, 46.43
and 46.74, 49.39, 51.91 and 51.97, 52.20 and 52.60, 54.36, 80.11,
115.72, 124.26, 125.57, 127.10, 129.13, 129.87, 132.17, 134.70,
146.16 and 146.34, 151.13, 153.97 and 154.03, 154.88, 166.95,
196.93 ppm; IR (ATR): =2974 (w), 2937 (w), 2876 (w), 1682 (s), 1624
(m), 1581 (m), 1567 (m), 1496 (m), 1454 (m), 1377 (s), 1256 (w),
1167 (m), 1121 (m), 1098 (m), 1070 (m), 1007 (m), 976 (w), 877 (m),
839 (m), 796 (m), 769 (m), 749 (m), 688 cmÀ1 (m); HR-ESI-MS: m/z
General procedure D (GP-D) for the deprotection of pyrrolidines
A solution of pyrrolidine 13 (1 equiv) in THF (1 mL) at 08C was
treated with 2m aq. HCl (1 mL) and stirred for 1–48 h at 228C, until
LCMS showed complete conversion. The mixture was evaporated,
triturated with Et2O (2 mL), and evaporated. HPLC (amino phase:
LiChrospher 100 NH2 (5 mm); 15 mLminÀ1, MeOH/CHCl3 10:90),
evaporation, treatment with 1.25m HCl in MeOH (1 mL) at 08C,
evaporation, precipitation from Et2O, and drying under HV afforded
the pyrrolidines as HCl salts.
Diethyl [2-(4-bromophenyl)-2-oxoethyl]phosphonate (4j): GP-A
(10 mL THF) starting from ethyl 4-bromobenzoate (346 mL,
2.07 mmol). FC (SiO2; EtOAc/heptane 4:1) gave phosphonate 4j
(614 mg, quant.) as a colorless oil. Rf =0.26 (SiO2; EtOAc/heptane
4:1); 1H NMR (400 MHz, CDCl3): d=1.30 (td, J=7.1 Hz, 4J(H,P)=
2
2.1 Hz, 6H; 2 OCH2CH3), 3.60 (d, J(H,P)=22.8 Hz, 2H; H-C(1)), 4.09–
(%): 514.1168 (69, [M+H]+, calcd for C25H2781BrN3O4+: 514.1160),
4.19 (m, 4H; 2 OCH2CH3), 7.63, 7.90 ppm (AA’MM’, J=9.1 Hz, 4H;
+
512.1183 (73, [M+H]+, calcd for C25H2779BrN3O4
: 512.1179),
3
C6H4); 13C NMR (100 MHz, CDCl3): d=16.26 (d, J(C,P)=6.2 Hz, 2 C;
2 OCH2CH3), 38.67 (d, 1J(C,P)=128.4 Hz, 1 C; C(1)), 62.74 (d,
2J(C,P)=6.4 Hz, 2 C; 2 OCH2CH3), 129.07 (C(4’)), 130.51 (2 C; C(2’)),
458.0540 (99), 456.0556 (100).
Data of (À)-13j: tR (prep. HPLC)=58 min; [a]D20 =À23.1 (c=0.1 in
131.94 (2 C; C(3’)), 135.24 (d, 3J(C,P)=1.7 Hz,
1
C; C(1’)),
CHCl3); 1H NMR and HR-ESI-MS data consistent with (+)-13j.
190.95 ppm (d, 2J(C,P)=6.5 Hz, 1 C; C=O); 31P NMR (162 MHz,
(+)-(3R,4S)-3-(4-Bromobenzoyl)-4-(4-oxo-3,4-dihydroquinazolin-
7-yl)pyrrolidinium Chloride ((+)-1j): GP-D starting from pyrroli-
dine (+)-13j (18 mg, 0.04 mmol) gave pyrrolidinium (+)-1j (11 mg,
72%) as a white solid. M.p. 1478C (decomp.); [a]D20 = +66.2 (c=
2
3
CDCl3): d=19.33 ppm (tquint., J(P,H)=24.1 Hz, J(P,H)=8.1 Hz, 1 P;
PO3); IR (ATR): =2982 (w), 2932 (w), 1679 (m), 1585 (m), 1568 (w,
sh), 1484 (w), 1443 (w), 1395 (w), 1248 (s), 1200 (w), 1163 (w), 1098
(w), 1052 (s, sh), 1019 (s), 998 (s), 956 (s), 872 (w), 802 (s), 738 (m),
704 cmÀ1 (w); HR-MS (ESI): m/z (%): 337.0027 (100, [M+H]+, calcd
for C12H1781BrO4P+: 337.0022), 335.0044 (88, [M+H]+, calcd for
C12H1779BrO4P+: 335.0042).
1
0.1 in MeOH); H NMR (400 MHz, (CD3)2SO): d=3.36–3.50 (m, 2H;
Ha-C(2), Ha-C(5)), 3.75–3.96 (m, 2H; Hb-C(2), Hb-C(5)), 3.91 (dd, J=
17.2, 10.0 Hz, 1H; H-C(4)), 4.62 (t, J=8.9 Hz, 1H; H-C(3)), 7.60 (dd,
J=8.4, 2.0 Hz, 1H; H-C(6’’)), 7.68, 7.74 (AA’MM’, J=8.4 Hz, 4H;
C6H4Br), 7.68 (d, J=1.6 Hz, 1H; H-C(8’’)), 8.07 (d, J=8.2 Hz, 1H; H-
C(5’’)), 8.25 (s, 1H; H-C(2’’)), 9.60 and 9.86 (2 br. s, 2H; H2N+), 12.0–
13.0 ppm (br. s, 1H; H-N(3’’)); 13C NMR (100 MHz, (CD3)2SO, signal
of C(2) not visible due to fast exchange, assignment based on
HSQC and HMBC spectra of (À)-1i): d=46.79 (C(4)), 47.80 (C(2 or
5)), 51.05 (C(2 or 5)), 51.45 (C(3)), 122.02 (C(4a’’)), 125.90 (C(8’’)),
127.00 (C(5’’)), 127.17 (C(6’’)), 128.67 (C(4’)), 130.84 (2 C; C(2’)),
132.39 (2 C; C(3’)), 135.07 (C(1’)), 145.44 (C(7’’)), 146.84 (C(8a’’)),
160.61 (C(4’’)), 197.80 ppm (C=O); IR (ATR): =3600–2200 (m, with
3359 (w), 3030 (m), 2900 (m), 2864 (m), 2709 (m), 2595 (m)), 1713
(s), 1680 (s), 1652 (s), 1621 (m, sh), 1581 (m), 1498 (w), 1463 (w),
1397 (m), 1287 (m), 1244 (m), 1069 (m), 1005 (m), 839 (m), 781 (m),
734 (w), 689 cmÀ1 (m); HR-ESI-MS: m/z (%): 426.9491 (100),
(2E)-1-(4-Bromophenyl)-3-(4-methoxyquinazolin-7-yl)prop-2-en-
1-one (10j): GP-B1 starting from phosphonate 4j (110 mg,
0.33 mmol); 20 h at 228C. MPLC (80 g SiO2; 120 mLminÀ1, MeOH/
CH2Cl2 0:100 ! 2:98) gave impure alkene 10j (77 mg, 64%; (E/Z)
99:1) as an off-white solid, of which a small amount was purified
by HPLC (Reprosil; 15 mLminÀ1, hexane/iPrOH/CHCl3 72:3:25) to
afford alkene 10j as an off-white solid. The compound appears to
react upon concentration and was therefore used for the next step
without further purification. Rf =0.38 (SiO2; MeOH/CH2Cl2 1:19);
1
m.p. 203–2068C; H NMR (400 MHz, CDCl3): d=4.23 (s, 3H; OMe),
7.69 (d, J=15.8 Hz, 1H; CH=CH-C=O), 7.70, 7.95 (AA’MM’, J=
8.4 Hz, 4H; C6H4Br), 7.84 (dd, J=8.5, 1.6 Hz, 1H; H-C(6)), 7.97 (d, J=
15.6 Hz, 1H; CH=CH-C=O), 8.17 (d, J=1.2 Hz, 1H; H-C(8)), 8.22 (d,
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
220
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
+
400.0486 (69, [M+H]+, calcd for C19H1781BrN3O2
: 400.0479),
[20] E. Persch, O. Dumele, F. Diederich, Angew. Chem. Int. Ed. 2015, 54,
3290–3327; Angew. Chem. 2015, 127, 3341–3382.
[21] J. E. Davoren, D. L. Gray, A. R. Harris, D. M. Nason, W. Xu, Synlett 2010,
16, 2490–2492.
398.0501 (68, [M+H]+, calcd for C19H1779BrN3O2+: 398.0499).
[22] M. Joucla, J. Mortier, J. Chem. Soc. Chem. Commun. 1985, 1566–1567.
[23] G. SolladiØ, N. Wilb, C. Bauder, C. Bonini, L. Viggiani, L. Chiummiento, J.
Org. Chem. 1999, 64, 5447–5452.
Acknowledgements
This work was supported by a grant from the ETH Research
Council and by F. Hoffmann-La Roche, Basel. O.D. thanks the
German Fonds der Chemischen Industrie for a KekulØ fellow-
ship. We acknowledge Dr. Alexander Pflug who initiated the
measurements at the Norwegian Structural Biology Centre,
Tromsø.
[24] C. E. F. Rickard, W. R. Roper, F. Tutone, S. D. Woodgate, L. J. Wright, J. Or-
ganomet. Chem. 2001, 619, 293–298.
[25] Z. E. Wilson, J. G. Hubert, M. A. Brimble, Eur. J. Org. Chem. 2011, 3938–
3945.
[26] N. Ghavtadze, R. Narayan, B. Wibbeling, E.-U. Würthwein, J. Org. Chem.
2011, 76, 5185–5197.
[27] Enantiomers (+)-1a–p and (À)-1a–p were further purified on an
amino phase HPLC column, treated with 1.2m HCl in MeOH, and subse-
quently precipitated from Et2O as the mono-HCl salts; see: A. Thaqi, A.
McCluskey, J. L. Scott, Tetrahedron Lett. 2008, 49, 6962–6964.
[28] E. Giorgio, M. Roje, K. Tanaka, Z. Hamersak, V. Sunjic, K. Nakanishi, C.
Rosini, N. Berova, J. Org. Chem. 2005, 70, 6557–6563.
Keywords: glycine-rich loop · interaction networks · kinases ·
molecular recognition · pyrrolidines · water replacement
[29] P. F. Cook, M. E. Neville, Jr., K. E. Vrana, E. Kent, F. T. Hartl, R. Roskoski Jr.,
Biochemistry 1982, 21, 5794–5799.
[1] D. A. Walsh, J. P. Perkins, E. G. Krebs, J. Biol. Chem. 1968, 243, 3763–
3765.
[30] a) F. Biedermann, W. M. Nau, H.-J. Schneider, Angew. Chem. Int. Ed. 2014,
53, 11158–11171; Angew. Chem. 2014, 126, 11338–11352; b) S. G.
Krimmer, M. Betz, A. Heine, G. Klebe, ChemMedChem 2014, 9, 833–846.
[31] B. Kuhn, J. E. Fuchs, M. Reutlinger, M. Stahl, N. R. Taylor, J. Chem. Inf.
Model. 2011, 51, 3180–3198.
[2] S. S. Taylor, J. Wang, J. Wu, N. M. Haste, E. Radzio-Andzelm, G. Anand,
Biochim. Biophys. Acta Proteins Proteomics 2004, 1697, 259–269.
[3] a) E. Šberg, B. Lund, A. Pflug, O. A. B. S. M. Gani, U. Rothweiler, T. M.
de Oliveira, R. A. Engh, Biol. Chem. 2012, 393, 1121–1129; b) T. M. de Oli-
veira, R. Ahmad, R. A. Engh, J. Phys. Chem. A 2011, 115, 3895–3904.
[4] P. Akamine, Madhusudan, J. Wu, N.-H. Xuong, L. F. Ten Eyck, S. S. Taylor,
J. Mol. Biol. 2003, 327, 159–171.
[32] M. Harder, M. A. Carnero Corrales, N. Trapp, B. Kuhn, F. Diederich, Chem.
Eur. J. 2015, 21, 8455–8463.
[33] Orthogonal dipolar interactions with the DFG-loop of the Bcr-Abl
kinase have previously been observed in the X-ray co-crystal structure
with the inhibitor Nilotinib (PDB ID: 3CS9); E. Weisberg, P. W. Manley, W.
Breitenstein, J. Brüggen, S. W. Cowan-Jacob, A. Ray, B. Huntly, D.
Fabbro, G. Fendrich, E. Hall-Meyers, A. L. Kung, J. Mestan, G. Q. Daley, L.
Callahan, L. Catley, C. Cavazza, A. Mohammed, D. Neuberg, R. D. Wright,
D. G. Gilliland, J. D. Griffin, Cancer Cell 2005, 7, 129–141.
[34] a) J. Olsen, D. W. Banner, P. Seiler, B. Wagner, T. Tschopp, U. Obst-Sander,
M. Kansy, K. Müller, F. Diederich, ChemBioChem 2004, 5, 666; b) T. Vogt,
M. Salwicek, B. Koksch in “Fluorine in Pharmaceutical and Medicinal
Chemistry” (Eds.: V. Gouverneur, K. Müller), Imperial College Press,
London, 2012, Chapter 2, pp. 33–90.
[35] S. Riniker, L. J. Barandun, F. Diederich, O. Krämer, A. Steffen, W. F. van
Gunsteren, J. Comput. Aided Mol. Des. 2012, 26, 1293–1309.
[36] A. Amadasi, J. A. Surface, F. Spyrakis, P. Cozzini, A. Mozzarelli, G. E. Kel-
logg, J. Med. Chem. 2008, 51, 1063–1067.
[37] For a report on a similar radiation damage in the co-crystal structure of
a protein-bound aryl halide ligand, see: C. Koch, A. Heine, G. Klebe, J.
Synchrotron Radiat. 2011, 18, 782–789.
[38] K. A. Brameld, B. Kuhn, D. C. Reuter, M. Stahl, J. Chem. Inf. Model. 2008,
48, 1–24.
[39] H. V. Le, D. D. Hawker, R. Wu, E. Doud, J. Widom, R. Sanishvili, D. Liu,
N. L. Kelleher, R. B. Silverman, J. Am. Chem. Soc. 2015, 137, 4525–4533.
[40] G. E. Garrett, G. L. Gibson, R. N. Straus, D. S. Seferos, M. S. Taylor, J. Am.
Chem. Soc. 2015, 137, 4126–4133.
[41] C. Bleiholder, D. B. Werz, H. Kçppel, R. Gleiter, J. Am. Chem. Soc. 2006,
128, 2666–2674.
[42] R. E. Rosenfield, R. Parthasarathy, J. D. Dunitz, J. Am. Chem. Soc. 1977,
99, 4860–4862.
[5] S. S. Taylor, J. Biol. Chem. 1989, 264, 8443–8446.
[6] G. Manning, D. B. Whyte, R. Martinez, T. Hunter, S. Sudarsanam, Science
2002, 298, 1912–1934.
[7] a) P. Cohen, D. R. Alessi, ACS Chem. Biol. 2013, 8, 96–104; b) http://
www.brimr.org/PKI/PKIs.htm.
[8] S. S. Staal, J. W. Hartley, W. P. Rowe, Proc. Natl. Acad. Sci. USA 1977, 74,
3065–3067.
[9] D. P. Brazil, Z.-Z. Yang, B. A. Hemmings, Trends Biochem. Sci. 2004, 29,
233–242.
[10] K.-M. V. Lai, M. Gonzalez, W. T. Poueymirou, W. O. Kline, E. Na, E. Zlot-
chenko, T. N. Stitt, A. N. Economides, G. D. Yancopoulos, D. J. Glass, Mol.
Cell. Biol. 2004, 24, 9295–9304.
[11] Z.-Z. Yang, O. Tschopp, M. Hemmings-Mieszczak, J. Feng, D. Brodbeck,
E. Perentes, B. A. Hemmings, J. Biol. Chem. 2003, 278, 32124–32131.
[12] T. G. Davies, M. L. Verdonk, B. Graham, S. Saalau-Bethell, C. C. F. Hamlett,
T. McHardy, I. Collins, M. D. Garrett, P. Workman, S. J. Woodhead, H.
Jhoti, D. Barford, J. Mol. Biol. 2007, 367, 882–894.
[13] M. Gaßel, C. B. Breitenlechner, P. Rüger, U. Jucknischke, T. Schneider, R.
Huber, D. Bossemeyer, R. A. Engh, J. Mol. Biol. 2003, 329, 1021–1034.
[14] J. E. Walker, M. Saraste, M. J. Runswick, N. J. Gay, EMBO J. 1982, 1, 945–
951.
[15] a) A. K. H. Hirsch, F. Fischer, F. Diederich, Angew. Chem. Int. Ed. 2007, 46,
338–352; Angew. Chem. 2007, 119, 342–357; b) S. S. Taylor, A. P. Kornev,
Trends Biochem. Sci. 2011, 36, 65–77.
[16] P. R. Gerber, K. Müller, J. Comput-Aided Mol. Des. 1995, 9, 251–268.
[17] A. Donald, T. McHardy, M. G. Rowlands, L.-J. K. Hunter, T. G. Davies, V.
Berdini, R. G. Boyle, G. W. Aherne, M. D. Garrett, I. Collins, J. Med. Chem.
2007, 50, 2289–2292.
[18] a) E. A. Meyer, R. K. Castellano, F. Diederich, Angew. Chem. Int. Ed. 2003,
42, 1210–1250; Angew. Chem. 2003, 115, 1244–1287; b) L. Salonen, M.
Ellermann, F. Diederich, Angew. Chem. Int. Ed. 2011, 50, 4808–4842;
Angew. Chem. 2011, 123, 4908–4944; c) B. R. Beno, K.-S. Yeung, M. D.
Bartberger, L. D. Pennington, N. A. Meanwell, J. Med. Chem. 2015, 58,
4383–4438.
[43] L. A. Hardegger, B. Kuhn, B. Spinnler, L. Anselm, R. Ecabert, M. Stihle, B.
Gsell, R. Thoma, J. Diez, J. Benz, J.-M. Plancher, G. Hartmann, D. W.
Banner, W. Haap, F. Diederich, Angew. Chem. Int. Ed. 2011, 50, 314–318;
Angew. Chem. 2011, 123, 329–334.
[44] M. Harder, B. Kuhn, F. Diederich, ChemMedChem 2013, 8, 397–404.
[19] a) R. Paulini, K. Müller, F. Diederich, Angew. Chem. Int. Ed. 2005, 44,
1788–1805; Angew. Chem. 2005, 117, 1820–1839; b) K. Müller, C. Fäh, F.
Diederich, Science 2007, 317, 1881–1886.
Received: September 5, 2015
Published online on November 18, 2015
Chem. Eur. J. 2016, 22, 211 – 221
www.chemeurj.org
221
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim