Phage display was designed as a molecular sieve: billions of peptide variants are incubated with an immobilized target, non-binders are washed off, and binders are eluted, amplified and re-challenged, until dominant sequences emerge. For cyclic peptides, the approach is uniquely powerful: randomized DNA is cloned such that the encoded peptide is displayed on the phage surface, and upon chemical or enzymatic cyclisation in vitro, the same DNA molecule serves as a barcode for rapid identification of active conformations. Because selection occurs under aqueous, physiological conditions, the resulting macrocycles often retain the binding affinity, proteolytic stability, and cell permeability that rival synthetic libraries, yet are discovered in a fraction of the time. The technique has therefore become a gold-standard discovery engine, yielding everything from integrin antagonists to antiviral entry inhibitors, and continues to evolve through new coat-protein formats, controlled valency, and post-selection cyclisation chemistries.
Phage display is based on the finding that foreign peptides fused to the coat proteins of filamentous bacteriophage do not compromise infectivity, and thereby form a physical link between a peptide sequence and its encoding DNA. The platform, introduced in 1985, has since developed from a proof-of-concept tool to discover short linear epitopes into a mature discovery engine for disulfide- or chemically-stapled cyclic peptides that can match antibodies in affinity, and exceed them in tissue penetration. Cyclic formats are particularly appealing since the constrained topology lowers the entropic penalty upon target binding, often resulting in nanomolar affinities and long serum half-lives. Libraries of over a billion unique clones can be prepared in a single electroporation step, and multiple orders of magnitude enrichment of binders over the course of a week can be achieved by repeated rounds of affinity selection (biopanning). Critically, because the phage particle can undergo post-translational cyclisation (either disulfide formation or chemical cross-linking) with the DNA inside and protected, the sequence of the resulting macrocycle can be read out directly. This combination of diversity, speed and genotype–phenotype linkage is the reason why phage display remains unsurpassed for de novo cyclic peptide discovery.
The concept of phage display is based on the ability of bacteriophages to display a large number of peptide variants on their surfaces. A wide variety of peptide libraries can be generated by fusing the desired peptide sequences to the genes that encode phage coat proteins. Individual phage display libraries can then be exposed to the target of interest, and peptides with affinity to the target are enriched by iterative rounds of selection. The key to phage display is that the peptide displayed on the surface of the phage can be physically linked to its corresponding gene, making it possible to efficiently identify high-affinity binders. Phage display has been used for many years to identify peptides with high affinity to a range of target molecules, in particular cyclic peptides, and remains one of the most popular approaches for peptide discovery.
Filamentous M13 phage are the workhorses of cyclic-peptide selection, largely due to their unique non-lytic life cycle, which ensures that displayed peptides are accessible to the extracellular environment throughout infection. The foreign sequence is typically cloned into the N-terminus of minor coat protein pIII, of which three to five copies are present on the virion and mediate host recognition. Each pIII can carry peptides of several kilodaltons without adversely affecting phage infectivity, allowing the display of constrained macrocycles or non-canonical amino acids introduced through nonsense suppression or post-translational modification. In contrast, major coat protein pVIII is reserved for extremely short motifs (≤ seven residues) and usually requires "g-type" fusions that maintain virion length. Nevertheless, recent engineering of mosaic coats have made possible mixed display, allowing for high valency to maximize initial capture with low valency to enable stringent off-rate selection. The fusion gene is then packaged within the phage, and the peptide is expressed during bacterial secretion and presented on the virion surface. The phage library is then ready for affinity selection against immobilized or soluble targets. After washing and elution, the enriched phage population can be easily amplified by infecting fresh bacteria. In less than a day, a complete panning round can be run, allowing for iterative enrichment of cyclic binders.
Fig. 1 The structure of the M13 filamentous phage, and the schematic illustration of the genotype–phenotype link M13 is a rod-shaped bacteriophage with a single-stranded DNA (ssDNA) genome encapsulated by five coat proteins.1,5
Diversity is created by solid-phase synthesis of degenerate oligonucleotides encoding randomized peptide cassettes flanked by unique restriction sites. For cyclic libraries, it is common to fix two cysteine codons at symmetrical positions, such that oxidative folding upon phage propagation will result in a disulfide-bridged macrocycle. The intervening residues can be fully randomized (NNK codons) or biased towards hydrophilic amino acids in order to increase aqueous solubility during selection. Library sizes often exceed 109 independent clones, large enough to sample every possible 7-mer and most 10-mers. Larger diversity can be achieved with trinucleotide phosphoramidites that are selected to avoid stop codons and rare tRNAs, or with split-pool synthesis to introduce defined motifs (e.g. RGD, PWLI) within otherwise randomized loops. Following selection, the DNA insert is amplified by PCR and high-throughput sequenced, yielding thousands of unique cyclic consensus sequences that can be chemically synthesized for validation. Notably, the cysteine constraint required for phage display may be replaced by a lactam or thioether bridge during chemical re-synthesis, such that disulfide vs. non-redox-stable topologies can be compared directly without the need to re-screen. In this way, phage display is used as both a discovery engine and a sequence-to-structure training set for downstream AI-guided design.
Fig. 2 Principle of filamentous bacteriophage M13 phage display using a phagemid vector.2,5
Cyclisation of peptides displayed on phage can be an important maturation step to stabilize a peptide, convert a linear floppy epitope into a constrained protease resistant macrocycle and improve target affinity. In the phage display pipeline the cyclisation chemistry chosen impacts not only the stability of the selected peptide, but also the chemistry set for further chemical elaboration and translation into therapeutics. The most widely used and evolutionary inspired cyclisation chemistry is the formation of disulfide bonds through oxidation of genetically encoded cysteines, but their redox lability precludes disulfides for intracellular and systemic applications. Alternative chemistries that are orthogonal to the genetic encoding strategy and introduced post-phage selection are frequently used, including chemical crosslinkers to form non-reducible lactam or thioether bonds and enzymatic strategies such as sortase-mediated ligation to effect head to tail cyclisation in mild aqueous conditions. These post-display approaches increase the chemical space amenable to phage selection and permit discovery of macrocyclic peptides with both the binding specificity refined by biological selection and the pharmacokinetic properties needed for further development.
Cyclisation by disulfide linkage has been achieved using phage display. In this approach, cysteines are encoded at predetermined positions within the library, often at the termini or on either side of a random sequence loop. The thiolate anions are deprotonated, and the radicals are coupled in the oxidizing environment of the bacterial periplasm during phage propagation to form a disulfide bridge that forces the peptide to fold into a hairpin shape. This process is relatively efficient, and phage-displayed cysteine-containing peptides can display paired disulfides with high yields, often higher than those of synthetic samples. This might be the result of a chaperone-like effect of the phage coat protein and steric isolation of individual peptide chains. The great benefit of this method is its simplicity; the library just has to include two cysteine codons, and cyclisation will be accomplished during phage production autonomously. Thus, the link between peptide phenotype and genotype is maintained and no additional chemical steps are necessary. The resulting disulfide-linked macrocycles display often increased binding affinities over linear peptides, since the entropy loss upon target binding is reduced and the side chain moieties are displayed in a pre-organized fashion that is often similar to that of natural epitopes. In addition, the disulfide bond may form a hydrogen bond network that also stabilizes the complex. However, because disulfide bonds are labile in reducing conditions (e.g., inside mammalian cells or in the presence of glutathione), the stability of disulfide-cyclized peptides can be an issue, although it has prompted the generation of libraries with nonreducible cysteine surrogates or other cross-linking motifs. Nonetheless, disulfide cyclisation is often used in discovery settings since it is a straightforward method and there is a large amount of structural data relating disulfide connectivity to binding motifs.
Chemical post-display cyclisation is an attractive and flexible alternative to disulfide-based approaches as it enables the synthesis of macrocycles that are non-reducible yet maintain the binding selectivity seen in the phage but are much more stable in physiological conditions. The selected enriched sequences are chemically synthesized with reactive handles installed for cyclisation (N-terminal cysteines and C-terminal thioester for native chemical ligation or side-chain lysine and aspartate residues for lactam bridge formation, for example) and the topology can be controlled through the use of linkers and crosslinking reagents, so head-to-tail, side-chain-to-side-chain or backbone-to-side-chain linkages can be installed as needed for the target (see below for examples). For example, bifunctional linkers with an activated ester or maleimide could be used to crosslink to cysteines to form non-reducible thioether bonds, or orthogonal protecting groups can be used to selectively deprotect and activate specific residues as needed.
An enzymatic post-display approach, sortase based ligation, is a more modern approach that utilizes the high specificity of an evolved biocatalyst to perform head-to-tail cyclisation under aqueous conditions which do not require the use of any harsh reagents or introduce potential side reactions. Sortase A is a transpeptidase enzyme that is found in Gram-positive bacteria and recognizes the pentapeptide motif LPXTG and cleaves the bond between threonine and glycine to form an acyl-enzyme thioester intermediate. The reaction is then driven to completion by nucleophilic attack from the N-terminal oligoglycine or cysteine, which forms a native peptide bond.
Biopanning is the workhorse of phage display. The primary objective is to harness, with each cycle of binding, washing, and amplification, the extraordinary molecular diversity presented by billions of individual phage clones in order to achieve a library size reduction by multiple orders of magnitude to a smaller population of high-affinity cyclic peptide ligands. This system interrogates all library members in parallel (versus screening a single variant at a time with a biochemical assay). The genotype and phenotype are physically tethered. Rare binders can be enriched in several cycles of biopanning, by repeatedly performing rounds of binding to the immobilized target (a purified protein, membrane fragment, or small-molecule hapten), washing and eluting phage that are bound to the target. Elution is optimized to not denature the phage. The eluted phage are then amplified in a bacterial culture and used as the input library for the next cycle of biopanning. Biopanning is repeated three to five times, each cycle being more stringent than the previous one. The off-rate, specificity, and binding affinity are improved in each cycle. The whole procedure can take less than 1 week, at the end of which the DNA sequences of the final population of phage-displayed peptides are easily determined by Sanger or next-generation sequencing. It is possible to sample billions of molecules of any cyclic topology (disulfide-constrained, head-to-tail, or side-chain-linked) by biopanning and thus discover potent and robust ligands that can be used as therapeutics.
Solid-phase immobilization of the target is the critical step for successful biopanning as the orientation, density and stability of the bait determine the quality of selected cyclic peptides. The most widely used platform is the high-binding polystyrene microtiter plate, on which the target protein is passively adsorbed overnight at 4 °C in a neutral buffer; although this is easy to use and amenable to automation, it orients the target at random, potentially burying binding epitopes or exposing non-native surfaces that can enrich for artifactual binders. Covalent coupling of the target to activated esters (NHS-ester) or maleimide chemistry on functionalized plates or magnetic beads results in more consistent orientation and higher surface density of target protein, reducing background and allowing more stringent washing without target loss. A gold standard for oriented immobilization is the biotin–streptavidin system: the target is enzymatically biotinylated at a single, engineered tag (AviTag) and captured on streptavidin-coated beads, with the result that the binding interface is presented uniformly and washing is very efficient. Immobilization of membrane proteins or whole cells can be achieved by indirect capture onto liposome-coated beads or by direct panning against intact cell monolayers, both of which require careful blocking with irrelevant proteins to prevent non-specific phage adherence. Affinity selection cycles typically start by incubating the naive phage library with the immobilized target for a period of time sufficient to allow binding equilibrium to be reached. This often takes one to two hours at room temperature with gentle agitation. The buffer used for binding is important: physiological salt concentrations (150 mM NaCl) and neutral pH are used to maintain the native conformation of the target, and the inclusion of low concentrations of non-ionic detergent (0.1 % Tween 20) reduces hydrophobic contacts without disrupting specific interactions. After binding, extensive washing with the same buffer removes weakly bound phage. The number and duration of washes are increased for each panning round to increase stringency. Bound phage are eluted either competitively, with excess soluble target or a low-pH glycine-HCl buffer (pH 2.2), or biologically, by direct infection of host bacteria. Competitive elution selects for fast on-rate binders, and elution at low pH captures the total bound population including slower off-rate binders. The eluted phage are neutralized immediately and amplified, completing one round of biopanning and setting the stage for iterative enrichment.
The cycling of selection rounds is what powers the process of phage display from a library into a pool of enriched specific cyclic peptide binders. Each round of selection acts as a sieve, through which most of the background are removed, while a small population of the desired, rare, high-affinity clones are able to pass through and be amplified. After the first panning, the phage that are recovered from the well represent a very small proportion of the original library (usually 10- to 100-fold enriched); this is magnified with each subsequent round (e.g., 10-fold per cycle) such that by round 3 or 4, 1 clone is likely to be massively enriched compared to the starting library. The stringency of selection can be increased at each step by reducing the concentration of the target, increasing the wash time or increasing the ionic strength (all of which will selectively remove low-affinity phage, which are those that have a high-off rate). A negative selection step can be included at later rounds of selection (phage library pre-incubated with immobilised irrelevant proteins/binding partners and blocking agents to remove non-specific binders which may bind to the solid support, rather than the target). Amplified phage at each round are titered to give an indication of recovery (which is semi-quantitative indication of enrichment). If the titer is going up with each round, then this suggests that the selection process is working and the target is being captured, with each round removing more of the background and is plateauing/going down then the target is not being captured in sufficient numbers or is being washed off. Enrichment at the sequence level can be monitored by Sanger sequencing of individual clones picked from a final round of selection, where one can see if there is a consensus motif indicating that the library has converged upon a binding solution, or more comprehensively by next-generation sequencing (NGS) of the entire library after each round to get a more holistic view of the pool diversity and dynamic process of enrichment of each clone.
Hits are identified by DNA sequencing to determine the identity of the displayed peptides. After several rounds of selection, the bound phage particles are isolated, and their DNA is extracted. The DNA is then sequenced to identify the amino acid sequences of the displayed peptides. High-throughput sequencing methods are commonly used to process the large number of variants in phage display libraries. After sequencing, the identified sequences of potential hit peptides are re-synthesized and validated to confirm their binding affinity and specificity. Validation is important to ensure that the hits truly bind to the target of interest and have the desired biological activity. A variety of assays, such as binding assays and functional assays, can be used to validate cyclic peptide hits.
Phage display has been successfully used to identify cyclic peptides that serve as enzyme inhibitors, receptor ligands, and protein–protein interaction (PPI) inhibitors. A major advantage of the phage display approach in all of these applications is the ability to screen extremely large libraries under conditions that are likely to be physiologically relevant. In particular, phage display has been used to identify inhibitors of enzymes with shallow, flexible active sites or allosteric pockets that are not easily targeted with small molecules. The constrained peptide loop presents a target protein with a molecular shape that closely resembles the desired transition state or substrate leading to selection of inhibitors that often have high specificity and slow off-rates. For example, cyclic peptides that inhibit carbonic anhydrase isoforms have been isolated from phage libraries encoding sulfonamide-linked cyclic peptides. Selected peptides not only bind the catalytic zinc ion but also make contacts in surrounding pockets on the protein surface. Binding to these adjacent pockets leads to selectivity between very similar isoforms that is hard to achieve using conventional small molecules. Phage-derived inhibitors that block the activity of carbonic anhydrases have been shown to have nanomolar potency and can be stable in serum.
Phage display for cyclic peptide discovery provides an attractive set of advantages of ultra-high diversity, fast selection, and direct genotype–phenotype link that is not available with any other display platform. However, its utility is limited by a number of practical factors, including biological bias, library design, and downstream chemical maturation. The primary advantage of phage is library size, with 109–1010 unique clones routinely being produced, which is thought to cover sequence space completely for cyclic peptides up to 12 residues in length. This is combined with a strong selection performed in aqueous buffer which leaves the structure of the displayed macrocycle intact, allowing enrichment of binders under conditions similar to the physiological milieu. Finally, the direct link between genotype and phenotype eases hit identification, as sequence information is stored in the phage DNA, so the exact amino-acid composition of enriched clones can be identified by DNA sequencing without the need for Edman degradation or MS de novo sequencing.
On the other hand, phage display also has several intrinsic limitations that must be accounted for in order to be successful. First, library composition is biased by both the genetic code and the host tRNA pool; NNK randomization is convenient but leads to an under-representation of some amino acids and non-proteinogenic residues cannot be encoded directly. Second, phage infection and amplification efficiency is modulated by the displayed peptide sequence, which can lead to the enrichment of "parasitic" clones that can amplify rapidly even in the absence of binding to the target. This amplification bias is typically accounted for by performing a control panning experiment without target, followed by bioinformatic subtraction of the parasitic clones. This can be computationally expensive and requires a well-tuned statistical model. Finally, phage display is limited to amino-acid sequences, and post-translational modifications such as phosphorylation, glycosylation, or N-methylation must be added after selection, which can be synthetically challenging and potentially modify the binding conformation found during panning.
Next-generation sequencing (NGS) integration is quickly turning phage display from the low-throughput clone-picking exercise it has been for 30 years to a high-throughput, deep-sequencing-enabled discovery machine that can survey and take advantage of the full complexity of the selected population. Rather than picking a dozen clones at the end of a final panning round and Sanger sequencing them, one can now quantitatively track the frequency of every peptide sequence through every selection cycle using NGS, creating enrichment trajectories that not only track the most abundant binders but also pick out rare high-affinity clones that are only transiently enriched as they are outcompeted by other parasitic sequences in the library. Bioinformatic enrichment scoring and motif clustering deconvolutes the selection landscape to find consensus enrichment patterns that would otherwise be overlooked by Sanger sequencing and pick out rare sequences that are enriched at a statistically significant level. This has the dual benefit of allowing for stricter filtering of false positives, while also accelerating progress from the discovery phase to validation. Furthermore, data from NGS-selected libraries can be fed into machine learning algorithms that predict binding affinity from sequence, creating a predictive feedback loop that can guide future library design and reduce the number of required panning rounds. Phage-mRNA hybrid display platforms combine the stability and simplicity of phage particles with the ultra-high diversities achievable by in vitro translation systems, and may ultimately address some of the size and composition constraints of traditional phage display. The encoding mRNA is also selected on and captured in the same particle in hybrid phage-mRNA selection platforms, and can be directly sequenced without an intervening DNA reverse transcription step, while still displaying the peptide of interest on the surface of the particle by a fusion coat protein. This effectively bypasses the transfection bottleneck in library size that yeast and bacterial display methods have, enabling theoretical diversities in excess of 1013 unique clones and still retaining the physical robustness and ease of handling of phage particles. In vitro translation also enables the use of non-canonical amino acids and chemical modifications during the selection process. The incorporation of ribosomal synthesis with phage-like particles also enables continuous evolution approaches, where selected peptides are directly reverse-transcribed, amplified, and re-translated in a microfluidic device to accomplish Darwinian selection on the scale of hours instead of days. These hybrid platforms are still in their relative infancy, but as they mature they are expected to open up whole new classes of cyclic peptides that fill the therapeutic gap between synthetic macrocycles and biologics.
Phage display remains one of the most powerful technologies for discovering high-affinity cyclic peptides. Our platform integrates library design, cyclization strategies, and screening workflows to help you identify potent binders quickly and efficiently.
We support:
Whether you are targeting enzymes, receptors, or PPIs, our discovery workflows deliver diverse, high-quality cyclic peptide hits ready for optimization.
References
