Efficient Experimental Assessments Of And Improvements On The Foldability Of De Novo-designed Proteins

One of the most vexing problems facing de novo protein design is how to efficiently select well-folded proteins. Thorough structure analysis on each of a large number of designed sequences is neither realistic nor warranted. Here we established an experimental approach that can be used to efficiently assess or improve the foldability of designed proteins. In this approach, the structural stability of the protein of interest （POI） has apositive correlation with the antibiotic resistance of bacteria cells, the POI with flanking linker sequences as a guest segment inserted into TEM1-b-lactamase forming a sandwich-like structure. Periplasmic proteases can specifically recognize unfolded proteins and POIs that are not well-folded are prone toproteolysis, leading to weak antibiotic resistance of host cells. This system canhelp us not only to assess the foldability of designed proteins but also to select mutations that can rescue an initially problematic design.The design target, the backbone of the natural protein thioredoxin （Protein Data Bank ID:1r26）, has a five-stranded P-sheet surrounded by four a-helices with 113 amino acid residues. In the initial design, Dv₁r26 shows a low level of antibiotics resistance with the TEM1-β-lactamase-based in vivo system, suggesting a not so well-folded structure, which is consistent with the corresponding 1H-NMR spectra result. Three rounds of directed evolution of Dv₁r26 using the antibiotic-resistance-based screening system were carried out. A number of mutant sequences that exhibited strong antibiotic resistance have been identified. Three point mutations, V31D, V61D and W105R, are observed with high frequencies in these mutants. One of the mutant protein Dv₁r26_M1, has been selected for further analysis. This mutant contains four point mutations including the three high frequency ones plus the mutation L23P. The’H chemical shift spectra as well as the 1H-15N HSQC spectra of Dv₁r26_M1, both spectra showing good dispersion and indicating a well-folded structure. A more thorough NMR structural analysis were conducted on this protein, which indicated that the β-strands are well-formed and consistent with the desired arrangement of the strands within the sheet.However, most of the N-terminal helix is disorderedand the position of the C-terminal helix was not well-defined.Based on analyzing possible problems of sequences designed with the initial computational model, some subtle but important revisions of the model have been introduced. New protein sequenceswere designed using revised model, one of them named E₁r26, which lead to high antibiotics resistant in the TEM1--lactamase-based in vivo system without directed evolution. Subsequently, E₁r26 has been subjected to NMR structural characterization. Result shows that theNMR mode ofE 1r26 is in excellent agreement with the design target 1r26, the root mean square deviation of all backbone heavy atoms is only around 1.7 A between the lowest energy model of E₁r26 and the design target.The TEM1-β-lactamase-based selection system was found to be not free of negative positives, namely, mutants associated with strong in vivo antibiotics resistance but inferior structural properties in subsequent vitro analyses. To identity a complementary system, an GFP-basedinsertion reporter system has been tested for its sensitivity to assess protein foldability. The experimental results show that there is no obvious difference between three target proteins.