Improved Stability and Brightness Following Iterative Redesign of a De Novo Biliprotein

Bilins, linear tetrapyrroles derived from heme catabolism, are ubiquitous across biological systems. The constraints imposed by binding to a protein confer valuable optical properties on bilins, which are used for sensing and harvesting light, and offer potential for fluorescence imaging, optogenetics, and biosensing. Recently, bilins have been used to evaluate methods for designing ligand binding sites, and novel biliproteins (BPs) that bind phycoerythrobilin (PEB) emerged from computational design using RoseTTAFold Diffusion All-Atom (RFDiffusionAA) and LigandMPNN. Here, we find much lower stability, extinction coefficients and fluorescence quantum yields for de novo designs compared to the native CpcA-PEB biliprotein, so we used a LigandMPNN-AlphaFold 3 (AF3)-Rosetta Relax pipeline to redesign one BP, C11, which can be implemented quickly and without in-house GPUs. The top three ligand confidence scores emerging from this procedure yielded C11-578, C11-620, and C11-756 redesigns that retain 50%, 52%, and 57% identities, respectively, with C11. AF3-Rosetta modeling of the BP redesigns predicted similar conformations of the PEB, decreased solvent-accessible surface area of the bilin binding site, an additional predicted hydrogen bond in C11-620, and a more rigid protein backbone. The new constraints on bilin binding improved the stability, absorption properties and fluorescence yields for C11-578, C11-620, and C11-756, and the brightness of the redesigns improved up to 13-fold, approaching the values for the native CpcA-PEB BP. Further design iterations can be used to create a range of absorbing and emitting BPs, including oligomeric assemblies housing internal energy cascades.