Discovery of triazolopyrimidine-based PDE8B inhibitors: Exceptionally ligand-efficient and lipophilic ligand-efficient compounds for the treatment of diabetes

Article abstract of DOI:10.1016/j.bmcl.2012.06.079

PDE8B is a cAMP-specific isoform of the broader class of phosphodiesterases (PDEs). As no selective PDE8B inhibitors had been reported, a high throughput screen was run with the goal of identifying selective tools for exploring the potential therapeutic utility of PDE8B inhibition. Of the numerous hits, one was particularly attractive since it was amenable to rapid deconstruction leading to inhibitors with very high ligand efficiency (LE) and lipophilic ligand efficiency (LLE). These triazolopyrimidines were optimized for potency, selectivity and ADME properties ultimately leading to compound 42. This compound was highly potent and selective with good bioavailability and advanced into pre-clinical development.

Full text of DOI:10.1016/j.bmcl.2012.06.079

Bioorganic & Medicinal Chemistry Letters 22 (2012) 5721–5726
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Discovery of triazolopyrimidine-based PDE8B inhibitors:
Exceptionally ligand-efficient and lipophilic ligand-efficient
compounds for the treatment of diabetes
,
⇑
Michael P. DeNinno , Stephen W. Wright, John B. Etienne, Thanh V. Olson, Benjamin N. Rocke,
Jeffrey W. Corbett, Daniel W. Kung, Kenneth J. DiRico, Kim M. Andrews, Michele L. Millham,
Janice C. Parker, William Esler, Maria van Volkenburg, David D. Boyer, Karen L. Houseknecht à,
Shawn D. Doran
Pfizer Global Research & Development-Groton Laboratories, Groton, CT 06340, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
PDE8B is a cAMP-specific isoform of the broader class of phosphodiesterases (PDEs). As no selective
PDE8B inhibitors had been reported, a high throughput screen was run with the goal of identifying selec-
tive tools for exploring the potential therapeutic utility of PDE8B inhibition. Of the numerous hits, one
was particularly attractive since it was amenable to rapid deconstruction leading to inhibitors with very
high ligand efficiency (LE) and lipophilic ligand efficiency (LLE). These triazolopyrimidines were opti-
mized for potency, selectivity and ADME properties ultimately leading to compound 42. This compound
was highly potent and selective with good bioavailability and advanced into pre-clinical development.
Ó 2012 Elsevier Ltd. All rights reserved.
Received 2 May 2012
Revised 19 June 2012
Accepted 25 June 2012
Available online 6 July 2012
Keywords:
Phosphodiesterase 8B
PDE8B
Diabetes
Ligand efficiency
Lipophilic ligand efficiency
The inhibition of phosphodiesterases (PDEs) has remained an
active area of drug research for over twenty-five years. PDE8B is
a relatively recent addition to the cAMP-specific PDE family, which
also includes PDE 4 and 7.1–3 The enzyme’s presence in pancreatic
b-cells suggested a potential role for PDE8B inhibitors as insulin
secretagogues, perhaps in combination with drugs which act
through the GLP-1 receptor.4,5 Early on it became evident that
expression levels and distribution of PDE8B varied greatly between
species. The strategy implemented to overcome this issue was to
validate the target using a human in vitro system (primary human
pancreatic islets), then rapidly move into the clinic to achieve proof
of mechanism. To achieve the latter goal, a diabetic clamp study
was planned using an iv infusion of drug to better control plasma
levels and return a more informative PK/PD relationship. With this
development path in mind, we targeted an inhibitor with high po-
tency and selectivity, and with solubility sufficient for parenteral
delivery. Moreover, the compound would have good oral bioavail-
ability, to enable the seamless transition to the PO route of admin-
istration once proof of mechanism was achieved. Very little was
known about the function of PDE8A, and although achieving selec-
tivity over this isoform would be desirable, it was not deemed crit-
ical for advancement. Herein we describe the discovery and
optimization of highly potent and selective triazolopyrimidine
PDE8B inhibitors.
As no potent PDE8B inhibitors had been reported, a high
throughput screen was run using baculoviral-expressed PDE8B
protein lysate from Sf9 insect cells at 10 lM concentration. One
of the hits identified was from a triazolopyrimidine series (com-
pound 1, Fig. 1). Although this hit was not very attractive from a
ligand efficiency (LE)6 or lipophilic ligand efficiency (LLE)7,8 per-
spective, further file mining around this structure identified addi-
tional hits including compound 2.
The hit 2 was an ideal starting point with low molecular weight
and cLogD. It also contained three sites of diversity with which to
modulate potency, selectivity and ADME properties. The synthesis
of this class of compounds was straightforward as shown in
Scheme 1. The C7 substituent was the most readily modified, and
it was quickly determined that the preferred substituent at the po-
sition was simply NH2 (9a, Fig. 2).9 Although it removed one of the
three substitution vectors, this result significantly lowered the
⇑
Corresponding author. Tel.: +1 858 404 6664; fax: +1 858 404 6719.
E-mail
addresses:
mike_deninno@sd.vrtx.com,
deninnomp@gmail.com
(M.P. DeNinno).
Present address: Vertex Pharmaceuticals Inc., 11010 Torreyana Rd., San Diego,
CA 92121, USA.
à
Present address: University of New England, 716 Stevens Ave., Portland,
ME 04103, USA.
0960-894X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmcl.2012.06.079

5722
M. P. DeNinno et al. / Bioorg. Med. Chem. Lett. 22 (2012) 5721–5726
Scheme 2. Prepararation of C5 analogs. (a) RCN (solvent), HCl (g) in dioxane,
100 °C, 3 h.
Table 1
C5-analogs
Figure 1. Hit identification.
NH2
N
N
N
N
N
R
O
PDE8B IC50 MW/cLogDa LE/LLEb
HLM CLint
c
Compd #
R
9a
12
13
14
15
16
17
18
CF3
H
Me
14 nM
1400 nM
45 nM
302/2.0
234/0.6
248/0.8
274/2.1
278/0.5
354/0.6
310/2.3
316/2.6
0.52/5.9 ND
0.48/5.3 <7.2
0.57/6.6 ND
0.56/5.9 <7.2
0.49/6.5 <7.2
0.41/6.5 ND
0.41/4.5 17
0.40/4.0 100
Cyclopropyl 10 nM
CH2OMe
CH2SO2iPr
Phenyl
91 nM
84 nM
160 nM
243 nM
Cyclohexyl
a
cLogD was determined using an in-house model built from over 100,000
experimental LogD determinations of Pfizer project compounds.
b
LE = À1.4log(IC50)/#heavy atoms. LLE = Àlog(IC50)–cLogD.
c
HLM, and in future tables, RLM refer to the rate of turnover in human liver
microsomes and rat liver microsomes respectively, calculated from six time points
taken over a one hour incubation period. Units for CLint
determined.
= lL/min/mg. ND = not
Scheme 1. Synthesis of triazolopyrimidines. (a) Diethyl malonate (1 equiv), NaOMe
(3 equiv), MeOH, 0 °C to 25 °C, 18 h; (b) 90% HNO3 (5 equiv), AcOH, 0 °C to 20 °C,
1 h; (c) POCl3 (5 equiv), PhNEt2 (1.5 equiv), 105 °C, 3 h; (d) Fe powder (4 equiv),
AcOH, 50 °C, 3 h; (e) R2NH2 (1.2 equiv), DIEA (2 equiv), NMP, 120 °C, 18 h; (f) NaNO2
(1.2 equiv), AcOH, H2O, CH2Cl2, 20 °C, 1.5 h; (g) R3NH2 (excess), dioxane, 85 °C, 3 h.
In all, about thirty C-5 analogs were prepared, and a brief sum-
mary is given in Table 1.9 Substitution at C5 was tolerated by the
enzyme, but did not provide additional benefit beyond what could
be achieved with small alkyl substituents. On the contrary, the lar-
ger groups increased the MW and hydrophobicity resulting in com-
pounds with higher microsomal/oxidative metabolism. Thus,
future SAR was performed on three triazolopyrimidine templates
containing methyl, cyclopropyl, or trifluoromethyl at the C5
position.
The focus then turned to the remaining substituent at N3.
Starting from the dichloropyrimidines 5, a three step protocol to
the final compounds was developed in parallel format. In this
manner, extensive exploration of this region was rapidly achieved
(Table 2). The SAR was fairly permissive and a wide range of func-
tional groups were tolerated.9 This result was fortuitous since the
ability to modify compound properties now largely resided in this
region of the molecule. A close analog of compound 9a, the chiral
tetrahydro furan 9b, looked very promising from an in vitro per-
spective. It was very potent with good selectivity except for PDE4C
and 4D (ꢀ20-40x) and it was progressed into iv pharmacokinetic
studies in rats. Despite showing no turnover in the rat liver micro-
some (RLM) assay, in vivo clearance was at hepatic blood flow.11
Other analogs, such as 26, which displayed improved iv clearance,
had low oral bioavailability which was thought to be the result of
poor aqueous solubility.11 The poor solubility of these compounds
would also hinder execution of our clinical plan (vide supra) using
parenteral dosing and therefore needed to be addressed.
Figure 2. Lead molecules 9a and 9b.
molecular weight and increased the potency providing a lead mol-
ecule with candidate-like LE and LLE.
Modification at the C5 position was the most synthetically labo-
rious since it was incorporated at the very earliest stages of the
synthesis. To circumvent this issue and to allow for a greater diver-
sity of analogs to be made, a novel one step protocol was developed
from the cyano-triazole 1010 (Scheme 2).

M. P. DeNinno et al. / Bioorg. Med. Chem. Lett. 22 (2012) 5721–5726
5723
Table 2
N3-analogs
NH2
N
N
N
N
N
R
Compd
#
R
PDE8B
IC50
LE/LLE
RLM/HLM
CLint
O
O
0.56/
6.1
19
20
25 nM
10 nM
<28/11.9
93/<7.2
Figure 3. Core modifications.
0.56/
5.9
but only through significant structural modifications to other re-
gions of the molecule which suggested an alternate binding mode.
Another strategy for improving solubility that also has prece-
dent from the PDE5 literature13 is to incorporate an ionizable
group in the molecule. It was already known from the early hits
that a basic amine could be positioned off of the C7 nitrogen. How-
ever, potency suffered with such analogs unless properties (LE and
LLE) were compromised. Instead, it was decided to further develop
0.48/
5.1
21
22
23
150 nM
925 nM
86 nM
35/<7.2
ND/ND
ND/<7.2
O
O
0.42/
4.4
Table 3
Improving solubility
0.50/
3.8
NH2
Cl
N
N
N
H
N
N
N
N
0.52/
6.5
24
6.8 nM
ND/<7.2
R
Compd
#
R
PDE8B
IC50
RLM/HLM
CLint
N
N
0.51/
5.1
29
92 nM
244/52
230/30
25
26
136 nM
ND/<7.2
<14/24.6
N
N
N
N
30a
206 nM
S
N
0.55/
4.7
93 nM
N
S
N
31
23 nM
625/303
ND/150
N
N
N
One strategy to improve solubility involved modifications to the
core. It is not hard to recognize the highly nitrogenous triazolopyr-
imidine ring system as perhaps the main contributor to the poor
solubility of the compounds. It is flat, has both hydrogen bond do-
nors and acceptors, and a high polar surface area. Given the prec-
32b
3620 nM
edent for such core changes in other PDE isoforms (e.g. PDE5)12
,
we were optimistic that replacing one or more of the nitrogens
in the triazole ring would be tolerated. In the event, replacement
of either N1 or N2 with carbon resulted in over a 100 fold loss of
potency ( Fig. 3). Potency could be restored to these templates,
a
Inhibition of PDE4C = 62% @ 1
Inhibition of PDE4C = 60% @ 1
l
l
M; 4D = 52 % @ 1
M; 4D = 64 % @ 1
l
l
M.
M.
b

5724
M. P. DeNinno et al. / Bioorg. Med. Chem. Lett. 22 (2012) 5721–5726
Figure 4. Combining features.
the SAR at the N3 position given that the library work had shown
this region to be tolerant to polar functionality with regards to
maintaining potency. Initial efforts focused on pyrrolidine and
piperidine analogs with aryl or aryl alkyl substitution on the basic
nitrogen (Table 3). Despite having some success maintaining po-
tency, the compounds suffered from poor microsomal stability
and reduced selectivity over PDE4.
Careful analysis of all the N3 SAR highlighted the importance of
the oxygen atom in the tetrahydro pyran and tetrahydro furan
derivatives (analog 20 vs. 21 and 22). Its presence at the proper po-
sition imparted both increased potency and selectivity to the mol-
ecules. A key experiment was designed to combine the key features
of compounds 20 and 32 (Fig. 4). The resulting analog 33 was po-
tent and selective with improved pH-dependent solubility (data
not shown).
The key enantioenriched amine 38 (>99% ee) could be prepared
in three steps including the resolution by dibenzoyl tartaric acid14
and is readily scalable (Scheme 3).
The N-benzyl substituent in 33 imparted poor microsomal sta-
bility, but was quickly optimized with a small library via reductive
Scheme 3. Preparation of chiral amine 38: (a) toluene, 20 °C, 18 h; then KOtBu
(1 equiv), 0 °C, 3 h; (b) LAH (2 equiv), THF, 0 °C, 2 h; (c) (+) dibenzoyl tartaric acid,
EtOH, H2O, 20 °C (two recrystallizations).
Table 4
Top compounds
NH2
N
N
N
N
R1
N
O
N
R2
Compd #
R1
R2
PDE8B IC50
1.6 nM
LE/LLE
RLM/HLM CLint
36/36
Rat CLa/F
121/ND
Dog CLa
8.5
N
39
CF3
0.44/7.1

M. P. DeNinno et al. / Bioorg. Med. Chem. Lett. 22 (2012) 5721–5726
5725
Table 4 (continued)
Compd #
R1
R2
PDE8B IC50
0.9 nM
LE/LLE
RLM/HLM CLint
50/20
Rat CLa/F
72/20%
Dog CLa
N
40
41
c-Propyl
0.49/7.6
ND
7.0
S
N
c-Propyl
23 nM
0.40/5.7
75/16
124/ND
N
N
42
43
Me
Me
1.5 nM
4.2 nM
0.52/8.2
0.45/7.0
<14/<8
22/15
23/63%
24/45%
3.5
5.6
S
N
S
F
44
Me
19 nM
0.4/5.2
80/36
28/12%
4.8
F
a
Units for rat and dog iv clearance: ml/min/kg.
Figure 6. Insulin secretion from human pancreatic islet cells
= 3 mM glucose;
= 8 mM glucose.
inhibitor chosen to move into pre-clinical development was the
one with the highest LLE, compound 42.
The data for 42 is shown in Figure 5. It has high solubility, espe-
cially at lower pH. It has high potency and more than 1000-fold
selective over all targets tested except PDE8A. This lack of selectiv-
ity is not surprising given the high homology of the catalytic site.16
In fact, across five distinct scaffolds explored in this project, includ-
ing our previously published work17, no significant selectivity over
PDE8A was achieved. The compound has good permeability and rat
and dog PK. In primary human pancreatic islets, it was shown to
increase insulin levels in a glucose-dependent manner (Figure 6).
Further development of this compound was halted due to what
is believed to be a mechanism-based increase in blood pressure
and heart rate in a dog telemetry study at concentrations above
10x the free IC50. It is not clear whether the 8A or 8B PDE isoform
is responsible, but it appears unlikely to be a non-PDE mediated ef-
fect given the low plasma concentrations involved and the fact that
the effect could be reproduced in the rat with several structurally
distinct17 analogs.
Figure 5. Data for compound 42.
amination of the debenzylated intermediate. The best replace-
ments were then made with the trifluoromethyl and methyl C5
substituents. From that matrix set of compounds, six were chosen
for further profiling based on activity, properties and structural
diversity (Table 4). The potency and selectivity of the inhibitors,
as well as the LE and LLE scores were stellar. To aid in the selection
process, all six compounds were tested in the full PDE panel as well
as a broad receptor screening panel. Additionally, both rat and dog
PK was obtained and genetic toxicity was assessed in in vitro
micronucleus and Ames assays.15 Perhaps not surprisingly, the

5726
M. P. DeNinno et al. / Bioorg. Med. Chem. Lett. 22 (2012) 5721–5726
3. Soderling, S. H.; Bayuga, S. J.; Beavo, J. A. J. Biol. Chem. 1998, 273, 15553.
In summary, a high throughput screening hit was deconstructed
4. Dov, A.; Abramovitch, E.; Warwar, N.; Nesher, R. Endocrinology 2008, 149, 741.
5. Horvath, A.; Giatzakis, C.; Tsang, K.; Greene, E.; Osorio, P.; Boikos, S.; Libè, R.;
Patronas, Y.; Robinson-White, A.; Remmers, E.; Bertherat, J.; Nesterova, M.;
Stratakis, C. A. Eur. J. Genet. 2008, 16, 1245.
6. Hopkins, A. L.; Groom, C. R.; Alex, A. Drug Discov. Today 2004, 9, 430.
7. Leeson, P. D.; Springthorpe, B. J. Med. Chem. 2004, 47, 6338.
8. Edwards, M. P.; Price, D. A. Annu. Rep. Med. Chem. 2010, 45, 381.
9. The activity of additional examples of N3, C5 and C7 analogs can be found in
the Supplementary data.
10. L’Abbe, G.; Beenaerts, L. Bull. Soc. Chim. Belg. 1989, 98, 421.
11. Rat PK for compound 9b: CL = 61 ml/min/kg; Vdss = 1.7 L/kg; F = 17%; Dog PK
for compound 9b: CL = 22 ml/min/kg; Vdss = 1.4; F = 13%. Rat PK for compound
26: CL = 29 ml/min/kg; Vdss = 5.5 L/kg; F = <5%. Dog PK for compound 26:
CL = 11 ml/min/kg; Vdss = 0.9; F = <5%.
to optimize for ligand efficiency against PDE8B. SAR development
was aimed at improving potency, selectivity and solubility, while
maintaining high ligand and lipophilic-ligand efficiency metrics.
Compound 42 emerged as the top analog in the series and dis-
played very high levels of potency and selectivity with good oral
bioavailability in the rat and dog. The compound moved into pre-
clinical development but was halted due to significant hemody-
namic effects in the dog.
Supplementary data
12. Rotella, D. Nat. Rev. Drug Disc. 2002, 1, 674.
13. Terrett, N. K.; Bell, A. S.; Brown, D.; Ellis, P. Bioorg. Med. Chem. Lett. 1819, 1996,
6.
14. Sakurai, N.; Sano, M.; Hirayama, F.; Kuroda, T.; Uemori, S.; Moriguchi, A.;
Yamamoto, K.; Ikeda, Y.; Kawakita, T. Bioorg. Med. Chem. Lett. 1998, 8, 2185.
15. All genetic toxicology results were negative.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2012.06.
079.
16. Such isoform selectivity has been a major issue in past PDE research, due to a
highly conserved catalytic domain within each PDE family; see: Ke, H.; Wang,
H. Curr. Top. Med. Chem. 2007, 7, 391.
17. DeNinno, M. P.; Wright, S. W.; Visser, M. S.; Etienne, J. B.; Moore, D. E.; Olson, T.
V.; Rocke, B. N.; Andrews, M. P.; Zarbo, C.; Millham, M. L.; Boscoe, B. P.; Boyer,
D. D.; Doran, S. D.; Houseknecht, K. L. Bioorg. Med. Chem. Lett. 2011, 21, 3095.
References and notes
1. Hayashi, M.; Matsushima, K.; Ohashi, H.; Tsunoda, H.; Murase, S.; Kawarada, Y.;
Tanaka, T. Biochem. Biophys. Res. Commun. 1998, 250, 751.
2. Soderling, S. H.; Bayuga, S. J.; Beavo, J. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95,
8991.