Protein Engineering, Design & Selection vol. 27 no. 2 pp. 41– 47, 2014 Published online January 8, 2014 doi:10.1093/protein/gzt060
A general strategy for antibody library screening via conversion of transient target binding into permanent reporter deposition Alexander Maaß1†, Tim Heiseler1†, Franziska Maaß1, Janine Fritz1, Thomas Hofmeyer1, Bernhard Glotzbach1, Stefan Becker2 and Harald Kolmar1,3 1
Institute for Organic Chemistry and Biochemistry, Technische Universita¨t Darmstadt, D-64287 Darmstadt, Germany and 2Merck Serono, Protein Engineering and Antibody Technologies, D-64293 Darmstadt, Germany
3
To whom correspondence should be addressed. E-mail: [email protected]
Edited by: Arne Skerra
We report here a generally applicable method for the selective covalent attachment of a reporter molecule to a replicating entity that allows one to obtain specific binders from a single round of library screening. We show that selective biotinylation of phage particles displaying a binder to any given target can be achieved by application of a coupled enzyme reaction on the surface of the target-binding phage particles that includes a peroxidase, an oxidase and a catalase. Due to the covalent linkage of biotin together with the tight and stable interaction of biotin with streptavidin, very stringent wash conditions for removal of nonspecific binders can be applied. The method termed 3CARD (triple catalytic reporter deposition) was successfully applied to singleround screening of a phage display library of camelid singledomain antibodies against three different target proteins. Keywords: antibody engineering/catalysed reporter deposition/peroxidase/phage display screening/oxidase
Introduction The phage display technology was originally introduced by George Smith more than 25 years ago (Smith, 1985) and has become an indispensible tool in molecular biology and molecular medicine to identify peptides and particularly antibody molecules with predefined binding characteristics from large libraries. Since then, many improvements have been reported and screening concepts are now available to isolate binding molecules against virtually any target protein (Bradbury et al., 2011). Moreover, in recent years bacterial and particularly yeast surface display have become increasingly popular alternatives for phage display library screening applications (Pepper et al., 2008; van Bloois et al., 2011). Library screening generally involves exposure of a population of binding molecules on the surface of a replicating entity and contacting them with an interaction partner of interest such as an antigen molecule, in the case of antibody display. †
These authors contributed equally to this work.
# The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
41
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Received August 15, 2013; revised October 31, 2013; accepted November 21, 2013
The experimental setup is designed such that those phage particles or microbial cells displaying a target-specific variant can be isolated from the population by methods such as biopanning or cell sorting. Ideally, a single round of screening would be required to obtain the desired protein variants. However, the enrichment of specific binders is limited by non-specific binding that requires recursive rounds of screening and amplification to obtain the best binders from the library (Vodnik et al., 2011). The first screening round is particularly critical, since it can be expected for large libraries that specific binders constitute only a tiny fraction of the initial population such that non-specific binders largely dominate over those that interact specifically. As a consequence, conditions have to be carefully controlled and optimized for every single target and screening strategies devised that are stringent enough to remove nonspecific binders but sufficiently mild to keep the target binding of specifically interacting library members intact (Liu et al., 2009). Since the probability of obtaining high-affinity binders increases with library size (Steiner et al., 2008), progressively larger libraries have been established that can exceed 1010 variants (Rothe et al., 2008; Brockmann et al., 2011). However, repeated amplification of libraries as it occurs in each screening round is known to reduce diversity and limits the number of candidates that can be identified in a library screening experiment (Derda et al., 2011). As a consequence, it is highly desirable to reduce the number of screening rounds to a minimum— ideally a single round. Conceptually, this can be achieved by setting up a method where the transient and relatively weak interaction of target and library member translates into covalent attachment of a labeling molecule that can be used as a purification handle and allows for application of more stringent conditions to remove non-specifically bound library members. Catalysed reporter deposition (CARD) has been widely used in immune histochemistry for protein biotinylation (Bobrow et al., 1989, 1991, 1992). It relies on the horseradish peroxidase (HRP)-mediated formation of a biotin tyramide radical that readily reacts with proteins in close proximity, resulting in the formation of a covalent linkage between biotin and tyrosine residues on proteins (Fig. 1). To establish a selective biotinylation of phage or cell surface display library members which are bound to a target molecule, we make us of the fact that HRP-mediated biotinylation requires the presence of hydrogen peroxide (H2O2). The 3CARD system described here combines the activity of three different enzymes, namely a peroxidase, an oxidase, and a catalase. First, HRP is coupled onto all members of a starting population, which is demonstrated here in the experimental setting of a phage display library (Fig. 1a). Phage biotinylation by HRP requires presence of biotin-S-S-tyramide and H2O2. In the 3CARD setup, H2O2 is generated by a second enzyme that is conjugated to the antigen of interest (Fig. 1b and c). Sugar oxidases such as glucose oxidase catalyse the oxidation of sugars by electron transfer to oxygen, which results in the formation of H2O2. H2O2
A.Maaß et al.
is generated at higher concentrations on those phage particles to which the antigen–oxidase complex is bound since the reaction mixture contains as a third enzyme catalase, which very efficiently breaks down H2O2. As a consequence, H2O2 transfer to HRP and subsequent phage biotinylation should only occur for antigen-bound phage that has the oxidase in close proximity to HRP. Specifically labeled phages can then be rescued by biopanning to immobilized streptavidin and subjected to very stringent wash conditions prior to infection of bacteria. Here we describe the application of 3CARD for singledomain antibody library screening and present straightforward methods for coupling HRP onto the phage surface and conjugating oxidase with target protein such that a combined enzyme reaction on phagemid virions results in their selective biotinylation and isolation in a single round of screening. Materials and methods
Sodium periodate oxidation of HRP and GOX 18.8 mg Sodium periodate (87.8 mmol, Sigma-Aldrich) were diluted in 1 ml H2O. The protein to be oxidized, either HRP (Sigma-Aldrich, Typ VI-A) or glucose oxidase, respectively, was diluted in H2O to a final concentration of 10 mg/ml and 200 ml protein solution was mixed with 20 ml of the sodium periodate stock solution. After 20 min incubation at room temperature (RT) in the dark, the reaction mixture was purified with a PD10 column (GE Healthcare) pre-equilibrated with 100 mM sodium borate buffer pH 9.1. Subsequently, 1.4 ml of the eluate containing the oxidized protein was filled up to 2 ml with sodium borate buffer and stored in aliquots at 2208C. The protein concentration in the eluted fractions was determined spectrophotometrically at 403 nm assuming a molar extinction coefficient of 100 for HRP and at 280 nm for GOX assuming a molar extinction coefficient of 16.7.
Target protein coupling to glucose oxidase 0.1 mg Of periodate oxidized glucose oxidase at a concentration of 1 mg/ml (6 mM; Sigma Type VI-A) was incubated with an equimolar amount or less of the target protein in 42
200 ml phosphate buffered saline (PBS) and incubated at room temperature in the dark for 180 min, followed by gel filtration chromatography using Superdex 200 pg 16/60 column (GE Healthcare) and collection of the early eluting fractions (Supplementary Fig. S1). The protein concentration in the eluted fractions was determined spectrophotometrically at 280 nm assuming an 1 : 1 ratio of enzyme to target and taking into account the molar extinction coefficient of GOX (16.7) and of the respective conjugated protein. Target proteins LipH and matriptase-1 were purified according to published procedures (Becker et al., 2007; Avrutina et al., 2012). Cetuximab was provided by Merck Serono.
Phage library propagation The pAK200-VHH library consists of 108 variants that were generated by randomization of CDR1 (5 residues), CDR2 (8 residues), and CDR3 with loop variegation from 7 to 17 residues using a NNS coding scheme. pAK200 mediates chloramophenicol resistance and contains the coding sequence for a PelB leader peptide followed by the VHH gene fused to shortened pIII. The gene fusion is expressed under lac promoter control. Due to the expected high occurrence of amber stop codons within the randomized loop regions, no additional stop codon was introduced between pIII and the fused VHH coding sequence. Each phage particle displays on average not more than one VHH fragment due to amber codon suppression in Escherichia coli ER2738. The phage library was generated by infection of 1011 pAK200 containing E. coli cells using 1013 M13VCS helper phage particles.
HRP coupling to phagemid virions Prior to HRP coupling, streptavidin binders were removed from the phage library. To achieve this, 100 ml pAK200-VHH phagemids (2.1011 phagemid virions) were incubated with 10 ml Streptavidin-Dynabeads (T1, Invitrogen) for 30 min at RT in a shaker. After separation of the beads using a magnetic particle concentrator, the supernatant was transferred into a 1.5-ml reaction tube and 100 ml oxidized HRP (1 mg/ml in
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 1 General outline of the 3CARD phage display screening procedure. (a) Covalent deposition of HRP to the surface of phage particles that display a library of antibody variants. (b) Binding of phage displayed antibody (AB) to protein target (P) which is coupled to glucose oxidase (GOX). (c) GOX-mediated generation of H2O2. (d) Catalase (CAT) breakdown of excess H2O2. (e) Activation of HRP mediated by H2O2. (f,g) Formation of a biotin tyramide radical by HRP and conjugation to the phage surface. (h) Biopanning of biotinylated phage using to streptavidin-coated magnetic beads.
General strategy for antibody library screening
100 mM borate buffer pH 9.1) was added. The reaction mixture was incubated at room temperature in the dark under slow agitation for 45 min. To remove uncoupled HRP, 200 ml phage precipitation buffer (20% PEG 6000, 1.5 M NaCl) was added, followed by incubation on ice for 45 min. After centrifugation in a table top centrifuge (20 min, 13.000 rpm), the phage pellet was dissolved in 50 ml PBS.
collector was used to separate beads from the supernatant. The beads were washed 10 times with 1 ml PBS-Tween (0.05%) buffer. Finally, beads were collected and resuspended in 200 ml 0.2 M glycine-HCl pH 2.2 and incubated for 8 min to release phage particles form the beads followed by magnetic separation and neutralization by addition of 1 ml 1 M Tris-HCl pH 9.0. Infection of E. coli cells was performed as described above.
Coupled enzyme reaction on phage
Conventional phage display screening Target protein biotinylation was performed as described (O’Brien and Aitken, 2002) using Sulfo-NHS-SS-biotin (Thermo Scientific) followed by removal of unbound biotin by gel filtration using a PD10 column (GE Healthcare). Biotinylation was verified using the HABA-avidin assay (Green, 1970, 1975; Janolino et al., 1996). To 2 1011 phagemid virions, biotinylated target protein was added at a final concentration of 4.5 nM in a final volume of 200 ml. After 20-min incubation, 20 ml streptavidin coupled magnetic beads (Dynabeads T1, Invitrogen) were added to the sample and incubated by agitation for 30 min. A magnetic particle
Results For the phage display screening procedure described above to be functional, it had to be shown that a coupled reaction of an oxidase and peroxidase can be performed on phage particles using an experimental setting, where the oxidase is coupled to the phage surface via antigen/antibody interaction and HRP is chemically conjugated to the phage surface such that the H2O2 generated leads to selective labeling of the bound phage and thereby to antigen –antibody– phage linkage. In addition, a simple method for conjugation of HRP to phage and oxidase to the target protein of interest had to be established. To address these points, a model experiment was performed that relies on the phage display of a camelid VHH single-chain antibody fragment that was obtained by conventional phage display library screening and selectively binds Pseudomonas aeruginosa LipH chaperon (Wilhelm et al., 2007) with an Kd of 20 nM. Presentation of the anti-LipH VHH domain on phage was achieved by genetic fusion of the VHH fragment to truncated pIII using phagemid display vector pAK200 (Habicht et al., 2007).
Conjugation of oxidase to target protein For conjugation the commercially available glucose oxidase was used. Since this enzyme is glycosylated, protein conjugation with the target protein can easily be achieved by periodate oxidation of terminal sugar moieties followed by Schiff base formation of the resulting aldehyde with a lysine residue residing on the surface of the target protein (Hermanson, 1996). For this purpose, oxidized glucose oxidase was incubated with target protein and the fusion was purified from unconjugated oxidase by gel filtration chromatography (Supplementary Fig. S1).
Coupled enzyme reaction on phage To investigate whether HRP can be coupled to the surface of filamentous phage without negative interference with antigen binding, phage particles displaying anti-LipH VHH and control phages displaying an unrelated peptide were incubated with oxidized HRP and purified by severalfold precipitation using polyethylene glycol. Wells of a microtiter plate were coated with LipH target protein. Binding of HRP-coated phages that displayed an anti-LipH VHH antibody to immobilized antigen could be confirmed by measurement of peroxidase activity using the chromogenic peroxidase substrate TMB (Fig. 2a). Since HRP was used at a 10 000-fold molar excess over phagemid particles, each phagemid most likely carries numerous HRP molecules on its surface. It remained to be shown whether a coupled reaction can occur on the phage surface where the oxidase delivers the H2O2 required by HRP. Therefore, a microtiter plate was coated with a LipH – oxidase conjugate. HRP-coated phages displaying anti-LipH VHH or control phages displaying an unrelated antigen were added. After washing and addition of substrate, peroxidase activity 43
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fifty microliters antigen –oxidase conjugate (9 nM) was added to 50 ml HRP conjugated phage suspension (1011 phagemid virions) and incubated for 90 min at 48C with slight agitation. The reaction volume was then raised to 500 ml by addition of PBS. One microliter catalase stock solution (2 mg/ ml) and 2.5 ml biotin-S-S-tyramide stock solution (11.5 mM) were added. The enzyme reaction was started by addition of 10 ml 10 mM glucose. The reaction mixture was incubated for 15 min at RT (without agitation) in the dark. After addition of 200 ml catalase stock solution and incubation for 3 min at RT, 700 ml phage precipitation buffer was added and the mixture was incubated on ice for 45 min. The phage pellet was collected by centrifugation in a table top centrifuge (20 min, 48C) and resuspended in 200 ml PBS. Twenty microliters of streptavidin coupled magnetic beads (Dynabeads T1, Invitrogen) were added to the sample and incubated by agitation for 30 min. Then the beads were washed five times with 200 ml PBS. A magnetic particle collector was used to separate beads from the supernatant. Afterward, the collected beads were suspended in 200 ml PBS and transferred to a new tube to eliminate potential plastic-binding antibody-phagemids. Beads were collected and resuspended in 200 ml 0.2 M glycine-HCl pH 2.2 and incubated for 8 min to remove nonspecifically bound phages from the beads. This procedure was repeated two times (without incubation). Finally, beads were washed five times using PBS Tween buffer (0.05% Tween 20 in PBS) and five times with PBS. After the last wash step, streptavidin-coated beads with bound phage were resuspended and incubated 10 min in 200 ml PBS containing 50 mM DTT to reduce the disulfide bond in biotin-S-S-tyramide. After magnetic separation, the supernatant was used to infect E. coli ER2738 via transfer to 20 ml bacterial liquid culture at an OD600 of 0.5 and incubation for 30 min at 378C without shaking. Cells were collected by centrifugation, suspended in 1.5 ml PBS, streaked out in 500 ml aliquots on agar plates containing chloramphenicol (25 mg/ml), and incubated overnight at 378C. Presence of 50 nM DTT did not negatively interefere with phage infectivity (data not shown). The total amount of eluted phage was determined by E. coli ER2738 infection of a serial dilution and plating of an aliquot of the infected cells.
A.Maaß et al.
Fig. 2 Coupled enzyme reaction on phage. (a) Proof of HRP coupling to the surface of phage particles while maintaining antigen binding capability. LipH protein was coupled to the wells of a microtiter plate. Phage displaying an anti-LipH VHH or unrelated protein to serve as a control were decorated with HRP (þHRP) and added to the wells (dark gray bars). Unconjugated anti-LipH VHH phages served as a control (light gray bar). For the undecorated anti-LipH-VHH phages, target protein binding was verified by addition of a peroxidase-conjugated anti-pVIII antibody. HRP activity was detected by addition of TMB and measurement of absorbance at 450 nm. (b) Oxidase-mediated activation of HRP on phage. Wells of a microtiter plate were coated with LipH– oxidase conjugate. After addition of HRP-decorated anti-LipH VHH phages (dark grey bar) or control phages (light gray bar), respectively, and plate wash, oxidase substrate was added and HRP activity was measured at 405 nm 20 min after addition of ABTS. Error bars represent the standard deviation from three measurements.
could be detected using substrate ABTS (Fig. 2b), indicating a coupled reaction took place on phage.
Validation of selective phage biotinylation upon antigen binding For the validation of selective phage bitinylation, anti-LipH VHH presenting phages were mixed with control phages presenting an anti-matriptase VHH at a ratio of 1 : 100 000. To this end, phages were grown separately, titered and mixed. After HRP conjugation to 1011 phages and addition of the LipH–oxidase complex (5 nM) and catalase, the coupled enzyme reaction was initiated by addition of glucose and biotin-S-S-tyramide. After a 15-min incubation, phages were PEG precipitated to remove excess biotin, resuspended in PBS and captured by binding to streptavidin-coated magnetic beads. After sequential washing, first with PBS, then PBS-Tween, then glycine buffer pH 2.2, phage particles were released from the magnetic beads by addition of 50 mM DTT which reduces the internal disulfide bond in biotin-S-S-tyramide. After magnetic separation, the supernatant was used to infect E. coli cells. As a control and to investigate 44
the possibility that biotinylation and immobilisation of antigen is involved in the observed enrichment, a conventional phage display experiment was performed, where the biotinylated LipH antigen (5 nM) was incubated with phages followed by addition of magnetic beads and separation. Enrichment of anti-LipH VHH-presenting phages was verified by polymerase chain reaction (PCR) analysis using specific primers for the anti-LipH VHH-coding sequence (Fig 3). PCR analysis of individual clones after screening round 1 using 3CARD revealed that 5 of 16 clones of the 1 : 100 000 dilution contained a 451-bp fragment which is indicative of presence of the anti-LipH VHH gene while no positive clones were obtained by conventional phage display screening of the same phagemid library. This mixing experiment was repeated with similar results using different control phages to exclude the possibility that growth advantages of the Lip-VHH phage over the control phage or differences in infectivity might account for the observed enrichment (Supplementary Fig. S2). Enrichment depends on the presence of both biotin tyramide and glucose (Supplementary Fig. S3). LipH–VHH phage enrichment was also observed after two consecutive screening
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 3 Enrichment of binders by 3CARD screening. (a) Phage enzyme-linked immunosorbent assay of clones from a 1 : 100 000 mixture (total input number 1011) of anti-LipH and anti-matriptase control phages screened via 3CARD (black bar) or conventional phage display (gray bar) for LipH binding phages before (round 0) and after screening rounds 1 and 2. Phage binding to microtiter plate-coated LipH was determined using an peroxidase-conjugated anti-pVIII antibody and chromogenic HRP substrate by measurement of absorbance at 450 nm. Error bars, s.d. n ¼ 3. (b) PCR screen of 16 individual phagemids obtained after screening round 1 of 3CARD or conventional phage display screening (c) using a primer pair that is specific for the anti-LipH VHH-coding sequence. For the anti-LipH VHH-presenting phagemid a 451-bp fragment was expected. P: PCR product from anti-LipH VHH-presenting phagemid that served as a PCR control.
General strategy for antibody library screening
rounds using a mixing ratio of 1 : 106 and 1 : 107, respectively (Supplementary Fig. S4). 3
Discussion In this paper we show that screening of diverse libraries can be significantly accelerated using 3CARD via conversion of a transient interaction of an antibody with its antigen into a permanent label, i.e. covalent coupling of biotin to the surface of phage. This was achieved by application of a coupled enzyme reaction on the phage surface, such that target-bound oxidase generates the H2O2 that is required by phage-bound peroxidase to activate biotin tyramide. Biotin tyramide has been used extensively in immunohistochemistry and also for selectively labeling enzyme displaying cells. Due to its short half-life in aqueous solutions, the tyramide radical formed by the enzyme reacts at the site of formation thus allowing for the site-specific covalent deposition of biotin (Becker et al., 2007, 2008; Bobrow et al., 1989, 1991, 1992; Lipovsek et al., 2007). For a genotype – phenotype linkage to occur, the biotinylation reaction has to be carefully controlled and restricted to the population of rare library members that are able to interact specifically with a predetermined interaction partner. To this end, a coupled enzyme reaction was established were the activity of
45
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
CARD VHH library screening. To investigate whether CARD phage display screening can be used for the fast selection of antigen binders from a phage library, a camelid VHH phage display library was used. The library contains 108 variants of camelid variable heavy chains with randomly variegated hypervariable loop sequences (Habicht et al., 2007). Three different target proteins were used for screening: P. aeruginosa LipH; cetuximab, an anti-epidermal growth factor receptor antibody (Robert et al., 2001); matriptase-1, a cancer relevant type II serine protease (Choi et al., 2009). Phagemid virions were screened for binders using 3CARD. In addition, a conventional phage display screen was performed for LipH and cetuximab using biotinylated target protein followed by enrichment of binders via phagemid sorption to streptavidin-coated magnetobeads. After screening round 1 (for phagemid titers see Supplementary Table S1), phagemid virions were generated from 96 individual clones and probed for target protein binding by phage enzyme-linked immunosorbent assay, which indicated for the 3CARD screen accumulation of binding phagemid clones for all targets (Fig. 4). No specific binders to target proteins LipH and cetuximab were detected in the initial library prior to screening (data not shown) or after round 1 of conventional phage display screening (Fig. 4b and d). The coding sequences of putative binders from 3CARD screening round 1 were determined (Supplementary Fig. S5). From each target screen, the VHH gene of one selected clone was transferred to expression vector pMX and the respective maltose-binding protein–VHH fusion proteins were produced and purified from E. coli cell lysate (Wentzel et al., 1999). Interaction with the respective target protein was verified by biolayer interferometry (Supplementary Fig. S6), corroborating the notion that the screening scheme is robust enough to obtain antibody binders from phage libraries against several unrelated targets in one screening round. No efforts were made to remove oxidase, peroxidase, or phage protein binders from the initial phage library prior to screening that therefore may make up a significant fraction of the selected phage population. 3
tyramide radical forming HRP is controlled by the formation of H2O2 via an oxidase. In this case, the antigen – oxidase conjugate in solution is in vast excess over phage particles. However, production of excess H2O2 that is required by HRP to activate tyramide can be avoided by addition of catalase, which efficiently and continuously breaks down excess H2O2. As a consequence, only HRP molecules that are in close proximity to an oxidase have a chance to capture a H2O2 molecule before it is decomposed by catalase. All together, the system provides a double layer of security to ensure selective biotin labeling of target-bound phage both by making H2O2 exclusively available to HRP on antigen-binding phage and by generating a tyramide radical that, due to its short half-life, only acts locally. To our knowledge, this is the first time that a coupled enzyme reaction has been applied to phage display screening. A singleenzyme approach to permanently label target-binding bacteriophages has been described by McCafferty and coworkers over a decade ago (Bradbury and McCafferty, 2011). It was termed ‘pathfinder technology’ and relies on the HRP-mediated biotinylation of antibody-displaying phage at a specific site on a cell surface or target protein (Osbourn et al., 1998). In this experimental setup, the peroxidase was not directly coupled to phage particles. Instead, a peroxidase-conjugated ligand to the target protein, e.g. an antibody was used that binds the target protein apart from the phage antibody-binding site. Colocalization of target-bound phage and HRP on the target results in phage biotinylation. This method has been applied exclusively to cellsurface exposed target proteins, which makes it possible to remove excess peroxidase–ligand conjugate and unbound phage particles, whether biotinylated or not, by cell centrifugation and washing prior to phage capturing by binding to streptavidin beads. The method described here can be successfully applied to the screening of soluble targets, most likely due to the strongly reduced cross-biotinylation of phage particles by addition of catalase that eliminates excess H2O2. Conversion of the transient interaction of binder and target protein into a covalent biotin label provides a siginificant advantage in that very stringent conditions for removal of phages that nonspecifically bind to the adsorption matrix can be applied. We have used extensive wash procedures that include usage of detergents as well as lowering the pH to 2.2 since it is known that the biotin – streptavidin interaction is stable under those conditions (Hofmann et al., 1980). It can be expected that at low pH the phage-bound target protein complex denatures and therefore dissociates, which may be benefical for phage infection of E. coli cells that require an unmasked pIII protein. Usage of a biotinylation substrate with a cleavable linker allows for the release of streptavidin-bound phages from the magnetic beads by addition of a reducing agent which may further reduce background of nonspecifically bound phages. The phage library used here reportedly allowed for VHH isolation applying conventional phage display strategies (Lipovsek et al., 2007), yielding medium and low affinity binders. At present it remains to be experimentally elucidated whether 3CARD phage display screening allows for the discrimination between low-, medium-, and high-affinity binders and can therefore be used for affinity maturation screening. Since the concentration of oxidase-target conjugate can be varied freely it may be feasible to screen for high-affinity binders by affinity selection by reduction of target concentration or by off-rate selection (Hawkins et al., 1992). For affinity
A.Maaß et al.
selection, phages can be mixed with small amounts of soluble oxidase conjugated antigen such that the antigen is in excess over phages but with the concentration of antigen lower than the dissociation constant (Kd) of the antibody. Those virions bound to antigen are then labeled by 3CARD and isolated by capture to magnetic beads. For off-rate selection, antibodies can be preloaded with oxidase-coupled antigen and diluted into excess of unconjugated antigen prior to 3CARD capture on streptavidin-coated magnetic beads. In conclusion, the strategy described here may be particularly useful for the initial screening of very large libraries exceeding 109 variants, where it is important in the first screening rounds to enrich the rare target binders against a large background of nonspecific binding phages. The genotype –phenotype linkage via coupled enzyme reaction and reporter deposition may also be applicable to the screening of libraries that are displayed on the surface of bacterial and yeast cells with interesting ramifications for future applications. 46
Supplementary data Supplementary data are avaible at PEDS online. Acknowledgements We are grateful to Svenja Wiechmann for validating the reproducibility of the screens and control experiments. This work was supported in part by BMBF projects BIOPUR and NANOKAT. Author contributions: T.H. and A.M. conceived the experiments; A.M. and T.H. performed the experiments and analyzed the date; F.M. performed gene cloning and validated the reproducibility of the screens; T.H. and F.M. produce and purification the antigen– glucose oxidase products. J.F. and T.H. conceived protein production, analyses and purification B.G. produced matriptase-1 protein and supervised the screening for matriptase-1 binders. H.K. and S.B. planned and supervised the project. H.K. wrote the article. Conflict of interest: none declared.
Competing financial interests The authors declare no competing financial interests.
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 4 Phage enzyme-linked immunosorbent assay of individual clones after 3CARD (a,c,e) or conventional phage display (b,d) screening against the respective target. Microtiter plates were coated with the target proteins LipH (a,b), cetuximab (c,d), or matriptase-1 (e), respectively. Phage binding was detected by addition of a peroxidase-conjugated anti-pVIII antibody using the chromogenic HRP substrate TMB and measurement of absorbance at 450 nm. The mean value of absorbance that is indicative of non-binding was determined from 10 clones of each experiments with the lowest OD450 values. Phage clones exceeding this value at least by a factor of 2.5 were considered to be potential target protein binders and are displayed as black bars. Clones that have been selected for expression as soluble protein are marked by an asterisk with the Kd of target protein binding given in the table (f ).
General strategy for antibody library screening
Funding This work was supported by BMBF through grant BIOPUR 0315588. References
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Avrutina,O., Fittler,H., Glotzbach,B., Kolmar,H. and Empting,M. (2012) Org. Biomol. Chem., 10, 7753– 7762. Becker,S., Michalczyk,A., Wilhelm,S., Jaeger,K.E. and Kolmar,H. (2007) Chembiochem, 8, 943– 949. Becker,S., Ho¨benreich,H., Vogel,A., Knorr,J., Wilhelm,S., Rosenau,F., Jaeger,K.E., Reetz,M.T. and Kolmar,H. (2008) Angew Chem. Int. Ed. Engl., 47, 5085– 5088. Bobrow,M.N., Harris,T.D., Shaughnessy,K.J. and Litt,G.J. (1989) J. Immunol. Methods, 125, 279–285. Bobrow,M.N., Shaughnessy,K.J. and Litt,G.J. (1991) J. Immunol. Methods, 137, 103– 112. Bobrow,M.N., Litt,G.J., Shaughnessy,K.J., Mayer,P.C. and Conlon,J. (1992) J. Immunol. Methods, 150, 145–149. Bradbury,A.R., Sidhu,S., Dubel,S. and McCafferty,J. (2011) Nat. Biotechnol., 29, 245–254. Brockmann,E.C., Akter,S., Savukoski,T., et al. (2011) Protein Eng. Des. Sel., 24, 691–700. Choi,S.Y., Bertram,S., Glowacka,I., Park,Y.W. and Pohlmann,S. (2009) Trends Mol. Med., 15, 303– 312. Derda,R., Tang,S.K., Li,S.C., Ng,S., Matochko,W. and Jafari,M.R. (2011) Molecules, 16, 1776– 1803. Green,N.M. (1970) Methods Enzymol., 18, 418–424. Green,N.M. (1975) Adv. Protein Chem., 29, 85– 133. Habicht,G., Haupt,C., Friedrich,R.P., et al. (2007) Proc Natl Acad Sci U S A, 104, 19232–19237. Hawkins,R.E., Russell,S.J. and Winter,G. (1992) J. Mol. Biol., 226, 889– 896. Hermanson,G.T. (1996) Bioconjugation Techniques, Academic Press. Hofmann,K., Wood,S.W., Brinton,C.C., Montibeller,J.A. and Finn,F.M. (1980) Proc Natl Acad Sci U S A, 77, 4666– 4668. Janolino,V.G., Fontecha,J. and Swaisgood,H.E. (1996) App. Biochem. Biotech., 56, 1– 7. Lipovsek,D., Antipov,E., Armstrong,K.A., Olsen,M.J., Klibanov,A.M., Tidor,B. and Wittrup,K.D. (2007) Chem. Biol., 14, 1176– 1185. Liu,Y., Adams,J.D., Turner,K., Cochran,F.V., Gambhir,S.S. and Soh,H.T. (2009) Lab. Chip, 9, 1033– 1036. O’Brien,P.M. and Aitken,R. (2002) Antibody Phage Display: Methods and Protocols, Humana Press. Osbourn,J.K., Derbyshire,E.J., Vaughan,T.J., Field,A.W. and Johnson,K.S. (1998) Immunotechnology, 3, 293–302. Pepper,L.R., Cho,Y.K., Boder,E.T. and Shusta,E.V. (2008) Comb. Chem. High Throughput Screen., 11, 127–134. Robert,F., Ezekiel,M.P., Spencer,S.A., et al. (2001) J Clin Oncol, 19, 3234–3243. Rothe,C., Urlinger,S., Lohning,C., et al. (2008) J. Mol. Biol., 376, 1182–1200. Smith,G.P. (1985) Science, 228, 1315– 1317. Steiner,D., Forrer,P. and Pluckthun,A. (2008) J. Mol. Biol., 382, 1211–1227. van Bloois,E., Winter,R.T., Kolmar,H. and Fraaije,M.W. (2011) Trends Biotechnol., 29, 79–86. Vodnik,M., Zager,U., Strukelj,B. and Lunder,M. (2011) Molecules, 16, 790–817. Wentzel,A., Christmann,A., Kratzner,R. and Kolmar,H. (1999) J. Biol. Chem., 274, 21037–21043. Wilhelm,S., Rosenau,F., Becker,S., Buest,S., Hausmann,S., Kolmar,H. and Jaeger,K.E. (2007) Chembiochem, 8, 55– 60.
47
A general strategy for antibody library screening via conversion of transient target binding into permanent reporter deposition Alexander Maaß1†, Tim Heiseler1†, Franziska Maaß1, Janine Fritz1, Thomas Hofmeyer1, Bernhard Glotzbach1, Stefan Becker2 and Harald Kolmar1,3 1
Institute for Organic Chemistry and Biochemistry, Technische Universita¨t Darmstadt, D-64287 Darmstadt, Germany and 2Merck Serono, Protein Engineering and Antibody Technologies, D-64293 Darmstadt, Germany
3
To whom correspondence should be addressed. E-mail: [email protected]
Edited by: Arne Skerra
We report here a generally applicable method for the selective covalent attachment of a reporter molecule to a replicating entity that allows one to obtain specific binders from a single round of library screening. We show that selective biotinylation of phage particles displaying a binder to any given target can be achieved by application of a coupled enzyme reaction on the surface of the target-binding phage particles that includes a peroxidase, an oxidase and a catalase. Due to the covalent linkage of biotin together with the tight and stable interaction of biotin with streptavidin, very stringent wash conditions for removal of nonspecific binders can be applied. The method termed 3CARD (triple catalytic reporter deposition) was successfully applied to singleround screening of a phage display library of camelid singledomain antibodies against three different target proteins. Keywords: antibody engineering/catalysed reporter deposition/peroxidase/phage display screening/oxidase
Introduction The phage display technology was originally introduced by George Smith more than 25 years ago (Smith, 1985) and has become an indispensible tool in molecular biology and molecular medicine to identify peptides and particularly antibody molecules with predefined binding characteristics from large libraries. Since then, many improvements have been reported and screening concepts are now available to isolate binding molecules against virtually any target protein (Bradbury et al., 2011). Moreover, in recent years bacterial and particularly yeast surface display have become increasingly popular alternatives for phage display library screening applications (Pepper et al., 2008; van Bloois et al., 2011). Library screening generally involves exposure of a population of binding molecules on the surface of a replicating entity and contacting them with an interaction partner of interest such as an antigen molecule, in the case of antibody display. †
These authors contributed equally to this work.
# The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected]
41
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Received August 15, 2013; revised October 31, 2013; accepted November 21, 2013
The experimental setup is designed such that those phage particles or microbial cells displaying a target-specific variant can be isolated from the population by methods such as biopanning or cell sorting. Ideally, a single round of screening would be required to obtain the desired protein variants. However, the enrichment of specific binders is limited by non-specific binding that requires recursive rounds of screening and amplification to obtain the best binders from the library (Vodnik et al., 2011). The first screening round is particularly critical, since it can be expected for large libraries that specific binders constitute only a tiny fraction of the initial population such that non-specific binders largely dominate over those that interact specifically. As a consequence, conditions have to be carefully controlled and optimized for every single target and screening strategies devised that are stringent enough to remove nonspecific binders but sufficiently mild to keep the target binding of specifically interacting library members intact (Liu et al., 2009). Since the probability of obtaining high-affinity binders increases with library size (Steiner et al., 2008), progressively larger libraries have been established that can exceed 1010 variants (Rothe et al., 2008; Brockmann et al., 2011). However, repeated amplification of libraries as it occurs in each screening round is known to reduce diversity and limits the number of candidates that can be identified in a library screening experiment (Derda et al., 2011). As a consequence, it is highly desirable to reduce the number of screening rounds to a minimum— ideally a single round. Conceptually, this can be achieved by setting up a method where the transient and relatively weak interaction of target and library member translates into covalent attachment of a labeling molecule that can be used as a purification handle and allows for application of more stringent conditions to remove non-specifically bound library members. Catalysed reporter deposition (CARD) has been widely used in immune histochemistry for protein biotinylation (Bobrow et al., 1989, 1991, 1992). It relies on the horseradish peroxidase (HRP)-mediated formation of a biotin tyramide radical that readily reacts with proteins in close proximity, resulting in the formation of a covalent linkage between biotin and tyrosine residues on proteins (Fig. 1). To establish a selective biotinylation of phage or cell surface display library members which are bound to a target molecule, we make us of the fact that HRP-mediated biotinylation requires the presence of hydrogen peroxide (H2O2). The 3CARD system described here combines the activity of three different enzymes, namely a peroxidase, an oxidase, and a catalase. First, HRP is coupled onto all members of a starting population, which is demonstrated here in the experimental setting of a phage display library (Fig. 1a). Phage biotinylation by HRP requires presence of biotin-S-S-tyramide and H2O2. In the 3CARD setup, H2O2 is generated by a second enzyme that is conjugated to the antigen of interest (Fig. 1b and c). Sugar oxidases such as glucose oxidase catalyse the oxidation of sugars by electron transfer to oxygen, which results in the formation of H2O2. H2O2
A.Maaß et al.
is generated at higher concentrations on those phage particles to which the antigen–oxidase complex is bound since the reaction mixture contains as a third enzyme catalase, which very efficiently breaks down H2O2. As a consequence, H2O2 transfer to HRP and subsequent phage biotinylation should only occur for antigen-bound phage that has the oxidase in close proximity to HRP. Specifically labeled phages can then be rescued by biopanning to immobilized streptavidin and subjected to very stringent wash conditions prior to infection of bacteria. Here we describe the application of 3CARD for singledomain antibody library screening and present straightforward methods for coupling HRP onto the phage surface and conjugating oxidase with target protein such that a combined enzyme reaction on phagemid virions results in their selective biotinylation and isolation in a single round of screening. Materials and methods
Sodium periodate oxidation of HRP and GOX 18.8 mg Sodium periodate (87.8 mmol, Sigma-Aldrich) were diluted in 1 ml H2O. The protein to be oxidized, either HRP (Sigma-Aldrich, Typ VI-A) or glucose oxidase, respectively, was diluted in H2O to a final concentration of 10 mg/ml and 200 ml protein solution was mixed with 20 ml of the sodium periodate stock solution. After 20 min incubation at room temperature (RT) in the dark, the reaction mixture was purified with a PD10 column (GE Healthcare) pre-equilibrated with 100 mM sodium borate buffer pH 9.1. Subsequently, 1.4 ml of the eluate containing the oxidized protein was filled up to 2 ml with sodium borate buffer and stored in aliquots at 2208C. The protein concentration in the eluted fractions was determined spectrophotometrically at 403 nm assuming a molar extinction coefficient of 100 for HRP and at 280 nm for GOX assuming a molar extinction coefficient of 16.7.
Target protein coupling to glucose oxidase 0.1 mg Of periodate oxidized glucose oxidase at a concentration of 1 mg/ml (6 mM; Sigma Type VI-A) was incubated with an equimolar amount or less of the target protein in 42
200 ml phosphate buffered saline (PBS) and incubated at room temperature in the dark for 180 min, followed by gel filtration chromatography using Superdex 200 pg 16/60 column (GE Healthcare) and collection of the early eluting fractions (Supplementary Fig. S1). The protein concentration in the eluted fractions was determined spectrophotometrically at 280 nm assuming an 1 : 1 ratio of enzyme to target and taking into account the molar extinction coefficient of GOX (16.7) and of the respective conjugated protein. Target proteins LipH and matriptase-1 were purified according to published procedures (Becker et al., 2007; Avrutina et al., 2012). Cetuximab was provided by Merck Serono.
Phage library propagation The pAK200-VHH library consists of 108 variants that were generated by randomization of CDR1 (5 residues), CDR2 (8 residues), and CDR3 with loop variegation from 7 to 17 residues using a NNS coding scheme. pAK200 mediates chloramophenicol resistance and contains the coding sequence for a PelB leader peptide followed by the VHH gene fused to shortened pIII. The gene fusion is expressed under lac promoter control. Due to the expected high occurrence of amber stop codons within the randomized loop regions, no additional stop codon was introduced between pIII and the fused VHH coding sequence. Each phage particle displays on average not more than one VHH fragment due to amber codon suppression in Escherichia coli ER2738. The phage library was generated by infection of 1011 pAK200 containing E. coli cells using 1013 M13VCS helper phage particles.
HRP coupling to phagemid virions Prior to HRP coupling, streptavidin binders were removed from the phage library. To achieve this, 100 ml pAK200-VHH phagemids (2.1011 phagemid virions) were incubated with 10 ml Streptavidin-Dynabeads (T1, Invitrogen) for 30 min at RT in a shaker. After separation of the beads using a magnetic particle concentrator, the supernatant was transferred into a 1.5-ml reaction tube and 100 ml oxidized HRP (1 mg/ml in
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 1 General outline of the 3CARD phage display screening procedure. (a) Covalent deposition of HRP to the surface of phage particles that display a library of antibody variants. (b) Binding of phage displayed antibody (AB) to protein target (P) which is coupled to glucose oxidase (GOX). (c) GOX-mediated generation of H2O2. (d) Catalase (CAT) breakdown of excess H2O2. (e) Activation of HRP mediated by H2O2. (f,g) Formation of a biotin tyramide radical by HRP and conjugation to the phage surface. (h) Biopanning of biotinylated phage using to streptavidin-coated magnetic beads.
General strategy for antibody library screening
100 mM borate buffer pH 9.1) was added. The reaction mixture was incubated at room temperature in the dark under slow agitation for 45 min. To remove uncoupled HRP, 200 ml phage precipitation buffer (20% PEG 6000, 1.5 M NaCl) was added, followed by incubation on ice for 45 min. After centrifugation in a table top centrifuge (20 min, 13.000 rpm), the phage pellet was dissolved in 50 ml PBS.
collector was used to separate beads from the supernatant. The beads were washed 10 times with 1 ml PBS-Tween (0.05%) buffer. Finally, beads were collected and resuspended in 200 ml 0.2 M glycine-HCl pH 2.2 and incubated for 8 min to release phage particles form the beads followed by magnetic separation and neutralization by addition of 1 ml 1 M Tris-HCl pH 9.0. Infection of E. coli cells was performed as described above.
Coupled enzyme reaction on phage
Conventional phage display screening Target protein biotinylation was performed as described (O’Brien and Aitken, 2002) using Sulfo-NHS-SS-biotin (Thermo Scientific) followed by removal of unbound biotin by gel filtration using a PD10 column (GE Healthcare). Biotinylation was verified using the HABA-avidin assay (Green, 1970, 1975; Janolino et al., 1996). To 2 1011 phagemid virions, biotinylated target protein was added at a final concentration of 4.5 nM in a final volume of 200 ml. After 20-min incubation, 20 ml streptavidin coupled magnetic beads (Dynabeads T1, Invitrogen) were added to the sample and incubated by agitation for 30 min. A magnetic particle
Results For the phage display screening procedure described above to be functional, it had to be shown that a coupled reaction of an oxidase and peroxidase can be performed on phage particles using an experimental setting, where the oxidase is coupled to the phage surface via antigen/antibody interaction and HRP is chemically conjugated to the phage surface such that the H2O2 generated leads to selective labeling of the bound phage and thereby to antigen –antibody– phage linkage. In addition, a simple method for conjugation of HRP to phage and oxidase to the target protein of interest had to be established. To address these points, a model experiment was performed that relies on the phage display of a camelid VHH single-chain antibody fragment that was obtained by conventional phage display library screening and selectively binds Pseudomonas aeruginosa LipH chaperon (Wilhelm et al., 2007) with an Kd of 20 nM. Presentation of the anti-LipH VHH domain on phage was achieved by genetic fusion of the VHH fragment to truncated pIII using phagemid display vector pAK200 (Habicht et al., 2007).
Conjugation of oxidase to target protein For conjugation the commercially available glucose oxidase was used. Since this enzyme is glycosylated, protein conjugation with the target protein can easily be achieved by periodate oxidation of terminal sugar moieties followed by Schiff base formation of the resulting aldehyde with a lysine residue residing on the surface of the target protein (Hermanson, 1996). For this purpose, oxidized glucose oxidase was incubated with target protein and the fusion was purified from unconjugated oxidase by gel filtration chromatography (Supplementary Fig. S1).
Coupled enzyme reaction on phage To investigate whether HRP can be coupled to the surface of filamentous phage without negative interference with antigen binding, phage particles displaying anti-LipH VHH and control phages displaying an unrelated peptide were incubated with oxidized HRP and purified by severalfold precipitation using polyethylene glycol. Wells of a microtiter plate were coated with LipH target protein. Binding of HRP-coated phages that displayed an anti-LipH VHH antibody to immobilized antigen could be confirmed by measurement of peroxidase activity using the chromogenic peroxidase substrate TMB (Fig. 2a). Since HRP was used at a 10 000-fold molar excess over phagemid particles, each phagemid most likely carries numerous HRP molecules on its surface. It remained to be shown whether a coupled reaction can occur on the phage surface where the oxidase delivers the H2O2 required by HRP. Therefore, a microtiter plate was coated with a LipH – oxidase conjugate. HRP-coated phages displaying anti-LipH VHH or control phages displaying an unrelated antigen were added. After washing and addition of substrate, peroxidase activity 43
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fifty microliters antigen –oxidase conjugate (9 nM) was added to 50 ml HRP conjugated phage suspension (1011 phagemid virions) and incubated for 90 min at 48C with slight agitation. The reaction volume was then raised to 500 ml by addition of PBS. One microliter catalase stock solution (2 mg/ ml) and 2.5 ml biotin-S-S-tyramide stock solution (11.5 mM) were added. The enzyme reaction was started by addition of 10 ml 10 mM glucose. The reaction mixture was incubated for 15 min at RT (without agitation) in the dark. After addition of 200 ml catalase stock solution and incubation for 3 min at RT, 700 ml phage precipitation buffer was added and the mixture was incubated on ice for 45 min. The phage pellet was collected by centrifugation in a table top centrifuge (20 min, 48C) and resuspended in 200 ml PBS. Twenty microliters of streptavidin coupled magnetic beads (Dynabeads T1, Invitrogen) were added to the sample and incubated by agitation for 30 min. Then the beads were washed five times with 200 ml PBS. A magnetic particle collector was used to separate beads from the supernatant. Afterward, the collected beads were suspended in 200 ml PBS and transferred to a new tube to eliminate potential plastic-binding antibody-phagemids. Beads were collected and resuspended in 200 ml 0.2 M glycine-HCl pH 2.2 and incubated for 8 min to remove nonspecifically bound phages from the beads. This procedure was repeated two times (without incubation). Finally, beads were washed five times using PBS Tween buffer (0.05% Tween 20 in PBS) and five times with PBS. After the last wash step, streptavidin-coated beads with bound phage were resuspended and incubated 10 min in 200 ml PBS containing 50 mM DTT to reduce the disulfide bond in biotin-S-S-tyramide. After magnetic separation, the supernatant was used to infect E. coli ER2738 via transfer to 20 ml bacterial liquid culture at an OD600 of 0.5 and incubation for 30 min at 378C without shaking. Cells were collected by centrifugation, suspended in 1.5 ml PBS, streaked out in 500 ml aliquots on agar plates containing chloramphenicol (25 mg/ml), and incubated overnight at 378C. Presence of 50 nM DTT did not negatively interefere with phage infectivity (data not shown). The total amount of eluted phage was determined by E. coli ER2738 infection of a serial dilution and plating of an aliquot of the infected cells.
A.Maaß et al.
Fig. 2 Coupled enzyme reaction on phage. (a) Proof of HRP coupling to the surface of phage particles while maintaining antigen binding capability. LipH protein was coupled to the wells of a microtiter plate. Phage displaying an anti-LipH VHH or unrelated protein to serve as a control were decorated with HRP (þHRP) and added to the wells (dark gray bars). Unconjugated anti-LipH VHH phages served as a control (light gray bar). For the undecorated anti-LipH-VHH phages, target protein binding was verified by addition of a peroxidase-conjugated anti-pVIII antibody. HRP activity was detected by addition of TMB and measurement of absorbance at 450 nm. (b) Oxidase-mediated activation of HRP on phage. Wells of a microtiter plate were coated with LipH– oxidase conjugate. After addition of HRP-decorated anti-LipH VHH phages (dark grey bar) or control phages (light gray bar), respectively, and plate wash, oxidase substrate was added and HRP activity was measured at 405 nm 20 min after addition of ABTS. Error bars represent the standard deviation from three measurements.
could be detected using substrate ABTS (Fig. 2b), indicating a coupled reaction took place on phage.
Validation of selective phage biotinylation upon antigen binding For the validation of selective phage bitinylation, anti-LipH VHH presenting phages were mixed with control phages presenting an anti-matriptase VHH at a ratio of 1 : 100 000. To this end, phages were grown separately, titered and mixed. After HRP conjugation to 1011 phages and addition of the LipH–oxidase complex (5 nM) and catalase, the coupled enzyme reaction was initiated by addition of glucose and biotin-S-S-tyramide. After a 15-min incubation, phages were PEG precipitated to remove excess biotin, resuspended in PBS and captured by binding to streptavidin-coated magnetic beads. After sequential washing, first with PBS, then PBS-Tween, then glycine buffer pH 2.2, phage particles were released from the magnetic beads by addition of 50 mM DTT which reduces the internal disulfide bond in biotin-S-S-tyramide. After magnetic separation, the supernatant was used to infect E. coli cells. As a control and to investigate 44
the possibility that biotinylation and immobilisation of antigen is involved in the observed enrichment, a conventional phage display experiment was performed, where the biotinylated LipH antigen (5 nM) was incubated with phages followed by addition of magnetic beads and separation. Enrichment of anti-LipH VHH-presenting phages was verified by polymerase chain reaction (PCR) analysis using specific primers for the anti-LipH VHH-coding sequence (Fig 3). PCR analysis of individual clones after screening round 1 using 3CARD revealed that 5 of 16 clones of the 1 : 100 000 dilution contained a 451-bp fragment which is indicative of presence of the anti-LipH VHH gene while no positive clones were obtained by conventional phage display screening of the same phagemid library. This mixing experiment was repeated with similar results using different control phages to exclude the possibility that growth advantages of the Lip-VHH phage over the control phage or differences in infectivity might account for the observed enrichment (Supplementary Fig. S2). Enrichment depends on the presence of both biotin tyramide and glucose (Supplementary Fig. S3). LipH–VHH phage enrichment was also observed after two consecutive screening
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 3 Enrichment of binders by 3CARD screening. (a) Phage enzyme-linked immunosorbent assay of clones from a 1 : 100 000 mixture (total input number 1011) of anti-LipH and anti-matriptase control phages screened via 3CARD (black bar) or conventional phage display (gray bar) for LipH binding phages before (round 0) and after screening rounds 1 and 2. Phage binding to microtiter plate-coated LipH was determined using an peroxidase-conjugated anti-pVIII antibody and chromogenic HRP substrate by measurement of absorbance at 450 nm. Error bars, s.d. n ¼ 3. (b) PCR screen of 16 individual phagemids obtained after screening round 1 of 3CARD or conventional phage display screening (c) using a primer pair that is specific for the anti-LipH VHH-coding sequence. For the anti-LipH VHH-presenting phagemid a 451-bp fragment was expected. P: PCR product from anti-LipH VHH-presenting phagemid that served as a PCR control.
General strategy for antibody library screening
rounds using a mixing ratio of 1 : 106 and 1 : 107, respectively (Supplementary Fig. S4). 3
Discussion In this paper we show that screening of diverse libraries can be significantly accelerated using 3CARD via conversion of a transient interaction of an antibody with its antigen into a permanent label, i.e. covalent coupling of biotin to the surface of phage. This was achieved by application of a coupled enzyme reaction on the phage surface, such that target-bound oxidase generates the H2O2 that is required by phage-bound peroxidase to activate biotin tyramide. Biotin tyramide has been used extensively in immunohistochemistry and also for selectively labeling enzyme displaying cells. Due to its short half-life in aqueous solutions, the tyramide radical formed by the enzyme reacts at the site of formation thus allowing for the site-specific covalent deposition of biotin (Becker et al., 2007, 2008; Bobrow et al., 1989, 1991, 1992; Lipovsek et al., 2007). For a genotype – phenotype linkage to occur, the biotinylation reaction has to be carefully controlled and restricted to the population of rare library members that are able to interact specifically with a predetermined interaction partner. To this end, a coupled enzyme reaction was established were the activity of
45
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
CARD VHH library screening. To investigate whether CARD phage display screening can be used for the fast selection of antigen binders from a phage library, a camelid VHH phage display library was used. The library contains 108 variants of camelid variable heavy chains with randomly variegated hypervariable loop sequences (Habicht et al., 2007). Three different target proteins were used for screening: P. aeruginosa LipH; cetuximab, an anti-epidermal growth factor receptor antibody (Robert et al., 2001); matriptase-1, a cancer relevant type II serine protease (Choi et al., 2009). Phagemid virions were screened for binders using 3CARD. In addition, a conventional phage display screen was performed for LipH and cetuximab using biotinylated target protein followed by enrichment of binders via phagemid sorption to streptavidin-coated magnetobeads. After screening round 1 (for phagemid titers see Supplementary Table S1), phagemid virions were generated from 96 individual clones and probed for target protein binding by phage enzyme-linked immunosorbent assay, which indicated for the 3CARD screen accumulation of binding phagemid clones for all targets (Fig. 4). No specific binders to target proteins LipH and cetuximab were detected in the initial library prior to screening (data not shown) or after round 1 of conventional phage display screening (Fig. 4b and d). The coding sequences of putative binders from 3CARD screening round 1 were determined (Supplementary Fig. S5). From each target screen, the VHH gene of one selected clone was transferred to expression vector pMX and the respective maltose-binding protein–VHH fusion proteins were produced and purified from E. coli cell lysate (Wentzel et al., 1999). Interaction with the respective target protein was verified by biolayer interferometry (Supplementary Fig. S6), corroborating the notion that the screening scheme is robust enough to obtain antibody binders from phage libraries against several unrelated targets in one screening round. No efforts were made to remove oxidase, peroxidase, or phage protein binders from the initial phage library prior to screening that therefore may make up a significant fraction of the selected phage population. 3
tyramide radical forming HRP is controlled by the formation of H2O2 via an oxidase. In this case, the antigen – oxidase conjugate in solution is in vast excess over phage particles. However, production of excess H2O2 that is required by HRP to activate tyramide can be avoided by addition of catalase, which efficiently and continuously breaks down excess H2O2. As a consequence, only HRP molecules that are in close proximity to an oxidase have a chance to capture a H2O2 molecule before it is decomposed by catalase. All together, the system provides a double layer of security to ensure selective biotin labeling of target-bound phage both by making H2O2 exclusively available to HRP on antigen-binding phage and by generating a tyramide radical that, due to its short half-life, only acts locally. To our knowledge, this is the first time that a coupled enzyme reaction has been applied to phage display screening. A singleenzyme approach to permanently label target-binding bacteriophages has been described by McCafferty and coworkers over a decade ago (Bradbury and McCafferty, 2011). It was termed ‘pathfinder technology’ and relies on the HRP-mediated biotinylation of antibody-displaying phage at a specific site on a cell surface or target protein (Osbourn et al., 1998). In this experimental setup, the peroxidase was not directly coupled to phage particles. Instead, a peroxidase-conjugated ligand to the target protein, e.g. an antibody was used that binds the target protein apart from the phage antibody-binding site. Colocalization of target-bound phage and HRP on the target results in phage biotinylation. This method has been applied exclusively to cellsurface exposed target proteins, which makes it possible to remove excess peroxidase–ligand conjugate and unbound phage particles, whether biotinylated or not, by cell centrifugation and washing prior to phage capturing by binding to streptavidin beads. The method described here can be successfully applied to the screening of soluble targets, most likely due to the strongly reduced cross-biotinylation of phage particles by addition of catalase that eliminates excess H2O2. Conversion of the transient interaction of binder and target protein into a covalent biotin label provides a siginificant advantage in that very stringent conditions for removal of phages that nonspecifically bind to the adsorption matrix can be applied. We have used extensive wash procedures that include usage of detergents as well as lowering the pH to 2.2 since it is known that the biotin – streptavidin interaction is stable under those conditions (Hofmann et al., 1980). It can be expected that at low pH the phage-bound target protein complex denatures and therefore dissociates, which may be benefical for phage infection of E. coli cells that require an unmasked pIII protein. Usage of a biotinylation substrate with a cleavable linker allows for the release of streptavidin-bound phages from the magnetic beads by addition of a reducing agent which may further reduce background of nonspecifically bound phages. The phage library used here reportedly allowed for VHH isolation applying conventional phage display strategies (Lipovsek et al., 2007), yielding medium and low affinity binders. At present it remains to be experimentally elucidated whether 3CARD phage display screening allows for the discrimination between low-, medium-, and high-affinity binders and can therefore be used for affinity maturation screening. Since the concentration of oxidase-target conjugate can be varied freely it may be feasible to screen for high-affinity binders by affinity selection by reduction of target concentration or by off-rate selection (Hawkins et al., 1992). For affinity
A.Maaß et al.
selection, phages can be mixed with small amounts of soluble oxidase conjugated antigen such that the antigen is in excess over phages but with the concentration of antigen lower than the dissociation constant (Kd) of the antibody. Those virions bound to antigen are then labeled by 3CARD and isolated by capture to magnetic beads. For off-rate selection, antibodies can be preloaded with oxidase-coupled antigen and diluted into excess of unconjugated antigen prior to 3CARD capture on streptavidin-coated magnetic beads. In conclusion, the strategy described here may be particularly useful for the initial screening of very large libraries exceeding 109 variants, where it is important in the first screening rounds to enrich the rare target binders against a large background of nonspecific binding phages. The genotype –phenotype linkage via coupled enzyme reaction and reporter deposition may also be applicable to the screening of libraries that are displayed on the surface of bacterial and yeast cells with interesting ramifications for future applications. 46
Supplementary data Supplementary data are avaible at PEDS online. Acknowledgements We are grateful to Svenja Wiechmann for validating the reproducibility of the screens and control experiments. This work was supported in part by BMBF projects BIOPUR and NANOKAT. Author contributions: T.H. and A.M. conceived the experiments; A.M. and T.H. performed the experiments and analyzed the date; F.M. performed gene cloning and validated the reproducibility of the screens; T.H. and F.M. produce and purification the antigen– glucose oxidase products. J.F. and T.H. conceived protein production, analyses and purification B.G. produced matriptase-1 protein and supervised the screening for matriptase-1 binders. H.K. and S.B. planned and supervised the project. H.K. wrote the article. Conflict of interest: none declared.
Competing financial interests The authors declare no competing financial interests.
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Fig. 4 Phage enzyme-linked immunosorbent assay of individual clones after 3CARD (a,c,e) or conventional phage display (b,d) screening against the respective target. Microtiter plates were coated with the target proteins LipH (a,b), cetuximab (c,d), or matriptase-1 (e), respectively. Phage binding was detected by addition of a peroxidase-conjugated anti-pVIII antibody using the chromogenic HRP substrate TMB and measurement of absorbance at 450 nm. The mean value of absorbance that is indicative of non-binding was determined from 10 clones of each experiments with the lowest OD450 values. Phage clones exceeding this value at least by a factor of 2.5 were considered to be potential target protein binders and are displayed as black bars. Clones that have been selected for expression as soluble protein are marked by an asterisk with the Kd of target protein binding given in the table (f ).
General strategy for antibody library screening
Funding This work was supported by BMBF through grant BIOPUR 0315588. References
Downloaded from http://peds.oxfordjournals.org/ by guest on August 16, 2015
Avrutina,O., Fittler,H., Glotzbach,B., Kolmar,H. and Empting,M. (2012) Org. Biomol. Chem., 10, 7753– 7762. Becker,S., Michalczyk,A., Wilhelm,S., Jaeger,K.E. and Kolmar,H. (2007) Chembiochem, 8, 943– 949. Becker,S., Ho¨benreich,H., Vogel,A., Knorr,J., Wilhelm,S., Rosenau,F., Jaeger,K.E., Reetz,M.T. and Kolmar,H. (2008) Angew Chem. Int. Ed. Engl., 47, 5085– 5088. Bobrow,M.N., Harris,T.D., Shaughnessy,K.J. and Litt,G.J. (1989) J. Immunol. Methods, 125, 279–285. Bobrow,M.N., Shaughnessy,K.J. and Litt,G.J. (1991) J. Immunol. Methods, 137, 103– 112. Bobrow,M.N., Litt,G.J., Shaughnessy,K.J., Mayer,P.C. and Conlon,J. (1992) J. Immunol. Methods, 150, 145–149. Bradbury,A.R., Sidhu,S., Dubel,S. and McCafferty,J. (2011) Nat. Biotechnol., 29, 245–254. Brockmann,E.C., Akter,S., Savukoski,T., et al. (2011) Protein Eng. Des. Sel., 24, 691–700. Choi,S.Y., Bertram,S., Glowacka,I., Park,Y.W. and Pohlmann,S. (2009) Trends Mol. Med., 15, 303– 312. Derda,R., Tang,S.K., Li,S.C., Ng,S., Matochko,W. and Jafari,M.R. (2011) Molecules, 16, 1776– 1803. Green,N.M. (1970) Methods Enzymol., 18, 418–424. Green,N.M. (1975) Adv. Protein Chem., 29, 85– 133. Habicht,G., Haupt,C., Friedrich,R.P., et al. (2007) Proc Natl Acad Sci U S A, 104, 19232–19237. Hawkins,R.E., Russell,S.J. and Winter,G. (1992) J. Mol. Biol., 226, 889– 896. Hermanson,G.T. (1996) Bioconjugation Techniques, Academic Press. Hofmann,K., Wood,S.W., Brinton,C.C., Montibeller,J.A. and Finn,F.M. (1980) Proc Natl Acad Sci U S A, 77, 4666– 4668. Janolino,V.G., Fontecha,J. and Swaisgood,H.E. (1996) App. Biochem. Biotech., 56, 1– 7. Lipovsek,D., Antipov,E., Armstrong,K.A., Olsen,M.J., Klibanov,A.M., Tidor,B. and Wittrup,K.D. (2007) Chem. Biol., 14, 1176– 1185. Liu,Y., Adams,J.D., Turner,K., Cochran,F.V., Gambhir,S.S. and Soh,H.T. (2009) Lab. Chip, 9, 1033– 1036. O’Brien,P.M. and Aitken,R. (2002) Antibody Phage Display: Methods and Protocols, Humana Press. Osbourn,J.K., Derbyshire,E.J., Vaughan,T.J., Field,A.W. and Johnson,K.S. (1998) Immunotechnology, 3, 293–302. Pepper,L.R., Cho,Y.K., Boder,E.T. and Shusta,E.V. (2008) Comb. Chem. High Throughput Screen., 11, 127–134. Robert,F., Ezekiel,M.P., Spencer,S.A., et al. (2001) J Clin Oncol, 19, 3234–3243. Rothe,C., Urlinger,S., Lohning,C., et al. (2008) J. Mol. Biol., 376, 1182–1200. Smith,G.P. (1985) Science, 228, 1315– 1317. Steiner,D., Forrer,P. and Pluckthun,A. (2008) J. Mol. Biol., 382, 1211–1227. van Bloois,E., Winter,R.T., Kolmar,H. and Fraaije,M.W. (2011) Trends Biotechnol., 29, 79–86. Vodnik,M., Zager,U., Strukelj,B. and Lunder,M. (2011) Molecules, 16, 790–817. Wentzel,A., Christmann,A., Kratzner,R. and Kolmar,H. (1999) J. Biol. Chem., 274, 21037–21043. Wilhelm,S., Rosenau,F., Becker,S., Buest,S., Hausmann,S., Kolmar,H. and Jaeger,K.E. (2007) Chembiochem, 8, 55– 60.
47
