reporter protein are taken (e.g. the DNA binding domain and transcription activation region of the GAL4 transcription factor) and fused to a “bait” and “prey” protein respectively (Fields & Song, 1989). If the bait and prey interact, the biological circuit is complete and an output occurs. In the GAL4 system, the transcription/translation of a reporter gene such as !-galactosidase is used, which converts exogenous X-gal from a colourless compound into a blue product. Blue yeast colonies therefore indicate a successful interaction. A similar system utilising part of a native RNA polymerase has been used in E. coli (Joung et al., 2000). For bioactive peptide screening, the prey protein gene may be replaced by a DNA library to find potential peptides that interact with a defined bait protein; or an existing protein-protein (i.e.
bait-prey) interaction may be targeted for disruption by an additionally expressed peptide library. Peptides found to bind the MyD88 adaptor-like protein (see Section 1.2.2.3) were identified using such a two-hybrid approach (Watt, 2009). However, false positive rates are reportedly high, presumably due to non-specific bait-prey interactions (Brückner et al., 2009).
Other recombinant screening techniques include mRNA display (Roberts &
Szostak, 1997), ribosome display (Zahnd et al., 2007) and CIS display (Odegrip et al., 2004). All of these techniques involve in vitro transcription and translation as opposed to phage display. In mRNA display, a DNA library is used to generate an mRNA library (no stop codons included), which subsequently has a puromycin moiety (a mimic of tyrosyl-tRNA) ligated to its 3’ end. In vitro translation is then conducted, and once the ribosome reaches the incorporated puromycin, translation is disrupted (puromycin serves as an antibiotic in this manner). Crucially, the puromycin is ligated by the ribosome to the C-terminus of the peptide, thus physically coupling genotype with phenotype. Ribosome display proceeds in a similar manner, except that a spacer sequence is used instead of puromycin at the 3’ end of the DNA library, and a lack of stop codon in the resulting transcript causes the ribosome to stall on this spacer. However, the peptide has enough “chaser”
sequence to ensure it is properly displayed rather than being masked inside the ribosome exit tunnel. Finally, CIS display (CIS being derived from cis-acting) couples the peptide directly to its library DNA template via fusing the coding sequence upstream to that of the bacterial plasmid protein RepA, which when
template. If transcription and translation are coupled in vitro, the nascent RepA protein is able to perform this function while the DNA template, RNA polymerase, mRNA and ribosome are in a complex; thus each library peptide is directly linked to its DNA template. An alternative library DNA/corresponding peptide association can also be achieved through in vitro compartmentalisation techniques such as water/oil droplet emulsions that encapsulate single library DNA members (Miller et al., 2006).
Example uses of these in vitro approaches include screening for single-chain antibodies with improved target binding affinities (Hanes & Plückthun, 1997; Fukuda et al., 2006), as well as finding peptide ligands that bind enzymes such as lysozyme (Odegrip et al., 2004).
Peptides produced by all of these methods may then be screened for specific binding or function, and their genotype recovered for further analysis or subsequent enrichment (i.e. screening again). It is possible to use such methods to isolate peptides with very low dissociation constants (<10 nM) to a particular target (Sato et al., 2006a).
1.2.3.2 Advantages and disadvantages of recombinant peptide screens
There are general problems to consider related to transcribing and translating peptide libraries: sequence length, promoter efficiency, codon bias, product solubility and activity in differing physiological conditions are all contributing factors (Ingham & Moore, 2007). The choice of expression system can help with this, as well as the use of larger protein scaffolds or tags in order to aid stability, solubility or downstream purification of a peptide. In addition, such a single-peptide approach necessarily limits the complexity of what may be identified. Any peptide requiring co-factors for activity (unless serendipitously provided by the expression system) or post-translation modification will be missed, as well as those that may show synergistic action with other peptides. Along with the limitations of uncontrolled insertion orientation and reading frame, this leads to a narrowed window of discovery when using DNA libraries to screen for bioactive peptides. Despite this, such an approach is still capable of finding useful peptides, as previous examples cited in Section 1.2.3.1 have illustrated. Screens will inherently ignore peptides that are poorly expressed or insoluble, which serves to remove entities unsuitable for further development. Any peptide that is identified as a hit should therefore be
physiochemically suitable for future production. It is also worth noting that discovery of even a single peptide with limited function may be an important first step in elucidating little-known or new pathways to do with the activity being assayed for.
1.2.3.3 Screen considerations
Because of a peptide’s limited bioavailability and stability (at least before optimisation; see Section 1.1.2), it may be more worthwhile pursuing peptide agonists than antagonists. An agonist of a particular receptor is typically only required at low concentration for activation, and a short half-life is not problematic due to the ensuing signal cascade that such an activation typical evokes (Lien &
Lowman, 2003). In contrast, an antagonist is competing with the native ligand for its binding site, and thus requires higher concentrations and binding site residence time for efficacy. In general, an antagonist must occupy over 50% of the receptor population to have the desired effect, in comparison to 5-20% occupation for an agonist (Vlieghe et al., 2010). Despite this knowledge, many screens are targeted towards identifying antagonists rather than agonists because screening for inhibitors of natural ligand binding is more straightforward than the functional screens that agonists require (Lien & Lowman, 2003). Common antagonist targets are specific receptors or enzymes, which have ligand binding constants measured in vitro by utilising techniques such as surface-plasmon resonance or isothermal titration calorimetry (Rich & Myszka, 2000; Pierce et al., 1999). In contrast, finding peptides that provoke a desired response through other modes of action, like disrupting protein-protein interactions or allosteric activation/inhibition, require more complicated, whole-cell functional screens. This dissertation is concerned with whole-cell approaches.
Whole-cell phenotypic screening is a more holistic way of looking for potential bioactive peptides. Much like classical forward genetic screens, such an approach looks for the desired phenotype/output and then studies the agent responsible (Watt, 2009). For recombinantly-produced bioactive peptides, where the input genotype is linked to an output phenotype in a discrete manner, identification of the casual agent is straightforward. If the interaction partner(s) of an active peptide is unknown, modern proteomic techniques such as chemical cross-linking, pull-down
A screen relying on positive selection, i.e. an output is given or its level is increased, is more desirable than negative selection, i.e. a decrease in an output from an observed baseline, as this allows for rare peptides to be more easily identified from large pools. For instance, if the desired activity in an in vivo screen is coupled to an output of antibiotic resistance, organisms that can grow when challenged by this antibiotic represent hits – those that are negative fail to appear. Coupling the desired peptide activity to a selectable phenotype in this manner gives a screen enormous resolution power (cf. phage display in Section 1.2.3.1). In contrast, locating and isolating clones that fail to grow is more difficult. Such a positive selection approach may be required if the hit rate is low (i.e. less than one per million) (Watt, 2009).