Conformationally biased P3 amide replacements of β-secretase inhibitors

Article abstract of DOI:10.1016/j.bmcl.2005.10.032

We have synthesized and evaluated a series of conformationally biased P3 amide replacements based on an isophthalamide lead structure. The studies resulted in the identification of the β-secretase inhibitor 7m which has an in vitro IC₅₀ = 35 nM

Full text of DOI:10.1016/j.bmcl.2005.10.032

Bioorganic & Medicinal Chemistry Letters 16 (2006) 641–644
Conformationally biased P3 amide replacements
of b-secretase inhibitors
Shawn J. Stachel,a,* Craig A. Coburn,a Thomas G. Steele,a Min-Chi Crouthamel,b
Beth L. Pietrak,b Ming-Tain Lai,b M. Katharine Holloway,c Sanjeev K. Munshi,d
Samuel L. Grahama and Joseph P. Vaccaa
aDepartment of Medicinal Chemistry, Merck Research Laboratories, West Point, PA 19486, USA
bDepartment of Biological Chemistry, Merck Research Laboratories, West Point, PA 19486, USA
cDepartment of Molecular Systems, Merck Research Laboratories, West Point, PA 19486, USA
dDepartment of Structural Biology, Merck Research Laboratories, West Point, PA 19486, USA
Received 15 September 2005; revised 4 October 2005; accepted 12 October 2005
Available online 2 November 2005
Abstract—We have synthesized and evaluated a series of conformationally biased P3 amide replacements based on an isophthal-
amide lead structure. The studies resulted in the identification of the b-secretase inhibitor 7m which has an in vitro IC50 = 35 nM.
The synthesis and biological activities of these compounds are described.
ꢀ 2005 Elsevier Ltd. All rights reserved.
Alzheimerꢀs disease (AD) is a neurodegenerative disease
of the brain that is characterized by the progressive
formation of insoluble amyloid plaques and fibrillary
tangles.1 These plaques and tangles consist primarily
of aggregated Ab, a 40–42 amino acid peptide that is
formed by the proteolytic processing of b-amyloid
precursor protein (bAPP) by two enzymes, b- and
c-secretase.2 The rate-limiting enzyme in this catabolic
process is b-secretase (b-site APP cleaving enzyme or
BACE-1), a novel aspartyl protease whose identity was
uncovered only a few years ago.3 As such, b-secretase
inhibition is considered an attractive therapeutic target
for the treatment and prevention of AD.
5-Substituted isophthalamides, as represented by struc-
ture 1, emerged as an early lead class that was investigat-
ed for inhibition of b-secretase.4 Inhibitor 1 possessed
the desirable properties of both good in vitro
(IC50 = 15 nM) and cell-based potencies (IC50 = 29 nM)
against BACE-1 and was found to be 10-fold selective
versus the homologous enzyme BACE-2. However,
problems associated with its pharmacodynamic proper-
ties precluded it as a candidate for advancement. A main
concern was that 1 was found to be a substrate of P-gly-
coprotein (P-gp) transport. Since a b-secretase inhibitor
must inevitably exhibit activity intracranially, it is
imperative that it be able to cross the blood–brain
barrier and achieve sustainable drug levels. It is general-
ly accepted that a greater number of functional groups
with a close proximity of hydrogen bond donors/accep-
tors to each other, i.e., amide bonds, increase the likeli-
hood that a compound will act as a P-gp substrate.5 This
said, we became interested in the replacement of the
bulky a-methylbenzamide P3 appendage present in 1
with a smaller, non-amide pharmacophore.
S2
O
O
S
N
Thr231
OH
H
N
H
O
N
NH
O
S3
S1
Gly230
1
Keywords: b-Secretase; P3; Amide replacement.
Corresponding author. Tel.: +1 215 652 2273; fax: +1 215 652
3971; e-mail: shawn_stachel@merck.com
*
The co-crystal structure of b-secretase with a statin
derivative, based on a peptidic fragment of APP, was
0960-894X/$ - see front matter ꢀ 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.10.032

642
S. J. Stachel et al. / Bioorg. Med. Chem. Lett. 16 (2006) 641–644
first published in 2000.6 In general, APP, or peptide
fragments thereof, are arranged in the standard zig-zag
conformation in the BACE-1 active site. However, the
catalytic domain of BACE-1 in comparison with those
of other aspartyl proteases (e.g., pepsin, HIV protease)
is more open and less hydrophobic. The S3 site is one
of the more hydrophobic domains in the active site
and its preferred amino acid subunit is valine.7 Within
the S3 domain of inhibitor 1 the corresponding amide
linker is involved in two distinct hydrogen bonding
interactions, namely NH to GLY230 and C@O to
THR231. We hypothesized that this P2/P3 amide group
serves mainly to orient the peptide backbone for presen-
tation of the amino acid side chains and that amide
removal could be achieved without abrogating activity,
as long as the substrate is conformationally biased to
orient the side chain toward the S3 pocket.
Aldehyde formation was achieved by the two-step pro-
cess of Weinreb amide formation and subsequent Dibal
reduction. Addition of the appropriate Grignard reagent
to aldehyde 4 produced benzylic alcohols 5 that could be
carried through the scheme or oxidized to the corre-
sponding ketones 6 using manganese oxide. Hydrolysis
and BOP mediated amide coupling to the dihydrochlo-
ride salt of (2R,3S)-N-1-cyclopropyl-2-hydroxy-4-phen-
ylbutane-1,3-diamine4 afforded the target inhibitors
7(a–n). It should be noted that secondary Grignard re-
agents that contained a b-hydrogen (i.e., cyclopentyl,
isopropyl) displayed a competitive, and often chemo-
selective, 1,2-reduction of aryl aldehyde 4 resulting in
depressed yields for compounds 7i–k.
The general schemes outlined above allowed access to
P3 substituents and the IC50 values for these inhibitors
versus BACE-1 are shown in Table 1.8 The results ob-
The truncated methyl amide 7a was chosen as a starting
point to investigate the feasibility of removing the amide
functionality. Direct replacement of the amide nitrogen
with a methylene group (7b), thereby removing the NH
hydrogen bonding contact, resulted in only a 2-fold loss
in potency. This result was encouraging since it demon-
strated that amide removal was tolerated without signif-
icant loss of activity. Next we focused on regaining
potency toward the enzyme by accessing the S3 binding
domain with a lipophilic appendage.
Table 1. SAR of P3 amide replacements
O O
S
N
OH
H
N
H
N
R
O
*
Compound
R
IC50 (nM)
980
H
N
7a
The general synthetic scheme for the synthesis of com-
pounds 7b–f is shown in Scheme 1. Mesylation of com-
mercially available dimethyl 5-amino isophthalate using
methanesulfonyl chloride in pyridine gave the desired
sulfonanilide in high yield. N-methylation was accom-
plished with sodium hydride and methyl iodide which
was followed by saponification to give benzoic acid 3.
O
O
7b
7c
7d
7e
7f
1500
1400
5500
98
Ph
O
Ph
Ph
OH
O
S
O
NH2
N
O
d, e
MeO
OMe
a, b, c
Ph
Ph
HO
OMe
5900
1300
4630
450
O
O
OH
O
O
2
3
7g
7h
7i
O
O
O O
S
S
N
N
Ph
g
f
H
OMe
R
OMe
O
O
OH
O
4
5
O
7j
240
O
S
O
O
O
O
N
S
h, i
N
7k
7l
180
OH
H
N
H
O
R
N
R
OMe
730
O
O
O
O
6
7
7m
7n
35
Scheme 1. Reagents: (a) MsCl, pyridine, CH2Cl2, 75%; (b) NaH, MeI,
DMF, 98%; (c) 0.1 N LiOH, THF/H2O, 67%; (d) EDC, HOBT, N,O-
dimethylhydroxylamine HCl, NaHCO3, 81%; (e) DIBAL-H, 94%; (f)
RMgX; (g) MnO2, CH2Cl2; (h) 1 N NaOH, THF/MeOH/H2O (i) BOP,
hydroxyethyldiamine, diisopropylethylamine.
420
*IC50ꢀs reported as n = 1.

S. J. Stachel et al. / Bioorg. Med. Chem. Lett. 16 (2006) 641–644
643
O
S
O
tained from deoxybenzoin 7c, which was predicted to
N
access the S3 domain, were less than anticipated and
only a nominal change in potency was achieved. It was
reasoned that significant amounts of enol tautomers in
solution led to either a non-optimal binding conforma-
tion or unfavorable interactions with the enol hydroxyl
group. Reduction of the ketone 7c to alcohol 7d effec-
tively prevented enolization and resulted in further
reduction in potency. We assumed the hydroxyl group
presented an unfavorable interaction with the enzyme,
so we focused on reducing the flexibility of the benzyl
group by synthesizing benzophenone 7e and a 15-fold
increase in potency was obtained. Intriguingly, the re-
duced benzophenone 7f resulted in a loss of potency.
This result coupled with that of 7d implies that loss of
activity is most likely due to an unfavorable interaction
of the benzyl hydroxyl group with the enzyme. Analysis
of the co-crystal structure of 7e in the BACE-1 active
site clearly showed the phenyl ring oriented directly into
the S3 pocket and the benzophenone carbonyl oxygen
pointing toward the solvent front.
a
4
OMe
O
8
O
S
O
NO2
NH2
N
b,c
d,e
OMe
OMe
I
OMe
I
O
O
O
9
10
11
f
g
O
O
O O
S
S
N
N
OMe
OMe
O
O
E-12
Z-12
In order to validate the ꢁorientingꢀ effects of the benzo-
phenone carbonyl group, which seemed to lack a hydro-
gen bonding contact with the enzyme, the isosteric olefin
7g was synthesized. While we anticipated this compound
to display a comparable binding conformation to 7e, a
13-fold loss in potency resulted. Nevertheless, this com-
pound exhibited a 4-fold enhancement over the sp3 ana-
log 7h. It can be reasoned that selective orientation is a
factor that contributes to potency but the ketone oxygen
gained additional stabilization through electrostatic
contacts, most likely with the solvent boundary.
Scheme 2. Reagents: (a) Cyclopropylmethyl triphenylphosphonium
bromide, n-BuLi, 46%; (b) TfOH, NIS, 37%; (c) SnCl2 Æ H2O, THF/
EtOH, 96%; (d) MsCl, pyridine, CH2Cl2, 85%; (e) NaH, MeI, DMF,
98%; (f) cyclopropylacetylene, 9-BBN, PdCl2 Æ (dppf) CH2Cl2, AsPh3,
Cs2CO3, 73%; (g) cyclopropylacetylene, InCl3, DIBAL-H, Et3B,
Pd(dba)2 Æ CHCl3, P(furyl)3, 93%.
modest, 37%, the remaining material was unreacted ben-
zoate. Subsequent reduction of the iodoarene with stan-
nous chloride afforded aniline 10. The aniline was then
mesylated and methylated using the previously
described conditions. Access to trans isomer 7l was real-
ized via a Suzuki reaction between 11 and the organobo-
rane derived from the hydroboration of cyclopropyl
acetylene resulting in E-12. Synthesis of the cis isomer
(7m) employed an indium(II) catalyzed cross coupling
reaction recently reported by Oshima.10 In that regard,
triethylborane induced hydroindination of cyclopropyl-
acetylene produced the (Z)-alkenylindium species which
smoothly underwent palladium mediated cross coupling
to yield the (Z)-alkene Z-12 exclusively in 93% yield. We
were pleased to find that the cis-olefin 7m did in fact ori-
ent itself in the S3 pocket with excellent binding activity.
Figure 1 shows an overlay of the X-ray crystal structure
of 111 with a model of 7m. The figure depicts how the
cis-alkene effectively mimics the s-trans amide binding
orientation present in 1. In contrast, the trans-olefin
(7l) displayed a 20-fold reduction in activity versus the
cis-olefin. In fact, the trans-olefin (7l) was even less
favorable than the fully saturated analog 7n reinforcing
the effect that conformational bias has on potency.
To further improve the potency of benzophenone 7e,
various aryl replacements were investigated with a goal
toward filling the S3 binding domain with the optimal
ligand (7i–k). While molecular modeling predicted
cyclopentyl derivative 7k to have the optimal fit, it was
proven experimentally to be slightly less potent than
the phenyl analog 7e. One possible explanation is that
p-stacking within the S3 binding pocket could provide
the additional increase in binding energy.
The above results established that omission of the P2/P3
amide can be achieved while retaining potency only if
the P3 appendage is properly oriented. As such, we
turned our attention to an alternative means of orienting
the P3 group, that of a geometrically defined internal
olefin. We had predicted through modeling a priori that
the cis-alkenyl isomer (7m) would be most conforma-
tionally biased to access S3 thereby mimicking the s-
trans amide bond geometry. Initial attempts to synthe-
size the desired E/Z olefins via a Wittig reaction resulted
in a 2:1 mixture of cis/trans isomers that proved to be
chromatographically inseparable (Scheme 2, cf. 4 ! 8).
In order to access a geometrically pure material, a mod-
ified synthetic route that employed iodide 11 was de-
vised (Scheme 2). Methyl 3-nitrobenzoate was
iodinated with N-iodosuccinimide in triflic acid to pro-
duce the iodoarene.9 The strongly acidic iodination pro-
cedure was necessary due to the electron-deficient nature
of the aryl ring. While the yield for the reaction was
As noted previously, compound 1 had significant phar-
macokinetic liabilities associated with it. While 1 was
considered to be a severe P-gp substrate, the apparent
permeability (Papp) was so low (0.6 · 10ꢀ6 cm/s) that
accurate P-gp efflux values were indeterminable. In con-
trast, while compounds such as 7m, containing an olefin-
ic linkage, were still found to be moderate substrates for
P-gp efflux (B/A-A/B mdr1a = 8) the Papp values
(14 · 10ꢀ6 cm/s) were significantly increased as to permit

644
S. J. Stachel et al. / Bioorg. Med. Chem. Lett. 16 (2006) 641–644
Acknowledgment
We thank Lixia Jin for P-gp analysis.
References and notes
1. Glenner, G. G.; Wong, C. W. Biochem. Biophys. Res.
Commun. 1984, 120, 885.
2. Wolfe, M. S. J. Med. Chem. 2001, 44, 2039.
3. For a recent review see: Vassar, R. J. Mol. Neurosci. 2001,
17, 157.
4. Stachel, S. J.; Coburn, C. A.; Steele, T. G.; Loutzenhe-
iser, E. F.; Gregro, A. R.; Rajapakse, H. A.; Lai, M.-T.;
Crouthamel, M. C.; Xu, M.; Tugusheva, K.; Lineberger,
B. L.; Pietrak, B. L.; Espeseth, A. S.; Shi, X.-P.; Chen-
Dodson, E.; Holloway, M. K.; Munshi, S. K.; Simon,
A. J.; Kuo, L. C.; Vacca, J. P. J. Med. Chem. 2004, 47,
6117.
Figure 1. Overlay of 1 (green) with a model of 7m (yellow) depicting S3
access using a conformationally defined olefin.
5. Schinkel, A. Adv. Drug Deliv. Rev. 1999, 36, 179.
6. Hong, L.; Koelsch, G.; Lin, X.; Wu, S.; Terzyan, S.;
Ghosh, A.; Zhang, X.; Tang, J. Science 2000, 290, 150.
7. Gruninger-Leitch, F.; Schlatter, D.; Kung, E.; Nelbock,
P.; Dobeli, H. J. Biol. Chem. 2002, 277, 4687.
8. Pietrak, B. L.; Crouthamel, M.-C.; Tugusheva, J. E.;
Lineberger, M.-X.; DiMuzio, J. M.; Steele, T. G.;
Espeseth, A. S.; Stachel, S. J.; Coburn, C. A.; Graham,
S. L.; Vacca, J. P.; Shi, X.-P.; Simon, A. J.; Hazuda, D. J.;
Lai, M.-T. Anal. Biochem. 2005, 342, 144.
9. Olah, G.; Wang, Q.; Sandforda, G.; Prakash, G. K. S.
J. Org. Chem. 1993, 58, 3194.
10. Takami, K.; Yorimitsu, H.; Oshima, K. Org. Lett. 2002, 4,
2993.
11. The PDB file for BACE1/inhibitor 1 complex has been
deposited with the Protein Data Bank (PDB identifier
2B8L).
measurement. It should be noted that 7m still contains a
hydroxyethyl amine moiety, an amide bond, and a sul-
fonamide, all contributors to P-gp efflux.
In conclusion, we have demonstrated that the P2/P3
amide is non-critical for BACE-1 activity. The amide
functionality acts mainly to orient the P3 ligand and
the amide hydrogen bonding contacts do not significant-
ly contribute to the binding energy. This was demon-
strated through the synthesis of compounds that are
conformationally biased to access S3. The ability to re-
place amide linkages with more hydrophobic homologs
results in compounds possessing both lower polar sur-
face area and less hydrogen bond donors and acceptors,
two critical properties that contribute to P-gp efflux.