Structural Characterization of a Computationally Designed De Novo Protein Library by NMR

Abstract

The design of new proteins is a powerful way to gain understanding about the world of natural proteins, while simultaneously expanding that world to include new sequences, structures, and functions. Most design processes reported to date focus on designing a single sequence to meet a given set of criteria and, though successful, are slow and computationally expensive. This study builds on previous work toward developing a method to combine the power of modern computational tools with the sampling and efficiency of combinatorial library design. Two libraries of 4 helix bundle proteins have been synthesized using a simple binary patterning scheme based on the polarity of amino acid side chains. In previous work, this library was been screened for well folded proteins and characterized at a basic level [3]. Here, we use NMR spectroscopy to characterize the structures of several proteins from these libraries in greater detail. Based on the first stage of 1D proton NMR, roughly half of the proteins isolated from the library appear to be well-folded relative to similar natural proteins. In the second stage, one sequence, R4HB_01, was studied in greater detail using uniform \(^{15}\)N isotopic labeling. Data from HSQC, TOCSY, and NOESY experiments reveal a highly stable and well-folded structure, and were able to allow us to make some preliminary determinations of connectivity. To fully test the success of this design methodology, the final step is to solve a complete solution structure of a library protein with geometric constraints. Using uniform 13C and 15N double isotopic labeling with the L4HB_02_Ala sequence, the structure of a well folded 4 helix bundle was predicted based on backbone and C_ chemical shift values. This structure is a highly favored energetic minimum according to the CS-Rosetta folding algorithm, and has an RMSD of 1.54 Åcompared to the structure determined by ab initio methods. This design technique has successfully produced a variety of stable, well-folded, and completely de novo 4 helix bundles, and we look forward to exploring the ways in which it can be expanded to new types of structures.