Biomedical Engineering Reference
In-Depth Information
cores of the two compounds, one of which gave a better correlation to the
experimental data. The binding site was distal to the binding site of the second
subunit of CBF, Runx1, and the compounds acted as allosteric inhibitors. In
an attempt to step beyond the state of the art, Gonz´ lez-Ruiz and Gohlke have
developed a method that quantitatively exploits amide proton CSPs for
protein-ligand docking.
51
This method combines standard scoring by the
DrugScore function
52
(which describes the protein-ligand interactions) with
scoring the ligand poses with respect to their agreement with experimental CSP
data. The comparison is achieved by back-calculating the CSP based
exclusively on ring currents. After testing the approach on a set of crystal
structures, the authors applied it to three real-world cases for which CSP data
was available. In two of the three cases the docking pose was quite good,
however the third differed substantially. The authors' analysis of the reasons
for the differences highlight the complexities of the approach, which include
the effects of hydrogen bonds between the ligand and protein and CSPs
resulting from conformational and/or dynamic changes that occur upon ligand
binding. In yet another example, both CSPs and intermolecular NOEs were
used to guide the docking of the prodrug Losartan to glycoprotein VI, a target
for anti-thrombotics.
53
Instead of trying to derive direct structural constraints from CSPs, one can
analyse the data in terms of binding models. Auto-FACE (Auto-Fast chemical
Exchange analyser) fits titration data from an e.g.,[
15
N,
1
H] HSQC experiment
to multiple possible models of protein-ligand interaction.
54
The models include
simple single-site binding, multiple-site with or without sequential binding and
allosteric contributions. The authors claim that residues directly involved in
ligand binding can be differentiated from those affected by e.g., conforma-
tional changes, by analysing the 'initial rate of perturbation'. ITC suggested a
four-state binding model for the system being investigated where the last state
was non-specific. The NMR titration data was analysed in terms of the binding
constant and the initial rate and magnitude of the perturbation. This data
suggested two distinct binding sites with different affinities for the compound,
which agreed well with the ITC data (free, low affinity and high affinity
complex). Interestingly, the two sites correspond to two sites predicted by
JSURF
55
using CSPs from low and high concentrations of the ligand. For a
more extensive review of the use of chemical shift information for protein
structure elucidation see the recent comprehensive review by Mulder and
Filatov.
56
Solving the NMR structure of a large protein remains a challenge. Recently,
Schwalbe and colleagues achieved specific resonance assignment of roughly
half of the primary structure of DXR (DOXP reductoisomerase), an 87 kDa
homodimer that is a potential anti-infective target,
57
by using 3D hetero-
nuclear experiments with uniform
15
N,
13
C and
2
H labelling. The authors
measured intermolecular NOEs between the protein and the cofactor NADPH
thus potentially enabling the structure of the complex to be determined using
the known crystal structure. However, no intermolecular NOEs to a bound
