(6ec) Constructing, Screening, and Evolving Therapeutic Proteins | AIChE

(6ec) Constructing, Screening, and Evolving Therapeutic Proteins

Authors 

Van Deventer, J. A. - Presenter, Massachusetts Institute of Technology
Kelly, R. L., Massachusetts Institute of Technology
Le, D. N., Massachusetts Institute of Technology
Zhao, J., Massachusetts Institute of Technology
Rajan, S., University of Toronto
Sidhu, S., University of Toronto
Wittrup, K. D., Massachusetts Institute of Technology

Protein-based therapeutics are beginning to come of age, but the development of proteins tailored for specific diagnostic and therapeutic applications remains an ongoing challenge. Yeast surface display (YSD) and other display technologies provide powerful approaches to the high-throughput isolation and evaluation of proteins with characteristics suitable for therapeutic settings. However, these techniques require judicious library construction, reformatting of candidate binding proteins prior to most biological characterizations, and are normally limited to screening with the chemical diversity encoded in the 20 canonical amino acids. The work presented here describes the combination of YSD with synthetic antibody libraries and noncanonical amino acids (ncAAs). These efforts have yielded 1) an extremely efficient approach to isolation of binding proteins in conventional YSD format; 2) a switchable display/secretion technology for rapid isolation and immediate production and characterization of candidate binding proteins in soluble form; and 3) a platform for the construction and evaluation of bioconjugates on the surface of yeast.

The discovery of new antibodies remains an important aspect of therapeutic discovery. Yeast surface display (YSD) and other in vitro technologies have emerged as powerful alternatives to conventional immunization and hybridoma technologies, but these in vitro approaches require the construction of antibody libraries with amino acid diversities and loop length variations that enable the efficient discovery of high affinity antibodies. Synthetic antibody libraries described by Sidhu and coworkers have previously been shown to yield high affinity binding proteins when employed in a phage display format. We utilized one of these libraries to construct a billion-member synthetic antibody library in a YSD format in order to take advantage of the quantitative screening capabilities afforded by YSD. Initial screening with model antigens indicated that the Sidhu library yielded full-length binding proteins more efficiently than a YSD nonimmune antibody library previously constructed in the Wittrup laboratory. We have used the synthetic antibody library to isolate a suite of antibodies targeting cancer-associated fibroblasts (CAFs), an understudied component of many solid tumors. We screened the synthetic antibody library for proteins recognizing both the human and murine forms of fibroblast activation protein (FAP), a membrane-bound protease selectively overexpressed on the surface of CAFs. Our screening campaign yielded more than 50 unique antibodies recognizing FAP, with close to 30 exhibiting crossreactivity for both the human and murine form of the protein. Secondary screens for binding enabled efficient identification of candidates recognizing multiple epitopes of FAP, and further development has resulted in a suite of full-length murine antibodies. Pharmacokinetic experiments in mice have confirmed that these antibodies are suitable for in vivo work, and further evaluation of the antibodies as agents for the therapeutic targeting of CAFs is underway. The general antibody isolation and characterization approaches used in this work are accessible to many single-investigator laboratories and should be applicable to the development of antibodies against numerous therapeutic targets.

While the combination of conventional YSD and high-quality synthetic antibody libraries is quite powerful, conventional display technologies are limited by the requirement of converting displayed proteins into soluble form prior to characterization at the end of a screening campaign. Toward this end, we have developed a switchable display/secretion platform that enables the rapid production of soluble reagents immediately following screening efforts without the use of any cloning steps. In our implementation, we use a ncAA-mediated amber suppression event to control the display of an antibody-like scFv-Fc on the yeast surface. Flow cytometry results have confirmed that the presence of the ncAA in the induction media enables display of full-length scFv-Fcs on the yeast surface. This system has the necessary characteristics for screening and soluble reagent production: in model enrichment experiments, a single round of bead-based selections enables approximately 500-fold enrichment of clones binding the target antigen, while yields of secreted scFv-Fcs are consistently upwards of one milligram per liter. Encouraged by these initial results, we have constructed synthetic antibody libraries in the switchable format and screened them against epidermal growth factor receptor (EGFR), a validated cancer antigen. ScFv-Fcs secreted from enriched populations of the library specifically recognize mammalian cells expressing EGFR, confirming both the success of our screening and the recognition of EGFR in its native form. These results suggest that this switchable screening format will enable rapid characterization of additional biological effects of candidate binding proteins (e.g., receptor clustering or downregulation, cellular proliferation, cellular differentiation, or apoptosis) in the future. The implementation of the switchable display/secretion technology enables cloning-free approaches to the generation of specific binding reagents and functional evaluation of these proteins for cancer targeting and other applications.

Although established display technologies have yielded a startling array of engineered proteins with therapeutic properties, the limited chemical repertoire normally found within proteins can impede efforts to tailor proteins for use in specific therapeutic applications. For example, conventional display technologies are rarely suitable for the characterization or engineering of antibody-drug conjugates or other bioconjugates. To overcome this limitation, we have adapted the switchable system described above to enable the incorporation of ncAAs with reactive functional groups into proteins displayed on the yeast surface. We first used the display/secretion system to confirm that several ncAAs, including ncAAs containing ketone or azide groups, can suppress the amber codon in the switchable system in cells expressing an aminoacyl-tRNA synthetase of appropriate specificity; this provides indirect proof of the display of proteins containing reactive side chains on the surface of yeast. Flow cytometry experiments further indicate that cells displaying proteins containing these functional groups can be selectively modified using oxime ligations (ketones), copper-catalyzed azide-alkyne cycloadditions (azides), or strain-promoted cycloadditions (azides) as determined by detection of reaction products on the surface of yeast. Thus, this approach enables both efficient display of proteins containing ncAAs on the surface of yeast and facile, site-specific protein modification, giving this system all of the necessary components for constructing bioconjugates on the surface of yeast. We anticipate that this approach will offer unprecedented throughput for constructing, evaluating, and evolving antibody-drug conjugates and other small molecule-protein conjugates, opening up new possibilities for the development of therapeutics that simultaneously exploit the favorable properties of both proteins and small molecules.