Combinatorial and Experimental Design

Finally, a real challenge for the design of new proteins by a combinatorial approach is to start from fully random proteins without introducing any constraint except their length. The sequence space that corresponds to fully random proteins of 80 amino acids in length for example, contains an astronomical number of sequences (2080). Thus, the fraction of this sequence space that one can explore with the most powerful method of selection is obviously very small (for in vitro techniques about 1013 clones at best). Given that, a question immediately arises: is there a substantial chance of finding a folded or even a functional protein in such combinatorial libraries?

To answer this question, Cho and colleagues [76] constructed a random library of about 1013 clones in which every position had an equal probability of encoding each amino acid. Arbitrarily, the length of proteins was fixed to 80 amino acid residues and the targeted function was ATP binding. This library was used for selection by mRNA display using immobilized ATP. After several rounds of selection, binders were isolated and showed no sequence homology with known ATP binding motifs. Their dissociation constants for ATP were in the range of 100 nM to 10 mM. Competition with ATP analogues showed that the recognition was specific for some parts of the ATP molecule. This binding was EDTA concentration dependent, and could be restored after dialysis in the presence of Zn2+. One family of binders indeed contained a CXXC motif characteristic of zinc binding domains. Furthermore, a core of only 45 amino acid residues was determined to be sufficient for ATP binding. Finally the monitoring of selection led the authors to conclude that the ATP binding property probably occurs at a frequency of 10~n.

Selected clones were only soluble when fused to the very soluble maltose binding protein. To improve their solubility, a pool of previously selected binders was submitted to rounds of selection in the presence of increasing concentrations of a denaturant over successive rounds. The population was thus selected for the ability to bind ATP in presence of 3 M guanidinium hydrochloride. One clone was shown to have a defined folded structure according to circular dichroism, trypto-phan fluorescence and NMR studies [77].

Can this strategy lead to new folds? A structural study by X-ray crystallography of one previously isolated clone was undertaken [78]. The three dimensional structure obtained confirmed that the in vitro evolved protein binds a Zn2+ ion (confirmed by X-ray fluorescence) via a CXXC motif. Although the protein has no apparent sequence homolog, the binding site of ADP/ATP showed features typical of adenine binding proteins and revealed a novel a/b fold (Fig. 9.9). This

Fig. 9.9 Structure of a de novo designed ATP/ADP binding protein selected from a fully random protein library (PDB ID: 1UW1 [78]). Binder = blue; ADP molecule = yellow; Zn2+ = grey.

demonstrates that a protein with a tailored function and with the properties of natural proteins can be designed from unconstrained protein libraries.

0 0

Post a comment