Libraries of phages displaying diverse peptides are typically surveyed by affinity selection, using immobilized biomolecules as selectors. After exposing the library to the selector and washing away unbound phages, the bound phages are enriched for clones displaying selector binding peptides. Those phages are recovered by release from the selector and propagation in fresh host cells. Release is generally achieved by weakening the peptide-selector interaction without impairing phage infectivity. A perennial concern with this mode of release is recovery bias—that is, underrepresentation of the highest-affinity peptides because they are not effectively released. Here we argue for trypsin digestion as a superior release mode. It requires that the displayed peptide be connected to the phage body through a trypsin-sensitive tether, and exploits the resistance of the phage itself to that protease. We show that trypsin release is nearly complete even when phages are captured by multiple irreversible bonds, which implies little or no recovery bias.
Phage display is widely used to discover ligands from large phage peptide libraries (1). The recombinant phages in these libraries display cloned foreign peptides or protein domains (both called peptides herein regardless of size) fused to one of the phage coat proteins. Affinity selection is the prime method of ligand discovery. It consists of bringing a library of peptide-displaying virions (phage particles) in contact with a selector (e.g., a biomolecule immobilized in an ELISA well, a semiconductor surface, etc.) in order to allow formation of peptide-selector bonds. The reaction mixture is then processed to remove virions that are not bound to the selector (e.g., by washing ELISA wells). In this way, virions displaying selector-binding peptides are physically separated from the bulk of the library.
Ordinarily, the affinity-selected virions are released from the selector in order to propagate them in preparation for further rounds of selection or for analysis of individual clones. Release is usually accomplished by breaking bonds (usually noncovalent but sometimes covalent) between selector and displayed peptide (1). A perennial concern with this procedure is the possibility of recovery bias; that is, under-representation of the strongest binding peptides in the output because the virions that bear them cannot be released effectively, especially virions that are attached through multiple peptide-selector bonds (2,3). Almost always, these are precisely the peptides that are the most desired output of selection. Proteolytic release, in contrast, cleaves a protease-sensitive tether that has been engineered between the virion body and the displayed peptide, rather than breaking the bonds between the peptide and the selector (4). The tether incorporates cleavage sites for proteases such as trypsin and chymotrypsin that the phage has evolved to resist (5,6). Proteolytic cleavage at such a tether is an appealing mode of release when recovery bias is a concern, since it is presumably independent of the strength of peptide-selector interaction. In some cases, proteolytic release has the added advantage of greatly increasing the infectivity of the virions by eliminating interference from the displayed peptide (7). This article reports a quantitative assessment of trypsin release, focusing specifically on whether or not it allays concerns about recovery bias.
In order to assess trypsin release, we constructed two tester virions that display biotinylated peptides tethered through a trypsin-sensitive linker to a recombinant coat protein subunit. The trypsin-release vectors and their corresponding biotinylated testers are detailed in the Supplementary Materials. One of the testers was a Type 3 construct (8), in which a biotinylated peptide was fused to all five subunits of the pIII coat protein at one tip of the thread-shaped virion. The other tester was a Type 88 construct (8), in which an average of 10 biotinylated peptides were fused to recombinant pVIII subunits distributed randomly along the entire length of the tubular outer sheath, the bulk of the sheath being composed of about 3900 normal, non-recombinant pVIII subunits. Recovery of such multiply biotinylated testers from immobilized streptavidin (SA) is a stringent test of trypsin release. The reason is that biotin binds SA irreversibly, so that release is achieved only when all tethers connecting SA-bound biotins to the virion body are cleaved by trypsin. Release of the testers, therefore, is a model for release of virions that have been captured by an immobilized selector through multiple covalent or superstrong noncovalent interactions.
Trypsin-mediated release of the biotinylated testers from an SA-coated ELISA dish was quantified as detailed in the Supplementary Materials. Briefly, serial dilutions of the biotinylated tester virions were applied to the wells and allowed to bind. After washing to remove unbound virions (if any), wells were treated with trypsin or with buffer alone as a control. Released virions were titered as tetracycline-resistant colony-forming units (cfu; cells that are successfully infected by these phages are not killed and become resistant to tetra-cycline) by in-well infection of Escherichia coli host cells.
Overall recoveries (output cfu per input virion) for individual wells are graphed in Figure 1A. From wells treated with trypsin (× and + markers), recoveries were ~1.6% (results for control wells are discussed below). The primary reason for low recovery is the low intrinsic infectivity of the entire family of phages to which these constructs belong (9,10); even when the virions are infected directly into host cells with no intermediate capture and release, infectivity (cfu/virion) is only ~5% (see Supplementary Materials). The recoveries reported in Figure 1A are therefore roughly a third of the theoretical maximum. The reasons why recovery was not higher are largely unknown, but probably do not include incomplete capture, since much less than 1% of the initial input biotinylated virions could be detected in solution 20 h later (data not shown). It is also unlikely that trypsin reduces overall recovery by directly affecting intrinsic infectivity, since incubation of phages in solution for >4 h under the conditions of the in-well infection had little or no effect on cfu titer (data not shown). Most importantly of all, the data in Figure 1B show that incomplete overall recovery does not reflect incomplete release by trypsin. In order to quantify release, after in-well infection, the ELISA dish was washed and the virions that remained bound were quantified as described in the Supplementary Materials. The results in Figure 1B indicate that trypsin released >90% of the virions, allaying fears that incomplete release strongly biases the output against virions that are connected to multiple immobilized selectors. This is a key finding, since in many affinity selections, such multiply connected virions will be those whose displayed peptides have the highest affinity for the selector and are thus the most highly valued clones in the input population. More generally, since release from SA requires complete removal of all biotins, the results imply that trypsin strips all displayed peptides from nearly all input virions. This largely eliminates the possibility of recovery bias, which cannot discriminate against particular phage clones on the basis of peptides they no longer display. Finally, whatever the reasons for incomplete recovery, the observed overall recovery of 0.016 output infectious units (cfu in this case) per input virion compares favorably to typical recoveries for other modes of release.
The recoveries from the control wells (not treated with trypsin; open circles and diamonds in Figure 1A) were ~4–10× lower than recoveries from the trypsin-treated wells. Nonzero recoveries from these wells were entirely expected. They are typical of “elution by infection,” in which selector-bound virions are recovered by direct infection without prior release (11,12). Elution by infection not only gives lower recoveries, but also fails to result in a detectable decrease in the number of virions remaining bound to the well (Figure 1B). Therefore, for elution by infection, concerns about recovery bias against the highest-affinity (and usually most valuable) clones—concerns that have been allayed in the case of trypsin release by the present work—remain very much in force.
In summary, this article greatly strengthens the case for trypsin digestion as a favorable method for releasing affinity-selected virions from immobilized selectors. Its advantages include not only the benign or positive effects on infectivity that have already been noted by others (5-7); but also elimination of potential recovery bias against the most valuable virions, even when they are bound to selectors through multiple irreversible bonds.
This work was supported by a National Cancer Institute (NCI) Center Grant (grant no. P50-CA-10313, to Wynn A. Volkert) and an NCI research grant (grant no. R21CA127339, to G.P.S.). W.D.T was supported by a University of Missouri Life Sciences Fellowship. We acknowledge the expert technical help of Robert Davis. This paper is subject to the NIH Public Access Policy.
The authors declare no competing interests.
Address correspondence to George P. Smith, Division of Biological Sciences, Tucker Hall, University of Missouri, Columbia, MO, 65211, USA. email: [email protected]