232
J. M. Kriegl et al. / Bioorg. Med. Chem. Lett. 25 (2015) 229–235
complex with maraviroc.22 Human CCR3 and CCR5 share
a
sequence identity of more than 50%, thus making CCR5 a well sui-
ted template for the generation of a human CCR3 homology model.
The homology modeling procedure is based on the sequence align-
ment shown in Figure S1 (cf. Supporting material). The compara-
tive modeling step was done by Modeller23 following standard
homology modeling procedures. The resulting homology model
was protonated with the Protonate3D procedure as implemented
in MOE.24 In a next step an energy optimization of the receptor
with high tethers on the heavy atoms was conducted, with a sub-
sequent reduction of the tethers. In a final step the protein back-
bone was kept fixed, and the side chains were energy minimized.
All minimizations were done with the MMFF94x force field as
implemented in MOE. Placement of the ligand (compound 27) into
the receptor was done manually by generating an ensemble of low
energy conformations of the ligand, followed by manual docking of
the various conformations into the receptor. Thereby the ionic
interaction between the positively charged center of the ligand
and Glu287 on transmembrane helix (TM) 7 was enforced. Confor-
mations which caused major clashes were discarded. This ionic
interaction is reported to be formed for most positively charged
chemokine receptor ligands and the corresponding glutamate in
TM 7.25 As a first plausibility check, the shape overlay with marav-
iroc in hCCR5 was used. As a second check we requested the forma-
tion of a salt bridge between the acidic part of the compound and
the positively charged His97 on TM 2, which we considered as the
most likely interaction partner for a negatively charged moiety in
this part of the TM binding region. Finally, the most plausible bind-
ing pose underwent stepwise geometry optimization (as previ-
TM4
TM1
Glu 287
TM7
A
His 97
ously described) and was followed by
a 60 ns molecular
dynamics simulation of the resulting complex in a DMPC lipid
bilayer employing the TIP3P water model with the Amber99sb
force field using GROMACS 4.5.26 The ligand did not undergo major
rearrangements, suggesting stability of the modeled binding mode.
Figure 3 shows the modeled binding mode of compound 27 in
the homology model of human CCR3, overlaid with the crystal
structure of human CCR5 in complex with maraviroc22 (Fig. 3 A).
The modeled binding mode displays the second salt bridge which
is formed between the carboxylate of compound 27 and the pro-
tonated His97. A sequence alignment (cf. Supplementary material)
indicates that rat and mouse CCR3 carry an asparagine at this posi-
tion, which is a much less favorable interaction partner for the neg-
atively charged carboxylate of 27 than the histidine in the human
receptor orthologue. This is reflected in the SAR around the aryl
group at position R4. The unsubstituted benzene ring (compound
22) resulted in a loss in potency on all three species orthologues.
Introduction of an acidic group (compounds 24, 25, and 27)
regained potency on human CCR3, but not on mouse or rat CCR3.
The presence of a salt bridge with His97 in the human orthologue,
which cannot be formed in rat and mouse CCR3, can explain this
observation. Substitution of the carboxylate by a neutral group
such as the corresponding ethyl ester should in general reduce
the potency shift between human and rodent receptors. This was
indeed observed in a consistent manner for several matched pairs
of carboxylates and ethyl esters, as summarized in Table 4. The
only exception was found when the carboxylate or the correspond-
ing ethyl ester is linked via a methyl spacer to the indole ring (com-
pounds 23 and 28). However, the reduced species selectivity for
the aryl ethyl ester analogues is accompanied by a reduced overall
potency on human CCR3. In the context of the CCR3 homology
model this can be attributed to the missing ionic interaction with
histidine His97.
B
Figure 3. (A) Superimposition of CCR5 in complex with maraviroc (grey; PDB entry
4MBS) and the homology model of CCR3 with compound 27 modeled into the TM
binding pocket (magenta). (B) Surface representation of the CCR3 binding pocket.
Key residues are shown as sticks.
it appears very unlikely that polar and/or large substituents at this
position would lead to very potent compounds, as also seen in the
SAR data (Table 2).
On the opposite end of the molecule, that is, at the R1 position,
the prediction of the correct binding mode is challenging. Various
placements of the flexible fluorophenylsulfanyl-propyl tail in the
receptor model are possible, and it is difficult to generate clear
evidence for the biologically relevant binding pose. However, some
observations provide support for our hypothesis. With the piperi-
dine-4-yl-1H-indole part in our proposed binding pose being fixed
between TMs 1, 2, 3 and 7, the fluoro-phenylsulfanyl-propyl tail
protrudes deeply into the predominantly hydrophobic transmem-
brane region of TMs 3, 6 and 7 and overlaps with the benzene ring
of maraviroc when overlaid with the hCCR5—maraviroc complex
(see Fig. 3A). The model suggests that there is a hydrophobic sub
pocket which offers only little space for moieties larger than fluo-
rophenyl and where the introduction of polar substituents might
be less favorable. This is indeed reflected in the SAR trends
observed around R1, in particular for mouse and rat CCR3 (cf.
Table 1). An increase in size and polarity resulted in compounds
that are less active than the reference molecule 3. The picture is
less consistent for human CCR3, where this trend is more obvious
for nonpolar substituents. Elimination of the fluorophenylsulfanyl-
propyl tail resulted in inactive compounds, suggesting that the
hydrophobic interactions formed by this part of the molecule con-
tribute significantly to the overall free energy of binding. However,
the model as such is not accurate enough to explain the beneficial
The homology model can be further used to interpret SAR at
positions R1 and R3 of the piperidine-4-yl-1H-indole series. It sug-
gests that the R3 position points towards a small, lipophilic sub
pocket between TM 7 and TM 1. From the surrounding residues