Chemistry Reference
In-Depth Information
limited applicability to typical pharmaceutical target proteins and the high demands on
resources. Most therapeutically important target proteins have molecular weights above
30 kDa, and obtaining sequential resonance assignments for such proteins is a major task
in terms of both resources and time, despite recent advances in data acquisition schemes.
An indirect and more limited approach would be to assign only resonances of active site
residues by the use of known substrates or ligands. There are, in any case, substantial
demands on the protein production. Milligram amounts of isotopically labeled target pro-
tein (e.g.
13
C,
15
N,
2
H) are to be produced in a suitable expression host. The purified protein
must be soluble, stable for at least a couple of weeks at room temperature and remain
monodisperse at the high concentrations needed. Even if it were possible to obtain the
sequence-specific resonance assignments for a target protein of 30-40 kDa, the overlap in
the 2D spectrum would be considerable due to both severe line-broadening effects and the
large number of peaks in the spectrum. Consequently, it would be very difficult to reliably
re-assign the peaks that have moved upon ligand binding. Therefore, efforts have been
dedicated to develop protein-detected methods that are more generally applicable to larger
proteins.
A passable way would be to decrease the number of peaks in the 2D spectrum. One
approach is to record 2D
1
H-
13
C correlation spectra on target proteins with selective
13
C
labeling of the methyl groups of valine, leucine and isoleucine.
[
108
]
In this approach, the
sensitivity is increased almost threefold for small proteins (MW
<
20 kDa) compared with
2D
1
H-
15
N correlation spectra on uniformly
15
N-labeled targets. For larger proteins, the
13
C
labeling should be combined with
2
H labeling in order to decrease dipole-dipole relaxation.
In another labeling strategy, termed site-selective screening,
[
109, 110
]
it is possible to screen
selectively for binding to a selected epitope without the need for sequence-specific assign-
ments. First, pairs of sequential amino acid residues that reside in the binding site of interest
(e.g. the active site) and are unique, i.e. they only appear once in the protein sequence, are
identified. Such a unique amino acid pair, XY, is then selectively labeled so that amino acid
X is labeled with
13
C and amino acid Y with
15
N. Performing an HNCO-type correlation
spectrum of a protein labeled in this way will result in only one signal. Thus, chemical
shift perturbations upon addition of a ligand that binds in the vicinity to the labeled amino
acid pair are easily detected even for large proteins due to the reduced spectral complexity.
However, this simplicity of the resulting spectrum is both the strength and weakness of
this method. With only one probe for the binding site, it is not possible to separate direct
binding in the vicinity of the amino acid pair from small shifts due to indirect effects from
locations other than the binding site. To minimize this risk, several unique amino acid pairs
should be identified and selectively labeled.
4.5.3 Choice of NMR Technique for Primary Fragment Screening
From the survey of the NMR techniques above, the general recommendation would be to
use a ligand-detected technique for the primary fragment screen.Aprotein-detected method
could be considered in cases where the isotopically labeled target protein is readily produced
in large quantities and with known 3D structure. Preferably also the sequential resonance
assignments of the target protein should be known or it should at least be clear from inspec-
tion of the 2D
1
H-
15
N correlation spectrum that the assignments will be straightforward
to obtain. Assuming that these conditions are fulfilled, one case where a protein-detected
