Screening of structurally diverse bicyclic peptide libraries by phage display

Bicyclic peptides are an attractive molecule format for the development of therapeutics and research tools. Our laboratory is specialized on the development of bicyclic peptide ligands by phage display. In brief, libraries of randompeptides each containing three cysteines are displayed on the surface of a phage and cyclized with a trivalent thiol-reactive chemical reagent. Ligands of a target of interest are isolated by affinity selection and their sequence identified by DNA sequencing. The first goal of my thesis was to test if more diverse bicyclic peptide libraries can be generated if linear peptides on phage are cyclized in parallel with multiple different chemical cyclization linkers. Towards this end, we cyclized peptide libraries of the format XCXnCXnCX (X = random amino acid, C = cysteine, n = 3 or 4) with three different chemicals and panned the library against the protease urokinase-type plasminogen activator (uPA). Interestingly, different peptide sequences were enriched when the phage peptide library was cyclized with the different chemical linkers, indicating that a larger diversity of bicyclic peptides is generated. Screening the larger library generated binders with a higher binding affinity. In the second project I carried out, I isolated bicyclic peptide ligands to a protein of therapeutic interest, b-catenin. b-catenin is the main translational co-activator of theWnt cell signaling pathway. Pathological upregulation of b-catenin is associated with the majority of colorectal cancer. Blocking b-catenin activity is thus an attractive therapeutic approach. However, so far the development of b-catenin inhibitors has been challenging due to the flat and featureless surface of the protein, lacking binding pockets for synthetic ligands. Screening a phage display library comprising >12 billion bicyclic peptides yielded binders for different protein epitopes. The binding site of four peptides was identified to correspond to that of ICAT (inhibitor of b-catenin and Tcf), which is a prime target site on b-catenin for therapeutic intervention and to which no synthetic ligands could be isolated so far. In the third project, we isolated and characterized bicyclic peptides binding to globular actin (G-actin). G-actin constitutes the fundamental building block of microfilaments (also named F-actin) in eukaryotic cells. Beyond this structural function, G-actin has a role in gene transcription and chromatin remodeling. While specific F-actin probes are available for in vitro applications, G-actin probes are lacking. To fulfil this demand, we screened a library of bicyclic peptides for G-actin binders and isolated ligands with dissociation constants in the low nanomolar range. The isolated peptides were found to bind to the same surface region as thymosin b4 and a range or marine toxins. The bicyclic peptide G-actin ligands allowed measurement of the binding affinity of natural G-actin ligands in fluorescence anisotropy competition assays.

New methods for developing (bi)cyclic peptides by phage display

Camille Villequey

Cyclic peptide ligands are a promising molecular format for the development of therapeutics. They combine some of the advantages of large protein therapeutics (high affinity and specificity) and of small molecule drugs (accessibility to chemical synthesis and low immunogenicity). With phage display, large combinatorial libraries of more than one billion mono- and polycyclic peptides can be developed and screened against virtually any target. During phage affinity selections, the physical link between phenotype (encoded peptide) and genotype (phage DNA) allows the identification of enriched peptide binders by sequencing of the phage vector DNA. Based on these concepts, our laboratory has established a robust chemistry in which phage displayed peptides containing two or three cysteines are cyclized or bi-cyclized with thiol-reactive linkers. This procedure has been generally applied for the isolation of potent and selective (bi)cyclic peptide ligands with therapeutic potential. The isolation of binders with optimal properties is highly dependent on the possibility to generate (bi)cyclic peptide libraries with high structural diversity and on the ability to thoroughly decode the phage selection output. In my PhD work, I developed new methods to generate large and structurally diverse libraries and to decode more efficiently phage selected peptides. In my first project, I explore the cyclization of peptides with two chemical bridges as a method to provide rapid access to phage-encoded libraries with high scaffold diversity. To this end, a range of twelve structurally diverse chemicals were tested for thiol-reactivity and phage compatibility. Subsequently, peptide libraries of the format XCXnCXnCX (X = random amino acid, C = cysteine, n = 3 or 4) were cyclized with six of these chemical linkers and panned against the protease plasma kallikrein. The selection yielded inhibitors with remarkable affinity (sub-nM KI), target selectivity (>1000-fold) and proteolytic stability (t1/2 in plasma > 3 days) despite a relatively small molecular mass (~ 1200 Da). In the second project, I developed phage-encoded libraries of small cyclic peptides in order to identify ligands that have the potential to be orally absorbed. A major challenge for the generation of such library is the usually low diversity of compounds due to the limited number of amino acids. In this work, we addressed this limitation by taking advantage of a recently discovered peptide cyclization reaction in which the thiol of a cysteine is ligated with the N-terminal amino group using a chemical reagent. We applied this strategies to two phage-encoded peptide libraries of four and five amino acids (XXXC and XXXXC) that were cyclized with eight chemical linkers yielding a collection of more than 1.3 million compounds. Panning the libraries against the two disease targets plasma kallikrein and coagulation factor XIa yielded ligands with single-digit micromolar affinity. In a third project, I developed a strategy to bypass the bacterial infection step in phage display. Bacterial infection is usually crucial for the propagation and identification of enriched clones by DNA sequencing. However, mutations or chemical modifications of the phage coat proteins, sometimes lead to decreased infection rates and therefore compromise the entire procedure. Consequently, there is an interest for developing methods in which this step could be completely omitted.

EPFL2017

Phage Selection of Cyclic Peptides for Application in Research and Drug Development

Christian Heinis, Xudong Kong

Cyclic peptides can bind to protein targets with high affinities and selectivities, which makes them an attractive modality for the development of research reagents and therapeutics. Additional properties, including low inherent toxicity, efficient chemical synthesis, and facile modification with labels or immobilization reagents, increase their attractiveness. Cyclic peptide ligands against a wide range of protein targets have been isolated from natural sources such as bacteria, fungi, plants, and animals. Many of them are currently used as research tools, and several have found application as therapeutics, such as the peptide hormones oxytocin and vasopressin and the antibiotics vancomycin and daptomycin, proving the utility of cyclic peptides in research and medicine. With the advent of phage display and other in vitro evolution techniques, it has become possible to generate cyclic peptide binders to diverse protein targets for which no natural peptides have been discovered. A highly robust and widely applied approach is based on the cyclization of peptides displayed on phage via a disulfide bridge. Disulfide-cyclized peptide ligands to more than a hundred different proteins have been reported in the literature. Technology advances achieved over the last three decades, including methods for generating larger phage display libraries, improved phage panning protocols, new cyclic peptide formats, and high-throughput sequencing, have enabled the generation of cyclic peptides with ever better binding affinities to more challenging targets. A relatively new cyclic peptide format developed using phage display involves bicyclic peptides. These molecules consist of two macrocyclic peptide rings cyclized through a chemical linker. Compared to monocyclic peptides of comparable molecular mass, bicyclic peptides are more constrained in their conformation. As a result, they can bind to their targets with a higher affinity and are more resistant to proteolytic degradation. Phage-encoded bicyclic peptides are generated by chemically cyclizing random peptide libraries on phage. Binders are identified by conventional phage panning and DNA sequencing. Next-generation sequencing and new sequence alignment tools have enabled the rapid identification of bicyclic peptides. Bicyclic peptide ligands were developed against a range of diverse target classes including enzymes, receptors, and cytokines. Most ligands bind with nanomolar affinities, with some reaching the picomolar range. To date, several bicyclic peptides have been positively evaluated in preclinical studies, and the first clinical tests are in sight. While bicyclic peptide phage display was developed with therapeutic applications in mind, these peptides are increasingly used as research tools for target evaluation or as basic research probes as well. Given the efficient development method, the ease of synthesis and handling, and the favorable binding and biophysical properties, bicyclic peptides are being developed against more and more targets, ever increasing their potential applications in research and medicine.

2017

In Vitro-Evolved Peptides Bind Monomeric Actin and Mimic Actin-Binding Protein Thymosin-β4

Davide Bertoldo, Raphael Johannes Gübeli, Christian Heinis, Ganesh Kumar Mothukuri, Jonas Alfred Karl Wilbs

Actin is the most abundant protein in eukaryotic cells and is key to many cellular functions. The filamentous form of actin (F-actin) can be studied with help of natural products that specifically recognize it, as for example fluorophore-labeled probes of the bicyclic peptide phalloidin, but no synthetic probes exist for the monomeric form of actin (G-actin). Herein, we have panned a phage display library consisting of more than 10 billion bicyclic peptides against G-actin and isolated binders with low nanomolar affinity and greater than 1000-fold selectivity over F-actin. Sequence analysis revealed a strong similarity to a region of thymosin-β4, a protein that weakly binds G-actin, and competition binding experiments confirmed a common binding region at the cleft between actin subdomains 1 and 3. Together with F-actin-specific peptides that we also isolated, we evaluated the G-actin peptides as probes in pull-down, imaging, and competition binding experiments. While the F-actin peptides were applied successfully for capturing actin in cell lysates and for imaging, the G-actin peptides did not bind in the cellular context, most likely due to competition with thymosin-β4 or related endogenous proteins for the same binding site.