Design of Cyclic Peptides - Peptide and Protein Design for Biopharmaceutical Applications - page 151

and conformations in a (cyclic) peptide. Depending on the techniques
used for structure determination, the resulting conformational models
often differ. In particular, the structures in crystal and solution may be
different [106]. Since the solution structure is usually closer to the recep-
tor-bound conformation than the X-ray structure, we clearly prefer the
application of liquid-state NMR for structure elucidation. The methods
for yielding NMR structures of high quality are well advanced and have
recently been reviewed in the literature [107].
4.3.4.2 Reliability of structural models
Sometimes 3D structures in solution are presented on the basis of insuffi-
cient data material. For example, small chemical-shift changes of certain
NH protons in temperature gradient measurements are interpreted as
evidence for the presence of a b-turn. However, such a result only indi-
cates shielding from solvent and not necessarily an internal hydrogen
bond, which is found in said b-turns. For a precise investigation, a careful
analysis by special NMR experiments is essential, which can directly
identify hydrogen bridges via J-coupling from 13 CO to the 15 N involved
in a hydrogen bond [108]. As a drawback, this requires expensive isotopic
labelling of the cyclic peptide.
In addition, there are many other sources of error. NMR structures of
cyclic peptides are often calculated via methods originally developed for
protein structure determination. Since spin-diffusion processes have
only minimal influence on the NOESY/ROESY cross-peak volumes of
small- and medium-sized cyclic peptides, distance information should
be used as distinct values (upper and lower bounds) and not divided into
distance classes. In addition, it is strongly recommended that NMR
coupling constants are included for the structure determination,
because they provide important information (e.g. dynamics) not avail-
able from NOE cross-peak volumes. Furthermore, NMR peptide struc-
tures derived fromDG calculations are often refined byMD simulations
without incorporation of explicit solvent molecules, thus resulting in
strongly distorted backbone conformations. This is especially true
for cyclic peptides because of the distinct and before-mentioned volume:
surface ratio. Finally, 3D models of cyclic peptides determined by
X-ray crystallography often display artificial backbone arrangements,
since forces resulting from crystal packing usually dominate the
conformation.