Trifluoromethyl Dihydrothiazine-Based β-Secretase (BACE1) Inhibitors with Robust Central β-Amyloid Reduction and Minimal Covalent Binding Burden

Article abstract of DOI:10.1002/cmdc.201900478

The β-site amyloid precursor protein cleaving enzyme 1 (BACE1, also known as β-secretase) is a promising target for the treatment of Alzheimer's disease. A pK_a lowering approach over the initial leads was adopted to mitigate hERG inhibition and

Full text of DOI:10.1002/cmdc.201900478

Accepted Article
Title: Trifluoromethyl Dihydrothiazine-Based β-Secretase (BACE1)
Inhibitors with Robust Central Aβ Reduction and Minimal
Covalent Binding Burden
Authors: Ken-ichi Kusakabe, Kosuke Anan, Yasuyoshi Iso, Takuya
Oguma, Kenji Nakahara, Shinji Suzuki, Takahiko Yamamoto,
Eriko Matsuoka, Hisanori Ito, Gaku Sakaguchi, Shigeru Ando,
Kenji Morimoto, Naoki Kanegawa, Yasuto Kido, Tomoyuki
Kawachi, Tamio Fukushima, Ard Teisman, Vijay Urmaliya,
Deborah Dhuyvetter, Herman Borghys, Nigel Austin, An Van
Den Bergh, Peter Verboven, Francois Bischoff, Harrie J. M.
Gijsen, and Yoshinori Yamano
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemMedChem 10.1002/cmdc.201900478
Link to VoR: http://dx.doi.org/10.1002/cmdc.201900478
A Journal of
www.chemmedchem.org

ChemMedChem
10.1002/cmdc.201900478
Trifluoromethyl Dihydrothiazine-Based β-Secretase (BACE1)
Inhibitors with Robust Central Aβ Reduction and Minimal
Covalent Binding Burden
†
†
†
†
†
Kosuke Anan, Dr. Yasuyoshi Iso, Dr. Takuya Oguma, Dr. Kenji Nakahara, Shinji Suzuki,
†
,#
†
†
†
Dr. Takahiko Yamamoto, Eriko Matsuoka, Dr. Hisanori Ito, Dr. Gaku Sakaguchi, Shigeru
‡
‡
‡
‡
‡
Ando, Dr. Kenji Morimoto, Dr. Naoki Kanegawa, Dr. Yasuto Kido, Tomoyuki Kawachi, Dr.
‡
§
§
§
Tamio Fukushima, Dr. Ard Teisman, Dr. Vijay Urmaliya, Dr. Deborah Dhuyvetter, Dr.
§
║,◊
║
║
Herman Borghys, Dr. Nigel Austin, Dr. An Van Den Bergh, Dr. Peter Verboven, Dr.
║
║
†
Francois Bischoff, Dr. Harrie J. M. Gijsen, Dr. Yoshinori Yamano, and Dr. Ken-ichi
Kusakabe*,†
†
‡
Discovery Research Laboratory for Core Therapeutic Areas, Research Laboratory for
Development, Shionogi Pharmaceutical Research Center, 1-1 Futaba-cho 3-chome, Toyonaka,
Osaka 561-0825, Japan
§
║
Non-Clinical Safety, Discovery Sciences, Janssen Research & Development, Turnhoutseweg
3
0, B-2340 Beerse, Belgium
Abstract
The β-site amyloid precursor protein cleaving enzyme 1 (BACE1, also known as β-secretase) is a
promising target for the treatment of Alzheimer’s disease. A pK
a
lowering approach over the initial
leads was adopted to mitigate hERG inhibition and P-gp efflux, leading to the design of 6-CF
3
This article is protected by copyright. All rights reserved.
1

ChemMedChem
10.1002/cmdc.201900478
dihydrothiazine
8
(N-(3-((4S,6S)-2-amino-4-methyl-6-(trifluoromethyl)-5,6-dihydro-4H-1,3-
thiazin-4-yl)-4-fluorophenyl)-5-cyanopicolinamide). Optimization of 8 led to the discovery of 15
N-(3-((4S,6S)-2-amino-4-methyl-6-(trifluoromethyl)-5,6-dihydro-4H-1,3-thiazin-4-yl)-4-
(
fluorophenyl)-5-(fluoromethoxy)pyrazine-2-carboxamide) with an excellent balance of potency,
hERG inhibition, P-gp efflux, and metabolic stability. Oral administration of 15 elicited robust Aβ
reduction in dog even at 0.16 mg/kg, leading to an EC50 value of 29 ng/mL. Reflecting the reduced
hERG inhibitory activity, no QTc prolongation was observed at high doses. The potential for
1
4
reactive metabolite formation of 15 was realized in a nucleophile trapping assay using [ C]-KCN
in human liver microsomes. Utilizing covalent binding (CVB) in human hepatocytes and the
maximum projected human dosage (73 mg), the daily CVB burden of 15 was calculated to be at
an acceptable value of below 1 mg/day. However, hepatotoxicity was observed when 15 was
subjected to a 2-week tolerance study in dog, which prevented further evaluation of this compound.
Introduction
Alzheimer’s disease (AD) is a neurodegenerative disease and the most prevalent form of
dementia, accounting for 60 to 80% of all cases. In 2016, 47 million people were living with
1
dementia worldwide, and the number is predicted to increase to 131 million by 2050. In the United
States, AD was the sixth leading cause of death in 2017, and there were 5.5 million people living
with the disease. The total cost for patients with AD or other dementias was estimated at $259
2
billion. In addition to the mortality and healthcare burden, current treatment options, such as
donepezil and memantine, only afford suboptimal symptomatic effects and do not slow the disease
progression. Thus, there is an urgent need to develop therapeutics with preventive or disease-
modifying potential.
This article is protected by copyright. All rights reserved.
2

ChemMedChem
10.1002/cmdc.201900478
The hallmarks of AD are amyloid-β (Aβ) plaques and neurofibrillary tangles of
3
hyperphosphorylated tau proteins. The Aβ peptides consisting of 36 to 43 amino acids are
generated by proteolysis of the amyloid precursor protein (APP) via sequential cleavage of β-site
amyloid precursor protein cleaving enzyme 1 (BACE1), also known as β-secretase, and γ-secretase,
providing Aβ40 as the most prevalent form of Aβ, as well as minor amounts of the more toxic
Aβ42.3,4 On the basis of the amyloid hypothesis as well as accumulating genetic and biomarker
evidence, BACE1 has become a prime target for treatment of AD.4,5 Indeed, familial AD is
associated with mutations in APP resulting in increased Aβ production such as in the case of the
Swedish mutation (KM670/671NL) and a rare A673V mutation located near the BACE1 catalytic
4
6
site, ultimately causing AD where the ages of dementia onset are similar between generations.
In contrast, the protective mutation of A673T reduces the Aβ production by 40% which leads to
7
protection against developing AD and cognitive decline in elderly individuals. Such evidence
points to the development of BACE1 inhibitors as a promising strategy for modifying or
preventing AD.
8
Since the identification of BACE1 in 1999, academic and industrial institutions have
9
extensively pursued orally available and brain penetrant BACE1 inhibitors. Although initial
BACE1 inhibitors included peptide mimetic structures with low cellular potency and high P-gp
recognition resulting in limited oral availability and central efficacy, identification of amidine-
9
based inhibitors addressed such issues, leading to multiple clinical BACE1 inhibitors, such as
verubecestat (MK-8931),10 LY2811376,11a LY2886721,11b LY3202626,11c lanabecestat
(
AZD3293),12 elenbecestat (E2609),13 umibecestat (CNP520),14 and atabecestat (JNJ-
1
5
5
4861911). Disappointingly, some of the clinical compounds were not acceptable due to safety
and efficacy reasons. For example, Eli Lilly suspended clinical trials of LY2811376 following
This article is protected by copyright. All rights reserved.
3

ChemMedChem
10.1002/cmdc.201900478
preclinical toxicology issues which were likely caused by inhibition of cathepsin D (CatD),11a,16
and development of LY2886721 and atabecestat was stopped after serious elevations of liver
enzymes were observed in phase II and phase III trials, respectively.11b The most advanced clinical
BACE1 inhibitor verubecestat did not reduce cognitive decline in patients with mild-to-moderate
AD.10d Merck terminated the phase III clinical study of verubecestat enrolling prodromal AD
patients as it failed to show a positive risk-benefit profile after the interim analysis.10e Clinical
studies of autosomal dominant (familial) and sporadic AD suggest that Aβ accumulation in the
brain begins at least 20 years before the onset of dementia and reaches a plateau at the onset,
indicating that administration of BACE1 inhibitors need to be initiated at the start of Aβ deposition.
Targeting patients with clinical symptoms, such as prodromal or mild-to-moderate AD, is likely
1
7
to be too late to achieve a significant clinical outcome. To successfully deliver BACE1 inhibitors
to the market, relevant clinical designs such as intervention at a very early stage of AD using
asymptomatic individuals as well as inhibitors with safer profiles are required given the failures of
the clinical compounds tested thus far.
We previously described a series of oxazine-based BACE1 inhibitors originating from a high-
1
8
throughput screening of our compound library (Figure 1). Initial efforts identified a cellular
potent oxazine lead 3, which was also a strong P-gp substrate and hERG inhibitor. This issue was
addressed by a pK
a
lowering approach culminating in oxazines such as 1 and 2. During this effort,
less basic inhibitors displayed reduced P-gp efflux and hERG inhibition, although the drop in
basicity, in particular compounds with pK
a
< 6.5, was associated with a loss in cellular potency,
< 7.4 for BACE1 inhibitors.18a The
which led to the proposal of a basicity guideline of 6.5 < pK
a
early efforts also identified dihydrothiazine 4 as a lead, which displayed a higher potency than the
corresponding dihydro-oxazine 3, though exhibiting high hERG inhibition and P-gp efflux like 3.
This article is protected by copyright. All rights reserved.
4

ChemMedChem
10.1002/cmdc.201900478
Considering the difference in pK
.0, respectively) as well as the pK
design dihydrothiazines with the optimal pK
medicinal chemistry efforts that led to a 6-CF
a
between the dihydro-oxazine 3 and dihydrothiazine 4 (9.8 and
9
a
guideline above, alternative substituents were required to
a
range to address the issue. Herein, we detail our
3
dihydrothiazine 15 with improved hERG inhibition
and P-gp efflux, achieving robust Aβ reduction in both mouse and dog at doses as low as 1 mg/kg.
Although the dihydrothiazine 15 exhibited a potential for reactive metabolite formation in an
1
4
electrophile trapping assay using [ C]-KCN, its covalent binding (CVB) burden, in which the
maximum predicted human daily dose and the metabolic turnover rates were considered, was in
an acceptable range of below 1 mg/day.19,20
Head
Tail
F
F
H
H N
O
H N
O
F
2
2
NC
F
F
O
N
N
H
N
H
N
N
N
N
O
O
F
=
6.9)
= 7.2)
1
2
(pK
(
pK
a
a
H N
X
F
H N
O
CF3
2
2
NC
NC
N
N
H
N
N
H
N
N
O
O
F
X = O (pK = 9.8)
(pK = 7.5)
3
4
:
:
5
a
a
X = S (pK = 9.0)
a
Figure 1. BACE1 Inhibitors
Results and Discussion
Design of the 6-CF
3
Dihydrothiazine Head Group. As described above, we identified the
dihydrothiazine 4 as an initial lead, which was a potent BACE1 inhibitor with a biochemical IC50
This article is protected by copyright. All rights reserved.
5

ChemMedChem
10.1002/cmdc.201900478
of 9.9 nM and cellular Aβ40 IC50 of 0.63 nM, displaying inhibition of hERG (91% inhibition at 3
μM) and being a P-gp substrate (efflux ratio = 3.1). Considering our experience that lowering the
1
8
pK
to optimize the dihydrothiazine series. With detailed knowledge of the optimal pK
amidine moiety (6.5 < pK < 7.4) to reduce hERG inhibition and P-gp efflux as well as retaining
a
worked well in previous SAR efforts to address these issues, we adopted a similar strategy
a
range of the
a
cellular potency,18a our design efforts focused on incorporating electron withdrawing substituents
into the dihydrothiazine head group. During this work, an elegant molecular design was disclosed
by Hilpert et al. at Roche to reduce the basicity leading to the oxazine 5 with a CF group at the 6-
3
position, which lowered the amidine pK
a
by 2.3 units relative to the non-substituted oxazine 3
(
Figure 2).21 Interestingly, this modification also improved BACE1 potency in our BACE1
biochemical assay. Given the pK
a
value of 9.0 in the dihydrothiazine 4, incorporation of the CF
3
group could achieve an optimal pK
a
value of 6.7 leading to the design of a 6-CF dihydrothiazine
3
8
.
This article is protected by copyright. All rights reserved.
6

ChemMedChem
10.1002/cmdc.201900478
=
–2.3
∆
H N
O
pK
a
H N
O
CF3
2
2
N
N
3
5
=
=
9.8
7.5
= 9.5 nM
pK
a
pK
a
=
118 nM
BACE1 IC50
hERG inhib at 3 µM = 66%
BACE1 IC50
hERG inhib at 3 µM = 68%
=
–0.8
∆
pK
a
H N
S
H N
S
CF3
2
2
N
N
4
8
=
9.0
pK
a
This work
=
6.7
=
9.9 nM
predicted pK
a
BACE1 IC50
hERG inhib at 3 uM = 91%
Figure 2. Design of a 6-CF
3
dihydrothiazine BACE1 inhibitor
Exploration of Fluoro-alkyl at the 6-Position on the Dihydrothiazine. The 6-CF
3
dihydrothiazine 8 indeed displayed an experimental pK value of 6.6 as predicted (Table 1). The
a
reduced basicity in the amidine group was translated into significantly reduced hERG inhibition
at 3 μM and P-gp efflux (41% and 0.80, respectively) when compared with those of the non-
substituted dihydrothiazine 4 (91% and 3.1, respectively). No improvement in potency was
observed with this modification (4 to 8), unlike that in the oxazine series (3 to 5; Figure 2). To
further test our design hypothesis, difluoroalkyl substituents, such as difluoromethyl (6) and 1,1-
difluoroethyl (7), were incorporated. As expected, difluoro-alkyl 6 and 7 had higher pK
a
values
(7.5 and 7.4, respectively) than 8 resulting in increased hERG inhibition and P-gp efflux (~67%
and ~2.3, respectively). Furthermore, this change was also accompanied by reduced microsomal
stability in both rat and human, despite their decreased LogD values (2.7) relative to that in 8 (3.8).
Most importantly, the CF dihydrothiazine 8 demonstrated an improved hERG profile when
3
This article is protected by copyright. All rights reserved.
7

ChemMedChem
10.1002/cmdc.201900478
compared with the corresponding oxazine 5, which was thought to be attributed to the higher pK
of 5.
a
To confirm whether the reduced P-gp efflux translated into improved brain penetration, we
investigated the non-substituted 4, difluoromethyl 6, and 8 in our rat pharmacokinetic (PK)
screening studies using a cassette dosing method (Table 2). The compounds were dosed orally and
intravenously (iv) at 1 mg/kg and 0.5 mg/kg, respectively, and brain penetration was measured at
0
.5 h in the iv studies. The initial lead 4 with a P-gp efflux of 3.1 displayed a modest total brain-
to-plasma ratio (K ) of 0.67 and unbound brain-to-plasma ratio (Kp,uu) of 0.12, whereas the 6-CF
dihydrothiazine 8 did achieve improved brain penetration (K = 2.3, Kp,uu = 1.1). The in vivo effect
of P-gp efflux was evaluated using mdr1a knockout (KO) mice. The KO and wild-type (WT) mice
were administered an oral dose of 2 mg/kg of compounds 4 and 8, where the KO/WT ratios (K in
the KO mice/K in the WT mice) were determined 2 h after dosing. Indeed, for 8, a significant
p
3
p
p
p
reduction of the KO/WT ratio was observed relative to that for 4 (KO/WT ratio = 2.3 vs 31,
respectively). Given the good passive permeability of both 4 and 8 (Papp = 19.1 and 13.5 × 10-6
cm/s) in LLC-PK cells, the increased brain penetration of 8 was attributed to the reduced P-gp
efflux. The decreased metabolic stability observed with CF
2
H 6 was reflected in a high clearance
(CL) value of 47.6 mL/min/kg. The improved brain penetration observed with 8 prompted us to
optimize the amide tail group to further improve the hERG inhibition.18
This article is protected by copyright. All rights reserved.
8

ChemMedChem
10.1002/cmdc.201900478
Table 1. Exploration at the 6-Position on the Dihydrothiazine
H N
S
R
2
NC
6
N
H
N
N
O
F
(nM)a
IC50
HLM hERG inhib
pKag
LogDh
P-gp
BACE1b
c
RLMd at 3 µMe
compd
R
H
ERf
Cellular Aβ
4
9.9
14
12
0.63
104
91
67
61
41
3.1
9.0
1.2
66
6
7
CF H
0.29
0.68
69
2.0
7.5
2.7
2
53
CF Me
71
2.3
7.4
2.7
2
52
8
5
CF3
—
9.9 ± 2.2 0.79 ± 0.39
96
0.80
6.6
3.8
82
9.5
1.4
103
00
68
1.5
7.5
2.7
1
a
Values represent the mean value of at least two determinations. Compound 8 includes SD values.
b
c
Biochemical homogeneous time-resolved fluorescence (HTRF)-based assay. HTRF-based assay
measuring secreted Aβ40 levels in human APP-transfected neuroblastoma (SH-SY5Y) cells. %
remaining after 30 min incubation in human and rat liver microsomes. % inhibition at 3 μM
d
e
measured in CHO cells transfected with hERG channels using an automated patch clamp system.
f
The value for compound 4 was determined at 5 μM. Efflux ratio measured in LLC-PK1 cells
g
h
transfected with human MDR1. pK
a
determined by capillary electrophoresis. LogD determined
in 1-octanol/phosphate buffer at pH 7.4.
This article is protected by copyright. All rights reserved.
9

ChemMedChem
10.1002/cmdc.201900478
Table 2. Rat Pharmacokinetic Profiles for Key Compounds
rat, iv, 0.5 mg/kg, n = 2a
rat, po, 1 mg/kg, n = 2b
AUC (ng·h/ml)h Cmax (ng/ml)i
CL (ml/min/kg)d Vdss (
L/kg)e
Kpf
Kp,uug
F (%)j
compd fu,s / fu,bc
4
6
8
9
2
4
5
0.41 / 0.071
13.5
47.6
12.7
5.7
8.8
6.6
0.67 0.12
328
83.8
32
16
38
61
24
62
0.11 / NT
1.8
2.3
1.5
ND
0.037 / 0.017
0.024 / 0.014
0.17 / 0.039
0.057 / 0.014
0.055 / 0.013
1.1
819
427
23.8
26.3
31.1
15.2
6.6
2.9
5.1
6.1
0.88
38
17
14
31
60
10
16
27
1
1
1
0.76 0.17
65.0
84.9
301
2.9
3.7
0.71
0.88
a
Dosed as a solution of test compounds in N,N-dimethylacetamide (DMA)/propylene glycol
b
c
(
=
PG) = 1/1. Dosed as a suspension of test compounds in 0.5% methylcellulose (MC, 400 cP). fu,s
fraction unbound in rat serum. fu,b = fraction unbound in rat brain. Total clearance. Volume of
distribution at steady state. Total brain-to-plasma ratio. Unbound brain-to-plasma ratio. Plasma
area under the concentration-time curve. Maximal plasma concentration. Oral bioavailability.
d
e
f
g
h
i
j
Table 3. Pharmacokinetic Profiles of Key Compounds in Wild-type (C57BL/6J) and mdr1a(-
/
-) Mice.
c
po, 2 mg/kg, 2 h, n = 3
compd
P-gp Pappb
ER
moused
Kpe
KO/WTf
a
4
8
9
3.1
19.1
WT
KO
WT
KO
0.44 ± 0.02
13.5 ± 2.2
2.6 ± 0.3
6.1 ± 0.2
2.0 ± 0.1
7.2 ± 1.2
0.47 ± 0.04
5.8 ± 0.4
31
0.80 13.5
2.3
3.6
12
1.0
17.0
19.7
WT
KO
WT
KO
1
2
5
2.0
1
0.70
8.38
WT
KO
2.9 ± 0.3
5.9 ± 0.9
2.0
a
b
Efflux ratio measured in LLC-PK1 cells transfected with human MDR1. Permeability in wild-
c
d
type LLC-PK1 cells. Dosed as a suspension of test compounds in 0.5% MC. WT = wild-type
This article is protected by copyright. All rights reserved.
1
0

ChemMedChem
10.1002/cmdc.201900478
e
(
C57BL/6J) mouse. KO = mdr1a (-/-) mouse. Total brain-to-plasma ratio at 2 h time point.
f
KO/WT = (K
p
in KO)/(K in WT)
p
Optimization at the Tail Group in 8 Leading to 15. We had previously pursued the tail
analogues to reduce hERG inhibition as well as P-gp efflux, in which incorporation of substituents
at the ortho-position of the carbonyl on the pyridine or replacement of the pyridine with a pyrazine
1
8
mitigated these liabilities. With these findings in mind, we focused on optimizing the tail group
in 8 (Table 4). Adding a methyl to the ortho-position on the pyridine (9) reduced hERG inhibition
as expected; however, it was associated with reduced metabolic stability. Changing the CN to a Cl
group (10) also mitigated hERG inhibition relative to 8, although again a somewhat reduced
microsomal stability was observed. We anticipated that the reduced metabolic stability observed
in 10 could be attributed to its increased LogD value of 4.8 relative to that of 8 (LogD = 3.8). To
this end, we replaced the pyridine ring with a pyrimidine resulting in compound 11, which indeed
reduced lipophilicity (LogD = 3.4) and improved metabolic stability in rat but showing no
significant improvement in humans. Further reduction in lipophilicity was realized by replacing
the Cl in 11 with a methoxy group leading to 12 (LogD = 2.6). Although this change displayed
poor microsomal stability in rats, it resulted in a significant improvement for human. In addition,
the changes to the pyrimidine caused an increased P-gp efflux. Moving a nitrogen in 12 to the 1-
position provided pyridazine analogue 13 exhibiting a significantly reduced potency associated
with an increased LogD value. Finally, incorporation of a pyrazine ring achieved an optimal
balance in terms of potency, metabolic stability, hERG inhibition, and P-gp efflux. Compound 14
with a methoxypyrazine had a minimal hERG inhibitory activity at 3 μM of 13%. Consistent with
1
8
our previous structure-activity relationship (SAR), adding a fluorine onto the methoxy in 14 (15)
improved metabolic stability in both humans and rats as well as potency but retained other profiles
This article is protected by copyright. All rights reserved.
1
1

ChemMedChem
10.1002/cmdc.201900478
such as hERG inhibition and P-gp efflux. Like the comparison between the cyanopyridine
analogues 5 and 8, the CF dihydrothiazine 15 showed relatively reduced hERG inhibitory activity
3
compared with the corresponding oxazine 16 together with the slightly higher potency in both
biochemical and cellular assays.
We subsequently performed rat screening PK studies of compounds 9, 12, 14 and 15 (Table 2).
With a relatively high RLM stability, 15 exhibited a low CL value of 15.2 mL/min/kg. The low P-
gp efflux ratios observed for 9, 14, and 15 were translated into high Kp,uu values close to 1, whereas
1
2 with a relatively high P-gp efflux resulted in a low Kp,uu value of 0.17. Notably, the PK studies
of 9, 12 and 15 using mdr1a KO mice also highlighted the involvement of P-gp in brain penetration
within these compounds. The KO/WT ratio in 15 was found to be 2.0, compared to a ratio of 12
for 12. Based on the reduced hERG inhibitory activity, the high HLM stability, and the increased
brain penetration derived from the reduced P-gp efflux, compound 15 was selected for further
characterization.
This article is protected by copyright. All rights reserved.
1
2

ChemMedChem
10.1002/cmdc.201900478
Table 4. Exploration of the 6-CF
3
Dihydrothiazine at the Tail Group and Profiles of Oxazine
1
6
H N
S
F
CF3
H N
O
F
CF3
2
2
F
O
N
H
N
N
H
N
R
N
N
O
O
16
(
nM)a
IC50
BACE1b
f
LogDg
HLM hERG inhib
RLMd at 3 µMe
P-gp ER
c
compd
R
Cellular Aβ
NC
NC
N
N
8
9
9.9 ± 2.2 0.79 ± 0.39
96
41
23
0.80
1.0
3.8
3.8
82
6.4
6.0
0.79
1.6
71
68
Cl
Cl
N
N
1
1
1
0
1
2
81
26
30
17
0.81
1.7
4.8
3.4
2.6
64
8.1 ± 1.4 1.3 ± 0.33
81
87
N
O
O
O
N
N
N
9.7
0.91
95
68
2.0
N
N
1
1
1
1
3
4
5
6
86
15
25
92
30
13
13
22
ND
3.8
3.8
4.0
2.9
56
1.4
86
0.60
0.70
1.4
N
N
68
F
O
N
9.1 ± 1.7 0.82 ± 0.12
19 ± 3.1 3.6 ± 1.1
94
94
109
93
This article is protected by copyright. All rights reserved.
1
3

ChemMedChem
10.1002/cmdc.201900478
a
Values represent the mean value of at least two determinations. Compounds 8, 11, 15, and 16
b
include SD values. Biochemical homogeneous time-resolved fluorescence (HTRF)-based assay.
HTRF-based assay measuring secreted Aβ40 levels in human APP-transfected neuroblastoma
SH-SY5Y) cells. % remaining after 30 min incubation in human and rat liver microsomes. %
c
d
e
(
inhibition at 3 μM measured in CHO cells transfected with hERG channels using an automated
f
patch clamp system. Efflux ratio measured in LLC-PK1 cells transfected with human MDR1.
g
LogD determined in 1-octanol/phosphate buffer at pH 7.4.
Crystal Structure of 15 Bound to BACE1. The crystal structure of 15 bound to human BACE1
confirmed that the dihydrothiazine moiety forms hydrogen bonding interactions with Asp32 and
Asp228 residues, the amide hydrogen forms a hydrogen bond with the carbonyl oxygen of Gly230,
and the fluoromethoxy pyrazine occupies the S3 pocket. As was observed with the oxazine
analogues 2 and 5,18b,21 the CF
group at the 6-position does not engage in any interactions with
3
BACE1 (Figure 3).
Figure 3. Structure of compound 15 bound to human BACE1.
This article is protected by copyright. All rights reserved.
1
4

ChemMedChem
10.1002/cmdc.201900478
In Vitro ADMET Profiles of 15. The crystalline form of 15 was obtained as a free base, which
showed moderate thermodynamic solubility of 785 μg/mL and 2.7 μg/mL at pH 1.2 and 6.8,
respectively. The pK of 15 was measured to be 6.8. In addition to human and rat microsomes, the
a
metabolic stability of 15 was examined in mouse, dog, and monkey microsomes after 30 min
incubation (69%, 100%, and 64%, respectively). The in vitro plasma protein binding of 15 was
determined in preclinical species and humans by equilibrium dialysis using the Dianorm apparatus
at a concentration of 2 µM, where the protein binding values (% unbound) were 2.34%, 2.86%,
2
.92%, 4.34%, and 2.84% in mice, rats, dogs, monkeys, and humans, respectively. Compound 15
exhibited moderate to weak inhibition over cytochrome P450 (CYP) isoforms with IC50 values of
20 μM, >20 μM, 7.6 μM, 6.6 μM, and 5.6 μM for 1A2, 2C19, 2C9, 2D6, and 3A4, respectively.
The hERG IC50 value was found to be 15.6 μM using a manual patch clamp method. Compound
5 did not show any significant activity over cathepsin D (CatD; IC50 = >100 μM) and also no
>
1
selectivity over a close homolog of BACE2 with a biochemical IC50 value of 15.2 nM. Finally, the
potential of genetic toxicity was evaluated with the Ames test using mutant strains of TA98,
TA1537, TA100, TA1535 and TA102 in the absence and presence of rat liver metabolic activation
(S9-mix); 15 gave negative results (below 2-fold increase in the number of colonies).
In Vivo Pharmacology of 15 in Mouse and Dog. Pharmacodynamic (PD) profiles of compound
1
5 for the ability to lower Aβ levels in the brain were determined in mouse and dog. For the PD
assay in mouse, total Aβ levels in male ICR mice (n = 4) orally administered with compound 15
at doses of 1 and 3 mg/kg were measured at 2, 4, and 6 h time points and compared to those treated
with vehicle (Figure 4). A significant and dose-dependent decrease of total Aβ was seen, where 15
reduced total Aβ by 76% (4 h) at a free brain concentration of 4.1 ng/mL (8.9 nM) at 1 mg/kg.
Consistent with the Kp,uu in rat as shown in Table 2, 15 demonstrated a Kp,uu value close to 1 (Kp,uu
This article is protected by copyright. All rights reserved.
1
5

ChemMedChem
10.1002/cmdc.201900478
=
1.3) at 1 mg/kg (4 h time point). At 3 mg/kg, a maximum Aβ reduction of 87% was achieved at
a free brain concentration of 9.5 ng/mL (21 nM).
Inhibition of Aβ production in the cerebrospinal fluid (CSF) in beagle dogs (n = 4) was also
2
2
investigated (Figure 5). Compound 15 was dosed at 0.08, 0.16, 0.31, 0.63 and 2.5 mg/kg with
plasma samples being taken at 0.5, 1, 2, 4, 8, 25, and 49 h after dosing. CSF from conscious animals
was also sampled from the lateral ventricle before dosing and at 4, 8, 25, and 49 h post dose. As
shown in Figure 4, 15 caused dose-dependent and robust CSF Aβ42 reduction by 50% even at 0.16
mg/kg. In terms of brain penetration in dog, the Kp,uu (CSF) value (CCSF/Cp,u) at 4 and 8 h (0.16
mg/kg) was determined to be 0.62. Finally, the unbound plasma concentrations to reduce CSF Aβ42
by 50% (EC50,u) and 90% (EC90,u) were calculated to be 0.85 ng/mL (1.8 nM) and 7.7 ng/mL (17
nM), respectively (total plasma EC50 = 29 ng/mL; EC90 = 262 ng/mL), demonstrating the ability
of 15 to reduce Aβ at low concentrations.
This article is protected by copyright. All rights reserved.
1
6

ChemMedChem
10.1002/cmdc.201900478
Figure 4. Total Aβ reduction in the brain in male ICR mice (n = 4) after oral dosing of 15 as a
solution of 20% hydroxypropyl-β-cyclodextrin (HPBCD): (A) 15 was dosed at 3 mg/kg; (B) 15 at
1
mg/kg.
Figure 5. Aβ42 reduction in the CSF in beagle dog (n = 4) after oral dosing of 15 at 0.08, 0.16,
0
.31, 0.63, and 2.5 mg/kg as a solution of 20% HPBCD.
Human Dose and PK/PD Projection of 15. Prior to the human dose projections, PK profiles
of 15 were investigated in mouse, rat, dog, and monkey following intravenous and oral
administrations (Table 5). Reflecting the high microsomal stability in dog, the dog CL value was
low, whereas the volume of distribution at steady state was high across the species, resulting in a
long half-life (t1/2) of 21.5 h in dog. Compound 15 also achieved moderate to high exposure levels
and Cmax at po doses, translating into excellent bioavailability across multiple species. Utilizing
the PK parameters as well as the PD data for dog, the Dedrick approach was applied to generate
human dose and PK/PD projections (See Supporting Information). The target trough plasma
This article is protected by copyright. All rights reserved.
1
7

ChemMedChem
10.1002/cmdc.201900478
concentration (Ctrough) was set at the level to reduce CSF Aβ42 by 90% (EC90 = 262 ng/mL). For
this level of reduction, 15 was predicted to require a once-daily dose of 73 mg with an expected
C
max of 472 ng/mL, an AUC of 8,539 ng/mL·h, and a t1/2 of 22 h.
Table 5. PK Parameters of 15 in Mouse, Rat, Beagle Dog, and Cynomolgus Monkey.
iva
a
po
(h)
dose
CL
Vdss
t1/2
dose
AUC
Cmax
F (%)
(mg/kg) (ml/min/kg)
(mg/kg)
(ng·h/ml)
(L/kg)
(ng/ml)
mouse
rat
2.5
1
20
20
5.6
4.7
4.8
4.7
3.8
2.5
10
5
7420
1902
941
408
87
45
63
85
dog
1
2.7
10
21.5
7.6
5
19300
6442
1500
868
monkey
1
5
aDosed as a solution of 15 in 20% HPBCD.
Cardiovascular Safety of 15. The cardiovascular (CV) safety profile of 15 was assessed using
pentobarbital-anesthetized guinea pigs (n = 7), where doses of 0.25, 0.5, 1, 2, 4, and 8 mg/kg were
infused over 5 min at 15 min intervals, and these were compared to the vehicle (20%
hydroxypropyl-β-cyclodextrin) treated group. No QTc prolongation was observed up to 8 mg/kg,
at which 15 reached a compound level of 6,950 ng/mL (964 nM free; f in guinea pig = 0.064),
u
resulting in a safety margin of 33-fold (a free concentration-based margin) over the human
projected Cmax of 472 ng/mL (29 nM free). Based on the favorable CV safety profile in guinea pig,
work on 15 progressed to CV safety studies in lofentanil/scopolamine anesthetized beagle dogs.
The compounds were intravenously infused with doses of 0.16, 0.32, 0.63, 1.25, 2.5, and 5 mg/kg.
Consistent with the results in guinea pigs, even at the highest dose of 5 mg/kg (18,950 ng/mL;
This article is protected by copyright. All rights reserved.
1
8

ChemMedChem
10.1002/cmdc.201900478
1
,199 nM free), 15 did not induce relevant effects on QTc intervals, indicating low potential for
QTc prolongation.
Covalent Binding Burden of 15 in Human Hepatocytes. Avoiding reactive metabolite (RM)
formation with drug candidates could mitigate the risk of drug-induced liver injury (DILI).23
Functional groups leading to RM formation via oxidative enzymes are referred to as “structural
alerts”, and eliminating them is a strategy to address the issue of mitigating DILI. However, as
2
0
almost half of the top 200 drugs in 2009 have structural alerts, it is important to quantitatively
estimate to what extent the candidates can form RM. Compound 15 possesses an isothiourea
substructure, which is susceptible to metabolic activation like thioureas. To investigate the
potential to form RM, 15 was incubated in human liver microsomes with exogenously added
14
trapping agents of dansylated glutathione (GSH) and [ C]-KCN. Compound 15 did not form
1
4
dansyl-GSH conjugates, whereas reactive intermediates trapped with CN amounted to be 807
pmol/mg protein, indicating the potential for bioactivation of 15 to reactive intermediates in the
liver. In contrast, the corresponding oxazine analogue 16 did not provide any detectable dansyl-
GSH and 14CN adducts in the trapping assay, implying that the predominant bioactivation
mechanism of 15 could involve oxidation of the sulfur atom on the dihydrothiazine followed by
nucleophilic reactions with liver proteins leading to RM formation.
To further investigate the potential for RM formation of 15, the covalent binding (CVB)
1
4
capability to human hepatocyte using [ C]-15 was assessed. The level of CVB at a concentration
of 10 μM for 15 was measured to be 32.2 pmol drug equivalents/mg hepatic protein. Troglitazone,
2
0
which is known to form RM and was withdrawn due to DILI, has been reported to show levels
of CVB of 153 pmol/mg protein. The level of RM for 15 was less than that of troglitazone and was
more or less the same as pioglitazone (11.5 pmol/mg protein), which also forms RM but very rarely
This article is protected by copyright. All rights reserved.
1
9

ChemMedChem
10.1002/cmdc.201900478
causes DILI. Although it is unclear to what extent the levels of CVB need to be reduced to avoid
DILI, an estimation of the body burden of CVB (CVB burden), where the levels of CVB were
corrected for metabolic turnover and maximum human daily dose, can distinguish drugs that can
1
9
cause DILI. Thus, we utilized a daily dose of 73 mg/day as the maximum, because it was
estimated to retain the Ctrough that can reduce Aβ by 90% in human. As shown in Table 6, the CVB
burden for 15 was calculated to be 0.56 mg/day, whereas those for troglitazone and pioglitazone
were 4.4 and 0.050 mg/day. Thompson et al. suggested a CVB burden of 1 mg/day as a threshold
value to provide effective discrimination between safe and unsafe drugs concerning DILI,19
demonstrating that 15 was in an acceptable range regardless of the potential for RM formation
observed in the CVB and CN trapping studies. The acceptable CVB burden value as well as other
favorable in vitro and in vivo profiles led 15 being subjected to in vivo toxicology studies.
Table 6. Covalent binding (CVB) and estimation of CVB burden for 15 and reference
compoundsa
(
pmol/mg protein)c
Daily dose (mg)b CVBNET
73
CVB Burden (mg/day)d
compd
1
3
3
4
15
C]-
32.2
0.56
4.4
[
[
[
600
153
11.5
H]-troglitazone
H]-pioglitazone
45
0.050
a
14
The CVB was measured in human hepatocytes at 10 μM of [ C]-15 and the reference
b
compounds. The maximum daily dose used for the reference compounds. For 15, a predicted
human dose at which 15 can reach dog EC90 as a trough level. The amount of the covalent binding
after incubation of [ C]-15 and the reference compounds, from which the results of the 0 min
incubation samples were subtracted. Covalent binding burden based on the daily dose.
c
1
4
d
This article is protected by copyright. All rights reserved.
2
0

ChemMedChem
10.1002/cmdc.201900478
In Vivo Toxicology in Rat and Dog. Prior to repeated dose studies, single dose toxicology
studies were conducted for Sprague-Dawley rats and beagle dogs including both males and females.
In rats (po; n = 3/gender/dose), 15 was well tolerated and showed no adverse clinical findings up
to 1000 mg/kg, where a difference in toxicokinetic (TK) profiles between males and females was
observed as shown in Table S1. In dogs (po; n = 2/gender/dose), 15 was orally administered at
doses of 50, 100, and 300 mg/kg (Table S3). Although decreased food consumption was seen in a
female dog at both 100 and 300 mg/kg, 15 was well tolerated at 50 mg/kg.
On the basis of the tolerated doses in the single dose studies, we subsequently performed
repeated dose toxicology studies for 14 days with male rats (n = 5/dose; 0, 60, 150, and 500
mg/kg/day) and male and female dogs (n = 2/gender/dose; 10, 30, and 100 mg/kg/day). In rats, no
adverse clinical and histopathological findings were observed up to 150 mg/kg/day, whereas at
5
00 mg/kg/day a decrease in body weight gain and decreased food consumption along with
hematopoietic system features were considered to be toxicologically relevant. The medium dose
of 150 mg/kg/day was the no observed adverse effect level (NOAEL) of 56,200 and 11,800
ng/mL·h (AUC0-24h) in males and females, respectively, where the safety margin of the NOAEL
(AUC0-24h) over the projected human AUC (8,539 ng/mL·h) calculated above was 6.6- and 1.4-
fold, respectively (Table S2, see Supporting Information). In dogs, minimal hyperplasia of the bile
duct was seen in one male and one female at 30 mg/kg/day and in all males and females at 100
mg/kg/day. In addition, 15 caused alkaline phosphatase (ALP) elevation and cellular infiltration
of mixed cells at 30 and 100 mg/kg/day and hepatocellular single-cell necrosis in a male at 30
mg/kg/day. Thus, there were no adverse findings up to 10 mg/kg/day reaching AUC0-24h values of
9
1
0,100 and 43,500 ng/mL·h in males and females (NOAEL), which resulted in safety margins of
1- and 5.1-fold in males and females, respectively.
This article is protected by copyright. All rights reserved.
2
1

ChemMedChem
10.1002/cmdc.201900478
Contrary to the findings in the rat studies, the repeated dose toleration studies in dogs detected
toxicity of the hepatic and biliary functions, such as bile duct hyperplasia, ALP elevation, and
hepatocellular single-cell necrosis. Although the minimal daily CVB burden of 15 (0.56 mg/day)
was confirmed, it did exhibit a potential of RM formation in the CVB study at 10 μM concentration,
which was above the Cmax observed in the repeated dose study at 30 mg/kg/day in dog and in the
same range as the Cmax at 10 mg/kg/day. Nevertheless, it is unclear whether the hepatic toxicities
observed in dogs could be attributed to the CVB of 15 measured in human hepatocytes. As
discussed by Thomson et al., enhanced prediction of the potential of DILI in human may require
additional in vitro assessments including measuring cytotoxicity derived from RM, mitochondrial
toxicity, and inhibition of the activity of the human bile salt export pump (BSEP) and the rat
multidrug resistance-associated protein 2 (Mrp2), generating the integrated in vitro hazard matrix
which achieved discrimination between safe and unsafe drugs in terms of DILI using 36 drugs
19
expressing different DILI patterns in humans. Given the lack of confidence in how the hepatic
toxicity in dog associated with the potential of RM formation translates to the clinical setting as
well as the recent Phase III failure of our clinical BACE1 inhibitor, atabecestat, due to the DILI,15b
compound 15 was considered to not contribute to the de-risking of the liver toxicity observed in
atabecestat and thus led to discontinuation of its further study.
Chemistry
Preparation of the 6-nonsubstituted dihydrothiazine 4 is outlined in Scheme 1. The synthesis
commenced with the well-established diastereoselective alkylation to Ellman’s imine24 for
construction of its C4 chiral stereocenter. Addition of a Reformatsky reagent, readily prepared
21
from ethyl bromoacetate, to Ellman’s imine 20 afforded the t-butyl sulfinyl amine 21. The amine
This article is protected by copyright. All rights reserved.
2
2

ChemMedChem
10.1002/cmdc.201900478
was treated with HCl to remove the tert-butyl sulfinyl group to give the primary amine 22, which
was subsequently protected with Boc O. After reduction of the ethyl ester to a primary alcohol
with LiBH , the Boc group in 23 was removed to afford the amino alcohol 24, which was then
2
4
reacted with benzoyl isothiocyanate to give 25. Cyclization of 25 was accomplished with 1-chloro-
N,N,2-trimethylprop-1-en-1-amine (Ghosez’s reagent)25 to form the dihydrothiazine 26. The
benzoyl group of the dihydrothiazine was cleaved by hydrazine to give 27. Regioselective para-
nitration on the fluorophenyl group of 27 and subsequent reduction of the nitro group with iron
afforded the aniline 28, which was successfully condensed with 5-cyanopicolinic acid in the
presence of 1 equivalent of HCl to furnish the 6-nonsubstituted dihydrothiazine 4.
Scheme 1. Synthesis of the 6-nonsubstituted dihydrothiazine 4a
O
NH2
O
NHBoc
S
a
b
c,d
S
N
F
O
NH
O
OEt
OH
OEt
F
F
F
21
22
23
20
S
BzHN
S
F
NH2
F
e
f
g
h
BzHN
NH
N
OH
OH
F
24
25
26
H N
S
H N
S
F
H N
S
F
2
2
2
NC
i,j
k
N
N
N
H
N
H N
N
2
O
F
27
28
4
This article is protected by copyright. All rights reserved.
2
3

ChemMedChem
10.1002/cmdc.201900478
aReagents and conditions: (a) Zn, BrCH
quantative; (c) Boc O, EtOH, 70 °C (d) LiBH
f) benzoyl isothiocyanate, DCM, rt, quantative; (g) Ghosez’s reagent, DCM, 0 °C, 76%; (h)
hydrazine hydrate, EtOH, rt, 80%; (i) H SO , HNO , TFA, -20 °C; (j) Fe, NH Cl, toluene/water,
0 °C, 76% in 2 steps; (k) HCl, 5-cyanopicolinic acid, EDC·HCl, MeOH, 0 °C, quantative.
CO
Et, THF, 97%; (b) HCl, MeOH, 0 °C to rt,
2
2
2
4
, THF, 0 °C to rt, 99% in 2 steps; (e) HCl, rt, 97%;
(
2
4
3
4
8
A series of the 6-fluoroalkyl-substituted dihydrothiazines as in Table 1 was prepared with the
synthetic strategy used for 4. Diastereoselective alkylation to Ellman’s imines and cyclization of
N-hydroxypropyl thiourea using N,N-diethylaminosulfur trifluoride (DAST)26 enabled
construction of six-membered dihydrothiazine cores. Scheme 2 presents the preparation of the 6-
difluoromethyl dihydrothiazine 6. The ethyl ester 21 derived from Ellman’s imine 20 was
subjected to diethyl (difluoromethyl)phosphonate to introduce the difluoromethyl group, followed
by removal of the t-butyl sulfinyl group to afford the amino alcohol 30. Using benzoyl
isothiocyanate, 30 was transformed to the N-hydroxypropyl thiourea 31. The cyclization was
accomplished by our previously reported DAST-mediated dihydrothiazine selective cyclization
26
without forming the corresponding oxazine. After removal of the benzoyl group of 32, the
synthesis of the 6-difluoromethyl dihydrothiazine 6 was elaborated via nitration, reduction, and
condensation with 5-cyanopicolinic acid, in accordance with the same procedure used for 4.
This article is protected by copyright. All rights reserved.
2
4

ChemMedChem
10.1002/cmdc.201900478
Scheme 2. Synthesis of the 6-difluoromethyl dihydrothiazine 6a
NH2 OH
S
a,b
S
c
d
O
NH
F
O
O
NH OH
F
F
OEt
F
F
F
F
30
21
29
F
F
S
BzHN
S
H N
S
e
f
2
g,h
F
F
BzHN
NH OH
N
N
F
F
F
F
F
31
32
33
F
F
F
H N
S
F
H N
S
F
2
2
F
NC
i
N
N
H
N
N
H N
2
O
34
6
aReagents and conditions: (a) LDA, diethyl (difluoromethyl)phosphonate, THF, -78 °C; (b)
NaBH , i-PrOH, 30% in 2 steps; (c) HCl, dioxane, rt, 98%; (d) benzoyl isothiocyanate, DCM, 0 °C,
4%; (e) DAST, DCM, rt, 81%; (f) hydrazine hydrate, EtOH, rt, 96%; (g) H SO , HNO , TFA, -
0 °C; (h) iron, NH Cl, toluene/water, 80-90 °C, 87% in 2 steps; (i) HCl, 5-cyanopicolinic acid,
4
9
1
2
4
3
4
EDC·HCl , MeOH, 0 °C to rt, 79% in 2 steps
Synthesis of the 6-(1,1-difluoroethyl)dihydrothiazine 7 was carried out with a modified strategy
2
1
of 6. The known Ellman’s imine-derived aldehyde 35 was reacted with a Reformatsky reagent
to afford 36. The t-butyl sulfinyl group of 36 was removed by HCl followed by introduction of the
N-benzoyl thiourea group with benzoyl isothiocyanate to give 37. Construction of the
dihydrothiazine core was accomplished via DAST-mediated cyclization under the same conditions
as that employed for the cyclization of 31 to 32. The ethyl ester group of 38 was converted to the
This article is protected by copyright. All rights reserved.
2
5

ChemMedChem
10.1002/cmdc.201900478
alkyl iodide 39 with NaBH
4
and a subsequent Appel reaction, followed by radical reduction with
Bu
3
SnH and 2,2'-azobis(isobutyronitrile) (AIBN) to obtain the dihydrothiazine 42 possessing the
difluoroethyl moiety. After removal of the benzoyl group, followed by nitration, reduction with
iron, the resulting aniline 42 was acylated with 5-cyanopicolinic acid to afford the 6-(1,1-
difluoroethyl)dihydrothiazine 7.
Scheme 3. Synthesis of the 6-(1,1-difluoroethyl)dihydrothiazine 7a
S
a
b,c
d
S
S
NH OH
BzHN
O
NH
F
O
O
NH OH
COOEt
COOEt
H
F F
F F
F
F
35
36
37
F F
COOEt
F F
F F
BzHN
S
BzHN
S
F
I
BzHN
S
e,f
g
h
N
N
N
F
F
38
39
40
F F
F F
F F
H N
S
H N
S
H N
S
F
2
2
2
NC
i, j
k
N
N
N
H
N
H N
N
2
O
F
F
41
42
7
aReagents and conditions: (a) Zn, ethyl 2-bromo-2,2-difluoroacetate, THF, rt, 24%; (b) HCl,
dioxane, rt; (c) benzoyl isothiocyanate, DCM, 0 °C, 18%; (d) DAST, DCM, rt, 59%; (e) NaBH
EtOH, 0 °C, (f) PPh , imidazole, iodine, THF, rt, 80 °C, 59% in 2 steps; (g) Bu SnH, AIBN, toluene,
0°C, quantative; (h) hydrazine hydrate, EtOH, rt, 66%; (i) H SO , HNO , TFA, -10 °C; (j) iron,
Cl, toluene/water, 80-90 °C, 54% in 2 steps; (k) HCl, 5-cyanopicolinic acid, EDC·HCl ,
4
,
3
3
8
2
4
3
NH
4
MeOH, 0 °C to rt, 79% in 2 steps.
This article is protected by copyright. All rights reserved.
2
6

ChemMedChem
10.1002/cmdc.201900478
Synthesis of 6-trifluoromethyl dihydrothiazines 8−15 was accomplished by acylation of aniline
2
6
4
3, which was prepared according to our previous report, with corresponding carboxylic acids in
the presence of 1 equivalent of HCl with chemical yields of 72 to 91% (Scheme 4).
Scheme 4. Synthesis of the 6-trifluoromethyl dihydrothiazines 8-15a
H N
S
CF3
H N
S
CF3
2
2
a
N
H
N
H N
R
N
2
O
F
F
43
8-15
NC
NC
O
Cl
Cl
F
N
N
N
N
N
8:
9:
10:
11:
15:
N
O
N
O
O
N
N
N
12:
13:
14:
N
N
N
aReagents and conditions: (a) HCl, RCO
H, EDC·HCl, MeOH, 0 °C to rt, 72−91%.
2
Conclusions
Utilizing a pKa lowering strategy, a 6-CF dihydrothiazine-based BACE1 inhibitor was
3
identified, leading to an early lead 8 possessing reduced hERG inhibition and P-gp efflux relative
to those of the corresponding oxazine 5. Further mitigation of hERG inhibition was achieved when
the cyanopyridine tail group in 8 was replaced with other hetero aromatics, such as pyrimidines,
pyridazines, and pyrazines. This work culminated in compound 15 with a fluoromethoxypyrazine,
yielding an optimal balance of cellular potency, hERG inhibition, P-gp efflux, as well as metabolic
stability. The improved P-gp efflux translated into a low KO/WT ratio in vivo using mdr1a KO
This article is protected by copyright. All rights reserved.
2
7

ChemMedChem
10.1002/cmdc.201900478
mice. Oral administration of 15 reduced dog CSF Aβ42 by 50% even at 0.16 mg/kg leading to
EC50,u and EC90,u of 1.8 nM and 17 nM, respectively. Reflecting the reduced hERG inhibition, no
QTc prolongation was observed even at unbound concentrations of approximately 10 μM in dog
and guinea pig. The quantitative measurements of bioactivation in human liver microsomes using
[14
C]-KCN revealed the potential of 15 to form RM. For risk-benefit analysis of reactive metabolite
1
4
exposure, we assessed the CVB ability using [ C]-15 in human hepatocytes. Calculation of the
CVB burden based on the CVB value, the compound turnover, and the predicted maximum human
dosage provided an acceptable daily CVB burden of below 1 mg/day. However, despite the
acceptable CVB burden, compound 15 caused liver toxicities, such as ALP elevations, in a 2-week
tolerance study in dog. Based on the adverse events observed in the liver and the failure of our
clinical compound, atabecestat, due to hepatotoxicity, we elected not to advance 15 into further
evaluation studies.
Experimental Section
General Chemistry. All commercial reagents and solvents were used as received without
further purification. Flash column chromatography was carried out on an automated purification
1
13
system using Yamazen or Fuji Silysia prepacked silica gel columns. H and C NMR spectra
were recorded on a Bruker Advance 400 and 100 MHz spectrometers, respectively. Spectral data
are reported as follows: chemical shift (as ppm referenced to tetramethylsilane), integration
value, multiplicity (s = singlet, d = doublet, dd = double doublets, ddd = double double doublet,
dt = double triplet, t = triplet, q = quartet, m = multiplet, br = broad peak), and coupling constant.
Analytical LC/MS was performed on a Shimadzu Shim-pack XR-ODS (C18, 2.2 µm, 3.0 × 50
mm, a linear gradient from 10% to 100% B over 3 min and then 100% B for 1 min (A = water +
This article is protected by copyright. All rights reserved.
2
8

ChemMedChem
10.1002/cmdc.201900478
0
.1% formic acid, B = MeCN + 0.1% formic acid), flow rate 1.6 mL/min) using a Shimadzu
UFLC system equipped with a LCMS-2020 mass spectrometer, LC-20AD binary gradient
module, SPD-M20A photodiode array detector (detection at 254 nm), and SIL-20AC sample
manager. The purity of all compounds used in the bioassays was determined by this method to be
>95%. High resolution mass spectra were recorded on a Thermo Fisher Scientific LTQ Orbitrap
using electrospray positive ionization.
Ethyl (S)-3-((tert-butoxycarbonyl)amino)-3-(2-fluorophenyl)butanoate (21). To a stirred
suspension of Zn (2.84 g, 43.5 mmoL) in THF (30 mL) was added CuCl (1.23 g, 12.4 mmol).
After stirring for 30 min at 50 °C, ethyl 2-bromoacetate (3.46 mL, 31.1 mmol) in THF (10 mL)
was added to the reaction mixture dropwise over 15 min. After stirring for 20 min at the same
temperature, the reaction mixture was cooled to -30 °C, and 20 (3.00 g, 12.4 mmol) in THF (20
mL) was added dropwise over 15 min. After stirring for 160 min at 0 °C, the reaction was
quenched with sat. aqueous NH
4
Cl. The aqueous layer was extracted with EtOAc and the organic
. The crude product was purified by silica-
layer was washed with brine, and dried over MgSO
4
gel column chromatography (n-hexane/EtOAc, gradient: 30−60% EtOAc) to afford 21 (3.97g,
1
1
2.1 mmol 97%) as a colorless oil. H NMR (400 MHz, CDCl
3
) δ 1.12 (3H, t, J = 7.2 Hz), 1.29
(
9H, d, J = 8.9 Hz), 1.85 (3H, s), 3.10 (1H, dd, J = 16.3, 1.3 Hz), 3.37 (1H, d, J = 16.3 Hz), 4.02
2H, q, J = 7.2 Hz), 5.50 (1H, s), 7.01 (1H, ddd, J = 12.8, 8.2, 1.1 Hz), 7.12 (1H, ddd, J = 7.6,
(
7
.6, 1.1 Hz), 7.23−7.29 (1H, m), 7.50 (1H, ddd, J = 8.2, 8.2, 1.7 Hz). MS-ESI (m/z): 330 [M +
H]+.
(
S)-N-((2-(2-Fluorophenyl)-4-hydroxybutan-2-yl)carbamothioyl)benzamide (22). To a
stirred solution of 21 (3.97 g, 12.1 mmol) in MeOH (8.0 mL) was added HCl (4 M in dioxane,
.52 mL, 18.1 mmol) dropwise over 3 min at 0 °C. After stirring for 45 min at room temperature,
4
This article is protected by copyright. All rights reserved.
2
9