Library screening methods

Overview

Cell surface display systems are valuable tools for protein engineering when coupled with a method for high-throughput screening of large libraries (Boder and Wittrup 1998; Daugherty et al. 1998). While flow cytometry remains a common method for screening cell surface display libraries, other methods such as magnetic bead sorting and panning have also been developed. The library screening strategies discussed in this section are generally applicable to multiple hosts and display formats.

Fluorescence-Activated Cell Sorting (FACS)

Flow cytometry, a mature technology analogous to fluorescence microscopy, is used in both research and clinical settings (Shapiro 2003). When used to isolate distinct populations from mixtures of cells, it is often referred to as FACS. Briefly, the central concept involves hydrodynamically focusing a cell suspension into a stream of single cells. Laser light is then directed onto the cells as they flow one at a time through a detection chamber, which allows simultaneous measurement of light scatter and flu orescence. Cells that match user-defined optical criteria are then isolated by charge deflection into a collection tube for further propagation.

Polypeptide libraries displayed on phage, mRNA, or ribosomes cannot be screened by FACS because these particles are too small to produce the light scattering necessary for detection. In contrast, bacteria, yeast, insect, and mammalian cells are all large enough for detection. For protein engineering projects aimed at improving binding affinity, such as with antibody-antigen interactions, cell surface-displayed libraries are first incubated with a fluorescently labeled target. FACS is then used to isolate fluorescently stained cells resulting from the desired binding interaction. Screening stringency can be controlled through altering ligand concentration, the number of wash steps, and incubation time in the presence of competitive binders. Both equilibrium and kinetic binding screens have been successfully employed using FACS (Boder and Wittrup 1998). For equilibrium binding screens, the yeast-displayed library is incubated with fluorescent ligand at a concentration below the Kd that is desired for affinity enhancement. For kinetic binding screens, mutants with a decreased off-rate of binding can be isolated using an extended wash step, followed by incubation with excess unlabeled ligand prior to sorting. Optimization of these variables, in combination with multiple rounds of FACS, can be used to isolate rare library members with particularly interesting properties (Boder and Wittrup 1998). Library screening criteria can be adjusted based upon cell data collected in real-time. For example, in affinity maturation experiments, cell sorting criteria can be clearly defined to retain only the top 1% of clones with high binding affinity. In FACS, cells are analyzed at speeds of 20,000-50,000 cells/second, depending on the instrumentation used. This creates an upper practical limit of about 108 cells that can be sorted in a typical session. A 10-fold excess of the library diversity should be sampled to ensure recovery of rare clones; thus, only library sizes of up to 107 can reasonably be covered. While enrichment for high affinity clones has been observed after only one round of cell sorting, it is more common to perform several rounds of FACS to enable enrichment of extremely rare clones. Between sort rounds, the pool of collected yeast cells can be analyzed by flow cytometry to measure enrichment of clones with the desired phenotype. After sufficient enrichment, individual mutants are isolated and sequenced, and binding parameters are often analyzed by flow cytometry while the proteins are still tethered to the cell surface, removing the need for soluble protein expression and purification.

When sorting yeast-displayed libraries, it is common to perform dual-color FACS, in which two separate fluorescent signals are measured simultaneously (Boder and Wittrup 2000b; Colby et al. 2004c). Bacterial display systems that employ epitope tags are also compatible with dual-color FACS (Kenrick et al. 2007; Rice and Daugherty 2008). As shown in Figure 2.4, one signal results from a terminal epitope tag that is bound to a fluorescently labeled antibody (used to measure display levels of the protein of interest), while the other signal results from binding of a fluorescently labeled ligand (used to measure protein-ligand binding interactions). Simultaneous detection and, hence, normalization of ligand binding with protein expression enables high-affinity clones to be isolated regardless of their expression level. Moreover, this normalization prevents inadvertent collection of clones with apparent high binding resulting from high expression levels rather than strong binding to the target ligand.

FIGuRE 2.4 Flow cytometry dot plot of a yeast surface-displayed protein library incubated with a fluorescently labeled antibody that binds an epitope tag (x-axis) and a fluorescently labeled target (y-axis). Protein expression and target binding on the cell surface are simultaneously measured through the use of two different fluorophores. The region defined by the polygon indicates cells with maximum target binding at each protein expression level. Cells in this gated region would be collected in an affinity maturation sort to yield the top ~1% of binding clones within the total population.

FIGuRE 2.4 Flow cytometry dot plot of a yeast surface-displayed protein library incubated with a fluorescently labeled antibody that binds an epitope tag (x-axis) and a fluorescently labeled target (y-axis). Protein expression and target binding on the cell surface are simultaneously measured through the use of two different fluorophores. The region defined by the polygon indicates cells with maximum target binding at each protein expression level. Cells in this gated region would be collected in an affinity maturation sort to yield the top ~1% of binding clones within the total population.

Thus, one of the greatest advantages of FACS is quantitative control with single-cell resolution, as demonstrated in one example where fine affinity discrimination of clones with only a twofold difference in binding affinity was achieved (VanAntwerp and Wittrup 2000).

Magnetic Bead Sorting

If one does not have easy access to FACS instrumentation, cell surface display libraries can also be screened against a target that is immobilized on magnetic beads (Christmann et al. 1999; Yeung and Wittrup 2002; Miller et al. 2008). In a typical experiment, biotinylated targets are covalently coupled to magnetic beads that are coated with streptavidin. Cell surface display libraries are usually first incubated with uncoupled magnetic beads to deplete the library of members that recognize streptavidin or bind nonspecifically to the bead surface. The nonbound cells are then recovered and incubated with beads presenting the target ligand of interest. Significant library oversampling is often required to compensate for a high probability of losing binding clones, particularly in early rounds of sorting where individual clones may be represented in relatively small numbers. Following incubation and washes, a magnet is used to attract the beads, which are bound to cells displaying mutants that recognize the ligand. The supernatant, which contains the nonbind-ing clones, is discarded. Magnetic beads with bound cells are transferred to growth media to propagate clones for subsequent sort rounds.

Magnetic bead sorting may also play a complementary role in situations where FACS is available yet may not be practical for use as a sole screening method. For example, since magnetic bead screens can readily accommodate 109-1010 cells per sort, this can be used as a first step to reduce large libraries to a size more manageable for FACS, while simultaneously enriching for desired clones (Siegel et al. 2004). Also, naive libraries that are not based on a pre-existing binding interaction may have only a small subpopulation that recognizes the desired target, and it is likely that most of these clones will also have very weak binding affinity. In these situations, initial rounds of screening should be performed with relaxed stringency to collect as many clones as possible, including ones with weak binding (KD ~ pM). However, FACS may not be the best method to use for screening naive libraries due to difficulty in discriminating between the low signal produced by a weak binding interaction and the background signal produced by clones that are devoid of binding. Furthermore, target can rapidly dissociate from weak-binding mutants during the unavoidable lag time between the final wash and loading onto the flow cytometer, and also during the time it takes to perform the sort. Since FACS often requires a much greater amount of target relative to magnetic bead sorting for isolating clones with weak binding affinity, magnetic bead sorting can also reduce the amount of target required for early sort rounds.

Despite the lack of quantitative control with magnetic bead sorting, sort stringency can be tuned by optimizing parameters such as cell suspension density and incubation time in order to enrich for desired clones (Yeung and Wittrup 2002). Magnetic bead sorting also carries the risk of enriching for clones that show apparent binding as a result of high expression levels and avidity effects rather than 1:1 mutant-target stoichiometry. Therefore, if possible, magnetic bead sorting should be interspersed with rounds of FACS to normalize binding with protein expression levels through immunofluorescent detection of an appropriate epitope tag.

Panning

Panning, a technique used extensively for screening phage display libraries, has been adapted for screening bacterial and yeast surface display libraries against mammalian cell targets (Wang and Shusta 2005; Zitzmann et al. 2005; Wang et al. 2007; Li et al. 2008; Yang et al. 2008). With this method, cell surface display libraries are incubated with living cell lines, and following wash steps, bound library members are recovered and propagated for future rounds of screening. Unlike FACS and magnetic bead sorting, cell panning does not require the production and purification of soluble target, which can be tedious or challenging if the target of interest is difficult to express or poorly behaved. In addition, cell panning allows libraries to be screened against targets in their native cell membrane environment. This screening method can also be used as a potent strategy for discovering binders that recognize previously unknown targets. For example, a yeast-displayed scFv library of ~109 clones was panned against brain endothelial cells to identify scFvs that bound tightly to cells, and in some cases, were internalized into cells (Wang and Shusta 2005; Wang et al. 2007). This finding was important because antibodies that engage endocytosis pathways have potential for delivering payloads, such as transporting therapeutic agents through the blood-brain barrier.

Density centrifugation is a common technique in immunological studies for separating mixtures of cell types into pure populations, and has recently been applied to screen cell surface display libraries against intact cells (Richman et al. 2006). In this method, a yeast-displayed scTCR library was incubated with a suspension of mammalian cells that expressed peptide-bound forms of the MHC protein on their surface. The yeast cell-mammalian cell conjugates were then separated from unbound yeast by centrifugation through a density gradient. Yeast-displaying scTCRs of interest were enriched 1000-fold after a single round of density centrifugation. This approach led to the isolation of scTCRs that were specific for either class I or II MHC proteins, and has important implications for protein engineering due to difficulties in obtaining soluble peptide-class II MHC ligands for library screening. Moreover, this screening methodology could potentially be applied to engineer other ligand-receptor interactions using a variety of cell surface display formats.

Despite the advantages just described, several important factors should be considered when panning cell surface display libraries against whole cells. As in magnetic sorting, cell panning lacks the rigorous quantitation provided by FACS, and does not allow discrimination of clones based on expression levels or avidity effects. Another concern is that mammalian cells express many more proteins besides the target on their surfaces; thus, there is a high likelihood of isolating clones that bind to other things. Therefore, negative screening steps should be incorporated whenever possible, such as depleting the library of clones that bind to similar mammalian cells that do not express the target. Moreover, this screening method is limited in that some membrane-bound targets may be inaccessible for binding to cell surface display library members, or may be expressed at levels that are too low for efficient recovery and propagation of binding clones.

Cell surface display libraries have also been panned against inorganic materials with the goal of discovering peptide motifs that are capable of binding to solid surfaces. These peptides could be used as novel affinity tags to immobilize proteins for biomaterials applications as an alternative to chemical conjugation, and would allow site-specific protein attachment for control of proper surface orientation. Early reports described the display of random peptide sequences within endogenous E. coli surface proteins for the identification of sequences that bind to materials such as iron oxide (Brown 1992), metallic gold and chromium (Brown 1997), and zinc oxide (Kjaergaard et al. 2000). More recently, the bacterial FliTrx cell surface display system has been used to identify disulfide-constrained dodecapeptides that bind semiconducting metal oxides (Thai et al. 2004). Due to their larger cell size compared to bacteria, yeast offer additional advantages for studying peptide-sur-face interactions through the ability to use conventional light microscopy in combination with techniques based on the application of mechanical force. Panning a yeast-displayed scFv library against CdS resulted in the identification of short scFv fragments, which were rare in the original library (Peelle et al. 2005a). Subsequent studies used yeast-displayed peptide libraries to pan against II-IV semiconductor and Au surfaces (Peelle et al. 2005b), as well as metal oxide surfaces (Krauland et al. 2007). Along with identifying peptides that bound to these materials, insight into the molecular interactions responsible for recognition was obtained through sequence-

activity analysis of the surface-binding peptides. Uncovering the molecular details in these systems will aid nanobiotechnology efforts, and provide a better understanding of natural biomineralization mechanisms.

Was this article helpful?

0 0

Post a comment