I. T. Raheem et al. / Bioorg. Med. Chem. Lett. 26 (2016) 126–132
127
As schizophrenic patients display dysfunction in these signaling
pathways, abnormal striatal output has been implicated in the
pathophysiology of the disease. It is hypothesized that inhibition
of PDE10A will increase the intracellular levels of cGMP and cAMP,
thereby increasing striatal output and restoring behavioral
inhibition.7
Since the seminal disclosures of PQ-10, TP-10, and MP-10
beginning in 2006,3,9 at least 20 pharmaceutical companies have
initiated discovery programs targeting PDE10A inhibitors, includ-
ing Merck,3b,c with at least seven compounds entering human
clinical trials.10 Preclinically, structurally diverse PDE10A
inhibitors demonstrate efficacy in assays predictive of antipsy-
chotic activity.3b,c These include the psychostimulant-induced rat
hyperlocomotion assay (LMA), the conditioned avoidance
response assay (CAR), and the pre-pulse inhibition assay (PPI).8
Furthermore, PDE10A inhibitors display efficacy in rodent models
of negative symptoms8a as well as both rodent and non-rodent
models of cognition.8c,d These inhibitors also display fewer
preclinical adverse events compared to typical and atypical
antipsychotic treatments.
A summary of our efforts to optimize the selectivity pocket lin-
ker is shown in Table 1. Aside from potency against PDE10A, com-
pounds were generally assessed for their metabolic clearance via
in vitro and/or in vivo studies. We began our efforts by replacing
the quinoline ring with a less lipophilic pyridine (entry 1). While
this structural change provided a significant improvement in com-
pound physicochemical properties, it imparted a >100-fold loss in
potency (PDE10A Ki = 1.1 nM). Transposition of the oxygen atom to
the C3 position of the propyl chain as well as altering the linker
atom identity to carbon or nitrogen (entries 2–4) provided similar
compound potency (PDE10A Ki = 0.4–4.1 nM). We next sought to
functionalize selected positions in the aliphatic chain with the goal
of blocking possible sites of metabolism. Unfortunately, methyl
branching as well as installation of unsaturation provided no ben-
efit with respect to compound potency (entries 5–7). However, we
realized significant improvements by incorporation of a cyclo-
propane constraint (entries 8 and 9). While the potency of the
cis-cyclopropane propane linker was compromised relative to
compounds with more flexible linkers (PDE10A Ki = 15.5 nM), the
corresponding trans-cyclopropane isomer provided a dramatic
improvement in compound potency (PDE10A Ki = 0.07 nM). Inter-
estingly, the all-carbon trans-cyclopropane linker was markedly
less potent (entry 10, PDE10A Ki = 1.7 nM). Resolution of the race-
mic O-linked trans-cyclopropane compound revealed that nearly
all potency resided with the S,S isomer (entries 11 and 12).
With a highly optimized selectivity pocket linker in hand, we
sought to interrogate the identity of the selectivity pocket binder
(western arene group, Table 2). Both published9,3c and internal
crystallographic evidence suggested that this group needed to
maintain a strong hydrogen bonding interaction with Tyr693 of
the selectivity pocket in the PDE10A catalytic active site. We inves-
tigated a number of 2-linked aza-arenes and learned that not only
was the selectivity pocket tolerant of significant structural varia-
tion at this position, but potencies could be further improved by
appending or fusing lipophilic functionality. Indeed, functionalized
pyridines, imidazole, and fused bicyclic arenes provided measur-
able improvements in compound potency (entries 1–6). Not
surprisingly, modifications that presumably destabilized the requi-
site hydrogen-bonding interaction to Tyr693 were detrimental to
potency. Such changes included transposition of the H-bonding
nitrogen atom (entries 7 and 8) and modification of the nitrogen
pKa (entries 9 and 10). Finally, we interrogated a number basic
amine functional groups, and in all cases found potency to be com-
promised, likely due to either a poor H-bonding vector or a subop-
timal pKa imparted by this group (entries 11 and 12). Despite
realizing slight improvements in compound potency with some
of these structural alterations, we chose to move forward with
the original pyridine-containing cyclopropane linker.
X-ray crystallography has greatly enabled some of these early
drug discovery efforts, providing clarity around the binding
requirements within the enzyme active site, and allowing medici-
nal chemists to employ rational approaches toward compound
design. Of particular importance has been the identification of a
PDE10A selectivity pocket, first reported by scientists at Pfizer.9,3c
This is a region of the enzyme active site into which highly potent
and selective inhibitors bind, forming
a strong H-bonding
interaction with Tyr693, a residue unique to PDE10A. In the
present manuscript, we present the discovery of Pyp-1, a novel,
potent and highly selective pyrazolopyrimidine inhibitor of
PDE10A.
Our laboratories recently disclosed a successful fragment-based
drug discovery (FBDD) effort beginning with 2-chloro-3-
methylpyrimidine fragment 1, identified from a screen of the
Merck Fragment Library (Fig. 1).11 This initial hit had a PDE10A
Ki = 8.7 l
M and very high ligand binding efficiency (LBE)12 of
0.57. Rational optimization of the pyrimidine fragment focused
on two critical areas of chloropyrimidine core, which was enabled
by both efficient X-ray crystallographic support and parallel library
synthesis, leading initially to the identification of compound 2.
Installation of an eastern aminomethyl dimethylthiazole and a
western quinoline propoxy group, which engaged the selectivity
pocket, led to the identification of lead compound 3, a highly
potent and selective PDE10A inhibitor. Unfortunately, despite
promising potency, selectivity, and in vivo efficacy, this lead com-
pound suffered from a poor preclinical pharmacokinetic (PK) pro-
file (high clearance, low oral bioavailability), very low aqueous
solubility, high plasma protein binding, a number of problematic
off-target liabilities, including competitive binding with MK-499
and PXR activation, and inhibition of several cytochrome P450
enzymes.
Given the overall acceptable profile of compound 2, we attribu-
ted the poor physicochemical properties and off-target activities of
3 to the pendant propoxy quinoline group. Further, we carried out
an in vitro metabolite identification study to better understand
potential liabilities contributing to its poor PK. As summarized in
Figure 2, major metabolites identified after incubation with rat
hepatocytes included oxidation/glucuronidation of the quinoline,
O-dealkylation of the alkoxy chain, and oxidation of the
dimethylthiazole ring. As we knew that a selectivity pocket lin-
ker/binder was required for high PDE10A potency and selectivity,
we turned our attention to optimization of the quinoline propoxy
group, with the goal of replacing the lipophilic quinoline ring (bin-
der) as well as stabilizing the required propoxy chain (linker)
toward metabolic transformation.
Holding the selectivity pocket linker and binder constant as the
trans cyclopropyl pyridine, we carried out a second round of SAR
on the eastern aminomethyl arene group utilizing parallel synthe-
sis and relying on the SAR generated in previous optimization
efforts to guide monomer selection.11 These efforts culminated in
the identification compound 4 bearing a less lipophilic N-methyl
pyrazole group. While this change resulted in a 20-fold loss in
potency, 4 represented the best overall compound profile to date
in the 2-chloro-3-methylpyrimidine series (Fig. 3). Compound 4
is a sub-nanomolar PDE10A inhibitor with excellent selectivity
over other PDEs, and has good aqueous solubility with a signifi-
cantly improved rat PK profile (moderate clearance and high oral
bioavailability) over earlier leads. Unfortunately, a number of off-
target liabilities remained with this advanced lead, particularly
PXR activation and reversible inhibition of CYP2C9 and CYP3A4
enzymes. A retrospective analysis of >1000 compounds prepared
in the 2-chloro-3-methylpyrimidine class revealed that these two
specific off-targets were common to nearly every compound