In vivo screening identifies a highly folded beta-hairpin peptide with a structured extension.

The conventional approach for developing folded peptides involves chemical synthesis of systematically modified peptide variants and individual characterization by circular dichroism and NMR spectroscopy. The use of synthetic peptides and reliance on low-throughput characterization techniques necessarily restricts the sequence diversity that can be explored, though efforts have been made to overcome this limitation. We have recently described an approach that enables us to asses the ability of peptides to fold into b-hairpins in the cytoplasm of live cells. Our strategy entails the recombinant expression of a peptide gene fused in frame between flanking genes encoding a cyan fluorescent protein (CFP) and a yellow fluorescent protein (YFP). If a particular peptide sequence adopts a folded structure, CFP and YFP are brought into closer proximity and exhibit a higher efficiency of fluorescence resonance energy transfer (FRET). Higher FRET efficiency enhances the YFP (acceptor) fluorescence at the expense of the CFP (donor) fluorescence. Imaging of plates harboring colonies of transformed bacteria provides the YFP and CFP fluorescence emission intensities for each colony, and thus peptides that are highly folded in vivo can be distinguished from those that are not. As will be described in this manuscript, this FRET-based approach also provides a versatile method for screening large libraries of peptide sequences for highly structured variants. We have applied this strategy to the development of a version of a “tryptophan zipper” (trpzip)-type b-hairpin with a structured extension. Trpzips are a class of highly folded b-hairpins that are defined by the presence of cross-strand diagonally oriented Trp/Trp pairs on one face of the hairpin. The minimal and highly stable trpzip structure has emerged as a preferred model system for computational and experimental studies of protein folding. We have proposed that trpzip-type peptides (or tandem fusions of such peptides) could serve as a minimal protein scaffold for molecular recognition in the cytoplasm of live cells. With this goal in mind, we sought to employ our screening strategy to identify a candidate trpziptype peptide with high fold stability in vivo. A similar approach has previously been used to increase the thermal stability of engineered immunoglobulin VL domains. [15] The template for our initial “extended trpzip” library was a 20-mer version of the highly folded 16-mer trpzip HP5W4. The 20-mer contained two additional pairs of residues, one random and one threonine, genetically inserted after what would otherwise have been the second and 14th residues of HP5W4. The sequence of the resulting “peptide” portion of the 400-member protein library was KKXTWTWNPATG KWTWTXQE ; here X represents all 20 amino acids, and the residues inserted relative to HP5W4 are underlined. Assuming b-strand conformation, the randomized positions would be directed towards the face of HP5W4 that harbors the interdigitated Trp side chains. Escherichia coli was transformed with the gene library, and ~6F10 colonies on 10 Petri dishes were subject to fluorescence imaging in order to identify those exhibiting the highest ratio of YFP-to-CFP fluorescence emission. We have previously demonstrated the ability of this imaging system to reliably distinguish colonies that express FRET constructs of various FRET efficiencies. The fluorescent brightness of each colony, considered commensurate with peptide solubility, was determined by direct excitation and imaging of YFP. Five individual colonies that exhibited both a high ratio of YFP-to-CFP fluorescence emission and high brightness were cultured overnight, and the plasmid DNA was purified. DNA sequencing revealed a striking consensus at the randomized positions: in all five sequences, position 3 was Trp and position 18 was either Lys or Arg (Table 1). This consensus sequence was used as the new