Rational Design of Novel 1,3-Oxazine Based β-Secretase (BACE1) Inhibitors: Incorporation of a Double Bond to Reduce P-gp Efflux Leading to Robust Aβ Reduction in the Brain

Article abstract of DOI:10.1021/acs.jmedchem.8b00002

Accumulation of Aβ peptides is a hallmark of Alzheimer's disease (AD) and is considered a causal factor in the pathogenesis of AD. β-Secretase (BACE1) is a key enzyme responsible for producing Aβ peptides, and thus agents that inhibit BACE1 should be bene

Full text of DOI:10.1021/acs.jmedchem.8b00002

Subscriber access provided by Kaohsiung Medical University
Article
Rational Design of Novel 1,3-Oxazine Based #-Secretase
(BACE1) Inhibitors: Incorporation of a Double Bond to Reduce
P-gp Efflux Leading to Robust A# Reduction in the Brain
Kouki Fuchino, Yasunori Mitsuoka, Moriyasu Masui, Noriyuki Kurose, Shuhei Yoshida,
Kazuo Komano, Takahiko Yamamoto, Masayoshi Ogawa, Chie Unemura, Motoko Hosono,
Hisanoro Ito, Gaku Sakaguchi, Shigeru Ando, Shuichi Ohnishi, Yasuto Kido, Tamio Fukushima,
Hirofumi Miyajima, Shuichi Hiroyama, Kiyotaka Koyabu, Deborah Dhuyvetter, Herman
Borghys, Harrie J.M. Gijsen, Yoshinori Yamano, Yasuyoshi Iso, and Ken-ichi Kusakabe
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b00002 • Publication Date (Web): 07 May 2018
Downloaded from http://pubs.acs.org on May 7, 2018
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 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 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.
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 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Koyabu, Kiyotaka; Shionogi Pharmaceutical Research Center, Research
Laboratory for Development
Dhuyvetter, Deborah; Janssen Pharmaceutica NV
Borghys, Herman; Janssen Pharmaceutica NV
Gijsen, Harrie; Johnson&Johnson Pharmaceutical Research&Development,
Medicinal Chemistry; Janssen Pharmaceutica NV, Neuroscience Medicinal
Chemistry
Yamano, Yoshinori; Shionogi Pharmaceutical Research Center, Discovery
Research Laboratory for Core Therapeutic Areas
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
Iso, Yasuyoshi; Shionogi Pharmaceutical Research Center, Discovery
Research Laboratory for Core Therapeutic Areas
Kusakabe, Ken-ichi; Shionogi & Co., Ltd., Medicinal Chemistry
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 56
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
Rational Design of Novel 1,3ꢀOxazine Based βꢀ
Secretase (BACE1) Inhibitors: Incorporation of a
Double Bond to Reduce Pꢀgp Efflux Leading to
Robust Aβ Reduction in the Brain
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
†
,∞
†,∞
†,∞
†
†
Kouki Fuchino, Yasunori Mitsuoka, Moriyasu Masui, Noriyuki Kurose, Shuhei Yoshida,
†
†
†
†
†
Kazuo Komano, Takahiko Yamamoto, Masayoshi Ogawa, Chie Unemura, Motoko Hosono,
†
†
‡
‡
‡
Hisanori Ito, Gaku Sakaguchi, Shigeru Ando, Shuichi Ohnishi, Yasuto Kido, Tamio
‡
‡
‡
‡
║
Fukushima, Hirofumi Miyajima, Shuichi Hiroyama, Kiyotaka Koyabu, Deborah Dhuyvetter,
║
§
†
†
Herman Borghys
,
Harrie J. M. Gijsen, Yoshinori Yamano, Yasuyoshi Iso, and Kenꢀichi
*
,†
Kusakabe
†
‡
Discovery Research Laboratory for Core Therapeutic Areas, Research Laboratory for
Development, Shionogi Pharmaceutical Research Center, 1ꢀ1 Futabaꢀcho 3ꢀchome, Toyonaka,
Osaka 561ꢀ0825, Japan
§
║
Neuroscience Medicinal Chemistry, Discovery Sciences, Janssen Research & Development,
Turnhoutseweg 30, Bꢀ2340 Beerse, Belgium
ACS Paragon Plus Environment
1

Page 3 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
ABSTRACT
Accumulation of Aβ peptides is a hallmark of Alzheimer’s disease (AD) and is considered a
causal factor in the pathogenesis of AD. βꢀSecretase (BACE1) is a key enzyme responsible for
producing Aβ peptides, and thus agents that inhibit BACE1 should be beneficial for diseaseꢀ
modifying treatment of AD. Here we describe the discovery and optimization of novel oxazineꢀ
based BACE1 inhibitors by lowering amidine basicity with the incorporation of a double bond to
improve brain penetration. Starting from a 1,3ꢀdihydoꢀoxazine lead 6 identified by a hitꢀtoꢀlead
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
SAR following HTS, we adopted a pK lowering strategy to reduce the Pꢀgp efflux and the high
a
hERG potential leading to the discovery of 15 that produced significant Aβ reduction with long
duration in pharmacodynamic models and exhibited wide safety margins in cardiovascular safety
models. This compound improved the brainꢀtoꢀplasma ratio relative to 6 by reducing Pꢀgp
recognition, which was demonstrated by a Pꢀgp knockout mouse model.
INTRODUCTION
Alzheimer’s disease (AD) is the most common type of dementia. The histopathological
hallmarks of AD are neurofibrillary tangles, aggregated and hyperꢀphosphorylated tau proteins,
and amyloid plaques composed of amyloid β (Aβ) peptides of 38−43 amino acids. Patients with
AD die within 3−9 years after diagnosis. In addition to shortening of life expectancy, AD has a
significant impact on quality of life both for the patients and their family caregivers. In 2015,
over 46 million people were living with dementia worldwide, with AD being the most common
1
form, and the number is estimated to increase to 132 million by 2050. Dementia imposes an
enormous economic burden with the total estimated worldwide cost of 818 billion dollars, and
1
the estimated cost for 2018 to be a trillion dollars. In spite of these serious issues, the standard
ACS Paragon Plus Environment
2

Journal of Medicinal Chemistry
Page 4 of 56
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
therapy of Donepezil and Memantine only temporarily relieves disease symptoms. The
development of diseaseꢀmodifying drugs has thus far been unsuccessful.
Strategies targeting inhibition of Aβ peptide production should be important for diseaseꢀ
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
modification of AD. The βꢀsite amyloid precursor protein cleaving enzyme 1 (BACE1), also
known as βꢀsecretase or memapsin2, is an aspartic protease and the rateꢀlimiting enzyme in the
4
Aβ peptide production. First, the amyloid precursor protein (APP) is cleaved by BACE1 to
generate the Nꢀterminus of sAPPβ and the membraneꢀbound Cꢀterminus of C99. Second, C99 is
cleaved by γꢀsecretase to generate the Cꢀterminus of the Aβ peptides including various isoforms
of 38−43 amino acid length, of which the Aβ42 consisting of 42 amino acids is the most toxic
5
form as it readily aggregates to generate toxic oligomers, which ultimately result in the
deposition of amyloid plaques. Although γꢀsecretase is also considered to be a promising target
for reducing Aβ production, its concomitant inhibition of Notch processing has hampered the
6
development of γꢀsecretase inhibitors for clinical use. BACE1 knockout mice are viable and
3
exhibit a moderate phenotype of morphologic and functional deficit, while deletion of the
PSENꢀ1 gene, the catalytic site of γꢀsecretase, is embryonically lethal, making BACE1 a more
6
attractive target for Aβꢀlowering therapeutics.
Mutations found in APP also provide genetic evidence for the importance of BACE1 in AD. For
example, the Swedish mutation (K670N and M671L) is found near the BACE1 cleavage site and
makes APP more susceptible to cleavage by BACE1, which ultimately causes familial AD
3
(FAD). In contrast, the A673T mutation is less effectively cleaved by BACE1 and results in an
approximately 40% reduction in Aβ peptide formation leading to a protective effect against
3
,7
AD. Such evidence partly demonstrates the clinical proof of concept by inhibition of BACE1.
ACS Paragon Plus Environment
3

Page 5 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Significant efforts have been directed at exploring orally available and brain penetrant BACE1
8
,9
inhibitors since the identification of BACE1 in 1999. Initial BACE1 inhibitors were designed
based on the transition state or other aspartic proteases such as HIV protease and renin. However,
these efforts failed to identify brain penetrant inhibitors with significant in vivo efficacy. Their
high molecular weight, polar surface area, and rotatable bond number resulted in compounds
with low permeability and high Pꢀgp efflux, which ultimately limited CNS exposure. Identifying
small molecules and nonꢀpeptidomimetic BACE1 inhibitors was considered to be very
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
8
challenging due to the large size of the catalytic site in BACE1. The first examples of potent and
nonꢀpeptide inhibitors with amidineꢀbased scaffolds were reported by scientists at Scheringꢀ
8
,10
Plough and Wyeth. Since their discovery, many pharmaceutical companies pursued analogs.
Although early amidineꢀbased inhibitors retaining high Pꢀgp efflux had low potency in vivo, the
challenge was addressed successfully by utilizing structureꢀbased design and controlling
physicochemical properties, leading to compounds with robust Aβ reduction in preclinical
8
animal models. Finally, minimizing the risks associated with the high basicity and lipophilicity
observed with these analogs, such as high hERG activity and CYP2D6 inhibition, culminated in
1
1
12
the discovery of clinical BACE1 inhibitors including 1 (LYꢀ2811376), LYꢀ2886721, 2 (MKꢀ
1
3
14
15
16a
8
931, verubecestat), 3 (Eꢀ2609, elenbecestat), RGꢀ7129, lanabecestat (AZDꢀ3293),
16b
17
CNP520, and JNJꢀ54861911. Unfortunately, phase I/II studies of several of these compounds
1
8
were discontinued, most likely due to offꢀtarget toxicity. Therefore, identifying BACE1
inhibitors in diverse structural classes was considered to be worthwhile.
ACS Paragon Plus Environment
4

Journal of Medicinal Chemistry
Page 6 of 56
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
Chart 1. AmidineꢀBased Clinical 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
In 2003, our early research started with a highꢀthroughput screening (HTS) campaign utilizing a
cellular Aβ production assay, which successfully identified a thiazineꢀbased hit compound 4
without cellular toxicity (Chart 2). Our early efforts revealed that incorporation of amide groups
1
9
on the phenyl ring in 4 significantly improved potency leading to the discovery of 5.
Furthermore, replacement of the sulfur in 5 with an oxygen atom followed by an amide screen
led to 6, which exhibited comparable cellular potency, while reduced activity was observed in
the biochemical assay relative to 5. In parallel with our efforts, Roche also identified the same hit
2
0
compound of 4 in their HTS campaign. Their efforts utilizing fluorine scans by reducing
basicity (pK ) on amidine moieties culminated in the CF ꢀsubstitued dihydroꢀoxazine 7 with in
a
3
2
1
vivo efficacy (Chart 3). In spite of the excellent profile described for 7, the Roche clinical
2
2
compound RGꢀ7129, an analog of 7, was terminated in Phase I.
Chart 2. HTS Hit 4 and Initial Leads 5 and 6
H N
S
H N
S
F
H N
O
F
2
2
2
Cl
NC
N
N
N
H
N
H
N
N
N
O
O
O
4
5
6
BACE1 IC : 73 ꢀM
BACE1 IC : 6.7 nM
BACE1 IC : 118 nM
Cell A IC50: 3.7 nM
5
0
50
50
Cell A IC : 2.6 ꢀM
Cell A IC : 0.84 nM
5
0
50
ACS Paragon Plus Environment
5

Page 7 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Controlling pK is a wellꢀknown approach for reducing Pꢀgp efflux as well as mitigating hERG
a
2
3
inhibitory activity, particularly in exploring BACE1 inhibitors as with compounds 7, 8, and 9
9
h,21,24
(Chart 3).
For instance, to lower the pK of amidine or amine moieties that interact with the
a
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
catalytic aspartates in BACE1, incorporation of electron withdrawing groups such as fluorine,
13,20,21,24
oxygen, carbonyl, and sulfone has been successfully applied.
Our aim in this work was to
improve the Pꢀgp and the hERG liabilities observed with 6 by lowering the pK . We describe an
a
alternative approach to reducing the pK of the dihydroꢀoxazine 6 by incorporating a double
a
bond, which successfully lowered the pK by 2 log units, leading to oxazine 11 with reduced Pꢀ
a
gp efflux. Further modification by fineꢀtuning the pK led to the discovery of compound 15 that
a
showed robust Aβ reduction in mouse and dog, as well as good cardiovascular safety margins in
guinea pig and dog over the efficacy level (EC ) in dog.
5
0
Chart 3. BACE1 Inhibitors 7, 8, and 9 Identified Utilizing a pK ꢀLowering Strategy
a
RESULTS AND DISCUSSION
Design of Novel 1,3ꢀOxazine BACE1 Inhibitors by Incorporating a Double Bond. The initial
lead 6 displayed moderate to good Aβ inhibitory activity in cells (IC = 3.7 nM); however, it
5
0
exhibited a high Pꢀgp efflux ratio of 36 in MDCK cells resulting in a low total brainꢀtoꢀplasma
ratio (B/P or K ) of 0.29 as well as a hERG inhibitory activity of 78% at 5 ꢁM (Table 1). Such
p
profiles resulted in a lack of significant Aβ reduction in mouse at 10 mg/kg and could raise
ACS Paragon Plus Environment
6

Journal of Medicinal Chemistry
Page 8 of 56
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
cardiovascular concerns if used clinically. Therefore, we initiated our efforts to improve the Pꢀgp
and hERG values in order to obtain centrally active BACE1 inhibitors with acceptable
cardiovascular safety.
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
For central nervous system (CNS) drugs, optimal pK ranges of basic amines (the conjugate acids
a
23
of amines), such as 7.5 < pK < 10.5 or pK < 8, were proposed. According to these analyses,
a
a
lowering pK achieved a decreased likelihood of Pꢀgp recognition and hERG inhibition, while
a
the lower limits were estimated based on data sets including compounds targeting monoamine
23
receptors or transporters. Considering this, we decided it would be important to determine the
lower limit of pK utilizing BACE1 inhibitors alone, due to the difference in optimal pK ranges
a
a
for retaining potency among these targets. Indeed, amine protonation is critical for inhibiting
BACE1 enzyme at a given pH in the assay. Our BACE1 biochemical assay was conducted at pH
5
.0, while our cellular assay was at the physiological pH of 7.4. Because the cellular potency
reflects Aβ reduction in vivo, understanding the lower limit of amidine pK in our cellular assay
a
was important for designing BACE1 inhibitors. For the analysis, we utilized our proprietary
BACE1 inhibitors at that time, most of which included thiazine derivatives reported previously.
25
A good correlation between biochemical and cellular IC values was observed except for
50
compounds with a pK <6.5 and a biochemical IC <100 nM. The discrepancy came from the
a
50
difference in pH used in our biochemical and cellular assays of 5.0 and 7.4, respectively. A box
plot analysis using compounds with a biochemical IC <100 nM clearly indicated that
5
0
compounds with a pK <6.5 reduced the cellular activities regardless of having good biochemical
a
potency data (Supporting Information, Figure S1). Given that Pꢀgp and the hERG channel
recognize the ionization state at the physiological pH of 7.4, the optimal upper limit was
ACS Paragon Plus Environment
7

Page 9 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
considered to be 7.4. The analysis and hypothesis led to a proposed guideline of 6.5 < pK < 7.4
a
for our research.
To develop BACE1 inhibitors meeting our pK guideline, we designed the oxazine head group B
a
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
inserting a double bond into the dihydroꢀoxazine head group A. Because good alignment
between the measured and ACD predicted pK values in A was observed, we utilized ACD
a
26
software to predict the pK of B. The result was an acceptable value of 7.0, most likely due to
a
delocalization of the lone pairs on the nitrogen atoms, which initiated our efforts to synthesize
the oxazine analogs.
H N
O
F
H N
O
F
2
2
N
N
A
B
a
ACD pK :
10.1
9.8
7.0
a
pK measured:
this work
a
a
Figure 1. Design of novel oxazine B from dihydroꢀoxazine A. The pK values were calculated
a
2
6
with ACD/Percepta as the imine tautomer.
Optimization of the 1,3ꢀOxazine Head Group. The designed compound 11 showed a reduced
pK of 7.7 (ꢂpK of 2.1 over 6) and an attenuated Pꢀgp efflux ratio of 7.5, while exhibiting a
a
a
slightly increased hERG inhibitory activity, 2ꢀfold decrease in potency in the biochemical and
cellular assays, and a significantly increased LogD when compared with 6 (Table 1).
Additionally, we observed a discrepancy between the measured and predicted pK s (7.7 vs 7.0)
a
lying outside the range of the guideline for pK . To mitigate hERG inhibitory activity as well as
a
ACS Paragon Plus Environment
8

Journal of Medicinal Chemistry
Page 10 of 56
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
improve potency, we investigated the incorporation of substituents on the double bond in 11.
Compound 13, possessing a methyl group at the 6ꢀposition, was 3 times more potent than 11 in
the biochemical assay, while moving the methyl group to the 5ꢀposition led to 12 with reduced
potency, indicating that the 6ꢀmethyl oxazine 13 could serve as a key scaffold for fineꢀtuning the
pK to within the optimal range. To further reduce the pK , we incorporated a fluorine atom at the
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
a
a
4
ꢀmethyl group, leading to compound 14 with improved hERG liability. However, despite
biochemical activity comparable to 13, decreased cellular potency was observed. This result can
be explained by the pK of 6.3, which violates the lower limit of our analysis. Alternatively,
a
substitution of the 6ꢀmethyl group in 13 with a fluorine led to compound 15 with an optimal pKa
of 6.9, which retained the potency associated with slightly reduced hERG inhibition relative to
the 6ꢀmethyl 13. Replacement of the 4ꢀmethyl group with a cycloplopyl group (16) was not
tolerated. Incorporation of methoxy as an alternative electron withdrawing group (17) led to a
pK of 7.4. Although 17 had a pK value in the acceptable range, it showed reduced cellular
a
a
potency and was found to be a strong Pꢀgp substrate. On the other hand, its hERG potency was
better than the fluorine analog 15 despite its higher pK . Although both the fluorine in 15 and the
a
methoxy in 17 function as electronꢀwithdrawing groups, the difference in hydrogen bonding
capacity and lipophilicity between 15 and 17 can explain these discrepancies. The methoxy in 17
may have a favorable interaction with Pꢀgp relative to the fluorine in 15, most likely because of
its hydrogen bond accepting feature; in contrast, the more lipophilic fluorine in 15 may have
better affinity for the hERG channel than the methoxy in 17. Introduction of an additional
fluorine in 15 led to 6ꢀdifluromethyl 18 with a pK of 6.4. Like compound 14, its reduced pK
a
a
was associated with impaired cellular potency. Consistent with the above analysis, such a low
pK value was translated into improvements in Pꢀgp and hERG levels. Except for 18 with a lower
a
ACS Paragon Plus Environment
9

Page 11 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
pK of 6.5, the compounds in Table 1 were more potent in the cellular assay than in the
a
biochemical assay. This could be explained by accumulation of basic inhibitors in the acidic
4a
endosomes, where BACE1 resides and coꢀlocalizes with APP, and the pH is maintained in a
2
7
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
range of 4.5 to 6.5.
The rat pharmacokinetic (PK) profiles of selected analogs 6, 14, 15, and 17 using our preliminary
screening model (cassette dosing, n = 2) were assessed (Table 2). The initial lead 8 was stable in
rat liver microsomes (RLM, 100% remaining after 30 min incubation), while a high fraction
unbound in rat serum (f in rat) of 0.41 resulted in a moderate to high clearance of 37
u,s
mL/min/kg. Although 6 had a high Vd of 6.1 L/kg, a low total brainꢀtoꢀplasma (B/P) ratio of
ss
0.29, and an unbound brainꢀtoꢀplasma ratio (Kp,uu) of 0.057 were observed. Compounds 14 and
1
5 with reduced pK values exhibited improved B/P ratios of 1.30 and 0.90 (Kp,uus of 0.52 and
a
0
.37), respectively, which can be explained by their reduced Pꢀgp efflux ratios. Likewise, the
reduced brain penetration of the methoxy analog 17 (B/P = 0.19) can be attributed to its high Pꢀ
gp efflux. Despite their varied metabolic stability in RLM, all the compounds in Table 2 had
favorable oral exposures. Taken together, 6ꢀfluoromethyl 15 was notable in that it displayed a
pronounced improvement in Pꢀgp efflux resulting in a good B/P ratio as well as retaining cellular
potency, although the hERG potency still required further mitigation.
Table 1. Exploration of the Oxazine Head Group
ACS Paragon Plus Environment
10

Journal of Medicinal Chemistry
Page 12 of 56
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
ACS Paragon Plus Environment
11

Page 13 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
b
Values represent the mean values of at least two determinations. Biochemical homogeneous
c
timeꢀresolved fluorescence (HTRF)ꢀbased assay. IC determined by measuring the levels of
secreted Aβ40 in human APPꢀtransfected human neuroblastoma (SHꢀSY5Y) cells via an HTRFꢀ
based assay. Efflux ratio measured in MDCK cells transfected with human MDR1 at Absorption
Systems. % Inhibition at 5 ꢁM measured in HEK293 or CHO cells transfected with hERG
channels using an automated patch clamp system. LogD determined in 1ꢀoctanol/phosphate
buffer at pH 7.4. pK determined by capillary electrophoresis (CE).
5
0
d
e
f
g
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
a
Table 2. Pharmacokinetic Properties of Compounds 6, 14, 15 and 17 in SpragueꢀDawley
Rat Using a Cassette Method
a
b
Dosed as a suspension of test compounds in 0.5% methylcellulose (MC, 400 cP). Dosed as a
c
solution of test compounds in N,Nꢀdimethylacetamide (DMA)/propylene glycol (PG) = 1/1. %
remaining in rat liver microsomes after 30 min incubation. f = Fraction unbound in rat serum.
fu,b = Fraction unbound in rat brain. NT = not tested. Total clearance. Volume of distribution at
steady state. Total brainꢀtoꢀplasma ratio measured at 0.5 h. Unbound brainꢀtoꢀplasma ratio
measured at 0.5 h. ND = not determined. Plasma area under the concentrationꢀtime curve.
Maximal plasma concentration. Oral bioavailability.
d
u,s
e
f
g
h
i
j
k
Optimization of the Tail Group in 15. In an effort to improve the hERG inhibitory activity of
5, we focused on optimizing the tail group of R (Table 3). In addition to its suboptimal hERG
1
profile, 15 was found to be somewhat unstable in rat plasma, relative to other compounds
synthesized at our laboratory. Although this instability may be limited to the rat due to the highly
28
expressed profile of carboxylesterases (CES) in the plasma compared to nonꢀrodents, we first
investigated compounds with substituents at the orthoꢀposition of the amide group such as 20, 21,
ACS Paragon Plus Environment
12

Journal of Medicinal Chemistry
Page 14 of 56
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
and 22, in order to prevent hydrolysis by CES. As expected, these compounds were found to
improve rat plasma t1/2 to more than 50 h. Interestingly, incorporating the orthoꢀsubstituents also
reduced the hERG inhibitory activity relative to 15, particularly in orthoꢀchloro 21 and orthoꢀ
methoxy 22 (30% and 22% inhibition, respectively), while exhibiting an increased Pꢀgp efflux
which would lead to reduced brain penetration. Although the methyl 20 provided reduced hERG
inhibition (50%) and a reasonable B/P ratio (0.62), decreased human liver microsomal (HLM)
stability was observed compared to 15 (63% vs 88% remaining).
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
For further improvements in hERG as well as retaining acceptable ADMET profiles, a variety of
6
ꢀmembered heterocycles was explored. Among them, pyrazine derivatives such as 23−26 were
found to reduce hERG inhibitory activity. Although the methoxy pyrazine 23 had a reduced
hERG inhibition of 30% at 5 ꢁM, it was associated with decreased plasma stability (t1/2 of 14.3
h) and poor RLM stability (13% remaining after 30 min) resulting in a high CL value of 64.9
ml/min/kg. Replacement of the methoxy group in 23 with difluoromethyl (24) or fluoromethoxy
(25) provided molecules with improved RLM stability (44% and 52% remaining after 30 min,
respectively), while these were still unstable in rat plasma (plasma t1/2 = 9.2 h and 10.5 h,
respectively). Furthermore, such analogs (23−25) displayed only modest potency in biochemical
and cellular assays (cellular IC = 22 nM, 72 nM, and 31 nM, respectively). To mitigate these
5
0
problems observed with the pyrazine tails, a small set of substituents, such as methyl (data not
shown), methoxy (data not shown), and amino, was incorporated on the pyrazine in 25 at the
orthoꢀposition of the amide group leading to aminopyrazine 26. Indeed, 26 was 5ꢀfold more
potent than the corresponding nonꢀamino 25 (cellular IC of 6.6 nM in 26), and plasma stability
5
0
was also improved relative to the nonꢀsubstituted 23, 24, and 25, but was comparable to the
cyanopyridine 15 (plasma t1/2 of 22.3 h in 26). Despite adding the polar amino group, an
ACS Paragon Plus Environment
13

Page 15 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
increased LogD was unexpectedly observed for 26 in comparison to the nonꢀamino 25, resulting
in reduced metabolic stability in RLM and HLM (35% and 67% remaining after 30 min,
respectively). However, although introduction of NH increased the hydrogen bond count, 26
2
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
provided a comparable Pꢀgp efflux to the nonꢀamino 25, very likely due to the intramolecular
hydrogen bond of the amide carbonyl with the NH . We postulated that the improved potency
2
observed in 26 might come from the increased planarity between the amide and the pyrazine ring
derived from the intramolecular hydrogen bond of the carbonyl with the amino group. Therefore,
we hypothesized that reducing the ring size, which can maintain planarity and decrease
lipophilicity, could lead to retaining potency as well as improving metabolic stability. Indeed,
compounds with an oxazole ring such as 27 and 28 had a comparable cellular activity to the
aminopyriazine 26. Like the pyrazines, changes to the isoxazoles retained the low hERG
inhibitory potential. Unfortunately, despite their reduced LogD values relative to 26, the
isoxazoles 27 and 28 still displayed suboptimal metabolic stability in RLM (57% and 39%
remaining after 30 min, respectively). Furthermore, this modification was associated with
increased Pꢀgp recognition resulting in poor brain penetration. In the end, although the
compounds in Table 2 did lower hERG inhibition compared to cyanopyridine 15, the
improvement came at the cost of reduced potency, low metabolic stability, or high Pꢀgp
recognition. The balanced profile of 15 led to its being subjected to further study.
Table 3. Exploration of the Oxazine Head Group
ACS Paragon Plus Environment
14

Journal of Medicinal Chemistry
Page 16 of 56
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
ACS Paragon Plus Environment
15

Page 17 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
b
Values represent the mean values of at least two determinations. Biochemical homogeneous
c
timeꢀresolved fluorescence (HTRF)ꢀbased assay. IC determined by measuring the levels of
secreted Aβ40 in human APPꢀtransfected human neuroblastoma (SHꢀSY5Y) cells via an HTRFꢀ
based assay. Efflux ratio measured in MDCK cells transfected with human MDR1 at Absorption
Systems. % inhibition at 5 ꢁM measured in CHO cells transfected with hERG channels using an
automated patch clamp system. % Remaining after 30 min incubation with rat liver (RLM) and
human liver microsomes (HLM). SpragueꢀDawley rats. Total clearance dosed iv at 0.5 mg/kg
as a solution in DMA/PG = 1/1 (n = 2). Plasma halfꢀlife in vitro. Total brainꢀtoꢀplasma ratio
determined by iv dose at 0.5 h. LogD determined in 1ꢀoctanol/phosphate buffer at pH 7.4.
5
0
d
e
f
g
h
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
i
j
k
Cocrystal Structure of 15 Bound to BACE1. The Xꢀray structure of 15 complexed with human
BACE1 was determined at a resolution of 2.30 Å (Figure 2). Compound 15 bounds to a flapꢀ
closed BACE1 structure, in which the Tyr71 side chain hydroxyl on the flap forms a hydrogenꢀ
8,9
bond interaction with the NH of the Trp76 side chain. Like other amidineꢀbased inhibitors, the
amidine moiety of 15 interacts with the two catalytic aspartate residues 32 and 228. The two
aromatic rings of the fluorophenyl and the cyanopyridine occupy the S1 and S3 pockets of the
enzyme, respectively. In addition, the fluorine on the phenyl forms a van der Waals contact with
the side chain of Tyr71. Indeed, the corresponding nonꢀfluorine analog was 2−3 times less potent
than 15 in the biochemical assay (data not shown) indicating the importance of the fluorine.
Finally, the amide NH hydrogen forms a hydrogen bond with the backbone carbonyl oxygen of
Gly230 as well as with the pyridine nitrogen in 15.
ACS Paragon Plus Environment
16

Journal of Medicinal Chemistry
Page 18 of 56
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
Figure 2. Cocrystal structure of the oxazine 15 bound to the active site of human BACE1
determined at 2.30 Å resolution (PDB: 5YGY). Purple mesh shows |Fo| ꢀ |Fc| map contoured at
2
.0 σ in which ligand coordinates are omitted from the calculation. Key residues of the catalytic
residues Asp228 and Asp32 (green), Tyr71 (purple) on the Flap region, and Gly230 (gray) are
shown in stick style. Compound 15 is shown in ballꢀandꢀstick style with carbon atoms colored
yellow. Nitrogen, oxygen, and fluorine atoms are colored blue, red, and sky blue, respectively.
Dashed lines represent hydrogen bond interactions between 15 and BACE1.
In vitro ADMET Profiles of 15. The free base of compound 15 formed a crystalline powder
that had good aqueous thermodynamic solubility of 3220 ꢁg/mL and 28.7 ꢁg/mL at pH 1.2 and
6
.8, respectively. Consistent with the favorable LogD at pH 7.4 of 2.3, the permeability in wildꢀ
ꢀ6
type MDCK cells was in an acceptable range for CNS drugs (Papp = 20.4 × 10 cm/s). Prior to
PK studies in multiple species, liver microsomal stability and the fraction unbound in serum (fu,s)
were determined. Like the stability in HLM, compound 15 was found to be stable in mouse, dog,
and monkey microsomes (69%, 87%, and 79% remaining after 30 min, respectively), and its fu,s
ACS Paragon Plus Environment
17

Page 19 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
was moderate to high in rat, mouse, dog, monkey, and guineaꢀpig (0.16, 0.30, 0.11, 0.22, and
0
.42, respectively). Brain tissue binding in rat was found to be moderate (f = 0.066) as shown
u,b
in Table 2. Compound 15 was also evaluated for its ability to inhibit a panel of cytochrome
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
P450s and exhibited IC s of more than 20 ꢁM over 1A2, 2C9, 2C19, 2D6 and 3A4. Additionally,
5
0
1
5 did not show timeꢀdependent inhibition of CYP3A4 up to 20 ꢁM. The hERG IC value using
50
a manual patch clamp method was determined to be 2.0 ꢁM. The close analog 4ꢀfluoromethyl 14
exhibited hERG IC of 3.9 ꢁM reflecting the difference in % inhibition at 5 ꢁM. Compound 15
5
0
was tested against other aspartyl proteases, cathepsin D (CatD) and BACE2. Although 15
showed limited selectivity over BACE2 (IC = 86.3 nM), no significant inhibition was observed
5
0
in CatD (IC >100 ꢁM). Solid state stability testing for 15 was conducted at 40 °C/89% relative
5
0
humidity (RH), where a degradation product increased by 0.06%, indicating that 15 was almost
stable. Because 15 has a potential to generate anilines as metabolites (a structural alert), the
Ames test was conducted to estimate the genetic toxicity of this series. Five mutant strains of
TA100, TA1535, TA1537, TA98, and WP2uvrA were used in the presence and absence of the
rat S9 fraction. Negative results were observed for 15 at doses up to 5 mg/plate. Also, the
29
corresponding aniline dihydrochloride was found to be negative in the Ames test.
Brain Penetration in WildꢀType and Mdr1a Knockout Mouse. From exploration of the head
group, compound 15 provided lower efflux ratios in MDCK cells transfected with human MDR1
relative to the initial lead 6. To investigate the potential impact of Pꢀgp in vivo, we utilized
mdr1a knockout (KO) mouse to determine B/P ratios in knockout versus wildꢀtype (WT). For
comparison, oxazine 14 with a lower Pꢀgp efflux of 9.0 and dihydroꢀoxazine 6 with a higher Pꢀ
gp efflux of 36 were included in the tests. The compounds were administered orally at doses of 2
and 10 mg/kg (2 mg/kg for 6 and 15, 10 mg/kg for 14). Clearly, the Pꢀgp efflux ratios were
ACS Paragon Plus Environment
18

Journal of Medicinal Chemistry
Page 20 of 56
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
translated into the KO/WT ratios of the B/P values. As shown in Table 4, the B/P ratios in WT
mouse were comparable to those in rat in Table 2, while increased B/P ratios were observed in Pꢀ
gp KO mouse, indicating that these compounds were Pꢀgp substrates in mouse. Despite the low
ꢀ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
lipophilicity of the initial lead 6 (LogD = 0.024) and the low permeability (Papp = 0.32 × 10
cm/s), the B/P ratio in the KO mouse significantly increased to 4.7 (from 0.15 in the WT
counterpart), which highlighted the important role of Pꢀgp in brain penetration. The B/P ratios in
the KO mouse for 14 and 15 were 2.5 and 4.4 resulting in KO/WT ratios of 2.5 and 3.6,
respectively, demonstrating the improvement in Pꢀgp efflux for 14 and 15 in vivo relative to 6
(KO/WT = 32). According to the analysis at Pfizer for successful CNS drugs using Pꢀgp KO
3
0
mouse, KO/WT ratios of 2 to 3 appeared to be in an acceptable range. Thus, although both 14
and 15 were still found to be Pꢀgp substrates in mouse, the magnitude of the potential clinical
effects for 14 and 15 may be low.
Table 4. Brain Penetrations of 6, 14, and 15 in Wildꢀtype (C57BL/6J) and mdr1a (ꢀ/ꢀ)
Mouse.
a
b
Dosed as a suspension of test compounds in 0.5% methylcellulose (MC). Efflux ratio
c
measured in MDCK cells transfected with human MDR1 at Absorption Systems. KO = mdr1a (ꢀ
ꢀ) mouse; WT = wildꢀtype (C57BL/6J) mouse (vehicle = 0.5% MC). Brain concentration at 2 h
d
/
ACS Paragon Plus Environment
19

Page 21 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
e
f
time point. Plasma concentration at 2 h time point. Brainꢀtoꢀplasma total concentration ratio
g
(
C /C ). KO/WT = (B/P in KO)/(B/P in WT).
b p
NeuroꢀPharmacokinetic Studies of 15 in Multiple Species. Investigations of pharmacokinetic
properties of 15 were conducted using SpragueꢀDawley rats, beagle dogs, and cynomolgus
monkeys. As shown in Table 5, exposure and plasma levels were high in rats after oral
administration of 15 at doses of 1, 3, and 10 mg/kg. At 1 mg/kg, the maximum concentration
(Cmax) and the area under the curve (AUC) were at similar levels to those in the cassette method
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
(
Table 5 vs Table 2). Also, the intravenous administration at 2 mg/kg provided comparable
results to those from the cassette method of 0.5 mg/kg. The doseꢀproportionality of 15 was
maintained between the given doses. Compound 15 showed high exposure, moderate clearance,
and moderate to high volume of distribution. To assess neuroꢀpharmacokinetic profiles of 15,
compound levels in the brain and CSF were also measured at the po dose of 3 mg/kg (Table 6).
Total B/P ratios (C /C , 0.70 to 0.83) were similar to unbound brainꢀtoꢀplasma ratios based on
b
p
CSF levels (CCSF/C , 0.51 to 0.78), while those based on unbound brain levels (C /C ; C =
p,u
b,u p,u
b,u
C × f ; 0.29 to 0.34) decreased 2ꢀfold relative to the total B/P and C /C ratios. The higher
b
u,b
CSF p,u
CCSF/C ratio could be rationalized by the observed Pꢀgp effect in mouse, because Pꢀgp works
p,u
2
3
as an active transporter at the bloodꢀCSF barrier.
Table 7 summarizes the pharmacokinetic data for beagle dog and cynomolgus monkey. In dog,
5 displayed low clearance, moderate to high Vd , a decent exposure level (AUC), and high
1
ss
bioavailability following an iv dose of 1 mg/kg and a po dose of 0.75 mg/kg (day 1). After
repeated daily po doses, slight increases in Cmax and AUC were seen on day 7. In contrast to dog,
a higher clearance and decreased bioavailability were observed in monkey, most likely due to
ACS Paragon Plus Environment
20

Journal of Medicinal Chemistry
Page 22 of 56
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
decreased microsomal stability and higher f as shown in Table 7. To compare rat and dog CNS
u,s
availability, compound levels in CSF were measured at 10 and 24 h after dosing on day 7, where
comparable CCSF/C ratios were obtained between rat and dog. On the basis of the good brain
p,u
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
penetration observed, 15 was subjected to pharmacokinetic/pharmacodynamics (PK/PD) study.
Table 5. Pharmacokinetic Properties of 15 at Multiple Doses in Rat
a
b
c
Dosed as a suspension of 15 in 0.5% MC. Maximal plasma concentration. Plasma area under
d
e
f
the curve (po). Oral bioavailability. Dosed as a solution of 15 in 0.5% MC. Total clearance.
Volume of distribution at steady state. Halfꢀlife (iv). Plasma area under the curve (iv).
g
h
i
Table 6. CNS Availability Data for 15 in Rat
ACS Paragon Plus Environment
21

Page 23 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
b
c
Dosed as a suspension of 15 in 0.5% MC. Plasma concentration. Brain concentration.
d
d
e
Unbound brain concentration (Cb,u = C × f ). Cerebrospinal fluid (CSF) concentration. Total
brainꢀtoꢀplasma ratio (B/P or K ). Unbound brainꢀtoꢀplasma ratio (Kp,uu). Cb,u = C × f
unbound brain fraction in rat). Cp,u = C × f (unbound plasma fraction in rat). CSFꢀtoꢀ
unbound plasma ratio (Kp,uu (CSF)).
b
u,b
f
p
b
u,b
g
(
p u,s
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
Table 7. Pharmacokinetic Properties of 15 in Beagle Dog and Cynomolgus Monkey
a
b
%
remaining in dog or monkey liver microsomes after 30 min incubation. Fraction unbound
c
in dog or monkey serum. Dosed as a solution of 15 in DMA/PG/ hydroxypropylꢀβꢀcyclodextrin
HPBCD) = 1/3/1 for dog and in HPBCD for monkey. NT = not tested. Total clearance.
Volume of distribution at steady state. Halfꢀlife (iv). Dosed as a suspension of 15 in 0.5% MC.
Plasma area under the curve (po). Maximal plasma concentration. Oral bioavailability. ND =
d
(
e
f
g
h
i
j
k
not determined. After 7 days of once daily dosing, plasma samples were collected for up to 24 h.
Table 8. CNS Availability Data For 15 in Dog
a
After 7 days of once daily dosing (po, 0.5% MC), cisternal CSF samples were collected from
b
anesthetized dogs at 10 and 24 h time points. Plasma concentration measured in the experiment
in Table 7. Unbound plasma concentration (C = C × f (unbound plasma fraction in dog)).
Cerebrospinal fluid (CSF) concentration. CSFꢀtoꢀunbound plasma ratio (Kp,uu (CSF)).
c
p,u
p
u,s
d
e
ACS Paragon Plus Environment
22

Journal of Medicinal Chemistry
Page 24 of 56
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/PD Studies of 15 in Mouse and Dog. To assess BACE1 inhibition in vivo, we used Crlj:CD
3
1
(
ICR) mouse and beagle dog. As shown in Figure 3, 15 demonstrated robust and sustained
reduction of total Aβ levels in the brain over 8 h following a po dose of 10 mg/kg. The maximum
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
Aβ reduction of 69% was observed at 4 h, where the brain level reached 295 ng/mL (C = 51
b,u
nM), which was 14ꢀfold higher than a cellular IC value of 3.6 nM. Even at 8 h, 15 achieved
5
0
1
.9ꢀfold higher unbound brain level than the IC , explaining the sustained Aβ reduction in
50
mouse.
Figure 3. Total Aβ reduction in mouse after an oral dose of 15 at 10 mg/kg (n = 4). Compound
1
5 was dosed as a solution of 20% 2ꢀhydroxypropylꢀβꢀcyclodextrin (HPBCD).
Dog models for assessing the potential of inhibiting Aβ production are useful intermediates
between models for mice and humans, because pathological changes of Aβ production in dogs
are similar to those in humans, and dogs also develop Aβ plaques thereby being considered as a
32
natural model of ageꢀdependent cognitive dysfunction. To determine the plasma level that can
achieve Aβ reduction by 50% (EC ), a PK/PD study in dog was conducted at multiple doses of
5
0
ACS Paragon Plus Environment
23

Page 25 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
32a
0
.31, 1.25, and 5 mg/kg. The plasma samples were taken at 0.5, 1, 2, 4, 8, and 25 h after
dosing, and the CSF was sampled in conscious dogs from the lateral ventricle before dosing and
at 4, 8, and 25 h after dosing. As shown in Figure 4, 15 demonstrated a doseꢀdependent and
sustained decrease in CSF Aβ 40 and 42. Importantly, even at 1.25 mg/kg, a robust CSF Aβ
reduction of more than 70% was achieved at 8 h. Finally, the EC value in plasma was
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
determined to be 62 ng/mL (EC50, unbound = 18 nM); the EC value in CSF was estimated to be 10
50
nM using the CCSF/C ratios of 0.56 in Table 8, which was 3ꢀfold higher than the cellular IC
p,u
50
value of 3.6 nM. For comparison, a dog PK/PD study for 14 was also done, and the EC value
5
0
for Aβ42 was found to be 145 ng/mL (EC50, unbound = 45 nM, f in dog = 0.12), reflecting the
u,s
difference in the cellular IC values between 14 and 15 of 17 nM and 3.6 nM, respectively.
5
0
Figure 4. CSF Aβ1−40, and Aβ1−42 reduction in beagle dog after oral doses of 15 at 0.31, 1.25,
and 5 mg/kg (n = 4). Compound 15 was dosed as a solution of 20% HPBCD.
Cardiovascular Safety of 15 in Anesthetized GuineaꢀPig and Conscious Dog. To investigate
the cardiovascular profile of 15, we first utilized an anesthetized guineaꢀpig model. Increasing
doses of 15 (3, 10, and 30 mg/kg) were administered intravenously over a period of 10 min at 30
min intervals to urethaneꢀanesthetized guineaꢀpigs. The average unbound plasma levels of 15,
ACS Paragon Plus Environment
24

Journal of Medicinal Chemistry
Page 26 of 56
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
measured at the end of the intravenous infusions of 3, 10, and 30 mg/kg, were 0.907, 4.31, and
1
3.4 ꢁM, respectively. Compared to the vehicle, 15 significantly increased the duration of QTc
interval (13.9%) and decreased the heart rate at 30 mg/kg. The unbound noꢀobservedꢀadverseꢀ
effect level (NOAEL) for cardiovascular safety in guineaꢀpig is 4.31ꢁM, and the safety margin
over the dog EC50,unbound was calculated to be 239ꢀfold—wide safety margin.
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
The cardiovascular effects of 15 in conscious dogs were then studied. On the basis of the dog PK
study in Table 7 and the NOAEL in guineaꢀpigs, 15 was orally administered at doses of 30 and
1
00 mg/kg. No significant effects on arterial blood pressure, heart rate, PR interval, and QTc
interval were observed at 30 mg/kg, while statistically significant increases in heart rate and QTc
interval (16.1%) were observed at 100 mg/kg. The unbound maximum concentrations
(Cmax,unbound) for 30 and 100 mg/kg were 1.40 and 3.21 ꢁM, respectively; 15 had a NOAEL of
1
.40 ꢁM, which was close to the hERG IC value of 2.0 ꢁM, establishing the safety margin of
5
0
78ꢀfold over dog EC50,unbound. Although the margin was lower than observed for guinea pigs, it
was still in an acceptable range.
CHEMISTRY
The synthesis of oxazine 11 is outlined in Scheme 1. Condensation of acetophenone 30 with (R)ꢀ
3
3
tertꢀbutysulfinamide using Ti(OEt) afforded ketimine 31. This was followed by reduction of
4
the nitro group and trifluoroacetylation with trifluoroacetic anhydride (TFAA) to provide 32.
Addition of allyl Grignard reagent to the ketimine 32 afforded compound 33 as a single
diastereomer, which was followed by ozonolysis to give aldehyde 34. The tertꢀbutylsulfinyl
group in 34 was removed with HCl in MeOH to give acetal 35, which was then reacted with
triphosgene followed by diallylamine, yielding urea 36. The oxazine ring was formed using
ACS Paragon Plus Environment
25

Page 27 of 56
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
34
Burgess reagent in the presence of pyridinium pꢀtoluenesulfonate (PPTS) to provide 37.
Deprotection of the trifluoroacetyl group in 37 and amide coupling with 5ꢀcyanopicolinic acid
followed by deprotection of the diallyl group in 39 in the presence of palladium catalyst afforded
the oxazine 11. The stereochemistry of 11 was confirmed by single crystal Xꢀray analysis of
compound 18, which was synthesized via the intermediate 33 (see Supporting Information).
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
a
Scheme 1. Synthesis of Oxazine 11
a
Reagents and conditions: (a) (R)ꢀ2ꢀmethylpropaneꢀ2ꢀsulfinamide, Ti(OEt) , THF, 65 °C, 65%;
4
(
b) (i) Fe, NH Cl, toluene−H O, 80 °C, (ii) TFAA, Et N, THF, −25 °C, 78%; (c) allylmagnesium
4 2 3
bromide, THF, −78 °C, 40%; (d) O , DCM, −78 °C, 88%; (e) 2 N HCl in MeOH, rt, 56%; (f)
3
triphosgene, Et N, diallylamine, H O−EtOAc, 0 °C, 82%; (g) (i) aq H SO , acetone, (ii) Burgess
3
2
2
4
reagent, PPTS, THF, reflux, 24%; (h) K CO , THF−MeOH−H O, 40 °C, 87%; (i) EDC, HOBt,
2
3
2
ACS Paragon Plus Environment
26

Journal of Medicinal Chemistry
Page 28 of 56
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
DMAP, 5ꢀcyanopicolinic acid, DMF, rt, 92%; (j) 1,3ꢀdimethylbarbituric acid, Pd(PPh ) , DCM,
reflux, 58%.
3
4
The 6ꢀfluoromethyl substituted oxazines 15 and 20‒28 were synthesized via the route outlined in
Scheme 2. Sulfinyl ketimine 40, prepared from the commercially available 2ꢀfluoroacetophenone,
was reacted with allyl Grignard reagent to provide compound 41 as a single diastereomer, which
was followed by cleavage of the sulfinyl group to afford allyl amine 42. Urea formation was
done with 4ꢀnitrophenyl chloroformate and bis(2,4ꢀdimethoxybenzyl)amine to give urea 43,
which was then oxidized to the hydroxy ketone 44. Use of 4ꢀnitrophenyl chloroformate, instead
of triphosgene, under the SchottenꢀBaumann reaction condition avoided formation of a side
product affording 43 with improved yield. Unlike 36, cyclization of the ketone 43 was
accomplished without PPTS using Burgess reagent, furnishing oxazine 45. Replacement of the
mesylate group with NaI to 46 followed by fluorination using AgF led to 6ꢀfluoromethyl oxazine
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
4
7. After cleavage of the bisꢀ2,4ꢀdimethoxybenzyl group using TFA, 48 was subjected to
nitration reaction, which was followed by reduction of the nitro group and subsequent Boc
protection resulting in aniline 51. Finally, amide formation followed by deprotection delivered
the 6ꢀfluoromethyl substituted oxazines 15 and 20‒28. The absolute configuration was confirmed
by the Xꢀray structure of 15 bound to BACE1 (Figure 2).
a
Scheme 2. Synthesis of Oxazines 15 and 20‒28
ACS Paragon Plus Environment
27

Page 29 of 56
Journal of Medicinal Chemistry
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
a
Reagents and conditions: (a) allylmagnesium bromide, Et O, −60 to −20 °C, 72%; (b) 4 N
2
HCl in EtOAc, MeOH, 0 °C, quant; (c) 4ꢀnitrophenyl chloroformate, K CO , bis(2,4ꢀ
2
3
dimethoxybenzyl)amine, EtOAc−H O, 0 °C to rt, 82%; (d) KMnO , AcOH, acetone−H O, rt,
2
4
2
68%; (e) (i) Ms O, DCM, 0 °C, (ii) Burgess reagent, THF, rt, 40%; (f) NaI, acetone, rt, 95%; (g)
2
AgF, MeCN, rt, 98%; (h) anisole, TFA, 80 °C, quant; (i) HNO , H SO −TFA, −20 °C, 94%; (j)
3
2
4
Boc O, DMAP, THF, 0 °C, 90%; (k) Fe, NH Cl, EtOH−THF−H O, 60 °C, 85%; (l) (i) HATU,
2
4
2
RCO H, DIEA, DCM, rt, (ii) HCO H, rt, 61−95%.
2
2
CONCLUSION
ACS Paragon Plus Environment
28