Binding mode and structure-activity relationships around direct inhibitors of the Nrf2-Keap1 complex

Article abstract of DOI:10.1002/cmdc.201300525

An X-ray crystal structure of Kelch-like ECH-associated protein (Keap1) co-crystallised with (1S,2R)-2-[(1S)-1-[(1,3-dioxo-2,3-dihydro-1H-isoindol-2-yl) methyl]-1,2,3,4-tetrahydroisoquinolin-2-carbonyl]cyclohexane-1-carboxylic acid (compound (S,R,S)-1 a) was obtained. This X-ray crystal structure provides breakthrough experimental evidence for the true binding mode of the hit compound (S,R,S)-1 a, as the ligand orientation was found to differ from that of the initial docking model, which was available at the start of the project. Crystallographic elucidation of this binding mode helped to focus and drive the drug design process more effectively and efficiently. To dock or not to dock? Nrf2 has become an attractive neuroprotective target, as the Nrf2 pathway provides a natural cell defense mechanism against damage. Targeting its physiological negative modulator Keap1 with small molecules may allow Nrf2 to play its protective role. To this end, an X-ray structure of Keap1 co-crystallised with compound (S,R,S)-1 a was obtained, elucidating its binding mode, which in turn helped to drive the drug design process.

Full text of DOI:10.1002/cmdc.201300525

CHEMMEDCHEM
COMMUNICATIONS
DOI: 10.1002/cmdc.201300525
Binding Mode and Structure–Activity Relationships around
Direct Inhibitors of the Nrf2–Keap1 Complex
Eric Jnoff,*[a] Claudia Albrecht,[b] John J. Barker,[b] Oliver Barker,[b] Edward Beaumont,[b]
Steven Bromidge,[b] Frederick Brookfield,[b] Mark Brooks,[b] Christian Bubert,[b] Tom Ceska,[c]
Vincent Corden,[d] Graham Dawson,[b] Stephanie Duclos,[b] Tara Fryatt,[b]
Christophe Genicot,[a] Emilie Jigorel,[a] Jason Kwong,[b] Rosemary Maghames,[b]
Innocent Mushi,[b] Richard Pike,[b] Zara A. Sands,[a] Myron A. Smith,[b]
Christopher C. Stimson,[b] and Jean-Philippe Courade[a]
An X-ray crystal structure of Kelch-like ECH-associated protein
(Keap1) co-crystallised with (1S,2R)-2-[(1S)-1-[(1,3-dioxo-2,3-di-
hydro-1H-isoindol-2-yl)methyl]-1,2,3,4-tetrahydroisoquinolin-2-
carbonyl]cyclohexane-1-carboxylic acid (compound (S,R,S)-1a)
was obtained. This X-ray crystal structure provides break-
through experimental evidence for the true binding mode of
the hit compound (S,R,S)-1a, as the ligand orientation was
found to differ from that of the initial docking model, which
was available at the start of the project. Crystallographic eluci-
dation of this binding mode helped to focus and drive the
drug design process more effectively and efficiently.
tion, which maintains low levels of Nrf2 protein. Upon oxida-
tive stress, Nrf2 dissociates from Keap1 and translocates into
the nucleus, where it forms heterodimers with a variety of
transcriptional regulatory proteins.[7] These protein complexes
then associate with the ARE to induce gene transcription of cy-
toprotective enzymes to prevent toxicity. A number of articles
have described the exciting preclinical neuroprotective effects
of Nrf2 not only in Parkinson’s[8–10] and Alzheimer’s disease,[11,12]
but also in multiple sclerosis[13] and amyotrophic lateral sclero-
sis.[14] Moreover, our research group recently reported the im-
portance of Nrf2 in epilepsy;[15] we demonstrated that Nrf2
mRNA is significantly increased in human epileptic hippocam-
pal tissue. Additionally, Nrf2 and the expression levels of its
downstream genes [heme oxygenase 1 (HO-1), NAD(P)H qui-
none oxidoreductase 1 (NQO1), glutathione S-transferases
(GSTs)] increased progressively in a spontaneously occurring
seizure model of epilepsy in mice. Importantly, mice that over-
express Nrf2 displayed fewer seizures with a profound de-
crease in microglia activation and preservation of hippocampal
neurons. Almost all known ARE activators behave as indirect
inhibitors of the Nrf2–Keap1 interaction; they are believed to
disrupt this protein–protein interaction by oxidation or cova-
lent modification of cysteine residues on Keap1, thus releasing
Nrf2.[16]
Neurodegenerative diseases are conditions with high unmet
medical needs for patients, and currently available treatments
only provide symptomatic relief. In recent years, nuclear factor
erythroid 2-related factor 2 (Nrf2) has become an attractive
neuroprotective target, as the Nrf2 pathway represents a natu-
ral cell defense mechanism.[1] Nrf2 is a transcription factor and
member of the Cap’N’collar family of proteins.[2] It regulates
the antioxidant response element (ARE)-mediated transcription
of a myriad of antioxidant and protective genes to counteract
the harmful effects of reactive oxygen species or xenobiotic
damage.[3–6] Under basal conditions, Nrf2 is tightly regulated in
the cytoplasm by Kelch-like ECH-associated protein (Keap1)
and is targeted for ubiquitin-mediated proteasomal degrada-
Very recently, Hu et al.[17,18] de-
scribed (S,R,S)-1a as the first cys-
teine-independent activator of
Nrf2 to act as a direct inhibitor of
[a] Dr. E. Jnoff, Dr. C. Genicot, Dr. E. Jigorel, Dr. Z. A. Sands, Dr. J.-P. Courade
UCB Pharma, UCB NewMedicines
Nrf2–Keap1 complex formation.
This single diastereomer, contain-
Chemin du Foriest, 1420 Braine-l’Alleud (Belgium)
E-mail: eric.jnoff@ucb.com
ing three chiral centers, originated
from a mixture of diastereomers of
[b] Dr. C. Albrecht, Dr. J. J. Barker, Dr. O. Barker, Dr. E. Beaumont,
Dr. S. Bromidge, Dr. F. Brookfield, Dr. M. Brooks, Dr. C. Bubert,
Dr. G. Dawson, Dr. S. Duclos, Dr. T. Fryatt, Dr. J. Kwong, R. Maghames,
I. Mushi, Dr. R. Pike, Dr. M. A. Smith, Dr. C. C. Stimson
Evotec, 114 Innovation Drive, Milton Park
1a, which was identified as a hit
from a high-throughput screen of the NIH MLPCN library of
small molecules. Prior to this publication, we had independent-
ly mined the NIH screening data, which was in the public
domain, and selected 1a as worthy of confirmation and decon-
volution. Subsequent preparation and biological evaluation of
each pure stereoisomer confirmed that the most active com-
pound is (S,R,S)-1a. Herein we describe how structural biology
informed our understanding of the binding pose, which in
turn helped to drive the drug design process.
Abingdon, Oxfordshire OX14 4RZ (UK)
[c] Dr. T. Ceska
UCB Pharma, UCB NewMedicines
208 Bath Road, Slough, Berkshire SL1 3WE (UK)
[d] V. Corden
CEM Microwave Technology Ltd.
2 Middle Slade, Buckingham, Buckinghamshire MK18 1WA (UK)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201300525.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 699

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
Compound (S,R,S)-1a was docked into two crystal structures
of the human Keap1 Kelch domain, PDB ID: 2FLU (co-crystal
structure with Nrf2 fragment)[19] and PDB ID: 1ZGK (apo struc-
ture),[20] as these structures showed differing conformations for
Arg380 and Arg415. Visual inspection of the docked poses and
comparison of the intermolecular protein–ligand interactions
revealed that the top-ranked pose in 2FLU mimics several in-
teractions observed between Keap1 and Nrf2. Interestingly, our
top pose posits the cyclohexyl group of (S,R,S)-1a in a similar
pocket to the exemplary docked pose published by Hu and
co-workers,[17] with the remainder of the molecule forming
quite different interaction patterns. In particular, our solution
allowed the acid to form both a salt bridge with Arg415 and
a hydrogen bond with Ser508, in a similar manner to one of
the Nrf2 acidic moieties. In addition to these Nrf2-like interac-
tions, the cyclohexane ring, which adopts a conformation with
the larger amide substituent positioned equatorially, makes
a hydrophobic contact with Tyr525 (not shown), and the car-
bonyl group of the amide linker between the tetrahydroisoqui-
noline (THIQ) moiety and the cyclohexane ring interacts with
Ser555. The THIQ group interacts with Tyr572 (p-stacking) and
the six-membered ring of the phthalimide moiety forms a hy-
drophobic contact with Tyr334. Finally, one of the phthalimide
carbonyl groups interacts by a hydrogen bond with Arg415
(Figure 1a).
While this pose appeared promising, the eventual co-crystal
structure revealed a very different mode of binding (PDB ID:
4L7B).[21] In the presence of ligand 1a, we found that the
Arg415 side chain adopts a conformation different from that
observed in the Nrf2-bound Keap1 structure (PDB ID: 2FLU)
used for the initial docking efforts pose. Surprisingly, the cyclo-
hexane ring observed in the Keap1–1a co-crystal structure
adopts an unusual conformation, with the larger amide sub-
stituent positioned axially. The top-ranked docked pose for the
1ZGK structure (data not shown) was significantly closer to the
crystallographically observed binding mode, but was initially
less favored, as the interactions with Keap1 appeared less Nrf2-
like.
Initial attempts at obtaining ligand complexes with (S,R,S)-
1a using the published mouse and human Keap1 Kelch
domain crystal structures were unsuccessful due to occlusion
of the Nrf2 binding site in the crystal lattice by loop residues.
A panel of rationally designed mutants were explored to
create a system with different crystal packing that would be
amenable to co-crystallisation or soaking with small molecules.
The human Keap1 Kelch domain (321–609) with the R354D
point mutation and the histidine tag relocated to the C termi-
nus of the construct subsequently co-crystallised with (S,R,S)-
1a (Figure 1b).
The conformation of (S,R,S)-1a in the crystal structure shows
the aromatic ring of the THIQ group further into the central
pore, and the cyclohexane carboxylic acid/phthalimide moiet-
ies extending outward, making interactions with Tyr572,
Ser602, Arg415, and Asn414. Compared with the apo structure
(PDB ID: 1ZGK) both Arg380 and Arg415 side chains are re-ori-
entated away from the central pore closer to the protein-
aceous aspect of the protein, in order to accommodate
Figure 1. a) Docking pose and b) co-crystal structure of (S,R,S)-1a with
Keap1.
(S,R,S)-1a. The carboxylic acid group fits into a groove formed
by Arg415, which extends into the center of the pore, and
Asn414 with hydrogen bonds to both residues. Arg415 is itself
positioned between the acid and the carbonyl group that links
the THIQ moiety to the cyclohexane ring. The phthalimide car-
bonyl group is hydrogen bonded to Ser602, and the phenyl
ring forms a face-to-face p-stack with Tyr572.
Finally, there also appears to be a water-mediated hydrogen
bond between the second phthalimide carbonyl oxygen atom
and Ser555. The binding pocket of Keap1 that accommodates
(S,R,S)-1a is the one used by the Nrf2 peptide and has recently
been involved with other small molecules co-crystallised with
Keap1.[22] As mentioned above, the THIQ moiety sits at the top
of the central pore, suggesting some potential chemical space
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 700

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
subsequent Friedel–Crafts acylation provided tetrahydroquino-
linone 6, which was reduced to afford 7. The cyclic carbamate
7 was cleaved to provide (S)-1-hydroxymethyl THIQ 8a; follow-
ing selective benzylation of the nitrogen atom, the phthali-
mide group was installed by Mitsunobu reaction under stan-
dard conditions. The N-benzyl group was cleaved by hydroge-
nolysis, and coupling with the optically pure (1R,2S)-cyclohex-
ane-1,2-dicarboxylic acid monobenzyl ester 13a yielded, after
hydrogenolysis of the benzyl group, compound (S,R,S)-1a. A
range of optically pure cyclic mono-protected diacid moieties,
13a–d, were prepared from 12a–d according to the published
procedure,[23] in which the corresponding cis-cyclic anhydrides
were desymmetrised by quinidine-mediated stereoselective
ring opening. Coupling reaction of 13b–d or 14 with THIQ 11
yielded the five-, four-, and three-membered ring analogues
1b–d and the des-acid compound 16, respectively. Conversion
of 1a into the corresponding primary amide 17 was carried
out under standard conditions (Scheme 1).
Figure 2. Extension from the THIQ into the central channel.
Tetrazole analogue 22 was prepared by coupling THIQ 11
with (1R,2S)-2-cyanocyclohexanecarboxylic acid 20 followed by
conversion of the cyanide moiety into the tetrazole group.
Compound 20 was prepared in three steps by starting from
13e, which was obtained in a similar manner to 13a but by
using quinine rather than quinidine as chiral catalyst in order
to obtain the opposite enantiomer (Scheme 2).
available around the aromatic ring for incorporating new inter-
actions to increase affinity toward Keap1 (Figure 2).
Optically pure THIQ 1a was prepared as shown in Scheme 1.
Treatment of (S)-phenylglycinol 2 with triphosgene gave the
cyclic carbamate 3, which was subsequently alkylated with
ethyl bromoacetate to give compound 4. Saponification and
Amino acid 32 was prepared
by coupling N-protected piperi-
dine 29 and THIQ 11 followed
by two deprotection steps. Syn-
thesis of the optically pure pi-
peridine 29 was performed ac-
cording to a published proce-
dure[24] by starting from (R)-phe-
nylglycinol,
and
conversion
toward piperidine 32 was per-
formed as depicted in Scheme 3.
Access to various substituted
THIQ structures was achieved by
a modification of the initial Mit-
sunobu reaction. Formation of
the cyclic sulfamates 33a,b was
carried out by reaction of the
corresponding 1-hydroxymethyl
THIQ 8a and 8b with thionyl
chloride, followed by oxidation
using ruthenium trichloride/
sodium periodate. The ring-
opening step was performed
with a wide range of nucleo-
philes, including phenyl magne-
sium bromide, pyridone, and the
anions of amides. Subsequent
coupling of the respective THIQs
34–38 with 13a afforded, after
hydrogenolysis, the fully elabo-
rated compounds 40, 43, 44, 46,
and 48 (Scheme 4).
Scheme 1. Reagents and conditions: a) triphosgene, CH2Cl2, TEA, 08C; b) ethyl bromoacetate, K2CO3, DMF; c) NaOH,
THF, H2O; d) thionyl chloride, CH2Cl2 then AlCl3, 08C; e) H2, Pd/C, EtOH; f) NaOH, EtOH, reflux; g) benzyl bromide,
DIPEA, THF; h) DIAD, PPh3, THF, phthalimide; i) H2, Pd/C, EtOH; j) quinidine, benzyl alcohol, toluene, À558C, 96 h;
k) T3P, 13a–d, CH2Cl2, TEA; l) cyclohexane carbonyl chloride, pyridine, CH2Cl2; m) H2, Pd/C, THF, EtOAc, EtOH;
n) (COCl)2, CH2Cl2, NH3(g)
.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 701

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
Disruption of the Nrf2–Keap1
protein–protein complex in-
duced by our small molecules
was measured by fluorescence
polarisation assay using Keap1
Kelch domain/Nrf2-ETGE peptide
and is expressed as IC50 values
(Table 1). Under these assay con-
ditions, the initial starting hit
1a displayed an IC50 value of
2.3 mm. The pendent phthali-
mide moiety linked to the cen-
tral THIQ core is well known to
be a potential problem with re-
spect to the developability of
molecules, so our initial task was
to carry out investigations to de-
termine whether this group
could be replaced. Interestingly,
isoxindole 43 retained a good
level of potency (IC50: 1.1 mm)
despite the absence of one hy-
drogen bond acceptor group.
The X-ray co-crystal structure
displayed interactions between
the carbonyl group of the isoxin-
dole and Ser602 in addition to
a putative p–p stacking of the
aromatic ring with Tyr572 (PDB
ID: 4N1B;[25] Figure 3a). We were
surprised to discover that succi-
nimide 52 retained potency
(IC50: 2 mm) and that removal of
Scheme 2. Reagents and conditions: a) (COCl)2, NH3, CH2Cl2; b) SOCl2, MeCN; c) BBr3, CH2Cl2, À788C; d) SOCl2 then
11, pyridine; e) Bu3SnN3, xylene, 1408C.
Scheme 3. Reagents and conditions: a) Pd/C, H2, MeOH; b) BH3·DMS, THF; c) Pd(OH)2/C, H2, MeOH; d) Boc2O,
NaHCO3, dioxane; e) BnBr, DBU, CH2Cl2; f) LiOH, THF; g) 11, HATU, DIPEA, DMF; h) H2, Pd/C, MeOH; i) HCl, Et2O.
a
further carbonyl group, as
shown with pyrrolidinone 40,
retained activity (IC50: 1.2 mm).
Starting from the THIQ intermediate 9, access to the succini-
mide 52 was achieved using the same synthetic route as for
compound 1a. Thus, coupling of succinimide with the hy-
droxy-THIQ under Mitsunobu conditions provided the fully
protected intermediate 49. Debenzylation, coupling with acid
13a, and final deprotection of the benzyl ester provided the
succinimide 52 as shown in Scheme 5.
These results indicate that either the p–p interaction does not
provide significant binding energy, or that the desolvation
penalty incurred by placing a lipophilic group in a solvent-ex-
posed area of the protein negates any positive binding energy
from the p-stacking interaction. On the other hand, complete
removal of the hydrogen bond acceptors whilst keeping the
While introduction of a methyl group at position 6 of the
THIQ moiety (in 44) was readily performed as outlined in
Scheme 4 by starting from (S)-2-amino-2-para-tolylethanol, an
alternative pathway was followed to prepare the isomer 59.
Acylation of 2-ortho-tolylethylamine 53 and subsequent Bis-
chler–Napieralski reaction allowed the synthesis of 3,4-dihy-
droisoquinoline 55. Introduction of the isoxindole moiety by
nucleophilic substitution, reduction of the imine group, and
HATU-mediated coupling of the resulting THIQ 57 with 13a
yielded compound 58 as a pair of diastereomers, which were
readily separated by column chromatography. Hydrogenolysis
of the benzyl group provided the optically pure 5-methyl-sub-
stituted derivative 59 (Scheme 6).
Table 1. Affinity of compounds for the Keap1 Kelch domain.
Compd
IC50 [mm][a]
Compd
IC50 [mm][a]
1a
1b
1c
1d
16
17
21
22
2.3Æ0.06
2.2Æ0.44
8.0Æ0.09
20.8Æ3.79
32
40
43
44
46
48
52
59
69.7Æ9.95[b]
1.2Æ0.14
1.1Æ0.19
7.1Æ0.51
>100
>100
>100[c]
2.0Æ0.38
0.75Æ0.32
>100
>100
7.4Æ0.89
[a] IC50 values determined by FP assay, and are the meansÆSEM (n=3–
5); active compounds inhibit with 80–100% control relative to the Nrf2
reference peptide. [b] Inhibition with 60% control. [c] 30% inhibition at
100 mm.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 702

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
putative interaction with Tyr572
significantly decreases activity, as
shown with compound 48,
which displayed only 30% inhib-
ition at 100 mm. The pyridyl
ether 46, which was expected to
place a hydrogen bond acceptor
in the correct region of space,
lost all measurable activity. It is
likely that the very specific direc-
tionality of the available lone
pair of the pyridine ring does
not allow the necessary hydro-
gen bond to form.
An investigation around the
saturated ring bearing the acid
function revealed that the cyclo-
pentyl analogue 1b is essentially
equipotent to the cyclohexyl de-
rivative 1a, while potency de-
creased progressively for the
four- and three-membered ring
compounds 1c and 1d. Intro-
duction of nitrogen atom into
the six-membered ring, leading
to piperidine analogues, was in-
vestigated. Among the various
isomers, only the a-amino acid
32 was found to be active, but
with a significant decrease in po-
tency (IC50: 69.7 mm). As the ac-
tivity of this compound class is
expected to occur in the brain,
modulation of the carboxylic
acid moiety was a key goal; it
was hoped that this would help
enable adequate levels of brain
exposure. However, neither re-
moval of the carboxylic acid
function (in 16) nor its replace-
ment by carboxamide (17) or ni-
trile (21) yielded active com-
pounds. Only replacement of the
acid function with a tetrazole
group (compound 22) was ac-
ceptable, leading to a small de-
crease in potency (IC50: 7.4 mm).
The crystal structure of this tet-
razole compound 22 bound to
Keap1 showed an analogous
pose to that of 1a. However, al-
though the tetrazole moiety
packs between Arg380 and
Scheme 4. Reagents and conditions: a) SOCl2, imidazole, TEA, CH2Cl2, 08C then RuCl3/NaIO4, EtOAc/H2O, 08C;
b) NaH, THF, isoxindole or pyrrolidone; c) 2-pyridone, Cs2CO3, THF, 08C; d) PhMgBr, THF, 08C; e) 13a, SOCl2 then
pyridine or HATU, DIPEA, DMF; f) H2, Pd/C, EtOH.
Scheme 5. Reagents and conditions: a) succinimide, DIAD, PPh3, THF; b) H2, Pd/C, EtOH; c) 13a, EDC, TEA, DMAP,
CH2Cl2; d) H2, Pd/C, EtOH
Arg415,
a different hydrogen
bonding pattern is observed
(PDB ID: 4L7C;[26] Figure 3b).
Scheme 6. Reagents and conditions: a) ClCH2COCl, K2CO3, CH2Cl2; b) P2O5, xylenes; c) isoxindole, Cs2CO3, DMF;
d) NaBH4, MeOH; e) 13a, HATU, DIPEA, DMF; f) H2, Pd/C, THF/H2O.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 703

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
methyl group likely acts as a lipophilic “plug”, giving a moder-
ate boost in potency over the original hit 1a. Further modifica-
tions that target that area of the molecule are currently under
investigation.
In addition to an exploration of structure–activity relation-
ships, assessment of in vivo brain exposure was performed.
Acid compound 1a was restricted to the peripheral compart-
ment in mice (unbound brain-to-plasma (Bu/Pu) ratio <0.01),
being a P-glycoprotein (P-gp) substrate, as predicted by a high
efflux ratio (ER value of 20) observed in an MDCK-MDR1 assay.
The efflux was mitigated by using Mdr1a/1b/Bcrp knock-out
mice, in which 1a did achieve an improved level of free brain
exposure (0.18 mm) and a much better Bu/Pu ratio (0.4), con-
firming the P-gp substrate property of 1a. This Bu/Pu ratio is
very similar to the biologically inactive non-acid compound 16
administered in wild-type mice, but with lower free brain expo-
sure. Neither replacement of the carboxylic acid group by the
tetrazole bioisostere 22 nor the zwitterionic amino acid 32 dis-
played improved profiles over that of 1a (Table 2). Despite the
Table 2. Physicochemical and DMPK properties.
[d]
Compd
LogD[a]
PSA [ꢁ2][b]
95
ER[c]
20
Bu/Pu
Cu [mm][e]
1a
1.36
<0.01
0.4[f]
0.3
NT[g]
<0.01
NT[g]
0.9
<0.01
0.18[f]
0.003
NT[g]
<0.01
NT[g]
16
22
32
43
48
3.59
1.15
0.69
1.11
2.06
58
112
107
78
0.7
28
NT[g]
7.5
58
0.7
0.2
[a] Measured at pH 7.4. [b] Polar surface area. [c] Efflux ratio in MDCK-
MDR1 cells (10 mm incubated up to 120 min). [d] Unbound brain-to-
plasma ratio measured in mice. [e] Unbound brain concentration mea-
sured in mice at Cmax. [f] Measured in Mdr1a/1b/Bcrp knock-out mice.
[g] Not tested.
P-gp liability related to the carboxylic acid function, screening
of additional compounds that exhibit lower polar surface area
(PSA) values than that of 1a was performed. In this area, whilst
the isoxindole compound 43 exhibited a trend toward de-
creased efflux, phenyl compound 48 was completely devoid of
efflux in MDCK-MDR1 cells (ER value of 0.7) with high passive
permeability (Papp value of 375 nmsÀ1) and Bu/Pu ratio close to
1 (Table 2). This result demonstrates that in this chemical
series, mitigation of efflux and improvement of in vivo brain
exposure is achievable in the presence of the acidic function
by decreasing the PSA or decreasing the number of hydrogen
bond acceptor groups.
Figure 3. Co-crystal structures of Keap1 with compounds a) 43 and b) 22.
As mentioned above, the crystal structure places the phenyl
ring of the THIQ on top of the central pore of the Kelch
domain and appears to show potential for extending groups
into the pore, thus opening up two possibilities for further ex-
ploration: The first was to nestle a lipophilic group into the
opening, taking advantage of van der Waals interactions and
a good shape fit to increase potency. The second possibility
was to extend a portion of the molecule into the pore, thereby
affording us an opportunity to create hydrogen bond interac-
tions with a number of backbone carbonyl groups, or to dis-
place or interact with water molecules observed inside the
pore. Methylation of the 6-position of the THIQ leading to 44
was tolerated, but not optimal relative to a methyl group at
the 5-position, providing ligand 59 (PDB ID: 4L7D),[27] which
showed increased potency (IC50: 0.75 mm). This result suggests
a good shape fit of the molecule toward the pore where the
In conclusion, solving the structure of the (S,R,S)-1a–Keap1
complex was a breakthrough that provided experimental evi-
dence for the true binding mode of the hit compound (S,R,S)-
1a. The ligand orientation was found to differ from that of the
initial docking model, which was available at the start of the
project. The X-ray crystal structure helped to focus and drive
the drug design process more effectively and efficiently. The
design and synthesis of a further round of compounds is
needed for delivering active CNS-penetrant compounds.
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 704

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
[14] M. R. Vargas, D. A. Johnson, D. W. Sirkis, A. Messing, J. A. Johnson, J.
Neurosci. 2008, 28, 13574–13581.
Acknowledgements
[15] M. Mazzuferi, G. Kumar, J. van Eyll, B. Danis, P. Foerch, R. M. Kaminski,
Ann. Neurol. 2013, 74, 560–568.
[16] A. J. Wilson, J. K. Kerns, J. F. Callahan, C. J. Moody, J. Med. Chem. 2013,
56, 7463–7476.
[17] L. Hu, S. Magesh, L. Chen, L. Wang, T. A. Lewis, Y. Chen, C. Khodier, D. In-
oyama, L. J. Beamer, T. J. Emge, J. Shen, J. E. Kerrigan, A. N. Kong, S.
Dandapani, M. Palmer, S. L. Schreiber, B. Munoz, Bioorg. Med. Chem. Lett.
2013, 23, 3039–3043.
[18] L. Hu, S. Magesh, L. Chen, T. Lewis, B. Munoz, L. Wang, (Rutgers Univer-
sity and The Broad Institute Inc.), Int. PCT Pub. No. WO2013067036 A1,
2013.
The authors thank Dr. T. Wayne (Evotec), Dr. C. Wilson (Evotec),
C. Rospo (UCB) and her team for pharmacokinetic analysis, and
R. Ward (Evotec) for chiral chromatographic purifications.
Keywords: CNS · co-crystal structures · docking · Keap1 ·
Nrf2 · structure–activity relationships
[1] S. Magesh, Y. Chen, L. Hu, Med. Res. Rev. 2012, 32, 687–726.
[2] J. Alam, D. Stewart, C. Touchard, S. Boinapally, A. M. K. Choi, J. L. Cook, J.
Biol. Chem. 1999, 274, 26071–26078.
[3] J. W. Kaspar, S. K. Niture, A. K. Jaiswal, Free Radical Biol. Med. 2009, 47,
1304–1309.
[19] S. C. Lo, X. Li, M. T. Henzl, L. J. Beamer, M. Hannink, EMBO J. 2006, 25,
3605–3617.
[20] L. J. Beamer, X. Li, C. A. Bottoms, M. Hannink, Acta Crystallogr. Sect. D
2005, 61, 1335–1342.
[21] PDB ID: 4L7B: E. Jnoff, F. Brookfield, C. Albrecht, J. J. Barker, O. Barker, E.
Beaumont, S. Bromidge, M. Brooks, T. Ceska, J-P. Courade, T. Crabbe, S.
Duclos, T. Fryatt, E. Jigorel, J. Kwong, Z. A. Sands, M. A. Smith.
[22] D. Marcotte, W. Zeng, J.-C. Hus, A. McKenzie, C. Hession, P. Jin, C. Ber-
geron, A. Lugovskoy, I. Enyedy, H. Cuervo, D. Wang, C. Atmanene, D.
Roecklin, M. Vecchi, V. Vivat, J. Kraemer, D. Winkler, V. Hong, J. Chao, M.
Lukashev, L. Silvian, Bioorg. Med. Chem. Lett. 2013, 21, 4011–4019.
[23] C. Bolm, I. Atodiresei, I. Schiffers, Org. Synth. 2005, 82, 120–125.
[24] M. Calosso, M. Wagner, T. Gendrineau, M. Petit, C. Kadouri-Puchot, L. De-
choux, Lett. Org. Chem. 2007, 4, 4–6.
[25] PDB ID: 4N1B: M. A. Smith, S. Duclos, E. Beaumont, J. Kwong, M.
Brooks, J. J. Barker, E. Jnoff, F. Brookfield, J-P. Courade, O. Barker, T.
Fryatt, C. Albrecht, S. Bromidge.
[26] PDB ID: 4L7C: E. Jnoff, F. Brookfield, C. Albrecht, J. J. Barker, O. Barker, E.
Beaumont, S. Bromidge, M. Brooks, T. Ceska, J-P. Courade, T. Crabbe, S.
Duclos, T. Fryatt, E. Jigorel, J. Kwong, Z. A. Sands, M. A. Smith.
[27] PDB ID: 4L7D: E. Jnoff, F. Brookfield, C. Albrecht, J. J. Barker, O. Barker, E.
Beaumont, S. Bromidge, M. Brooks, T. Ceska, J-P. Courade, T. Crabbe, S.
Duclos, T. Fryatt, E. Jigorel, J. Kwong, Z. A. Sands, M. A. Smith.
[4] J. M. Lee, J. A. Johnson, J. Biochem. Mol. Biol. 2004, 37, 139–143.
[5] T. Nguyen, P. J. Sherratt, C. B. Pickett, Annu. Rev. Pharmacol. Toxicol.
2003, 43, 233–260.
[6] B. M. Hybertson, B. Gao, S. K. Bose, J. M. McCord, Mol. Aspects Med.
2011, 32, 234–246.
[7] K. Itoh, N. Wakabayashi, Y. Katoh, T. Ishii, K. Igarashi, J. D. Engel, M. Ya-
mamoto, Genes Dev. 1999, 13, 76–86.
[8] A. I. Rojo, N. G. Innamorato, A. M. Martin-Moreno, M. L. De Ceballos, M.
Yamamoto, A. Cuadrado, Glia 2010, 58, 588–598.
[9] N. G. Innamorato, A. Jazwa, A. I. Rojo, C. Garcia, J. Fernandez-Ruiz, A.
Grochot-Przeczek, A. Stachurska, A. Jozkowicz, J. Dulak, A. Cuadrado,
PLoS One 2010, 5, e11838.
[10] P. C. Chen, M. R. Vargas, A. K. Pani, R. J. Smeyne, D. A. Johnson, Y. W.
Kan, J. A. Johnson, Proc. Natl. Acad. Sci. USA 2009, 106, 2933–2938.
[11] C. P. Ramsey, C. A. Glass, M. B. Montgomery, K. A. Lindl, G. P. Ritson, L. A.
Chia, R. L. Hamilton, C. T. Chu, K. L. Jordan-Sciutto, J. Neuropathol. Exp.
Neurol. 2007, 66, 75–85.
[12] K. Kanninen, R. Heikkinen, T. Malm, T. Rolova, S. Kuhmonen, H. Leino-
nen, S. Ylꢂ-Herttuala, H. Tanila, A. L. Levonen, M. Koistinaho, J. Koistina-
ho, Proc. Natl. Acad. Sci. USA 2009, 106, 16505–16510.
[13] D. A. Johnson, S. Amirahmadi, C. Ward, Z. Fabry, J. A. Johnson, Toxicol.
Sci. 2010, 114, 237–246.
Received: December 13, 2013
Published online on February 6, 2014
ꢀ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 699 – 705 705