β-secretase (BACE1) inhibitors with high in vivo efficacy suitable for clinical evaluation in Alzheimer's disease

Article abstract of DOI:10.1021/jm400225m

An extensive fluorine scan of 1,3-oxazines revealed the power of fluorine(s) to lower the pK_a and thereby dramatically change the pharmacological profile of this class of BACE1 inhibitors. The CF₃ substituted oxazine 89, a potent and

Full text of DOI:10.1021/jm400225m

Subscriber access provided by University of South Dakota
Article
#-Secretase (BACE1) Inhibitors with High In Vivo Efficacy
Suitable for Clinical Evaluation in Alzheimer’s Disease
Hans Hilpert, Wolfgang Guba, Thomas J. Woltering, Wolfgang Wostl, Emmanuel Pinard, Harald Mauser,
Alexander V Mayweg, Mark Rogers-Evans, Roland Humm, Daniela Krummenacher, Thorsten Muser, Christian
Schnider, Helmut Jacobsen, Laurence Ozmen, Alessandra Bergadano, David W. Banner, Remo Hochstrasser,
Andreas Kuglstatter, Pascale David-Pierson, Holger Fischer, Alessandra Polara, and Robert Narquizian
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/jm400225m • Publication Date (Web): 16 Apr 2013
Downloaded from http://pubs.acs.org on April 26, 2013
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 44
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
βꢀSecretase (BACE1) Inhibitors with High In Vivo
Efficacy Suitable for Clinical Evaluation in Alzheimer’s
Disease
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Hans Hilpert,*,† Wolfgang Guba,† Thomas J. Woltering,† Wolfgang Wostl,† Emmanuel Pinard,† Harald
Mauser,† Alexander V. Mayweg,† Mark RogersꢀEvans,† Roland Humm,† Daniela Krummenacher,†
Thorsten Muser,† Christian Schnider,† Helmut Jacobsen,‡ Laurence Ozmen,‡ Alessandra Bergadano,‡
David W. Banner,§ Remo Hochstrasser,§ Andreas Kuglstatter,§ Pascale DavidꢀPierson,# Holger Fischer,#
Alessandra Polara,∥ Robert Narquizian∥
†Discovery Chemistry, ‡DTA Neuroscience, §Discovery Technologies, #Drug Metabolism and
∥
Pharmacokinetics, Roche Partnering, F. HoffmannꢀLa Roche Ltd, pRED, Pharma Research & Early
Development, Grenzacherstrasse 124, Basel, Switzerland, CHꢀ4070
ABSTRACT: An extensive fluorine scan of 1,3ꢀoxazines revealed the power of fluorine(s) to lower the
pKa and thereby dramatically change the pharmacological profile of this class of BACE1 inhibitors. The
CF3 substituted oxazine 89, a potent and highly brain penetrant BACE1 inhibitor, was able to reduce
significantly CSF Aβ40 & 42 in rats at oral doses as low as 1 mg/kg. The effect was long lasting,
showing a significant reduction of Aβ40 & 42 even after 24 h. In contrast to 89, compound 1b lacking
the CF3 group was virtually inactive in vivo.
INTRODUCTION
Alzheimer’s disease (AD) is a progressive neurodegenerative disease, and the major cause of dementia
in the elderly. The clinical symptoms of AD are progressive memory loss, learning impairment and
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 44
behavioral and psychiatric disturbances. The predominant pathological hallmarks of the disease are 1.
amyloid plaques, i.e., extracellular deposits of polymerized amyloidꢀβ peptides (Aβ), and 2. intracellular
aggregates of misꢀfolded tau protein. Formation of these pathologies can precede the clinical disease
symptoms by years. The final disease pathology is further characterized by pronounced
neuroinflammation and widespread neuronal loss. As no causative treatment is currently available, there
is an urgent need for new, efficacious therapies.1 Amyloid aggregates are considered to be the earliest
pathology, and according to the amyloid cascade hypothesis, the ultimate cause of the disease.2ꢀ5 Their
main constituents, the amyloidꢀβ peptides, are derived from the βꢀamyloid precursor protein (APP) via
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
proteolytic processing by βꢀsecretase (BACE1) and γꢀsecretase. Both APP and BACE1 are typeꢀ1
membrane proteins highly expressed in the brain. BACE1 is an aspartic protease of 501 amino acids
length, organized into an extracellular domain with signal peptide, prodomain and catalytic domain, a
transmembrane domain and a short cytoplasmic stretch of 22 amino acids. Its principal cleavage site in
the APP extracellular domain forms the Nꢀterminus of Aβ (position D672 of the longest APP isotype); a
secondary cleavage site is localized at position E682. Processing of APP occurs predominantly in
endosomal vesicles, in agreement with the “acidic pH” optimum” of the enzyme.6ꢀ8 The unique and
essential role of BACE1 for the generation of Aβ has been convincingly shown in mice where both
alleles have been ablated by genetic means. These mice do not form Aβ, nor do they cleave APP at the
known BACE1ꢀpositions.9, 10 It has furthermore been demonstrated that upon crossing BACE1 knockꢀ
out mice and APPꢀtransgenic mice, the formation of brain amyloid is inhibited. Interestingly, ablation of
only one BACE1 allele was enough to reduce amyloidosis significantly. These data suggest that
pharmacological inhibition of BACE1 should also reduce amyloidosis, and that partial inhibition should
be sufficient to achieve therapeutic efficacy.11, 12 Taken together this makes BACE1 a prime target for
antiꢀamyloid therapy. Partial inhibition of BACE1 may also be the preferred therapeutic target, since a
number of additional BACE1 substrates have been described. Of special interest is neuregulin 1 (NRG1)
processing, since close examination of BACE1 knockꢀout mice has revealed that in these animals the
ACS Paragon Plus Environment

Page 3 of 44
Journal of Medicinal Chemistry
myelination of peripheral and possibly central nerves is severely reduced, presumably due to the absence
of NRG1 processing.13, 14 Importantly, this adverse effect was not observed in animals that retained one
functional allele.
1
2
3
4
5
6
7
8
9
The aspartyl protease BACE2 is a close homolog of BACE1. BACE2, however, cleaves APP
preferentially at a position downstream of the BACE1 cleavage sites at F691 of APP, which does not
produce amyloidogenic Aβ peptides. BACE2 is primarily expressed in peripheral tissue, in particular in
pancreatic islets.15 Its physiological function was poorly understood for some time,8 but recent evidence
suggests that it may play a role in glucose homeostasis.15
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Soon after the discovery of BACE1 in 1999,16ꢀ20 the enzyme became a prime target of both academic
and industrial research. Initial inhibitor design focused on statine type structures engaging with the two
aspartic acids in the catalytic site. While potent in vitro, the peptidic nature of these inhibitors rendered
them poorly brain penetrant. To overcome this shortcoming, less peptidic hydroxyethylamine (HEA)
isosters subsequently became the focus of research. Good brain penetration, however, remained the
exception rather than the rule. This has been ascribed to the flexibility of HEAs and their high number of
hydrogen bond donors and acceptors.21
Our research focused on cyclic, conformationally rigid structures that interact with the two aspartic
acids of BACE1. An HTS campaign yielded a small fragment of a weakly active (IC50=41.2 ꢀM) 2ꢀ
aminoꢀ4Hꢀ5,6ꢀdihydroꢀ[1,3]thiazine derivative A (Figure 1) which could be coꢀcrystallized with
BACE1, offering unique opportunities for optimization. This heterocycle was first synthesized at Roche
more than 30 years ago in a project looking for novel analgesics,22 and has later been explored by
various research groups23ꢀ28 as a head group interacting with the aspartic acids of BACE1.
Previously we described the identification and characterization of a number of cyclic amidines that
were devoid of substantial in vivo activity due to low penetration into the CNS.29 In this paper we
disclose our efforts to convert our initial lead compound, the 2ꢀaminoꢀ4Hꢀ5,6ꢀdihydroꢀ[1,3]oxazine
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 44
derivative 1a into fluorine containing oxazines B, a compound class characterized by high in vivo
1
2
3
activity and good overall properties.
4
5
6
7
8
Figure 1. Lead Compound 1a and Targeted Fluoroꢀoxazines B.
9
S
O
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
H2N
H2N
HCl
H2N
O
R2
R3
N
N
N
H
N
H
R3
Cl
N
N
R1
F
F
O
O
MeO
A
1a
B
R1: heteroaromatic residue
R2: H, CF3; R3: Fn (n=0-2)
BACE1 (enzyme): BACE1 (enzyme): IC50: 0.137 uM
IC50: 41.2 uM
Cell-based (Aβ40), IC50: 0.010 uM
CHEMISTRY
The syntheses of oxazines 1a-b and difluoroꢀoxazines 14a-v (Scheme 1) started with elegant work
developed by Ellman.30 Condensation of the chiral auxiliary (R)ꢀtertꢀbutanesulfinamide and
acetophenone 2 in the presence of Ti(OEt)4 gave the sulfinyl imine 3. Transmetalation of the lithium
enolate of ethyl acetate using TiCl(OiPr)3 generated the titanium enolate which reacted with the sulfinyl
imine 3 to furnish the ester 4 as a single diastereoisomer. Alternatively, the Reformatsky reaction of the
sulfinyl imine 3 and ethyl bromodifluoroacetate31 furnished the α,αꢀdifluoroester 5.
Scheme 1. Synthesis of Oxazines 1a-b and Difluoro-oxazines 14a-v.a
ACS Paragon Plus Environment

Page 5 of 44
Journal of Medicinal Chemistry
O
tBu
O
tBu
O
OH
R
S
1
2
3
tBu
S
COOEt
S
O
N
O
HN
R
R
HN
a
b or c
d
R
+
S
tBu
O2N
O2N
O2N
O2N
H2N
4
F
F
F
F
5
6
2
3
4: R=H
5: R=F
6: R=H
7: R=F
7
8
e
9
10
11
12
13
H2N
O
H2N
O
H2N
O
OH
R
N
R
R
N
R
R
N
R
R
H2N
h
g
f
H
R
N
R1
H2N
O2N
O2N
F
F
F
F
O
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1a: R=H, R1=5-chloro-pyridin-2-yl
1b: R=H, R1=5-cyano-pyridin-2-yl
12: R=H
13: R=F
10: R=H
11: R=F
8: R=H
9: R=F
1
14a: R=F, R =5-chloro-pyridin-2-yl
14b-v: R=F (R1 see Table 2)
aReagents and conditions: (a) Ti(OEt)4, THF, 50ꢀ70 °C. (b) 3ꢀ>4: MeCOOEt, LDA, TiCl(OiPr)3, THF, ꢀ78 °C. (c) 3->5:
BrCF2COOEt, Zn, CuCl, THF, 35 °C. (d) LiBH4, THF, 0 °C. (e) HCl/1,4ꢀdioxane/MeOH, 22 °C. (f) BrCN, EtOH, 85 °C. (g)
H2, Pd/C, EtOH, 22 °C. (h) 4ꢀ(4,6ꢀDimethoxy[1.3.5]triazinꢀ2ꢀyl)ꢀ4ꢀmethylmorpholinium chloride hydrate (DMTMM),
MeOH, 22 °C.
The ester group in 4 and 5 was then reduced to the alcohols 6 and 7 with LiBH4 which was followed by
cleavage of the chiral auxiliary under acidic conditions to give the amino alcohols 8 and 9.32 Cyclization
of the amino alcohols 8 and 9 was readily accomplished with cyanogen bromide to give the oxazines 10
and 11 which were reduced to the corresponding anilines 1233 and 13 by catalytic hydrogenation. Amide
formation was accomplished without prior protection of the amidine moiety using 4ꢀ(4,6ꢀ
dimethoxy[1.3.5]triazinꢀ2ꢀyl)ꢀ4ꢀmethylmorpholinium chloride hydrate,34 furnishing the target
compounds 1a-b and 14a-v (Table 2). The absolute configurations were assigned based on Xꢀray
structures of 1b and 14d complexed to BACE1.
The commercially unavailable acids 23-26 (Scheme 2), required to introduce R1 residues in difluoroꢀ
oxazine 14 (Table 2), were made available from pyridine and pyrazine derivatives 15 and 17 via the
alkylated esters 19-22. Attempted alkylation of the hydroxyꢀpyrazine ester 16 with 3ꢀbromopropyne did
not lead to the expected Oꢀalkylated product 22 but furnished the Nꢀalkylated product (5ꢀoxoꢀ4ꢀpropꢀ2ꢀ
ynylꢀ4,5ꢀdihydroꢀpyrazineꢀ2ꢀcarboxylic acid methyl ester, structure not shown). Compound 27 was
prepared directly from the acid 18 without proceeding via the corresponding ester. Finally, chlorination
of the pyrazole ester 28 with Nꢀchlorosuccinimide followed by saponification delivered the chloroꢀ
pyrazole derivative 30.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 44
1
2
3
4
Scheme 2. Synthesis of Commercially Unavailable
Building Blocks 23-27 and 30.a
5
6
7
R2
R2
R1
O
O
N
N
N
a
b
OMe
OMe
OH
X
X
X
8
9
O
O
O
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
23: X= CH, R2= CF3
24: X= CH, R2= CF2CHF2
25: X= N, R2= CF3
15: X=CH, R1=OH
16: X=N, R1=OH
17: X=N, R1=Cl
19: X= CH, R2= CF3
20: X= CH, R2= CF2CHF2
21: X= N, R2= CF3
26: X= N, R2= ethynyl
22: X= N, R2= ethynyl
N
Cl
c
OH
27: X= N, R2= prop-1-yn-1-yl
N
F
O
18
F
F
F
F
F
N
N
N
N
N N
d
b
OMe
OMe
OH
O
O
O
Cl
29
Cl
30
28
aReagents and conditions: (a) 15ꢀ>19: CF3CH2OTf, NaH, DMF,
22 °C; 15ꢀ>20: CHF2CF2CH2OTf, K2CO3, acetone; 17ꢀ>21:
CF3CH2OH, Cs2CO3, DMF; 17ꢀ>22: propꢀ2ꢀynꢀ1ꢀol, tertꢀBuOK,
1,4ꢀdioxane, 22 °C. (b) LiOH or NaOH, THF/H2O, 22 °C to
reflux. (c) Butꢀ2ꢀynꢀ1ꢀol, tertꢀBuOK, DMF, 65 °C. (d) Nꢀ
chlorosuccinimide, DMF, 50 °C.
To access fluoroꢀoxazines 66-70 (Scheme 3) having fluorines attached at positions R2ꢀR4 the
synthetic strategy was modified. Occasionally the alcohol group in 6-9 (Scheme 1) can undergo an
intramolecular aromatic substitution of the fluorine, which is activated by the nitro group, to give the
corresponding nitroꢀchromane derivatives (structures not shown). We therefore introduced the nitro
group at a later stage of the synthesis. The commercially unavailable ketones 34 and 35 (Scheme 3) were
made available by fluorination of the alcohol 31 or by acylation of 32 using ethyl difluoroacetate.
Sulfinyl imines 36ꢀ38, prepared as described in Scheme 1, were then subjected to the Reformatsky
reaction to afford the esters 39-43. The reaction of 36 with BrCHFCOOEt yielded a separable 4:1ꢀ
mixture of the epimers 39 and 40. Conversion of 37 with BrCHFCOOEt gave a separable 8:1ꢀmixture of
41 as the major isomer and the corresponding epimer (R3=H, R4=F). Oxazines 49-51 and 53-54,
prepared in analogy to Scheme 1, were then nitrated with fuming nitric acid, followed by reduction of
ACS Paragon Plus Environment

Page 7 of 44
Journal of Medicinal Chemistry
the nitro group and amide formation in the usual manner, to give the targeted fluorinated oxazines 66-
70. The relative and absolute configurations were assigned based on Xꢀray structures of 66ꢀ70
complexed to BACE1.
1
2
3
4
5
6
7
8
9
Scheme 3. Synthesis of Fluoro-oxazines 66-70.a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
O
O
OH
tBu
O
OH
S
tBu
COOEt
S
a
R4
R4
O
N
O
H2N
HN
R2
R2
e, f
F
3
3
c
d
R2
R
R2
R
+
R1
R1
R1
R1
S
tBu
31
H2N
F
F
F
F
4
33: R1=H, R2=CH3
34: R1=H, R2=CH2F
35: R1=Br, R2=CHF2
36: R1=H, R2=CH3
37: R1=H, R2=CH2F
38: R1=Br, R2=CHF2
39: R1=H, R2=CH3, R3=F, R =H
40: R1=H, R2=CH3, R3=H, R4=F
41: R1=H, R2=CH2F, R =F, R =H
42: R1=Br, R2=CHF2, R3=R4=H
43: R1=H, R2=CH2F, R3=R4=F
44: R1=H, R2=CH3, R3=F, R =H
45: R1=H, R2=CH3, R3=H, R4=F
46: R1=H, R2=CH2F, R3=F, R =H
47: R1=Br, R2=CHF2, R3=R4=H
48: R1=H, R2=CH2F, R3=R4=F
4
b
Br
F
3
4
4
32
g
H2N
O
H2N
O
H2N
O
H2N
O
R4
2 R3
NC
R4
2 R3
R4
R4
N
N
N
N
N
OH
k
j
i
3
3
H
R
R2
R2
N
R
R
N
H2N
O2N
R1
NC
R
O
F
F
F
F
O
49: R1=H, R2=CH3, R3=F, R =H
50: R1=H, R2=CH3, R3=H, R4=F
51: R1=H, R2=CH2F, R3=F, R =H
52: R1=Br, R2=CHF2, R3=R4=H
53: R1=H, R2=CHF2, R3=R4=H
54: R1=H, R2=CH2F, R3=R4=F
4
4
4
66: R2=CH3, R3=F, R =H
67: R2=CH3, R3=H, R4=F
68: R2=CH2F, R3=F, R =H
69: R2=CHF2, R3=R4=H
70: R2=CH2F, R3=R4=F
55: R2=CH3, R3=F, R =H
56: R2=CH3, R3=H, R4=F
60: R2=CH3, R3=F, R =H
61: R2=CH3, R3=H, R4=F
4
65
4
3
4
4
62: R2=CH2F, R =F, R =H
63: R2=CHF2, R3=R4=H
64: R2=CH2F, R3=R4=F
57: R2=CH2F, R =F, R =H
58: R2=CHF2, R3=R4=H
59: R2=CH2F, R3=R4=F
3
4
h
aReagents and conditions: (a) 31ꢀ>34: Nonafluoroꢀnꢀbutansulfonyl fluoride, triethylamine trihydrofluoride, CH2Cl2, 0 °C to
22 °C. (b) 32ꢀ>35: LDA, F2CHCOOEt, THF, ꢀ70 to ꢀ10 °C. (c) Ti(OEt)4, THF, 70 °C. (d) 39, 40 and 41: BrCHFCOOEt, 42:
BrCH2COOEt, 43: BrCF2COOEt, Zn, CuCl, THF. (e) LiBH4, EtOH, 22 °C. (f) aq. HCl, THF, 22 °C. (g) BrCN, iꢀPrOH or
EtOH, 80 °C. (h) H2, Pd/C, NaOAc, MeOH, 22 °C. (i) HNO3, H2SO4, TFA, 0 °C. (j) H2, Pd/C, Et3N, EtOH, 22 °C. (k)
DMTMM, MeOH, 0 °C to 22 °C.
The synthesis of trifluoromethyl substituted oxazines 88 and 89 (Scheme 4) required a different
strategy. Ready access to CF3 substituted 3ꢀaminoꢀpropanꢀ1ꢀol derivatives 80 and 81 is achieved using a
dipolar cycloaddition reaction of a nitrile oxide and an olefin35 as the key step. Utilizing carefully
optimized conditions, the nitrile oxide was slowly generated in situ from 2ꢀchloroꢀ2ꢀhydroxyꢀimino ester
71 in a heterogeneous mixture of NaHCO3 in ethyl acetate in the presence of 3,3,3ꢀtrifluoropropene (72)
to give the 4,5ꢀdihydroꢀisoxazole ester 73 in quantitative yield. Reduction of the ester 73 followed by
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 44
fluorination of the alcohol 74 with DeoxoꢀFluor® furnished the desired fluoro compound 75. 3ꢀMethylꢀ
4,5ꢀdihydroꢀisoxazole derivative 76 was directly accessible by reacting nitroethane and phenyl
isocyanate to generate the nitrile oxide in situ in the presence of 72. Arylation of the imines 75 and 76
with 1ꢀbromoꢀ2ꢀfluoroꢀbenzene (77) and nꢀBuLi occurred from the sterically less hindered side to give
the racemic isoxazolidines 78 and 79, which were separated by chiral HPLC affording the chiral
compounds 78 and 79. Transfer hydrogenation employing ammonium formate and Pd/C led to the
amino alcohols 80 and 81, which were then converted to the CF3ꢀsubstituted oxazines 88 and 89
following the procedures employed in Scheme 3. The relative and absolute configurations were assigned
based on Xꢀray structures of 88 and 89 complexed to BACE1.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 4. Synthesis of Trifluoromethyl-oxazines 88 and 89.a
CF3
HO
H2N
CF3
R
O
HN
CF3
CF3
+
O
O
CF3
OH
N
O
Br
CF3
a
b
c
e
f
+
N
N
N
Cl
R
F
EtOOC
71
EtOOC
73
HO
R
F
F
72
74
75: R=F
76: R=H
77
78: R=F
79: R=H
80: R=F
81: R=H
d
g
H2N
O
CF3
H2N
O
CF3
H2N
O
CF3
R
H2N
O
CF3
R
j
i
h
N
NC
N
N
N
N
OH
+
H
N
NC
R
R
N
H2N
O2N
O
F
F
F
F
O
65
88: R=F
89: R=H
86: R=F
87: R=H
84: R=F
85: R=H
82: R=F
83: R=H
aReagents and conditions: (a) NaHCO3, EtOAc, ꢀ65 °C to 22 °C. (b) NaBH4, EtOH, 0 °C. (c) DeoxoꢀFluor®, CH2Cl2, ꢀ78 °C
.
to 22 °C. (d) 72ꢀ>76: Nitroethane, phenylisocyanate, Et3N (cat.), Et2O, ꢀ78 °C to 22 °C. (e) THF, nꢀBuLi, BF3 Et2O, ꢀ78 °C to
22 °C. (f) HCOONH4, Pd/C, EtOH, 22 °C. (g) BrCN, EtOH, 85 °C. (h) H2SO4, HNO3, 0 °C. (i) H2, Pd/C, EtOH, Et3N, 22 °C.
(j) DMTMM, MeOH, 0 °C to 22 °C.
The last set of CF3ꢀsubstituted oxazines 109-111, not accessible by methods hitherto described, again
required a different synthetic strategy. Esters 90 and 39 were reduced with DIBAH at ꢀ76 °C to give the
aldehydes 91 and 92 which were reacted with trimethyl(trifluoromethyl)silane (RuppertꢀPrakash
reagent) yielding the alcohols 93ꢀ96. The reaction of 91 afforded a separable 1:2ꢀmixture of epimers 93
and 94, aldehyde 92 gave a separable 2:1ꢀmixture of epimers 95 and 96 with inverse selectivity.
ACS Paragon Plus Environment

Page 9 of 44
Journal of Medicinal Chemistry
Conversion of alcohols 93, 95 and 96 to the target compounds 109ꢀ111 followed wellꢀestablished
procedures as described in previous Schemes. The relative and absolute configurations were assigned
based on Xꢀray structures of 109ꢀ111 complexed to BACE1.
1
2
3
4
5
6
7
8
9
Scheme 5. Synthesis of Fluoro-oxazines 109-111.a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
O
OH
R2
R3
R2
R3
R1
tBu
O
tBu
O
tBu
O
HO
H2N
S
tBu
COOEt
R1
CHO
R1
S
S
S
N
HN
HN
HN
R1
a
b
c
d
F
F
F
F
F
36
90: R1=H
39: R1=F
91: R1=H
92: R1=F
93: R1=H, R2=CF3,R3=H
94: R1=H, R2=H, R3=CF3
95: R1=F, R =CF3, R3=H
97: R1=H, R2=CF3,R3=H
98: R1=F, R2=CF3, R3=H
99: R1=F, R2=H, R3=CF3
2
96: R1=F, R =H, R3=CF3
2
e
R2
R3
R1
R2
R3
R1
R2
R3
R1
R2
H2N
O
H2N
O
H2N
O
H2N
O
R3
R1
N
NC
N
N
N
N
OH
h
g
f
H
N
NC
N
H2N
O2N
O
F
F
F
F
O
109: R1=H, R2=CF3,R3=H
110: R1=F, R2=CF3, R3=H
111: R1=F, R =H, R3=CF3
106: R1=H, R2=CF3,R3=H
107: R1=F, R =CF3, R3=H
108: R1=F, R =H, R3=CF3
103: R1=H, R2=CF3,R3=H
104: R1=F, R2=CF3, R3=H
105: R1=F, R2=H, R3=CF3
100: R1=H, R2=CF3,R3=H
101: R1=F, R =CF3, R3=H
102: R1=F, R =H, R3=CF3
65
2
2
2
2
2
aReagents and conditions: (a) 36->39: BrCHFCOOEt, Zn, CuCl, THF; 36->90: BrCH2COOEt, Zn, CuCl, THF. (b) DIBAH,
toluene, ꢀ76 °C. (c) Trimethyl(trifluoromethyl)silane, THF, TBAF, 0 °C. (d) aq. HCl, THF, 22 °C. (e) BrCN, EtOH, 85 °C.
(f) H2SO4, HNO3, 0 °C. (g) H2, Pd/C, EtOH, Et3N, 22 °C. (h) DMTMM, MeOH, 0 °C to 22 °C.
RESULTS AND DISCUSSION
At the outset of the BACE project, we envisaged both BACE1 and BACE2 as viable targets for AD and
type 2 diabetes, respectively. The recently described oxazine 1a29, 36, 37 (for other 1,3ꢀoxazine BACE1
inhibitors see26, 27, 38) looked particularly suited for the diabetes target, which is primarily expressed in
peripheral tissue. Compound 1a showed a high enzymatic inhibitory activity at BACE2 (Table 1,
IC50=0.052 ꢀM) and was hardly brain penetrant as evident from the low brain/plasma ratio in mice (0.13
at 30 mg/kg) resulting in a marginal activity of 29% reduction of Aβ40 at 30 mg/kg in our wild type
mice model (vide infra) we use for the AD indication. An undesired potential liability of 1a, however,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 44
was the high basicity (pKa=9.8) giving rise to potential phospholipidosis39 and polypharmacology.40, 41
To overcome these liabilities, we introduced fluorines into the head group42 as depicted in difluoroꢀ
oxazine 14a.43 Indeed, this modification reduced the pKa by 3.5 log units compared to 1a and improved
the enzymatic potency at BACE1 and BACE2 5 and 3ꢀfold, respectively. Unexpectedly, however, 14a
readily penetrated the brain (brain/plasma ratio in mice=2.3 at 30 mg/kg) and therefore showed good in
vivo activity in mice with a 69% reduction of Aβ40 at 30 mg/kg. Obviously, this serendipitously found
effect of fluorinated oxazines now rendered this compound class much more appropriate for the AD
indication than type 2 diabetes.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1. In Vitro and In Vivo Profiles of Oxazines 1a, 14a and Thiazines 112, 113.
H2N
X
N
R
R
H
N
Cl
N
F
O
IC50 (ꢀM)a
hBACE2
enzyme
0.052
hBACE1
enzyme
0.137
0.026
0.016
Aβ40
shift BACE1
brain Aβ40 reduction in
compd
X
R
pKab
cellꢀbasedc enzyme vs. cellꢀbased
mice at 30 mg/kgd
1a
O
O
S
H
F
H
F
9.8
6.3
9.0
6.1
0.010
0.017
0.001
0.004
14
2
16
3
29%
69%
92%
14a
112
113
0.015
0.009
0.010
S
0.014
0%
aIC50 values are means of at least 2 independent experiments. pKa determined by capillary electrophoresis. HEK293 cells
transfected with wildꢀtype human APP. dCompound given orally, n=3ꢀ4, 4h post dose, mean reduction as compared to vehicle
control.
b
c
For comparison, the thiazine 11223 and its fluorine congener 11332, both described by Shionogi, were
also prepared. Interestingly, thiazine 112 was very potent in vivo (92% Aβ40 reduction at 30 mg/kg)
while the oxazine 1a was not. Most surprisingly, the fluorine congener 113, expected to be even more
potent than 112, was inactive in vivo. Another remarkable difference was the impact of basicity on the
shift between enzymatic and cellular BACE1 potency. Compounds 1a and 112 with high basicity
(pKa=≥9) showed a significantly higher cellular activity (14 and 16ꢀfold, respectively) whereas
compounds 14a and 113 with low basicity (<6.5) did not (2 and 3ꢀfold, respectively). The oxazine 14a
served as a basis for further optimization of the in vivo activity.
ACS Paragon Plus Environment

Page 11 of 44
Journal of Medicinal Chemistry
SAR of R1 substituents. Next we explored various R1 aromatic residues all having a nitrogen atom
1
2
3
4
5
in ortho position to the amide group (Table 2) to direct the aromatic residue into the S3 pocket (vide
infra).
6
7
8
9
Table 2. SAR of R1 Substituents and In Vivo Properties of Difluoroꢀoxazines 14a–v.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
R1 =
O
H2N
N
N
N
N
N
N
N
Cl
F
NC
MeO
N
F
F
H
N
R1
14a
14b
14c
14d
14e
14f
14g
F
O
F
F
F3C
F
N
N
N
N
N
N
N
Cl
Cl
F3CO
O
O
O
Cl
F
N
Cl
14h
14i
14j
14k
14l
14m
14n
F
F
F3C
F
F
N
N
N
N
MeO
O
O
N N
N N
N N
N N
O
N
N
N
N
Cl
Cl
14o
14p
14q
14r
14s
14t
14u
14v
IC50 (ꢀM)a
IC50 (ꢀM)a
Aβ40
cellꢀbasedb
0.017
0.230
0.013
0.021
0.087
0.024
0.061
0.054
0.480
0.290
0.100
0.016
0.013
0.023
0.022
0.110
0.001
0.001
0.110
0.131
0.017
0.013
hBACE1
enzyme
0.026
0.294
0.066
0.023
0.087
0.032
0.051
0.057
0.184
0.108
0.042
0.017
0.026
0.067
0.027
0.025
0.013
0.006
0.446
0.151
0.036
0.019
hBACE2
enzyme
0.015
0.036
0.022
0.024
0.030
0.047
0.045
0.245
0.904
7.490
9.440
0.010
0.012
0.023
0.119
0.943
0.175
0.101
0.106
0.031
0.018
0.011
selectivity
shift BACE1
enzyme vs. cellꢀbased
brain Aβ40 reduction in
mice at 30/10 mg/kgc
69% / 28%
ꢀ / ꢀ
compd
BACE2/1
0.6
0.1
0.3
1
14a
14b
14c
14d
14e
14f
14g
14h
14i
2
1
5
1
1
1
1
1
ꢀ3
ꢀ3
ꢀ2
1
2
3
ꢀ / ꢀ
84% / 78%
ꢀ / ꢀ
27% / ꢀ
64% / ꢀ
64% / ꢀ
ꢀ / ꢀ
0.3
1
1
4
5
14j
14k
14l
69
ꢀ / ꢀ
ꢀ / ꢀ
ꢀ / ꢀ
225
0.6
0.5
0.3
4
14m
14n
14o
14p
14q
14r
14s
14t
91% / 77%
ꢀ / ꢀ
95% / 66%
ꢀ / ꢀ
98% / 93%
94% / 28%
ꢀ / ꢀ
ꢀ / ꢀ
ꢀ / ꢀ
57% / ꢀ
1
38
13
17
0.2
0.2
0.5
0.6
ꢀ4
13
6
4
1
14u
14v
2
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 44
aIC50 values are means of at least 2 independent experiments. HEK293 cells transfected with wildꢀtype human APP
b
cCompound given orally, n=3ꢀ4, 4 h post dose, mean reduction as compared to vehicle control.
1
2
3
4
Out of a number of 6ꢀmembered heterocycles explored, pyridines, pyrimidines and pyrazines showed
promising in vitro activities. Pyridines 14a–k having one substituent in para position to the amide
showed enzymatic inhibition at IC50s values below 0.3 uM, most of them with good cellular activity. The
nitrile 14d displayed already an improved in vivo activity compared to the chloride 14a with Aβ40
reductions of 78 and 28%, respectively at a dose of 10 mg/kg. In addition, favorable mouse PK data for
14d (Cl=22 ml/min/kg, T1/2=1.8 h, Vss=2.9 L/kg, F=52%) encouraged us to explore further substitution
patterns. The oꢀ and pꢀdisubstituted pyridines 14l and 14m were virtually equipotent to the mono
substituted chloride 14a both in the enzymatic as well as the cellꢀbased assay, but 14m showed a higher
in vivo activity compared to 14a (77% vs. 28% reduction of Aβ40 at 10 mg/kg). The pyrimidine 14n
was 3ꢀfold less active in terms of enzymatic potency compared to the pyridine 14a. Pyrazines 14o–14r
showed good enzymatic inhibition activity (IC50=0.027ꢀ0.006 ꢀM), and examples 14o and 14p were
clearly more potent in the enzyme/cellꢀbased assays compared to their corresponding pyridine
counterparts 14g (2/3ꢀfold) and 14j (4/3ꢀfold), respectively. Interestingly, the propynyloxyꢀpyrazine 14q
was very potent in vivo (93% reduction of Aβ40 at 10 mg/kg). Out of a number of 5ꢀmembered aromatic
heterocycles explored, pyrazoles looked particularly promising. A first example, the methyl pyrazole
14s, was only moderately active in the enzyme/cellꢀbased assays (IC50=0.446/0.110 ꢀM) and showed
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
very low metabolic stability in mouse microsomes (Cl=200 ꢀL/min/mg protein). Potential labile
positions, i.e. the methyl group and position 4 of the pyrazole, were therefore blocked by fluorines and a
chlorine substituent, respectively. As expected, metabolic stability in mouse microsomes improved
considerably in 14t-v (Cl=98, 59 and 9 ꢀL/min/mg protein, respectively) and higher enzymatic potencies
were also observed (IC50=0.151, 0.036 and 0.019 ꢀM, respectively). Disappointingly, however, the in
vivo activity of 14v remained rather modest, with a 57% reduction of Aβ40 at 30 mg/kg. The shift
between enzyme and cellꢀbased activity was in the majority of examples rather small (6ꢀfold and less,
ACS Paragon Plus Environment

Page 13 of 44
Journal of Medicinal Chemistry
with the exception of a 13ꢀfold shift of the alkyne derivative 14q) and in general in favor of higher
cellular activity. The (fluoro)alkoxy derivatives 14iꢀ14k and 14p, however, were found to have
“negative” shifts (2 to 4ꢀfold) in favor of higher enzymatic potencies clearly demonstrating that not only
the pKa but also the nature of R1 substituents had an impact on the shift.
1
2
3
4
5
6
7
8
9
X-ray structures and BACE2 selectivity. The catalytic aspartate residues D93 and D289 form the
central recognition motif in the BACE1 binding site (Figure 2). The active site is partially covered by a
flexible hairpin loop commonly referred to as the “flap” (Y129 ꢀ E138).44 In apo structures of BACE1 or
in complexes with small molecule inhibitors containing an amidine binding motif, the flap backbone is
approximately 12 Å away from the sideꢀchains of D93 and D289 (the “open” position). In complexes
with peptidomimetic transition state analogs the flap moves approx. 4 Å closer to the binding site (the
“closed” position).44 There is an additional flexible region, the “10s loop”44 (G69 – G74) forming one
side of the S3 pocket. In BACE1 the 10s loop is either in an “up” or “down” conformation,45 whereas in
BACE2 the 10s loop has been observed in a down conformation by Ostermann et al.46 and by ourselves
in the complex with 14i (Figure 3). The S3 pocket is much larger in the up conformation, so this
difference can be exploited to modulate BACE1/2 selectivity as well as selectivity against other single
chain aspartyl proteases.44 The binding pocket in S1 and the region at the catalytic aspartates are almost
identical between BACE1 and BACE2. Conformational variation around the active site of BACE1 has
recently been discussed in detail by Xu et al:47 we will analyze BACE2 flexibility in detail separately.48
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 44
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2. 14d bound to the BACE1 binding site. The amidine
recognition motif is tightly bound to the catalytic aspartates
D289 and D93 (partially hidden) in the bottom of the pocket.
The fluorine atoms in the head group and the oꢀF substituent of
the phenyl ring (marked in green) form a network of vdW
interactions to Y132 and to F169. An additional Hꢀbond is
established between the NH of the amide bond of the ligand
and the backbone carbonyl oxygen of G291. The para nitrile
substituent displaces bound water molecules from the S3
pocket.
In Figure 2 the topology of the BACE1 active site bound to 14d is highlighted. Tight hydrogen bonds
are formed between the protonated amidine motif in the oxazine head group and the catalytic aspartates.
The fluorine atoms in position 5 and the oꢀF substituent of the phenyl ring (S1 pocket) are engaged in a
network of vanꢀderꢀWaals interactions to Y132 located in the flap and to F169 at the rim of the S1
pocket. A hydrogen bond is formed between the NH of the amide bond in 14d and the backbone
carbonyl oxygen of G291. The para nitrile substituent projects into the S3 pocket and displaces bound
water molecules seen in the apo structure, thereby enhancing binding affinity. The proper orientation in
the S3 pocket is supported by a stabilization of the conformation of the pyridine ring via a favorable
electrostatic interaction between the pyridyl N and the NH of the amide bond in 14d.
ACS Paragon Plus Environment

Page 15 of 44
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Figure 3. 14i bound to BACE1 (yellow, ball and stick
representation) and BACE2 (magenta, stick representation).
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The molecular surface has been generated for BACE1, the 10s
loops of BACE1 and BACE2 are depicted separately. The
ethoxy sidechain in 14i prefers a planar orientation around the
CꢀO bond. When bound to BACE2 the corresponding torsion
angle deviates significantly from planarity, since the S3 pocket
in BACE2 is tighter than in BACE1 due to the “down”
conformation of the 10s loop in BACE2.
The role of the 10s loop in modulating BACE1/2 selectivity can be seen in the ratios of inhibition
constants of the compounds 14g-k (Table 2). As it is evident in Figure 3, the S3 pocket in BACE2 is
smaller than in BACE1, because the 10s loop in BACE2 is observed here in a down conformation. The
IC50 ratio between BACE2 and BACE1 increases with the growing chain length of the S3 alkyl
substituent. Longer chain lengths can only be accommodated in the BACE2 S3 pocket by adopting an
unfavorable ligand conformation as depicted in Figure 3. The ethoxy sidechain in 14i prefers a planar
orientation around the CꢀO bond. When bound to BACE2 the corresponding torsion angle deviates
significantly from planarity, and this conformational strain translates into a reduced inhibition of
BACE2. A similar rationale applies for the increase in selectivity with an increasing number of F
substituents. In order to avoid repulsive contacts between Fꢀsubstituents and carbonyl oxygen atoms of
the protein, the tighter S3 pocket in BACE2 requires more strained conformations of the S3 substituent
than the more spacious pocket in BACE1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 44
Fluorine scan of oxazines. From various R1 aromatic residues explored, we selected the nitrile 14d
for further optimization based on its balanced profile, i.e. good in vitro activities, no shift between
enzymatic and cellꢀbased assay and a favorable in vivo activity. The crucial role of the pKa on the
pharmacological profile as found in the difluoroꢀoxazine 14a prompted us to investigate extensively the
impact of fluorines on various parameters. Oxazines with fluorines appended at different positions in the
head group (Table 3) were prepared spanning a pKa range of almost five log units. The head groups in
Table 3 are ordered according to decreasing pKa values. Below we discuss the impact of fluorines on the
pKa, the lipophilicity, the enzymatic potency, the shift of enzyme vs. cellꢀbased activity, Pꢀgp efflux and
hERG inhibition, most of these parameters eventually affecting in vivo activity.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 3. Head Group Fluorine Scan and In Vitro/In Vivo Properties of Fluorinated Oxazines.
O
O
O
O
O
O
H2N
CF3
H2N
H2N
H2N
CF3 H2N
H2N
N
N
N
N
N
N
O
H2N
F
F
F
F
F
N
F
H
N
NC
66
67
69
109
89
68
N
F
O
H2N
CF3
O
O
O
H2N
H2N
H2N
CF3
F
O
H2N
CF3
F
O
1b
N
N
N
F
F
N
F
N
F
F
F
111
110
88
14d
70
IC50 (ꢀM)a
hBACE1 Aβ40
ꢁ
hERG
brain Aβ40
ꢁ
shift BACE1
Pꢀgp
compd pKab pKa logDc logP
inhibition reduction in mice
logD/logP enzyme cellꢀbasedd enzyme vs. cellꢀbased ERe
to 1b
at 10ꢀM at 10/3/1 mg/kgf
1b
66
67
69
9.7
0
0.0 2.3
2.3
0.8
0.3
0.2
0.3
0.1
0.1
0
0
0
0
0
0.051
0.028
0.435
0.054
0.776
0.012
0.077
0.019
0.013
0.040
0.023
0.049
0.014
0.001
0.027
0.009
0.230
0.002
0.025
0.013
0.010
0.150
0.021
0.400
4
28
16
6
3
6
3
1
1
ꢀ4
1
14.8
4.9
2.1
1.7
ꢀ
1.9
2.6
1.6
1.5
1.5
1.5
1.6
88% 13% /
85% 89% / 43% /
91% 2% /
85% 68% / 34% /
ꢀ
/
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
8.1 ꢀ1.6 1.1 1.9
7.4 ꢀ2.3 1.3 1.6
7.3 ꢀ2.4 1.7 1.9
ꢀ
/
109 7.3 ꢀ2.4 2.8 3.1
ꢀ
ꢀ
/
ꢀ
/
89
68
88
7.0 ꢀ2.7 2.8 2.9
6.7 ꢀ3.0 1.6 1.7
6.3 ꢀ3.4 3.0 3.0
75% 95% / 88% / 78%
65% 55% / 26% /
39% 88% / 64% / 11%
48% 93% / 71% / 41%
ꢀ
111 5.9 ꢀ3.8 3.1 3.1
110 5.8 ꢀ3.9 3.1 3.1
14d 5.8 ꢀ3.9 2.5 2.5
ꢀ
ꢀ
/
ꢀ
/
ꢀ
43% 78% / 39% / 4%
14%
70
5.1 ꢀ4.6 2.3 2.3
ꢀ8
ꢀ
/
ꢀ
/
ꢀ
aIC50 values are means of at least 2 independent experiments. pKa determined by capillary electrophoresis. LogD
b
c
d
e
determined in 1ꢀoctanol/phosphate buffer, pH=7.4. HEK293 cells transfected with wildꢀtype human APP. ER=efflux ratio
ACS Paragon Plus Environment

Page 17 of 44
Journal of Medicinal Chemistry
f
in LLCꢀPK1cells stably expressing human MDR1. Compound given orally, n=3ꢀ4, 4 h post dose, mean reduction as
compared to vehicle control.
1
2
3
4
5
6
7
8
The configuration of a fluorine atom attached directly to the oxazine ring has a strong impact on the
pKa. Thus, the “down” configured fluorine in 66 showed a pKa decrement of 1.6 (ꢁpKa of 1b and 66)
whereas the “up” configured fluorine in 67 induced a larger pKa decrement of 2.3. Similar effects were
also observed in fluoroꢀpiperidines.49, 50 The difluoro oxazine 14d showed exactly the sum of both
contributions, i.e. a pKa decrement of 3.9. The two fluorines in compound 69,37, 51 also located at the βꢀ
position to the ring nitrogen, induced a lower pKa decrement of 2.4 compared to those of difluoro
oxazine 14d. It appears that fluorines with limited conformational flexibility as present in 14d induced a
larger pKa decrement compared to those in 69 having more flexibility. The CF3 group in compounds
109 and 89 still showed a remarkably large pKa decrement of 2.4 and 2.7, respectively, despite the
longer distances through which the fluorines exert their impact on the nitrogens. This observation can be
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
rationalized based on the equidistance of the fluorines to both nitrogens in the δꢀposition having an
impact not just on one but on both nitrogens of the amidine moiety. Fluorines located at more than one
position (as in 68, 88, 111, 110 and 7051) showed a more complex picture with pKa decrements not
being additive.
The lipophilicity of compounds having 1ꢀ3 fluorines attached to the head group is reduced or similar
(logP=1.6–2.5 in 66-70 and 14d) compared to the nonꢀfluorinated oxazine head group 1b (logP =2.3).
Compounds substituted by a CF3 group showed increased lipophilicity (logP=2.9 – 3.1 in 88, 89, 109–
111) compared to 1b. As expected, the ꢁ logP/D approaches a value of 0 at a pKa of 6.3.
The enzymatic potency was strongly affected by fluorines introduced into the oxazine head group. A
down configured CF3 group in 89 improved the binding affinity considerably (IC50=0.012 ꢀM)
compared to unsubstituted 1b (IC50=0.051 ꢀM). According to the Xꢀray structure, the CF3 group does
not interact with BACE1 but points towards the solvent. A possible rationale for this unexpected
increase in binding affinity might be the pKa of 89, which is 2.7 log units lower than for 1b. Therefore,
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 44
1b should be more strongly solvated than 89 due to its higher basicity. The lower binding affinity of 1b
compared to 89 could be rationalized by the higher desolvation enthalpy upon binding to the active site.
The configurations of the fluorine in 66 and the CF3 group in 89 strongly affected the binding affinity,
which increased 16ꢀfold and 65ꢀfold, compared to their epimers 67 and 109, respectively. This effect can
be rationalized by conformational strain of the head groups. The head groups of the Xꢀray structures
were submitted to a semiempirical geometry optimization52 (PM3 hamiltonian) outside the binding
pocket (unbound geometry). As shown in Figure 4, the coordinates of bound as well as unbound 66 and
89 remain almost unchanged, therefore, the oxazine rings are without conformational strain. The
geometry optimization of 109 resulted in a different ring pucker, indicating that the bound ligand is not
in its lowest energy conformation. This conformational strain results in a reduced binding affinity. The
geometry optimization of 110 also yields a strained conformation in the bound state. However, the effect
on binding affinity is much lower than expected. A reason for this result might be the involvement of the
Fꢀsubstituent in the network of vanꢀderꢀWaals interactions involving Y132, which could
overcompensate the conformational energy penalty. For 67 the position of the methylene carbon next to
the oxazine oxygen atom cannot be clearly assigned because of weak electron density. This is probably
due to the presence of multiple ring conformations indicating that 67 is not well bound in the active site.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. Xꢀray structures (yellow) of fluorinated head groups with the S1 phenyl ring were energy minimized in the unbound
state by a semiempirical geometry optimization with the PM3 hamiltonian (magenta). The coordinates of 66 and 89 remain
almost unchanged as compared to the bound conformation. Therefore, the oxazine rings are without conformational strain.
The geometry optimization of unbound 109 and 110 resulted in different ring puckers indicating that those ligands are not in
their lowest energy conformations. For 67 the position of the marked Cꢀatom cannot be clearly assigned because of weak
electron density. This indicates the presence of multiple ring conformations.
ACS Paragon Plus Environment

Page 19 of 44
Journal of Medicinal Chemistry
The shift of cellular vs. enzymatic activity depends on the pKa: basic compounds with a pKa ≥7
1
2
show a trend for higher activity in the cellꢀbased assay whereas oxazines with a pKa <7 are more
balanced or can even revert to have a higher potency in the enzymatic assay.53 BACE1 inhibitors with
enhanced basicity have been associated with increased cellular potency by favoring the acidic
intracellular compartment where BACE1 is located.54
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Pꢀgp mediated efflux is clearly correlated with the pKa.55 Lowering of the pKa from 9.7 to 8.1 to 7.4
in compounds 1b, 66 and 67 resulted in a continuous reduction of the efflux ratio from 15 to 5 to 2,
respectively, which remained low at a pKa <7.4. There was in general no species differences between
human and mouse Pꢀgp efflux, e.g. for compound 89 (efflux ratio=1.9/1.9, respectively) and compound
14d (efflux ratio=1.5/2.3, respectively).
hERG inhibition is strongly affected by the pKa.56 Reduction of the pKa from 9.7 to 5.1 correlated
well with hERG inhibition which declined from 88 to 14% at 10 ꢀM.
The low in vivo activity of compound 1b (13% reduction of Aβ40 at 10 mg/kg) is now better
understood because of the high Pꢀgp efflux ratio of 15. Attachment of one fluorine in the relaxed down
configuration in compound 66 boosted the cellular activity 14ꢀfold and reduced Pꢀgp efflux 3ꢀfold as
compared to 1b resulting in good in vivo activity (43% reduction of Aβ40 at a lower dose of 3 mg/kg).
The cellular activity of compounds 88 and 111 (IC50=0.013 and 0.010 ꢀM) was lower compared to 66
(IC50=0.001 ꢀM) but the 3ꢀfold lower Pꢀgp efflux of 1.6/1.5 may account for the higher in vivo activity
(64 and 71% reduction of Aβ40 at 3 mg/kg) compared to 66. Most gratifyingly, the CF3 substituted
oxazine 89 with combined favorable properties (cellular IC50=0.002 ꢀM, Pꢀgp efflux of 1.9) was most
potent showing a reduction of Aβ40 of 78% at a dose as low as 1 mg/kg.
In vivo Pharmacokinetics (PK) and Pharmacodynamics (PD) of Oxazine 89. The oxazine 89 was
selected for further investigation of pharmacokinetic and dynamic parameters. Aβ40 lowering was
determined after oral administration in wildꢀtype mice (C57Bl/6J mice, n=3ꢀ4) 4 h post dosing. Wildꢀ
type mice were used as they have a normal physiological level of APP expression and are thus expected
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 44
to provide a better physiological model for Aβꢀpeptide production than APPꢀtransgenic organisms.
1
2
3
4
5
6
7
8
9
Compound 89 showed a doseꢀdependent inhibition of Aβ40 production in the brain of mice (Figure 5a)
with a brain/plasma ratio of 1ꢀ1.5 between 0.3 and 3 mg/kg gradually increasing to 2.3 and 3.4 at the
higher doses of 10 and 30 mg/kg. In a rat PK experiment (Table 4), oxazine 89 showed a low to
moderate systemic plasma clearance (16.8 ml/min/kg), a high volume of distribution (11.5 L/kg) and a
long terminal halfꢀlife time (9ꢀ11 h). This, together with a good oral bioavailability (68%) and a
moderately high fraction unbound (20.5% in plasma) would predict a favorable PD effect in rats.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 4. Pharmacokinetic Properties of Oxazine 89
in Rats.a
dose
AUCinf
Cl
VdSS
t1/2
F
Fuc
.
(mg/kg) (ng h/mL) (ml/min/kg) (L/kg) (h) (%) (%)
1.2 mg ivb
5.2 mg pob
1195
3520
16.8
ꢀ
11.5
ꢀ
2
8.9
ꢀ
ꢀ
10.9 68 20.5
aAll values given are means of
experiments. bPo:
Microsuspension of 89 in gelatine/aqueous NaCl; Iv: Solution of
89 in aqueous NaCl/Nꢀmethylꢀ2ꢀpyridone. cFu=fraction unbound.
Wildꢀtype Wistar rats were used with a cannula inserted into the Cisterna magna and the kinetics of
changes in the CSF Aβ40/42 in an acute treatment study was followed.57 An oral dose as low as 1 mg/kg
(Figure 5b) was sufficient to induce a long lasting effect, with CSF levels remaining most inhibited 4 h
post dosing and still showing a significant reduction of Aβ40/42 after 24 h as a result of favorable PK
properties.
The changes in plasma Aβ40/42 (Figure 5b) were analyzed in relation to the measured plasma exposure
in rats (Table 4). For the modeling of the plasma exposure data, a 1ꢀcompartment model with firstꢀorder
drug absorption and elimination was considered. For the modeling of Aβ40/42 changes in plasma, an
indirect response model was considered with the zeroꢀorder production rate being modulated by the
pharmacological effects of the oxazine 89. The proposed model fitted the data well and the in vivo
potency was estimated to be 0.020 ꢀM free, which compares favorably to the in vitro measurement of
IC50=0.012 ꢀM in the hBACE1 enzyme assay. Thus, it can be concluded that the free plasma
concentration of 89 equals the free brain concentration, which can be considered as the pharmacological
relevant concentration.
ACS Paragon Plus Environment

Page 21 of 44
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Figure 5. (a) Dose dependent inhibition of brain Aβ40 production in mice by compound 89 determined 4 h oral post dose,
n=3ꢀ4, mean reduction as compared to vehicle control. (b) Time dependent inhibition of CSF Aβ40/42 production in rats
after an oral dose of 1 mg/kg of compound 89, n= 2ꢀ3. CSF was sampled at ꢀ1 and 0 h to define an average Aβ40/42 level for
each individual rat; all postꢀtreatment Aβꢀlevels were expressed as percentage of this pretreatment average.
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
In terms of its in vitro toxicological profile, compound 89 did not inhibit CYP450 3A4, 2D6 and 2C9
(IC50 > 25ꢀM) nor was there a timeꢀdependent inhibition of CYP450 3A4. Furthermore, 89 was very
selective against other aspartyl proteases such as human Cathepsin D/E and the peptidases renin and
pepsin (IC50 > 200ꢀM).
SUMMARY AND CONCLUSION
Starting from a very basic, in vivo marginally active, 1,3ꢀoxazine 1a, the introduction of two fluorine
atoms into the head group reduced the pKa considerably and led unexpectedly to good in vivo activity,
as found in compound 14a. Encouraged by this finding, we performed an extensive fluorine scan to
study the impact of fluorine(s) on the pKa and a number of other parameters. Lowering the pKa had a
profound effect on enzymatic and cellular activity, hERG activity, Pꢀgp efflux and eventually on in vivo
activity. The knowledge acquired by the fluorine head group scan delivered crucial information on how
to improve in vivo efficacy and reduce in vitro toxicological liabilities. The CF3ꢀsubstituted oxazine 89
was found to be an orally active BACE1 inhibitor with an excellent PD profile in mice and a long lasting
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 44
Aβ40/42 reduction in the CSF of rats at a dose as low as 1 mg/kg. A compound from this chemical class
1
2
3
is currently undergoing clinical evaluation for the treatment of Alzheimer’s Disease.
4
5
6
7
8
EXPERIMENTAL SECTION
9
General. All solvents and reagents were obtained from commercial sources and were used as received.
All reactions were followed by TLC (TLC plates F254, Merck) or LCMS (liquid chromatographyꢀmass
spectrometry) analysis. Silica gel chromatography was either performed using cartridges packed with
silica gel (ISOLUTE® Columns, TELOSTM Flash Columns) or silicaꢀNH2 gel (TELOSTM Flash NH2
Columns) on ISCO Combi Flash Companion or on glass columns on silica gel 60 (32ꢀ60 mesh, 60 Å).
Proton NMR spectra were obtained on Bruker Avance 300, 400 or 600 MHz spectrometer with chemical
shifts (δ in ppm) reported relative to tetramethylsilane or the residual solvent peak as the internal
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
reference (i.e. CDCl3 = 7.26 ppm, DMSOꢀd6 = 2.50 ppm) as the internal reference. H resonances are
reported to the nearest 0.01 ppm. NMR abbreviations are as follows: s, singlet; d, doublet; t, triplet; q,
quadruplet; quint, quintuplet; sext, sextuplet; m, multiplet; br, broadened. Coupling constants (J) are
reported to the nearest 0.1 Hz. Purity was analyzed by reverse phase HPLC and for specific compounds
by elemental analysis. HPLC was performed on Finnigan LTQ (Thermo Fisher Scientific) and Agilent
RRLC 1200 equipment. Column: Agilent XDB C15, 30 mm x 4.6 mm, 3.5 ꢁm. Analytical conditions:
gradient used: 5% to 95% acetonitrile in water containing 0.1% trifluoroacetic acid in 3 min. Flow: 4.5
mL/min. UVꢀDetector: DAD 190ꢀ400 nm. Sample solvent: in water/acetonitrile (8/2). The UV detection
was an averaged signal from wavelengths of 190ꢀ400 nm. Elemental analyses were performed by
Solvias AG (Mattenstrasse, Postfach, CHꢀ4002 Basel, Switzerland). Mass spectra were recorded on an
SSQ 7000 (FinniganꢀMAT) spectrometer for electron impact ionization. The purities of final test
compounds as measured by HPLC were found to be above 95%. Melting points were determined on a
Büchi Melting Point Bꢀ540: heating rate 1 °C/min starting 15 °C below melting point. LCꢀHRMS
spectra were recorded with an Agilent LCꢀsystem consisting of an Agilent 1290 high pressure system, a
CTC PAL auto sampler and an Agilent 6520 QTOF. The separation was achieved on a Zorbax Eclipse
ACS Paragon Plus Environment

Page 23 of 44
Journal of Medicinal Chemistry
Plus C18 1,7 ꢂm 2.1*30mm column at 55 °C; eluent A=0.01% formic acid in Water; B= 0.01% formic
acid in acetonitrile : 2ꢀpropanol (8:2); flow: 1 mL/min. Gradient: 0 min 5% of B, 0.3 min 5% of B, 4.5
min 99 % of B, 5 min 99% of B. The injection volume was 2 ꢂL. All assay compounds had a purity of
>95%. LCꢀMS (ESI, positive or negative ion) data were recorded on Waters UPLCꢀMS Systems
equipped with Waters Acquity, a CTC PAL auto sampler and a Waters SQD single quadrupole mass
spectrometer using ES ionization modes (positive and/or negative). The separation was achieved on a
Zorbax Eclipse Plus C18 1,7 ꢂm 2.1*30mm column at 50 °C; A=0.01% formic acid in water, B=
acetonitrile at flow 1; gradient: 0 min 3%B, 0.2 min 3%B, 2 min 97 %B, 1.7 min 97%B, 2.0 min 97%B.
The injection volume was 2 ꢂL. MS (ESI, positive or negative ion): FIA (flow injection analysis)ꢀMS
were recorded on a AppliedBiosystem API150 mass spectrometer. Sample introduction was made with a
CTC PAL auto sampler and a Shimadzu LCꢀ10ADVP Pump. The samples were directly flushed to the
ESI source of the mass spectrometer with a flow 50 ꢂL/min of a mixture of acetonitrile and 10 mM
ammonium acetate (1:1) without a column. The injection volume was 2 ꢂL.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(R)-2-Methyl-propane-2-sulfinic acid [1-(2-fluoro-5-nitro-phenyl)-(E)-ethylidene]-amide (3).32
To a solution of the (R)ꢀ(+)ꢀtertꢀbutylsulfinamide (10.88 g, 89.8 mmol) in THF (400 mL) was added
subsequently 1ꢀ(2ꢀfluoroꢀ5ꢀnitroꢀphenyl)ꢀethanone (2) (16.45 g, 89.8 mmol) and titanium(IV)ethoxide
(40.70 g, 178.4 mmol) and the solution was stirred at reflux temperature for 5 h. The mixture was
cooled to 22 °C, treated with brine (400 mL), the suspension was stirred for 10 min and filtered over
Dicalite®. The layers were separated, the aqueous layer was extracted with EtOAc, the combined organic
layers were washed with water, dried and evaporated. The residue was purified by flash column
chromatography on silica gel using a mixture of cyclohexane/EtOAc (2:1) to give the title compound 3
1
(21.56 g, 84%) as a pale yellow solid: m.p. 83ꢀ85°C. H NMR (600 MHz, CDCl3) δ ppm 8.57 (dd,
J=6.3, 2.9 Hz, 1 H), 8.33 (ddd, J=9.0, 3.2, 3.1 Hz, 1 H), 7.30 (t, J=9.5 Hz, 1 H), 2.82 (d, J=3.5 Hz, 3 H),
1.34 (s, 9 H). MS (ISP): m/z = 287.0 [M+H]+. Anal. (C12H15FN2O3S) calcd C 50.34%, H 5.28%, N
9.78%, S 11.20%; found C 50.49%, H 5.02%, N 9.94%, S 11.31%.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 44
(R)-2,2-Difluoro-3-(2-fluoro-5-nitro-phenyl)-3-((R)-2-methyl-propane-2-sulfinylamino)-butyric
acid ethyl ester (5).32 In a dry apparatus a suspension of freshly activated zinc powder (37.5 g, 574
mmol) and CuCl (5.99 g, 60.5 mmol) in dry THF (100 mL) was heated under an inert atmosphere to
reflux for 30 min. The suspension was cooled to 22 °C treated with a solution of ethyl 2ꢀbromoꢀ2,2ꢀ
difluoroacetate (30.7 g, 151 mmol) in dry THF (50 mL) over 1 h, and stirring was continued for 30 min.
The black mixture was treated with a solution of (R)ꢀ2ꢀmethylꢀpropaneꢀ2ꢀsulfinic acid [1ꢀ(2ꢀfluoroꢀ5ꢀ
nitroꢀphenyl)ꢀ(E)ꢀethylidene]ꢀamide (3) (17.5 g, 60.5 mmol) in THF (50 mL), and stirring was continued
at 22 °C for 2h. The suspension was filtered through Dicalite® and washed with tꢀbutyl methyl ether.
The combined filtrates were subsequently washed with aqueous citric acid (5%) and aqueous saturated
NaHCO3 solution, the organic layer was dried over Na2SO4, filtered and the filtrate was evaporated. The
residue was purified by flash column chromatography on silica gel using a mixture of nꢀheptane/EtOAc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
(2:1) to give the title compound 5 (17.24 g, 69%) as a yellow oil. H NMR (300 MHz, CDCl3) δ ppm
8.51 (dd, J=6.6, 2.7 Hz, 1 H), 8.29 (ddd, J=9.1, 3.8, 3.0 Hz, 1 H), 7.26 (dd, J=11.3, 9.1 Hz, 1 H), 4.72 (s,
1 H), 4.34 (qd, J=7.2, 1.5 Hz, 2 H), 2.09ꢀ2.17 (m, 3 H), 1.32 (t, J=7.3 Hz, 3 H), 1.32 (s, 9 H). MS (ISP):
m/z = 411.2 [M+H]+. Anal. (C16H21F3N2O5S) calcd C 46.82%, H 5.16%, N 6.83%, S 7.81%; found C
46.73%, H 4.89%, N 6.81%, S 7.66%. The configuration of the newly introduced stereogenic center was
confirmed by Xꢀray crystallographic analysis of the target compound 14d complexed with BACE1
(3ZMG).
(R)-2-Methyl-propane-2-sulfinic acid [(R)-2,2-difluoro-1-(2-fluoro-5-nitro-phenyl)-3-hydroxy-1-
methyl-propyl]-amide (7).32 To a solution of (R)ꢀ2,2ꢀdifluoroꢀ3ꢀ(2ꢀfluoroꢀ5ꢀnitroꢀphenyl)ꢀ3ꢀ((R)ꢀ2ꢀ
methylꢀpropaneꢀ2ꢀsulfinylamino)ꢀbutyric acid ethyl ester (5) (3.78 g, 9.2 mmol) in dry THF (24 mL)
was added at 0 °C lithium borohydride (9.0 mmol), and stirring was continued at 0 °C for 1.5 h. The
reaction mixture was quenched by addition of acetic acid (1 mL) and water, the mixture was extracted
with EtOAc, the organic layers were washed with brine, dried over Na2SO4 and evaporated. The residue
was purified by flash column chromatography on silica gel using a mixture of nꢀheptane/EtOAc (3:2) to
give the title compound 7 (2.90 g, 86%) as a pale yellow solid: 1H NMR (600 MHz, CDCl3) δ ppm 8.46
ACS Paragon Plus Environment

Page 25 of 44
Journal of Medicinal Chemistry
(dd, J=6.8, 2.8 Hz, 1 H), 8.25 (ddd, J=9.0, 3.7, 2.9 Hz, 1 H), 7.25 (dd, J=11.7, 9.0 Hz, 1 H), 5.36 (br s, 1
H), 4.51–5.03 (br s, 1 H), 4.21 (ddd, J=25.0, 13.4, 2.7 Hz, 1 H), 3.84 (td, J=12.8, 7.9 Hz, 1 H), 2.19 (s, 3
H), 1.31 (s, 9 H). MS (ISP): m/z = 369.0 [M+H]+. A small sample was recrystallized from EtOAc/nꢀ
heptane: m.p. 154ꢀ155 °C. Anal. (C14H19F3N2O4S) calcd C 45.65%, H 5.20%, N 7.60%, S 8.70%; found
C 45.60%, H 5.07%, N 7.79%, S 8.78%.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(R)-3-Amino-2,2-difluoro-3-(2-fluoro-5-nitro-phenyl)-butan-1-ol (9).32 A solution of (R)ꢀ2ꢀ
methylꢀpropaneꢀ2ꢀsulfinic
acid
[(R)ꢀ2,2ꢀdifluoroꢀ1ꢀ(2ꢀfluoroꢀ5ꢀnitroꢀphenyl)ꢀ3ꢀhydroxyꢀ1ꢀmethylꢀ
propyl]ꢀamide (7) (3.79 g, 10.3 mmol) in MeOH (50 mL) was treated with a solution of HCl in 1,4ꢀ
dioxane (4 M, 13 mL), and stirring was continued at 22 °C for 1 h. The mixture was evaporated, the
residue partitioned between EtOAc and aqueous Na2CO3ꢀsolution (2 M), the organic layers were washed
with brine, dried over Na2SO4, filtered and evaporated to give the crude title compound 9 (2.50 g, 92%)
as a pale yellow solid which was used without further purification . 1H NMR (600 MHz, CDCl3) δ ppm
8.49 (dd, J=7.0, 2.8 Hz, 1 H), 8.25 (ddd, J=9.0, 3.8, 2.9 Hz, 1 H), 7.25 (dd, J=11.5, 9.0 Hz, 1 H), 4.02
(ddd, J=19.7, 12.7, 8.6 Hz, 1 H), 3.86 (dd, J=23.5, 11.8 Hz, 1 H), 2.01–3.63 (3 H), 1.85 (d, J=0.9 Hz, 3
H). MS (ISP): m/z = 265.1 [M+H]+. A small sample was triturated with 2ꢀpropanol: m.p. 110ꢀ111 °C.
Anal. (C10H11F3N2O3) calcd C 45.46%, H 4.20%, N 10.60%; found C 45.67%, H 4.20%, N 11.03%.
(R)-5,5-Difluoro-4-(2-fluoro-5-nitro-phenyl)-4-methyl-5,6-dihydro-4H-[1,3]oxazin-2-ylamine
(11).33 To a solution of (R)ꢀ3ꢀaminoꢀ2,2ꢀdifluoroꢀ3ꢀ(2ꢀfluoroꢀ5ꢀnitroꢀphenyl)ꢀbutanꢀ1ꢀol (9) (1.5 g, 5.7
mmol) in EtOH (30 mL) at 22 °C was added under argon cyanogen bromide (0.91 g, 8.6 mmol), and the
light yellow solution was stirred in a sealed tube at 85 °C for 24 h. The mixture was cooled to 22 °C,
evaporated and the residue partitioned between EtOAc and saturated aqueous Na2CO3ꢀsolution. The
organic layers were dried over Na2SO4, filtered, evaporated and the residue was purified by flash column
chromatography on silica gel using a mixture of nꢀheptane/EtOAc (5:2) to afford the title compound 11
1
(1.30 g, 79%) as a pale yellow solid: H NMR (600 MHz, CDCl3) δ ppm 8.59 (dd, J=6.6, 2.9 Hz, 1 H),
8.21 (ddd, J=9.0, 3.8, 3.0 Hz, 1 H), 7.20 (dd, J=10.6, 8.9 Hz, 1 H), 4.25–4.45 (2 H), 4.27 (ddd, J=23.2,
11.8, 3.8 Hz, 1 H), 4.06 (ddd, J=11.7, 10.4, 8.2 Hz, 1 H), 1.79 (t, J=2.3 Hz, 3 H). MS (ISP): m/z = 290.2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 44
[M+H]+. A small sample was recrystallized from EtOAc/nꢀheptane: m.p. 134ꢀ135 °C. Anal.
(C11H10F3N3O3) calcd C 45.68%, H 3.49%, N 14.53%; found C 45.63%, H 3.41%, N 14.65%.
(R)-4-(5-Amino-2-fluoro-phenyl)-5,5-difluoro-4-methyl-5,6-dihydro-4H-[1,3]oxazin-2-ylamine
(13).43 To a solution of (R)ꢀ5,5ꢀdifluoroꢀ4ꢀ(2ꢀfluoroꢀ5ꢀnitroꢀphenyl)ꢀ4ꢀmethylꢀ5,6ꢀdihydroꢀ4Hꢀ
[1,3]oxazinꢀ2ꢀylamine 11 (1.29 g, 4.47 mmol) in EtOH (35 mL) was added at 22 °C under an inert
atmosphere palladium on carbon (10% Pd, 238 mg) and the mixture was stirred under a hydrogen
atmosphere (balloon) at 22 °C for 1 h. The catalyst was filtered off, washed twice with EtOH, the
solvent was removed under reduced pressure to give the crude title compound 15 (1.14 g, 98%) as a
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
colorless foam which was used without further purification. H NMR (600 MHz, CDCl3) δ ppm 6.85
(dd, J=11.7, 8.6 Hz, 1 H), 6.79 (dd, J=6.5, 3.0 Hz, 1 H), 6.58 (ddd, J=8.6, 3.2, 3.1 Hz, 1 H), 4.10–4.90
(br s, 2 H), 4.10 (ddd, J=13.2, 11.5, 7.5 Hz, 1 H), 3.99 (ddd, J=18.2, 11.6, 5.9 Hz, 1 H), 3.55 (br s, 2 H),
1.75 (t, J=2.3 Hz, 3 H). MS (ISP): m/z = 260.1 [M+H]+. A small sample was purified by flash column
chromatography on silica gel using a mixture of dichloromethane/methanol (19:1) containing 1% of
ammonium hydroxide, followed by crystallization from tertꢀbutylmethyl ether: m.p. 122ꢀ123 °C. Anal.
(C11H12F3N3O) calcd C 50.97%, H 4.67%, N 16.21%; found C 51.03%, H 4.52%, N 16.34%.
General Procedure for the synthesis of the amides 14a and 14d. To a solution of the carboxylic
acid (0.23 mmol) in MeOH (5 mL) at 0 °C was added 4ꢀ(4,6ꢀdimethoxy[1.3.5]triazinꢀ2ꢀyl)ꢀ4ꢀ
methylmorpholinium chloride hydrate (DMTMM) (80 mg, 0.27 mmol). The colorless solution was
stirred at 0 °C for 30 min, then a solution of (R)ꢀ4ꢀ(5ꢀaminoꢀ2ꢀfluoroꢀphenyl)ꢀ5,5ꢀdifluoroꢀ4ꢀmethylꢀ5,6ꢀ
dihydroꢀ4Hꢀ[1,3]oxazinꢀ2ꢀylamine (13) (54 mg, 0.21 mmol) in MeOH (5 mL) was added dropwise at 0
°C and the mixture was stirred at 22 °C for 20 h. The mixture was partitioned between aqueous Na2CO3ꢀ
solution (1 M) and dichloromethane, the organic layers were washed with brine, dried over Na2SO4 and
evaporated. The residue was purified by flash column chromatography on silica gel using a mixture of
methanol/methylene chloride (0ꢀ10%) to give the pure amides 14a and 14d.
5-Chloro-pyridine-2-carboxylic acid [3-((R)-2-amino-5,5-difluoro-4-methyl-5,6-dihydro-4H-
[1,3]oxazin-4-yl)-4-fluoro-phenyl]-amide (14a).43 The condensation of 13 (110 mg, 0.42 mmol) and 5ꢀ
ACS Paragon Plus Environment

Page 27 of 44
Journal of Medicinal Chemistry
chloroꢀpyridineꢀ2ꢀcarboxylic acid (67 mg, 0.42 mmol) yielded the title compound 14a (122 mg, 72%) as
1
2
3
4
1
a colorless solid: H NMR (300 MHz, DMSOꢀd6) δ ppm 10.70 (s, 1 H), 8.79 (d, J=2.0 Hz, 1 H), 8.13–
5
6
7
8
8.22 (m, 2 H), 7.95 (dd, J=7.1, 2.6 Hz, 1 H), 7.83–7.91 (m, 1 H), 7.15 (dd, J=11.8, 8.8 Hz, 1 H), 4.29
(ddd, J=17.8, 12.0, 5.8 Hz, 1 H), 4.02 (ddd, J=17.8, 11.8, 5.8 Hz, 1 H), 3.30 (s, 2 H), 1.64 (s, 3 H). LCꢀ
9
+
HRMS: m/z 399.0839 [(M+H)+ calcd for C17H15ClF3N4O2 , 399.0830]. MS (ISP): m/z = 399.2 [M+H]+.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
5-Cyano-pyridine-2-carboxylic acid [3-((R)-2-amino-5,5-difluoro-4-methyl-5,6-dihydro-4H-
[1,3]oxazin-4-yl)-4-fluoro-phenyl]-amide (14d).43 The condensation of 13 (80 mg, 0.31 mmol) and 5ꢀ
cyanoꢀpyridineꢀ2ꢀcarboxylic acid (50 mg, 0.34 mmol) yielded the title compound 14d (78 mg, 65%) as a
1
colorless solid: H NMR (600 MHz, DMSOꢀd6) δ ppm 10.86 (s, 1 H), 9.20 (dd, J=2.1, 0.9 Hz, 1 H),
8.58 (dd, J=8.2, 2.1 Hz, 1 H), 8.28 (dd, J=8.2, 0.8 Hz, 1 H), 7.98 (dd, J=7.2, 2.8 Hz, 1 H), 7.88 (ddd,
J=8.8, 3.8, 2.9 Hz, 1 H), 7.17 (dd, J=11.8, 8.8 Hz, 1 H), 5.91 (br s, 2 H), 4.30 (ddd, J=18.5, 11.9, 5.4 Hz,
1 H), 4.03 (ddd, J=17.4, 11.8, 5.7 Hz, 1 H), 1.65 (s, 3 H). LCꢀHRMS: m/z 390.1183 [(M+H)+ calcd for
+
C18H15F3N5O2 , 390.1172]. MS (ISP): m/z = 390.2 [M+H]+. Xꢀray PDB accession code: 3ZMG.
3-Methyl-5-(trifluoromethyl)-4,5-dihydroisoxazole (76). To diethyl ether (640 mL) was added
subsequently at ꢀ78 °C 3,3,3ꢀtrifluoropropꢀ1ꢀene (65.3 g, 680 mmol), nitroethane (30.0 g, 400 mmol),
Et3N (405 mg, 4 mmol) and phenyl isocyanate (97.2 g, 816 mmol), the solution was warmed to 22 °C
and stirring was continued for 74 h. The suspension was filtered and the filtrate distilled from bulb to
bulb at 50ꢀ60 °C/1.0 mbar to give a 9:1 mixture of the title compound 76 and its corresponding position
1
isomer (37.3 g, 61%) as a pale yellow liquid. H NMR (600 MHz, CDCl3) δ ppm 4.80–4.87 (m, 1 H),
3.24 (ddq, J=17.8, 11.5, 1.0, 1.0, 1.0 Hz, 1 H), 3.08 (ddq, J=17.8, 5.8, 0.9, 0.9, 0.9 Hz, 1 H), 2.04 (t,
1
J=1.1 Hz, 3 H). H NMR signals of position isomer 3ꢀmethylꢀ4ꢀ(trifluoromethyl)ꢀ4,5ꢀdihydroisoxazole:
4.44–4.53 (m, 2 H), 3.83–3.91 (m, 1 H), 2.12 (t, J=0.9 Hz, 3 H). MS: m/z = 153 [M]+. Anal.
(C5H6F3NO) calcd C 39.22%, H 3.95%, N 9.15%; found C 39.27%, H 3.60%, N 9.34%.
(3S,5S)-3-(2-Fluoro-phenyl)-3-methyl-5-trifluoromethyl-isoxazolidine (79). To a stirred solution
of 1ꢀbromoꢀ2ꢀfluoroꢀbenzene (77) (11.2 g, 64 mmol) in THF (50 mL) and toluene (140 mL) was added
at ꢀ78 °C nꢀBuLi (1.6 M in hexane, 38 mL, 61 mmol) keeping the temperature below ꢀ70 °C and stirring
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 44
was continued at ꢀ78 °C for 1 h. To a solution of 3ꢀmethylꢀ5ꢀ(trifluoromethyl)ꢀ4,5ꢀdihydroisoxazole (76)
1
2
3
.
(4.68 g, 30.6 mmol) in toluene (330 mL) was added at ꢀ78 °C BF3 Et2O (8.68 g, 61.1 mmol) which was
4
5
6
7
8
followed by the addition of the phenyl lithium reagent prepared above using an insulated cannula over
10 min keeping the temperature below ꢀ70 °C. The mixture was stirred at ꢀ78 °C for 1 h, quenched with
saturated aqueous NH4Cl and extracted with EtOAc. The organic layers were washed with brine, dried,
evaporated and the residue was purified by flash column chromatography on silica gel using a mixture
of nꢀheptane/EtOAc (50:1) to give the racemic title compound 79 (4.275 g, 56 %) as a pale yellow
liquid. The racemate was resolved on a chiral HPLC column (Chiralpack AD) using nꢀheptane/EtOH
(95:5) to give the desired (3S,5S)ꢀ3ꢀ(2ꢀfluoroꢀphenyl)ꢀ3ꢀmethylꢀ5ꢀtrifluoromethylꢀisoxazolidine (79)
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
(1.719 g, 23%) as the slower eluting enantiomer with positive optical rotation and an ee >99.5%. H
NMR (300 MHz, CDCl3) δ ppm 7.78 (td, J=8.0, 1.3 Hz, 1 H), 7.22–7.31 (m, 1 H), 7.14 (td, J=7.7, 1.2
Hz, 1 H), 7.05 (ddd, J=12.0, 8.1, 1.1 Hz, 1 H), 5.67 (s, 1 H), 4.27 (sxt, J=7.4 Hz, 1 H), 3.10 (ddd,
J=13.4, 8.2, 1.0 Hz, 1 H), 2.37 (dd, J=13.3, 7.5 Hz, 1 H), 1.62 (s, 3 H). MS: m/z = 250.4 [M+H]+. The
relative and absolute configuration of the newly introduced stereogenic centers were confirmed by Xꢀray
crystallographic analysis of the target compound 89 complexed with BACE1 (4J1F). (3R,5R)ꢀ3ꢀ(2ꢀ
Fluoroꢀphenyl)ꢀ3ꢀmethylꢀ5ꢀtrifluoromethylꢀisoxazolidine (1.738 g, 23%) was obtained as the faster
eluting enantiomer with negative optical rotation and an ee >99.5%.
(2S,4S)-4-Amino-1,1,1-trifluoro-4-(2-fluorophenyl)pentan-2-ol (81). To a solution of (3S,5S)ꢀ3ꢀ
(2ꢀfluoroꢀphenyl)ꢀ3ꢀmethylꢀ5ꢀtrifluoromethylꢀisoxazolidine (79) (1.50 g, 6.02 mmol) in EtOH (35 mL)
was at 22°C ammonium formate (3.04 g. 48.2 mmol) and Pd/C (10%, 320 mg) and stirring of the
mixture was continued at 22 °C for 4 h. The suspension was filtered, the filtrate evaporated and the
residue was partitioned between EtOAc and saturated aqueous NaHCO3 solution. The organic layer was
dried and evaporated to give the crude title compound (1.50 g, 99%) as a colorless solid which was used
without further purification. 1H NMR (300 MHz, CDCl3) δ ppm 7.45 (td, J=8.1, 1.8 Hz, 1 H), 7.28–7.37
(m, 1 H), 7.19 (td, J=7.5, 1.2 Hz, 1 H), 7.09 (ddd, J=12.6, 8.1, 1.3 Hz, 1 H), 3.61 (dqd, J=11.4, 6.7, 6.7,
6.7, 2.0 Hz, 1 H), 2.50 (dt, J=14.3, 1.6 Hz, 1 H), 1.75–2.75 (br, 2 H), 1.82 (dd, J=14.3, 11.5 Hz, 1 H),
ACS Paragon Plus Environment

Page 29 of 44
Journal of Medicinal Chemistry
1.69 (d, J=1.2 Hz, 3 H) (OH not detectable). MS: m/z = 252.2 [M+H]+. A small sample was purified by
flash column chromatography on silica gel using a mixture of nꢀheptane/EtOAc: m.p. 113ꢀ115 °C. Anal.
(C11H13F4NO) calcd C 52.59%, H 5.22%, N 5.58%; found C 52.49%, H 5.24%, N 5.73%.
1
2
3
4
5
6
7
8
9
(4S,6S)-4-(2-Fluorophenyl)-4-methyl-6-(trifluoromethyl)-5,6-dihydro-4H-1,3-oxazin-2-amine
(83). To a solution of (2S,4S)ꢀ4ꢀaminoꢀ1,1,1ꢀtrifluoroꢀ4ꢀ(2ꢀfluorophenyl)pentanꢀ2ꢀol (81) (1.834 g, 7.30
mmol) in EtOH (38 mL) was added a solution of BrꢀCN (5 M in CH3CN, 2.2 mL, 11.0 mmol) and the
mixture was stirred in a sealed tube at 85 °C for 15 h. The mixture was evaporated and the residue
partitioned between EtOAc and saturated aqueous Na2CO3 solution, the organic layer was dried,
evaporated and the residue was purified by chromatography (silicaꢀNH2) using a mixture of nꢀ
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
heptane/EtOAc (5:1 to 0:1) to afford the title compound 83 (817 mg, 41%) as a colorless oil. H NMR
(600 MHz, CDCl3) δ ppm 7.38 (td, J=8.1, 1.9 Hz, 1 H), 7.24–7.29 (m, 1 H), 7.13 (td, J=7.6, 1.3 Hz, 1
H), 7.04 (ddd, J=12.4, 8.1, 1.3 Hz, 1 H), 4.35 (br s, 2 H), 3.97 (dqd, J=12.4, 6.0, 6.0, 6.0, 2.8 Hz, 1 H),
2.77 (dd, J=13.6, 2.8 Hz, 1 H), 1.86 (dd, J=13.6, 12.6 Hz, 1 H), 1.63 (d, J=1.3 Hz, 3 H). MS: m/z =
277.1 [M+H]+. A small sample was subjected again to flash column chromatography on silicaꢀNH2 gel
using a mixture of nꢀheptane/EtOAc: Anal. (C12H12F4N2O) calcd C 52.18%, H 4.38%, N 10.14%; found
C 52.04%, H 4.14%, N 10.05%.
(4S,6S)-4-(2-Fluoro-5-nitrophenyl)-4-methyl-6-(trifluoromethyl)-5,6-dihydro-4H-1,3-oxazin-2-
amine (85). To concentrated sulfuric acid (11 mL) was added portion wise (4S,6S)ꢀ4ꢀ(2ꢀfluorophenyl)ꢀ
4ꢀmethylꢀ6ꢀ(trifluoromethyl)ꢀ5,6ꢀdihydroꢀ4Hꢀ1,3ꢀoxazinꢀ2ꢀamine (83) (760 mg, 2.75 mmol) at 22 °C,
the solution obtained was cooled to 0 °C and treated with red fuming HNO3 (0.17 mL, 3.85 mmol ) over
20 min and stirring was continued at 0 °C for 30 min. The reaction mixture was slowly added to crushed
ice/water (150 mL), the pH was adjusted to 10 using NaOH, the aqueous layer was extracted with
EtOAc, the organic layer was dried, evaporated and the residue was chromatographed on silica gel using
a mixture of nꢀheptane/EtOAc (3:1) to afford the title compound 85 (800 mg, 91%) as a pale yellow oil.
1H NMR (300 MHz, CDCl3) δ ppm 8.37 (dd, J=7.0, 2.9 Hz, 1 H), 8.19 (ddd, J=8.9, 4.0, 3.0 Hz, 1 H),
ACS Paragon Plus Environment