Random peptide libraries

Many different random peptide libraries have been generated and proven effective for isolating novel targeting peptides. The main variable features of RPLs are summarised below.

(a) Phage coat protein on which peptides are dispiayed

Random peptide libraries in Ff phage generally contain fusions to pill or pVIII coat proteins. Since pill is located at the phage tip, fusion peptides of several

hundred residues are tolerated by this protein without significant loss of phage infectivity (Smith 1985). However the low copy number of pill ( 3 - 5 copies per particle) means that plll-fused peptides are presented at low valency. In contrast, fusions to pVIII can be represented at up to 2700 copies per phage and are distributed over the entire phage surface, rather than just at one tip. Since inserts of more than five amino acids interfere with the maturation of progeny virions (lannolo et al. 1995), display of larger peptides requires co­ packaging of phage with wild-type pVIII, such that capsids are composed of an approximately 10:1 mixture of wild-type and fusion pVIII (Felici et al. 1991). This way peptides of up to 20 amino acids can be displayed on pVIII without affecting phage infectivity, and are still represented at a high valency of

several hundred copies per phage particle (Barry etal. 1996).

Whilst both types of display have proven successful for isolating novel targeting peptides, the main difference between the two are in terms of avidity: unlike affinity, which is a measure the binding strength between two sites, avidity is the combined result of multiple interactions - each with a specific affinity - between two binding partners. Due to the increased valency of pVIII-mediated display, these phage will interact with the target substance with higher avidity than phage displaying identical peptides on pill. Hence pVIII-display has advantages when selecting for peptides which bind the target with relatively low affinity. Due to the proximity of pill proteins to one another, multivalent display on pill may lead to undesired interaction between displayed residues. For this reason pVIII-display is often preferred when a

greater distance between displayed peptides is required (Kay et ai. 1996).

interactions possible between pVIII-displayed peptides and cellular receptors,

although uptake following plll-display has also been demonstrated (Hart et a i

1994; Ivanenkov et a i 1999a; Larocca et a i 2001). The majority of

researchers have used plll-displayed libraries, possibly because of the lack of commercially available pVIII-display libraries to date.

(b) Phage or phagemid vector

RPLs are generated using either phage or phagemid vectors. Phage vectors are based on the genomes of bacteriophage M l3, fd or f1 that contain a fusion to the corresponding coat protein gene. For pVIII-display a second, wild-type copy of gVIII is usually inserted, to allow production of stable phage particles. Most phage vectors also contain an antibiotic resistance gene, which allows plasmid-style propagation of phage DNA in bacteria, and reduces selection against phage that replicate slowly due to their insert

sequence (Kay et a i 1996). Phagemid vectors are small plasmids that

encode just bacterial and phage origins of replication, a packaging signal and the fusion coat protein gene. All other phage proteins, including a wild-type copy of the corresponding coat protein, are supplied by helper phage - such as M13K07 - which contains a mutation to minimise packaging of its own genome. Most phagemid vectors also contain bacterial resistance genes.

Apart from the technical advantages of using phagemid vectors in terms of the ease of vector manipulation and phage titration, the main difference between phage- and phagemid-vectors lies in the size of phage particles generated: since phage vectors encode all genes required for phage propagation, their

genome size and subsequent phage particle size is approximately three-fold larger than that of phagemid vectors. This difference in size contributes to the increased level of receptor-mediated cellular uptake generally observed with

phagemid- vs. phage-encoded particles (Ivanenkov et al. 1999a; Larocca at

al. 2001). Phage display vectors have been classified into six groups by

George Smith, based on the coat protein chosen for display (pill or pVIII), the valency (mono- or multivalent) and vector type (phage or phagemid; Figure 1.5).

(c) Length and flexibility of the dispiayed peptide

Random peptide libraries generally contain between six and fifteen random residues, and are often flanked by one or two spacer residues such as glycine or alanine. Whilst long peptides generally have a greater structural stability and are capable binding with higher affinity, they also present a greater risk to

causing an immune response (Koopmann at al. 1996), which can affect future

in vivo applications. Furthermore the inefficiency of bacterial transformation currently limits library complexities to approximately 1 0^^ - 1 0^^ different

clones, which means that libraries containing more than nine random residues

Type 3 Type 8 Type 33 PHAGE Type 88 ^ o v m H ~~gTn>^ V—«^■T aZ nD -^ Type 3+3 Type 8+8 C-M(iverP° PHAG EM ID

FIGURE 1.5 Classification of Ff phage display vectors (Smith & Scott 1993): Multivalent phage vectors for display on pill and pVIII are termed type 3 and type 8 respectively, and contain no wild-type copy of the coat protein chosen for display. To increase phage stability, a wild-type copy of the coat protein containing the fusion is present in type 33 and 88 vectors. Type 3+3 and 8+8 phagemid vectors only encode the pill or pVIII fusion protein respectively. All other phage proteins, including wild-type pill and pVIII, are supplied by a helper phage, such that phage are packaged in a mixture of wt and recombinant coat proteins. Adapted from (Kay et al. 1996).

Displayed peptides have a free amino terminus, giving them flexibility similar to free peptides in solution. Due to the relatively short length of peptides in RPLs it is unlikely that they form stable secondary structures. Hence random peptide libraries often contain constrained peptides, which bind targets with increased affinity due to the lower entropy of displayed peptides (Ladner 1995). Flanking cysteine residues are usually used to constrain peptides, which form disulphide bridges and thus force the random peptide into a hairpin loop conformation (O'Neil etal. 1992; Figure 1.6)

X

X

X

X

X X X

X

X

X

X

X X

X

0 c

)

)'

FIGURE 1.6 Linear vs. constrained dispiay o f peptides: A linear stretch of seven random amino acids (X7) will have a relatively high entropy and thus a less stable conformation. In contrast, if a cysteine residue is inserted at either side of the random region, a disulphide bridge will form between the two cysteines, forcing the random residues into a hairpin loop of lower entropy and greater stability (McLafferty et ai. 1993).

Libraries containing one or more cysteine pair(s) within the variable region

have also been used, such as CX3CX3CX3C or X2CX4CX (C = cysteine; X =

any amino acid) libraries (Rajotte et al. 1998). Whilst stronger enrichment is

generally seen when constrained libraries are used, there is no evidence that

an increasing number of cysteine pairs yields better results (McLafferty et al.

1993; Rajotte et ai. 1998). Nevertheless it is impossible to predict the peptide length and degree/pattern of constraint that will yield the best results for a particular application (Bonnycastle et al. 1996; Irving et al. 2001). For this reason many researches screen with more than one library at a time (Rajotte etal. 1998; Kennel etal. 2000; Lee etal. 2001c).

Random peptide libraries displayed on the surfaces of lambda phage (Kuwabara et al. 1997), T7 phage (Essler & Ruoslahti 2002) have also been screened, and developments for display on MS2 phage are looking promising (Ernst ef a/. 1998).