For many years it has been almost dogma in the scientific community that in protein structures the planar peptide bond occurs predominantly in the trans conformation1. The occasional occurrence of a peptide bond in cis conformation was, in most cases, noted as a curiosity of the respective structure. This is remarkable since it became clear almost 20 years ago that the cis/trans-isomerization of peptide bonds on the N-terminal side of proline plays an important role in the folding process of a protein2. Systematic studies of peptide bond conformations have been hampered by the limited amount of structural information available3, and have so far mainly focused on proline residues4. With more three-dimensional structures of proteins at hand today, the notion is slowly emerging that cis peptide bonds are by no means a curiosity, and that they may even be important determinants for the function of proteins.

We have analyzed a non-redundant set of 571 proteins from the Brookhaven protein data base5, the structures of which have been determined crystallographically to a resolution of 3.5 Å or better. Within this set, only one in ∼360 peptide bonds is reported to be in the cis conformation (Table 1). Most of these instances (>90%) occur where the peptide bond is an imide (Xaa–Pro) rather than an amide bond (Xaa–nonPro).

Table 1 Occurrence and frequency of reported peptide bonds in the Brookhaven Protein Data Base (ref. 5)1
Full size table

An obvious discrepancy exists, however, between the fraction of cis peptide bonds observed and what can be predicted from free enthalpy values. Free enthalpy differences between the cis and trans conformations have been reported to lie anywhere between 2.0 kJ mol–1 for Xaa-Pro bonds and 10.0 kJ mol–1 for Xaa–nonPro bonds3. Based on these values and assuming thermodynamic equilibrium at 293 K, about 30% of all Xaa–Pro bonds in acyclic peptides should occur in the cis conformation, and so should about 1.5% of all Xaa–nonPro bonds. These numbers are larger — by a factor of 6 for Xaa-Pro and by a factor of ∼50 for Xaa–nonPro — than values determined from analyzing protein structures (Table 1).

It is noteworthy that there is a significant correlation between the resolution of the structure solved and the number of cis peptides detected. High resolution structures (<2.0 Å) contain almost twice the number of Xaa–Pro bonds than medium and low resolution structures (≥2.5 Å) and almost four times the number of Xaa–nonPro bonds. This striking resolution dependence as well as the above mentioned discrepancy between the fraction of cis bonds observed and expected lead us to suspect that many cis peptide bonds have not been recognized as such in the determination and refinement of the structures, especially at resolutions ≥ Å. Moreover, most of the refinement programs, which have been widely used in recent years (such as X-PLOR6), only allow for the possibility of a cis conformation of an Xaa–Pro bond but will, unless specified explicitly, force any other peptide bond into the trans conformation. Huber and Steigemann realized this as a potential problem as early as 19747.

In proteins in which non-proline cis peptide bonds have been unequivocally identified, they often occur at or near functionally important sites and are very likely involved in the function of the molecules. One example is coagulation factor XIII8, in which an Arg–Tyr cis peptide bond has been found near the active site, and a Gln–Phe cis peptide bond at the dimerization interface of the molecule. This strongly suggests a functional role for them, as has been proposed for the cis peptide bonds in carboxypeptidase A, dihydrofolate reductase and most recently in the intein gyrA9,10.

In conclusion, we would like to emphasize two points: first, the importance of the cis conformation of peptide bonds in protein structures, especially if it is a Xaa–nonPro peptide bond, and second, the possibility that many cis peptide bonds may have passed unnoticed due to the limited resolution of the data and to the refinement protocol used.