Design and Synthesis of Clinical Candidate PF-06751979: A Potent, Brain Penetrant, β-Site Amyloid Precursor Protein Cleaving Enzyme 1 (BACE1) Inhibitor Lacking Hypopigmentation

Article abstract of DOI:10.1021/acs.jmedchem.8b00246

A major challenge in the development of β-site amyloid precursor protein cleaving enzyme 1 (BACE1) inhibitors for the treatment of Alzheimer's disease is the alignment of potency, drug-like properties, and selectivity over related aspartyl proteases such as Cathepsin D (CatD) and BACE2. The potential liabilities of inhibiting BACE2 chronically have only recently begun to emerge as BACE2 impacts the processing of the premelanosome protein (PMEL17) and disrupts melanosome morphology resulting in a depigmentation phenotype. Herein, we describe the identification of clinical candidate PF-06751979 (64), which displays excellent brain penetration, potent in vivo efficacy, and broad selectivity over related aspartyl proteases including BACE2. Chronic dosing of 64 for up to 9 months in dog did not reveal any observation of hair coat color (pigmentation) changes and suggests a key differentiator over current BACE1 inhibitors that are nonselective against BACE2 in later stage clinical development.

Full text of DOI:10.1021/acs.jmedchem.8b00246

Subscriber access provided by Kaohsiung Medical University
Design and Synthesis of Clinical Candidate PF-06751979:
A Potent, Brain Penetrant, #-site amyloid precursor protein
cleaving enzyme 1 (BACE1) Inhibitor Lacking Hypopigmentation
Brian T. O'Neill, Elizabeth M. Beck, Christopher R. Butler, Charles E. Nolan, Cathleen Gonzales,
Lei Zhang, Shawn D Doran, Kimberly Lapham, Leanne M Buzon, Jason K Dutra, Gabriela Barreiro,
Xinjun Hou, Luis A Martinez-Alsina, Bruce N. Rogers, Anabella Villalobos, John C. Murray,
Kevin Ogilvie, Erik A. LaChapelle, Cheng Chang, Lorraine F Lanyon, Claire M. Steppan, Ashley
Robshaw, Katherine Hales, Germaine G. Boucher, Karamjeet Pandher, Christopher Houle,
Claude W. Ambroise, David Karanian, David Riddell, Kelly R. Bales, and Michael A. Brodney
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b00246 • Publication Date (Web): 03 Apr 2018
Downloaded from http://pubs.acs.org on April 3, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 111
1
2
3
4
5
6
7
8
9
Design and Synthesis of Clinical Candidate
PF-06751979: A Potent, Brain Penetrant, β-site
amyloid precursor protein cleaving enzyme 1
(BACE1) Inhibitor Lacking Hypopigmentation
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
Brian T. O’Neill,2 Elizabeth M. Beck,1 Christopher R. Butler,1 Charles E. Nolan,4 Cathleen
Gonzales,4 Lei Zhang,1 Shawn D. Doran,3 Kimberly Lapham,3 Leanne M. Buzon,2 Jason K.
Dutra,2 Gabriela Barreiro,1 Xinjun Hou,1 Luis A. Martinez-Alsina,2 Bruce N. Rogers,1 Anabella
Villalobos,1 John C. Murray,2 Kevin Ogilvie,2 Erik A. LaChapelle,2 Cheng Chang,3 Lorraine F.
Lanyon,6 Claire M. Steppan,6 Ashley Robshaw,4 Katherine Hales,4 Germaine G. Boucher,5
Karamjeet Pandher,5 Christopher Houle,5 Claude W. Ambroise,4 David Karanian,5 David
Riddell,4 Kelly R. Bales,4 and Michael A. Brodney*,1
2
1Medicine Design, Medicinal Chemistry, Pfizer Inc., Cambridge, MA 02139, USA; Medicine
3
Design, Medicinal Chemistry, Pfizer Inc., Groton, CT 06340, USA; Medicine Design,
4
Pharmacokinetics, Dynamics and Metabolism, Pfizer Inc., Groton, CT 06340, USA; Internal
5
Medicine, Pfizer Inc., Cambridge, MA 02139, USA; Drug Safety Research and Development,
6
Pfizer Inc., Groton, CT 06340, USA; Discovery Sciences, Primary Pharmacology, Pfizer Inc.,
Groton, CT 06340, USA;
1
ACS Paragon Plus Environment

Page 3 of 111
Journal of Medicinal Chemistry
1
2
3
4
RECEIVED DATE (to be automatically inserted after your manuscript is accepted if required
5
6
7
8
9
Keywords: BACE1, BACE2, hypopigmentation, depigmenation phenotype, PMEL17, Thiazole.
PF-06751979.
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
ABSTRACT
A major challenge in the development of β-site amyloid precursor protein cleaving enzyme 1
(BACE1) inhibitors for the treatment of Alzheimer’s disease is the alignment of potency, drug-
like properties, and selectivity over related aspartyl proteases such as Cathepsin D (CatD) and
BACE2. The potential liabilities of inhibiting BACE2 chronically has only recently begun to
emerge as BACE2 impacts the processing of the premelanosome protein (PMEL17) and disrupts
melanosome morphology resulting in a depigmentation phenotype. Herein we describe the
identification of clinical candidate PF-06751979 (64) which displays excellent brain penetration,
potent in vivo efficacy, and broad selectivity over related aspartyl proteases including BACE2.
Chronic dosing of 64 for up to 9-months in dog did not reveal any observation of hair coat color
(pigmentation) changes and suggests a key differentiator over current BACE1 inhibitors that are
non-selective against BACE2 in later stage clinical development.
Introduction
Alzheimer’s disease (AD) pathology is characterized by the presence of extracellular plaques
in the hippocampal and cortical regions of the brain, accompanied by intraneuronal
neurofibrillary tangles and extensive neuronal loss. Proteolytic cleavage of the amyloid
precursor protein (APP) by β-site amyloid precursor protein cleaving enzyme 1 (BACE1), a
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 111
1
2
3
4
5
6
7
8
9
member of the aspartyl protease family of enzymes, takes place within the endosome at low pH,
generating a soluble N-terminal ectodomain of APP (sAPP) and C-terminal fragment (C99).1
Subsequent cleavage of the membrane-bound C99 fragment by -secretase liberates the various
amyloid- (A peptide species, of which A40 and A42 are the predominant forms.2 A number
of mutations in APP, near the BACE1 cleavage site have been reported that either increase A
generation and are associated with early-onset AD, or decrease A generation and protect against
late-onset AD.3-4 Therefore, limiting the generation of A through inhibition of BACE1 is an
attractive approach for the treatment of Alzheimer’s disease (AD) and is being intensely pursued
by the pharmaceutical industry.5-6
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
Despite the considerable efforts to understand the impact of inhibiting BACE1, the
physiological consequences of inhibiting BACE2 has only recently begun to emerge. The
expression pattern of BACE2 is very low in brain and much greater in peripheral tissues.7
Identifying and understanding the role of BACE2 substrates and their inhibition is important
from an off-target perspective. Recently identified BACE2 substrates include transmembrane
protein 27 (TMEM27) and the pigment cell-specific melanocyte protein (PMEL17). In the case
of TMEM27, which is robustly expressed in pancreatic -cells, inhibition of BACE2 may
actually increase -cell mass and improve glucose tolerance in conditions of insulin resistence.8
On the other hand, disruption of PMEL17 processing in pigment cells, due to BACE2
knockdown, has profound effects on the formation of physiologically important, non-pathologial
amyloid fibers necessary for proper sequestration of melanin and ultrastuctural integrity of
melanosomes.9-11 Genetic mutation of BACE2 in zebrafish through zinc finger nuclease-
mediated genome editing generates a melanocyte migration phenotype, while failing to generate
any semblance of a hypomyelination phenotype seen with BACE1 silencing.12 Insights into the
3
ACS Paragon Plus Environment

Page 5 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
mechanism of this melanocyte migration phenotype comes from PMEL knockdown studies in
zebrafish retinal pigment epithelium, demonstrating a failure of cylindrical melanosomes to form
which consequently prevents the movement of melanosomes into apical processes.13 In addition,
BACE2 KO mice display a hypopigmented phenotype11, and several non-selective BACE1/2
inhibitors such as AZD3293 (5), NB-360 (6), and SCH1682496 have shown fur and skin
pigment loss in mice or dog following chronic dosing regimens.14-16
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
In recent years, several small molecule BACE1 inhibitors including MK-8931 (2), AZD3293
(5), E2609, CNP-520, and JNJ-54861911 have advanced into late stage clinical trials (Figure
1).17-18 This new class of amidine based inhibitors possess excellent BACE1 potency, adequate
central nervous system (CNS) penetration, robust lowering of A in the cerebral spinal fluid
(CSF) of humans, and overcame ocular toxicity associated with earlier development candidates
such as compound 1 likely due to the off-target inhibition of CatD.19-20 However, the current
cohort of clinical development candidates appear to inhibit both BACE1 and BACE2 to a similar
degree. This lack of selectivity is consistent with additional inhibitors reported in the literature
such as LY-2886721 (4), Merck (3), and NB-360 (6).15, 21-22 Despite the lack of selectivity over
BACE2, we are unaware of any potential safety concerns emerging thus far in late stage clinical
trials for the current cohort of Phase 3 molecules. For instance, the dosing of healthy human
volunteers for 14-day duration with AZD3293, a non-selective BACE1 inhibitor, showed no
signs of hypopigmentation.14
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 111
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 1: Representative Overview of BACE1 Inhibitors and Selectivity over BACE2
The rational design of BACE1 inhibitors with an improved balance of potency, CNS
penetration and reduced safety liabilities has been a major focus within medicinal chemistry
groups across the pharmaceutical industry. Balancing the compound attributes has been a direct
result of optimizing the underlying physico-chemical properties (pKa, MW, logD, hydrogen bond
donors) and removing undesired structural motifs to increase the probability for compound
survival.23 As BACE1 inhibition typically requires a basic amine for interaction with the
catalytic aspartic acids, the optimization of the amidine pKa has been critical in reducing P-gp
mediated efflux and driving central pharmacodynamic responses.24-27 In addition to reducing
efflux, a lower pKa has been associated with reduced safety liabilities such as inhibition of the
IKs channel (hERG) and reduction in potential drug-drug interactions (DDIs).28-29
5
ACS Paragon Plus Environment

Page 7 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
A common strategy to reduce pKa has been modification of the amidine functionality by
strategic placement of an electron withdrawing group such as fluorine (Figure 2).30 For example,
we recently disclosed a novel series of thioamidine-containing BACE1 inhibitors that possess
excellent overall properties such as high CNS penetration (Cub/Cup = 1), low clearance, reduced
hERG inhibition, and low projected DDI.31 A key design element of the series was strategic
incorporation of a pKa lowering group adjacent to the thioamidine moiety (compound 7 to 8).
The addition of fluorine reduced the pKa from 7.7 to 7.0, resulting in reduced inhibition of hERG
from an IC50 of 2.1 µM to 10.1 µM. Janssen and Novartis have described a related strategy for a
series of 1,4 oxazine-derived inhibitors that suffered from poor CNS penetration due to high
transporter efflux, hERG inhibition, and a weak pharmacodynamics response in brain, as
exemplified by compound 9. Tuning the pKa of the amidine with a trifluoromethyl group
(compound 10) reduced the pKa from 9.2 to 7.8 with a corresponding decrease in P-gp mediated
efflux and an increase in brain penetration.32 A similar strategy was taken by F. Hoffman-La
Roche30, wherein the modification of the amidine pKa by incorporation of the fluoro group
resulted in a well-balanced analog with robust central lowering of A due to increased exposure
in brain (compound 12 vs. 11).
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
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 111
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. The Impact of Reduced pKa on Key Properties
Despite the advances in aligning properties with potency, efforts in designing BACE1
inhibitors with selectivity over BACE2 have been minimal to date. A notable exception was the
incorporation of unique substituents into the S3 pocket that take advantage of greater flexibility
in the 10s loop for BACE1 over BACE2 as reported by F. Hoffman-La Roche.30 In this paper,
we describe efforts to modulate the basic pKa of the thioamidine moiety (represented by 13)
through modification of the P1 substituent while simultaneously obtaining selectivity over
BACE2 by modification of the inhibitor in the P3 and flap regions of BACE1 (represented by
14). We had previously described a stereospecific incorporation of electron withdrawing fluoro
or trifluoromethyl groups on the thioamidine backbone core that also reduced the pKa of this
functionality but that required lengthy complex synthesis and failed to deliver molecules with
7
ACS Paragon Plus Environment

Page 9 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
improved selectivity over BACE2.31 The present synthesis and characterization of thiazole
containing inhibitors (14) has resulted in the identification of compound 64, a clinical candidate
with an excellent balance of BACE1 potency, CNS penetration, robust central A lowering, long
predicted human half-life, and selectivity over BACE2 as determined by in vitro binding,
cellular/tissue pigmentation models, and long-term safety studies.
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 3: Design Strategy for Modulating pKa and Selectivity over BACE2
Results
We previously described a strategy to add aryl and heteroaryl substituents by straightforward
Lewis acid catalyzed anion addition to isoxazoline template 15 followed by elaboration to the
thioamidines 40-47 (Scheme 1).31 The chirality of the isoxazoline 15 originates from commercial
propylene oxide or a commercial homoallylic alcohol (S)-pent-4-en-2-ol that controls the
creation of two additional chiral centers. The compact ring structure contributes to the high
selectivity of the anion addition to the convex face. Following addition, the isoxazoline ring 16-
23 can be reductively cleaved to amino-alcohols 24-31 and the thioamidine ring formed by
addition of an acylisothiocyanate followed by ring closure to thioamides 32-39. Deprotection in
basic methanol provided the desired analogs 40-47 that were assessed for potency, pKa and
physical properties.
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 111
1
2
3
4
Scheme 1. Synthesis of aryl or heteroaryl containing thioamidines
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
Reagents and conditions: (a) ArX, 2.5M n-BuLi in hexane, BF3-Et2O, toluene:THF (10:1 v:v), -70°C then add 15 in
toluene:THF (10:1 v:v), 25-68%; (b) H2, Raney nickel, ethanol or IPA/THF, 50 °C; (c) 2eq BzNCS, ethyl acetate,
r.t. to 90 °C (no isolation of thiourea intermediate); (d) BzNCS, ethyl acetate, r.t., 2.5h (isolate thiourea), then
BzNCS, ethyl acetate, 90 °C, 26-66%; (e) DBU, methanol, 70 °C, 15-74%.
In addition, we prepared an array of analogs bearing various P1 moieties in a parallel
medicinal chemistry fashion illustrated in Scheme 2. Starting with a diverse set of amino
alcohols 48, that were prepared using the procedure described above, the desired analogs 50 were
obtained in two distinct steps followed by high throughput purification (Scheme 2). The success
rate for this parallel effort was 89% (40/45). The sum total of these efforts led us to the thiazole
series with the desired pKa and physical property profile.
9
ACS Paragon Plus Environment

Page 11 of 111
Journal of Medicinal Chemistry
1
2
3
4
Scheme 2. Array Chemistry to Generate P1 Aryl Groups
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
o
Reagents and conditions: (a) EtO2CN=C=S / THF then Scavenging Resin; (b) MeTHF / H2O / TFA, 60 C, 1h; (c)
120 oC, 4h (d) high throughput purification.
In order to fully evaluate the SAR of the thiazole containing series, we desired a robust
procedure to prepare the thiazole bromide 54 which could serve as a viable intermediate for
exploration of the P3 subpocket (Scheme 3). Selective C-2 halogen metal exchange of 2,4-
dibromothiazole followed by addition to 15 provided compound 51 which was poised for
elaboration into the P3 region. Alternatively, addition of the anion from halogen metal exchange
of 2-bromothiazole (or reductive removal of the previously mentioned C-4 bromide) afforded a
substrate that could be metalated at C-5 directed by the proximal sulfur. Due to the presence of
the bromide, an alternative method was required for selective cleavage of the N-O bond.
Reduction with a Mo (0) hexacarbonyl complex followed by a reductive work up afforded the
amino alcohol 52 without loss of the bromide. Addition of benzoyl isothiocyanate formed the
thiourea 53, which upon activation of the alcohol with Ghosez’s reagent33 yielded the key
thioamidine intermediate 54.
10
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 111
1
2
3
4
Scheme 3: Synthesis of Bromo-Thiazole 54
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
Reagents and conditions : (a) 2,4-dibromo-thiazole, 2.5M n-BuLi in hexane, BF3-Et2O, toluene:THF (10:1 v:v), -70
°C then add 15 in toluene:THF (10:1 v:v), 1h, 85%; (b) Mo(CO)6, CH3CN, H2O, NaBH4, 70 °C, 1h, 95%; (c)
benzoyl isothiocyante, CH2Cl2, rt, 24h, 66%; (d) Ghosez's reagent, CH2Cl2, rt, 1h, 77%.
The elaboration of the bromide 54 into the key thiazole-amides 56-68 and 70-71 is shown
in Scheme 4. The bromide was transformed to an amine and eventually to amides 55 by one of
several routes described below. Our initial protocol involved a copper-catalyzed azide addition
in the presence of ascorbic acid that conveniently reduced the azide in situ to an amine.
Acylation of the amine yielded the desired amides 55 which readily deprotected in basic
methanol to afford thioamidines 56-68. Amides 55 were also available via direct palladium
catalyzed amidation. For target 69, the bromide of 54 was removed via zinc reduction and a new
anion was then formed at C-5 with butyl lithium. The anion was then quenched with CO2 to give
the acid, which by suitable manipulation was coverted to a series of amides 69. Deprotection
afforded the C-5 amides 70 and 71 in good yield.
11
ACS Paragon Plus Environment

Page 13 of 111
Journal of Medicinal Chemistry
1
2
3
4
Scheme 4. Elaboration of Bromide 54 into Amide Analogs
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
Reagents and conditions : (a) trans-N,N'-Dimethylcyclohexane-1,2-diamine, NaN3, H2O, Sodium-L-ascorbate rt then CuI, 70
oC, 2h, 47%; (b) HATU, R-CO2H, DMF, rt, 1h or 2,4,6-Tripropyl-1,3,5,2,4,6-trioxatriphosphinane 2,4,6-trioxide,
o
pyridine or Et3N, rt, 12-24h; (c) ArC(O)NH2, Pd2dba3, tBuBrettPhos, LiO-t-Bu, 2-MeBuOH, 95 C, 1h, 61%; (d)
o
Methoxyl-amine, HCl, pyridine, 50oC, 16h or DBU, methanol, 70 °C; (e) Zn dust, THF, aq NaOH, 40 C, 16h,
86%; (f) n-BuLi, THF, CO2, -78 oC, 1h, 95%; (g) HATU, R-NH2, DMF; (h) MeNH2, EtOH, rt, 2-4h.
A synthesis of 64 on intermediate scale (25 g) is illustrated in Scheme 5. The Buchwald
coupling conditions arose from the need to replace the azide coupling on scale and provided for a
mild, selective deprotection with HCl. Amide formation with 72 was easily accomplished
through coupling with a variety of agents such as HATU or T3P. The final debenzoylation of 73
was delicate due to interference from the thioamidine. This was achieved in high yield using
basic methanol.
12
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 111
1
2
3
4
Scheme 5. Synthesis of Compound 64
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
Reagents and conditions : (a) 2,4-dimethoxy benzyl amine, Pd2(dba)3, t-BuONa, t-BuXPhos, 70 °C, 1h then HCl,
H2O, rt, 1h, 70%; (c) HATU, iPrNEt2, CH3CN, 5-(difluoromethoxy)picolinic acid, rt, 4h, 81%; (d) DBU, methanol,
70 °C, 1.5h, 55%.
Discussion and Conclusions
Our primary strategy to modulate the basic pKa of the amidine was accomplished by a
systematic exploration of the P1 region (Table 1). Our goal was to identify those aromatic
substituents or heteroaryl replacements that inherently reduced pKa to an optimal range of 6.5-7,
an optimal range to balance BACE1 potency with overall alignment of properties. This strategy
would obviate the need to fluorinate the core and reduce synthetic complexity. In previous SAR,
we had optimized the fluorination pattern on the P1 substituent for BACE1 potency and the
optimal substitution pattern was the 2,4 difluoro analog 40 which had a basic pKa of 8.1.31 As
13
ACS Paragon Plus Environment

Page 15 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
anticipated, removal of the fluorines produced analog 42 with weaker BACE1 potency in a
fluorescent polarization (FP) assay and a higher pKa (8.8). Attempts to further reduce the pKa
with an ortho substituted aryl group (41 or 43) resulted in a similar potency and pKa relative to
unsubstituted phenyl analog 42. We next investigated installation of a 2-pyridyl group (44) in P1
and were surprised that the pKa increased to 8.6 relative to analog 40. Further modification to 5-
membered ring heteroaryls such as thiophenes 45 or 46 resulted in reasonable BACE1 potency
but only the 2-thienyl analog 46 reduced the pKa to the desired range. This raised the possibility
that a thiazole in P1 would further reduce the pKa of the amidine due to the greater electron
withdrawing nature relative to thiophene. We were gratified to find that thiazole 47 resulted in a
BACE1 inhibitor with a pKa of 7 and displayed reasonable cellular activity (IC50 = 2.05 µM),
low P-gp mediated efflux, high passive permeability, and low clearance from human liver
microsomes. Despite the drop in potency for thiazole 47, we anticipated that appropriate
substitution on the thiazole ring would make favorable interactions with protein residues to
improve potency without increasing pKa. Across the P1 variations prepared in Table 1, each
analog displayed weak activity against CatD and the hERG channel.
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. Modification of the P1 Aryl Group
pKa
(hERG
IC50, µM)
HLM
(mL/min
/kg)f
BACE1
FP
RRCK
(x10-6
cm/s)e
CatD FP
IC50
BACE1
WCA
MDR
Erd
Cpd
R
IC50
(µM) b
IC50 (µM)c
(µM)a
14
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 111
1
2
3
4
5
6
8.1
(15.8)
40
0.704
> 72
0.039
1.19
17.5
< 8
7
8
9
8.8
(55.4)
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
> 79.3
> 100
> 100
>100
> 100
> 100
2.92
0.82
0.85
0.25
0.17
1.62
1.64
4.3
17.9
19.7
8.5
< 8
< 8
<8
41
42
43
44
45
8.8
(85.7)
1.33
6.06
9.1
(85.6)
8.6
(> 100)
9.16
5.17
2.4
7.6
< 8
< 8
8.4
(39.9)
1.4
13.2
7.9
(>100)
1.92
23.9
> 98
> 64
0.11
2.05
1.1
1.2
24.1
13.6
< 8
< 8
46
47
7.0
(> 100)
aIC50 values obtained from BACE1 Fluorescent Polarization (FP) Assay. IC50 values obtained from
CatD FP Assay. cIC50 values obtained from BACE1 Whole-Cell Assay (WCA). dRatio from the MS-based
quantification of apical/basal and basal/apical transfer rates of a test compound at 2 μM across contiguous
monolayers from MDR1-transfected MDCK cells. eRRCK–MDCK cells used to evaluate passive
permeability fHepatic clearance predicted from in vitro human microsomal stability study.
b
A key strategy to improve potency of thiazole 47 was installation of an aryl amide moiety
to engage the back-bone carbonyl of Gly230 while simultaneously placing an aryl substitutent
into the lipophilic S3 pocket. To guide the optimal placement of the amide group, we modeled
thiazole 47 using a co-crystal structure (PDB Code: 5CLM) (Figure 4a and 4b). Because of a
wider bond angle (150o vs 131o) induced by the sulfur atom in the thiazole ring, the vector off
the 5-position of the thiazole is not ideal for aligning the amide functionality within the P1-P3
15
ACS Paragon Plus Environment

Page 17 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
region of the protein (Figure 4a). On the other hand, the 4-position of the thiazole afforded a
more favorable angle for amide introduction (Figure 4b). Structural optimizations of the two
proposed phenyl amides demonstrated that 5-substituted phenyl amide would not make the key
hydrogen bond (Figure 4C), yet the 4-substituted phenyl amide would occupy the S3 pocket
nicely and make the desired hydrogen bond interaction to Gly230 (Figure 4d).
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. Molecular Modeling of Thiazole Regioisomers
Vectors to reach S3 pocket from 4 or 5-position of thiazole 40 aligned in co-crystal structure
(thin grey, PDB: 5CLM). Panels a and b: the initial vector of 4- or 5-substitutions and Panels c
and d: the modeled phenyl amides reaching S3 pockets.
We also evaluated the conformational energy profile of the thiazole orientation of
thiazole 47 to assess the potential energy penalty as different substitutions were introducted into
16
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 111
1
2
3
4
5
6
7
8
9
P3. The torsion of amidine and thiazole bond was subjected to a conformational analysis using
TetraChem with B3LYP/6-31G** in gas phase.34 We noticed that the “trans” conformation
(corresponding to Figure 4a) is at the lowest energy potential well, whereas the “cis”
conformation (corresponding to Figure 4b) is only at a slightly higher energy state (Figure 5),
giving us assurance that any conformational penalty of 4- vs. 5- substituted thiazoles would be
tolerated without a significant loss of potency.
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 5. Energy Confirmations for Thiazole 47
Since the conformational energies for thiazole 47 were relatively comparable, we
modeled four variations of aryl-amide linkers (74-77) to asses and rank-order their potential
binding to BACE1 (Figure 6, Panel A). Structure minimization using a variant of MM-GB/SA
calculations (PLIERS35) showed that 4-pheny amide 76 and 5-phenyl reversed amide 75 bound
preferably based on intermolecular energy, desolvation, and ligand strain energy components
(Figure 6, Panel B). An important consideration is the angle between the bicyclic core and the
amide group on the thiazole ring. The requirement to reverse the amide confirmation from 74 to
17
ACS Paragon Plus Environment

Page 19 of 111
Journal of Medicinal Chemistry
1
2
3
4
75 allowed for the optimization of the amide vector to P3 and alignment of a hydrogen bond
5
6
between Gly230 and the amide.
7
8
9
Figure 6 (Panel A-B). Aryl-amide linkers designed to reach S3 pocket with potential to
make H-bond to Gly230 backbone carbonyl.
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
Panel A: Amide regioisomers on the Thiazole Ring
Panel B: PLIERS calculations for analogs 74-77. Total energy (dE, kcal/mol) is the sum of all components in the
table. Only relative dE are are significant for rank ordering of these molecules.
Ligand
Conformational Total dE
Electrostatic van der Waals
GB/SA
Desolvation
(kcal/mol)
Molecule
Interaction
(kcal/mol)
Interaction
(kcal/mol)
Strain
(kcal/mol)
(kcal/mol)
-104.6
-109.4
-124.1
-123.4
-33.1
-33.7
-37.0
-35.7
76.9
76.1
93.4
94.0
8.3
3.2
2.6
4.1
-52.5
-63.8
-65.1
-61.0
74
75
76
77
As the amido-thiazoles 75 and 76 both appeared to satisfy the key parameters for binding,
synthesis and profiling of analogs 70-71 and 56-57 was initiated which contained a methoxy
group for additional potency (Table 2). For each case, molecular modeling suggested that a meta
or para substituted aryl group was optimal for BACE1 activity. We were gratified to find that
18
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 111
1
2
3
4
5
6
7
8
9
both analogs 56 and 70 showed considerable binding and cellular potency improvement relative
to the parent thiazole 47 while maintaining a similar pKa (6.8 and 6.5 respectively). Both
analogs exhibited low clearance in human liver microsomes and high passive permeability.
Unfortunately, the increase in hydrogen bond donor (HBD) count and molecular weight
increased the P-gp mediated efflux liability. Modification of the P3 aryl ring to a pyridine ring
has been previously demonstrated to reduce P-gp mediated efflux due to the potential for an
intramolecular hydrogen bond between the amide NH with nitrogen of the pyridine ring.31 In an
attempt to understand the impact of the pyridine ring on efflux and lower logD, we designed and
profiled analogs 71 and 57. For each matched molecular pair, the modification from phenyl to
pyridine improved potency, lowered P-gp mediated efflux, and reduced hERG binding without
negatively impacting clearance in human microsomes relative to the phenyl analogs 56 and 70.
Unfortunately, cross-species plasma instability with 5-amido substituted thiazoles 70 and 71
precluded further evaluation for efficacy or safety due to rapid turnover in rodent plasma.
Fortunately, thiazole-amides 56 and 57 were stable in cross species plasma stability assays.
Table 2: Profile for Aryl-Amides 52-55
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
Rat;
Human
Plasma
stability
T1/2
HLM
(mL/
min
/kg)f
pKa;
(hERG
IC50
BACE1
FP
CatD
FP
BACE1
WCA
IC50
RRCK
(x10-6
cm/s)e
MDR
Erd
Molecule
R
IC50
IC50
(µM)a
(µM)b
(nM)c
M)
(min)
19
ACS Paragon Plus Environment

Page 21 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
6.5,
6.2
20;
>360
70
71
0.27
0.11
> 100 0.11
8.9 14.7 <11.1
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
6.5;
4.8
2;
>360
> 100 0.033 6.4 13.5 < 11.3
6.9;
18.4
>360;
>360
56
0.99
0.26
> 97 0.034 4.4 12.9
< 8
6.9; >360;
13.1
57
> 95 0.014 1.8 15.7
< 8
>360
b
c
aIC50 values obtained from BACE1 FP Assay. IC50 values obtained from CatD FP Assay. IC50 values
d
obtained from BACE1 Whole-Cell Assay (WCA). Ratio from the MS-based quantification of
apical/basal and basal/apical transfer rates of a test compound at 2 μM across contiguous monolayers
from MDR1-transfected MDCK cells. eRRCK–MDCK cells used to evaluate passive permeability
fHepatic clearance predicted from in vitro human microsomal stability study.
With the 4-amido subsitutent on the thiazole ring providing a vector to probe potency and
selectivity by filling the P3 pocket, we profiled a range of 6-membered ring heteroaryl groups
such as pyridines and pyrazines containing a substituent in the 5-position (Table 4). In each
case, the regiochemistry of the heteroaryl ring placed a nitrogen atom ortho to carbonyl of the
20
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 111
1
2
3
4
5
6
7
8
9
amide as this was previously reported to both lock the confirmation of the amide and reduce P-gp
mediated efflux. With suitable binding assays in place, pyridine 58 gave comparable potency
and selectivity over CatD to the methoxy analog 56 while maintaining the low clearance, plasma
stability, and low P-pg efflux ratio (Table 3). As further potency improvements were needed, we
modified the fluoro substituent to a variety of analogs 60-64. We were gratified to identify
chloro, cyano, and difluoromethoxy substitutents improved potency in binding and whole cell
(IC50 < 10 nM) BACE1 assays relative to fluoro analog 58. In all cases, clearance was reduced
and the compounds showed high passive permeability. A second strategy to improve potency
was incorporation of an ortho methyl group resulting in analogs 59 and 65. Substitued pyrazines
(66-67) were well tolerated along with a 5-membered ring in P3 such as pyrazole 68. Using
BACE1 and BACE2 FP assays, compounds 60, 62, 64, and 67 displayed the highest levels of
selectivity over BACE2 of 5.6, 10, 6.4, and 5-fold respectively. The improved selectivity with
larger substituents in the P3 region may be explained by the greater flexibility in this region of
the protein relative to the rigid confirmation observed in BACE2. Unfortunately, poor plasma
stability was observed with compounds 60 and 66-68 which precluded further development.
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
Rat;Human
BACE1;
BACE2
FP
CatD
FP
BACE1
WCA
IC50
RRCK
(x10-6
cm/s)e
HLM
(mL/min
/kg)f
hERG
IC50
(µM)
Plasma
stability
(min)
MDR
Er
R
Cpd
IC50
IC50 (nM)a
(µM)b
(nM)c
21
ACS Paragon Plus Environment

Page 23 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
229;
233
>43
23.7
26.9
2.5 >360; >360 17.4
< 8
<8
10.7
17.9
58
59
178;
38.9
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
34.8
2.8
2.3
ND
14.9
8.5
60.2;
338
>300
>97
4.8
3.8
189/360
9.5
11.7
60
61
29.1;
21.6
2.3 >360; >360 19.3 < 11.6 9.0
107;
1072
163.6 12.5
2.8 >360; >360
7.4
20.7
10.1
62
21.8;
60.7
197.6
3.9
2.3 >360; >360 14.4
<8
12.1
9.9
63
64
36.9;
238
>95.3
5.1
2.2 >360; >360 12.6
<8
22.8;
70.5
9.6
43
>70.3
2.9
3.0 >360; >360 12.8
<8
<8
65
66
536;
1709
>96.7 47.6
2.1
225; >360
19.5
22
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 111
1
2
3
4
5
6
7
8
72.5;
367
23.6
9.3
>300
>300
7.2
1.8
5.8
45; >360
14.4
16.3
<8
67
68
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
9.3;
3.9
0.64
250; >360
< 8
aIC50 values obtained from BACE1 FP Assay. IC50 values obtained from CatD FP Assay. IC50 values
obtained from BACE1 Whole-Cell Assay (WCA). Ratio from the MS-based quantification of
apical/basal and basal/apical transfer rates of a test compound at 2 μM across contiguous monolayers
from MDR1-transfected MDCK cells. eRRCK–MDCK cells used to evaluate passive permeability
fHepatic clearance predicted from in vitro human microsomal stability study. ND = not determined.
b
c
d
A select group of compounds (61, 64, 65, 67) displayed an excellent alignment of properties,
potency, and were predicted to have good brain penetration due to the low P-gp mediated efflux
as measured in the Borst Cell Line. However, in vivo neuro-pharmacokinetics in mouse
suggested that all four analogs were brain impaired with Cub/Cup values ranging from 0.06 to
0.10 (Table 4). To build a more comprehensive understanding of the transporters limiting brain
penetration in rodent, we incorporated a higher expressing P-gp assay from the National Institute
of Health (NIH) along with a breast cancer resistant protein (BCRP) assay, an additional
transporter highly expressed at the blood-brain barrier. These analogs now displayed a range of
P-gp and BCRP mediated efflux which helped to explain the apparent in vitro to in vivo
disconnect. Since the expression of P-gp is highest in rodents and lower in NHP and humans, we
anticipated that NHP neuroPK would more accurately reflect brain penetration in humans.36 In
contrast to P-gp, BCRP expression in rodents is lower than NHP and unlikely to contribute to
differences in brain penetration when comparing mouse to NHP. To this end, brain penetration
studies in NHP were conducted and resulted in much improved Cub/Cup values ranging from 0.28
to 0.62 and CSF/Cup ranging from 0.39 to 0.76.
23
ACS Paragon Plus Environment

Page 25 of 111
Journal of Medicinal Chemistry
1
2
3
4
Table 4; Mouse vs NHP NeuroPK
5
6
7
Mouse NeuroPKa NHP NeuroPKa
BCRP
Erb
Borst NIH
MDR MDR
Cpd
R1
Cub/Cup;
CSF/Cup
Cub/Cup;
CSF/Cup
Erb
Erb
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
0.06;0.02
0.10;0.13
0.28;0.39
0.36;0.47
2.3
11.0
2.75
1.4
61
2.2
3.0
1.8
20.3
17.7
11.2
64
65
67
0.07;0.08
0.06;ND
0.62; 0.76
0.32;0.52
9.4
4.3
aUnbound brain to unbound plasma and CSF to unbound plasma area under the curve (AUC) ratios in
mice and non-human primates Ratio from the MS-based quantification of apical/basal and basal/apical
transfer rates of a test compound at 2 μM across contiguous monolayers from MDR1or BCRP-transfected
b
MDCK cells.
The in vitro and in vivo PK properties for compounds (61, 64, 65) are shown in Table 5. All
three compounds were characterized by low in vitro clearance in rat, dog and human microsomal
systems. As expected based on the low in vitro clearance, compounds 61, 64 and 65 all
exhibited low in vivo clearance in both rats and dogs. All three analogs were predicted to have a
long half-life in human (>12 h).
Table 5. In Vitro and in Vivo PK Properties of Representative Compounds
PK Properties
64 61 65
24
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 111
1
2
3
4
5
6
7
8
In vitro h-Clh (mL min-1 kg-1)a <8 <8 <8
In vitro r-Clh (mL min-1 kg-1)b <25 <25 <25
In vitro d-Clh (mL min-1 kg-1)c <18 <18 <18
Rat CL (mL/min/kg)d
Dog CL (mL/min/kg)e
Rat bioavailability (%)f
Dog bioavailability (%) g
31 10 2.9
1.8 5.5 1.2
33 43 100
49 34 60
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
aHepatic clearance predicted from in vitro human microsomal stability study. bHepatic clearance
predicted from in vitro rat microsomal stability study. cHepatic clearance predicted from in vitro dog
microsomal stability study. cdRat plasma clearance determined following a single intravenous bolus dose
of 1 mg/kg; eDog plasma clearance determined following a single intravenous bolus dose of 1 mg/kg.
fRat bioavailability determined following a single oral dose of either 3 or 5 mg/kg; gDog bioavailabilty
determined following a single oral dose of either 1 or 3 mg/kg.
Several compounds satisfied our criteria for advancement but only compound 64
combined excellent potency, modest selectivity (~5.6-fold) over BACE2, weak hERG inhibition,
stability in plasma, and high predicted brain penetration. To understand the in vivo potency of
compound 64, we assessed the compound against a panel of Aβ endpoints in brain, CSF and
plasma in 129/sve mice and dog. The exposure was also measured in these biological
compartments in order to understand PK/PD and predict efficacious concentrations (Ceff). For
mouse studies we used sensitive immunoassays to measure Ax-42 in brain homogenates, as
well as Ax-40 in CSF (Figure 7, Panel A and B). Acutely, robust lowering of brain
Apersisted in a dose and time dependent manner with maximum inhibition of 56%, peaking
at 5 to 7 h post dose. The 80 mg/kg dose did not produce any further reduction beyond the 50
mg/kg dose, but did demonstrate a delayed return to baseline at 14 hours. Acute administration
of 64 yielded a robust dose-responsive and time-dependent reduction of CSF Ax-40 with peak
inhibition at 3 hours of >77%. To determine if the reduction in brain and CSF A was
25
ACS Paragon Plus Environment

Page 27 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
maintained during sustained exposure to 64, a 5-day sub-chronic study was executed, dosing
once daily by subcutaneous (SC) administration (10 or 50 mg/kg/day). Brain and CSF samples
were collected on day 5, following the last dose. Compound 64 produced a dose-responsive and
time-dependent inhibition of A in mouse brain. At the 50 mg/kg/day dose, maximal brain
lowering was 63% at 7 to 9 hours (see SI, Figure 1). Administration of 64 (10 or 50 mg/kg/day
for 5 days) produced a dose-responsive and time-dependent inhibition of Ax-40 in mouse CSF
resulting in 77% inhibition of CSF at 3 hours post final 50 mg/kg dose.
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
Using cisterna magna cannulated Beagle dogs, the Ax-42 lowering by 64 was evaluated
in serial CSF samples collected over a 72 h period post-dose following oral doses of 1 or
3 mg/kg (Figure 7, Panel D). The results for each analyte measured in the CSF over a time
course of 72 h are expressed as absolute levels in picograms per milliliter. Following
administration of 64, the 3 mg/kg dose showed a long-lasting reduction in CSF levels of Ax-42,
with a maximum decrease of 74% versus baseline at 11 h post dose.
Figure 7: Effect of Compound 64 on Brain and CSF Amyloid-Beta Following
Administration in Mouse and Dog
150
M o u s e C S F A  x -4 0
150
M o u s e B ra in A  x -4 2
*
10 m g/kg
125
10 m g/kg
125
50 m g/kg
50 m g/kg
80 m g/kg
80 m g/kg
100
* *
100
75
50
25
0
* * *
* * * *
* *
* * * *
* * *
* * *
* *
75
50
25
0
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
* * * *
B LQ
B LQ
B LQ
B LQ
* *
B LQ
* * *
* *
* * *
0
5
10
15
0
5
10
15
T im e p o s t-d o s e (h )
T im e p o s t-d o s e (h )
26
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 111
1
2
3
4
5
6
7
8
9
1 0 0 0
D o g C S F A  x -4 0
150
125
100
75
5 D a y S u b c h ro n ic C S F A  x -4 0
8 0 0
6 0 0
4 0 0
2 0 0
10 m g/kg/day
50 m g/kg/day
* * * *
*
* * * *
*
*
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
*
†
*
*
50
†
*
* * * *
* * * *
†
* * * *
†
* * * *
25
†
†
†
* * * *
†
V ehicle
1
3
m g /k g C o m p o u n d # 6 4
m g /k g C o m p o u n d # 6 4
* * * *
0
0
5
10
15
-5 0
0
5 0
1 0 0
T im e p o s t-d o s e (h )
T im e p ost-d o se (h )
A) Compound 64 (10, 50 or 80 mg/kg) and vehicle were administered subcutaneously to male 129/sve mice. Brain
Ax-42 was assessed at 1, 3, 5, 7, 9 and 14 h post-dose. Data are expressed as the mean of % of vehicle of n = 5
animals per group  SEM.  p <0.0001.  p <0.001.  p <0.01.  p <0.05. B) Compound 64 (10, 50, or 80
mg/kg) and vehicle were administered subcutaneously to male 129/sve mice. CSF Ax-40 was assessed at 1, 3, 5,
7, 9 and 14 h post-dose. Data are expressed as the mean of % of vehicle of n = 5 animals per group  SEM. **** p
<0.0001. *** p <0.001. ** p <0.01. All values for the 50 mg/kg dose at 3 h were below BLQ which is defined as
<23 % of vehicle. All values for the 80 mg/kg dose groups at 3, 5, 7 and 9 h where BLQ which is defined as <36.3
% of vehicle. BLQ data points were interpolated from the standard curve and are represented in the graph. C) Male
129/sve mice were dosed subcutaneously once a day for 5 days with 64 (10 or 50 mg/kg/day) or vehicle. CSF
Ax-40 was assessed at 1, 3, 5, 7, 9 and 14 h after the last dose. Data are expressed as the mean percent of vehicle
of n = 5 animals per group  SEM.  p <0.0001;  p <0.001;  p <0.01;  p <0.05. D) Time course data for
the effect of 64 on CSF Ax-42 levels in dog. Compound 64 was orally dosed at 1 or 3 mg/kg at time = 0 h. Data
are the mean  SEM, n = 3 per group, * = p<0.05 for the 1 mg/kg group, † = p<0.05 for the 3 mg/kg group; Ax-40
= Amyloid- isoform containing amino acids 17-24 and ending in carboxy-terminal amino acid 40; Ax-42 =
Amyloid- isoform containing amino acids 17-24 and ending in carboxy-terminal amino acid 42; BLQ = Below
limit of quantitation; CSF = Cerebrospinal fluid; h = hour; mg/kg = Milligram per kilogram; SEM = Standard error
of the mean.
Our first goal was to establish a link between BACE2 inhibition in vitro and
accumulation of full length PMEL17 in primary melonocytes. To this end, we assessed two non-
selective BACE1 inhibitors (3 and 4) in primary human melanocytes from darkly pigmented
donors after overnight incubation with a dose response of inhibitor from 10 µM to 1 nM.
Accumulation of full length PMEL17 is apparent for both compounds (see SI, Figure 2)
consistent with inhibition of PMEL17 processing.
Furthermore, we were concerned that the measurements in the standard fluorescent
polarization (FP) assays which utilize an artificial substrate, truncated enzyme, and a non-
27
ACS Paragon Plus Environment

Page 29 of 111
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
physiological relevant pH (4.5) may not accurately reflect the selectivity for BACE1 over
BACE2 in a native environment. Due to these limitations, we established BACE1 and BACE2
binding assays using using full length human recombinant enzymes overexpressed in HEK293
cells where inhibition is determined by competition binding in cell paste at physiological pH
(6.0). Table 6 compares the selectivity for literature molecules 3 and 4 relative to thiazoles 61,
64, and 65. All inhibitors have excellent potency at BACE1 in binding or FP assay formats
along with cellular activity looking at production of sAPP in H4 cells. Remarkably, thiazole 64
showed improved selectivity over BACE2 in binding (27-fold) relative to the literature
examplars and across multiple chemical series in our BACE1 program (see SI, Figure 3). We
hypothesized that the improved selectivity over BACE2 was due to subtle non-binding
interactions in BACE1 in the P3 and flap region of the enzyme that were less tolerated in
BACE2.
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 6: Selectivity Profile for Analogs 3, 4, 61, 64, and 65
BACE1 BACE2
B2/B1
BACE1 BACE2
3D-Skin
sAPP
Binding Binding Binding
FP
FP
(nM)c
Compound
Model (nM)d
(nM)a
(nM)a
ratio
(nM)b
(nM)b
6.4
3.3
0.5
<2.2
42.0
4.5
< 0.54
38.4
2.44
38.5
3
25.7
17.8
0.7
42.9
4
8.2
7.3
8.7
20
194
57
2.4
26.6
6.5
29.1
36.9
22.8
21.6
238
4
5
3
234
2860
372
61
64
65
70.5
aIC50 values obtained from BACE1or BACE2 binding assay. bIC50 values obtained from BACE1 or
c
BACE2 FP assay. IC50 values obtained from whole cell assay measuring sAPP. dSpectrophotometric
28
ACS Paragon Plus Environment