Phage-displayed peptide libraries

Abstract

Over the past year, significant advances have been achieved through the use of phage-displayed peptide libraries. A wide variety of bioactive molecules, including antibodies, receptors and enzymes, have selected high-affinity and/or highly-specific peptide ligands from a number of different types of peptide library. The demonstrated therapeutic potential of some of these peptides, as well as new insights into protein structure and function that peptide ligands have provided, highlight the progress made within this rapidly-expanding field.

Introduction

The field of phage display was first begun by George P Smith in 1985 [1]. Yet, it was his idea for phage-displayed peptide libraries, published in 1988 [2], that caused a burst of activity in the field, culminating in the publication of three papers describing the first phage-display libraries in 1990 [3–5]. Since then, phage display has developed into a wide-ranging field, full of novel applications and responsible for significant advances in many areas of protein recognition. In preparing this review, we counted over 170 papers published in the field over the past year. (A complete bibliography of this literature search, including an extended, annotated bibliography that covers antibody [Ab], and site-directed mutant libraries, is available on the World Wide Web at URL: http://www.biol.sfu.ca/faculty/scott/phage97-98). Although we cannot do justice to most of this work, we present the results from a handful of selected papers covering the results from screening peptide libraries. Also presented ate advances in library construction and screening methods. To provide the reader an idea of the impact that phage-display technology is making on areas such as drug discovery and cancer therapy, we also include a few ‘follow-up’ reports on exciting, bioactive molecules that were previously selected from phage-display libraries.

Antibody-binding peptides and the structural basis of peptide recognition

A number of groups have screened peptide libraries with monoclonal (M) Abs were produced against protein and non-protein immunogens, and have isolated ‘peptide mimics’ that cross react with the MAbs. In several cases, these ligand peptides were also immunogenic mimics of the antigen ‘target’ against which the screening Ab was made; that is, if whole phage were used as immunogens, or if the synthetic counterpart to the phage-displayed peptide was used in a protein-conjugate to immunize, the resulting immune sera cross-reacted with the antigen target. Thus, the polyclonal Ab response induced by an immunogenic-mimic peptide will bind to the same epitope on the target antigen as that recognized by the MAb used to select the peptide. As described below, this type of experiment has now been performed successfully for a number of antigens. The concept of using peptides as ‘epitope-mimic’ haptens that produce highly-directed Ab responses has held great promise since its inception [6]; yet, the use of ‘designed’ peptides as immunogenic mimics has not had good success. It is beginning to emerge that small features within the peptide that are involved in Ab binding, but not necessarily in epitope mimicry, may often be responsible for the recent successes. Thus, given a vast set of peptide sequences to choose from, many Abs appear to select effective, immunogenic mimics of their corresponding target epitopes.

Demangel et al. [7] previously isolated clones beating linear-epitope-mimic peptides by screening a disulfide-constrained, 6-mer (CX₆C; single letter amino acid code where X may be any amino acid) peptide library with a malaria-specific MAb. On immunization, several clones elicited malaria-binding Abs, including two clones whose peptide sequence bore no homology with the presumed malarial epitope. More recently, this group isolated two MAbs from the malaria-binding, anti-phage response and compared their heavy chain variable region and light chain variable region sequences with those of the parent MAb that was initially used to isolate the phage clone [8•]. They found that there were significant differences in hypervariable-region sequences, although there were similarities in features contributing to the gross structure of the Ab combining site, such as variable-gene and canonical structure use. Thus, the phage-borne peptide and the malarial epitope may elicit Abs sharing gross structural features that allow cross-reactivity, but that differ in their finer specificities. Similarly, in collaboration with M Yu and P Talbot (Institute Armond Frappier, Laval, Quebec), we isolated a number of linear-epitope-mimic clones by screening a panel of 12 peptide libraries [9] with a MAb that neutralizes a murine coronavirus. In this case, all of the clones shared a strong consensus sequence matching a sequence on the viral coat, and all bound the MAb with similar strength; yet, when each of the clones was used to immunize mice, only one clone produced cross-reactive Abs that protected mice from intracerebral challenge with the virus (M Yu, JK Scott and P Talbot, unpublished data). These results indicate that relatively small sequence differences in the peptides (in this case, differences in the regions flanking the consensus sequence) are responsible for driving the fine specificity of the Ab response. Structural studies are imperative for understanding the complexities involved in this process. The molecular basis of immunogenic mimicry can best be revealed by comparing the target epitope and peptide mimic bound to both the screening MAb and to cross-reactive MAbs produced by peptide immunization (see [10] for an excellent example of such a comparison). As, in many experiments of this type, whole phage bearing cross-reactive peptides are used as immunogens, it would also be valuable to know the structure of a given peptide in the milieu of the phage, rather than as a synthetic analog. In some cases, the structure of the phage-borne peptide can be determined, as shown by the NMR study of Jelinek et al. [11••].

There is also evidence that immunogenic-mimic peptides can be isolated by polyclonal, serum Abs. In previous work, Mennuni et al. [12] isolated several diabetes-specific peptides by screening two peptide libraries with scrum IgGs from a number of pre-diabetic patients; in turn, several peptides from among these were bound by the serum Abs from a number of patients having full-blown diabetes. Mennuni et al. [13••] showed that, on immunization of rabbits, a phage clone bearing one of these peptides elicited Abs that recognized pancreatic-islet β cells. Thus, this immunogenic-mimic peptide may reveal new antigens (or epitopes) involved in the evolution of type-1 diabetes, especially if it appears to produce pathogenic Abs (i.e. Abs that direct inflammatory responses against pancreatic-islet β cells) as a hapten on its own.

There is some controversy as to whether immunogenic-mimic peptides always act as structural mimics of the epitope on the target antigen, by binding MAb through the same mechanism as the target epitope (i.e. by making identical contacts with the Ab combining site). Alternatively, it may be that the mechanism of binding a MAb differs between the target epitope and its corresponding immunogenic-mimic peptide (i.e. with each antigen making different contacts with the Ab combining site). This controversy has been most clearly defined with anti-carbohydrate and anti-DNA antibodies and their peptide ligands, as these types of targets are chemically very different from peptides, and thus would be most likely to bind Ab by a different mechanism from that of a peptide. Phalipon et al. [14••] screened two peptide libraries (X₉ and CX₉C) with two MAbs that had been produced against the lipopolysaccharide (LPS) of Shigella flexneri. There were 19 related sequences identified by the MAbs, with only one clone cross-reacting with both MAbs. The 19 clones were tested as immunogens, and two clones (the one that cross-reacted with both MAbs and one that did not) elicited significant lipopolysaccharide-binding activity. This indicates that structural mimicry may be involved in immunogenic mimicry (as one clone was recognised by both MAbs and, thus, is consistent with being a structural mimic of LPS); but that structural mimicry is not necessarily required for immunogenic mimicry to occur (as the other peptide was specific for only one of the two MAbs and thus must bind by a different mechanism from LPS and, yet, elicited LPS-binding Ab responses). Similarly, Sibille et al. [15••] isolated phage bearing consensus-sequence peptides with two different, polyreactive, anti-DNA MAbs, by screening an X₄CX₆CX₄ library. Although the phage isolated by each MAb cross-reacted only with its selecting MAb, several phage clones, on immunization, produced DNA-binding Abs. These studies show that peptides can act as Ab-specific ligands (as they can discriminate between two DNA-binding MAbs), and also act as effective immunogenic mimics of DNA (in eliciting DNA-binding Abs). Taken together, these two studies [14••,15••] support the notion that some immunogenic-mimic peptides do not function as structural mimics of their LPS and DNA targets in eliciting cross-reactive Ab production.

Fine-specificity discrimination by peptides may be common among anti-DNA and anti-carbohydrate Abs. With a panel of synthetic oligosaccharides, Harris et al. [16••] mapped the minimal epitope on the cell-wall polysaccharide of Group A Streptococcus for five, closely-related MAbs. They showed that all five MAbs recognize the same minimal structure, and so probably bind the same epitope on the cell-wall polysaccharide. Yet, when used to screen a panel of different peptide libraries, the MAbs selected different consensus sequences, and each MAb only bound the phage that it selected. This indicates that the mechanism of binding to peptide ligands differs from that of binding to the cell-wall polysaccharide, as all MAbs bind the same epitope on the cell-wall polysaccharide, yet each MAb would only bind the peptides that it selected. A study by Gaynor et al. [17•] also showed this type of fine-specificity recognition exhibited by peptides. They screened a 10-mer library with an anti-DNA MAb that causes immune-complex-mediated glomerular nephritis when injected into mice, and thus serves as a good model of this type of disease. Although one of the MAb-binding peptides isolated was able to block the deposition of MAb–DNA complexes in vivo, the peptide was shown to be MAb specific, being unable to bind another, closely related, anti-DNA MAb. These results indicate that the peptide is unlikely to be a structural mimic of DNA, and, consequently, will not function to suppress the polyclonal anti-DNA Ab response involved in this type of immune-complex disease.

Contrasting this are studies suggesting that peptides can act as structural mimics of carbohydrate epitopes. As opposed to the functional mimics described above, structural mimics offer the great advantage of being able to replace the target antigen in functions such as diagnostic assays and as inhibitors for a range of enzymes (so long as all of the proteins to be tested bind the target antigen by the same mechanism). Kieber-Emmons et al. [18••] immunized mice with three different peptides bearing the Aro–Aro–X–Aro (where Aro denotes an aromatic amino acid) motifs identified for three different carbohydrates (alpha-methyl mannopyranoside, Lewis Y and the major polysaccharide of N. meningitidis C). The resulting sera cross-reacted at low titers with human breast cancer and melanoma cells, and in another study [19•] with glycosylated HIV-1 envelope proteins, but not non-glycosylated ones. The far-ranging immunogenic mimicry observed in these studies suggests a role for this peptide-sequence motif as a ‘universal’ structural mimic of a variety of carbohydrate targets. Two other recent studies suggest a role for structural mimicry in peptide-carbohydrate cross-reactivity. In two separate screenings of an X₁₅ library, Taki et al. [20•] isolated peptides that bound to another MAb that is specific for its isomer; a consensus-sequence motif that is common to both MAbs was identified from these two sets of peptides. Synthetic peptides bearing this motif bound the lectin from Ricinus communis and also affected the activity of Jack Bean β-galactosidase; one of the two glycosphingolipid isomers is the native ligand of both proteins. Thus, the MAbs, lectin and enzyme may recognize peptides bearing this sequence motif via a shared binding mechanism.

Several interesting studies on the capsular polysaccharide of Cryptococcus neoformans have revealed that cross-reactive peptides are recognized by the sera of multiple individuals, and that the mechanism of cross-reaction is not necessarily structural mimicry. Zhang et al. [21••] showed broad cross-reactivity of phage bearing the consensus sequence QT(G/T)(L/D); these clones were isolated by a human MAb against the Cryptococcal capsular polysaccharide. Phage bearing a representative consensus peptide were used in a competition assay to block binding of Abs from sera of people with HIV-1 infections and without such infections to the Cryptococcal polysaccharide. HlV-negative sera, but not HIV-positive sera, were partially inhibited from binding the polysaccharide (with inhibition levels of 27–41%), indicating that the peptide recognizes a common Ab species among non-infected individuals. These findings arc understandable, given the highly-restricted nature of anti-carbohydrate responses in most individuals, (in which limited numbers of V_H and V_L genes are usually expressed in response to a given carbohydrate antigen), and the diversity of Ab production caused by the polyclonal activation of B-cell responses in HIV-1 infected individuals (in which a large variety of V_H and V_L genes are expressed in the absence of specific immunization with antigen). That partial inhibition occurred with multiple sera indicates that this peptide-specific sub-species of Ab is produced by most people in response to the Cryptococcal capsule, and, thus, suggests the application of peptides in diagnostics; not as carbohydrate mimics, but as indicators of specific Ab sub-species in the Ab response against the Cryptococcal capsule.

In a related study, Valadon et al. [22] isolated decapeptides bearing the consensus sequence TPXW(M/L)(M/L) with a murine MAb against the Cryptococcal capsular polysaccharide. A synthetic peptide bearing this motif was able to elicit Abs having the correct idiotype, but that cross-reacted only weakly with the capsular polysaccharide. Interestingly, these Abs bore the same light-chain sequence as the screening MAb, but several different heavy chains. The crystal structute of the peptide bound to the MAb that selected it [23••] revealed that the peptide associates mostly with the light chain, whereas binding to the Cryptococcal antigen depends upon the sequence of the third hypervariable loop of V_H (H3). This indicates that the peptide most probably binds by a different mechanism from that of the capsular polysaccharide. Thus, the mechanism of cross-reactivity between peptide-mimics and their target antigens may largely depend upon the Ab used to select the peptide, with the selection of structural versus functional mimics being a matter of the peptides having the best fit with the Ab in question. These studies also emphasize the necessity of structural studies in determining the mechanism of cross-reactivity. So far, there is no direct evidence, from crystallographic or other physical studies, of structural mimicry being a mechanism of cross-reactivity of immunogenic mimicry between peptides and non-proteinaceous epitopes. Understanding of the basis of cross-reactivity is essential for applications of peptides in diagnostics, therapeutics and vaccines. Using structural information as a guide, one can decide whether a given peptide should be used in a vaccine or diagnostic application as a structural mimic of the target epitope (as probably occurs for some peptide mimics of linear epitopes on folded proteins; see below) or as an Ab-specific reagent (as appears to be the case for many non-proteinaceous epitopes).

Peptide agonists and antagonists

Peptide ligands have recently been found for receptors (thrombopoietin, melanocortin receptor, CD80, and a hantaviral receptor), receptor ligands (angiogenin, α-bungarotoxin [24••,25]), and folded domains from within larger proteins (SH2 [26•], SH3 and WW domains [27••]). In their second discovery of peptide agonists that are active in the dimeric form, the Affymax group [28••] isolated linear peptides that bind to the thrombopoietin receptor and compete with the natural ligand for binding. An optimzed peptide was covalently dimerized, and in this form was equipotent to thrombopoietin in vitro, and stimulated platelet production in vivo. In several cases, receptor ligands have been isolated by novel methods. The melanocortin (MC)1 receptor is a G-protein-coupled receptor that is activated by a peptide hormone, which also activates other MC receptor subtypes. Szardenings et al. [29••] constructed a ‘sublibrary’ of the hormone-core sequence, flanked by randomized residues, and screened it on whole insect cells that overexpress the MCI receptor. Peptides were isolated having moderately-high affinity (as compared to the native hormone) and very high (3–4 log) selectivity for the MC1 receptor. Thus, peptide libraries can be used to find highly selective receptor ligands by screening on whole cells. Fukumoto et al [30••] used a novel approach to identify a receptor-ligand using a MAb against the CTLA4 signalling protein on T cells; this MAb mimics the CD80 counter-receptor on antigen-presenting cells. They identified from an X₁₅ library two candidate peptides that bind this MAb, one of which binds to CD80 and, interestingly, potentiates T-cell proliferative responses. Yet another novel screening approach is described by Heiskanen et al. [31••], who screened a CX₇C peptide library on whole Puumala hantavirus using ‘competitive elution’ with neutralizing, antiviral MAbs. They identified peptides that neutralize viral infection in vitro at nanomolar concentrations, which is similar to the neutralizing activity of the MAbs. Gho et al. [32••] identified a cyclic octapeptide that binds the factor angiogenin (by competitive elution with actin), and showed that this peptide could block angiogenin activity in a neo-vascularization assay. Furthermore, the peptide blocked angiogenesis induced by a tumor cell line that secretes angiogenin. Following a nearly-identical protocol, Choi et al. [33•] identified 12-mer peptides bearing a different angiogenin-binding sequence from that described by Gho et al. [32••], yet both groups’ peptides bind to a similar or identical site on angiogenin, because they were both obtained by competitive elution with actin.

Phage that home to tumor-associated blood vessels were characterized by Arap et al. [34••], who used mice bearing tumors for the in vivo screening of peptide libraries. Phage bearing two types of cell-adhesion sequence motif (RGD and NGR) were isolated from tumor tissues after intravenous injection of cyclic-peptide libraries that was followed by perfusion to remove unbound phage from the vascular tree. Peptides bearing homing sequences were covalently coupled to the anti-cancer drug doxorubicin, injected into mice beating human breast cancer tumors and shown to cause tumor regression [34••]. Earlier work by the same group [35•] showed that phage displaying the RGD sequence home to a variety of tumors via the alpha-v integrins present in tumor-associated blood vessels.

Several other previously-identified receptor-binding peptides have now demonstrated biological activity, and others have been improved by rational design. In previous work, Bottger et al. [36] identified a peptide ligand for the site on the MdM2 protein that binds the p53 tumor suppressor. As described in a follow-up paper [37••], this group engineered thioredoxin to display a loop bearing the peptide, and showed that the presence of this fusion within cells causes the accumulation of p53. Similarly, Stauffer et al. [38••] improved the selectivity of previously-identified peptides for the SH3 domain of the Lyn tyrosine kinase, and showed that they inhibit Lyn function when electro-porated into mast-cell-like leukemia cells. Acyclic, 20-merpeptide, which, in the dimeric form, binds and acts as an agonist of the erythropoietin receptor [39,40], was further characterized with regard to its critical-binding residues, and was minimized to 13 residues [41•]; its potency was also improved by covalent dimerization [42•].

Substrates and inhibitors have also been found for several proteases. Whereas several new substrate peptides have been identified using the novel ‘substrate phage’ approach of Matthews and Wells [43], inhibitors have been identified by screening phage libraries directly. O’Boyle et al. [44••] constructed three different substrate phage peptide libraries and screened them with the protease of Type 1-herpes simplex virus. An optimized 8-mer peptide, having a different amino acid sequence from the native site, was defined and shown to be cleaved with the same efficiency as the native site. Ke et al. [45•,46•] isolated peptides that are specific substrates for the tissue-type and urokinase-type plasminogen activators, or t-PA and u-PA, respectively. Ploug et al. [47•] further characterized a 15-mer peptide that inhibits the urokinase-type plasminogen activator. In contrast to studies identifying substrate sequences, two papers from Sollazzo’s group describe inhibitors of the hepatitis C virus NS3 protease, which were identified from three different ‘scaffold libraries’ based on a camelized V_H domain [48••], a serine-protease inhibitor and a ‘minibody’ [49•].

Increasing the range of receptors that cross-react with peptides

Several problems are associated with failed screens in which low-affinity or non-binding phage are isolated. The affinity of weakly-binding phage can sometimes be improved by constructing ‘sub-libtanes’ of two types. If a consensus sequence is observed but the peptides bind with low affinity, a new library can be constructed in which the consensus residues are fixed, and the residues flanking them randomized. Alternatively, when there is only a single weakly-binding peptide, a library can be constructed in which the nucleotide sequence encoding a binding peptide is doped (during oligonucleotide synthesis) with lesser amounts of the remaining three bases at each nucleotide in the sequence, to yield variability at every amino acid residue. Such libraries can be re-screened under more stringent selection conditions for tighter binders. Another means of improving the apparent affinity (rather than the instrinsic affinity) of a weakly-binding peptide is to make it multivalent. Terskikh et al. [50•] devised the ‘peptabody’ which displays peptides in a pentavalent array. Alternatively, highly-multivalent phage libraries can be screened. Similar to the ‘organic landscape’ phage libraries of Petrenko et al. [51], Iannolo et al, [52•] constructed an octapeptide library in which the peptides were fused to every copy of the pVIII major coat protein. (Filamentous bacteriophage are composed of thousands of copies of pVIII, with the number of molecules per virion varying with the length of the genome; one pVIII molecule for every 2.3 nucleotides in the single stranded genome.) Thus, this library of high-density peptides was screened with an organic dye to find dye-binding peptides. Such multivalent systems have particular application for targets that are themselves multivalent; furthermore, the phage, being relatively large filaments, can also be used as a cheap means to a selective adsorbant.

As for the problem of negative screens, in which no binding phage are found, it is probable that the existing peptide libraries do not cover enough sequence space to include tight-binding peptides for targets that require more than 5–6 critical binding residues (these are residues that form high-energy contacts, and are usually few in number). In some cases, ‘gene-fragment’ libraries may provide a better way of identifying epitope sequences for protein-binding molecules. In a comparative study, Fack et al. [53••] mapped the epitopes for a panel of four MAbs using two random-peptide libraries (X₆ and X₁₅) versus a gene-fragment library constructed from fragments of cDNA encoding the target antigen. The MAbs always selected cross-reactive phage from the gene-fragment library, but not the random-peptide ones. This approach has limited application as a new library must be made for each screening target, and the target has to be a cloned protein. Alternatively, ribosome-display technology [54,55], which can accommodate libraries of 10¹³–10¹⁴ sequences, may allow the screening of a much larger sequence space in the future.

Whereas the approach to epitope mapping of Fack et al. [53••] is recommended for immunoblot-reactive Abs that presumably recognize linear epitopes, there is still the problem of identifying discontinuous epitopes, whose structures cannot be mimicked by simple peptides. This is in part due to the gross topology of Ab combining sites, which can be related to the types of epitope that Abs can recognize. Anti-peptide Abs bear grooved sites having particular dimensions, whereas Abs against discontinuous epitopes arc larger and flatter [56,57]. Craig et al. [58••] showed, from screening a panel of peptide libraries with polyclonal Ab responses against two folded proteins, that peptide mimics of linear epitopes predominate in these screens, even though Abs against linear epitopes constitute only a small fraction of the total Ab response against folded proteins [59]. Linear-epitope-mimic peptides are much more abundant in peptide libaries than discontinuous-epitope mimics, as the number of different clones isolated for a given linear epitope is relatively large compared to the number of clones isolated for a discontinuous epitope. Unpublished work from our lab with a panel of MAbs known to bind discontinuous epitopes has shown that 60–70% of these types of MAbs isolate rare clones from the same panel of libraries, and these ‘discontinuous-epitope mimics’ often bind their MAbs relatively tightly. Thus, in comparison to linear-epitope mimics, peptides that cross-react with discontinuous protein epitopes are rare or altogether absent from standard peptide libraries. Jespers et al. [60••] have circumvented the problem of mapping discontinuous epitopes with peptides by a strategy in which a library of mutants of a protein antigen is produced, and then ‘negatively selected’ on an a MAb against the discontinuous epitope of interest. Phage that do not bind the MAb are then ‘positively selected’ for global folding by binding to a ligand for the antigen or MAb against a separately-located, discontinuous epitope on the antigen. This approach was extended to mapping the major epitopes recognized by a set of patients in their polyclonal Ab response against a protein antigen [61•].

The view that peptides only sporadically cross-react with discontinuous epitopes on protein antigens can be extended to include many other types of receptor-binding sites, especially those on protein ligands. With the intention of better mimicking discontinuous epitopes, several groups [62–64] have produced libraries that form a surface on a folded-protein scaffold. Such scaffolds are expected to form a more constrained framework for displaying the randomized (X) residues in a library, and, when displayed on separate adjacent strands of a folded protein, the X residues can assemble into a library of discontinuous epitopes. Such libraries are reminiscent of site-directed, protein-mutant libraries, such as protease-inhibitor [65,66] and zinc-finger libraries [67,68], which are targeted to specific classes of ligands, such as proteases and DNA, respectively. Two new libraries have been constructed based on different protein scaffolds [69••,70•]; both groups based their design on very stably-folding scaffolds that would allow modeling of the ligands isolated from them. Nord et al. [69••] designed a three-helix bundle to display 13 randomized (X) positions on its surface. From this library, they isolated ligands for three very different proteins: a polymerase, insulin and an apolipoprotein. Similarly, Smith et al. [70•] constructed a library using the knottin-like, cellulose-binding domain and identified ligands for cellulose and alkaline phosphatase, but not for α-amylase or β-glucuronidase.

Although the chances are slim that a single ‘universal’ constrained-scaffold library will contain ligands for most receptors and Abs, it is quite possible that a panel of different libraries will.

It is clear from several studies [9,28••,39] that the screening of a variety of peptide libraries with a given receptor or Ab increases the chances of identifying ligands for it. Besides the number and types of library screened, the design of the selection experiment is crucial to successful ligand identification; however, few studies exist in which different screening methods have been compared side-by-side. Levitan [71•] has devised a model of phage-library screening to clarify the effects of several parameters in the probability of selecting phage bearing ligand peptides. For instance, we routinely perform screenings on about 16 different peptide libraries that are mixed into seven or 11 pools. Following the general principles outlined by Levatin [71•], we firstly remove high-background phage by pre-adsorbing our libraries on plastic plates before each round of selection, secondly, perform the first round of selection by ‘biopanning’ [2,3] with solid-phase target [9] to ensure the capture of all target-binding phage, thirdly, introduce high-stringency, biopanning selections, [3,72] in later rounds of screeing, by capturing page-target-molecule complexes out of solution and using target-molecule concentration to control selection stringency, and finally, using phage yield (i.e. the yield of binding phage over a control phage) during the selection experiment and a functional assay (usually ELISAs) to choose phage pools from which clones will be selected for further evaluation. As the phage yield and ELISA show the enrichment of target-binding phage for a given round of selection, selection stringency can be increased by dropping target concentration (usually by 1–2 orders of magnitude) and repeating that round of selection.

Conclusions

As new libraries are constructed their value would be mote clearly demonstrated were they tested side-by-side along with several other different libraries in screenings against a variety of receptors, enzymes and Abs. Futhermore, optimization of screening methods should be performed for each target molecule screened to ensure the highest probability of identifying ligand peptides from a given library. As it stands now, most researchers only test a single library at a time using screening methods that arc usually not optimized. The results of side-by-side comparisons between different libraries would clarify the types of libraries that most effectively produce ligands for a given type of receptor or Ab, and whether there are any general rules that govern the types of libraries that contain ligands for a particular class of receptor or Ab. In the longer view, the combination of these types of functional studies with structural studies that determine the mechanisms of peptide binding (i.e. functional versus structural mimicry) would be of great benefit in determining the rules governing protein cross-reactivity.

Abbreviations

Ab

antibody

LPS

lipopolysaccharide

M

monoclonal

MC

melanocortin

V_H

heavy chain variable region

V_L

light chain variable region