Computational Analysis of Threonine Ladders on Distinct β-Solenoid Scaffolds, with Implications for the Design of Novel Antifreeze Proteins

Cold-adapted organisms frequently express antifreeze proteins (AFPs) that facilitate their survival at low temperatures, with some especially potent insect AFPs exhibiting β-solenoid structures with ice-binding threonine ladders. β-solenoids exist in nature in numerous forms and emerging protein design technologies may afford opportunities to diversify them further, suggesting the possibility of developing a variety of new AFPs by installing a threonine ladder on non-AFP natural or designed β-solenoids. However, early attempts at such engineering, combined with differences observed between AFPs and structurally similar ice-nucleating proteins, have raised a critical question: Will a threonine ladder show essentially the same behavior regardless of the β-solenoid scaffold that hosts it, or does the specific solenoid scaffold significantly affect a threonine ladder's structural characteristics (and thus potentially alter its suitability for ice binding)? We set out to address this question by creating distinct variants of a simplified model β-solenoid for <i>in silico</i> analysis <i>via</i> structure prediction and molecular dynamics simulations. Our findings indicate that local structural details such as the distance between the hydroxyl groups of adjacent threonines in a TXT motif can vary depending on the β-solenoid scaffold. In the most extreme example among our model solenoids, we observed in simulations that differences in only inward-facing residues of the scaffold were sufficient to influence the presence of ordered channel waters between the threonines, a noted feature of natural ice-binding threonine surfaces such as that of TmAFP. While additional studies will be necessary to expand on how such distinctions affect activity, these results emphasize that the impact of a particular β-solenoid scaffold on the local geometry of a threonine ladder may be a pertinent consideration in future efforts to design novel hyperactive AFPs to support applications ranging from biomedical cryopreservation to food science. We conclude our present investigation with a preliminary exploration of how this and other considerations manifest in a proposed workflow for generating predicted AFP-like β-solenoids using AlphaFold and ProteinMPNN.