Editing Phage Display from Synthetic Library (section)

=== Discussion ===

When conducting ''in vitro'' antibody selections, the two primary determinants of selection success are 1) the isolation of well-folded antigenic targets for selecting upon and 2) the availability of a high functional diversity antibody library. In many cases, the availability of sufficient quantities of well-folded, full-length protein can be limiting. One approach to overcoming this limitation is to use domains identified from bioinformatic analysis<sup>22</sup> and synthesized for insertion into a bacterial expression vector with an appropriate affinity tag.

There are a variety of tags that are used in biotech applications to increase solubility, facilitate purification and to stabilize unstable domains.<sup>11</sup> The ability to synthesize virtually any desired insert sequence enables a highly versatile antigen generation platform. Although the GST-tagged format is common, successful results have been previously obtained using hexahistidine, Fc, Halo, maltose binding protein and site-specific biotinylation tags and it is believed that myriad other affinity tags may be optimized for use in a similar platform. However, affinity tags can have both beneficial and unintended negative consequences in downstream applications<sup>11</sup> and should be thoroughly investigated before establishing a pipeline based on a particular format.

Often new users will pose the question of how to assess the quality of an antibody library and though there is no straightforward answer, one should try to assess several key features (whether natural or synthetic) including overall diversity, relative display levels, the nature of and extent of randomization (i.e., library design versus library content determined by sequencing) and the physicochemical properties of the antibody versus intended application. Although beyond the scope of this protocol, a more detailed treatment of these features and how they can affect the selection of antibodies can be found here<sup>2,23</sup>. Knowledge of library diversity is, in particular, important to ensure adequate coverage of the diversity during selections. Typically, a phage concentration that provides 100–1,000 copies of each phage clone in the library is desirable and should ensure that any true positive binders present in the library can be recovered. However, non-specific binding of phage to microplate surfaces increases dramatically above 10<sup>13</sup> phage/ml, and for very diverse libraries, more than a single well per antigen may be needed to ensure sufficient coverage. On the other hand, if the input library is too dilute, the absolute concentration of true positive binders will be so far below the KD (typically nM) that they will not be recovered. Given these constraints, phage libraries should generally be resuspended to 10<sup>12-13</sup> phage/ml for the first round of selections. At this concentration, a 100 L aliquot of phage represents a 1,000–100 fold coverage of a 10<sup>9</sup>-10<sup>10</sup> diversity library and can routinely yield binding clones to diverse antigens. At lesser concentrations, or increased diversities, it may be necessary to increase the number of wells of antigen selected against in the first round (see Section 2.9). Subsequent to round 1, however, most of the non-binding diversity of the library has been removed; excess coverage of the round 1 output diversity is most often easily achievable using a single well. In cases of parallel selections in 96 well plates, this can reduce the number of antigen selection plates required to just one, subsequent to the first round.

During actual selections, phage titration can offer objective measures to characterize and monitor progress, while helping to identify areas of concern particularly in the early rounds of selection where enrichment is not sufficient to be detected. This can be particularly useful during scaling of selections to determine protocol performance. In general, input phage titres (''i.e.,'' naïve library or post-amplification of eluted phage) should remain relatively constant (10<sup>12</sup>–10<sup>13</sup> phage/ml) between rounds, with poor cell growth / low phage titres indicating potential contamination and/or poor output from the previous round. Although output phage titres are expected to increase during enrichment in later rounds of selection, a range of 10<sup>3</sup>–10<sup>5 </sup>cfu/ml in the first two rounds is expected. Deviation from this range can indicate potential issues with the selection such as contamination or excess wash stringency. In later rounds of selection (''i.e.,'' rounds 3 and 4), the enrichment of clones that specifically bind the target antigen can be assessed via pooled ELISAs that measure binding of the phage pool against both target antigen and negative control protein (See Section 4).

After 3-4 rounds of selection (or once enrichment on target protein has reached a target antigen:negative control protein signal ratio of 2:1 or higher in comparison to background protein), infected bacterial cells are plated to isolate single phagemid colonies for sequencing, initial characterization and cloning if necessary. At this time, it is highly advisable to convert Fab-phage to free Fab before additional characterization. Although Fab-phage can be used for initial characterization by diluting 10 µl of phage supernatant (described in Section 5.1) in 20 µl PBT buffer and detecting with anti-M13-HRP antibodies (described in Steps 4.4 - 4.5), in our experience, binding properties can change substantially in anywhere from 5-20% of clones upon liberation from the phage particle and early conversion can obviate issues with false positives. The specific methodology used for conversion will depend on the unique architecture of the phagemid and thus will not be dealt with explicitly here. However, in the most commonly used vectors, this can generally be accomplished by 1) transformation of purified phagemid (or phagemid pool) in to a competent non-suppressor ''E. coli'' strain such as HB2151 or 55244 if an amber stop (TAG) codon intervenes the Fab and pIII protein, 2) insertion of an amber stop codon between the Fab and pIII proteins to allow expression of soluble antibody fragments in a non-suppressor strain or 3) by cloning immunoglobulin genes (from individual phagemid or phagemid pool) in to a suitable expression vector. To further streamline selection and characterization, pooled cloning methods may offer effective means for converting binding clones ''en masse'' in to expression constructs, both eliminating the potential for false-positive binding of Fab-phage particles and where expression lysates are used for early characterization, even the cost and effort of purification can be initially avoided.

For clones isolated directly from Library F (described in<sup>24</sup>), antibody variable domains can be sequenced using 1-2 μl of phage supernatant as template to PCR amplify complementarity determining regions (CDRs) (See '''Table 2''' of Materials for Library F-directed primers). Using these primers, the heavy and light chain variable regions will yield products of ~700 and ~500 bp, respectively covering all three CDRs. Primers include annealing sites for M13 primers such that products can be sequenced after cleanup (as described in Section 5.4.5) using standard primers available in most sequencing core facilities. Alternately, for Fab-phage pools that have been cloned in to an expression vector and transformed directly in to ''E. coli'' for expression, single colonies can be isolated and resuspended in 15 µl of 2YT and 1-2 µl of the cell suspension used directly as template for single colony PCR using appropriate primers.

Automating selections will require optimization and validation of several key robotic functions. In general, pipette-based liquid handling must be accurate and consistent across all 96 channels, plate washing must remove unbound phage efficiently without stripping adsorbed antigen or bound Fab-phage from the selection plate. Effective decontamination of the plate washer is also necessary between rounds to allow enrichment and effective counter-selection, and between subsequent selections to prevent cross-contamination. To help optimize the high-throughput selection procedure, simple approaches for evaluating these functions have been used previously and are described below and detailed within the protocol.

To evaluate the consistency of volumes delivered by all channels of a liquid dispensing / aspiration head, a colored solution is pipetted to a flat-bottom 96 well plate and the absorbance in each well measured on a plate reader. A chromophore such as bromophenol blue is helpful for dyeing solutions similar to those used for real selections to evaluate the effects of different surface tension and cohesion properties on volumes delivered. After confirming that all channels are dispensing consistent volumes, accuracy of the total delivered volume can be determined by measuring the mass of liquid transferred to each plate.

Automation of plate washing to remove unbound clones primarily requires optimization of the rate at which wash buffer is to be dispensed and the number of washes to be employed. A reasonable method of evaluating this parameter employs positive controls (specific Fab-phage binding clones) to assess the effects of varying the dispense rate and/or number of washes to determine whether 1) adsorbed antigen or bound phage is being removed by overly stringent washing or 2) non-specific binding to non-cognate proteins is excessive due to wash conditions that are not sufficiently stringent. Residual/bound phage can be detected and quantified either by infection into bacteria and plating for single colony counts, or by ELISA using secondary antibodies specific for antibody affinity tags or phage coat proteins. Negative controls are used to assess residual levels of non-binding Fab-phage to target proteins or specific binding Fab-phage to non-cognate/background proteins, which may be high if washing is not sufficiently stringent. In this case, we used immobilized BSA as a control for non-specific binding and an anti-FLAG antibody that recognizes a FLAG epitope tag on the Fab-phage as a positive control. We compared eight and sixteen wash cycles using the robot versus washing by hand, at a medium flow rate, to show that these techniques were equivalent. Alternately, measuring input and output phage titres (See Section 7) during ongoing selections can also help to make fine adjustments in the wash parameters.

Finally, decontamination of plate washers used for selections must be effective in order to eliminate carry-over of phage from prior selections. In our experience, freshly diluted 0.5% sodium hypochlorite is fast, effective and cheap. Detergents such as SDS are also effective but require much more extensive washing to remove from the system. Check reagent compatibility for the system prior to any tests. To assess decontamination, we test for the presence of residual phage in solutions handled by the fluidics system and interpret results according to '''Table 3'''.

In summary, this protocol provides a scalable method for isolating Fabs from combinatorial libraries by parallel ''in vitro'' selections against protein domains and offers techniques for validating an automated phage selection system. Cost and infrastructure barriers that extensive animal facilities can pose are mitigated since this method does not rely upon the immunization of animals. Further, by integrating antigen generation with antibody-phage selection, quality factors that affect selection success are more easily monitored and adjusted. When the time frame from antigen expression to selecting and initial characterization is on the order of 6-8 weeks, a more responsive system of production may be realized. Antibodies isolated from''in vitro'' selections are also both easily modified (due to genotype-phenotype linkage) and renewable, requiring only stored DNA of the expression construct to repeatedly and reproducibly re-generate antibody. Finally by showing that this protocol can be conducted using either a quasi-manual or fully automated approach that employs a custom-designed, robotic selection system, throughput can be scalable and accessible to small academic and industrial labs alike.

Using the high-throughput methods described above and the Library F synthetic antibody library<sup>24</sup>, high affinity (low nM) and specificity Fab antibody fragments can be routinely obtained and are functional in a variety of immuno-applications including Western blotting, immunoprecipitation and immunofluorescence of both''in vitro''- and''in situ''-expressed proteins. Although success rates for any given selection will depend both on the quality (purity, foldedness, mono-dispersity etc.) of the antigen targets as well as the quality and diversity of the antibody library, it has been observed that anywhere from 25-80% of antigens will yield binders for a high throughput selection. Importantly, it is worth noting that clones with the ability to modulate target function can also often be selected. This library is constructed using a fully human framework, which makes these antibodies amenable to potential therapeutic application<sup>25</sup>, thus underscoring the importance of this pipeline as an alternate approach for the generation of affinity reagents with broad-based value.

In summary, the robustness and versatility of the approach presented in this protocol will continue to aid in the optimization, industrialization and ultimate widespread use of this technology as a viable means of achieving the long-term goal of generating affinity reagents to virtually all members of the human proteome.

Subscription Required. Please recommend JoVE to your librarian.