1318
J. Evenäs et al. / Bioorg. Med. Chem. Lett. 24 (2014) 1315–1321
The initial SAR around 1 is summarized in Table 1. Removal of
A similar binding mode, with overlap between the pyrimidone
and Febuxostat, was also proposed by the Rao group.10a
the aminoacetyl group from 1 led to 6 showing a 20-fold drop in
potency. Only methyl substitution was accepted on the exocyclic
amino group of 6 (7, pIC50 = 5.1), whereas other residues had no ef-
fect (see 8), or were detrimental (compounds 9 and 10). Methyla-
tion of N-3 yielded inactive 11, pointing to a critical role for that
group. Introduction of a substituent in the 5-position, such as
chloro (13) or bromo (14), or a nitrile (15) led to a more than 10-
fold gain in potency. Interestingly, a fluorine was not effective
(12, pIC50 = 4.9). A methyl group was slightly worse than hydrogen,
whereas the introduction of a phenyl group had an effect similar to
halogen. Overall, the only modification of 6, which led to a signif-
icant increase in potency and LLE was achieved by the introduction
of Cl and CN in the 5-position.
Based on these initial results, we docked compound 13 into a
homology model of hXO constructed from the X-ray crystal struc-
ture of bXO in complex with Pyraxostat and Febuxostat (Fig. 4).21
According to this model, 13 blocks access to the catalytic center
similar to Febuxostat, and their ring systems overlap significantly.
Compound 13 establishes several hydrogen bond and hydrophobic
interactions. The amino group in the 2-position of the pyrimidone
ring donates H-bonds to one of the Mo-Pt-oxido(ꢁ1) oxygens and
Glu802. There appears to be room for a substituent no larger than
methyl (R1 in Table 1), explaining the drop in activity for com-
pounds 9 and 10. These substituents would clash with E1261.
The N-3 and the carbonyl (or hydroxyl enol) mimics the carboxyl
group of Febuxostat and maintains H-bonding interactions with
Arg880 and Thr1010 in similar manner. It is also stacked between
Phe914 and Phe1009 (not shown). Methylation of N-3 would be
expected to be detrimental for these interactions, as was observed
for compound 11. Substituents in the 5-position point towards a
sub-pocket, which is formed by the interdomain loop (id-loop)
containing Thr1010 and Val1011. It is large enough to accommo-
date halogens and groups of similar size such as methyl or cyano.
Electronegative substituents in the 5-position lower the pKa of the
hydrogen-bond donors on the pyrimidone (Table 1) and this is ex-
pected to strengthen the hydrogen bond network and thus increase
potency.22 This could account for the different effect of a chloro or
cyano versus a methyl group on potency. Fluorine would not fill
the sub-pocket, maybe explaining why despite the observed lower-
ing of the pKa of the hydrogen bond donors, a gain in potency was
not observed. To accommodate the larger phenyl residue would re-
quire movement of the id-loop (see Supplementary material).
The overlay of 13 with Febuxostat suggested applying the sub-
stitution pattern of the phenyl ring to the pyrimidone series, hop-
ing it would pick up the interaction between the 3-cyano group
and Asn768 and the hydrophobic contacts of the 4-isobutoxy tail.
Compound 18, boosting the potency of 6 by nearly 3 orders of mag-
nitude, confirmed this assumption (Table 2). In the NMR assay 18
showed reversible and competitive binding versus Febuxostat. Fur-
ther SAR on the aromatic ring revealed that ortho-substitution,
which would disturb the nearly coplanar arrangement of the phe-
nyl and the pyrimidone ring, was not tolerated (data not shown).
Substitution in the meta-position with a cyano, trifluoromethyl or
a trifluoromethoxy group enhanced potency and also LLE signifi-
cantly (compounds 19–21). A simple methyl group (22) was not
significantly better than 6, but the combination with the 3-cyano
substituent further enhanced potency and LLE with respect to 19
(23, pIC50 = 7.0, LLE = 5.5). The combination of two electron-with-
drawing groups such CF3 and CN (24, pIC50 = 6.1) did not provide
any advantage. In the para-position a variety of substituents such
as esters, acetamides or reversed acetamides were tolerated or
led to some potency gain. Eventually, we settled on the ether moi-
ety. Relatively simple residues such as benzyl, isobutyl or neopen-
tyl led to a more than 10-fold boost in potency, but not to an
increase in LLE. As a consequence of the higher lipophilicity, com-
pounds 25–27 showed rather low solubility. Addition of a solubi-
lizing group such as an alcohol (28) or a tetrahydrofuryl moiety
(29) not only increased solubility, but also maintained potency
with an increase in LLE with respect to 26. Most of the compounds
showed good stability in human liver microsome preparations,
with the exception of 26 and 29, which had higher intrinsic
clearance.
The final round of SAR combined the most ligand-efficient res-
idues identified previously (Table 3). Addition of the nitrile or
chloro-substituent in the 5-position of 18 led to a more than 10-
fold increase in potency. Nitrile 30 inhibited hXO with subnanom-
olar potency, whereas the chloro analogue 31 was about 3-fold less
potent. Different from the simpler analogue 7, N-methylation was
not tolerated and resulted in a 100-fold loss in potency (32,
pIC50 = 6.3). Due to the overall good profile with respect to solubil-
ity and metabolic stability, 30 was dosed to Han-Wistar rats in
preparation for the hyperuricemic rat model.24 Low plasma clear-
ance and long half-life were observed after intravenous dosing,
but oral bioavailability was low. Testing in the Caco2 assay re-
vealed low permeability and high efflux for 30, whereas the chloro
analogue 31 showed improved permeability and a lower efflux
ratio.25 Alcohol 33 showed better solubility than 31, with only a
2-fold loss in potency, but permeability and efflux were not im-
proved. To obtain an optimal balance of the desired properties,
changes to the aryl substitution pattern were investigated in the
cyano and the chloro series. Compounds 34 and 35 were found
to be potent, soluble and stable, but demonstrated the same low
permeability and high efflux, whereas compound 36 was rather
poorly active. An exception was 37, which showed improved per-
meability and efflux ratio with respect to 31. An analysis of the
SAR obtained so far indicated that 5-unsubstituted and 5-chloro
substituted pyrimidones were superior to 5-cyano compounds
with respect to permeability, and that the polar surface area
(PSA) should ideally be below 100 Å2 (Table 3). Our efforts in the
5-chloro subseries confirmed this, where compounds such as 38
with a PSA of 107 Å2 had low permeability, but analogs 39–41
demonstrated enhanced permeability, although significant efflux
was still observed. Based on their overall profile, compounds 40
and 41 were dosed to rats and displayed improved oral bioavail-
ability compared to 31. As Table 3 shows, combination of the most
ligand-efficient residues resulted on the whole in a further gain in
Figure 4. Febuxostat (black carbons) and compound 13 (magenta carbons) docked
into a homology model of human XO (Prime/Maestro, Schrodinger Inc.). Protein
residues are displayed in green when they are different from bXO, otherwise in
grey. The loop connecting the 2-layer sandwich domains is denoted ‘id-Loop’.
Further details on model building, coordinates and an additional view can be found
in the Supplementary material.