(626f) Leveraging Disulfide Bonds to Stabilize Small Protein Scaffolds during Extensive Diversification

Cysteines play a unique role in natural and engineered proteins. Its high reactivity and susceptibility to oxidation gives cysteine the ability to drive molecular interactions as well as provide structural support. In protein design, cysteines are often avoided because of potential difficulties during production in bacteria as well as to enable site-specific conjugation via genetic introduction of a sole cysteine. While being mindful of these benefits afforded by cysteine-free proteins, cysteines also elicit stabilizing features upon formation of disulfides. This trade-off becomes especially important during protein library design. Heavily diversifying a protein in pursuit of novel function or improved biophysical properties can result in the accumulation of numerous destabilizing mutations within individual library variants. However, appropriate placement of cysteines can compensate for such destabilization by forming covalent disulfide bonds. In this study, we examined the impact of disulfides in the context of multiple small protein scaffolds with diverse secondary structures. We constructed combinatorial libraries within three protein scaffolds – fibronectin beta sandwich, affibody three-helix bundle, and Gp2 alpha/beta topology – and sorted them for specific, high affinity binders to a multitude of targets. Deep sequencing datasets corresponding to these high affinity binders were probed for cysteine content. Variants isolated from combinatorial protein libraries based on each of the three scaffolds gave rise to a varying prevalence of cysteine. Importantly, libraries composed of overly-diversified positions (i.e. first generation designs that failed to exclude detrimental amino acids at library positions) produced a strong increase (>7%) in cysteine at one-third of diversified sites. These less refined library designs demonstrated dramatic up-regulation, from 20% to 45%, in clones containing either two or four cysteines while showing a 15% decrease in the prevalence of clones containing either one or three cysteines. Characterization of cysteine-rich variants revealed disulfides contributing up to 21°C in thermal stability. Conversely, second generation library designs that acted on site-specific information to guide amino acid bias at each position, reducing the average Shannon entropy of library sites from 0.97 to 0.74, actually showed a down-regulation in cysteine content among high affinity variants. This depletion of cysteines was accompanied by a 15°C increase in thermal stability. Our results provide evidence for both the importance of site-wise amino acid bias as well as the potential for disulfides to compensate for overly-mutated, destabilized proteins.