Structural and Energetic Analysis of Disulfide Bonds in the Protein Data Bank

Disulfide bonds between cysteine residues play a key role in determining structure and regulating function of biological molecules. These bonds generally stabilize native protein structures along with other secondary and tertiary structures. For example, a disulfide formed between two cysteine residues separated in sequence space can establish their proximity in three-dimensional space; thus, a disulfide bond provides biological systems with a way to branch covalent interactions that tether different regions of peptide backbone. However, unlike rotations about the single bonds of a polypeptide chain(φ,ψ), there is a significant barrier to rotation about a disulfide bond. The energetic cost of rotating about the disulfide bond restrains available conformations further and provides a potential source of energy for allosteric transformations. For instance, if a protein fold rotates a disulfide bond up hill in energy that energy can be released via reduction of the disulfide bond. Strained disulfides have been observed and characterized in the protein data bank for X-ray and NMR structures. In this study we seek to elucidate the structural information associated with disulfide bonds, strained or unstrained, and quantify their effects on protein structure. To this end, we constructed a database of cysteine information extracted from protein data bank files containing disulfide bonds. We analyzed 6238 X-ray structures (with resolution better than 3.0 ̊A) and 1452 NMR structures, each with sequence identity no greater than 90%. For high resolution X-ray structures (resolution less than 1.7 ̊A), we identified shifts in the Ramachandran plots due disulfide strain; these shifts are muted for low resolution and NMR structures. Further analysis of the internal coordinates of disulfide bonds revealed key differences between high resolution X-ray structures and NMR structures. NMR and low resolution X-ray structures exhibited significantly higher strain across these bonds, compared to high/ultra-high resolution X-ray structures. Thus, many highly strained bonds present in these structures can be attributed to the characteristics of their experimental determination, and our analysis reveal evidence of modeling artifacts, especially in NMR structures. The results of this study can have a direct impact on increasing the accuracy of models constructed from structural biology experiments.