Chemistry Reference
In-Depth Information
docking to generate protein-ligand co-structures. Since it is not practical to use
NMR to screen the large library of compounds typically utilized by HTS or virtual
screening, a more focused approach with a smaller compound library is employed.
Fragment-based screening utilizes a significantly smaller library consisting of
simple, low molecular-weight (
<
250-350 Da) molecules [
15
,
20
-
22
]. These frag-
ment-like molecules typically have weaker binding affinities (millimolar range)
compared to hits found in high-throughput screens (micromolar range), but NMR is
sensitive enough to detect these weak protein-ligand interactions. Importantly,
fragment-based libraries are more efficient in covering chemical space. Simply,
the number of possible compounds decreases drastically as the number of atoms is
reduced. Thus, a smaller chemical library actually covers a larger percentage of
chemical space. An even greater structural diversity can be achieved by chemically
linking multiple fragments. This also results in an additive improvement in
binding affinity. Evolving a drug from smaller fragments in this manner has the
added benefit of improving ligand efficiency, which typically results in a more
bioavailable compound that minimizes non-specific and unfavorable interactions
[
172
,
173
].
A recent study [
174
] by Barelier and colleagues utilized fragment-based screen-
ing by NMR and molecular docking in the investigation of the human peroxi-
redoxin 5 (PRDX5) ligands. Peroxiredoxins are important enzymes that catalyze
the reduction of hydroperoxides through a conserved cysteine. However, very few
ligands have been identified that bind these proteins despite the availability of
crystal structures for PRDX5 bound with benzoate (PDB ID: 1HD2, 1H40) [
175
].
A compound library of 200 fragment compounds was screened by NMR using STD
and WaterLOGSY experiments, where six fragments were identified as binders.
STD experiments were also used to calculate the binding affinities for the six
fragment molecules, which were in the 1-5 mM range. Since the 1D experiments
did not provide information about the location of the binding site, AutoDock 4 [
40
]
was used to dock the fragments to the PRDX5 protein structure. The docking was
done against the entire protein structure; a grid search focusing on the
benzoate ligand binding site was not used. Not surprisingly, ambiguous results
were obtained. The molecular fragments bound to several locations on the PRDX5
structure that were indistinguishable based on binding energies.
Of necessity, the NMR backbone assignments for PRDX5 were obtained to
enable the identification of the ligand binding site by monitoring CSPs in 2D
1
H-
15
N HSQC experiments. All the fragments were shown to generate a similar
set of CSPs consistent with a binding site that included the proposed catalytically
active cysteine. The docked binding conformation was also further confirmed from
CSPs for derivatives of these fragments. Analysis of the PRDX5 structure with the
docked fragments identified the presence of a potentially important hydroxyl
functional group that was pointed towards the catalytic cysteine (Fig.
9c
). Interest-
ingly, the benzoate compound found in the PRDX5 crystal structure did not show
binding by NMR. But, derivatives of benzoate that included a hydroxyl functional
group showed improved affinity, further indicating the importance of this hydroxyl
group in ligand binding to PRDX5. These results provide further validation of the
