Identification of a Potent, Highly Selective, and Brain Penetrant Phosphodiesterase 2A Inhibitor Clinical Candidate

Article abstract of DOI:10.1021/acs.jmedchem.7b01466

Computational modeling was used to direct the synthesis of analogs of previously reported phosphodiesterase 2A (PDE2A) inhibitor 1 with an imidazotriazine core to yield compounds of significantly enhanced potency. The analog PF-05180999 (30) was subsequen

Full text of DOI:10.1021/acs.jmedchem.7b01466

Subscriber access provided by READING UNIV
Article
Identification of a Potent, Highly Selective, and Brain
Penetrant Phosphodiesterase 2A Inhibitor Clinical Candidate
Christopher J. Helal, Eric P Arnold, Tracey L Boyden, Cheng Chang, Thomas A. Chappie, Ethan Fisher,
Mihaly Hajós, John F. Harms, William E. Hoffmann, John M. Humphrey, Jayvardhan Pandit, Zhijun Kang,
Robin J. Kleiman, Bethany L. Kormos, Che-Wah Lee, Jiemin Lu, Noha Maklad, Laura McDowell, Dina
McGinnis, Rebecca E O'Connor, Christopher J. O'Donnell, Adam Ogden, Mary Piotrowski, Christopher
J. Schmidt, Patricia Seymour, Hirokazu Ueno, Nichole Vansell, Patrick R. Verhoest, and Edward X. Yang
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01466 • Publication Date (Web): 02 Jan 2018
Downloaded from http://pubs.acs.org on January 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 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 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Identification of a Potent, Highly Selective, and
Brain Penetrant Phosphodiesterase 2A Inhibitor
Clinical Candidate
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a
a
a
b
Christopher J. Helal,* Eric Arnold, Tracey Boyden, Cheng Chang, Thomas A. Chappie,
a
a
b
a
a
Ethan Fisher, Mihaly Hajos, John F. Harms, William E. Hoffman, John M. Humphrey,
a
a
a
b
a
Jayvardhan Pandit, Zhijun Kang, Robin J. Kleiman, Bethany L. Kormos, CheꢀWah Lee,
a
a
a
a
a
Jiemin Lu, Noha Maklad, Laura McDowell, Dina McGinnis, Rebecca E. O’Connor,
a
a
a
b
Christopher J. O’Donnell, Adam Ogden, Mary Piotrowski, Christopher J. Schmidt, Patricia
a
a
a
b
a
A. Seymour, Hirokazu Ueno, Nichole Vansell, Patrick R. Verhoest, and Edward X. Yang
a
Pfizer Worldwide Research and Development, Eastern Point Road, Groton, CT 06340
Pfizer Worldwide Research and Development, 1 Portland Street, Cambridge, MA 02139
b
KEYWORDS: phosphodiesterase (PDE) 2A, cyclic guanosine monophosphate (cGMP), Nꢀ
methylꢀDꢀaspartate receptor antagonist, ketamine, MKꢀ801, cortical EEG, modified release
formulation
ABSTRACT. Computational modeling was used to direct the synthesis of analogs of previously
reported phosphodiesterase 2A (PDE2A) inhibitor 1 with an imidazotriazine core to yield
compounds of significantly enhanced potency. The analog PFꢀ05180999 (30) was subsequently
ACS Paragon Plus Environment

Page 3 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
identified as a preꢀclinical candidate targeting cognitive impairment associated with
schizophrenia. Compound 30 demonstrated potent binding to PDE2A in brain tissue, dose
responsive mouse brain cGMP increases, and reversal of NꢀmethylꢀDꢀaspartate (NMDA)
antagonistꢀinduced (MKꢀ801, ketamine) effects in electrophysiology and working memory
models in rats. Preꢀclinical pharmacokinetics revealed unbound brain/unbound plasma levels
approaching unity and good oral bioavailability resulting in an average concentration at steady
state (Cav,ss) predicted human dose of 30 mg once daily (QD). Modeling of a modified release
formulation suggested that 25 mg twice daily (BID) could maintain plasma levels of 30 at or
above targeted efficacious plasma levels for 24 hrs. which became part of the human clinical
plan.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Introduction
Phosphodiesterases (PDEs) play a central role in modulating cellular levels of the key second
messenger molecules cGMP and cAMP. Inhibition of PDEs as a means to selectively increase
discrete pools of cGMP and/or cAMP has emerged as strategy to treat a diverse set of human
medical conditions such as heart failure (PDE3 inhibitors), chronic obstructive pulmonary
disease, psoriatic arthritis (PDE4 inhibitors), and pulmonary hypertension and erectile
dysfunction (PDE5 inhibitors). We had a particular interest in identifying inhibitors of
phosphodiesterase 2A (PDE2A), a PDE that hydrolyzes both cGMP and cAMP, to treat
cognitive disorders. This approach was based upon the enrichment of PDE2A in striatum,
1
frontal cortex, and hippocampus, brain circuitry critical for selecting and enabling cognitive
programs, working memory and executive function. Additionally, data from rodent models
2,3
demonstrated efficacy for PDE2A inhibitors in different models of cognition. We recently
disclosed the potency, selectivity, pharmacokinetics, and pharmacology of a series of PDE2A
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
4
inhibitors represented by 1 (Figure 1), a pyrazolopyrimidine derivative. This data provided
support for PDE2A inhibition as a potential treatment for cognitive impairment associated with
schizophrenia (CIAS), a condition that is associated with hypoglutamatergic signaling, based
upon the ability of 1 to reverse the effects of NꢀmethyꢀDꢀaspartate (NMDA) antagonists, MKꢀ
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
01 and ketamine, in two different in vivo rodent models. The exquisite PDE2A selectivity is
proposed to be driven by the pꢀtrifluoromethyl phenyl inducing and occupying a pocket unique
5
,6
to PDE2A (Figure 2).
4
Figure 1. Profile of 1, a potent, selective, brain penetrant, in vivo active PDE2A inhibitor. HLM
Cl = intrinsic clearance apparent scaled determined from human liver microsomes (HLM); MDR
int
BA/AB = ratio of basolateralꢀtoꢀapical/apicalꢀtoꢀbasolateral flux in MadinꢀDarby Canine Kidney
(
MDCK) cells transfected with the multidrugꢀresistance gene (MDR1) that encodes for human Pꢀ
7
glycoprotein (Pꢀgp). C = unbound brain concentration. C = unbound plasma concentration.
bu
pu
ACS Paragon Plus Environment

Page 5 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2. Model of 1 bound to PDE2A based on xꢀray structure PDB 5U7L with pꢀ
trifluoromethylphenyl filling induced pocket.
While 1 met our target for potency, selectivity, and brain penetration for a preꢀclinical
candidate, we sought to reduce the estimated human dose (108 mg/day). The design strategy
was to reduce HLM Cl by targeting compounds of reduced lipophilicity as measured by shake
int
flask logD (SFlogD) (SFlogD of 1 = 2.9), while maintaining or improving the high potency and
low MDR BA/AB efflux associated with 1.
Results
Estimated human dose for a brain targeting compound (Equation 1a) was used as a multiꢀ
parameter optimization tool as compounds with lower human doses have been reported to be less
8
susceptible to idiosyncratic toxicology findings. In vitro data surrogates were used in place of
4
more labor intensive in vivo data to drive rapid decisions (Equation 1b) as described previously.
a) Estimated human dose, in vivo data = (Css,av,bu / (C /C )) x Cl x 24 h x 70 kg / Fa
bu pu
int
b) Estimated human dose, in vitro data = (10 x PDE2A IC x MDR BA/AB) x HLM Cl x 24
50
int
h x 70 kg
Equation 1. In vivo and in vitro human estimated dose equations
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
a) Css,av bu = targeted average unbound steady state brain concentrations of drug; C /C = unbound brain
bu pu
concentration/unbound plasma concentration; Clint = intrinsic clearance; Fa = fraction absorbed of the
drug. b) MDR BA/AB = ratio of basolateralꢀtoꢀapical/apicalꢀtoꢀbasolateral flux in MadinꢀDarby Canine
Kidney (MDCK) cells transfected with the multidrugꢀresistance gene (MDR1) that encodes for human Pꢀ
glycoprotein (Pꢀgp); HLM Clint = intrinsic clearance apparent scaled determined from human liver
microsomes (HLM).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Previous attempts to reduce lipophilicity in this series by incorporating azetidine derivatives
with polar appendages preserved PDE2A potency and afforded HLM Cl reductions (Figure 3,
int
4
3
), but significant, undesirable MDR BA/AB efflux was introduced. Alternatively, conversion
of the pendant pꢀsubstituted phenyl group into a heterocycle did reduce HLM Cl and retained
int
low MDR BA/AB efflux, but PDE2A potency suffered (4).
Figure 3. Attempts to reduce HLM Cl by reducing lipophilicity.
int
ACS Paragon Plus Environment

Page 7 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Subtle variations to the 1ꢀmethylꢀ1Hꢀpyrazolo[3,4ꢀd]pyrimidine core that would maintain key
proteinꢀligand contacts were next considered. Realization of a potency increase without altering
overall PDE selectivity and physical properties could allow for polarity enhancements of the
azetidine or pendant pꢀsubstituted aryl groups to align potency with desirable ADME properties.
We were inspired by a PDE5A example wherein alternative bicyclic cores provided differential
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
9
potency. For example, the PDE5A inhibitor vardenifil (5) is ca. 50x more potent than isomeric
6
, these compounds differing only by the juxtaposition of atoms in the bicyclic core that do not
10,11
make direct polar interactions with the PDE5A enzyme (Table 1).
Table 1. Impact of atom position on potency differences in 5 and 6
PDE5A IC50
ID
(nM)
5
0.17
6
8.9
We undertook a computational approach to see if the potency differences between 5 and
sildenafil (7) (6 with Nꢀmethyl vs. Nꢀethyl piperazine) could be explained by differential
hydrogen bonding interactions between the bicyclic cores and the PDE5A protein, with the
12
ultimate goal of predicting potency increases for our proposed PDE2A inhibitors. The Xꢀray
coꢀcrystal structures of 5 and 7 with the catalytic domain of PDE5A (Protein Data Bank (PDB)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
IDs 1XP0 and 1TBF, respectively) indicate that the imidazotriazinone and pyrazolopyrimidinone
bicyclic cores make hydrogen bond interactions with Gln817 (N(H)ꢀ3 and C(O)ꢀ4 in 5 and N(H)ꢀ
6 and C(O)ꢀ7 in 7), Gln775 (waterꢀmediated through C(O)ꢀ4 in 5 and C(O)ꢀ7 in 7) and Tyr612
sidechain/Asp764 backbone carbonyl (waterꢀmediated through Nꢀ6 in 5 and Nꢀ2 in 7).
Hydrogen bond strength calculations were computed for each individual residue or water
interaction with truncated versions of 5 (8) and 6 (9) then summed to determine an overall
contribution to the binding energy of these three hydrogen bond interactions (Table 2). In
summary, the relative binding energies correctly rank order the compounds with respect to
potency and also appear to be a reasonable approximation of the relative binding free energy
difference between the two compounds based on their potencies.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 2. Comparison of calculated hydrogen bonding strengths of truncated bicycle cores 8 and 9
with PDE5A versus measured PDE5 IC values for 5 and 6
5
0
-
O
6.4b
-5.0b
O
-
17.6a
-15.8a
N
HN
N -7.9c
HN
6.3c
N
-
N
N
N
8
9
Calculated
(Truncated System)
ꢀ
G
ꢀꢀG (kcal/mol)
(
kcal/mol)
8
9
ꢀ31.9
ꢀ27.1
0.0
4.8
Experimental
ꢀ13.6
5
6
0.0
2.3
ꢀ11.3
(a) Calculated Hꢀbond strength (kcal/mol) with highlighted atoms and Gln817. (b)
Calculated Hꢀbond strength (kcal/mol) with highlighted atom and structural water
interacting with Gln775. (c) Calculated Hꢀbond strength (kcal/mol) with highlighted atom
and structural water molecule interacting with Tyr612 and Asp764.
ACS Paragon Plus Environment

Page 9 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Encouraged by these promising results, we calculated hydrogen bond interaction strengths
for the pyrazolopyrimidine core (10) and an imidazotriazine core (11) to predict PDE2A
inhibitory potency (Table 3). The Xꢀray crystal structure of a compound with a
pyrazolopyrimidine core (PDB ID 5U7L, Figure 2) indicates that it hydrogen bonds with the
conserved Gln859, Gln812 and Tyr655/Asp808 backbone carbonyl (waterꢀmediated). As with
the PDE5 cores, hydrogen bond strength calculations were computed for each individual residue
or water interaction with 10 and 11, then summed to determine an overall contribution to the
binding energy of these three hydrogen bond interactions. As with the PDE5A example, these
calculations suggested that 11 would have stronger hydrogen bond interactions with PDE2A
versus 10. With this potential improvement in potency, we pursued synthesis of compounds of
general formula 12 (Figure 4).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 3. Comparison of calculated PDE2A potencies across bicyclic cores 10 and 11
-7.8b
-7.5b
N
N
N
N -6.1c
N
N -7.5c
N
N
-6.9a
-7.1a
N
N
10
11
Calculated (Truncated System)
ꢀG
ꢀꢀG
(kcal/mol)
1.3
(
kcal/mol)
1
1
0
1
ꢀ20.8
ꢀ22.1
0.0
See Figure 2 for proteinꢀligand contacts. (a) Calculated Hꢀbond strength (kcal/mol) with
highlighted atom and Gln859. (b) Calculated Hꢀbond strength (kcal/mol) with highlighted
atom and Gln812. (c) Calculated Hꢀbond strength (kcal/mol) with highlighted atom and
structural water molecule interacting with Tyr655 and Asp808.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. General imidazotriazine structure 12 targeted for synthesis
A key synthetic challenge for the imidazotriazine series was formation of the NꢀN bond
embedded within the bicyclic core. Initial routes sought to generate this bond via amination of
an imidazole (13 to 14, Figure 5), but yields were modest with side products and the most
1
3
effective amination reagent, Oꢀ(4ꢀnitrobenzoyl)hydroxylamine (15), was found to be
sufficiently energetic to preclude large scale usage based upon internal guidance.
Figure 5. Amination route of 13 to 14
To circumvent the NꢀN bond forming reaction, we sought an alternative disconnection with
this key bond derived from commercial starting materials (Figure 6). We felt 12 could be
generated from 16, wherein X is a leaving group and R is a functional group that would facilitate
introduction of the aryl/heteroaryl group. The intermediate 16 would be derived from 17 via
activation of the C=O group. Imidazotriazinone 17 would originate from Nꢀaminoimidazole 18,
which would derive from the condensation of αꢀbromoketone (19) with a protected (PG)
amidohydrazide 20, the source of the NꢀN bond that ultimately originates from readily available
1
4
hydrazine.
ACS Paragon Plus Environment

Page 11 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. Alternative imidazotriazinone retroꢀsynthesis starting with NꢀN bond formed.
In a forward sense, commercially available aminopyrazole 21 was converted to the 5ꢀbromoꢀ
pyrazole via treatment with copper (II) bromide and tertꢀbutyl nitrite followed by hydrolysis to
acid 22 (Scheme 1). This was followed by transformation of 22 to the acid chloride, then
treatment with trimethylsilyl diazomethane and hydrogen bromide to yield 23. Reaction of Nꢀ
amino amidine 24 (prepared from reaction between tertꢀbutoxycarbonyl hydrazide and Oꢀ
ethylacetimidate hydrochloride) with 23 occurred regioselectively to yield the targeted Nꢀ
aminoimidazole 25 and thus established the critical NꢀN bond of the subsequent imidazotriazine
ring system. Acidic removal of the BOC group followed by condensation with formamidine
acetate provided amidine 26. In order to construct the triazinone ring system, 26 was acylated
with carbonyl ditriazole (CDT) and heated, likely eliminating triazole to form an isocyanate
which resulted in intramolecular acylation to yield imidazotriazinone 27. Suzuki crossꢀcoupling
with (4ꢀ(trifluoromethyl)phenyl)boronic acid afforded 28, which was activated with
POCl /triazole, then treated with azetidine to afford 29.
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
Scheme 1. Synthesis of imidazotriazine analog 29. (a) CuBr , tertꢀbutyl nitrite, Et O, 65 C,
2
2
o
o
66% (b) LiOH, THF/MeOH/water, 23 C, 92% (c) i. NaOMe, MeOH; 23 C ii. Oxalyl chloride,
o
o
o
DMF (cat), CH Cl , 23 C iii. TMSꢀCHN , CH CN, 23 C; HBr/AcOH, 0 C, 78% (d) DIPEA, 2ꢀ
2
2
2
3
o
MeTHF/DME, reflux, 59% (e) TFA, CH Cl , 23 C, 96% (f) formamidine acetate, 2ꢀbutanol,
2
2
o
o
1
00 C, 72% (g) CDT, 1,4ꢀdioxane, 23 to 50 C, 66% (h) 4ꢀ(trifluoromethyl)phenyl boronic acid,
Pd(dppf)Cl ꢀCH Cl , Cs CO , iꢀPrOH/H O, reflux, 58% (i) i. 1Hꢀ1,2,4ꢀtriazole, POCl , Et N,
2
2
2
2
3
2
3
3
o
o
CH Cl , 70 C; ii. azetidine, Cs CO , DMF, 23 C, 85%.
2
2
2
3
The initial data with 29 showed a 5x increase in potency relative to 1 in line with
computational predictions (ꢀꢀG predicted = 1.3 kcal/mol; ꢀꢀG experimental = 1.1 kcal/mol),
but with the same SFLogD as 1, resulting in an improved lipophilic ligand efficiency (LipE = ꢀ
ACS Paragon Plus Environment

Page 13 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
15
log (PDE2A IC ) ꢀ SFLogD) (Table 4). Additionally, similar HLM Cl (93 ml/min/kg) and
1
0
50
int
MDR BA/AB (1.6) values were observed versus 1, suggesting that compound clearance and
transporter trends would translate from the pyrazolopyrimidine series to the imidazotriazine
series.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Table 4. Comparison of calculated hydrogen bond donor strengths of truncated systems (10 and
11) to measured ꢀG and ꢀꢀG values for 1 and 29.
Calculated (Truncated System)
ꢀ
G
ꢀꢀG
(
kcal/mol) (kcal/mol)
1
1
0
1
ꢀ20.8
ꢀ22.1
1.3
0.0
Experimental (Full System)
ꢀꢀG
PDE2A IC50
(kcal/mol) (kcal/mol) (nM) / (LipE)
ꢀG
1
ꢀ
12.3
13.4
1.1
0.0
1.6 (5.9)
0.3 (6.6)
2
9
ꢀ
The observed increase in potency enabled us to pursue analogs of calculated reduced
lipophilicity relative to 1 in our search for compounds of low clearance, acceptable human halfꢀ
life, and reduced estimated human dose (Table 5).
Table 5. Targeted imidazotriazine analogs of potentially lower lipophilicity
Cpd.
ID
ꢀNRR
ꢀAr
cLogP
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
30
1.6
3
3
4
1
9
0
1.3
3.0
1.9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A route similar to that for the synthesis of 29 was developed to access 30 and 31 using a CꢀH
arylation to install the trifluoromethyl and chloropyridine groups (Scheme 2). Thus, 4ꢀbromoꢀ1ꢀ
methylpyrazole (32) was lithiated and trapped with NꢀmethoxyꢀNꢀmethylacetamide to afford the
methyl ketone which was selectively monobrominated to yield αꢀbromoketone 34. In a manner
similar to that depicted in Scheme 1, 34 was converted to 35. Activation of 35 followed by
treatment with azetidine provided 36. A palladiumꢀcatalyzed CꢀH arylation on the pyrazole Cꢀ5
position with heteroaryl halides yielded final targets 30 and 31. It was found that CꢀH arylation
chemistry was superior to Suzuki coupling for heteroaromatic derivatives, likely due to
instability of the required heteroaryl boronates.
Scheme 2. Synthesis of imidazotriazine analogs 30 and 31 using a CꢀH arylation. (a) nꢀBuLi, Nꢀ
o
o
methoxyꢀNꢀmethyl acetamide, THF, ꢀ78 C, 57% (b) pyridiniumꢀBr , EtOH, CH Cl , 23 C,
3
2
2
o
8
9% (c) 24, DIPEA, 2ꢀMeTHF/DME, reflux, 69% (d) TFA, CH Cl , 23 C, 92% (e)
2 2
ACS Paragon Plus Environment

Page 15 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
o
formamidine acetate, 2ꢀbutanol, 100 C, 90% (f) CDI, THF, reflux, 70% (g) i. 1Hꢀ1,2,4ꢀtriazole,
o
POCl , Et N, CH CN, 70 C; ii. azetidineꢀHCl, Et N, 91%. (h) 2ꢀbromoꢀ5ꢀ
3
3
3
3
(trifluoromethyl)pyridine, allylpalladium(II) chloride dimer, K CO, dioxane, reflux, 93% (i) 2ꢀ
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
bromoꢀ5ꢀchloropyridine, allylpalladium(II) chloride dimer, K CO, dioxane, reflux, 52%.
2
The imidazotriazine 30 yielded a 6x PDE2A IC potency increase versus 4 with a similar
5
0
MDR BA/AB efflux ratio and slightly higher HLM Cl , resulting in an improved estimated
int
human dose of 49 mg/day (Table 6). The 5ꢀchloropyridine derivative 31 proved less potent than
3
0, but with a favorable reduction in lipophilicity afforded improved HLM Cl and provided a
int
lower human dose projection.
Table 6. PDE2A Potency, ADME, and Lipophilicity Measures for Imidazotriazine Analogs
4
30
31
39
40
PDE2A IC50
nM)
9
1.6
8.2
1.3
11
(
HLM Clint
(ml/min/kg)
MDR BA/AB
Estimated
3
0
52
13
33
9.5
1.6
1.5
1.4
1.1
1.2
human dose
178
49
45
20
65
(mg/day)
SFlogD
LipE
ꢀꢀ
ꢀꢀ
2.3
6.5
1.7
6.4
2.8
6.1
1.8
6.2
HLM Clint = intrinsic clearance apparent scaled determined from human liver microsomes (HLM).
MDR BA/AB = ratio of basolateralꢀtoꢀapical/apicalꢀtoꢀbasolateral flux in MadinꢀDarby Canine Kidney
(
MDCK) cells transfected with the multidrugꢀresistance gene (MDR1) that encodes for human Pꢀ
glycoprotein (Pꢀgp). See Equation 1b for estimated human dose. SFlogD = shake flask logD; LipE = (ꢀ
log (PDE2A IC ) – SFlogD).
10
50
In order to understand the clearance pathways to support designing lower clearance analogs,
metabolite identification studies in HLM on 30 were conducted (Figure 7). Results from this in
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
vitro evaluation revealed significant metabolism of the azetidine at both the 2 and 3 positions,
yielding the acid 37 and 3ꢀhydroxy azetidine 38, which were significantly less active than 30
(PDE2A IC = 72 nM and 12 nM, respectively), with both compounds likely to have low brain
5
0
ꢀ6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
availability (low passive permeability for 37, MDCK AB = 1.1 x 10 cm/sec; MDR BA/AB
efflux ratio = 12 for 38). Rat and dog liver microsomes and hepatocytes produced the same
metabolites as HLM and human hepatocytes (HHep) suggesting related clearance pathways
across species.
Figure 7. Structures of metabolites of 30 in HLM.
Efforts to simultaneously remove azetidine metabolically reactive sites while reducing
lipophilicity led to the preparation of Nꢀmethyl analogs 39 and 40 via routes related to previous
compounds (Scheme 3). Activation of 27 and reaction with methyl amine afforded 41 which
was subsequently subjected to a Suzuki crossꢀcoupling with 4ꢀ(trifluoromethyl)phenyl boronic
acid to yield 39. In the synthesis of 40, activation of 35 and S Ar with Nꢀmethylꢀ(pꢀ
N
methoxybenzyl)amine provided 42, the pꢀmethoxybenzyl (PMB) group being important in
facilitating the subsequent CꢀH arylation with 2ꢀbromoꢀ5ꢀ(trifluoromethyl)pyridine. Acidic
removal of the PMB proceeded cleanly to yield 40.
ACS Paragon Plus Environment

Page 17 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
o
Scheme 3. Synthesis of 39 and 40. (a) i. 1Hꢀ1,2,4ꢀtriazole, POCl , Et N, CH CN, 70 C; ii.
3
3
3
o
MeNH , Cs CO , DMF, 23 C, 78%. (b) 4ꢀ(trifluoromethyl)phenyl boronic acid, Pd(Ph P) ,
2
2
3
3
4
o
K PO , EtOH, H O, 75% (c) i. 1Hꢀ1,2,4ꢀtriazole, POCl , Et N, CH CN, 70 C; ii. MeNH(4ꢀ
3
4
2
3
3
3
o
methoxybenzyl), Et N, 23 C, 76% (d) 2ꢀbromoꢀ5ꢀ(trifluoromethyl)pyridine, allylpalladium(II)
3
chloride dimer, K CO , dioxane, reflux, 50% (e) TFA, anisole, CH Cl , reflux, 63%.
2
3
2
2
The analog 39 demonstrated a combination of potency, moderate microsomal clearance, and
low MDR BA/AB efflux to yield the lowest human dose projection (Table 6). The
corresponding 5ꢀtrifluoromethylpyridine 40 suffered in potency relative to 39 (ca. 10x), but
HLM Cl also dropped significantly which contributed to an acceptable estimated human dose.
int
Overall, these examples highlight the importance of considering both potency and ADME
properties for evaluation of compound drugꢀlikeness. Indeed, the LipE for the compounds in
Table 6 are all close as a result of the reduced lipophilicity of the lower potency compounds.
In terms of crossꢀPDE activity, targets 30, 31, 39, and 40 each maintained generally high
selectivity over other PDE families in line with previously observed data for the 1 (Table 6).
a
Table 6. PDE Selectivity for 30, 31, 39, and 40
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Compound
30
1.6
31
8.2
39
1.3
40
11
PDE2A IC (nM)
50
PDE Selectivity
PDE1B1
>56250
>56250
>56250
>56250
>56250
26969
>56250
>56250
2030
1556
2915
875
4647
>10976
3292
>10976
>10976
372
22945
>69231
>69230
>69230
>69230
34658
>69231
>69231
2370
7808
>8181.8
>56250
>8182
>8182
>8182
>8182
>8182
1049
PDE3A1
PDE4D3
PDE5A1
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PDE6 (Bovine)
PDE7B
PDE8B
PDE9A1
PDE10A1
PDE11A4
50090
3979
>69231
>8182
a) PDE selectivity defined as the ratio of PDEX IC / PDE2A IC . PDE selectivity color
5
0
50
legend – red: <1000; orange: 1000ꢀ5000; green: >5000.
The four compounds 30, 31, 39, and 40 were next evaluated in a rat oral exploratory toxicity
study for up to 15 days, at doses of 5, 50, and 250 mg/kg/day. Despite similarities in structure,
there were notable in vivo tolerability differences across this dose range and developmentꢀ
limiting microscopic findings in the heart and vasculature were observed with both 31 and 39.
Because of the degenerative nature of the findings and the low exposure margins from both the
maximal plasma concentrations (Cmax) and average plasma concentrations (C ) of active drug
av
relative to those expected to achieve efficacy (see Supporting Information), these compounds
were not advanced.
Compounds 30 and 40, however, were generally well tolerated at all doses and in particular,
there was no evidence of injury in the heart or vasculature which was attributed to either
compound in this study, despite high circulating plasma and brain exposures
achieved. Comparing offꢀtarget pharmacology (Cerep) did not show any significant differences
in 30, 31, 39, and 40 that would contribute to the varying toxicity findings. These results are
interesting as the small variations in structure from a trifluoromethyl to chloro (30 to 31) and
ACS Paragon Plus Environment

Page 19 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
phenyl to pyridyl (39 to 40), result in measurable in vivo tolerability differences and demonstrate
1
6
the challenge in predicting in vivo drug toxicity.
In a subsequent single dose dog study, 30 was tolerated at doses up to 30 mg/kg, and produced
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
no microscopic findings, including in the heart or vasculature, when administered at doses up to
1
5 mg/kg/day for 14 days. In contrast, 40 produced the microscopic finding of arteriopathy
within the arteries of the heart at high and intermediate doses (≥ 20 mg/kg/day). Exposure
margins at the no observable effect level (NOEL), the lowest dose tested, for arteriopathy, were
insufficient to continue advancement of 40 into development. During development, 30 was
administered to rats and dogs in studies up to 1 month in duration. Target organs identified in
these studies included skeletal muscle (rats) and the central nervous system (rats and dogs); no
effects were observed in the heart or vasculature. Based upon the lack of developmentꢀlimiting
in vivo safety findings with 30, further in vitro and in vivo pharmacological and pharmacokinetic
profiling was carried out.
The testing of 30 across a panel of kinases and other targets (Cerep) showed minimal activity
(see Supporting Information), demonstrating highly selective inhibition of PDE2A. We next
evaluated the ability of 30 to displace radiolabeled 2 in rat, dog, and monkey striatal tissue in
order to establish binding to native PDE2A which yielded a K = 4.2 – 8.4 nM (Table 7). These
i
values are within ca. 3ꢀ5 fold of the PDE2A enzyme IC .
5
0
Table 7. K values for displacement of 2 by 30 in striatal tissue and comparison to PDE2A IC
i
50
Species
Ki
nM)
Ki /
PDE2A
IC50
2.6
(
Rat
Dog
4.2
8.4
5.2
Monkey 5.5
3.4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
We have previously reported that the NMDA antagonist MKꢀ801 can disrupt low frequency
cortical delta oscillation in cortical electroencephalogram (EEG) recordings in rat brain, and that
4
PDE2A inhibition with 1 can reverse these deficits.
In this model, 30 showed complete
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
reversal of MKꢀ801’s effects at 0.1 mg/kg over the experiment time course (data not shown). At
.03 mg/kg, the reversal of MKꢀ801 was observed at early time points with the effect
0
diminishing over time and thus allowed for effective modeling of in vivo potency (Figure 8).
Similar activity in terms of both the extent of reversal of the MKꢀ801 effect and the similar fold
PDE2A IC C necessary to achieve the reversal as compared to 1 were achieved (Table 8).
5
0
bu
1
00
80
*
P < 0.05 vs. Control
P < 0.05 vs. MK-801
(n=4)
#
*
#
#
6
4
2
0
0
0
0
Control MK-801
5-15
15-25
i mi m af
25-35
os
35-45
45-55
(
0.07 mg/kg, IV)
T
T
e
e
a
t ef t
r
e
3
r
7
3
d
0
d
i
o
n gs i n( mi n( m. ) in)
g
Figure 8. 30 reverses the effect of MKꢀ801 on cortical EEG low frequency delta in rats, 0.03
mg/kg (iv dose). N = 4. Statistical significance was determined by means of twoꢀtailed paired
Student’s tꢀtests.
Table 8. Comparison of PDE2A IC and cortical EEG IC C values
5
0
80 bu
PDE2A IC50 EEG IC80
EEG IC80 /
PDE2A IC50
ID
(nM)
(C , nM)
bu
1
0
1.6
1.6
11
9
7
6
3
EEG = Electroencephalogram.
ACS Paragon Plus Environment

Page 21 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
As with 1, the examination of behavioral effects of 30 revealed no impact upon either
spontaneous or habituated locomotor activity in rats (sc dose range 0.32, 1.0 and 3.2 mg/kg, data
not shown). No effects of 30 (sc dose range 0.32, 1.0, and 3.2 mg/kg) were seen in the
conditioned avoidance responding assay, a measure used to predict antipsychotic potential, and
when dosed in combination with risperidone (0.032 and 0.32 mg/kg) did not alter the efficacy of
risperidone (data not shown). Compound 30 did, however, significantly attenuate the working
memory errors produced by ketamine in a working memory radial arm maze (RAM) model in
rats (0.032 to 0.32 mg/kg, SC dosing, Figure 9), demonstrating an ability to ameliorate deficits
produced by the NMDA antagonist. Satellite pharmacokinetic studies were used to estimate the
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
C levels associated with the efficacious doses, and ranged from 3.1 nM (0.1 mg/kg) to 12.5 nM
bu
(
0.32 mg/kg). At a dose of 1 mg/kg, however, the beneficial effect of 30 was significantly
attenuated, leading to a Uꢀshaped doseꢀresponse curve. A similar profile was also produced in
1
7
this assay in studies using an AMPA positive allosteric modulator. With regard to a potential
mechanistic understanding of this process, recent reports have shown that PDE2A inhibition can
18
potentiate aspects of dopamine D1 receptor signaling in vitro and it is well established that
working memory tasks typically demonstrate peak performance within an optimal range of
1
9
dopamine signaling. Thus, it is possible that the Uꢀshaped dose response curve by 30 in the
RAM test could be due to augmented D1 signaling exceeding an optimal range at higher doses,
which would be important to consider in a clinical setting.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
8
7
6
5
4
3
2
1
0
*
*
**
*
**
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
0
.76
6
.65
1.95
5.62
3.35
2.25
4.86
Veh + Veh
Veh +
Veh
Veh + Ketamine LY646(0.032) PF999 (0.032) + PF999 (0.1) + PF999 (0.32) + PF999 (1) + Ket
Veh +
Ketamine Ketamine
LY451646 +
37 (dose, mpk) + Ketamine
30
Ket Ket
5
.11
+Ket
Ket
0.032
0.1
0.32
1.0
Figure 9. The effect of 30 in a ketamineꢀdisrupted RAM working memory task in rats. N = 20ꢀ
3. Statistical method – Hochberg where *, ** = p <0.05, 0.01 vs. Vehicle + ketamine.
2
Administration of 30 to mice led to an acute and exposureꢀdependent elevation in the
accumulation of bulk levels of cGMP in the cortex, striatum and hippocampus (Figure 10), but
no changes in cAMP and the associated downstream phosphoꢀcAMP response elementꢀbinding
4
protein (pꢀCREB) as previously reported for 1. Given the established activity of PDE2 toward
cAMP, this suggests that the specific pool of cAMP regulated by PDE2A in these brain regions
may be relatively small relative to the total intracellular pool of cAMP measured in bulk
biochemical studies. Measurement of drug exposure in satellite animals was used to model the
cGMP response across brain regions. Following the dose response study, a time course study was
conducted to evaluate the temporal relationship between Cbu and cGMP response. Upon
confirmation of the direct response nature of cGMP levels to C of 30, results from the dose
bu
response study and the time course study were combined in order to estimate the potency of 30.
A linear model best described the relationship between Cbu and the corresponding cGMP
increase in all three brain regions using the current dose range and resulted in the estimated
potencies across the three brain regions that were quite comparable (Table 9). The Cbu E50
values in the mouse brain cGMP assay are significantly higher than the measured EEG IC and
8
0
ACS Paragon Plus Environment

Page 23 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
RAM C values. A positron emission tomography (PET) study in nonꢀhuman primates showed
bu
5
0% PDE2A occupancy by 30 at C = 57 nM, also severalꢀfold higher than the EEG IC and
bu 80
20
RAM C values. These results suggest that localized and/or small changes in cGMP may lead
bu
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
to these in vivo effects and that low measured levels of target occupancy could be sufficient for
human efficacy.
Figure 10. Mouse brain cGMP dose response for 30
Table 9. Estimated potency for cGMP increase of 30 across different brain regions
Brain
Hippoꢀ
campus
Striatum
Cortex
region
a
Slope
0
.51
0.43
0.63
79.4
(
%/nM)
Cbu
98.0
116.3
b
E50
a) Slope definition (%/nM): % increase in cGMP per nM of drug free in brain (C ). b)
bu
Projected C associated with 50% increase in brain region cGMP levels
bu
A variety of in vitro and in vivo pharmacokinetic assays were next used to profile 30. An
assessment for potential drugꢀdrug interactions (DDIs) with 30 showed no direct or timeꢀ
dependent inhibition of metabolism by representative human recombinant cytochrome P450
(CYP) enzymes (1A2, 2B6, 2C8, 2C9, 2C19, 2D6 and 3A) and no induction of CYP1A2. In
vitro, 30 was identified as a weak inducer of CYP3A4, however the risk of DDIs at the estimated
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
clinically relevant exposures was predicted to be low. Metabolism of 30 was primarily by
CYP3A4 and was expected to have a high risk eliciting DDIs via increased plasma levels of 30 if
coꢀadministered with both highly and moderately potent clinicallyꢀrelevant inhibitors of
CYP3A4, such as ketoconazole or erythromycin. This risk is mitigated, however, since the
majority of schizophrenia drugs which could be dosed with 30 are not typically CYP3A4
inhibitors and that large preꢀclinical safety margins were obtained with 30 from toxicology
studies in rats and dogs which could afford modest upward fluctuations in exposure levels upon
coꢀadministration.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The C /C and cerebrospinal fluid/Cpu ratio ((CSF)/C ) for 30 were evaluated across
bu pu
pu
different species from multiple studies (Table 10, see Supporting Information for full details). In
these studies, 30 demonstrated consistent access to brain and CSF, supporting the ability of 30 to
readily engage PDE2A in the central nervous system and built confidence in doing the same in
humans. These data, along with the high passive permeability in Ralph Russ Canine Kidney
ꢀ6
(RRCK) cells (26 x 10 cm/s) and low MDR BA/AB efflux ratio, suggest that it is reasonable to
21
use C /C = 1 in human dose projections based upon a recent analysis.
bu pu
Table 10. Mean C /C and CSF/C of 30 in Mouse, Rat, Dog, and Monkey
bu pu
pu
Species
C /C
bu pu
CSF/Cpu
a
b
c
a
a
d
c
c
CDꢀ1 Mouse
WistarꢀHan
Rat
Beagle Dog
Cynomolgus
Monkey
0.51
0.71
0.58
0.65
1.85
0.68
0.81
0
.60
Overall
Average
0
.65
0.98
(a) Determined from one study. (b) Average from seven different studies. (c) Average from
three different studies. (d) Average from two different studies. C /C = unbound brain
bu pu
ACS Paragon Plus Environment

Page 25 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
concentration/unbound plasma concentration; CSF/Cpu = cerebrospinal fluid/unbound
plasma concentration.
Following iv administration in rat, 30 demonstrated a high plasma clearance (CL ) of 122
p
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
mL/min/kg, which was not predicted by rat liver microsomes using the wellꢀstirred model (28.3
ml/min/kg). Combined with a moderate steady state volume of distribution (V ) of 3.9 L/kg, 30
ss
had a short halfꢀlife (t ) of 1.4 hours in rat (Table 11). In dog, the Cl was low (12.3
1
/2
p
mL/min/kg) and was consistent with the prediction from dog liver microsomal clearance (Clp
<13.4 ml/min/kg) which supported using HLM Cl to predict human Cl . No metabolism of 30
int
p
by aldehyde oxidase was observed and scaled clearance in HHep (HHep Cl = 44 ml/min/kg)
int
was similar to that in HLM, suggesting minimal nonꢀCYP mediated metabolism that provided
further support in using HLM Cl to predict human Cl .
int
p
Table 11. Mean Pharmacokinetics of 30 in Rat and Dog Following Single Oral and iv Dose
Administration
Species N
Dose
C
max
T
max
AUCinf
Cl
p
V
ss
T
1/2
%F
(mk/kg)/Route (ng/mL) (h) (ng•h/mL) (mL/min/kg) (L/kg) (hour)
b
Wistarꢀ 2
Han
5/Oral
363
0.08
ꢀꢀ
312
141
ꢀꢀ
ꢀꢀ
1.4
1.1
44
ꢀꢀ
c
d
1
/iv
1020
122
3.94
Rat
e
Beagle
Dog
2
5/Oral
1770
0.25
ꢀꢀ
4120
431
ꢀꢀ
ꢀꢀ
4.1
5.0
57
ꢀꢀ
b
d
0
.3/iv
293
12.3
2.89
ꢀꢀ = Not calculated; AUC = Area under the concentration time curve from time 0 to infinity;
inf
CL = Plasma clearance; C = Maximum observed concentration; % F = Bioavailability; N =
p
max
number of animals; Oral = Oral gavage; t = Terminal elimination phase halfꢀlife;SBECD =
½
sulfobutyletherꢀβꢀcyclodextrin; Tmax = Time of the first occurrence of Cmax; V = Steadyꢀstate
ss
volume of distribution.
a. F (%) = ([AUC (Oral) × Dose (IV)] / [AUC (IV) × Dose (Oral)]) × 100
inf
inf
b. Vehicle was 20% SBECD + 2 molar equivalent of 1N hydrochloric acid.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
c. Vehicle was 2 molar equivalent of 1N hydrochloric acid.
d. Extrapolated concentration to time zero (C0).
e. Vehicle was 0.5% methylcellulose + 2 molar equivalent of 1N hydrochloric acid
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The predicted CL of 30 in humans (8.8 ml/min/kg) was estimated using HLM Cl in the
p
int
wellꢀstirred model, adjusted for human plasma and microsomal protein binding and afforded an
estimated bioavailability of 32% (Table 12). Human predicted V was approximated using the
ss
rat and dog V values adjusted for differences in species plasma protein binding, and yielded a
ss
value of 1.9 L/kg. The estimated human t1/2 was 2.5 h. Using the aforementioned predicted
2
2
pharmacokinetic parameters with a Cav,ss of 11 ng/mL, a predicted human dose of 30 mg/day
was estimated, close to the value of 49 mg/day determined from in vitro values in Equation 1b.
Table 12. Human predictions of 30: Pharmacokinetics and efficacious dose based upon
animal data.
Clp
mL/min/kg)
8.8
F (%)
32
30
(
Cav,ss QD
Dose
mg/day)
1
2
.9
.5
V (L/kg)
ss
(
t (h)
1
/2
Cl = Plasma clearance; F% = Bioavailability; F = Fraction absorbed; t = half life; V =
p
a
1/2
ss
Volume of distribution at steady state.
To ensure that the mechanism of action is adequately evaluated for efficacy of a drug, it is
often desirable to administer doses that will allow minimum drug concentrations (C ) to remain
min
above an efficacious concentration threshold for the entire dosing interval versus maintaining an
average concentration. Since the data from both the mouse cGMP and rat cortical EEG models
suggested that these effects were the result of direct PDE2A inhibition, we utilized
ACS Paragon Plus Environment

Page 27 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
pharmacokinetic modeling using projected PK parameters presented in Table 12 to explore
human doses of 30 that would be projected to maintain Cmin at or above the Cav,ss target of 11
ng/mL for 24 hrs. The Cav,ss based 30 mg immediate release (IR) oral dose of 30 was predicted
to maintain exposure at or above the targeted Cmin for approximately 7 h (Figure 11a). A single
dose of 2700 mg IR could sustain the Cmin threshold over 24 h but it was predicted to have a Cmax
that was above the human Cmax exposure limit established from the exposureꢀlimiting finding of
convulsions in female rats (Figure 11b). We next modeled 90 mg, twiceꢀdaily dosing (BID),
which suggested that sustained plasma levels at or above Cmin were attainable for 24 h with an
acceptable plasma Cmax (Figure 11c). In order to further mitigate against adverse events while
maintaining Cmin exposure and minimizing total drug exposure as based on the area under the
concentrationꢀtime curve (AUC), a modifiedꢀrelease (MR) formulation was considered.
Importantly, the ADME and physical properties of 30, especially permeability and
thermodynamic solubility (340 ug/mL, pH 6.6 buffer), were suitable for development of an MR
formulation with a release duration of approximately 8ꢀ10 hours utilizing an osmotic release
mechanism. Subsequent modeling indicated that a 25 mg MR, BID dosing paradigm could
maintain Cmin plasma levels for 24 h (Figure 11d), reducing overall dose almost 4ꢀfold when
compared to the 90 mg IR, BID formulation. It was decided that this MR formulation would be
evaluated in human clinical trials and is the subject of future publications.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
a)
b)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 66
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
c)
d)
Figure 11. Predicted human plasma concentration profiles of 30 using different oral dosing
paradigms and formulations at steady state. a) 30 mg IR, QD. b) 2700 mg IR, QD. c) 90 mg
IR, BID. d) 25 mg MR, BID. Dotted lines = 11 ng/mL. Dashed lines = 374 ng/mL (human
th
Cmax exposure limit based on 1/10 the convulsion NOEL from 1ꢀmonth rat toxicology study).
BID = twice daily; IR = immediate release; NOEL= no observed effect level; QD = once daily.
Discussion and Conclusions
Starting with pyrazolopyrimidine lead 1, potential potency improvement through the use of an
imidazotriazine core was identified via application of hydrogen bond acceptor strength
calculations. These potency enhancements were realized upon synthesis of target molecules,
providing four compounds (30, 31, 39, 40) with low projected human dose, high PDE selectivity,
and reduced lipophilicity relative to 1. Multiꢀday in vivo toleration studies of these compounds
ACS Paragon Plus Environment

Page 29 of 66
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
revealed significant differences in histopathological findings with only small differences in
structure. The superior safety profile of 30 led to further evaluation of its in vitro and in vivo
PDE2A activity. Radioligand 2 was displaced from brain tissue by 30 with a K ranging from
i
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
.2ꢀ8.4 nM, with this higher value being close to the cortical EEG IC related C value of 9 nM
80 bu
and efficacious ketamine RAM C measures of 3.1 to 12.5 nM. In line with the PDE inhibitory
bu
activity of 30, mouse brain cGMP levels increased with increasing drug exposure in a direct,
linear manner in multiple brain regions. This pharmacology data is comparable to that observed
with 1 and supports the ability of 30 to have potential impact on functional measures disrupted
by NMDA receptor hypofunction, which could include cognitive endpoints impacted in
schizophrenia. Pharmacokinetic analyses across species built confidence in human brain
penetration with C /C approaching unity, along with good oral bioavailability, resulting in a
bu pu
predicted Cav,ssꢀbased dose in human of 30 mg/day. In order to ensure constant inhibition of
PDE2A via a Cmin dosing paradigm, pharmacokinetic modeling suggested that a ca. 50 mg/day
(25 mg, BID) MR formulation of 30 could achieve this and an extended release formulation
approach was planned to be explored in human volunteers.
Experimental Section
Chemistry
Experiments were generally carried out under inert atmosphere (nitrogen or argon), particularly
in cases where oxygenꢀ or moistureꢀsensitive reagents or intermediates were employed.
Commercial solvents and reagents were generally used without further purification, including
TM
anhydrous solvents where appropriate (generally SureꢀSeal
products from the Aldrich
Chemical Company, Milwaukee, Wisconsin). Mass spectrometry (MS) data is reported from
ACS Paragon Plus Environment