Minimization of drug-drug interaction risk and candidate selection in a natural product-based class of gamma-secretase modulators

Article abstract of DOI:10.1016/j.bmcl.2015.01.051

Early lead compounds in this gamma secretase modulator series were found to potently inhibit CYP3A4 and other human CYP isoforms increasing their risk of causing drug-drug-interactions (DDIs). Using structure-activity relationships and CYP3A4 structural information, analogs were developed that minimized this DDI potential. Three of these new analogs were further characterized by rat PK, rat PK/PD and rat exploratory toxicity studies resulting in selection of SPI-1865 (14) as a preclinical development candidate.

Full text of DOI:10.1016/j.bmcl.2015.01.051

Bioorganic & Medicinal Chemistry Letters 25 (2015) 1621–1626
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Minimization of drug–drug interaction risk and candidate selection
in a natural product-based class of gamma-secretase modulators
a,
Jed L. Hubbs a, , Nathan O. Fuller , Wesley F. Austin a, Ruichao Shen a,à, Jianguo Ma a,§, Zhen Gong b,
⇑
Jian Li b, Timothy D. McKee a,–, Robyn M. B. Loureiro a,k, Barbara Tate a, , Jo Ann Dumin a, Jeffrey Ives a,
Brian S. Bronk a,àà
a Satori Pharmaceuticals, Inc., 281 Albany Street, Cambridge, MA 02139, USA
b WuXi AppTec Co., Ltd, 288 Fute Zhong Road, Shanghai 200131, China
a r t i c l e i n f o
a b s t r a c t
Article history:
Early lead compounds in this gamma secretase modulator series were found to potently inhibit CYP3A4
and other human CYP isoforms increasing their risk of causing drug–drug-interactions (DDIs). Using
structure–activity relationships and CYP3A4 structural information, analogs were developed that mini-
mized this DDI potential. Three of these new analogs were further characterized by rat PK, rat PK/PD
and rat exploratory toxicity studies resulting in selection of SPI-1865 (14) as a preclinical development
candidate.
Received 19 November 2014
Revised 20 January 2015
Accepted 21 January 2015
Available online 10 February 2015
Keywords:
Alzheimer’s disease
Amyloid-beta
Ó 2015 Elsevier Ltd. All rights reserved.
Gamma secretase modulator
Cytochrome P450 3A4
Sterol
Natural product
Semisynthesis
Drug–drug-interactions
Gamma secretase modulators (GSMs) are promising agents for
the treatment of Alzheimer’s disease (AD).1,2 We have previously
disclosed the identification and optimization of a new class of
natural product-based GSMs as potential disease-modifying treat-
ments for AD.3–6 Our disclosure highlighted two triterpene-derived
compounds, 1 and 2, with promising GSM activity both in vitro and
in Sprague-Dawley rats (Fig. 1).3 These compounds also exhibited
promising pharmacokinetic properties in rodents. These and relat-
ed compounds were also found to potently inhibit cytochrome
P450 enzymes, particularly CYP3A4, and therefore carried an unac-
ceptably high potential to cause drug–drug interactions (DDIs),7
precluding any further advancement toward clinical studies. In this
disclosure we describe additional optimization that allowed us to
minimize the DDI potential and identify SPI-1865 (14) as our pre-
clinical candidate.
Compounds 1 (CYP3A4 IC50 = 1.1
lM) and 2 (CYP3A4 IC50 = 12
lM) and several closely related morpholine analogs that displayed
good to excellent inhibition of Ab42 production (91–400 nM) in
our cell-based assay were screened against the 5 major human
CYP isoforms for inhibition, as measured by their impact on the
rate of disappearance of known substrates (Table 1). The majority
of compounds inhibit CYP3A4 more strongly than other CYP iso-
forms. The strongest CYP 3A4 inhibitor (3, 99% inhibition at
Abbreviations: Ab, amyloid beta; AD, Alzheimer’s disease; DDIs, drug–drug
interactions; GSM, gamma-secretase modulator; iv, intravenous; mg/kg, milligrams
per kilogram of body weight; PD, pharmacodynamic; PK, pharmacokinetic; po, per
os (by mouth); SAR, structure–activity relationship.
⇑
Corresponding author at present address: ETH Zürich, HCI E306, Vladimir-
Prelog-Weg 3, 8093 Zürich, Switzerland. Tel.: +41 078 729 75 43.
10
l
M) does not contain a substituent on the morpholine nitrogen,
M; 12, 22% at 10 M)
while the weakest inhibitors (2, 29% at 10
l
l
E-mail address: jedhubbs@gmail.com (J.L. Hubbs).
contain a bulky basic group. From the SAR, we quickly noted that
the connectivity of the second basic amine was important as 9,
an ortholog of 12, (62% inhibition at 10 lM) was also a moderately
strong inhibitor. Adding a methyl or ethyl substituent to the mor-
pholine did not significantly decrease inhibition (4, 71% inhibition
at 10 lM; 5, 81% inhibition at 10 lM), especially as compared to 2
AstraZeneca Pharmaceuticals, 35 Gatehouse Drive, Waltham, MA 02451, USA.
Enanta Pharmaceuticals, 500 Arsenal Street, Watertown, MA 02472, USA.
939 Rue Chantilly, Mandeville, LA 70471, USA.
Biogen Idec, 14 Cambridge Center, Cambridge, MA 02139, USA.
à
§
–
k
Shire plc, 300 Shire Way, Lexington, MA 02421, USA.
Head of Biology, Rodin Therapeutics, 25 First Street, Cambridge, MA 02141, USA.
Sanofi, 640 Memorial Drive, Cambridge, MA 02139, USA.
àà
http://dx.doi.org/10.1016/j.bmcl.2015.01.051
0960-894X/Ó 2015 Elsevier Ltd. All rights reserved.

1622
J. L. Hubbs et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1621–1626
OH
OH
Me
H
Me
H
Me
Me
Me
Me
Me
Me
H
H
H
H
OEt
OEt
O
H
O
H
O
H
O
H
H
Me
H
Me
OH
OH
N
N
O
MeO
O
N
Me Me
Me Me
Me
1
2
Aβ42 IC50 = 270 nM
Aβ42 IC50 = 100 nM
CYP 3A4 IC50 = 1.1 μM
CYP 3A4 IC50 = 12 μM
Figure 1. Satori GSMs.
Table 1
Human CYP isoform screening data
OH
Me
Me
Me
Me
H
OEt
H
O
H
H
O
H
H
Me
OH
N
R
O
Me Me
Compd
R
Ab42
IC50
pKa
3A4 % inhibition @
10 mM
2C19 % inhibition @
10 mM
1A2 % inhibition @
10 mM
2C9 % inhibition @
10 mM
2D6 % inhibition @
10 mM
1
270
100
6.3
87
35
25
30
16
MeO
2
9.4*
29
42
20
0
33
N
Me
H
3
4
5
91
250
210
7.8
7.2
7.3
99
71
81
40
31
43
35
20
16
31
21
21
6
16
9
Me
Et
6
7
270
170
5.1
62
74
24
28
19
14
22
26
4
8
O
Me
N/A
Me
O
O
8
250
5.2
80
31
22
44
13
N
N
9
110
400
9.4*
5.2
62
49
59
31
35
5
2
49
2
10
14
O
Me
11
12
100
140
10.7*
9.6*
74
22
68
17
20
28
91
2
17
29
HN
Me
N
*
Most basic nitrogen on R-group.
or 12. Further reducing the basicity of the nitrogen atom (see cal-
culated pKa values) also failed to eliminate CYP inhibition (6, 62%
at 10
removing the amine completely by converting the morpholine
amine to an amide, as in compound 7 (74% inhibition at 10 M),
lM; 10, 49% inhibition at 10 lM). Finally, we found that
inhibition at 10
l
M; 1, 87% inhibition at 10
lM; 8, 80% inhibition
l

J. L. Hubbs et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1621–1626
1623
A
O
N
N
Cl
Cl
O
O
O
N
N
Me
CH2
Me
Me
N
N
N
Fe
N
Me
HO2C
CO2H
B
Me
Me
HO
Me
Me
H
AcO
O
O
Me
H
H
O
H
HO
N
Me
Me Me
CH2
Me
N
N
N
Fe
N
Me
Me
HO2C
CO2H
Me
Me
HO
Me
Me
C
H
AcO
O
O
Me
H
H
O
HO
N
N
Me Me
Me
Me
CH2
Me
N
N
Fe
N
N
Me
Me
HO2C
CO2H
Me
Me
HO
Me
Me
D
H
AcO
O
Me
O
N
Me
H
H
O
HO
N
Me Me
Me
CH2
Me
N
N
N
N
Fe
Me
Me
HO2C
CO2H
Figure 2. CYP3A4 docking: (A) ketoconazole, (B) 3, (C) 2, (D) 12.

1624
J. L. Hubbs et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1621–1626
OH
Me
Me
H
Me
Me
OH
Me
H
Me
Me
Me
H
H
1. Coupling partner,
conditions
2. HCl, MeOH, H2O
H
H
OEt
O
H
OEt
O
H
O
H
O
H
H
Me
OH
HN
H
Me
OH
N
O
R
O
Me Me
Me Me
3
compd
R
Coupling partner
O
Conditions
11
NaBH(OAc)3, 4 Å mol sieves
DCM, AcOH
N
HN
HN
Boc
Boc
N
O
13
NaBH(OAc)3
DCM, MeOH, AcOH
H
Scheme 1. Analog synthesis.
also did not eliminate CYP inhibition. Although we deemed the CYP
inhibition as being too potent for all of these analogs, these struc-
ture–activity relationships suggested to us that b-diamines such as
2 and 12 were most promising for overcoming CYP3A4 inhibition.
We turned to computational modeling to better understand the
interactions responsible for CYP3A4 inhibition (Fig. 2). The crystal
structure of human CYP3A4 bound to ketoconazole8 (Fig. 2A) was
used as a starting point for these studies. When the unsubstituted
morpholine compound 3 was docked into this model it was found
that the nitrogen atom of the secondary amine likely coordinates
to the heme iron in a manner analogous to the imidazole nitrogen
atom in ketoconazole (Fig. 2B). In contrast, docking compound 2
suggested a repulsive interaction between the positively charged
heme and less basic azetidine amine (Fig. 2C), likely protonated
at physiological pH. In addition to apparently disrupting a key
interaction with the heme iron, this structural change also
appeared to break a hydrogen bonding network with Thr 224
and the C24–C25 side chain. Docking compound 12 suggests that
the added morpholine substituent size prevents the nitrogen
atoms from approaching the heme iron atom (Fig. 2D). These com-
bined results suggest that a combination of positive charge and
steric bulk on the morpholine substituent reduce CYP inhibition.
Based on these results, we focused our attention on morpho-
line diamines containing an azetidine directly attached from the
three position to the morpholine nitrogen atom (analogs of 2)
or one linked with a methylene unit (analogs of 12). Our two
key intermediates (11 and 13) were readily prepared from the
secondary amine intermediate 3, as shown in Scheme 1. Com-
pound 11 was prepared by coupling N-boc-3-azetidinone with
the morpholine under reductive amination conditions, followed
by cleaving the carbamate with HCl in water and methanol.
Similarly, amine 13 was prepared by utilizing N-boc-3-
azetidinecarboxaldehyde in the reductive amination, followed
by carbamate hydrolysis.
(PK), efficacy (Ab42 lowering) and safety profile necessary for fur-
ther development. In order to more comprehensively assess the
DDI potential of each compound, we elected to look at both the
direct and time-dependent (metabolism-based) inhibition of
CYP3A4 in the presence of each compound. Time-dependent inhi-
bition of CYP3A4 was determined by measuring human liver
microsome metabolism of midazolam after 30 min in two study
arms. In the first arm (non-time dependent), the test compounds
were preincubated with microsomes before the addition of mida-
zolam and NAHPH at t = 0 of the metabolism study. In the second
arm (time dependent), NADPH was included in the preincubation
step prior to the addition of midazolam at t = 0. The first arm is
reported as the CYP 3A4 IC50 in Table 2 and the second arm is
reported as the time dependent CYP3A4 IC50. Compounds with a
CYP3A4 IC50:time dependent CYP3A4 IC50 of P2 were considered
time-dependent inhibitors. We found that analogs that contained
a diamine without a basicity-attenuating oxygen atom were gener-
ally less potent CYP inhibitors (12
lM for 2 to >100
lM for 15) than
those that contained such an oxygen atom (3
lM for 18 to 17 lM
for 25). We attribute this difference to a repulsion between the
positively charged protonated nitrogen atom and the heme iron
in the more basic compounds, that is, lacking in the less basic com-
pounds. Time-dependent inhibition was noted with compounds
containing the oxetane substituent (16, 7.5-fold IC50 difference;
22, 3.5-fold IC50 difference) and with those containing a methox-
yethyl group (17, 2.0-fold difference; 23, 2.8-fold difference). In
addition, 2 was a time-dependent inhibitor (3.1-fold IC50 differ-
ence). Although the mechanism of the time-dependent inhibition
has not been fully explored, one potential explanation relies on
metabolism to release 3 and/or 11, both of which show potent
3A4 inhibition. Based on these results, three compounds (14, 15,
and 24) were selected for further studies based on a combination
of structural diversity and lower levels of CYP inhibition.
The rat PK for compounds 14, 15, and 24 is shown in Table 3. All
compounds show low clearance (Cl; 0.5–2.5 mL/min/kg), high vol-
umes of distribution (Vd; 5.8–14.3 L/kg), a long half-life (79–129 h)
and moderate to good bioavailability (%F = 39% to ꢀ100%). Also
noteworthy is the delayed Tmax (40–44 h), which we attribute to
lymphatic absorption9 based on the high lipophilicity of the com-
pounds (14, clogP = 5.7; 15, clogP = 5.9; 24, clogP = 5.0).
Amines 11 and 13 were used to make a series of seven analogs
selected to influence the basicity and/or the steric environment
around the nitrogen atoms. Across both systems, we incorporated
alkyl substituents (Me, i-Pr and cyclobutyl) and oxygen-containing
groups (oxetane, methoxyethyl, 2-hydroxyisobutyl and hydrox-
yethyl) on the pendant azetidine amine. Our hypothesis was that
control of the steric and electronic environment in this set of ana-
logs would both minimize DDI risk and fine-tune key physico-
chemical properties, thereby maximizing the chances that one of
these compounds would possess the desired pharmacokinetic
These three compounds were next examined in rat pharmaco-
dynamic experiments. The doses for 14, 15, and 24 were selected
to give similar plasma exposure levels based on their individual
rat PK profile. Consistent with our expectations, all compounds

J. L. Hubbs et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1621–1626
1625
Table 2
Azetidinediamine analogs Ab42 IC50 and CYP3A4 inhibition
OH
Me
Me
H
Me
Me
H
H
OEt
O
H
O
H
H
Me
OH
N
R1
O
Me Me
R1
R2
Ab42 IC50
pKa
CYP3A4 IC50
12
(lM)
Time dependent CYP3A4 IC50
3.9
(
lM)
Ratio IC50:TDI
a
Compd
Me
Me
2
100
9.4
3.1
1.5
N
N
N
N
N
N
R2
R2
R2
R2
R2
R2
14
15
16
17
18
19
110
120
130
68
9.6
10.1
8.5
37
>100
12
25
54
Me
>1.9
7.5
1.6
2.7
2.7
3.2
O
MeO
8.4
5.3
2.0
Me Me
HO
120
ND
9.1
3.0
1.1
HO
Me
8.4
4.9
1.5
N
N
R2
R2
12
20
21
22
23
24
25
110
110
210
170
86
9.6
9.8
25
15
13
6.6
22
14
17
25
16
1.0
0.94
1.7
R2
Me
Me
N
N
N
N
N
N
R2
R2
R2
R2
R2
10.5
8.8
7.7
1.9
7.8
11
3.5
O
MeO
8.6
2.8
Me Me
HO
86
9.3
1.3
HO
120
8.6
17
1.0
a
CYP3A4 IC50 after pre-incubation with NADPH. ND: IC50 was not determined.
gave good brain exposure (B:P = 0.69 to 2.4) and statistically sig-
nificant lowering of Ab42 (Table 4). These data were not used to
differentiate between compounds but instead set the stage for
establishing a therapeutic window. Toward that goal, these com-
pounds were examined in a 14-day exploratory safety study in
the same strain of rats. In this study, compound 14 was tolerated
at 90 mg/kg/day, 40 M average plasma concentration (Cave); com-
pound 15 was tolerated at a dose of 15 mg/kg/day (5.1 M Cave);
and compound 24 was tolerated at a dose of 60 mg/kg on the first
day, followed by 10 mg/kg on days 2–14 (8.5 M Cave). Based on
l
l
l

1626
J. L. Hubbs et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1621–1626
Table 3
Author contributions: All authors have given final approval to the
final version of this manuscript.
Sprague-Dawley rat PK (iv/po) for compounds 14, 15, and 24
Notes: The authors note the following relevant financial inter-
ests: the authors are named as inventors on one or more patents
and patent applications related to compounds discussed in this
Letter.
Compd
14
15
24
Cl (iva, mL/min/kg)
Vd (iva, L/kg)
t1/2 (iva, h)
0.50
5.8
129
13.9
40
1.0
6.2
80
11.2
44
2.5
14.3
79
11.9
40
>100
Cmax (pob, mM)
Tmax (pob, h)
%F (pob)
Acknowledgments
50
39
a
1 mg/kg.
100 mg/kg.
We thank the CADD and DMPK departments at WuXi AppTec
for their contributions to this program.
b
Supplementary data
Table 4
Sprague-Dawley rat PK/PD for compounds 14, 15 and 24.
Supplementary data (synthesis and analytical data for interme-
diates and exemplary compounds. Details for the Ab, CYP inhibi-
tion, rodent pharmacokinetics, and rodent pharmacodynamics
studies) associated with this article can be found, in the online ver-
sion, at http://dx.doi.org/10.1016/j.bmcl.2015.01.051.
Compd
14
15
24
Dose (mpk)
Paradigm
Time point
[Plasma]a
[Brain]a
100
Single Dose
24 h
150
Single Dose
24 h
100
Single Dose
24 h
14
33
9.1
9.0
6.3
17
Ab42 levelb,c
50 15%⁄⁄⁄
29 8%⁄⁄
23 16%⁄
References and notes
a
b
c
l
M.
1. Hall, A.; Patel, T. R. Prog. Med. Chem. 2014, 53, 101.
2. Pettersson, M.; Stepan, A. F.; Kaufmann, G. W.; Johnson, D. S. Expert Opin. Ther.
Pat. 2013, 23, 1349.
3. Findeis, M. A.; Schroeder, F. C.; McKee, T. D.; Yager, D.; Fraering, P. C.; Creaser, S.
P.; Austin, W. F.; Clardy, J.; Wang, R.; Selkoe, D. J.; Eckman, C. B. ACS Chem.
Neurosci. 2012, 3, 941.
Reduction relative to vehicle, brain homogenate.
Statistically significant lowering indicated by
⁄
⁄⁄
⁄⁄⁄
or (p-value is <0.05, 0.01,
,
or 0.001, respectively). The number of animals used in the study was N = 12 per
group. The value reported is mean standard deviation. Using an ANOVA with a
Dunnett’s test, the p-value is <0.01.
4. Fuller, N. O.; Hubbs, J. L.; Austin, W. F.; Creaser, S. P.; McKee, T. D.; Loureiro, R.
M. B.; Tate, B.; Xia, W.; Ives, J. L.; Findeis, M. A.; Bronk, B. S. ACS Med. Chem. Lett.
2012, 3, 908.
5. Hubbs, J. L.; Fuller, N. O.; Austin, W. F.; Shen, R.; Creaser, S. P.; McKee, T. D.;
Loureiro, R. M. B.; Tate, B.; Xia, W.; Ives, J. L.; Bronk, B. S. J. Med. Chem. 2012, 55,
9270.
6. Austin, W. F.; Hubbs, J. L.; Fuller, N. O.; Creaser, S. P.; McKee, T. D.; Loureiro, R.
M. B.; Findeis, M. A.; Tate, B.; Ives, J. L.; Bronk, B. S. Med. Chem. Commun. 2013, 4,
569.
7. Lin, A. H.; Lu, A. Y. H. Clin. Pharmacokinet. 1998, 35, 361.
8. Ekroos, M.; Sjögren, T. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 13682.
9. Wasan, K. M.; Brocks, D. R.; Lee, S. D.; Sachs-Barrable, K.; Thorton, S. J. Nat. Rev.
Drug Disc. 2008, 7, 84.
these results combined with the efficacy and CYP inhibition data,
compound 14 was chosen as
a candidate for preclinical
development.10
We found that our initial lead compounds were likely DDI per-
petrators due to their potent CYP3A4 inhibition. Structure–activity
relationships aided by in silico modeling of these compounds
revealed options for avoiding these liabilities. Based on this infor-
mation, a series of compounds was designed and synthesized and
found to have reduced CYP3A4 inhibition while maintaining other
key attributes of the initial leads. Three were chosen for additional
evaluation in PK, PK/PD and safety studies resulting in 14 (SPI-
1865) being selected for development.11
10. Loureiro, R. M.; Dumin, J. A.; McKee, T. D.; Austin, W. F.; Fuller, N. O.; Hubbs, J.
H.; Shen, R.; Jonker, J.; Ives, J.; Bronk, B. S.; Tate, B. Alzheimers Res. Ther. 2013, 5,
19.
11. The development of SPI-1865 was halted.