Fragment-based discovery and optimization of BACE1 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2010.06.089

A novel series of 2-aminobenzimidazole inhibitors of BACE1 has been discovered using fragment-based drug discovery (FBDD) techniques. The rapid optimization of these inhibitors using structure-guided medicinal chemistry is discussed.

Full text of DOI:10.1016/j.bmcl.2010.06.089

Bioorganic & Medicinal Chemistry Letters 20 (2010) 5329–5333
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Fragment-based discovery and optimization of BACE1 inhibitors
a
b
b
a
James Madden a, , Jenny R. Dod , Robert Godemann , Joachim Kraemer , Myron Smith ,
*
Marion Biniszkiewicz b, David J. Hallett a, John Barker a, Jane D. Dyekjaer a, Thomas Hesterkamp b
a Evotec UK Ltd, 114 Milton Park, Abingdon OX14 4SA, UK
b Evotec AG, Schnackenburgallee 114, 22525 Hamburg, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
A novel series of 2-aminobenzimidazole inhibitors of BACE1 has been discovered using fragment-based
drug discovery (FBDD) techniques. The rapid optimization of these inhibitors using structure-guided
medicinal chemistry is discussed.
Received 10 May 2010
Revised 14 June 2010
Accepted 14 June 2010
Available online 27 June 2010
Ó 2010 Elsevier Ltd. All rights reserved.
Keywords:
BACE1
Alzheimer’s disease
FBDD
SPR
Crystallography
2-Aminobenzimidazole
Alzheimer’s disease (AD) presents one of the greatest unmet
medical needs facing aging populations worldwide. Beta secretase
(BACE1) catalyzes the formation of b-amyloid (Ab), which can form
amyloid plaques, one of the pathological hallmarks of AD. The dis-
covery of orally bioavailable, brain-penetrant, small molecule
inhibitors of BACE1 has been the goal of a number of drug discov-
ery companies. The widely acknowledged failure of high-through-
put screening (HTS) approaches to identify brain-penetrant BACE1
inhibitors1 has given extra impetus to alternative hit discovery
strategies such as fragment-based drug discovery (FBDD) screening
techniques such as NMR, crystallography, and surface plasmon res-
onance (SPR).2
Herein, we describe the use of a high-throughput BACE1 assay
in combination with orthogonal hit confirmation using SPR and
structural elucidation by X-ray crystallography to identify a num-
ber of fragment inhibitors of BACE1. We go on to describe the ini-
tial optimization and profiling of one such example.
Although SPR did not recapitulate the affinity seen in the func-
tional assay, all of the compounds did show some degree of binding
to BACE1 in the SPR experiment. Compounds were further ranked
for submission to crystallography according to their aqueous solu-
bility (data not shown). Top ranked compounds were submitted
either for co-crystallization or soaking with BACE1.2 Fragment 3
was successfully co-crystallized with BACE1 and the structure
solved to 1.8 Å resolution.5 Figure 2 shows 3 bound in the active
site of BACE1.
HO
O
N
N
Cl
N
H2N
H2N
N
H
N
N
Cl
N
Evotec’s fragment library of 20,000 compounds,3 with an aver-
age molecular weight of 250, was screened at 1 mM against
BACE1.4 The resulting hits were clustered by chemotype and
ranked according to their novelty with reference to known BACE1
inhibitors and their ligand efficiency. Exemplars from each cluster
were submitted for confirmation of binding to BACE1 using SPR.2
Figure 1 shows example structures of fragment hits and Table 1
their assay IC50 compared to their BACE1 inhibition in SPR.
1
2
3
N
N
O
O
N
N
OH
OH
N
N
Cl
N
4
5
6
* Corresponding author. Tel.: +44 (0)1235441368.
Figure 1. Example fragment inhibitors identified from high-throughput functional
E-mail address: james.madden@evotec.com (J. Madden).
assay.
0960-894X/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2010.06.089

5330
J. Madden et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5329–5333
Table 1
Comparison of functional assay with SPR
Compd id
Functional assay
Ligand
SPRc % inhibition
@ 2 mM
efficiencyb
a
IC50
(lM)
1
2
3
4
5
6
928
731
770
862
1008
776
0.25
0.40
0.29
0.27
0.28
0.34
36
34
63
29
20
27
a
b
c
n P 2.
LE = ꢀRT ln(IC50)/non-H atom count.
p <0.05.
Figure 4. Fragment 3 (green) overlaid with J&J inhibitor 7 (blue) in the active site of
BACE1.
sites of the BACE1 protein. The cyclohexyl occupies a largely
hydrophobic pocket in the prime side of the site while the pro-
panamide branches towards the S3 pocket. In silico docking of
analogs of 3 suggested that the hydroxypropyl moiety could be
modified or replaced to target the non-prime side of the active
site and improve affinity. Indeed, numerous groups have success-
fully employed a strategy to target the S3 pocket and its con-
Figure 2. Fragment 3 bound in the active site of BACE1.
served water molecules as
a
means of improving potency.8
However, before progressing fragment 3, the available published
data from J&J showed that compound 7 suffered from a hERG
liability (hBACE1 Ki 8 nM, hERG Ki 84 nM).9 Given the similarity
of fragment 3 with the J&J lead, our rationale was to improve the
affinity of compound 3 for BACE1 to a point where selectivity
over hERG could be measured and used as a decision point for
progressing the series. Our calculated pKa for fragment 3 of 7.7
was thought to be sufficiently different from that reported by
J&J (pKa 10.6)7 to permit a possible change in recognition of
our molecules in the hERG binding site.
Analogs of 3 were rapidly accessed using commercially avail-
able building blocks (BB) as shown in Scheme 1. SnAr of methyl
4-aminobutanoate with 3-chloro-6-fluoronitrobenzene, reduction
of the nitro group and aminoimidazole formation using cyanogen
bromide all proceeded cleanly. Saponification and concentration
afforded the final step precursor as the 2-aminobenzimidazole lith-
ium carboxylate salt. Efforts to couple this salt with amines re-
sulted in the intramolecular cyclization to the lactam by-product.
It was found that acidification of the crude saponification concen-
trate with excess HCl in dioxane and subsequent amidation using
the HCl salt of the 2-aminobenzimidazole acid resulted in minimal
intramolecular cyclization.
The endocyclic N in the 3-position forms a hydrogen bond with
Asp32. The exocyclic N forms a hydrogen bond with both catalytic
aspartates and the hydroxyl hydrogen bonds to Asp228. The aryl
unit forms a face-to-edge
p-interaction with Tyr71 and the 5-
chloro substituent sits in a pocket formed between Tyr71 and
Trp76. The BACE1 protein adopts the flap open conformation typ-
ical of many of the published small molecule-BACE1 structures.6
A comparison of the structure of fragment 3 with known BACE1
inhibitors highlighted the similarity to Johnson & Johnson’s BACE1
inhibitors.7 The published crystal structure of one of their com-
pounds,
(4S)-4-(2-amino-7-phenoxyquinazolin-3(4H)-yl)-N,4-
dicyclohexyl-N-methylbutanamide (7, Fig. 3), was overlaid with
that of 3 and is shown in Figure 4.
The J&J compound contains an aminoquinazoline core that
binds to both catalytic aspartates in a similar manner to 3. In
lieu of the 5-chloro substituent of 3 the J&J inhibitor features a
5-phenoxy substituent that forces Tyr71 up out of the active site
and allows the phenoxy to pick up a face-to-edge interaction
with Phe108. A branched chain at the 3-position of the quinaz-
oline accesses further interactions with the prime and non-prime
Using this chemistry a range of 12 analogs was prepared. Their
activities are given in Table 2.
Demonstrating a 29-fold increase in potency over fragment 3
and a small decrease in LE, compound 12b was selected for crystal-
lography and its structure was solved to 2.6 Å.3 Figure 5 shows 12b
overlaid with J&J compound 7.
The central 2-aminobenzimidazole core of 12b retains the same
position as 3 and the butanamide branches towards the non-prime
N
O
N
O
H2N
N
Figure 3. Johnson & Johnson’s BACE1 inhibitor.

J. Madden et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5329–5333
5331
F
O
O
O
Cl
NO2
9
OH
a,b
c,d
NH
N
N
+
O
NH2
11
Cl
NH2
H2N
Cl
O
10
8
e
O
R2
R1
N
N
N
NH2
Cl
12
Scheme 1. Synthesis of analogs of 3. Reagents and conditions: (a) Et3N, dioxane, 80 °C, 85%; (b) Fe powder, NH4Cl (aq), 60 °C; (c) BrCN, EtOH, 80 °C; (d) (i) LiOHꢁH2O, H2O/THF,
(ii) 4 M HCl in dioxane, 23% (three steps); (e) R1R2NH, EDCIꢁHCl, HOBtꢁH2O, DMF.
Table 2
BACE1 IC50 for analogs of 3
hydrophobic pocket close to Ile118 and Phe108. The amide does
Compd id R1
R2
Functional assay IC50
(l
M) (LE)b
a
not appear to form any H-bonding interactions with the protein.
The SAR summarized in Table 2 highlights a preference of the pro-
tein for medium sized hydrophobic amides. It was rationalized that
12a
12b
12c
12d
12e
12f
12g
12h
12i
Benzyl
Cyclohexyl
Isopropyl
2-Morpholin-4-ylethyl
2-Pyrrolidin-4-ylethyl
Pyridine-2-ylmethyl
Cyclohexylmethyl
1-Methylpiperidin-4-yl
3-Fluorobenzyl
H
Me
H
H
H
H
H
H
H
1403 (0.17)
26 (0.27)
193 (0.26)
827 (0.17)
444 (0.20)
426 (0.20)
83 (0.24)
325 (0.21)
249 (0.20)
2854 (0.17)
194 (0.20)
59 (0.24)
an additional
a-branching group targeting the prime side of the
protein in combination with a range of polar or non-polar amides
might yield additional gains in potency by interacting with resi-
dues or water molecules in or near the S3 pocket. In order to
quickly access compounds of this type, it was envisaged that com-
mercially available b-amino acids could be employed in chemistry
analogous to that described in Scheme 1. Although these building
blocks would afford compounds with 1 carbon fewer in the
branched chain, it was rationalized that the desired binding inter-
actions could still be achieved through the use of extended amine
inputs. Scheme 2 describes the route that was used to synthesize
these analogs. With this route the use of the HCl salt of the 2-ami-
nobenzimidazole acid, as with Scheme 1, was not effective in
reducing the formation of cyclic lactam by-product. Therefore the
route was redesigned such that b-amino acids were used in the ini-
tial SnAr and the amide diversity introduced at the 2nd step of the
synthesis.
12j
12k
12l
Cyclopropylmethyl
3-Fluorobenzyl
Cyclohexylmethyl
H
Me
Me
a
n P 2.
b
LE = ꢀRT ln(IC50)/non-H atom count.
Analysis of the crystal structure of 12b suggested that the pre-
ferred stereochemistry around the a
-branched carbon would be S10
and indeed this proved to be the case. However, initially two pairs
of analogs were synthesized using both enantiomers of 3-amino-
pentanoic acid and 3-amino-5-methylhexanoic acid. Once the pre-
ferred stereochemistry had been established, the S enantiomers of
both acids were utilized for further analog synthesis. The results
are shown in Table 3.
As with the non-branched analogs, the addition of polarity in
the non-prime site was not tolerated by BACE1 and small or med-
ium sized hydrophobic substituents appeared to be preferred. The
cyclohexylmethyl and adamantyl groups (14i and 14m) afforded
modest improvements in potency and LE over 12b. Attempts to
introduce a H-bond acceptor into 14i as in 14k resulted in a drastic
loss in potency. Compounds 14i and 14m were selected for crystal-
0
0
lography and their structures solved to 2.4 ÅA and 2.6 ÅA,
respectively.3
As shown in Figure 6 the binding mode of both compounds is
highly conserved. As expected, the a-branched ethyl occupies the
Figure 5. Compound 12b (blue) overlaid with J&J inhibitor 7 (green) in the active
prime side of the protein with both amide substituents posi-
tioned in the hydrophobic pocket close to Ile118 and Phe108.
Although the amide N–H’s are both within hydrogen bonding
distance of Gly230 at the edge of the S3 pocket, the torsional an-
gle forced on the amide by the location of the cyclohexylmethyl
site of BACE1.
side of the protein. Instead of picking up any interactions with the
protein, 12b is seen to fold in on itself; the cyclohexyl sits in a

5332
J. Madden et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5329–5333
R2
R2
N
R1
R1
N
R3
R3
F
a,b
c,d
NH
O
O
N
N
NH2
Cl
NO2
Cl
NO2
Cl
8
13
14
Scheme 2. Synthesis of analogs of 12b. Reagents and conditions: (a) b-amino acid, Et3N, dioxane, 80 °C; (b) R2R3NH, EDCIꢁHCl, HOBtꢁH2O, DMF; (c) SnCl2ꢁ2H2O, DMF, EtOH;
(d) BrCN, EtOH, 80 °C.
Table 3
BACE1 IC50 for analogs of 12b
a
Compd id
R1 (stereochem.)
R2
R3
Functional assay IC50
(
l
M) (LE)b
14a
14b
14c
14d
14e
14f
14g
14h
14i
Et (R)
Et (S)
iBu (R)
iBu (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
Et (S)
iBu (S)
iBu (S)
H
H
H
H
H
H
H
H
H
Me
H
H
H
H
H
H
H
H
H
H
H
Benzyl
Benzyl
Isopropyl
Isopropyl
516 (0.19)
72 (0.24)
>1000
246 (0.22)
695 (0.21)
440 (0.18)
243 (0.23)
270 (0.20)
7.0 (0.29)
375 (0.19)
>1000
Isopropyl
2-Morpholin-4-ylethyl
Cyclopropylmethyl
2-Pyrrolidin-4-ylethyl
Cyclohexylmethyl
Cyclohexylmethyl
Tetrahydropyran-4-ylmethyl
Pyridine-2-ylmethyl
2-Adamantyl
14j
14k
14l
>1000
14m
14n
14o
14p
14q
14r
14s
14t
14u
8.9 (0.26)
95 (0.24)
340 (0.19)
264 (0.20)
168 (0.18)
75 (0.20)
87 (0.21)
235 (0.19)
54 (0.22)
3,3-Dimethylbutyl
3-Fluorobenzyl
4-Fluorobenzyl
O-tbutyl-
O-tbutyl-
O-tbutyl-
L
-serineOMe
-serineOMe
-serineOH
D
L
Pyridine-2-ylmethyl
3-Fluorobenzyl
a
n P 2.
b
LE = ꢀRT ln(IC50)/non-H atom count.
compound is equipotent against hBACE2 and hERG and has slight
selectivity over Cathepsin D (see Table 4).
Despite a 100-fold improvement in the affinity of the series for
BACE1 over two iterations of synthesis, the activity of 14i and other
analogs (data not shown) at hERG led us to deprioritize this series
in favor of better fragment starting points and these will be the
subject of a future publication.
Table 4
Profile of 14i
H
N
O
N
N
NH2
14i
Cl
Profile
TPSA
Value
0
73 ÅA2
MW
pKa
363
7.7
Figure 6. Compound 14i (magenta) and 14m (green) in the active site of BACE1.
hBACE1 IC50
Ab secretion IC50
7.0
12
0
Negative
42 lM
8.0
7.5
5.5/6.7 min
12%
lM
lM
% Cytotox @ 50
Ames S9
Cathepsin D
hBACE2 IC50
hERG IC50
lM
or adamantyl in the hydrophobic pocket appears to preclude this
interaction.
Compound 14i was profiled and showed comparable activity in
the BACE1 cellular Ab secretion assay, was clean in Ames test with
and without S9 metabolic activation and was not cytotoxic. The
lM
lM
Human/rat liver microsomes t1/2
PAMPAmaxflux

J. Madden et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5329–5333
5333
crystals, for examples, 12b, 14i, and 14m were obtained at Evotec UK. For
details of the human BACE construct employed and crystallization conditions
see Ref. 4.
References and notes
1. Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. J. Med. Chem. 2008, 51,
3661.
2. For SPR see www.biacore.com. For a review of crystallography see Williams, S.
P.; Kuyper, L. F.; Pearce, K. H. Curr. Opin. Chem. Biol. 2005, 9, 371; For NMR see
Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science 1996, 274, 1531.
3. Brewer, M.; Ichihara, O.; Kirchoff, C.; Schade, M.; Whittaker, M. Assembling A
Fragment Library. Fragment-based Drug Discovery;
Wiley: Weinheim, 2008. p 39.
4. Godemann, R.; Madden, J.; Krämer, J.; Smith, M.; Fritz, U.; Hesterkamp, T.;
Barker, J.; Höppner, S.; Hallett, D.; Cesura, A.; Ebneth, A.; Kemp, J. Biochemistry
2009, 48, 10743.
6. PDB Examples: 2OHQ (Astex), 3BRA (Roche), 2Q15 (Johnson & Johnson), and
2QMD (Schering-Plough).
7. Baxter, E. W.; Conway, K. A.; Kennis, L.; Bischoff, F.; Mercken, M. H.; De Winter,
H. L.; Reynolds, C. H.; Tounge, B. A.; Luo, C.; Scott, M. K.; Huang, Y.; Braeken, M.;
Pieters, S. M. A.; Berthelot, D. J. C.; Masure, S.; Bruinzeel, W. D.; Jordan, A. D.;
Parker, M. H.; Boyd, R. E.; Qy, J.; Alexander, R. S.; Brenneman, D. E.; Reitz, A. B. J.
Med. Chem. 2007, 50, 4261. PDB: 2Q15.
8. Eder, J.; Hommel, U.; Cumin, F.; Martoglio, B.; Gerhartz, B. Curr. Pharm. Des.
2007, 13, 271; Ghosh, A. K.; Kumaragurubaran, N.; Hong, L.; Koelsh, G.; Tang, J.
Curr. Alzheimer Res. 2008, 5, 121.
A Practical Approach;
9. Ellen Baxter. Gordon Research Conference, August 2008.
10. Note change of priority when assessing stereochemistry of Evotec compounds
versus J&J compound.
5. Structure deposited with PDB. PDB ID Codes: 3MSJ (Fragment 3), 3MSK (Ex.
12b), 3MSL (Ex. 14i), 3MSM (Ex. 14m). The co-crystal for fragment 3 was
obtained by Proteros Biostructures GmbH, Martinsried, Germany. The co-