Protein engineering with natural and unnatural amino acids

In an attempt to rationally design a site-specific antibody protease, a conserved metal binding motif from the metalloprotease thermolysin was appended to the N-terminus of the alpha-CD40L antibody 5c8. In a collaborative effort with members of the Hellinga group at Duke University, the alpha-helical HEXXH zinc binding motif was computationally engineered such that upon folding it should have had the ability to bind a zinc ion. According to the design, after formation of the antibody-antigen complex, the zinc ion would be placed in proximity to a main chain carbonyl in CD40L which would then be activated for proteolysis by a zinc bound water molecule. Unfortunately, initial experiments suggested that the engineered protein lacked proteolytic ability, and potential reasons for this result were subsequently examined. A series of metal binding studies were carried out on the engineered constructs and appeared to indicate that the engineered constructs did not have metal binding ability. In a separate project, an unnatural amino acid containing a 7-hydroxycoumarin moiety was incorporated into the combining site of the 5c8 antibody, and was observed to be a direct sensor of the 5c8-CD40L interaction event. Complex formation was signaled by a 2-3 fold increase in fluorescence signal, and this change in fluorescence was used to determine a solution phase Kd of the antibody-antigen interaction. Control experiments suggested that the effect was specific to the CD40L antigen. In a final project, an attempt was made to generate a metallo beta-lactamase variant whose catalytic ability was dependent on the presence of an unnatural amino acid. Four libraries were constructed around the active site of the zinc dependent metallo beta-lactamase BCII from Bacillius cereus. The libraries were selected on beta-lactam antibiotics in the presence and absence of the metal binding unnatural amino acids (8-hydroxyquinolin-3-yl)alanine and (2,2'-bipyridin-5-yl)alanine. Initial selections did not identify any unnatural amino acid dependent enzyme variants. In collaboration with members of the Baker research group at the University of Washington, the libraries were computationally redesigned. Selection and analysis of these libraries constitutes an ongoing effort between our laboratories.