Discovery of a series of efficient, centrally efficacious bace1 inhibitors through structure-based drug design

Article abstract of DOI:10.1021/jm501833t

The identification of centrally efficacious β-secretase (BACE1) inhibitors for the treatment of Alzheimers disease (AD) has historically been thwarted by an inability to maintain alignment of potency, brain availability, and desired absorption, distribution, metabolism, and excretion (ADME) properties. In this paper, we describe a series of truncated, fused thioamidines that are efficiently selective in garnering BACE1 activity without simultaneously inhibiting the closely related cathepsin D or negatively impacting brain penetration and ADME alignment, as exemplified by 36. Upon oral administration, these inhibitors exhibit robust brain availability and are efficacious in lowering central Amyloid β (Aβ) levels in mouse and dog. In addition, chronic treatment in aged PS1/APP mice effects a decrease in the number and size of Aβ-derived plaques. Most importantly, evaluation of 36 in a 2-week exploratory toxicology study revealed no accumulation of autofluorescent material in retinal pigment epithelium or histology findings in the eye, issues observed with earlier BACE1 inhibitors.

Full text of DOI:10.1021/jm501833t

Subscriber access provided by UMKC Libraries
Article
Discovery of a Series of Efficient, Centrally Efficacious
BACE1 Inhibitors through Structure-Based Drug Design
Christopher R. Butler, Michael A. Brodney, Elizabeth M. Beck, Gabriela Barreiro, Charles E. Nolan, Feng Pan,
Felix F. Vajdos, Kevin Parris, Alison H. Varghese, Christopher J. Helal, Ricardo Lira, Shawn D. Doran, David
Riddell, Leanne M. Buzon, Jason K. Dutra, Luis A. Martinez-Alsina, Kevin Ogilvie, John C Murray, Joseph M.
Young, Kevin Atchison, Ashley Robshaw, Cathleen Gonzales, Jinlong Wang, Yong Zhang, and Brian T. O'Neill
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/jm501833t • Publication Date (Web): 19 Feb 2015
Downloaded from http://pubs.acs.org on February 23, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Discovery of a Series of Efficient, Centrally
Efficacious BACE1 Inhibitors through Structureꢀ
Based Drug Design
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
a
a
a
e
Christopher R. Butler* , Michael A. Brodney , Elizabeth M. Beck , Gabriela Barreiro , Charles E.
b
b
d
d
d
c
Nolan , Feng Pan , Felix Vajdos , Kevin Parris , Alison H. Varghese , Christopher J. Helal , Ricardo
c
e
b
c
c
c
Lira , Shawn Doran , David Riddell , Leanne Buzon , Jason Dutra , Luis MartinezꢀAlsina , Kevin
c
c
c
b
b
b
Ogilvie , John C. Murray , Joseph Young , Kevin Atchison , Ashley Robshaw , Cathleen Gonzales ,
f
f
c
Jinlong Wang , Yong Zhang and Brian T. O’Neill
a
b
Neuroscience Worldwide Medicinal Chemistry, Neuroscience Research Unit, Pfizer Worldwide
c
Research and Development, 610 Main St, Cambridge, MA, 02139, United States; Neuroscience
d
d
Worldwide Medicinal Chemistry, Center of Chemistry Innovation and Excellence, Pharmacokinetics,
Dynamics, and Metabolism, Pfizer Worldwide Research and Development, 445 Eastern Point Road,
e
Groton, CT, 06340, United States; Eurofarma Laboratorios S.A., Departamento de Inovação, Av.
f
Vereador José Diniz, 3465, Campo Belo, São Paulo, Brasil, WuXi AppTec 288 Fute Zhong Road
Waigaoqiao Free Trade Zone Shanghai 200131, China
RECEIVED DATE (to be automatically inserted after your manuscript is accepted if required
ABSTRACT
The identification of centrally efficacious βꢀsecretase (BACE1) inhibitors for the
treatment of Alzheimer’s disease (AD) has historically been thwarted by an inability to maintain
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
alignment of potency, brain availability, and desired ADME properties. In this paper, we
describe a series of truncated, fused thioamidines that are efficiently selective in garnering
BACE1 activity without simultaneously inhibiting the closely related Cathepsin D or negatively
impacting brain penetration and ADME alignment, as exemplified by 36. Oral administration of
these inhibitors exhibit robust brain availability and are efficacious in lowering central Aβ levels
in mouse and dog. In addition, chronic treatment in aged PS1/APP mice effects a decrease in the
number and size of Aβꢀderived plaques. Most importantly, evaluation of 36 in a twoꢀweek
exploratory toxicology study revealed no accumulation of autofluorescent material in retinal
pigment epithelium or histology findings in the eye, issues observed with earlier BACE1
inhibitors.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Introduction
Alzheimer’s disease (AD) is a devastating neurological disorder that brings about a slow
progression of cognitive decline, dementia, and ultimately death. Amyloid β (Aβ)ꢀderived
plaques in the hippocampal and cortical regions of the brain are hallmarks of this disease. The
amyloid hypothesis suggests that an increase in Aβ, engendered either through its increased
production or decreased clearance, initiates a molecular cascade that eventually leads to
1,2
neurodegeneration and AD. Aβ production occurs by the consecutive degradation of amyloid
3
precursor protein (APP) by β and γꢀsecretases. The initial proteolytic event occurs via βꢀsite
APP cleaving enzyme (BACE1) within the endosome, to afford a soluble Nꢀterminal ectodomain
4
of APP (sAPPβ) and the Cꢀterminal fragment C99. The membraneꢀbound C99 is subsequently
5
cleaved by γꢀsecretase to release the Aβ isoforms, including Aβxꢀ40 and Aβxꢀ42. Recently, an
ACS Paragon Plus Environment

Page 3 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
APP mutation has been reported that has been shown to have protective effects against AD; this
6
mutant protein is cleaved more slowly by BACE1. The ability to safely, but effectively, inhibit
BACE1 (and thus modulate the Aβ cascade) has therefore been a target of great interest for a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
7
number of groups over the last two decades.
One of the consistent challenges in the effective inhibition of BACE1 in vivo has been
8
overcoming significantly impaired CNS penetration. There has been a myriad of inhibitor
classes that exhibit excellent BACE1 potency but struggle to achieve in vivo efficacy at
clinically viable doses because they fail to adequately cross the bloodꢀbrain barrier (BBB). A
key factor underlying this common problem is the strategy that has been employed to engage the
two aspartates in the catalytic site of BACE1 has generally been to design inhibitors that contain
a basic amine, which engages these residues either directly, or through a waterꢀmediated
9
,10
hydrogen bond interaction.
This amine can be deleterious to brain penetration, as it often
contains two hydrogen bond donors, and its impact can be further exacerbated by a basic pK , as
a
both of these characteristics contribute to recognition by Pꢀglycoprotein (Pꢀgp) efflux
11
transporters. In this light, the recent report of compound 1 (Figure 1) advancing to the clinic as
1
2
an efficacious BACE1 inhibitor with a wellꢀbalanced ADME profile is quite notable.
Unfortunately, clinical examination of 1 has been discontinued due to retinal toxicity observed in
12
preclinical species. We and others postulated that such ocular toxicity was potentially due to
the offꢀtarget inhibition of Cathepsin D (CatD), a related aspartyl protease that has significant
1
3,13b
regulatory function in the eye.
In its initial disclosure, 1 was described to have ~100x
1
2
selectivity between BACE1 and CatD inhibition (250 nM and 25 ꢁM IC , respectively). In
5
0
similarly run cellꢀfree fluorescence polarization (FP) assays for BACE1 and CatD, however, 1
exhibited only a 6ꢀfold selectivity for BACE1 over CatD (Figure 1).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In contrast, there is a plethora of BACE inhibitors reported to have significant CatD
1
4
selectivity, including thioamidine 2 (Figure 1). This class of inhibitors is characterized by the
presence of an amide moiety that occupies the S1/S3 pockets of BACE1 and engages Glyꢀ230
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
Figure 2a), which has been a successful strategy to significantly increase BACE1 potency and
improve selectivity over CatD. However, this structural motif imparts a significant Pꢀgpꢀ
mediated efflux liability (MDR efflux ratio (MDR) = 5.3), presumably due to the third hydrogen
bond donor and increased molecular weight (320 ꢀ 369) of 2. This efflux liability translates to
1
5
substantially impaired brain penetration observed in rodents (C /Cu,p = 0.07).
Such
u,b
impairment requires a significantly increased body burden of inhibitor in order to achieve the
same degree of central efficacy, resulting in lower safety therapeutic indices relative to an
equivalent inhibitor with good CNS penetration.
H N
S
H2N
S
2
BACE1 FP IC : 0.9 uM
BACE1 FP IC : 0.08 uM
LipE: 6.42
5
0
50
N
N
LipE: 5.01
BACE1 WCA IC : 50 nM
BACE1 WCA IC : 3 nM
5
0
50
F
CatD FP IC : 5.3 uM
CatD FP IC : >92 uM
50
50
HLM: < 8.8
MDR: 1.5
RRCK: 24
Cu,b/Cu,p = 1
HLM: 21.4
MDR: 5.3
RRCK: 15
N
HN
F
N
N
O
Cu,b/Cu,p = 0.07
1
± 2
N
Figure 1. Literature thioamidineꢀderived BACE1 inhibitors and properties
Thus, our objective herein was to efficiently improve CatD selectivity without
simultaneously impacting CNS penetration. We have adopted a reductionist approach to this
challenge, truncating back to a minimally required pharmacophore within this series, and
evaluating opportunities to improve selectivity using structureꢀbased drug design. It was
ACS Paragon Plus Environment

Page 5 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
recognized that paring back in this fragmentꢀlike approach would generate a significant loss in
BACE1 potency that would need to be addressed. Although this strategy allows for selective
reconstruction in an efficient manner, it requires a committed investment to build inhibitors back
into reasonable potency space without disruption of the balanced ADME properties. This
approach is particularly attractive when considering the observed interplay between molecular
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
weight (MW), pK , the number of hydrogen bond donors (HBDs), and their coordinated impact
a
16
on Pꢀgpꢀmediated efflux liabilities. As the thioamidine inherently contains two hydrogen bond
donors, diligence in design is required throughout to introduce additional molecular weight only
where necessary, at the same time remaining cognizant of any impact on pK , in order to
a
mitigate efflux concerns.
Results and Discussion
We initiated our design strategy by truncating the pyrimidine of 1 that occupies the S3
17
pocket to prepare fragmentꢀlike 6 , synthesis of which can be quickly accomplished via simple
1
8
acidꢀcatalyzed cyclization of intermediate 5 (Scheme 1).
Scheme 1. Preparation of compound 6
Reagents and conditions: (a) 3ꢀbromobutꢀ3ꢀenꢀ1ꢀyl 4ꢀmethylbenzenesulfonate, (Ph P) Pd, CsF, THF, 70
3
4
o
o
o
C, 90%; (b) thiourea, CH CN, 75 C, 83%; (c) 6 N HCl, 95 C, 17% then chiral separation.
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
As expected, 6 is only weakly potent at BACE1 (IC , 36 µM) and has no measurable
5
0
activity at CatD (up to 100 µM). A coꢀcrystal structure of 6 bound to BACE1 was obtained and
the binding of 6 to CatD was modeled, in an effort to identify opportunities to impart selectivity
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
Figure 2). The BACE1 structure reveals that the amidine group engages in hydrogen bonds
with the catalytic aspartic acid residues 228 and 32, with the difluorophenyl ring occupying the
S1 pocket. Analyses of the BACE1 crystal structure and CatDꢀbound model of 6 afford three
potential points of differentiation that could be exploited to gain selectivity (Figure 2a). First, an
Arg – Asp salt bridge is present in CatD in the S3 pocket, but absent in BACE1 (Region I). Thus
molecules that fit deeply into the S3 pocket are likely to disrupt the salt bridge, explaining the
selectivity of molecules similar to 2. Secondly, the fold of Pheꢀ108 loop of BACE1 is
significantly different from the corresponding region of CatD (Region II). In particular, the
residues that correspond to Pheꢀ108 and Pheꢀ109 of BACE1 are Ile and Thr, respectively, in
CatD. More importantly, the residue that would correspond to the backbone carbonyl acceptor
Pheꢀ108 in BACE1 does not present its carbonyl to the S1 pocket in CatD. This is consistent
with the structure and activity relationship of previous spiropiperidine series, where engagement
1
0
with the Pheꢀ108 backbone carbonyl significantly improved selectivity over CatD.
Third,
although the flap loop conformation represented by BACE1 Tyrꢀ71 and Trpꢀ76 is similar to that
in CatD, the substitution of Trpꢀ76 by Leu (CatD) in the flap loop prevents hydrogen bond
formation to Tyrꢀ78 (by Trp). Instead, Tyrꢀ78 engages Trpꢀ40, which is not within the flap loop,
likely rigidifying this region in CatD (Region III). Thus, it was anticipated that building towards
the flap in BACE1 would be better tolerated than the corresponding perturbation to the CatD flap
region. This analysis suggested three subtle, but key structural differences that could potentially
be exploited to design in selectivity over CatD.
ACS Paragon Plus Environment

Page 7 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(
a)
(
b)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(c)
Figure 2. a) Xꢀray crystal structure of 6 with BACE1 (green ball & stick/light blue ribbon) with
a model of 6 in Cathepsin D (orange ball & stick/yellow ribbon). Residues in the area of
interests are labeled with BACE1 residue names and numbers (light blue) following by the
corresponding CatD residue names (yellow). b) Potential regions to exploit for BACE1 vs.
Cathepsin D selectivity. c) Strategic construction to explore S2’.
Investigation of each of these key regions of differentiation delineated three potential
strategies (Regions IꢀIII, Figure 2b) for series progression. In an effort to rank the efficiency of
these strategies, the distances between the center of 6 to the corresponding key residues for each
of the regions outlined above that are expected to impart selectivity were measured (Figure 2b).
Of these, the S2’ pocket (Region III) clearly provides the closest opportunity, suggesting that
filling this pocket would be preferred from an “efficient selectivity” perspective. In contrast,
disruption of the S3 salt bridge (Region I) would require a substantial molecular weight increase;
additionally, the most efficient method in the literature for building in this direction introduced
the aforementioned third HBD, significantly increasing the risk of impaired brain penetration.
Building toward Region II can be deprioritized as well, given that optimal interactions with the
rd
Pheꢀ108 carbonyl would also require a 3 HBD. Thus, the strategy of building efficiently
towards the S2’ pocket was selected as our preferred approach to enhance selectivity for BACE1
over CatD.
ACS Paragon Plus Environment

Page 9 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The crystal structure of 6 and BACE1 suggested that access to the S2’ pocket could be
achieved by building from the Cꢀ5 position of the thioamidine ring in I, as outlined in Figure 2c.
A methoxymethyl group in this position was initially selected, to target the desired region
without a dramatic increase in lipophilicity. We chose a strategy that could broaden the potential
range of SAR and simplify the initial synthesis. We used Ellman imine technology to control the
absolute stereochemistry of the quaternary carbon at proꢀCꢀ4 but left the eventual Cꢀ5 substituent
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
9
stereochemistry unrestrained, providing both diastereomers in the P2’ region.
We expected
that separation would be straightforward at a late stage in the synthesis and a more selective
synthesis was not necessary at this point. The key late steps of the synthesis are depicted in
Scheme 2. Thioamidine ring formation stemmed from the differentially protected propanediol 7
(supporting information) equipped with (S)ꢀ2ꢀmethylpropaneꢀ2ꢀsulfinimine functionality. Lewis
acidꢀmediated addition of lithioꢀ5ꢀbromoꢀ2,4ꢀdifluorophenyl anion provided 8 with the (S)
configuration at proꢀC4 and as a 1:1 mixture of diastereomers at proꢀCꢀ5. Concomitant
hydrolysis of the sulfinamide and removal of the silyl protecting group revealed the amino
alcohol 9 as a mixture of optically active diastereomers, which was easily acylated by benzoyl
isothiocyanate. The thiourea 10 was cyclized to the thioamidine 11 as a 1:1 mixture at Cꢀ5 by
o
selective activation of the alcohol with triflic anhydride at ꢀ20 C. The protecting group was
exchanged for tꢀBOC and the diastereomers 13a and 13b were chromatographically separated at
this point. Incorporation of the versatile bromide offered a viable route to SAR extension. For
our initial two targets, the bromide was removed through halogenꢀmetal exchange and
protonation, followed by acidic removal of the BOC group. The sequence provides compounds
14a and 14b in a 1:3 ratio.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
S
O
OTBS
OCH3
OH
OCH3
S
N
HN
H2N
a
b
H C
OTBS
OCH3
F
F
3
2
4%
7
5%
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Br
Br
7
8
9
F
F
O
O
HN
S
S
NH OH
c
d
N
OCH3
HN
OCH3
F
F
7
7%
Br
Br
1
0
11
F
F
H
N
H
N
H N
S
S
F
S
2
BOC
BOC
N
OCH3
N
OCH3
N
OCH3
e
f
+
+
F
F
F
Br
Br
Br
F
F
1
2
13a
13b
1
6% (3 steps)
42% (3 steps)
H N
S
H N
2
S
2
N
OCH3
N
OCH3
g
+
F
F
F
F
14a
14b
4
7% (2 steps)
77% (2 steps)
Scheme 2. Reagents and conditions: (a) 1,5ꢀdibromoꢀ2,4ꢀdifluorobenzene, nꢀBuLi, AlMe , ꢀ
0 C, tol; (b) HCl in dioxane, rt; (c) PhC(O)NCS, THF, rt; (d) Tf O, pyridine, ꢀ20 C, CH Cl ; (e)
DBU, MeOH, 80 C; (f) tꢀBOC O, aq NaHCO ꢀTHF,rt, then chromatography; (g) 13a and 13b
separately: nꢀBuLi, ꢀ78 C, THF then TFA, CH Cl , rt.
3
o
o
7
2 2 2
o
2
3
o
2
2
ACS Paragon Plus Environment

Page 11 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
As shown in Table 1, analog 14a, which has a transꢀrelationship between the
methoxymethyl and difluorophenyl substituents, maintains some potency in both the FP and
wholeꢀcell (WCA) BACE1 assays, but remains inactive at CatD and thus indicates a productive
direction for further exploration. The diminished activity observed for 14b reinforces the
relevance of stereochemistry in this series. Although clearly some benefit accrued from growing
in this direction, the multiple degrees of freedom retained by the simple methoxy substituent in
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
4a may be problematic. First, this relatively small substituent lacks sufficient bulk to enforce a
ring conformation placing the aryl group in the crucial axial position, which is required for
binding. It was supposed that a significant population exists in the inactive ring conformation,
leading to the observed suboptimal potency. Second, the methoxymethyl group itself has no
conformational bias to orient specifically towards S2’, again limiting the population of the most
active conformation
Table 1. In Vitro Data for 14a/14b
d
e
Cpd
BACE1 FP
IC50 (µM)
CatD FP
IC50 (µM) (fold)
BACE1 WCA
IC50 (nM)
MDR Er
HLM
(mL/min
kg)
LipE
LE
a
b
c
/
14a
21
88
>100 (5.5x)
>100
850
1.95
1.71
33
3.8
0.34
0.30
14b
3,940
<8
2.7
a
b
c
IC values obtained from BACE1 FP Assay. IC values obtained from CatD FP Assay. IC values
50
50
50
d
obtained from BACE1 WholeꢀCell Assay (WCA). Ratio from the MSꢀbased quantification of
apical/basal and basal/apical transfer rates of a test compound at 2 ꢁM across contiguous monolayers
from MDR1ꢀtransfected MDCK cells. Hepatic clearance predicted from in vitro human microsomal
e
stability study.
It was envisaged that the two conformational challenges identified above could be
overcome by fashioning the two alkyl substituents into a THF or THP ring, as fused bicyclic
2
0
variants of thioamidine 6 (II in Figure 2c). This would necessarily lock the aryl group into
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
an axial orientation with respect to the thioamidine ring and provide a template for further
precise elaboration towards S2’ from the newly constructed ring; this was anticipated to exploit
the flap differences outlined above. We settled on a synthetic strategy that would
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
accommodate an examination of ring size, as well as facile control of multiple Csp
stereocenters, in order to further probe the S2’ regions of the protein. An approach that
performed well for this class of inhibitor involved early ring construction via an intramolecular
[
3+2] cycloaddition to form a [3.3.0] or [4.3.0] fused ring system whose constraints enforce
stereocontrol along the ring fusion (Scheme 3). In these early cases, we also decided to use
chiral chromatography to separate enantiomers for testing in biological screens. Introduction
of the aryl function occurs predominantly on the convex face, with excellent stereocontrol; this
was followed by ring expansion to the thioamidine. In the forward sense, allylic or
homoallylic alcohols are elaborated to the oxime 18a,b, one of several possible precursors to a
nitrile oxide. Nitrile oxide cycloaddition occurs in a straightforward fashion for either the fused
tetrahydrofuran or tetrahydropyran ring, with a small amount of accompanying nitrile oxide
dimer or intermolecular isoxazoline formation.
Subsequent addition of the (2,4ꢀ
difluoro)phenyl anion was very sensitive to the conditions employed: at a minimum, the
process required imine activation with BF ꢀOEt and low temperature to avoid benzyne
3
2
formation from the (2,4ꢀdifluoro)phenyl anion. We also noticed that variable quantities of
unreacted isoxazoline often remained following the addition, presumably due to deprotonation
o
of the isoxazoline ring. The optimal conditions involved formation of the anion at ꢀ78 C,
followed by dropꢀwise addition of BF ꢀOEt and the isoxazoline while maintaining the reaction
3
2
o
temperature below ꢀ73 C. As expected, these reactions were highly stereoselective for the cisꢀ
ring fusion, with no detectable formation of the transꢀisomer. The thioamidine ring formation
ACS Paragon Plus Environment

Page 13 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
was initiated after reductive cleavage of the NꢀO bond of 20a,b with zinc metal and acylation
with a suitable isothiocyanate. The resulting thiourea 22a,b could be cyclized to the
21
thioamidine by the novel use of the Ghosez reagent, which selectively activated the primary
alcohol, or by treatment with aqueous acid, with loss of water. The FMOC or benzoyl
protecting groups were removed by standard methods to afford the target compounds 24a and
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
24b.
22
o
Scheme 3. Reagents and conditions: (a) n=1 : HOCH CH OH, KOH, THF, 60 C; n=2:
2
2
o
o
o
BrCH CH(OEt) , NaH, THF, 0 C then 68 C; (b) n=1: (COCl) , DMSO, CH Cl , ꢀ78 C; n=2:
aq HCl, THF, 70 C; (c) n=1, 2: NH OHꢀHCl, NaOAc, EtOHꢀH O, 60 C; (d) n=1,2: NaOCl,
2
2
2
2
2
o
o
2
2
Et N, CH Cl , rt; n=2: then chiral HPLC; (e) n=1, 2: nꢀBuLi, 1ꢀbromoꢀ2,4ꢀdifluorobenzene,
3
2
2
o
o
THFꢀtol 1:3, ꢀ78 C then BF ꢀOEt , 19a,b, toluene, ꢀ78 C – rt; (f) n=1,2: Zn, AcOH, rt; (g) n=1:
3
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
o
BzNCS, CH Cl rt; n=2: FMOCꢀisothiocyanate, CH Cl , rt (h) n=1: Ghosez reagent, tol, 70 C;
n=2: concd HCl (aq), MeOH, 70 C; (i) n=1: DBU, MeOH, 60 C, then chiral HPLC; n=2:
MeCN, piperidine, rt.
2
2,
2
2
;
o
o
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
As shown in Table 2, these compounds indeed show improved selectivity over CatD.
The tetrahydrofuran derivative 24a exhibits only modest improvements in BACE1 potency to
1
7.2 ꢁM, whereas the pyranꢀcontaining 24b is >20x selective for BACE1 over CatD, driven by a
significant increase in BACE potency. Moreover, 24b exhibits a lipophilic efficiency (LipE)
improvement over 14a, suggesting that building in this direction is indeed a productive direction,
rather than a simple improvement in potency due to increased lipophilicity. A modest increase in
ligand efficiency hinted that the substituent may partially be influencing the population of active
conformations for 24b, relative to the unsubstituted 14a. Additionally, the compounds maintain
preferred ADME properties (MDR Er < 2.5, HLM <10), reaffirming our strategy and providing
an excellent starting point for further functionalization.
Table 2. In Vitro Data for 24a/24b
d
e
Cpd
BACE1 FP
IC50 (µM)
CatD FP
IC50 (µM) (fold) IC50 (nM)
BACE1 WCA
MDR Er
HLM
LipE LE
a
b
c
(mL/min/
kg)
2
4a
17.2
5.8
>100 (>7x)
375
1.48
1.23
< 8
3.9
4.4
0.36
24b
>100 (>20x)
290
< 8
0.38
a
b
c
IC values obtained from BACE1 FP Assay. IC values obtained from CatD FP Assay. IC values
obtained from BACE1 WholeꢀCell Assay (WCA). Ratio from the MSꢀbased quantification of
apical/basal and basal/apical transfer rates of a test compound at 2 ꢁM across contiguous monolayers
50
50
50
d
e
from MDR1ꢀtransfected MDCK cells. Hepatic clearance predicted from in vitro human microsomal
stability study.
An Xꢀray crystal structure of 24b bound in the BACE1 active site reveals a significant
difference compared to what had been observed previously with 6 (Figure 3). In this structure,
ACS Paragon Plus Environment

Page 15 of 101
Journal of Medicinal Chemistry
1
2
3
Tyrꢀ71 has rotated, forced into a lower energy conformation by the addition of the fused ring
system. The hydrogen bond that had existed between Tyrꢀ71 and Trpꢀ76 is now replaced by a
waterꢀmediated hydrogen bond from Trpꢀ76 to the THP oxygen (red water, Figure 3). In
addition, this significant conformational change reveals a pocket adjacent to the space occupied
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
3
by the THP ring. Analysis using HydroSite suggests that there is a highꢀenergy, unstable water
exposed in this pocket (cyan water). It was proposed that the introduction of small alkyl
substituents into this lipophilic pocket should displace this calculated water, likely enhancing
BACE1 potency.
Figure 3. Xꢀray coꢀcrystal of 24b (magenta) in BACE1 in comparison with 6 (green). Tyrꢀ71 has
rotated to a different rotameric conformation, to accommodate the THP ring. A
crystallographically observed water is making a hydrogen bond to the oxygen of the THP. A
computationally predicted “unstable” water molecule is shown in cyan.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To direct substituents from the fused THP system into the S2’ pocket, our synthetic
strategy targeted the creation of versatile non–racemic homoallylic alcohols that would
ultimately set the remaining stereocenters in the molecule and provide a handle for lateꢀstage
modification. This strategy, in combination with the [3+2] cycloaddition approach described
above, resulted in the synthesis of chiral, lateꢀstage intermediates required for analog generation
in a stereoselective manner (Scheme 4). To prepare the initial analogs, addition of nucleophiles
to chiral epoxides or aldehydes was first explored. The first and most versatile R substituent was
derived from (R)ꢀ(−)ꢀglycidyl benzyl ether. Copperꢀcatalyzed addition of vinylmagnesium
bromide afforded the desired (R)ꢀhomoallylic alcohol 25a (Scheme 5) bearing the versatile
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
4
hydroxymethyl substituent in a protected form. Alternatively, when desired substituents could
not be generated late in the synthetic sequence, these groups were introduced beginning with
aldehydes. In the case where R is cyclopropyl, we found that addition of allylmagnesium
bromide to cyclopropyl carboxaldehyde was the most efficient approach to this homoallylic
2
5
substrate, albeit initially in racemic form (25b).
Scheme 4. Retrosynthetic analysis for analog generation
The alcohol was elaborated as previously described to afford the oxime intermediate
28a,b as a direct precursor to a nitrile oxide. Subsequent [3+2] cycloaddition was controlled by
establishment of an equatorial R group and proceeded with excellent diastereoselectivity to
afford an intermediate in which the benzyloxymethyl group (or cyclopropyl) of 29a,b was trans
ACS Paragon Plus Environment

Page 17 of 101
Journal of Medicinal Chemistry
1
2
3
to the ring junction of the newly formed fused isoxazoline ring system (confirmed by 2D NMR,
see supporting information for details). Introduction of the final chiral center through Lewis acidꢀ
catalyzed anion addition occurred with complete selectivity to form a cis ring junction in modest
to good yields. Thioamidine ring formation was initiated by NꢀO bond reduction of 30a,b using
zinc in acetic acid to afford 31a,b. Thiourea formation by reaction with a suitable isothiocyanate
was straightforward and was followed by thioamidine cyclization. Two complementary methods
were used for this purpose. The free alcohol of thiourea 32a and the cyclopropyl THP 32b were
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
each activated with triflic anhydride at ꢀ60 C followed by spontaneous thioamidine cyclization to
33a,b. Alternatively 32a’ was transformed with aqueous acid in methanol at elevated
temperature to afford the thioamidine 33a’. Following routine final transformations, this versatile
route afforded thioamidines 34, 35 and 35’ equipped with three chiral centers and versatile
protecting groups to accommodate later transformation in the sequence.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
Scheme 5. Reagents and conditions: (a) R=CH OBn: vinylMgBr, CuIꢀEt O, THF ꢀ78 Cꢀ rt; R =
2
2
24
o
o
o
cꢀC H ;(b) BrCH CH(OEt) , NaH, THF, 0 C then 68 C; (c) aq HCl, THF, 40ꢀ70 C; (d)
3
5
2
2
o
NH OHꢀHCl, NaOAc, EtOHꢀH O, 60 C; (e) aq NaOCl, Et N, CH Cl , rt; (f) BF ꢀOEt , IPEꢀtol
2
2
3
2
o
2
3
2
o
1:1, or tol, ꢀ78 C then 1ꢀbromoꢀ2,4ꢀdifluorobenzene, nꢀBuLi, ꢀ78 Cꢀrt; (g) Zn, AcOH, rt; (h) R
CH OBn: FMOCꢀNCS, CH Cl , rt or R = CH OBn: BzꢀNCS, CH Cl , rt; R = cꢀC H :
BzNCS, CH Cl , rt; (i) P=FMOC, R = CH OBn: concd HCl, aq, MeOH, 70 C; P=Bz, R =
=
2
2
2
2
2
2
3
5
o
2
2
2
o
o
CH OBn: Tf O, pyr, CH Cl , ꢀ60 C; R = cꢀC H : Tf O, pyr, CH Cl , ꢀ50 C; (j) R = cꢀC H :
2
2
2
2
3
5
2
2
2
3
5
o
o
DBU, MeOH, 60 C, then chiral HPLC; (k) P=FMOC, R = CH OBn: concd HCl, 120 C
2
followed by piperidine, DMF, rt then tꢀBOC) O, Et N, THF/MeOH, rt; P=Bz, R = CH OBn:
2
3
2
o
NaBrO , EtOAcꢀH O, 0 C.
3
2
ACS Paragon Plus Environment

Page 19 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
The versatile intermediate 35 was transformed into a variety of substituents aimed at
exploring the S2’ binding pocket. The alcohol group was immediately converted to the fluoride
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(
36) using (diethylamino)sulfur trifluoride (DAST) or to the methoxy group with methyl iodide
43); alternatively, the alcohol can be deoxygenated in two steps via chloride formation followed
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
by Superhydride reduction (42). Alcohol 35 can be reliably transformed into the aldehyde,
which becomes a precursor for alkyl substituents via a Wittig reaction followed by
hydrogenation (39,40). The stereochemistry of 39 was confirmed through 2ꢀD NMR studies and
the configuration of the three chiral centers of 36 was confirmed by single crystal xꢀray (see
supplementary material). The aldehyde also underwent difluorination with DAST to provide 38.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 6. Reagents and conditions: (a) P=tꢀBOC: DAST, pentaneꢀCH Cl 1:1, rt, 51% then
2
2
o
o
o
TFA, CH Cl 0 C, 74%; (b) P= Bz: MeI, NaH, THF, 40 C, 69% then DBU, MeOH, 70 C,
2
2,
o
7
2%; (c) P=Bz: SO ꢀPyr, DMSO, Et N, CH Cl ,rt, 71%; (d) DAST, CH Cl , ꢀ20 C, 58% then
DBU, MeOH, 60 C, 96%; (e) Ph P Me Br , tꢀBuOK, THF, ꢀ15 Cꢀrt, 82% then H , Pd/C, MeOH,
rt, 100% then DBU, MeOH, 80 , 58%; (f) SOCl , tol, 80 C, 90%; (g) LiEt BH, THF, 0 C, 81%
then DBU, MeOH, 80 C, 81%; (h) Ph P iꢀPr Br , nꢀBuLi, THF, 0 Cꢀrt, 38% then H , Pd/C,
3 3 2 2 2 2
o
+
ꢀ
o
3
2
o
o
o
2
ꢀ
3
o
+
o
3
2
o
MeOH, rt, 26% then DBU, MeOH, 60 C, 52%.
In all cases, the effect of small substituents on the THP ring improves the desired
selectivity over CatD by significantly enhancing BACE1 potency, relative to 24b (Table 3).
The introduction of even a simple methyl group (42) improves BACE1 potency 7ꢀfold in both
the FP and WCA assays. In addition, 42 maintains the excellent balance of ADME properties,
ACS Paragon Plus Environment

Page 21 of 101
Journal of Medicinal Chemistry
1
2
3
with low clearance, good permeability and no predicted Pꢀgp efflux. Increasing the bulk of the
alkyl group further improves potency in most cases, but begins to impact the balanced profile of
these compounds. The methyl ether (43) introduces a new site of metabolism, as evidenced by a
significant increase in HLM despite a decrease in logD, relative to the ethyl derivative 39 (HLM
Cl,int,app 67 vs <8). Larger alkyl substituents (40) tend to improve potency, but a trend of
increasing Pꢀgp efflux suggests a potential to encounter lower brain availability in this direction.
The cyclopropyl substituted analog (34) offers a nice balance of potency, clearance and MDR,
but unfortunately suffers from significant inhibition of CYP2D6 (IC , 1.4 ꢁM). It does,
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
0
however, raise the possibility that modifying the electronics of the substituent could further aid
in Pꢀgp efflux liability. Gratifyingly, the fluorineꢀcontaining compounds 36 and 38 maintain
similar potency and selectivity, at the same time retaining good predicted brain penetration
(MDR BA/AB ratio of 0.9 for each). Moreover, all of these compounds exhibit a significant
increase in CatD selectivity relative to 1.
Table 3. In Vitro Data for Compounds
1
BACE1 WCA MDR Erd HLMe
Cpd
R
BACE1 FP CatD FP
IC50 (µM)
LipE LE
a
b
c
IC50 (µM) (fold) IC50 (nM)
(mL/min
/
kg)
2
4
4b
2
H
5.1
>100 (>19.6x)
72.0 (102x)
265
39
1.3
1.2
<8
<8
4.4
5.1
0.38
0.42
Me
0.70
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
4
3
CH
Et
2
OMe
0.85
0.47
>91.4 (107x)
63 (134x)
30
59
1.6
1.7
67
<8
4.7
4.6
0.38
0.41
39
3
4
cꢀPr
0.28
0.15
45 (161 x)
24
25
1.4
<8
19
4.9
4.4
0.41
0.41
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
40
iꢀBu
25.1 (167x)
2.82
3
3
8
6
CHF
2
0.67
1.07
83.1 (124x)
>81 (>76x)
59
61
0.9
0.9
<8
<8
4.6
4.6
0.38
0.38
CH
F
2
a
b
c
IC values obtained from BACE1 FP Assay. IC values obtained from CatD FP Assay. IC values
obtained from BACE1 WholeꢀCell Assay (WCA). Ratio from the MSꢀbased quantification of
apical/basal and basal/apical transfer rates of a test compound at 2 ꢁM across contiguous monolayers
from MDR1ꢀtransfected MDCK cells. Hepatic clearance predicted from in vitro human microsomal
stability study.
5
0
50
50
d
e
The consistent increase in LipE for these compounds relative to the unsubstituted 24b
suggests that productive BACE1ꢀligand interactions are indeed introduced through exploration of
this vector. Moreover, the constant or improved LE through the THP substitution reinforces that
relative potency was garnered for the resultant increases in MW. A crystal structure of 36 and
BACE1 (Figure 4) confirms that the fluoromethyl substituent is directed toward the lipophilic
pocket revealed by the rotation of Tyrꢀ71 and, further, overlays with the unstable water predicted
from the structure of 24b. As expected, the bound compound maintains a lowꢀenergy
conformation that places the fluoromethyl substituent in a favorable equatorial orientation on the
THP ring. Thus, displacement of the calculated, unstable water resulted in a significant (5ꢀ7x)
improvement in BACE1 potency, as well as improvements in ligand and lipophilic efficiency.
ACS Paragon Plus Environment

Page 23 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Crystal structure of 36 and BACE1. The fluoromethyl substituent is directed toward
the lipophilic pocket revealed by the rotation of Tyrꢀ71.
Based on the improved CatD selectivity and balanced ADME profile, the ADME
properties of compounds 38, 42 and 36 were examined in more detail. As shown in Table 4, all
three compounds exhibit low in vitro clearance in human liver microsomes. Further examination
across rodent, dog and nonꢀhuman primate derivedꢀhepatocytes suggest a low rate of clearance in
all species. In addition to high passive permeability (RRCK > 10), all analogs showed excellent
CNS penetration (C /Cu,p >0.9) in mouse with no predicted Pꢀgp efflux. In addition, 36 was
u,b
further profiled in a dog neuropharmacokinetic study and showed a similar C /C ratio of 0.78,
u,b u,p
suggesting confidence in translating the brain availability across multiple species.
Table 4. In Vitro and in Vivo PK Properties of Representative Compounds
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
PK Properties
42
36
<8
38
ꢀ1
ꢀ1 a
In vitro hꢀCl (mL min kg ) <8
<8
h
ꢀ1
ꢀ1 b
In vitro rꢀCl (mL min kg ) <14 <14 <14
h
c
RRCK AB
17
20
32
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
d
MDR BA/AB
1.2 0.9 0.9
3.90 9.2 3.51
0.43 0.6 0.36
0.10 0.18 0.11
0.92 1.42 1.20
e
B/P (mouse)
f
F (plasma)
u
f
F (brain)
u
g
u,p
C /C
u,b
a
b
Hepatic clearance predicted from in vitro human microsomal stability study. Hepatic clearance
predicted from in vitro rat microsomal stability study. RRCK cells with low transporter activity were
isolated from MadineꢀDarby canine kidney cells and were used to estimate intrinsic absorptive
permeability; Ratio from the MSꢀbased quantification of apical/basal and basal/apical transfer rates of a
test compound at 2 ꢁM across contiguous monolayers from MDR1ꢀtransfected MDCK cells.
c
d
e
Determined from plasma and brain exposures in mice 60 min post 10 mg/kg subcutaneous dosing;
f
g
Determined from equilibrium dialysis. As assessed in CDꢀ1 Mice
In vivo pharmacology
To assess whether the improved pharmacokinetic properties of this class of thioamidines
translate into a pharmacodynamic (PD) response in vivo, acute BACE1 inhibition was examined
in male wildꢀtype 129/sve mice following an acute subcutaneous administration of 36 at a dose
of 30 and 150 mg/kg. Aβ lowering was measured in brain, cerebrospinal fluid (CSF) and plasma
over a time course of 1 to 30 h post dose (Figure 5). The corresponding exposures were also
measured in these three compartments in order to delineate the PK/PD relationship for BACE1
inhibition. From a PD perspective, a significant reduction of Aβ was seen in all three
compartments. There also appears to be a threshold of maximum inhibition in rodent brain that
cannot be breached despite the higher exposures achieved with the 150 mg/kg dose.
ACS Paragon Plus Environment

Page 25 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
Although the time and degree of maximal inhibition was equivalent for the two doses, the
decay in inhibition back to baseline was significantly delayed for the 150 mg/kg dose. In brain, a
sustained maximum reduction of Aβxꢀ40 and Aβxꢀ42 was observed at 3ꢀ5 h postꢀdose, at 69%
and 72% respectively. Efficacy in CSF was more robust with a maximum inhibition of 85%,
while the weakest efficacy was observed in plasma, with 54% inhibition. Hysteresis is also
clearly demonstrated, with a delay in peak efficacy (3 h) from Cmax (1 h). Utilization of
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
6
previously described methods allowed for a C determination of 371 nM (free brain).
eff
Figure 5. Time course effect of 36 on brain, plasma and CSF Aβ following acute administration
in mouse. Compound 36 was dosed at 30 or 150 mg/kg sc at time = 0. Data are the mean ± SEM,
n=5/group.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In addition to demonstrating brain, plasma, and CSF lowering of Aβ in mice, the efficacy
of 36 was also confirmed in biomarker compartments of higherꢀorder species. Using cisterna
magnaꢀcannulated beagle dogs, the acute effect of 36 was evaluated in serial CSF and plasma
samples collected over a 60 h period (Figure 6). Administration of an 18 mg/kg PO dose of 36
produced a sustained reduction in CSF concentrations of Aβxꢀ40 and Aβxꢀ42, with a maximum
decrease of 74% compared to vehicle, but at a Tmax of 8 h, rightꢀshifted from that seen in mouse.
A decrease was also observed in plasma Aβxꢀ40 (45% at Tmax = 4 h postꢀdose).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 27 of 101
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 6. Time course of effect of 36 on plasma and CSF Aβ following acute 18 mg/kg p.o.
administration in beagle dogs. Compound 36 was dosed at 18 mg/kg PO at time = 0. Data are the
mean ± SEM, n=4/group.
Evaluation of the efficacy of 36 in PS1/APP mice with respect to plaque burden
modulation was examined following two months of chronic treatment at 100 mg/kg (po) using 2ꢀ
27
photon excitation microscopy. Twoꢀphoton imaging provides the advantage of being able to
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 101
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
follow changes in individual plaques over time, with each mouse acting as its own control versus
prior baseline measurements. Thinnedꢀskull cranial windows were created over the
somatosensory cortex of PS1(G384A)APPSWE mice for imaging amyloid plaques by intravital
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
28
2
ꢀphoton microscopy. Images were acquired at baseline and at 1 month and 2 months after
initiating compound treatment. The total plaque load in the imaged region was increased by
1±13% over the 2 month period in vehicleꢀtreated animals, suggesting a steady growth of
5
plaques in these mice at the age of 6.5 ꢀ 8.5 months. The change rate of individual plaques
displayed significant variability, from no detectable difference to a severalꢀfold increase.
Surprisingly, twoꢀmonth treatment with 36 led to a significant regression of total plaque burden,
with an average decrease of 26.8% (Figure 7, n=6; p<0.05). These results suggest that BACE1
inhibition was effective not only in halting amyloid plaque load increase, but also resulted in
regression of plaque burden when administered during the plaque growth phase. Figure 8
illustrates representative images of preexisting parenchymal amyloid plaques and subsequent
reductions following two month treatment with 36.
The demonstration of reduced amyloid deposition in the form of extractable brain Aβ as
well as reduction of plaque load and plaque number, following chronic BACE1 inhibitor
2
9
treatment in amyloid mouse models, has been reported. However, to our knowledge, this is the
first definitive evidence that treatment with a BACE1 inhibitor has resulted in inhibition of de
novo plaque formation since 2ꢀphoton imaging enables the tracking of individual plaque growth
in a live mouse with time. This enzymatic inhibition, preventing the formation of Aβ from APP
does not rule out the possibility that other mechanisms or processes such as enhanced clearance
3
0
could be responsible for diminishment of amyloid plaques in this model. One recent report has
shown that diminishment of BACE1 activity, either genetically or pharmacologically, enhanced
ACS Paragon Plus Environment

Page 29 of 101
Journal of Medicinal Chemistry
1
2
3
clearance of axonal and myelin debris, as well as accelerated axonal regeneration following
peripheral nerve axotomy in mice, opening up the possibility of BACE1 inhibition as a
therapeutic approach for peripheral nerve damage.
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. Effect of 36 on amyloid plaqueꢀlowering efficacy in 8.5ꢀmonthꢀold PS1/APP mice.
Fluorescence intensity was quantitated from a defined cortical region (600 ꢁm x 600 ꢁm x 200
ꢁm) at baseline (6.5 months of age) and after 2 months of treatment (8.5 months of age) with
vehicle or 36. Data are the mean ± SEM, n=14; vehicle, n=6; compound 36.
ACS Paragon Plus Environment