Structure-based design and synthesis of novel P2/P3 modified, non-peptidic β-secretase (BACE-1) inhibitors

Article abstract of DOI:10.1016/j.bmcl.2010.01.139

Starting from peptidomimetic BACE-1 inhibitors, the P2 amino acid including the P2/P3 peptide bond was replaced by a rigid 3-aminomethyl cyclohexane carboxylic acid. Co-crystallization revealed an unexpected binding mode with the P3/P4 amide bond placed into the S3 pocket resulting in a new hydrogen bond interaction pattern. Further optimization based on this structure resulted in highly potent BACE-1 inhibitors with selectivity over BACE-2 and cathepsin D.

Full text of DOI:10.1016/j.bmcl.2010.01.139

Bioorganic & Medicinal Chemistry Letters 20 (2010) 1924–1927
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Structure-based design and synthesis of novel P2/P3 modified, non-peptidic
b-secretase (BACE-1) inhibitors
a
b
b
b
Stephen Hanessian a, , Zhihui Shao , Claudia Betschart , Jean-Michel Rondeau , Ulf Neumann ,
*
Marina Tintelnot-Blomley b
a Department of Chemistry, Université de Montréal, C. P. 6128, Succ. Centre-Ville, Montréal, P.Q., Canada H3C 3J7
b Novartis Institutes for BioMedical Research, Novartis Pharma AG, PO Box, CH-4002 Basel, Switzerland
a r t i c l e i n f o
a b s t r a c t
Article history:
Starting from peptidomimetic BACE-1 inhibitors, the P2 amino acid including the P2/P3 peptide bond was
replaced by a rigid 3-aminomethyl cyclohexane carboxylic acid. Co-crystallization revealed an unex-
pected binding mode with the P3/P4 amide bond placed into the S3 pocket resulting in a new hydrogen
bond interaction pattern. Further optimization based on this structure resulted in highly potent BACE-1
inhibitors with selectivity over BACE-2 and cathepsin D.
Received 10 December 2009
Revised 22 January 2010
Accepted 27 January 2010
Available online 2 February 2010
Ó 2010 Elsevier Ltd. All rights reserved.
Keywords:
Protease inhibitor
Peptidomimetic
Alzheimer’s disease
Genetic links point to an excess level of cerebral amyloid b pep-
tide (Ab) as a major culprit in the pathogenesis of Alzheimer’s dis-
ease (AD). Ab is generated by the sequential proteolytic action of
The removal of the P3/P4 amide segment followed by macro-
cyclization (Scheme 1) leading to compounds active in vivo has
been described recently.5 As known from the first X-ray structure
of OM99-2, the heptapeptide-derived inhibitor with BACE-1,6 and
from previous hydroxyethylene transition-state inhibitors,5a the
P2/P3 amide nitrogen is not involved in a direct hydrogen bond.
It forms a water-mediated interaction and can be methylated with-
out substantial reduction in activity in the enzyme assay. This
amide bond is therefore a target for the introduction of non-pep-
tidic features.
We report herein on a conceptually novel approach to the de-
sign of P2/P3 modified inhibitors of BACE-1, which also allows
for a wide variation in the P3 site. Since the simple exchange of
the amide bond by an alkyl chain was expected to introduce too
much flexibility, some rigidification was considered necessary. To
this end we envisaged the incorporation of a 3-aminomethyl cyclo-
hexane carboxylic acid motif to replace the P2/P3 amide (Scheme
1). Carbocyclic and heterocyclic motifs have been previously incor-
porated in peptidomimetics as constrained P10 BACE-1 inhibitors.7
In the present series, the conformationally pre-organized cyclo-
hexane scaffold was intended to serve as a rigid linker between P1
and P3, which conserves the position of both the P3 side chain as
well as the P3/P4 amide (Fig. 1). As illustrated in Table 1, the opti-
mal filling of the S3 pocket in the original compound series en-
hances potency by two orders of magnitude going from methyl
to iso-butyl (compounds 1–3).
two transmembrane proteins, b- and
c-secretase, which excise
mainly the 40 and 42 amino acid Ab fragments from the membrane
bound b-amyloid precursor protein (APP). A misbalance between
Ab production and its clearance from the brain, and subsequent
degradation leads to the aggregation of Ab into oligomers and pla-
ques.1 Especially oligomeric forms, starting from the size of dimers,
are believed to be the neurotoxic entities leading to neurodegener-
ation, synaptic loss, impaired neurotransmission, and finally cogni-
tive decline.2 Being responsible for the first endoproteolytic
cleavage step in the generation of Ab, the aspartyl protease b-
secretase (BACE-1) is an attractive target when aiming at the
reduction of Ab. There are a number of recent reviews in the liter-
ature on recent progress in the development of BACE-1 inhibitors.3
However, despite intensive efforts, it has been very challenging to
find compounds which combine high potency and selectivity with
good brain penetration. Therefore, few compounds have entered
early-stage clinical trials so far.4
In previous papers, we have described some of our efforts in
reducing the peptidic character of peptide-derived BACE-1 inhibi-
tors. In particular, we aimed at reducing P-glycoprotein-mediated
efflux, which limits brain exposure and is often observed with pep-
tide-like compounds.
The lipophilic carbocycle is placed in S2, a position that is filled
with a polar amino acid of the natural substrate. However, in line
* Corresponding author. Tel.: +514 343 6738; fax: +514 343 5728.
E-mail address: stephen.hanessian@umontreal.ca (S. Hanessian).
0960-894X/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2010.01.139

S. Hanessian et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1924–1927
1925
Table 1
P2/P3 SAR for some P3 to P20 spanning hydroxyethylene compounds
O
O
O
O
H
N
H
N
OH
H2N
N
H
N
H
R2
N
H
NH2
N
H
O
O
O
H
N
O
OH
O
O
N
H
N
H
N
H
O
OH
R2
R1
OM 99-2
P1'
O
HO
O
P4
HO
O
P3
P2
P1
P2'
P3'
P4'
Compound
R1
BACE-1
CathD
reduce size
a
IC50
,
lM
1
2
3
4
MeS–C2H4–
MeS–C2H4–
MeS–C2H4–
Me
Me
iPr
iBu
iBu
4.0
0.37
0.053
0.12
0.035
0.048
0.009
0.26
O
O
O
H
N
N
N
H
N
H
H
O
OH
remove
P3/P4 amide
4
remove
P2/P3 amide
a
Values are means of at least three experiments. IC50 values for human BACE-1
and human CathD inhibition were determined as described.5a
O
O
O
O
O
H
N
N
H
NHBu
N
N
H
NHBu
Table 2
H
OH
SAR of P3 to P20 spanning P2/P3 cyclohexyl hydroxyethylenes
O
OH
R1
Ph
R2
O
O
O
N
H
N
H
N
H
OH
R
O
O
O
O
N
N
H
N
H
NHBu
R1
R2
BACE-1
BACE-2
CathD
N
H
N
Compound
H
OH
O
OH
IC50
(lM), or % inhib. at 10 lM
Scheme 1. Concepts for amide replacements to reduce peptidic character.
5
6
7
8
9
Me
Et
Pr
iPr
iBu
Me
H
H
H
H
H
Me
8%
14%
24%
80%
0.17
42%
0.37
1.5
71%
64%
84%
0.15
43%
0.24
0.17
17%
15%
0.80
11%
3.7
with the published subsite specificity of BACE-1,9 the P2 amino
acid can be replaced by methionine, and even alanine derivatives
can lead to highly potent inhibitors (Table 1).
10
11
–(CH2)4–
2.7
Initially, we prepared a series of P1 Phe hydroxyethylene isos-
teres with varying substituents at the P3 site (Table 2).10
In the previous peptidic series, activity convincingly increased
from Me to iBu with groups filling the S3 pocket more tightly
(Table 1). This is in line with the co-crystal structure of compound
3 that shows the iBu group filling the S3 pocket. Surprisingly, more
bulky R1 substituents (going from Me to Et, Pr, iBu) at the novel
aminomethyl cyclohexane scaffold were virtually devoid of activity
(Table 2, compounds 5–7, 9). As an exception, the iPr derivative 8
showed submicromolar activity on BACE-1, and the closely related
enzymes BACE-2 and cathepsin D. Additionally, the dimethyl
compound 10 and the cyclopentyl derivative 11 showed micromo-
lar activity. Both observations were in contrast to the originally
assumed position of this feature in the S3 pocket.
flipped in comparison to the canonical peptidic interaction pattern.
In the binding pattern of peptides, or peptide mimics such as
OM99-2 or compound 3, there are two hydrogen bonds formed
with Thr232 of BACE-1, one between the P2/P3 amide carbonyl
and Thr232 NH, and an energetically less valuable one between
the P3/P4 amide NH and the Thr232 hydroxyl group. A comparable
orientation for compounds without a P2/P3 amide could only make
use of the weaker ligand NH interaction whilst keeping its carbonyl
unused. The binding pattern for compound 10 changed by flipping
the P3/P4 amide into the S3 pocket, resulting in strong interactions
of the amide carbonyl through hydrogen bonds with the Thr232
NH and Thr232 hydroxyl group.
With the structural information gleaned from the X-ray data,
we sought to introduce functional and spatial elements in the P3
site that would improve the activity and selectivity. To circumvent
stereochemical issues, we continued with symmetrically disubsti-
tuted gem-dimethyl derivatives of 10.
A co-crystal structure of the dimethyl compound 10 with BACE-
1 explained the so far puzzling data.11 As shown in Figure 2 the
positions of the P3 alkyl and P3/P4 amide of compound 10 are
Toward this objective, we prepared a series of N-acylated 3-(1-
amino-1-methyl-ethyl)-cyclohexane carboxylic acids starting with
the enantioenriched nitro intermediate 12 (Scheme 2).12 Final
compounds 26–28 were obtained from these building blocks and
fragment 23 by standard coupling procedures (analogous to
Scheme 3).
Modeling studies had also indicated that the S3 site could
accommodate 5- and six-membered lactams. These were prepared
as shown in Scheme 3.12 Thus, the N-allyl intermediate 18 was
acylated to 19 and 20, respectively, then subjected to ring-closing
metathesis in the presence of 5 mol % of Grubbs’ first generation
catalyst13 to give the unsaturated lactams 21 and 22 in excellent
yields. Catalytic hydrogenation and coupling with the amine frag-
ment 23 afforded 24 and 25 in good yields after purification.
Figure 1. Modeled P2–P3 cyclohexyl analogue of compound 2 (ball-and-stick) in
comparison to OM99-2 (stick), the ligand in the co-crystal structure used for
calculations. Modeling and docking were done in the Flo modeling package.8

1926
S. Hanessian et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1924–1927
1, Raney-Ni, H2,
H2PtCl6, MeOH
NO2
H
H
HN
H
CO2Me
H
CO2Me
Br
2,
NaHCO3, MeCN
64%
18
12
5 mol% Grubbs 1st
generation catalyst,
CH2Cl2, reflux
n
nCOCl
O
N
H
H
CO2Me
Et3N, CH2Cl2
19, n=0, (66%)
20, n=1, (61%)
1, 1% Pd/C, H2 MeOH
2, 1 M LiOH in H2O/MeOH (1:1)
3, PyBop, DIEA, CH2Cl2
Ph
n( )
O
N
H
H
CO2Me
21, n=1, (85%)
22, n=2, (80%)
H2N
N
H
OH
23
( )n
Ph
O
O
N
24, n=1, (62%)
25, n=2, (60%)
H
H
N
H
N
H
OH
Scheme 3.
Table 3
SAR in P3 for P2/P3 cyclohexyl hydroxyethylamines
O
R
N
H
N
H
OH
Compound
R
BACE-1
BACE-2
CathD
7.9
Ab40 CHO
a
IC50
,
lM
IC50
,
l
M
Figure 2. (A) Overlay of compound 3 (yellow) and compound 10 (green) co-crystal
structures. (B) Hydrogen bonding to Thr232: Thr232 hydroxyl is an acceptor in the
interaction with compound 3, a donor in the interaction with compound 10.
O
26
4.2
3.8
No data
N
H
O
O
27
28
24
0.58
0.15
3.7
7.1
No data
0.66
1. Raney-Ni, H2,
H2PtCl6, MeOH
N
H
NO2
H
NH
R
H
H
H
CO2Me
CO2Me
O
2. acylation,
Pyr, DMAP
0.20
0.96
0.41
0.56
N
13, R=Me, (95%)
14, R=Et, (89%)
15, R=OEt, (83%)
12
O
O
0.025
0.13
N
N
O
Me
N
1, LiAlH4
2, EtCOCl, aq. Na2CO3, CH2Cl2
H
H
15
25
0.0025
0.14
0.12
0.067
OH
16
O
Inhibition of cellular release of Ab40 was determined as described.5
a
Me
H
Trichloroisocyanuric acid
TEMPO, KBr, acetone
99%
N
H
CO2H
interactions in the S3 pocket. Consequently, amide alkylation as
in compound 28, results in improved activity (Table 3).
17
In this case, it is rather the balance of the different effects on the
three related enzymes BACE-1, BACE-2, and CathD, which is most
remarkable. An optimization for the tight fit in S3, especially by
connecting the ethyl group with the N-methyl to form the lactam
24, nearly exclusively improves BACE-1 activity, whilst the effects
on BACE-2 and CathD stay comparably constant. It was predicted
from the model, that the six-membered lactam 25 would fill the
S3 pocket even better than the five-membered lactam 24. Indeed,
going from 24 to the homolog 25, we again gained an order of mag-
nitude in activity toward BACE-1. A co-crystal structure of com-
pound 25 with BACE-1 confirmed the snug fit of the lactam in
the S3 pocket11 (Fig. 3). Intramolecular hydrophobic contacts
Scheme 2.
As previously reported,5 a change of the transition-state mi-
metic moiety from hydroxyethylene (HE) to hydroxyethylamine
(HEA) was done in anticipation of an improved in vitro as well as
cellular activity. This benefit also materialized in this series (Table
3). Whilst in the HE-series small acyl variations (e.g., from methyl
to ethyl) did not improve activity, small changes in the P3 site in
the HEA series led to a breakthrough in activity (Table 3, com-
pounds 26–28). As seen in the co-crystal structure of compound
10 (Figs. 2A and B), the amide NH is not utilized for direct interac-
tions with the protein, whilst there is additional room for tighter

S. Hanessian et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1924–1927
1927
X-ray data collection was performed at the Swiss Light Source,
Paul Scherrer Institut, Villigen, Switzerland, for the BACE com-
plexes with compounds 3 and 10. We are grateful to the machine
and beamline groups whose outstanding efforts have made these
experiments possible.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2010.01.139.
References and notes
1. Crouch, P. J.; Harding, S.-M. E.; White, A. R.; Camakaris, J.; Bush, A. I.; Masters, C.
L. Int. J. Biochem. Cell Biol. 2008, 40, 181.
2. Shankar, G. M.; Li, S.; Mehta, T. H.; Garcia-Munoz, A.; Shepardson, N. E.; Smith,
I.; Brett, F. M.; Farrell, M. A.; Rowan, M. J.; Lemere, C. A.; Regan, C. M.; Walsh, D.
M.; Sabatini, B. L.; Selkoe, D. J. Nat. Med. 2008, 14, 837.
3. For example, see: (a) Stachel, J. S. Drug Dev. Res. 2009, 70, 101; (b) Cole, S. L.;
Vassar, R. Curr. Alzheimer Res. 2008, 5, 100; (c) Ghosh, A. K.; Kumaragurubaran,
N.; Hong, L.; Koelsch, G.; Tang, J. Curr. Alzheimer Res. 2008, 5, 12; (d) Hills, I. D.;
Vacca, J. P. Curr. Opin. Drug Discov. Devel. 2007, 10, 383; (e) Durham, T. B.;
Shepherd, T. A. Curr. Opin. Drug Discov. Devel. 2006, 9, 7761.
Figure 3. Compound 25 from co-crystal structure with BACE-1—interaction surface
shown.
between the P3 lactam ring and the P1 phenyl may also contribute
to the enhanced potency through pre-organization of the confor-
mation of the active inhibitor.
4. For example, see: (a) Ghosh, A. K.; Gemma, S.; Tang, J. Neurotherapeutics 2009,
5, 399; (b) Rafii, M. S.; Aisen, P. S. BMC Med. 2009, 7, 7.
5. (a) Machauer, R.; Veenstra, S.; Rondeau, J.-M.; Tintelnot-Blomley, M.; Betschart,
C.; Neumann, U.; Paganetti, P. Bioorg. Med. Chem. Lett. 2009, 19, 1361; (b)
Machauer, R.; Laumen, K.; Veenstra, S.; Rondeau, J.-M.; Tintelnot-Blomley, M.;
Betschart, C.; Jaton, A.-L.; Desrayaud, S.; Staufenbiel, M.; Rabe, S.; Paganetti, P.;
Neumann, U. Bioorg. Med. Chem. Lett. 2009, 19, 1366.
The different effect on BACE-1 compared to CathD might be
understood by comparing the S3 pockets of the two enzymes. Whilst
there is a very tight fit towards the Ser10-Gly13 loop for compound
25 in BACE-1, this pocket (Ala13-Gln14), appears to be more open in
CathD and BACE-2,14 resulting in a less pronounced gain in activity
by slightly larger acyl replacements. Moreover, two distinct confor-
mations of the Ser10-Gly13 loop have been observed with BACE
complexes.15 While compound 3 binds to the ‘down’ conformation,
compounds 10 and 25 select the ‘up’ conformation of the Ser10-
Gly13 loop (see Fig. 2), thus achieving improved activity and selec-
tivity. However, the size and shape of the extended S3 subpocket
(S3sp)15c are not fulfilled with the P3 features in the above-described
series of BACE-1 inhibitors.
In summary, starting from peptidomimetic BACE-1 inhibitors
with counter selectivity for CathD, and going through several steps
of optimization, we identified a highly potent, non-peptidic BACE-1
inhibitor with about 50-fold selectivity over BACE-2 and CathD and
good cellular activity. As mentioned above, efflux by P-glycopro-
tein (located in the blood–brain barrier) often limits the brain
exposure of BACE-1 inhibitors. Unfortunately, testing our most ac-
tive compound 25 in the series in an in vitro model (MDCK cells
stably transfected with the gene for the human P-glycoprotein
transporter),16 resulted in a high efflux ratio (BA/AB = 97). Thus,
contrary to our initial expectations, the modification of the P2/P3
amide region exemplified by the low nanomolar prototype inhibi-
tor 25, did not overcome this challenging problem. Further efforts
will be required to optimize and improve upon these results.
6. PDB ID: 1FKN; (a) Hong, L.; Koelsch, G.; Lin, X.; Wu, S.; Terzyan, S.; Gosh, A. K.;
Zhang, X. C.; Tang, J. Science 2000, 290, 150; (b) Ghosh, A. K.; Bilcer, G.;
Harwood, C.; Kawahama, R.; Shin, D.; Hussain, K. A.; Hong, L.; Loy, J. A.; Nguyen,
C.; Koelsch, G.; Ermolieff, J.; Tang, J. J. Med. Chem. 2001, 44, 2865.
7. (a) Hanessian, S.; Yun, H.; Hou, Y.; Yang, G.; Bayrakdarian, M.; Therrien, E.;
Moitessier, N.; Roggo, S.; Veenstra, S.; Tintelnot-Blomley, M.; Rondeau, J.-M.;
Ostermeier, C.; Strauss, A.; Ramage, P.; Paganetti, P.; Neumann, U.; Betschart, C.
J. Med. Chem. 2005, 48, 5175; (b) Hanessian, S.; Yun, H.; Hou, Y.; Tintelnot-
Blomley, M. J. Org. Chem. 2005, 70, 6735; (c) Hanessian, S.; Hou, Y.;
Bayrakdarian, M.; Tintelnot-Blomley, M. J. Org. Chem. 2005, 70, 6746.
8. McMartin, C.; Bohacek, R. S. J. Comput. Aided. Mol. Des. 1997, 11, 333.
9. Turner, R. T., III; Koelsch, G.; Hong, L.; Castanheira, P.; Gosh, A.; Tang, J.
Biochemistry 2001, 40, 10001.
10. Details of the synthesis will be published elsewhere. See also Supplementary
data.
11. X-ray coordinates for the complex of BACE-1 with inhibitors 3, 10, and 25 have
been deposited in the Protein Data Bank (http://www.rcsb.org), and can be
accessed under PDB ID: 3K5D, 3K5F, and 3K5G.
12. See Supplementary data for full experimental details.
13. Schwab, P.; France, M. B.; Ziller, J. W.; Grubbs, R. H. Angew. Chem., Int. Ed. 1995,
34, 2039.
14. (a) CathD: unpublished data.; BACE-2: PDB ID: 2EWY; (b) Ostermann, N.; Eder,
J.; Eidhoff, U.; Zink, F.; Hassiepen, U.; Worpenberg, S.; Maibaum, J.; Simic, O.;
Hommel, U.; Gerhartz, B. J. Mol. Biol. 2006, 355, 249.
15. (a) Stauffer, S. R.; Stanton, M. G.; Gregro, A. R.; Steinbeiser, M. A.; Shaffer, J. R.;
Nantermet, P. G.; Barrow, J. C.; Rittle, K. E.; Collusi, D.; Espeseth, A. S.; Lai, M.-T.;
Pietrak, B. L.; Holloway, M. K.; McGaughey, G. B.; Munshi, S. K.; Hochman, J. H.;
Simon, A. J.; Selnick, H. G.; Graham, S. L.; Vacca, J. P. Bioorg. Med. Chem. Lett.
2007, 17, 1788; (b) McGaughey, G. B.; Colussi, D.; Graham, S. L.; Lai, M.-T.;
Munshi, S. K.; Nantermet, P. G.; Pietrak, B.; Rajapakse, H. A.; Selnick, H. G.;
Stauffer, S. R.; Holloway, M. K. Bioorg. Med. Chem. Lett. 2007, 17, 1117; (c)
Coburn, C. A.; Stachel, S. J.; Li, Y.-M.; Rush, D. M.; Steele, T. G.; Chen-Dodson, E.;
Holloway, M. K.; Xu, M.; Huang, Q.; Lai, M.-T.; DiMuzio, J.; Crouthamel, M.-C.;
Shi, X.-P.; Sardana, V.; Chen, Z.; Munshi, S.; Kuo, L.; Makara, G. M.; Annis, D. A.;
Tadikonda, P. K.; Nash, H. W.; Vacca, J. P.; Wang, T. J. Med. Chem. 2004, 47, 6117.
16. Wang, Q.; Rager, J. D.; Weinstein, K.; Kardos, P. S.; Dobson, G. L.; Li, J.; Hidalgo, I.
J. Int. J. Pharm. 2005, 288, 349.
Acknowledgments
We thank the NSERC of Canada and Novartis (Basel, Switzer-
land) for generous financial assistance through the Medicinal
Chemistry Chair program.