Cochran Lab
Cochran Lab

Cystine Knot Peptides (Knottins)

Please see our recent review articles for more information:
Moore and Cochran, (2012) Engineering Knottins as Novel Binding Agents, Methods in Enzymology, 503, 223-251
Moore et al., Knottins: Disulfide-bonded Therapeutic and Diagnostic Peptides, Drug Discovery Today, In press.

Engineering Cystine Knot Peptides (Knottins) with Novel Molecular Recognition Properties

Our research group has a strong interest in developing peptide-based alternatives to monoclonal antibodies for tumor-targeting applications. Towards this goal, we engineered cystine knot (knottin) peptides for high affinity molecular recognition against tumor-associated receptors, and established them as a new class of molecular imaging agents in living animals. The knottin family of peptides contains a disulfide-bonded core that confers outstanding proteolytic resistance and thermal stability. Knottins, which naturally function as protease inhibitors, antimicrobials, and toxins, are composed of several loops that possess diverse sequences amongst family members.

We first used rational and combinatorial methods to engineer knottin peptides that bind with high affinities to integrin receptors, which are overexpressed in a variety of human cancers (Figure 1). We engineered the C-terminal knottin domain of the Agouti-related protein (AgRP), a human neuropeptide, to bind with sub-nM affinity and high specificity to αvβ3 integrin (Silverman, et al., 2009 J. Mol. Biol. 385, 1064-75). In addition, we engineered the Ecballium elaterium trypsin inhibitor II (EETI-II), a knottin found in squash seeds, to bind with single-digit nM affinity to αvβ3/αvβ5 or αvβ3/αvβ5/α5β1 integrins (Kimura et al., 2009 Proteins 77, 359-69). Until now, the development of a single peptide that can bind all three of these integrins with high affinity has not been achieved. This broad specificity has important implications since many tumors co-express these integrins, and thus current drugs have low efficacy due to biological redundancies. Notably, through this work we elucidated which amino acid residues dictate binding specificity to closely-related integrin receptors, and propose molecular rationale for our findings, which is important from the standpoint of molecular recognition and drug design.

Figure 3Fig. 1. Examples of knottin peptides our group has engineered to possess novel molecular recognition properties. The structures of native EETI-II, from squash seeds, or a truncated form of AgRP, a human regulatory peptide, are shown. Using rational and combinatorial methods, the purple loop in EETI-II or the red loop in AgRP were engineered to confer high affinity binding to tumor-related integrin receptors with varying specificities as shown.

We showed that these engineered knottins specifically target integrin-expressing tumors in vitro and in vivo, inhibit tumor cell adhesion to extracellular matrix proteins, and are highly stable to serum proteases, all of which bode well for in vivo applications. Importantly, we showed that the small size of knottins (~3-5 kDa) and their high stability translated into desirable pharmacokinetic and biodistribution properties for molecular imaging applications, namely high tumor uptake and rapid clearance from non-target tissues. (Please see: “Engineering Proteins as Molecular Imaging Agents” for more details).

In addition, we engineered AgRP-based knottins with low nanomolar affinity and exquisite specificity towards the platelet-specific αiibβ3 integrin receptor and showed that these peptides are potent inhibitors of platelet aggregation, with therapeutic implications in thrombosis (Silverman et al., 2011 J. Molecular Recognit. 24, 127-135). This work also defined amino acid residues which impart selectivity to αiibβ3 versus αvβ3 integrin.

Interrogation and Prediction of Tolerated Sequence Diversity in Knottin Protein Folds

In addition to engineering cystine-knot peptides for novel molecular recognition properties, we developed a method to test the tolerance of the knottin fold to sequence modification (Figure 2) (Lahti, et al., 2009 PLoS Computational Biol. 5, e1000499). By combining directed evolution with computational covariance analysis, we elucidated guidelines for introducing modifications (both in amino acid sequence and loop length) into the loop regions of the cystine-knot scaffold. Using these guidelines, we then predicted amino acid sequences that when substituted into a loop region facilitated proper folding of the knottin. Remarkably, all twenty-five predicted protein sequences retained the cystine knot fold, while twenty-five randomly generated sequences did not. This study defined a set of “rules” that can be used to guide future knottin engineering efforts, and highlights the potential of combining directed evolution with bioinformatics for rational protein engineering.

Figure 3Fig. 2. Schematic for interrogating the tolerance of sequence diversity in knottin loops. (A) Six libraries of loop-substituted knottin variants were designed based on the wild-type sequence of EETI. Libraries were created by replacing cysteine-flanked loop 2 (green) or loop 3 (blue) sequences with peptides of randomized amino acids (X) and varying lengths (n). The trypsin binding loop (orange) was not replaced, but instead used as a handle to evaluate the proper folding of EETI loop-substituted clones. Disulfide bonds are shown in yellow. (B) The binding interaction between trypsin (light grey) and EETI (PDB 2eti and 1h9h) is mediated through the trypsin binding loop, and is dependent on the correct formation of all three disulfide bonds. This interaction was exploited for high-throughput isolation of properly folded EETI loop-substituted variants. From Lahti et al. PLoS Computational Biology 2009 doi:10.1371/journal.pcbi.1000499.g001