Monosubstituted γ-lactam and conformationally constrained 1,3-diaminopropan-2-ol transition-state isostere inhibitors of β-secretase (BACE)

Article abstract of DOI:10.1016/j.bmcl.2011.06.109

The synthesis, evaluation, and structure-activity relationships of a class of γ-lactam 1,3-diaminopropan-2-ol transition-state isostere inhibitors of BACE are discussed. Two strategies for optimizing lead compound 1a are presented. Reducing the overall size of the inhibitors resulted in the identification of γ-lactam 1i, whereas the introduction of conformational constraint on the prime-side of the inhibitor generated compounds such as the 3-hydroxypyrrolidine inhibitor 28n. The full in vivo profile of 1i in rats and 28n in Tg 2576 mice is presented.

Full text of DOI:10.1016/j.bmcl.2011.06.109

Bioorganic & Medicinal Chemistry Letters 21 (2011) 6916–6924
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Monosubstituted c-lactam and conformationally constrained
1,3-diaminopropan-2-ol transition-state isostere inhibitors of b-secretase
(BACE)
Kenneth M. Boy a, , Jason M. Guernon a, Jianliang Shi a, Jeremy H. Toyn b, Jere E. Meredith b,
⇑
Donna M. Barten b, Catherine R. Burton b, Charles F. Albright b, Jovita Marcinkeviciene c, , Andrew C. Good d,à
Andrew J. Tebben e, Jodi K. Muckelbauer f, Daniel M. Camac f, Kimberley A. Lentz g, Joanne J. Bronson a,
Richard E. Olson a, John E. Macor a, Lorin A. Thompson III a
,
a Department of Neuroscience Discovery Chemistry, Research and Development, Bristol-Myers Squibb, 5 Research Parkway, Wallingford, CT 06492, USA
b Department of Neuroscience Discovery Biology, Research and Development, Bristol-Myers Squibb, 5 Research Parkway, Wallingford, CT 06492, USA
c Department of Chemical Enzymology, Research and Development, Bristol-Myers Squibb, 311 Pennington-Rocky Hill Road, Pennington, NJ 08534, USA
d Department of Computer Assisted Drug Design, Research and Development, Bristol-Myers Squibb, 5 Research Parkway, Wallingford, CT 06492, USA
e Department of Computer Assisted Drug Design, Research and Development, Bristol-Myers Squibb, 3551 Lawrenceville Road, Princeton, NJ 08648, USA
f Department of Protein Science and Structure, Research and Development, Bristol-Myers Squibb, 3551 Lawrenceville Road, Princeton, NJ 08648, USA
g Department of Pharmaceutical Candidate Optimization: Metabolism and Pharmacokinetics, Research and Development, Bristol-Myers Squibb, 5 Research Parkway,
Wallingford, CT 06492, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 30 June 2011
The synthesis, evaluation, and structure–activity relationships of a class of
2-ol transition-state isostere inhibitors of BACE are discussed. Two strategies for optimizing lead com-
c
-lactam 1,3-diaminopropan-
pound 1a are presented. Reducing the overall size of the inhibitors resulted in the identification of -lac-
c
Keywords:
Alzheimer’s
Beta-secretase
BACE
Transition-state isostere
tam 1i, whereas the introduction of conformational constraint on the prime-side of the inhibitor
generated compounds such as the 3-hydroxypyrrolidine inhibitor 28n. The full in vivo profile of 1i in rats
and 28n in Tg 2576 mice is presented.
Ó 2011 Elsevier Ltd. All rights reserved.
Alzheimer’s disease (AD) is a progressive neurodegenerative
disease which currently affects 5 million Americans.1 Since age is
the single best predictor of AD, the growing elderly population is
expected to cause the ranks of the affected to grow dramatically
(est. 16 million cases in 2050) over the coming decades.2 Current
approved therapies for AD, including acetylcholinesterase inhibi-
tors and NMDA antagonists, do not slow the progression of AD,
and are only palliative in nature.3 Therapies which would directly
modify the course of the disease are urgently needed, and would
benefit patients, their families and their caregivers.
The hallmark features of an AD-affected brain are neurofibril-
lary tangles composed of hyperphosphorylated tau and plaques
of deposited beta-amyloid (Ab) peptide. While the precise se-
quence of events that lead to the development of AD is still under
active investigation, the dominant amyloid hypothesis4 holds that
the 42 amino acid form of Ab (Ab42) derived from proteolysis of
the amyloid precursor protein (APP) is toxic and is causative of
the downstream disease cascade. Originally the insoluble plaques,
composed of Ab monomers, were hypothesized to be the toxic
agent responsible for neurodegeneration. However, recent work
has suggested that the soluble oligomers of Ab42 are highly toxic
to neurons5 and are more likely to be the cause of the neurodegen-
eration seen in the disease. Inhibitors of either b-secretase or
c-secretase, two enzyme activities responsible for Ab production,
would be expected to slow disease progression by lowering the
pool of toxic, soluble Ab. BACE-1 cleaves APP to produce the N-ter-
minus of the Ab peptides, which is the first step in APP process-
ing.6,7 Removal of BACE-1 activity by chemical inhibition or gene
deletion leads to complete blockage of Ab production. Given that
BACE is mainly expressed in the CNS and that BACE knockout mice
are viable, inhibitors of BACE-1 are hypothesized to be safe, dis-
ease-modifying agents for the treatment of AD.
⇑
Corresponding author. Tel.: +203 677 6278, fax: +203 677 7702.
Previously,8 the development of
c-lactam peptidomimetic BACE
inhibitors (e.g., 1a) was described. Key features of this molecule in-
cluded the 1,3-diaminopropan-2-ol transition state isostere and
the conformational constraint inherent in the P3-bearing c-lactam.
E-mail address: boyk@bms.com (K.M. Boy).
Present address: Novartis Institutes for BioMedical Research, 9 Cambridge Center,
Cambridge, MA 02142, USA.
à
Present address: Genzyme Pharmaceuticals, 153 Second Avenue, Waltham, MA
02451, USA.
0960-894X/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2011.06.109

K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
6917
molecules. In this Letter, we describe our efforts to maintain po-
tency with smaller and more permeable molecules. Initially, three
main strategies were employed: (1) elimination of the acetamide
P2
functionality from the
c-lactam, (2) reduction of the size of the
P4
P2 homophenylalanine side chain, and (3) potency enhancement
via incorporation of an additional conformational constraint on
the prime side of the transition-state isostere.
P2'
O
O
O
OH
H
H
NH
N
N
N
The synthons for the synthesis of the
target compounds are described in Scheme 1. Utilizing two related
methods, the intermediate -allylcarboxylic acids 4a–c, as well as
the -allylcarboxylic acids 4d–e started from acylated Evans’ aux-
c-lactam portion of the
O
X
P3
a
1 X = H
1a X = F
P1
a
X
iliaries. Stereocontrolled allylation of 2a–c afforded the intermedi-
ates 3a–c, with excellent diastereocontrol. Hydrolysis of the
auxiliary with lithium peroxide provided carboxylic acids 4a–c.
The oxygenated carboxylic acid derivatives 4d and 4e were
prepared via the pentenoylated auxiliary 5 (opposite sense of
stereocontrol). This intermediate was deprotonated with NaHMDS
and alkylated with bromomethyl methyl ether and reductive re-
While 1a has acceptable potency and selectivity over other
aspartyl proteases (i.e., cathepsin D, cathepsin E and pepsin), the
large molecular weight and modest Caco-2 permeability were
key issues that needed to be addressed to make more drug-like
O
O
O
O
O
O
R3
R3
R3
a
b
c
O
N
O
HN
O
N
OH
2a-c
4a-c
3a-c
O
O
O
O
g
d
e,f
O
OH
HN
O
O
N
O
OH
4d
6
5
O
TIPSO
OR
O
O
OTIPS
f
k
h,i
O
N
OH
4e
5
OH
9
7 R = H
8 R = Si(i-Pr)3
j
2a-4a R3 = n-Bu
2b-4b R3 = i-Pr
2c-4c R3 = i-Bu
Scheme 1. Synthesis of monosubstituted
c
-lactam precursors 4a–e. Reagents and conditions: (a) n-BuLi, R3CH2C(O)Cl, THF, ꢀ78 °C to rt, 38–87%; (b) NaHMDS, allyl bromide,
THF, ꢀ78 °C, 41–66%; (c) LiOH, H2O2, H2O, THF, 0 °C to rt, 30–68%; (d) n-BuLi, 4-pentenoyl chloride, THF, ꢀ78 °C to rt, 97%; (e) NaHMDS, bromomethyl methyl ether, THF,
ꢀ78 °C; (f) LiBH4, THF, MeOH, 0 °C to rt, 79% from 8; (g) Jones reagent, acetone, rt, 37%, three steps from 5; (h) (1) 5, i-Pr2NEt, CH2Cl2, Bu2BOTf, 0 to ꢀ78 °C, (2) Et2AlCl, hexane,
CH2Cl2, ꢀ78 °C, PrCHO, (3) add (2) to (1), ꢀ78 °C; (i) 5:1 MeOH:30% H2O2, 40% from 5, dr ꢁ4:1; (j) 2,6-lutidine, CH2Cl2, TIPSOTf, 0 °C, then separate diastereomers, 89%; (k)
Dess–Martin periodinane, CH2Cl2, then NaClO2, NaH2PO4, 2-methylbutene, t-BuOH, H2O, >90%.
O
R2
O
R2
a
b,c
4a-e
R3
O
R3
O
R2
N
R
N
H
O
O
O
H2N
12a-j R = Me
13a-j R = H
11a-j
d,e
O
10a R2 = Me
10b R2 = -(CH2)2Ph
R2 = -(CH2)2Ph
R2 = Me
11a-13a P3 = n-Bu
11b-13b P3 = i-Pr
11c-13c P3 = i-Bu
11f-13f P3 = n-Bu
11g-13g P3 = i-Pr
11h-13h P3 = i-Bu
11i-13i P3 = CH2OMe
11d-13d P3 = CH2OMe
11e-12e R3 = (R)-CH(OTIPS)Pr
13e R3 = (R)-CH(OH)Pr
11j-12j R3 = (R)-CH(OTIPS)Pr
13j R3 = (R)-CH(OH)Pr
Scheme 2. Synthesis of monosubstituted
c-lactam carboxylic acid synthons 13a–j. Reagents and conditions: (a) 4a–e, HATU, N-methylmorpholine, DMF, 48–90%; (b) O3,
MeOH, then PPh3; (c) Et3SiH, TFA, CH2Cl2, 0 °C, 71–96% from 11; (d) for 11e and 11j: TBAF, THF, rt; (e) 2 M LiOH, THF, rt, >90% from 12.

6918
K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
moval of the auxiliary afforded alcohol 6. Oxidation with Jones re-
agent supplied carboxylic acid 4d. To produce intermediate 4e, an
aldol reaction between the boron enolate of 5 and butyraldehyde
activated with diethyl aluminum chloride afforded 7. The resulting
alcohol was protected as the TIPS ether. Removal of the auxiliary
with lithium borohydride gave alcohol 9, which was oxidized by
sequential treatment by Dess–Martin periodinane and buffered so-
dium hypochlorite to supply the carboxylic acid 4e.
followed by KMnO4 yielded the versatile alcohol 19. This alcohol
was hydrogenated to provide directly the amine 18b. The amine
18c containing the epimeric alcohol was generated by Mitsonobu
inversion (to form alcohol 20) followed by hydrogenation. Addi-
tionally, alcohol 20 was reacted under Mitsonobu conditions to ob-
tain, after hydrogenation, aryloxy derivatives 18d–f.
The final inhibitors were prepared as is shown in Scheme 4.
Condensation of the lactam carboxylic acids 13a–j (Scheme 2) and
21a–e10 with the known amino-alcohol 2210 using HATU provided,
after N-Boc deprotection with TFA, the 1,3-diaminopropan-2-ol
derivatives 1a–m. Likewise, 13a–j (Scheme 2) and 21a–e10 were
condensed with the elaborated amines 238 and 248 to afford, after
hydrogenation, the tetrahydroisoquinoline (THIQ) compounds
25a–d and 26.8 The proline derivatives 28a–y were prepared by
condensing carboxylic acids 13a–j (Scheme 2) and 21a–e10 with
amines 18a–f (Scheme 3) with subsequent hydrolysis of both the
TBS ether and the N-Boc protecting groups with HCl in dioxane.
Amino analogs 29 and 30 were generated from the protected
pyrrolidinol-lactams 27l and 27n by azide displacement of the cor-
responding mesylates, with inversion of configuration. Hydrolysis of
the TBS ether and N-Boc carbamate with HCl followed by hydroge-
nation of the azide yielded the final compounds 29 and 30. The
benzyloxy derivatives 31a–f were furnished by the alkylation of
27l with various benzyl bromides and (bromomethyl)pyridines.
The predominant feature of our 1,3-diaminopropan-2-ol isoste-
In Scheme 2, the acids 4a–e were then converted to the
tam carboxylic acids 13a–e by condensation with P2-bearing
aminoesters (10a–b), using HATU as the coupling agent, to yield
the -lactam precursors 11a–j. Ozonolysis of 11a–j, followed by
reduction of the resulting hemiaminal with triethylsilane in TFA
formed the -lactams 12a–j. Saponification with lithium hydroxide
c-lac-
a
-
c
c
afforded the carboxylic acids 13a–j.
The syntheses of protected pyrrolidine-based P1–P20 diamines
18a–f utilized the previously described protected aldehyde 159
(Scheme 3). Stereocontrolled aldol addition of the boron enolate
of (S)-4-benzyl-3-(3-(3,5-difluorophenyl)propanoyl)oxazolidin-2-
one (14) with protected aldehyde 15 followed by protection of
the newly formed alcohol as a TBS ether afforded compound 16.
During installation of the TBS group, the use of 2,6-lutidine in both
the substrate solution as well as in the TBS triflate solution was re-
quired to avoid undesired cleavage of the N-Boc protecting group
on 15. The chiral auxiliary was then hydrolyzed using lithium per-
oxide, and the resulting carboxylic acid was reacted with DPPA to
effect a Curtius rearrangement. The resulting isocyanate was
trapped by benzyl alcohol to produce the N-CBZ-protected inter-
mediate 17. While direct hydrogenation of 17 afforded the primary
amine 18a, the removal of the allyl group with Wilkinson’s catalyst
re chemotype (i.e., 1a) is a gem-disubstituted c-lactam group. As
observed from the X-ray crystal structure of the des-fluoro analog
1,8 the lactam orients the P3 isobutyl substituent into the S3 pocket
normally occupied by the valine side-chain in the natural substrate
(interaction with I110 and G230), while projecting the acetamide
O
Bn
O
N
OTBS
O
O
Boc
Boc
N
O
c,d,e
a,b
F
N
O
N
O
F
Bn
O
O
15
14
16
F
F
OTBS
OTBS
OTBS
Boc
Boc
Boc
OH
CBZHN
CBZHN
CBZHN
f,g
N
N
N
h,i
F
F
F
17
19
20
O
OH
F
F
F
j
j
j
OTBS
OTBS
OTBS
Boc
Boc
Boc
H2N
H2N
H2N
N
N
N
F
F
F
O
OH
OH
18c
18b
18a
F
F
F
OTBS
Boc
H2N
k,j
N
20
F
O
Ar
18d Ar = Ph
18e Ar = 2-pyridyl
18f Ar = 2-pyridyl
F
Scheme 3. Synthesis of functionalized pyrrolidine-based 1,3-diaminopropan-2-ol fragments 18a–f. Reagents and conditions: (a) (S)-4-benzyl-3-(3-(3,5-difluorophenyl)pro-
panoyl) oxazolidin-2-one, Bu2BOTf, DIPEA, CH2Cl2, ꢀ78 °C to rt, 74%, dr 3:1; (b) 2,6-lutidine, TBSOTf, CH2Cl2, see text, 78%; (c) LiOH, H2O2, THF/H2O, 73%; (d) DPPA, DIPEA,
PhMe, 75 °C; (e) PhCH2OH, 75 °C, 77% over two steps; (f) Wilkinson’s catalyst, EtOH/water, reflux; (g) KMnO4, EtOH, pH 10 buffer, rt, 60% over two steps; (h) DEAD, PPh3, THF,
4-nitrobenzoic acid, rt, 83%; (i) K2CO3, MeOH, rt, 58%; (j) H2, Pd/C, MeOH, 51–91% for 18a–c, 23–66% over two steps for 18d–f; (k) DEAD, PPh3, THF, ArOH, rt.

K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
6919
OH
O
R2
Boc
O
R2
OH
R4
R3
H
H
R4
R3
H2N
N
OH
N
N
O
a,b
N
N
O
F
O
O
F
13a-j R3: see tables 2,3
22
1a-m
R4 = H
21a-e R3: see tables 2,3
R4 = NHAc
F
F
OH
O
R2
OH
P
N
H
H
R4
R3
H2N
N
N
13 or 21
N
F
a,c
F
Y
O
Y
25a-d Y=H
26 Y=OMe
23 Y=H, P = p-methoxybenzyl
24 Y=OMe, P = diphenylmethyl
F
F
O
R2
OX
P
TBSO
H
N
R4
R3
Boc
13 or 21
N
N
H2N
a
N
F
O
F
F
Y
Y
27a-y
18a-f
Y: OPr, OH, OAr
(Scheme 3)
X=TBS, P=Boc
d
Y: see Tables 2,3
F
Ph
28a-y
X=P=H
Y: see Tables 2,3
O
OH
H
N
H
AcHN
N
e,f,d,g
27l (Y = (R)-OH)
27n (Y = (S)-OH)
N
F
O
NH2
30 Y = (S)-NH2
29 Y = (R)-NH2
F
Ph
O
OH
H
N
H
AcHN
N
h,d
N
27l (Y = (R)-OH)
O
F
F
O
Ar
31a-f
Scheme 4. Synthesis of final compounds 1, 25a–d, 26, and 28–31. Reagents and conditions: (a) HATU, N-methylmorpholine, DMF; (b) TFA, CH2Cl2, 48–81% over two steps; (c)
H2, Pd/C, MeOH, HOAc, 31–75% over two steps; (d) 4 N HCl in dioxane, water, 29–34% from 18, 8–32% from 27l; (e) MsCl, pyridine; (f) NaN3, DMF, 70 °C, 67–100% over two
steps; (g) H2, Pd/C, MeOH, 24–32% over two steps; (h) 2 equiv NaHMDS, ArCH2Br.
Figure 1. X-ray crystal structure of 1 in the BACE-1 active site.

6920
K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
Table 1
R2
OH
O
H
H
R4
R3
N
N
N
O
O
F
F
Compound
R4
R3
R2
HPLCa IC50 (nM)
HEKswb IC50 (nM)
1a
1bc
1c
1d
1e
1f
NHAc
NHAc
H
H
H
sBu
CH2cPr
nBu
iPr
iBu
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
6.1
11
38
75
20
32
n.d.
10
350
220
240
210
H
CH2OMe
OH
OH
1g
H
(CH2)2Ph
5.2
90
1h
1i
1j
1k
1l
NHAc
CH2cPr
nBu
iPr
iBu
CH2OMe
Me
Me
Me
Me
Me
46
60
170
89
480
27
81
240
67
400
H
H
H
H
1m
H
Me
68
27
a
b
c
n = 2.
n = 4.
1b is the des-fluoro analog at P1.
Table 2
R2
OH
Het
O
H
H
N
H
N
H
R4
R3
N
N
N
O
O
O
C
A
B
F
F
Compound
R4
R3
R2
Het
HPLCa IC50 (nM)
HEKswb IC50 (nM)
25a
25b
25c
25d
26
28a
28b
28c
28d
28e
28f
NHAc
NHAc
H
iBu
sBu
nBu
nBu
sBu
iBu
sBu
nBu
iPr
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
Me
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
A
A
A
A
B
C
C
C
C
C
C
8.6
54
9.5
110
1400
720
20
2.0
2.7
84
n.d.
n.d.
4.2
1.4
2.6
19
21
5.7
23
H
NHAc
NHAc
NHAc
H
H
H
99
69
89
iBu
CH2OMe
H
OH
OH
28g
H
(CH2)2Ph
C
4.7
48
28h
28i
28j
H
H
H
nBu
iPr
iBu
Me
Me
Me
C
C
C
87
160
35
68
390
110
28k
H
Me
C
70
55
a
n = 2.
n = 4.
b
group toward S4 (Fig. 1). The homophenylalanine sidechain has
hydrophobic contacts with T72 and R235 within S2, while the P1
benzyl group interacts with F108 and L30 at the bottom of the
S1 pocket. The 1,3-diaminopropan-2-ol transition-state isostere
fragment forms the key H-bonds with D32 and D228 (from the
ligand alcohol and ammonium moieties, respectively), while the
m-methoxybenzyl group is in contact with Y71 and Y198 within
the S20 binding pocket. In considering how the size of these inhib-
itors could be reduced, we made molecular models of several po-
tent
c-lactam inhibitors using information from the previously

K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
6921
described X-ray structure.11 We also solved additional moderate-
resolution crystal structures (unpublished results). Together, these
data suggested that two binding conformations (rotamers) for the
acetamide functionality were possible, with the T232 carbonyl
interacting with the NHAc in only one pose. This suggested that
the loss in potency upon removing this entire group (R4 = NHAc)
would be modest. Compounds made according to this hypothesis
were screened for BACE-1 enzyme inhibition12 and for their ability
to inhibit Ab1-40 production in cultured HEKsw cells.13 A four- to
six-fold loss in BACE-1 potency in the enzyme assay was observed
for the monosubstituted c-lactam 1c (38 nM) versus the reference
disubstituted lactams 1a and 1b (Table 1). Further exploration of
the P3 group revealed that branching of R3 was tolerated (1d), with
a slight improvement in potency seen the isobutyl analog 1e
(20 nM). Ether analog 1f had similar potency to 1c. Among the co-
hort of the monosubstituted lactams (R4 = H), alcohol 1g was supe-
rior in both in vitro (HPLC IC50) and cellular (HEKsw IC50) potency
(5.2 and 90 nM, respectively). To further reduce the molecular
weight of the inhibitors, the homophenylalanine P2 moiety was re-
placed with alanine (1i–m). Compounds 1i–k were a modest two-
to five-fold less potent than the homophenylalanine derivatives
(compare with 1c–e), however the oxygenated derivatives 1l and
1m lost more than an order of magnitude in potency relative to
Figure 2. Model of 28a docked into the BACE-1 active site.
Figure 3. X-ray crystal structure of 28a in the BACE-1 active site.

6922
K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
Table 3
R2
OH
O
H
N
H
R4
R3
N
N
O
Y
Z
F
F
Compound
R4
R3
R2
Y
Z
HPLCa IC50 (nM)
HEKswb IC50 (nM)
28l
NHAc
H
NHAc
H
sBu
nBu
sBu
nBu
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
H
H
OH
OH
OH
H
6.5
25
0.89
110
29
650
11
28m
28n
28o
OH
H
720
OH
OH
28p
H
(CH2)2Ph
OH
H
16
220
28q
28r
NHAc
H
sBu
nBu
Me
Me
OH
OH
H
H
31
96
>660
1400
28s
H
Me
OH
H
40
870
28t
28u
28v
28w
28x
28y
29
NHAc
NHAc
NHAc
H
H
H
NHAc
NHAc
NHAc
NHAc
NHAc
NHAc
NHAc
NHAc
sBu
sBu
sBu
nBu
iBu
iBu
sBu
sBu
sBu
sBu
sBu
sBu
sBu
sBu
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
Me
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
(CH2)2Ph
H
H
H
H
H
H
H
NH2
H
H
OPh
2.6
2.6
5.9
n.d.
n.d.
n.d.
34
1.4
2.1
2.5
360
510
330
270
190
12
O-2-pyridyl
O-3-pyridyl
O-2-pyridyl
O-2-pyridyl
O-2-pyridyl
NH2
30
H
OBn
50
31a
31b
31c
31d
31e
31f
n.d.
2.7
3.0
10
1.5
1.4
OCH2(mCF3C6H4)
OCH2(mCNC6H4)
OCH2(pFC6H4)
OCH2(2-C5NH4)
OCH2(3-C5NH4)
71
H
H
H
H
4.2
8.9
1.3
2.0
a
n = 2.
n = 4.
b
Table 4
Ab reduction and drug exposure in rat plasma and brain
Compound Rat plasma Ab % Rat plasma
Rat brain Ab %
B/P
ratio,
5 h
of control,a 5 h
exposure (nM),
5 h
of control,a 5 h
1i
28c
26
23
1600
4000
112
114
0.18
0.02
a
‘% of control’ = 100 ꢀ (% reduction in Ab).
stituted lactam 1b (11 nM in vitro, 10 nM in HEK cells). We viewed
the modest loss in potency for 1i compared to 1b favorably, since
1i had two fewer heteroatoms and 114 Da less molecular weight
than 1b. The in vivo profile of compound 1i will be discussed later
in this paper.
We then turned our attention toward increasing the potency of
our analogs by introducing a conformational constraint within the
1,3-diaminopropan-2-ol fragment. As previously described,8 tying
back the methoxybenzyl group into a tetrahydroisoquinoline
(THIQ) afforded potent analogs 25a and 26 (Table 2). In contrast
to the acyclic analogs 1, however, cellular potency diminished
greatly (>10-fold) when combined with mono-substituted lactams
(25c–d, 1400 and 720 nM cell IC50’s, respectively). Nevertheless,
the tolerance of the enzyme for the novel cyclic disubstituted
lactams 25a and 26 encouraged us to look for additional con-
strained templates, ideally with a lower molecular weight.
Removing the prime-side aryl group allowed us to focus on
smaller nitrogen heterocycles. Molecular modeling suggested that
the 4-hydroxypyrrolidine ring would be well tolerated, and the
convenient handle at C(4) offered the potential for further
Figure 4. Model of 28n docked in to the BACE-1 active site.
their homophenylalanine congener (e.g., from 32 nM (1f) to
480 nM (1l)). Importantly, we observed that 1i (60 nM) was essen-
tially equipotent with the disubstituted lactam 1h (46 nM) in vitro.
Furthermore, 1i was only sixfold less potent in vitro and eightfold
less potent in the cells (81 nM) than the homophenylalanine disub-

K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
6923
derivatization. Specifically, the propyloxypyrrolidine derivatives
28a–k were the first compounds examined due to the predicted
H-bonding interaction between Y198 and the C(4) ether oxygen
as well as the hydrophobic interaction between the n-propyl group
and Y71 (Fig. 2). The improved potency of the propyloxypyrroli-
dine disubstituted lactams 28a and 28b (Table 2, in vitro and cel-
lular IC50’s less than 3 nM) over the THIQ analogs 25a–b and 26
demonstrated the superiority of the propyloxypyrrolidine isostere.
The X-ray crystal structure of BACE-1 in complex with 28a14–16
confirmed both the predicted Y198-OPr H-bond and the propyl
ether hydrophobic contact with Y71 (Fig. 3). The group of mono-
substituted lactams 28c–g bearing various P3 groups were all more
than 10-fold more potent in cells than the THIQ comparator 25c.
The i-butyl and (R)-1-hydroxybutyl analogs 28e and 28g were
equipotent (ꢁ5 nM) to the disubstituted lactam 25a, and only
slightly less potent in vitro than 28a and 28b.
We reduced the molecular weight of our analogs further by
trimming the P2 group to an alanine residue. As seen in our acyclic
series, the potency of compounds 28h–k was diminished (four- to
14-fold, depending on the pair examined). However, compound 28j
was among the smallest compounds studied (509 Da), and was a
reasonably potent inhibitor of BACE-1 (HPLC IC50 = 35 nM).
Examination of molecular modeling suggested several addi-
tional analogs at the C(4) position of the pyrrolidine (Table 3).
The (R)-4-hydroxy analog 28l (6.5 nM) was threefold less active
than its propyloxy congener 28b, but a striking order-of-magni-
tude less potent in cells (29 nM). Consistent with our prior SAR,
the corresponding monosubstituted lactam 28m (R4 = H) was less
potent than the disubstituted lactam 28l (R4 = NHAc), again with
a larger reduction in cellular potency than in vitro potency (>20-
fold and 4-fold, respectively). Interestingly, the (S)-4-hydroxypyr-
rolidine analog 28n was ꢁ1 nM in vitro, but was only 11 nM in
cells. Molecular modeling (Fig. 4) predicted a unique H-bond be-
tween the C(4)–OH and the T72 hydroxyl group on the enzyme’s
flap. This interaction may also explain the impressive selectivity
of 28n over other aspartyl proteases (3400- to 6000-fold over
cathepsin D, cathepsin E, and pepsin), since T72 is not conserved
in these enzymes. Compounds lacking the interaction with T72
are less selective (i.e., 28b, 31- to 40-fold selective). Among the co-
hort of (S)-4-hydroxypyrrolidine monosubstituted lactams 28o–s,
comparator 28o (R3 = n-Bu) was the least potent (110 nM), while
28p (R3 = (R)-1-hydroxybutyl) was 16 nM. Additionally, the P2 ala-
nine derivatives 28q and 28s were also more potent than 28r (Ta-
ble 3, 96 nM, R3 = n-Bu). The subtle factors responsible for the
detrimental effect of the P3 n-butyl substituent in this sub-series
were not examined. The 4-aminopyrrolidine analogs 29 and 30
were each less potent than their hydroxyl counterparts.
that the lack of efficacy was likely a consequence of non-specific
binding in the central compartment.19 The propyloxy-pyrrolidine
analog 28c demonstrated similar peripheral Ab lowering to 1i, how-
ever the negligible central exposure for this compound was disap-
pointing. Although not determined for 28c, several close analogs
(28a, 28h, and 28j) were shown to be Pgp substrates in the bi-direc-
tional Caco-2 assay measuring the ratio of basolateral to apical per-
meability versus apical to basolateral permeability. The compounds
28a, 28h, and 28j had ratios of >12, >17, and 32, respectively, indicat-
ing significant active transport.
Given the in vitro potency and selectivity of 28n, we examined
this compound in male Tg 2576 mice (30 mpk, sc).20 At 5 h post-
dose, we observed a modest reduction in plasma Ab (66% of con-
trol) with
a plasma concentration of 780 nM (Tg mouse
%Fu = 10%, 78 nM free-drug). Although we didn’t observe brain Ab
lowering, we were quite interested to see a high B/P ratio (0.87,
680 nM). To try to understand the disconnect between the expo-
sure and lack of efficacy, a timecourse experiment was performed.
Figure 5 shows that, while peripheral exposure declined over time
in the expected fashion in conjunction with a return of peripheral
Ab to baseline, the central exposure remained ꢁ800 nM over 12 h
with no pharmacological effect seen at any timepoint. The reason
for the prolonged retention of 28n in the central compartment is
unknown, but could involve non-specific binding to brain tissue
or active transport since 28n is a substrate for Mrp.21 Regardless
of the mechanism, unbound drug levels in the brain of 28n were
Aβ lowering after 30 mpk SC dose of 28n in Tg 2576 mice
120
Plasma
100
Brain
80
60
40
20
0
0
5
10
15
Time After Dose (hours)
Several additional ether groups at the (R)-C(4) were equipotent
to the propyloxy parent 28b, including the phenoxy and pyridoxy
derivatives 28t–v, however the monosubstituted analogs 28w–y
all were two orders of magnitude less potent in cells. Several
benzylic ethers (31a–f) were also shown to be potent inhibitors
of BACE-1, but did not contribute the additional potency needed
to justify the added molecular weight.
Concentration of 28n (nM) vs. time (h)
100000
10000
1000
100
plasma
brain
To test our hypothesis that compounds of lower molecular
weight would result in greater brain penetrance, we evaluated PK
and PD effects in female CD-IGS rats.17 Monosubstituted lactams
1i and 28c were dosed at 30 mpk ip, and we determined both drug
levels and %Ab lowering 5 h post-dose (Table 4). Compound 1i low-
ered plasma Ab to 26% of control (74% Ab reduction) with a periph-
eral exposure of 1.6 l
M (rat free fraction 2.9%,18 46 nM free drug), in
reasonable agreement with the cellular IC50 of 81 nM. Despite mod-
erate central exposure of 270 nM (B/P 0.18) there was no effect on
central Ab levels. While we did not attempt to measure the % free
drug in the brain, it appears that 270 nM total drug was not sufficient
to cover the 81 nM free-drug IC50 at the target. This data suggests
10
0
5
10
15
Time (h)
Figure 5. PK/PD of 28n in Tg 2576 mouse plasma and brain.

6924
K. M. Boy et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6916–6924
Camac, D. M.; Muckelbauer, J. K.; Lentz, K. A.; Grace, J. E.; Burton, C. R.; Toyn, J.
H.; Marcinkeviciene, J.; Meredith, J. E.; Albright, C. F.; Macor, J. E. Bioorg. Med.
Chem. Lett. 2011, Accompanying Article.
insufficient to reduce brain Ab. For a discussion of similar observa-
tions with additional potent BACE-1 inhibitors, see Ref. 13.
In summary, we discovered a class of potent BACE-1 inhibitors
which were lower in molecular weight and heteroatom count than
our lead compound 1a. In animals, compound 1i was efficacious in
reducing plasma Ab, but did not lower brain Ab even though it has
a molecular weight 22% less than 1a. We also developed a versatile,
constrained pyrrolidine template containing a 1,3-diaminopropan-
2-ol isostere which exhibited improved potency versus its acyclic
congeners. Specifically, the hydroxypyrrolidine 28n was potent,
exquisitely selective against other aspartyl proteases, and CNS pe-
netrant. However, achieving Ab lowering in the brain still remains
a challenge, highlighted by the difficulty in obtaining central expo-
sure from relatively small compounds such as 1i and 28c as a result
of Pgp efflux, and additionally, what seems to be the high non-spe-
cific binding within brain tissues as seen for 28n.
9. Thompson, L. A., III.; Boy, K. M.; Shi, J.; Macor, J. E. U.S. Patent 7,388,007, 2008.
10. Decicco, C. P.; Tebben, A. J.; Thompson, L. A., III.; Combs, A. P. U.S. Patent
7,557,137, 2009.
11. The co-crystal structure of the acyclic compound 4-amino-5-((S)-3-sec-butyl-
1-((S)-1-((2S,3R)-3-hydroxy-4-(3-methoxybenzylamino)-1-phenylbutan-2-
ylamino)-1-oxo-4-phenylbutan-2-yl)-2-oxopyrrolidin-3-ylamino)-5-
oxopentanoic acid in BACE was used as the starting point for the substituted
pyrrolidine models. Initial conformations of 28a and 28n were generated by
manual manipulation of the acyclic compound to incorporate the
hydoxypyrrolidine ring variants. Conformational sampling of the
hyrdroxypyrrolidine was then carried out using Macromodel (version 9.8,
Schrodinger LLC, New York, NY) with the OPLSAA-2005 forcefield.
Conformations of the hydroxypyrrolidne ring were sampled, while the
remainder of the molecule was fixed. The resulting conformations were then
rigidly docked into the BACE/ligand active site using Glide (version 5.6,
Schrodinger). Top scoring poses were minimized using the refine module of
Prime (version 2.2, Schrodinger) and final models were selected based on low
energy and visual inspection.
12. Iben, L. G.; Kopcho, L.; Marcinkeviciene, J.; Zheng, C.; Thompson, L. A.; Albright,
C. F.; Toyn, J. H. Eur. J. Pharmacol. 2008, 593, 10.
Acknowledgments
13. Meredith, J. E., Jr.; Thompson, L. A., III; Toyn, J. H.; Marcin, L.; Barten, D. M.;
Marcinkeviciene, J.; Kopcho, L.; Kim, Y.; Lin, A.; Guss, V.; Burton, C.; Iben, L.;
Polson, C.; Cantone, J.; Ford, M.; Drexler, D.; Fiedler, T.; Lentz, K. A.; Grace, J. E.,
Jr.; Kolb, J.; Corsa, J.; Pierdomenico, M.; Joned, K.; Olson, R. E.; Macor, J. E.;
Albright, C. F. J. Pharm. Exper. Ther. 2008, 3262, 502.
The authors appreciate the contributions of the following
colleagues. Maria Pierdomenico, Kelli Jones, and Rudy Krause
conducted the rat and mouse in vivo experiments while Tracey
Fiedler and Jason Corsa executed the accompanying Ab assays.
Carol Krause and Cathy Kieras performed our HEKsw assays. Paul
Morin and Vidhyashankar Ramamurthy generated the BACE
constructs used in crystallographic studies.
14. Co-crystals of BACE-1 and a peptidic statine15 were soaked with 1 mM 28a in a
stabilizing buffer of 35% PEG8 K, 0.2 M ammonium sulfate, 0.1 M sodium
cacodylate pH 6.2 for 24 h.
15. Kornacker, M. G.; Copeland, R. A.; Hendrick, J. P.; Lai, Z.; Mapelli, C.; Witmer, M.
R.; Marcinkeviciene, J.; Metzler, W. J.; Lee, V. G.; Riexinger, D. J.; Muckelbauer, J.
K.; Chang, C. J.; Camac, D. M.; Morin, P. E. U.S. Patent Application Publication
2,007,149,763.
16. Coordinates for the complex of BACE-1 with inhibitor 28a have been deposited
in the Protein Data Bank (www.rcsb.org) under PDB ID 3R2F.
17. Rat experiments were performed in a manner similar to those for mice. See Ref.
13.
18. For free-fraction methods, see Ref. 13.
19. Maurer, T. S.; DeBartolo, D. B.; Tess, D. A.; Scott, D. O. Drug Metab. Dispos. 2005,
33, 175.
20. Tg 2576 mouse experiments were performed in a manner similar to those for
wt-mice, utilizing the human 26D6 antibody in place of the murine 252Q8
antibody. See Ref. 13.
References and notes
1. Maslow, K. Alzheimer Demen. 2008, 4, 110.
2. Alzheimer’s Foundation of America. http://www.alzfdn.org/AboutAlzheimers/
statistics.html (accessed April 20, 2010).
3. Palmer, A. M. Trends Pharm. Sci. 2002, 23, 426.
4. Hardy, J.; Selkoe, D. J. Science 2002, 297, 353.
5. Walsh, D. M.; Selkoe, D. J. J. Neuorochem. 2007, 101, 1172.
6. Zheng, H.; Koo, E. H. Mol. Neurodegen. 2006, 1, 5.
7. DeStrooper, B.; Annaert, W. J. Cell Sci. 2000, 113, 1857.
8. Thompson, L. A., III.; Shi, J.; Decicco, C. P.; Tebben, A. J.; Olson, R. E.; Boy, K. M.;
Guernon, J. M.; Good, A. C.; Liauw, A.; Zheng, C.; Copeland, R. A.; Combs, A. P.;
21. Cellular IC50 values were obtained in N2A cells with and without the Mrp
inhibitor MK-571. Compound 28n exhibited a threefold shift in potency.