|Home | About | Journals | Submit | Contact Us | Français|
A set of phage display sorting strategies and validation methodologies are presented that are capable of producing high performance synthetic antibodies (sABs) with customized properties. Exquisite control of antigen and conditions during the phage display selection process can yield sABs that: 1) recognize conformational states, 2) target specific regions of the surface of a protein, 3) induce conformational changes, and 4) capture and stabilize multiprotein complexes. These unique capabilities open myriad opportunities to study complex macromolecular processes inaccessible to traditional affinity reagent technology. We present detailed protocols for de novo isolation of binders, as well as examples of downstream biophysical characterization. The methods described are generalizable and can be adapted to other in vitro direct evolution approaches based on yeast or mRNA display.
Over the past decade, the accumulation of detailed knowledge of antibody structure and function has enabled phage-displayed antibody technology to emerge as a powerful in vitro alternative to traditional hybridoma methods [1–4]. In this context, antibody fragments are expressed and displayed on the phage surface as fusions to a coat protein. Phage pools containing billions of unique antibodies (~1010) are used in affinity selections (Fig. 1A) to isolate antibody variants against the antigens of interest, and the sequence of each antibody can be decoded by sequencing of the viral DNA of the isolated phages (Fig. 1B). Thus, the process directly links the genotype with its phenotype since the selection process not only yields functional antibodies, but also DNA sequences from which they are produced in bacteria. Combining rational in silico library design, modern protein engineering tools and sophisticated downstream characterization methods has resulted in generating customized antibodies that recognize conformations or discrete states of target molecules [5–7].
Below is discussed how combining in silico designed libraries with exquisitely controlled selection conditions can generate a class of high performance affinity reagents that can be exploited in myriad of biological and biophysical applications.
The antigen recognition domains of antibodies contain six loops that constitute the complementary determining regions (CDRs). The CDRs are the source of the sequence diversity that is central to their antigen binding properties; three of these loops are contained in the light chain (L1, L3, L3) and three in the heavy chain (H1, H2, H3). Although there are a number of antibody-based phage-display libraries available, a particularly powerful set of synthetic antibody (sAB) libraries is based on a novel “reduced genetic code” concept. Such libraries provide a way for a large number of positions in the CDR loops to be diversified compared to traditional phage-display libraries with no loss in function [8–11]. Another distinction is that antibodies from natural sources contain a considerable level of sequence diversity in their framework, making them each a unique scaffold [12,13], so it is difficult to predict how well they will express or what their stability will be.
Synthetic antibody libraries have been constructed using a scaffold from a humanized Fab 4D5 fragment engineered for high stability and good phage display (Fig. 1B) [8,14]. This allows modular design of sAB libraries where the scaffold can be enhanced for stability, expression or even efficient crystal lattice formation. Further, the antigen-binding interface can be maximized for binding potency step by step, similarly to “synthetic” approaches exploited by chemists. Such systematic and modular optimization is much more difficult with monoclonal antibodies and antibody phage-display libraries derived from natural immune repertoires. Natural repertoires are limited in generation of antibodies against self-antigens, but synthetic antibodies are constructed entirely in vitro and thus, are not biased against natural proteins, regardless of source or sequence. For therapeutic applications, optimized human frameworks can be used to minimize immunogenicity, thus obviating the need for humanization. Furthermore, design features can be incorporated to allow facile affinity maturation and adaptation to a high throughput format. For comprehensive review of synthetic antibodies please see Miersch S. et al. .
Economical production and permanent storage of DNA clones are some of the advantages of the sAB approach. Monoclonal antibodies can be reproduced, but the maintenance and large-scale culture of hybridoma cells are cumbersome and expensive. Properties of a monoclonal antibody from the same hybridoma line can gradually change due to the clonal drift . The requirement of sacrificing animals in large animal facility is controversial and in many cases wasteful. In contrast, sABs can be readily produced in E. coli and they can be stored economically in the form of an expression vector. Because the amino acid sequences of all sABs can be easily determined by DNA sequencing, any sAB can be reproduced even in the event that the expression clones are lost. Furthermore, the use of a single stable antibody fragment makes it straightforward to reformat a sAB into a full length IgG construct or a single chain Fv.
An important attribute of the sAB phage display approach is the ability to design selection strategies to generate antibodies with customized functions [6,7], which can be classified based on activity or mode of binding. For instance, it is possible to generate sABs that: 1) preferentially recognize a specific conformational state and thus, have the potential to induce a specified conformational change ; 2) target specific regions of the surface of the target protein (“regio-specific”) ; 3) specifically recognize multi-protein complexes (unpublished results), and 4) capture and stabilize weak protein-protein interactions (unpublished results) (Fig. 2).
sABs can be used in essentially any application where monoclonal antibodies are commonly used: western blots, ELISA and related plate based immunoassays, immunostaining, immunoprecipitation and chromatin immunoprecipitation (ChIP). However, a higher level of power is in their ability to perform more sophisticated functions that utilize the custom designed properties to probe both the static and dynamic features of a protein system [18,19]. Importantly, selection procedures can be designed to generate multiple unique sABs that target a single protein or its ligand bound state, a multi-protein complex, a nucleic acidprotein complex or folded RNAs [6,7,20]. Further, sAB scaffolds with a known structure can be used as effective crystallization chaperones of RNAs and proteins inherently recalcitrant to generate stable crystal lattices [17,20,21]. Moreover, it has been shown that once diffraction quality crystals are produced, the sAB can be used as a molecular replacement model to obtain the initial phasing information speeding up the process of structure solution.
sABs with desired properties can be obtained from two different types of selection: i) negative or ii) positive. The choice of type of selection is mainly dictated by the properties of the system under study and the desired outcome, since subtraction and competition schemes lead to different outputs.
The basis for subtractive selection is presented in Fig. 3A where negative and positive selections are applied in a sequential manner. First the library is “pre-cleared” to remove unwanted binders by incubating the phage library with the unwanted protein component immobilized on solid support (bead or plate). Most phage binders to the unwanted component are captured and effectively eliminated from the library. This pre-cleared library then is used for positive selection with the primary target without any further modification of the protocol.
In the competitive selection strategy, negative and positive selections are performed in a single step (Fig. 3B). The competitor (unwanted component) is introduced in a soluble form in large excess to the mixture of an affinity-tagged target antigen and the phage-sAB library. This effectively “sponges up” all binders to the competitor that might cross-react with the desired target antigen. The affinity-tagged target is then pulled down using magnetic beads coated with a capture agent, bringing along only target-specific sAB-phages. Because the competitor is not tagged, it and any of the sABs bound to it are removed in the wash steps. The competition selection scheme usually produces higher numbers of clones due to smaller number of steps required to complete the experiment and it is generally better suited to systems with clear separation of conformations or where high affinity components are used.
For competitive selections to work, it is essential that the “competitor” closely mimic the property that is being selected against. For discriminating the apo- form from the ligand-bound form of a protein, the competitor is the ligand-bound form of the protein added in solution with the apo- form immobilized on the solid surface, as described above. The strategy is reversed to generate sABs to the ligand-bound form (Fig. 3). The competition selection strategy can be picked only if dissociation of ligand is negligible on the time scale of the experiment.
The strategy to generate sABs that bind exclusively to a molecular complex comprising of targets A and B is very similar. Target A is immobilized on a magnetic bead and then a saturating concentration of target B is added. This ensures the formation of the A-B complex, even if the affinity is low, while the sABs binding to uncomplexed target B are not captured.
For generating regio-specific sABs several strategies can be used. If the structure is known, strategic mutations or truncations can be made and used as competitors to direct binding to a specific region on the protein surface. If the protein target has a known binding partner, there are two possible outcomes. When the partner is added in excess at the end of the selection round, it will compete off sABs that share its epitope, while leaving the targetbound sABs that bind to other epitopes. Thus, the sABs can be categorized as either sharing or having an independent binding epitope.
For phage display selections, we use synthetic Fab fragment libraries built on a single antibody fragment framework and standard phage display generation methods [8,22,23]. In these libraries, limited amino acid diversity is introduced into the 3 heavy chain CDRs: H1, H2, and H3 as well as the third CDR of the light chain: L3 (Fig. 1B). The combined naïve library contains >1010 unique clones. Alternative scaffolds and libraries with similar designs can also be used in the approach described here. Nevertheless quality of the library will strongly influence selection process and is essential to success of antibody generation.
This section describes the concepts and detailed protocols for isolation of synthetic antibodies from combinatorial phage-display libraries. Protocols for phage library preparation and subsequent affinity maturation have been previously described [8,22,24]. Below we focus on experimental design, phage display selection schemes, methodology for sAB expression, characterization and troubleshooting.
Fig. 4 illustrates the workflow involved in the sAB generation and validation pipeline. Prior to the phage display selection process (Section 6.2), the target has to meet a set of quality control (QC) criteria; in particular, homogeneity, stability and efficient bead capture. Once the QC is passed, the targets are ready to enter the phage-display sorting pipeline (Section 6.3) following the scheme set out in Section 2. If the selection process is deemed successful, based on increased round to round enrichment of specifically binding sABphages, primary validation can commence (Section 3.2). The goal of primary validation is to provide a rapid high-throughput test to distinguish good clones and triage poor ones. At this point, additional characterization can be performed to rank the relative performance of individual clones in designated high-end applications. The most promising clones are carried on to secondary validation in which their quality and versatility are tested to a much more stringent level (secondary screening - Section 3.3).
Effective selection relies on a specific capture and release process utilizing an affinity tag to maximize solution capture of phage displaying the desired sABs. It is crucial that the tag binds to the capture beads at an extremely high affinity so as to withstand the rigorous washing steps that are essential for removing nonspecific binders present in the library. One of the most generic affinity tags relies on chemical modification of amines or exposed Cys residues with biotin-derivatized reagents with linkers that can be cleaved by reducing agents (Section 6.1) . Random surface biotinylation allows accessing a higher number of epitopes by not restricting the orientation of molecule during the selection and screening process. Chemical biotinylation is the method of choice for proteins that are obtained from natural sources due to lack of recombinant protein or the presence of post translational modifications (PTM). Thus, if the protein sample is stable and readily available this method should be used as the first approach. Biotinylated protein samples have to be rigorously tested for the level of biotinylation in pull-down assay described in Section 6.2 (example on Fig. 5A).
A downside to this approach is that many proteins, eukaryotic ones in particular, are not stable enough to withstand chemical biotinylation or their solubility can be compromised due to introduction of hydrophobic biotin moiety. In select cases biotinylation can impair protein function inadvertently affecting its conformation. An alternative approach relies on in vivo or in vitro biotinylation by the biotin ligase BirA of an AVI-tag fused to the protein of interest . This method is recommended when working in a high throughput environment and has some advantages over chemical modifications. In either case, it is critically important to determine the state of the target protein before library sorting. Denatured and/or aggregated protein samples are the principal problem for why sAB generation fails. Thus, after biotinylation it is important to test for physical state and function. For soluble proteins, one-dimensional 1H NMR spectra can be useful to establish the overall quality of the sample. For enzymes, target validation should include enzymatic activity assays. Determining the oligomeric state by dynamic light scattering in combination with size exclusion chromatography is also an important analytical metric.
Phage library sorting (Section 6.3) is conducted based on the selection strategy chosen as outlined in Section 2. The process combines both automated and manual steps and usually takes several days depending on the number of sorting rounds required. The steps can be readily adapted to high-throughput formats that can greatly expand the scope of the selections. Not only can many more targets be included simultaneously, but variants of the targets and conditions can be screened side by side allowing for direct comparisons of subtle, but important differences that can lead to dramatic changes in the outcomes of the selections.
The process utilizes a KingFisher magnetic bead handler and involves the following steps: 1) capturing sAB-displaying phages that bind to the target which is immobilized on magnetic beads, 2) vigorous washing to eliminate non-binding phages, and 3) release of the captured phages by cleaving the disulfide linkage within the biotinylation reagent  (when a non-cleavable biotinylation reagent is used, phages are released with an acidic solution). The progress of selection can be monitored by phage titering to track the enrichment of targetbinding phages round to round relative to background (Fig. 5B). Phage titering provides simplest diagnostic readout where count of phage particles expressed in colony forming units or plaque forming units can be assessed and compared to the relevant controls (Section 4). Recovered phages are amplified in E. coli and the sorting cycle is repeated. The selective pressure that favors the tightest binders is produced by the successive reduction of the concentration of the target protein usually starting with 100 nM in the first round down to 10 nM in the final round. However, for difficult targets, 10 nM may be too stringent, requiring readjusting the initial and final target concentrations.
Typically, three rounds of library sorting are sufficient to generate a good repertoire of sABs. In cases where few binding clones are isolated in round 3, additional rounds of library sorting can be performed. The initial step to identify binders with desired properties usually employs standard competitive phage ELISA (Section 6.4) modified to include compounds or partner molecules present in the selection process. 96 individual (“monoclonal”) phages are tested. In some cases fewer clones can be picked, but with a concomitant loss in quality of the final pool. Usually single point competitive ELISA is performed with 20 nM competitor. This allows selection of clones with KD values of 20 nM or lower. In addition, the long ELISA incubations and multiple wash steps favor clones with the low koff values. Clones for secondary validation are chosen based on several different parameters derived from ELISA data: i) overall strength of ELISA signal (indicating affinity and display level), ii) ability to compete with WT protein (competition ratio - fraction of the signal retained upon protein coating the ELISA plate competing for sAB-phage binding sites with WT non-biotinylated protein added in solution; indicating affinity), and iii) combination of both these values measured for each competitor used in selection (Fig. 5C).
Clones that pass an established threshold are then analyzed by DNA sequencing to determine their uniqueness (Section 6.5). Determining their CDR H3 sequences is usually sufficient for this purpose because this loop is most diverse and should contain unique sequence signatures. If the project involves analyzing data from many targets, for instance in high throughput environments, then dedicated software linked to the database can be used, especially if closely related targets are screened and duplicate sequences need to be pruned across the data sets.
Phage displayed sABs obtained in primary screening need to be converted to the format that will allow efficient expression of protein in soluble form. This step is essential to ensure that properties of sABs observed in primary screening also translate to similar properties in protein format later used in all downstream applications. This task can be achieved either by introduction of a stop codon between sAB and phage protein P3 by the means of mutagenesis (Section 6.5) or cloning of the sAB fragment into the bacterial expression vector. To establish the performance of sAB clones several methods provide simple, but robust evaluation of the level of performance as customized affinity reagents. In most of the cases, sABs can be validated with methods similar to those routinely used for monoclonal antibodies .
Most of the initial validation steps can be performed using sABs expressed at small scale (Section 6.6) utilizing the phoA promoter system also used in phage display (Fig. 1B). However, full secondary validation usually requires large scale production of sABs (Section 6.7). Both small and large scale methods follow similar steps employing Protein A affinity purification yielding sABs of a least 95 % purity. Yields vary typically from 5 to 20 mg of pure sAB from 1 L of media. Large scale purification can be readily scaled up to 12 L for applications requiring a large amount of antibody. sABs can also be subcloned into other dedicated vectors to introduce additional tags aiding the validation process or even to convert them into a full IgG format.
The principal set of techniques relies on various biophysical methods, some of which are described below.
There are many applications where it is advantageous to have a collection of sABs that bind to non-redundant epitopes. Most of the methods in use look to establish whether two sABs can bind simultaneously to the antigen. sABs that recognize non-overlapping epitopes are linked in pairs. These pairs can subsequently be used to group sABs displaying similar binding profiles into bins. The sAB pipeline utilizes single point competition ELISA screening, as described above. Each individual phage-sAB is screened against every other sAB that is used as a competitor following the standard competitive ELISA protocol. An alternative approach is epitope binning using SPR. This experiment provides much higher fidelity of detection although it is more time consuming. For the example and description of the method see .
The performance of phage display selection is usually characterized by an enrichment parameter defined as the ratio of the phage count recovered from the target sorting compared to the count of background binding phages in the negative control. Significant enrichment after round 3 and 4 of the library sorting step combined with the results of primary screening can generally gauge the performance of the sAB generation process. As a general rule, the expected enrichment after round 3 should fall within 2 – 100 fold and 10 – 1000 after round 4 (Fig. 5B). Lower or no enrichment, especially in selections where a competitor was not used, indicate poor selection performance which in turn drastically affects success rates of obtaining sABs for the desired conformation, complex or region of the molecule.
The follow up primary screening is performed using a single point competitive ELISA and involves analysis of 96 clones. A successful experiment typically yields 25 – 80 % positive clones in this assay. Because the goal is to generate sABs that perform quite sophisticated functions, there are usually fewer clones that meet the stringent criteria compared to standard phage display selections that capture all binders. Depending on the set thresholds, a successful secondary assay usually yields 10 – 50 % of clones with the desired properties (Fig. 5C).
Antigen quality is a principal source of problems in selection. Prior knowledge of the biochemical properties of the target is necessary, especially if the selection is performed on an unstable protein or a protein that requires cofactors to maintain its activity. Cysteine-rich proteins can be problematic, as well as proteins with metal binding sites, especially if metals undergo electrochemical reactions with paramagnetic particles such as those used during the selection. It is absolutely essential that the selection should include a positive control using a “standard” target that has known characteristics and selection output. Problems in the selection can be detected earlier in the process if enrichment levels can be compared with the control. If several of the targets show substantially lower levels of enrichment than the control, it can be a signal that those targets are problematic and their preparation or handling has to be re-evaluated; however, there were instances where we were able to obtain good sABs where the enrichment was low.
For problematic targets, there is often a trade-off between affinity and specificity. For instance, a sAB may have good selectivity to a particular conformation, but the affinity may be too low to be useful. As a general rule, it is advised to perform another selection with optimized stringency to obtain binders with the desired characteristic directly from the affinity selection; however, low affinity clones can be “rescued” by performing an affinity maturation step. Since the overall structure and sequence of each isolated sAB is known, this process is straightforward. A library can be designed to introduce a higher level of diversity and several additional cycles of sorting can be run.
Besides affinity issues, there are other possible sources of low recovery of usable clones. A potential troublesome area is the efficiency of the pull-down steps (Fig. 5A) resulting from a non-optimal level of biotinylation. Both too little and too much biotinylation have undesirable consequences. This is where in vitro or in vivo biotinylation in bacteria during expression can provide an advantage. In vivo biotinylation provides high throughput means for biotinylation of proteins utilizing BirA enzyme coexpressed in the same cells . There are also cases where enrichment is high, but the sequence diversity of captured clones is low and dominated by one or a few clones that most likely have expression advantage over the rest of the clones in the pool. This can happen for difficult targets where the pool of “winners” is inherently small and more rounds (>3) were required to pull them out. Ideally clones should be characterized on earliest round possible to not compromise diversity of the pool.
Protein mutants or compounds that stabilize a particular conformational state might unintentionally interfere with the selection process. In such cases, further investigation is required to find a better mutant or compound. One might consider switching to a subtractive selection scheme to achieve a better separation of states. We have successfully used different compounds during screening, mainly small molecule drugs. Usually compounds that are stable and soluble in aqueous solutions do not pose significant problems in phage selections. Low affinity compounds should be avoided and possibly replaced with higher affinity variants if available. The use of organic solvents like DMSO at concentrations higher than 5 % is not recommended, especially if the impact on the target protein cannot be tested. Also, the use of compounds for the phage elution (Section 6.3) should be treated with caution, since bacteriostatic activity might affect phage propagation in the bacterial host. Additionally low affinity compounds might be ineffective in eluting sABs with sub nM affinity. In the case of membrane proteins many commonly used detergents can be used above their CMC without significant loss of infectivity of phage.
This recombinant system has multiple advantages over the traditional monoclonal based approaches. Nevertheless there are several types of protein systems that inherently produce fewer high affinity sAB clones.
We have devised strategies and generated extensive proof of concept data demonstrating the capabilities of a high throughput phage-display pipeline to produce multipurpose synthetic antibodies to challenging protein systems. The pipeline includes all steps from target preparation, to phage display selections and several levels of validation. The pipeline processes are organized to allow most steps to be automated and run in a high throughput mode. The duty cycle of all operations takes about two weeks, start to finish and can be multiplexed to potentially accommodate over one hundred targets at a time. However, the protocols are equally applicable to running in a mode where researchers have only one, or a few targets to screen.
We would like to thank Sachdev Sidhu, Frederic Fellouse and Brett Welch for many helpful discussions. We thank Robert Hoey, Vincent Lu and Michal Olszewski for their technical expertise. This work was supported by National Institutes of Health Grants: U01-GM094588, R01-GM072688, U54-GM087519, U54-HG006436.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.