Creating Molecular Recognition From Scratch: Stepwise Engineering of Knottin Peptides to Bind Tumor Marker CA IX

Using rational and combinatorial methods, we engineered stable cystine-knot peptides (knottins) with nanomolar affinity to the tumor marker Carbonic Anhydrase IX (CA IX).  Expression of membrane-bound CA IX correlates with poor prognosis in a variety of cancers, including breast, colon, and lung.  There is a critical need for non-invasive molecular imaging agents that target cell surface markers such as CA IX for tumor staging and to inform clinical treatment decisions.   Knottin peptides consist of at least three interwoven disulfide bonds and multiple solvent exposed loops.  Their small size (~4 kDa) allows them to penetrate tumors and clear from the body on a timescale suitable for imaging, two applications in which large antibodies (150 kDa) fall short.  We used a combination of phage display and yeast surface display to engineer CA IX-binding knottin peptides based on a novel Agatoxin knottin scaffold, derived from spider venom, which had no initial binding to this target.   First, a phage-displayed cyclic peptide library was screened against yeast surface-displayed CA IX to identify cyclic peptides with weak (µM) binding affinity.  Second, selected cyclic peptides were substituted into a knottin loop with similar structure, and an adjacent knottin loop was randomized in length and composition, yielding a library of 10⁷ mutants, which were displayed on the surface of yeast.  This yeast-displayed library was sorted by flow cytometry to isolate knottin variants with highest affinity for fluorescently labeled CA IX.  We determined that both mutated knottin loops contributed to the observed CA IX binding. Furthermore, we are able to graft functional binding loops from the engineered Agatoxin knottin peptides into a structurally similar Agouti Related Protein (AgRP) knottin, derived from humans.  We are currently preparing labeled versions of these engineered knottins to evaluate their use as molecular imaging agents in living subjects.  In summary, this work demonstrates that a small knottin scaffold is suitable for engineering high affinity binders from naïve libraries.  In addition, to our knowledge this is the first time that multiple loops of a knottin peptide have been simultaneously engineered against an exogenous target.  This approach also provides a method for transferring weak binding cyclic peptides isolated from phage display libraries to a knottin scaffold to further engineer higher affinity binders.

Funding: NIH NCI R21 CA143498, Center for Biological Imaging at Stanford (CBIS) Seed Grant, NSF and Stanford Graduate Fellowship Programs