4902
N. C. Ray et al. / Bioorg. Med. Chem. Lett. 17 (2007) 4901–4905
the FlexS program.5 Owing to the computational time
required by FlexS, it was necessary to create a focused
database of compounds for searching using the proper-
ties of the known antagonists as a guide. A number of
filtering and clustering steps were used to reduce a col-
lection of ca. 718,000 commercially available screening
compounds to a set of 862, which comprised the FlexS
database for the searches.
displaying an activity of 33 and 300 nM in the binding
and the functional GR assays, respectively. In selectivity
screens, 2 showed 100-fold selectivity over the mineralo-
corticoid receptor (MR) and 21-fold selectivity over the
progesterone receptor (PR).
A final round of follow-up searching was carried out
around 2 and 27 additional compounds were obtained
for screening. This led to the identification of 3
(Fig. 2), the most potent compound found by virtual
screening, with a Ki value of 16 nM in the GR binding
assay and 135 nM in the GR functional assay.
Low-energy reference conformers of the compounds in
Figure 1 were identified using conformational searches
carried out in Macromodel6 (Low Mode, MMFFs
forcefield, GB/SA solvent model). FlexS was then used
in its flexible superposition mode to overlay the data-
base compounds on each of the reference structures.
The highest scoring alignments from each of the
searches were visually assessed and promising com-
pounds selected for further consideration.
Synthetic chemistry efforts commenced with examination
of the 6-substituent using the route outlined in Scheme 1.
Reaction of the appropriate benzyl malonate with a suit-
ably substituted urea gave the pyrimidinetrione in mod-
erate (R1 = H) or good (R1 = Me) yields. Chlorination
under forcing conditions gave the 6-chloropyrimidined-
iones in approximately 50% yields. These were function-
alized by treatment with a range of amines, optionally in
a microwave synthesizer.
From the four searches, 123 compounds of interest were
identified. Following the removal of duplicates and
any compounds containing undesirable features, 91
remained. Clustering (0.85 Tanimoto similarity thresh-
old, Daylight fingerprints7) to remove redundant com-
pounds provided a set of 78 compounds. This list was
refined further by more detailed examination and consid-
eration of supplier reliability to give a final list of 18 com-
pounds from four suppliers. These compounds were
obtained and screened in the GR binding assay described
below, providing 1 (Fig. 2) with a Ki value of 4.5 lM.
Data for a range of compounds thus prepared are shown
in Table 1. All initial investigations were made around
the 5-unsubstituted benzyl scaffold, since 4 was shown
to have a similar level of potency to 3. A large lipophilic
group such as phenyl or benzyl at the 4-piperidyl position
was preferred, since a 4-methyl piperidine analogue was
shown to be inactive (13). The nature of attachment of
the phenyl ring to the piperidyl group was also impor-
tant: ring fusion (8), or chains longer than a methylene
spacer (14, 15) resulted in attenuated potencies. Replace-
ment of the methylene with an oxygen (6) or a carbonyl
(16) also reduced potency, indicating that the receptor
tolerates only lipophilic moieties at this position. Re-
moval of the methylene spacer completely to give 5
yielded an equipotent compound. A 3-methyl group
was not required as illustrated by 11 and 12, and methyl-
ation of the N-1 position was shown to be deleterious (7).
Affinity for GR was determined by ligand binding mea-
suring displacement of [3H]dexamethasone from recom-
binant baculovirus derived human GR.8 Functional
activity at human GR was determined in SW1353/
MMTV-5 cells transfected with a plasmid encoding fire-
fly luciferase located behind a glucocorticoid response
element (GRE).8 GR antagonist activity was measured
as inhibition of dexamethasone induced luciferase
expression. Selected compounds were tested for GR
agonist activity in the same functional screen by
performing the assay in the absence of dexametha-
sone—no compounds tested demonstrated any agonism.
Selectivity over the estrogen (ERa), androgen (AR),
mineralocorticoid (MR), and progesterone (PR) recep-
tors was determined by ligand binding assays.9
Exploration of the effects of modifying the 5-benzyl sub-
stitution pattern showed that once again (Table 2), more
polar groups were not tolerated, with 2- and 3-substitu-
tion favored over 4-substitution. It is interesting to note
2-D substructure and similarity searches were carried
out around 1 using the full 718,000-compound database
and a further 11 compounds were selected for purchase.
The screening of these led to the discovery of 2 (Fig. 2),
O
OEt
R
R
R1
O
(a)
(c)
N
N
O
N
O
O
O
OEt
CF3
R3
R
(b)
O
O
O
R
R
R1
O
R1
O
N
N
N
O
O
N
H
N
H
N
N
NR2
N
N
Cl
2
R3
R3
OH
(1)
(2) R = 2-Br
(3) R = 3-Cl
Scheme 1. Reagents and conditions: (a) R3HNC(@O)NHR1, NaOEt,
85 °C, 18 h; (b) POCl3, 87% H3PO4, 100 °C, 3h; (c) HNR2
MeOCH2CH2OH, 140 °C, 48 h.
,
2
Figure 2. Initial hit from virtual screening.
Ray, Nicholas C.
