Ligand-detected NMR remains one
of the most popular methods for identifying fragment binders. However, the
approach is mostly used qualitatively, especially at the earliest stage of a
project. It is possible to determine dissociation constants, but doing so
normally requires preparing and analyzing a series of samples at different concentrations.
Three years ago we highlighted an approach that could be run in a single NMR tube,
but it required a concentration gradient throughout the tube, starting from high
(millimolar) concentrations of ligand. A new open-access paper in Angew. Chem.
Int. Ed. by Simon RĂ¼disser, Gabriela Stadler, and Alvar Gossert presents a
potentially more general approach.
Chemical shift anisotropy KD
(CSAKD) uses fluorine NMR, a subject we’ve discussed extensively. The “transverse
relaxation rate” R2 of a 19F atom in a ligand can be
measured experimentally and is a function of multiple parameters, some of which
are intrinsic to the small molecule itself and some of which reflect interaction
of the ligand with a protein. One contributor is the chemical shift anisotropy,
or CSA, which varies depending on the electronic environment of the atom. Calculating
the R2,CSA of a ligand and comparing it to the measured R2
can, in theory, allow one to determine the fraction of bound ligand. The more ligand bound at a given concentration, of course, the higher the
affinity.
The theory quickly gets complicated:
the paper sports 11 equations (and 24 more in the supplementary information!)
which include such esoteric terms as the “generalized order parameter,” “spectral
densities,” and the “rhombicity of the diffusion tensor.” Fortunately, the “shape
of the CSA tensor” can be calculated using density functional theory (DFT),
though the calculations are onerous. In the new paper the researchers present a
more efficient approach based on machine learning that speeds up the process by
7000-fold.
So, how well does it work in
practice? The researchers used CSAKD to measure the affinities of one to three
different known fluorine-containing ligands against four proteins, for a total
of eight pairs, all of which had previously been characterized by NMR titration
or isothermal titration calorimetry. The KD values determined from CSAKD
ranged from 38 micromolar to 4.4 millimolar and were in good agreement with published
numbers; several were within the margins of error and none differed by more
than a factor of three. The researchers note that error bars are larger for
more tightly-bound ligands.
Although I can’t claim to fully understand
the theory, the approach does seem practical, and thus a worthy addition
to the NMR toolbox. Pre-calculating the CSA tensors for a set of
fluorine-containing ligands in a library would enable one not just to identify hits
from an initial screening campaign, but to calculate their affinities from the outset.
This will be the topic of my next post.
No comments:
Post a Comment