Scholarly article on topic 'Enhancement of antibody fragment secretion into the Escherichia coli periplasm by co-expression with the peptidyl prolyl isomerase, FkpA, in the cytoplasm'

Enhancement of antibody fragment secretion into the Escherichia coli periplasm by co-expression with the peptidyl prolyl isomerase, FkpA, in the cytoplasm Academic research paper on "Biological sciences"

CC BY
0
0
Share paper
Academic journal
Journal of Immunological Methods
OECD Field of science
Keywords
{Chaperones / Bacteria / Cytoplasm / "Phage display"}

Abstract of research paper on Biological sciences, author of scientific article — Raphael Levy, Kiran Ahluwalia, David J. Bohmann, Hoa M. Giang, Lauren J. Schwimmer, et al.

Abstract Improper protein folding or aggregation can frequently be responsible for low expression and poor functional activity of antibody fragments secreted into the Escherichia coli periplasm. Expression issues also can affect selection of antibody candidates from phage libraries, since antibody fragments displayed on phage also are secreted into the E. coli periplasm. To improve secretion of properly folded antibody fragments into the periplasm, we have developed a novel approach that involves co-expressing the antibody fragments with the peptidyl prolyl cis-trans isomerase, FkpA, lacking its signal sequence (cytFkpA) which consequently is expressed in the E. coli cytosol. Cytoplasmic expression of cytFkpA improved secretion of functional Fab fragments into the periplasm, exceeding even the benefits from co-expressing Fab fragments with native, FkpA localized in the periplasm. In addition, panning and subsequent screening of large Fab and scFv naïve phage libraries in the presence of cytFkpA significantly increased the number of unique clones selected, as well as their functional expression levels and diversity.

Academic research paper on topic "Enhancement of antibody fragment secretion into the Escherichia coli periplasm by co-expression with the peptidyl prolyl isomerase, FkpA, in the cytoplasm"

Contents lists available at SciVerse ScienceDirect

Journal of Immunological Methods

journal homepage: www.elsevier.com/locate/jim

Research paper

Enhancement of antibody fragment secretion into the ^cmssMark

Escherichia coli periplasm by co-expression with the peptidyl prolyl isomerase, FkpA, in the cytoplasm^

Raphael Levy *, Kiran Ahluwalia, David J. Bohmann, Hoa M. Giang, Lauren J. Schwimmer, Hassan Issafras, Nithin B. Reddy, Chung Chan, Arnold H. Horwitz, Toshihiko Takeuchi

Preclinical Research and Development, XOMA Corp., Berkeley, CA 94710, United States

ARTICLE INFO ABSTRACT

Improper protein folding or aggregation can frequently be responsible for low expression and poor functional activity of antibody fragments secreted into the Escherichia coli periplasm. Expression issues also can affect selection of antibody candidates from phage libraries, since antibody fragments displayed on phage also are secreted into the E. coli periplasm. To improve secretion of properly folded antibody fragments into the periplasm, we have developed a novel approach that involves co-expressing the antibody fragments with the peptidyl prolyl cis-trans isomerase, FkpA, lacking its signal sequence (cytFkpA) which consequently is expressed in the E. coli cytosol. Cytoplasmic expression of cytFkpA improved secretion of functional Fab fragments into the periplasm, exceeding even the benefits from co-expressing Fab fragments with native, FkpA localized in the periplasm. In addition, panning and subsequent screening of large Fab and scFv naïve phage libraries in the presence of cytFkpA significantly increased the number of unique clones selected, as well as their functional expression levels and diversity.

© 2013 The Authors. Published by Elsevier B.V. All rights reserved.

Article history:

Received 30 November 2012 Received in revised form 8 April 2013 Accepted 12 April 2013 Available online 23 April 2013

Keywords: Chaperones Bacteria Cytoplasm Phage display

1. Introduction

Several groups have attempted with varying degrees of success to improve bacterial production of antibody fragments

Abbreviations: Fab, fragment-antigen-binding; Fc, fragment crystalliz-able; Fd, fragment of antibody consisting of CH1 and VH; ScFv, single-chain variable fragment; VH, variable heavy; VL, variable light; PPIase, peptidyl prolyl cis-trans isomerase; PCR, polymerase chain reaction; IPTG, isopropyl (J-D-1-thiogalactopyranoside; OD, optical density; PBS, phosphate buffer saline; M.O.I., multiplicity of infection; PEG, polyethylene glycol; HRP, horseradish peroxidase; ELISA, enzyme-linked immunosorbent assay; TMB, 3,3',5,5'-tetramethylbenzidine; huINSR, human insulin receptor; BSA, bovine serum albumin; kd, dissociation constant; RT, room temperature; EC50, half maximal effective concentration; CFU, colony-forming units; SD, Shine-Dalgarno sequence; PVDF, polyvinylidene fluoride; SPR, Surface Plasmon Resonance; SDS, sodium dodecyl sulfate; SEM, Standard Error of the Mean.

☆ This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

* Corresponding author. Tel.: +1 510 204 7574; fax: +1 510 841 7805.

E-mail address: levy@xoma.com (R. Levy).

by co-expressing them with molecular chaperones or folding catalysts (Bothmann and Pluckthun, 1998; Strachan et al., 1999; Bothmann and Pluckthun, 2000; Levy et al., 2001; Mavrangelos et al., 2001; Maynard et al., 2005). The correct folding of scFv and Fab antibody fragments is highly dependent on the activity of peptidyl prolyl cis-trans isomerases (PPlases). Following the formation of variable and constant domain intra-chain disulphide bonds, peptidyl prolyl cis-trans isomer-ization reactions drive folding into the native conformation, allowing formation of the interchain disulphide bonds. PPIases also prevent aggregation of antibody fragments (Feige et al., 2010). Kappa light chain variable domains (Vk) contain two conserved prolines in the cis conformation at positions L8 and L95 (Bothmann and Pluckthun, 2000) unlike the frameworks of heavy chain variable (VH) and lambda light chain variable (V\) antibody domains which, based on evaluation of sequences in the PDB database, do not contain any cis-prolines (Horne and Young, 1995). Cis-trans isomerization at Pro-L95 is a rate limiting step in the folding of Vk domains and is essential for VL/VH docking and therefore for native protein conformation

0022-1759/$ - see front matter © 2013 The Authors. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/! 0.1016/j.jim.2013.04.010

(Suominen et al., 1987; Knappik and Pluckthun, 1995; Forsberg et al., 1997; Ramm and Pluckthun, 2000). Interestingly, co-expression of the periplasmic Escherichia coli PPIase, FkpA, resulted in a significant improvement in secretion into the bacterial periplasm of functional scFv fragments containing either Vk chains, which contain cis prolines, or V\ chains which do not contain cis-prolines, suggesting that it has both molecular chaperone and PPIase enzymatic activities (Bothmann and Pluckthun, 2000). Employing FkpA deletion mutants and functional assays, Saul et al. (2004) established that the FkpA carboxy and amino terminal domains carry independent PPIase and chaperone activities, respectively.

Previously, Missiakas et al. (1996) demonstrated that FkpA can act as a "global folding catalyst" that limits the levels of unfolded proteins in the outer membrane and periplasm. Periplasmic overexpression of FkpA facilitates the expression of multiple heterologous proteins, including an E. coli maltose binding protein misfolding mutant (Arie et al., 2001), single-chain antibodies and antibody fusions (Arie et al., 2001; Zhang et al., 2003; Padiolleau-Lefevre et al., 2006; Sonoda et al., 2010).

Another molecular chaperone in the E. coli periplasm is the 17 kDa Skp protein which forms a trimer with a central cavity. This cavity allows Skp to engulf native polypeptide substrates and prevents their subsequent aggregation (Walton et al., 2009). Co-expression of Skp with a poorly soluble single chain Ab resulted in its secretion into the E. coli periplasm as well as improved solubility and phage display of the antibody fragment and diminished the toxicity of the antibody for the host cells (Hayhurst and Harris, 1999). As observed with FkpA, other groups have demonstrated that co-expression of scFvs with Skp increased their secretion in E. coli (Sonoda et al., 2010). Previously, it also was shown that overexpression of Skp lacking its signal sequence significantly improved the yield of a correctly folded Fab produced by a trxBgor mutant E. coli strain that enables the production of disulphide bonds in the bacterial cytoplasm (Levy et al., 2001).

We report here improvement in functional Fab expression into the E. coli periplasm as a result of its co-expression with FkpA lacking a signal sequence (cytFkpA). The secretion of active Fabs into the periplasm was higher when co-expressed with cytFkpA either on a separate vector under control of an L-arabinose-inducible promoter, or as part of a tricistronic message that includes the chaperone, Fd and light chains on a single plasmid. We also examined the effect of cytFkpA expression on selection of scFv or Fab candidates from large phage libraries and have demonstrated increased expression levels and diversity of displayed antibodies targeting the selected antigens, resulting in selection of a larger number of functional, sequence-unique antibody fragments with slower dissociation constants.

2. Materials and methods

2.1. Strains

XL1-Blue cells (recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F proAB lacIqZAM15 Tn10 (Tetr)]) andTG1 cells (supE thi-1 A(lac-proAB) A(mcrB-hsdSM)5 (rK-mK-) [F' traD36proAB lacIqZAM15]) were purchased from Agilent (Santa Clara, CA).

2.2. Plasmids expressing bacterial chaperones

In order to generate the plasmids responsible for cytoplasmic expression of chaperones, the native signal sequences were excised from the genes encoding the chaperones FkpA (Swiss-Prot accession no. P65764) and Skp (Swiss-Prot accession no. P0AEU7). Chaperones were also allowed to express in the bacterial periplasm with their native signal sequences. To generate the plasmid constructs of the cytoplasmic or periplasmic versions of the chaperones Skp and FkpA, and the bicistronic Skp-FkpA, the chaperone gene fragments were amplified by PCR and then cloned into the plasmid vector pAR3 (ATCC accession no. 87026). The vector pAR3 (Perez-Perez and Gutierrez, 1995) contains the pBAD promoter and the cat gene which confers chloram-phenicol antibiotic resistance. This plasmid harbors the p15A origin of replication which is compatible with the origin ColE1 included in all the vectors co-expressing Fabs or scFvs in our experiments. Two different forward primers and one reverse primer were designed in order to amplify FkpA from XL-1Blue cells by PCR amplification with or without the native leader peptide. Similarly, two forward primers and one reverse primer were designed to amplify Skp from XL1Blue cells by PCR with or without its native signal sequence.

To generate the chaperone plasmid constructs pAR3-FkpA and pAR3-Skp for periplasmic expression and pAR3-cytFkpA and pAR3-cytSkp for cytoplasmic expression, the products of the previous PCR reactions were used as templates for PCR re-amplification using forward primers to incorporate a BglII restriction site followed by the enhancer sequence GAATTCA TTAAAGAGGAGAAATTAACT upstream from the chaperone encoding gene fragment. Reverse primers were used to incorporate the V5 tag sequence (GGTAAGCCTATCCCTAACC

CTCTCCTCGGTCTCGATTCTACG) into pAR3-Skp and pAR3-cytSkp

and the FLAG tag sequence (GACTACAAGGACGATGACGACAAG) into the pAR3-FkpA and pAR3-cytFkpA, followed by the restriction site HindIII.

To generate the bicistronic periplasmic pAR3-[Skp + FkpA] and cytoplasmic pAR3-cyt[Skp + FkpA] constructs, the monocistronic PCR products were reamplified. To reamplify Skp, forward primers were used to incorporate BglII, followed by the enhancer GAATTCATTAAAGAGGAGA AATTAACT and the periplasmic or cytoplasmic versions of Skp. A reverse primer was designed that anneals to the entire V5 sequence and to an optimized Shine-Dalgarno (SD) sequence driving the translation initiation of FkpA. To reamplify FkpA, forward primers were designed to anneal to the C-terminal portion of V5, the optimized SD, and the periplasmic or cytoplasmic versions of FkpA. A reverse primer was designed to add a HindIII-FLAG tag sequence to the C-terminal portion of FkpA. The Skp and FkpA PCR products were then gel-purified using Qiagen gel extraction kits (Valencia, CA) and used as templates for an overlap extension PCR reaction using the external forward Skp primer and an external reverse FkpA primer. Fig. 1a illustrates the resulting chaperone inserts. Ligations of the BglII-HindIII digested PCR products to the BglII-HindIII digested pAR3 vector were then transformed by electroporation into XL1-Blue cells. The final constructs were confirmed by DNA sequencing.

Fig. 1. Chaperone constructs utilized in this study. (a) Monocistronic vectors expressing the chaperones cytSkp, cytFkpA and cyt[Skp + FkpA] in the E. coli cytoplasm and Skp, FkpA, and [Skp + FkpA] expressed in the periplasm. Gene constructs were cloned into the multi-cloning site of the arabinose-inducible vector pAR3 (Perez-Perez and Gutierrez, 1995) between the BglII and HindIII restriction sites. (b) Tricistronic vector expressing the light chain (VL-CL) and heavy chain (VH-CH) of the human anti-EpCAM Fab ING1 with a C-terminal 6His-cmyc-V5 Tag (HMV) in the periplasm via DNA signal sequences (ss) together with the cytFkpA that carries a C-terminal FLAG tag. The tricistronic construct was then cloned into a lac-inducible phagemid described in the Materials and methods section. The ends of open reading frames are depicted by open circles.

2.3. Vectors expressing Fab fragments

All Fabs and scFv fragments used in this work were cloned into proprietary phagemid vectors (Schwimmer et al., 2013) harboring a triple 6His-cmyc-V5 tag, the beta lactamase gene conferring ampicillin resistance and the ColE1 origin of replication that is compatible with the p15A origin of the pAR3 vector (backbone for chaperone expressing vectors). The Fabs with kappa light chains used were: a) XPA23 (anti-IL1(3, human), b) ING1 (anti-EpCAM, human), c) 83-7 (anti-human insulin receptor (huINSR), murine), d) CF1 (anti-Tie-1, human), and e) BM7-2 (anti-kinase, human). We also tested the lambda light chain containing gastrin-specific Fabs C10, D1, and E6. The antigens huINSR, Tie-1-Fc, and IL1(3 were purchased from R&D Systems (Minneapolis, MN).

2.4. Tricistronic vector expressing ING1 Fab and cytFkpA

In the phagemid vector expressing the ING1 Fab in the E. coli periplasm under the influence of the lac promoter, the gene fragment encoding the M13 phage pIII protein was replaced with cytFkpA. In order to produce the tricistronic construct (Fig. 1b), two non-amber stop codons were added following the triple detection tag 6His-cmyc-V5. The gene fragment cytFkpA was amplified by PCR from the vector pAR3-cytFkpA, together with an upstream enhancer sequence and C-terminal FLAG tag. The final construct was cloned into the modified phagemid expressing the ING1 Fab.

2.5. Preparation of cytoplasmic and periplasmic E. coli extracts

TG1 electroporation-competent cells were co-transformed with the Fab expressing vectors that were previously described along with one of the chaperone-expressing vectors pAR3-FkpA, pAR3-Skp, pAR3-cytFkpA, pAR3-cytSkp, or the bicistronic pAR3-[Skp + FkpA], and pAR3-cyt[Skp + FkpA]. Co-expression of

the Fab-expressing vectors with the empty plasmid pAR3 served as a negative control.

In order to prepare cytoplasmic and periplasmic extracts expressing the chaperones alone, TG1 cells carrying the empty pAR3 plasmid (negative control) or the chaperone-expressing plasmid constructs were grown overnight at 37 °C in 2YT growth media supplemented with 34 |ag/ml chloramphenicol and 2% (w/v) glucose until the OD600 reached 0.5. The expression of chaperones was then induced with 0.2% arabinose (w/v) at 30 °C overnight. At that point, the OD600 was recorded and cultures were normalized to the same OD600. Cells were pelleted and resuspended in 10 ml ice-cold PPB buffer (30 mM Tris-HCl, pH 8.0, 1 mM EDTA, 20% sucrose) (Teknova, CA) at 1:4 dilution. Following incubation at 4 °C for 1 h, samples were centrifuged for 30 min and supernatants containing the periplasmic extracts were collected. Pellets were resuspended in 10 ml BugBuster® solution (Novagen, NJ) supplemented with one tablet of complete EDTA-free protease inhibitor cocktail (Roche, IN) and 2500 units benzonase nuclease (Novagen) in order to reduce the viscosity of the lysates. Following 1 hour incubation in ice, lysates were centrifuged at 16,000 g for 20 min at 4 °C and supernatants containing the cytoplasmic extracts were collected.

To prepare periplasmic extracts of cells expressing Fabs together with the chaperones, TG1 cells harboring the Fab and chaperone plasmid constructs (or pAR3 alone as negative control) were grown overnight at 37 °C in 2YT growth media supplemented with 34 ^g/ml chloramphenicol, 100 ^g/ml carbenicillin and 2% (w/v) glucose and subcultured in 100 ml flasks at 37 °C until the OD600 reached 0.5. Thirty minutes after the addition of 0.2% arabinose (w/v), isopropyl (3-D-1-thiogalactopyranoside (IPTG) was added to a final concentration of 1 mM and cultures were incubated overnight at 30 °C. At that point the OD600 was recorded and cultures were normalized to equal OD600. Cells were pelleted and resuspended in 10 ml ice-cold PPB sucrose buffer (Teknova) at 1:4 dilution and one tablet ofcomplete EDTA-free protease

inhibitor cocktail (Roche). Following incubation at 4 °C for 1 h, samples were centrifuged for 30 min and the superna-tants containing the periplasmic extracts were collected. Similarly, periplasmic extracts from TG1 cells expressing the ING1 Fab and cytFkpA from a single tricistronic vector were generated without chloramphenicol selection (only with carbenicillin) and simultaneous induction of ING1 Fab and cytFkpA with 1 mM IPTG.

2.6. Western blotting

Samples of periplasmic and cytoplasmic extracts were resuspended in SDS loading buffer with 0.7 M beta-mercaptoethanol, boiled and loaded in NuPAGE® 4-12% Bis-Tris precast gels (Invitrogen, CA) using NuPAGE MOPS SDS running buffer (Invitrogen). Proteins from reduced gels were then transferred to PVDF membranes using the Millipore-SNAP-i.d.® electroblotter (Millipore, CA). The membranes were blocked with 0.5% BSA/PBS and incubated for 15 min at room temperature with mouse anti-V5 antibodies (Sigma, MO) at 1:2000 dilution for the detection of Skp or with murine anti-FLAG primary antibodies (Sigma) at 1:1000 dilution in 0.5% BSA/PBS for the detection of FkpA. Subsequently, primary antibodies were detected with goat anti-mouse IgG (H + L) conjugated with horseradish peroxidase (HRP) (Jackson Immunoresearch, PA) at a 1:2000 dilution. Color was developed with 1-Step TMB-Blotting substrate solution (Pierce, IL).

2.7. Target and expression ELISAs

The amount of functional Fab binding to target antigens was determined by ELISA. Ninety six-well high binding MaxiSorp® assay plates (Nunc, NY) were coated with 1-3 |ag/ml antigen diluted in phosphate buffer saline (PBS). EpCAM (bound by ING-1 Fab), IL1(3 (bound by XPA23 Fab) and Tie-1-Fc (bound by CF1 Fab) antigens were coated at 3 |ag/ml. Kinase (bound by BM7-2 Fab) was coated at 2 |ag/ml. Human insulin receptor (huINSR) (bound by 83-7 Fab) was coated at 1 ng/ml. Biotinylated gastrin (a 14-mer peptide recognized by the C10, D1, and E6 Fabs) was coated at 1 |ag/ml in PBS on Reacti-Bind Streptavidin-coated 96-well plates (Thermo Scientific, MN).The coated plates were then incubated overnight at 4 °C and blocked with 5% non-fat dry milk (Nestlé, OH) in PBS buffer (no blocking was required for the streptavidin-coated plates). Plate washes were carried out in PBS with 0.05% TWEEN®-20. Dilutions of Fabs, and primary and secondary antibodies were performed in 5% non-fat dry milk in PBS. Fabs were allowed to bind to their blocked antigens for 1 h at room temperature. The presence of ING1, XPA23, CF1, BM7-2, C10, D1, and E6 Fabs was confirmed with goat-anti-human IgG [specific for F(ab')2] (Jackson Immunoresearch) at 1:2000 dilution, followed by donkey anti-goat IgG (H + L) conjugated with HRP (Santa Cruz Biotechnology, CA) at 1:10,000 dilution. The 83-7 Fab was detected using rabbit-anti-mouse IgG [specific for F(ab')2] (Jackson Immunoresearch) antibodies at 1:2000 dilution, followed by goat anti-rabbit IgG (H + L) conjugated with horseradish peroxidase (Jackson Immunoresearch) at 1:10,000 dilution. The assay was developed with TMB soluble substrate (EMD Chemicals, CA). The reaction was quenched with 4.5 N H2SO4 and read at 450 nm using a SpectraMax® Plus microplate reader (Molecular Devices, CA). The amount of total Fab

expressed in the periplasm was determined by ELISA. For the detection of ING1, XPA23, BM7-2 and CF1 human kappa Fabs, high binding MaxiSorp 96-well plates were coated with 3 |ag/ml goat-anti-human kappa IgG (Invitrogen) diluted in PBS. Similarly, the murine kappa 83-7 Fab was detected with 3 ^g/ml goat-anti-mouse kappa antibodies (Jackson Immunoresearch) and the human lambda C10, D1, and E6 Fabs with 3 ^g/ml goat-anti-human lambda IgG (Pierce). Coated plates were incubated, blocked and washed, as previously described. Fabs were detected using rabbit anti-V5 (Sigma) primary antibody at 1:2000 dilution, followed by goat anti-rabbit IgG (Fc-specific) conjugated with HRP (Jackson Immunoresearch) at 1:10,000 dilution. The development of the assay was performed as previously described. ELISA data represent two biological replicate experiments.

2.8. Quantitation of Fab by surface plasmon resonance

Quantitation of human Fab in periplasmic extracts was performed by Surface Plasmon Resonance (SPR) on a Biacore A4000 or Biacore 2000 instrument (GE Healthcare, NJ). A standard curve was generated by diluting human Fab (Jackson Immunoresearch) in two-fold serial dilutions into assay running buffer and used for the estimation of Fab concentrations (Supplementary methods, tables and figures). Fab standard and unknowns were injected over a goat-anti-human IgG [specific for F(ab')2] surface (Jackson Immunoresearch) immobilized at a high density on a Biacore CM5 Sensor chip (GE Healthcare). Data analysis was performed using the BIAevaluation software (GE Healthcare).

2.9. Phage panning and screening of scFv and Fab libraries

For the first round of phage panning using a naïve scFv library (Schwimmer et al., 2013), 4.7 x 1013 cfu of phage particles from a scFv kappa library or 1.6 x 1013 cfu of phage particles from a Fab lambda library combined with 2.2 x 1013 cfu of phage particles from a Fab kappa library were blocked for 1 h at RT in 5% non-fat dry milk (Marvel Premier Foods, UK) in PBS buffer with gentle rotation. Blocked phage was twice deselected for 45 min against streptavidin-coated magnetic Dynabeads® M-280 (Invitrogen Dynal AS, Oslo, Norway). The kinase antigen (R&D Systems, MN) that was used for the scFv panning, and Tie-2 antigen (R&D Systems) that was used for the Fab panning, were biotinylated with Sulfo-NHS-LC-Biotin (Pierce) using the manufacturer's protocol in 20-fold molar excess of biotin reagent and confirmed by ELISA. The biotinylated products were then incubated with blocked streptavidin-coated magnetic Dynabeads® M-280 for 1 h with gentle rotation in order to immobilize biotinylated antigen and remove unbiotinylated material. Antigen-captured beads were then washed twice with PBS. For the first, second and third rounds of selection, 100, 50 and 10 pmol of biotinylated kinase or Tie-2 were used, respectively. For the first round, the deselected phage was divided into two aliquots: one was used to infect TG1 cells, and the other was used to infect TG1 cells harboring pAR3-cytFkpA. The rescued, deselected phage was used to perform parallel first round pannings by incubation with biotinylated kinase streptavidin beads for 90 min at room temperature. The input phage for rounds two and three was generated with separate

rescues from either the round one TG1 infection or the round one TG1 with pAR3-cytFkpA.

For the first round of panning, beads were washed quickly (i.e., beads were pulled out of solution using a magnet and resuspended in 1 ml wash buffer) three times with PBS—0.05% TWEEN®-20, followed by three times with PBS. For the second round of panning, beads were washed for five times with PBS— 0.05% Tween followed by one 5-minute wash. Similar washes were performed with PBS. For the third round of panning, beads were washed quickly for 4 times with PBS—0.05% TWEEN®-20 followed by four 5-minute washes with PBS—0.05% TWEEN®-20. Similar washes were performed with PBS. Antigen-bound phage was eluted via incubation for 30 min with 100 mM triethylamine (TEA) at room temperature and subsequently neutralized with 1M Tris-HCl (pH 7.4). The phage eluted from each round of panning was used to infect either TG1 alone or TG1 with pAR3-cytFkpA cells when the OD60o was equal to 0.5. TG1 cells were grown in 2YT media and TG1 cells expressing cytFkpA (from plasmid pAR3-cytFkpA) were grown in 2YT media supplemented with 34 ^g/ml chloramphenicol.

Following infection for 1 h at 37 °C, TG1 cells were centrifuged and pellets were resuspended in 2YT growth media supplemented with 100 ^g/ml carbenicillin and 2% (w/v) glucose. Resuspended cells were then plated onto 2YT agar plates containing 100 ^g/ml carbenicillin and 2% glucose and incubated overnight at 30 °C. Similarly, TG1 cells expressing the chaperone cytFkpA were plated onto 2YT agar plates with 100 ^g/ml carbenicillin, 34 ^g/ml chloram-phenicol and 2% glucose and incubated overnight at 30 °C.

Phage was rescued with helper phage M13K07 at a multiplicity of infection (MOI) ~20. For this purpose, first and second round selection output clones were allowed to grow to an OD600 ~ 0.5. At that point, cells were infected with helper phage at 37 °C for 1 h while shaking at 100 rpm. Cell pellets were resuspended in 2YT media supplemented with 100 ^g/ml carbenicillin, 50 ^g/ml kanamycin and 0.2% (w/v) arabinose (only for TG1 cells expressing cytFkpA), and allowed to grow overnight at 25 °C. Phage in the supernatant was recovered by centrifugation and used for the next round ofpanning. In order to estimate the enrichment resulting from the phage selections, the amount of input and output phage was titered and plated on 2YT agar plates supplemented with the appropriate antibiotics.

Clones were picked in a 96 well plate from third round output and grown in 2YT media supplemented with carbenicillin, 0.1% glucose with or without chloramphenicol at 37 °C for expression. Induction was done by adding 1 mM IPTG. Clones from the chaperone panning output were first induced with 0.2% arab-inose for 30 min at 30 °C for the expression ofcytFkpA, followed by overnight induction with 1 mM IPTG at 30 °C.

The generation ofperiplasmic extracts was done as described above and ELISA screening was performed using kinase or biotinylated Tie-2 coated on MaxiSorp plates and Reacti-Bind™ streptavidin-coated 96-well plates, respectively. Only kinase-coated plates needed to be blocked for 1 h with 5% non-fat milk prepared in PBS. The scFv or Fab fragments were allowed to bind for 1 h and 30 min to the antigen. Detection was enabled using murine anti-V5 antibodies (1:2000) followed by goat-anti-mouse HRP (1:10,000) (Thermoscientific, Rockford, IL). The development and quenching of ELISAs were done as described earlier herein.

2.10. Evaluation of Fab display on the surface ofM13 phage

One liter cultures of E. coli carrying phagemid vectors expressing either lambda or kappa Fab libraries (Schwimmer et al., 2013) in TG1 cells or TG1 cells harboring the plasmid pAR3-cytFkpA were initiated at OD600 = 0.1. These four cultures were grown with shaking at 250 rpm until the OD600 reached 0.5 in 2YT media supplemented with 2% glucose (w/v) and 100 |ag/ml carbenicillin. Chloramphenicol (34 ^g/ml) was also added to cells carrying the pAR3-cytFkpA plasmid. The cells were then infected with M13K07 helper phage at an MOI of 20 for 1 h at 37 °C; 30 min without shaking and 30 min with shaking at 100 rpm. After infection, the media was changed to 2YT supplemented with 100 ^g/ml carbenicillin, 50 ^g/ml kanamycin, and the TG1/pAR3-cytFkpA cultures also had 34 |ag/ml chloramphenicol and 0.2% (w/v) arabinose to allow expression of cytFkpA. Samples (50 ml) were taken from each culture 25 h after the start of the infection with helper phage. These cultures were centrifuged and the supernatant was heated to 60 °C to eliminate bacteria. The samples taken at 25 h were precipitated with polyethylene glycol in order to concentrate the phage. The concentrated phage was stored in 15% glycerol at — 80 °C. Serial dilutions of these samples were made in 3% non-fat dry milk in PBS and applied for 1 h at RT to blocked MaxiSorp plates that had been coated with anti-M13 antibodies (GE Healthcare) at 1:1000 dilution in PBS or murine anti-V5 antibodies (Sigma) at 1:2000 dilution in PBS. The phage was detected with anti-M13 antibodies conjugated with HRP (GE Healthcare) at 1:5000 dilution in 3% milk/PBS for 1 h at RT. The assay was developed by the addition of TMB soluble substrate (KPL, MD). The reaction was quenched with 2N H2SO4 and read at 450 nm by a SpectraMax® Plus microplate reader. The EC50 for each set of dilutions was calculated by fitting a sigmoidal dose response curve using GraphPad Prism®. The relative level of Fab display was calculated by dividing the inverse of the EC50 from the anti-V5 ELISA (binding the V5-tag indicates the presence of a Fab molecule displayed on a phage) by the inverse of the EC50 from the anti-M13 ELISA and comparing each ratio to the ratio calculated for the 25 hour time point of the rescue in TG1 cells. This method is described by Soltes et al. (2003).

2.11. Determination of antibody fragment dissociation rates using SPR

Antibody fragment screening for dissociation rate was performed on a Biacore 4000. Fab fragments were screened on ligand covalently coupled to a CM5 Series S biosensor (GE Healthcare) via amine chemistry. Tie-2 was immobilized at varying surface densities on spots 1 and 2 of the biosensor. Blank spot three was used for reference subtraction. ScFv fragments were screened utilizing capture methodology. ScFv capture utilized monoclonal anti-V5 antibodies (Sigma). The capture antibody was immobilized on a CM5 Series S biosensor by standard amine coupling. Amine coupling was performed by activating the chip with EDC/NHS (GE Healthcare) for 5 min and injecting either Tie-2 or anti-V5 at 5 ^g/ml in pH4.5 acetate (GE Healthcare) for 5 min. Deactivation was performed with 1 M ethanolamine.

Periplasmic extracts of anti-Tie-2 Fabs were diluted with equal volumes of running buffer (HBS-EP (GE Healthcare),

10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.05% Surfactant P20 at pH 7.4 with 1 mg/ml BSA (Sigma)), and filtered through 0.22 ц multiscreen GV filter plates (Millipore). Filtered periplasmic extracts were injected over immobilized ligand for 3 min at 30 ^l/min. Dissociation was followed for 10 min. The surface was regenerated following each analyte injection with 10 mM glycine at pH 1.7. Data was double referenced by subtracting the reference spot within the flow cell which was an activated and deactivated blank surface, as well as subtracting out blank injections. Following referencing, the data were fit to a 1:1 dissociation model using Biacore 4000 evaluation software.

30 kDa

cytFkpA-FLAG

3. Results

3.1. Expression ofSkp and FlqpA in the E. coli cytoplasm

17 kDa

• cytSkp-V5

To express Skp and FkpA in the E. coli cytoplasm, DNAs encoding these chaperones lacking their signal sequences and containing V5 and FLAG tags, respectively, were amplified by PCR from the XL1-Blue E. coli chromosome. The gene products, designated as cytSkp-V5 and cytFkpA-FLAG, were cloned into the L-arabinose-inducible expression vector, pAR3 (Perez-Perez and Gutierrez, 1995) either separately (Fig. 1a), or as a bicistronic gene sequence cyt[Skp + FkpA] encoding both cytFkpA and cytSkp (Fig. 1b) for expression in the E. coli cytoplasm. Vectors were also constructed containing Skp and FkpA with their native signal sequences for expression in the E. coli periplasm (Fig. 1a). Plasmids containing cytSkp-V5 and/or cytFkpA-FLAG, were transformed into E. coli TG1 cell cultures, grown to log phase, induced with L-arabinose, and periplasmic and cytoplasmic extracts prepared. Western blot analysis using anti-V5 and anti-FLAG tag antibodies verified that cytSkp and cytFkpA expressed on the same or separate plasmids were produced in the cytoplasm ofTG1 cells (Fig. 2, Lanes 3 and 5). Lower amounts of cytSkp and cytFkpA also were observed in an E. coli periplasmic extract (Fig. 2, Lanes 2 and 4) which may be due to escape of the chaperones through the inner membrane during the generation of the extracts. The two bands that appear upon overexpression of cytSkp in E. coli (Fig. 2b) could be attributed to an incomplete processing of Skp corresponding to the precursor and mature forms of Skp. Other scientists have previously demonstrated similar results when probing Skp using anti-Skp antisera (Volokhina et al., 2011).

3.2. Enhancement by chaperones of Fab secretion into the E. coli periplasm

We first tested the effect of co-expressing FkpA and Skp on secretion into the bacterial periplasm of Fabs containing kappa light chains. Initially, two human kappa Fabs, ING1 (anti-EpCAM) and XPA23 (anti-IL1ß) and a murine antihuman insulin receptor kappa Fab, 83-7 (Soos et al., 1986) were expressed in TG1 cells in the presence or absence of cytoplasmically or periplasmically-expressed FkpA or Skp, either alone or in combination. The level of Fab in the periplasm capable of binding to EpCAM and IL1ß was assessed by ELISA. The results of this experiment demonstrate that the expression of functional ING1 and XPA23 Fabs was increased when co-expressed with cytFkpA (Fig. 3a, b, fifth dark gray column from the left). By contrast, with the exception of the condition

Fig. 2. Western blots of soluble cytoplasmic (lanes 3, 5) or periplasmic fractions (lanes 2, 4) of TG1 E. coli cells expressing cytFkpA or cytSkp alone (lanes 2, 3) or together (lanes 4, 5) from the bicistronic construct cyt [Skp + FkpA]. (a) Detection of cytFkpA was performed with murine anti-FLAG mAbs. (b) Detection of cytSkp was performed with murine anti-V5 anti-sera. Whole extract of TG1 cells not expressing cytSkp or cytFkpA was used as negative control (lane 1).

in which it was co-expressed with cytFkpA, most of the XPA23 Fab expressed with or without chaperones was non-functional, as evidenced by the low amount of binding in the target-specific ELISA (ELISA absorbance at 450 nm was less than 0.1). The amount of functional murine 83-7 Fab expressed in the periplasm, assessed by target ELISAs (Fig. 3c, dark gray columns) was improved when co-expressed with cytFkpA (Fig. 3c, fifth set of columns from the left). Since the above results demonstrated that co-expression with cytFkpA and, in very few cases, cyt[Skp + FkpA] provided the greatest benefit for Fab secretion, we evaluated the effects of these chaperones on two additional human kappa Fabs, BM7-2 and CF1, which bind a human tyrosine kinase and Tie-1, respectively. Total and functional amounts of BM7-2 or CF1 Fab in the periplasm were measured by expression (Fig. 4, light gray columns) and target (Fig. 4, dark gray columns) ELISAs, respectively. The cytFkpA chaperone construct improved the functional BM7-2 and CF1 Fab expression (Fig. 4a and b, respectively), but to a lesser extent than XPA23 or ING1 Fabs.

Unlike kappa light chains, lambda light chains do not contain framework proline residues in the cis conformation. Since in addition to its catalytic proline isomerization activity, FkpA functions as a molecular chaperone, we measured levels of total and functional gastrin-specific Fabs, C10, D1, and E6, which contain lambda light chains, co-expressed with cytFkpA or cyt[Skp + FkpA]. The benefit of cytFkpA expression on secretion of functional Fabs containing lambda light chains was less apparent than with kappa Fabs in that C10, D1, and E6 Fab periplasmic expression did not benefit from co-expression with cytFkpA (Fig. 5). It appears that simultaneous expression of cytSkp and cytFkpA decreased the expression of C10, D1, and E6 Fabs (Fig. 5) possibly due to negative influence of Skp expression in the bacterial cytoplasm.

Fab expression also can be quantified by SPR by first capturing Fab fragments with anti-human Fab antiserum

¡¡■<0 =5 2

s= w 4 n ® 4

o > ^ 1 o o

2 <0 o

-9 ^ a a

o o "u £

1 . i li -L 1 1

^ / ^ Z5 ^ </ </

5 2 5 1 5 0

II i i i

* ^ -V^V J* J*

, 1, 1111

/ J* ^v* </ </

¿r Ax<f /vX <f ^

* ** # /

. jO «

Fig. 3. Effect of co-expression of cytoplasmically produced chaperones cytSkp and cytFkpA generated alone or together (cyt[Skp + FkpA]) or effect of their periplasmically expressed counterparts Skp, FkpA, and [Skp + FkpA] on the accumulation of (a) ING1, (b) XPA23, and (c) 83-7 Fabs in the periplasm of E. coli TG1 cells. Cells harboring only the empty plasmid pAR3 were used as negative control. Cells were grown and induced as described in the Materials and methods section. The concentration of functional Fabs (dark gray columns) and the total amount of soluble periplasmic Fab (light gray columns) were monitored by ELISA. Binding is represented as fold over absorbance (450 nm) of Fab alone with empty vector pAR3, in the absence of chaperones. Error bars represent SEM values.

immobilized on a Biacore sensor chip. For this study, we tested levels of Fab in the periplasm upon co-expression with the chaperone constructs that generated more substantial expression improvements based on ELISA results. To quantify Fab levels, a standard curve was generated using a control human Fab; periplasmic Fab concentrations were estimated

2 2.5 ffi

a. 0 a 2

■■g 0.5 0

Fig. 4. Effect of co-expression of the cytoplasmically produced chaperone cytFkpA generated alone or together with cytSkp (cyt[Skp + FkpA]) on the accumulation of the BM7-2 Fab (a) or the CF1 Fab (b) in the periplasm of TG1 cells. Cells harboring only the empty plasmid pAR3 were used as negative controls. Cells were grown and induced as described in the Materials and methods section. The concentration of the total amount of soluble periplasmic Fab (light gray columns) and functional (dark gray columns) Fabs was monitored by ELISA. Binding is represented as fold over absorbance (450 nm) of Fab alone with empty vector pAR3, in the absence of chaperones. Error bars represent SEM values.

.£¡8 120 p 1.00 -£3 <D

i- o 0.60 g -

(3 o 0.40 ~o t= o 5 0.20

ill ill Hi

f J* f J* jf ,,<f jP J?

Fig. 5. Effect of co-expression of the cytoplasmically produced chaperone cytFkpA generated alone or together with cytSkp (cyt[Skp + FkpA]) on the accumulation of the anti-gastrin lambda chain containing C10, D1, and E6 Fabs in the periplasm of TG1 cells. Cells harboring only the empty plasmid pAR3 were used as negative controls. Cells were grown and induced as described in the Materials and methods section. The concentration of the total amount of soluble periplasmic Fab (light gray columns) and functional (dark gray columns) Fabs was monitored by ELISA. Binding is represented as fold over absorbance (450 nm) of Fab alone with empty vector pAR3, in the absence of chaperones.

based on SPR resonance units (RUs) in relation to the standard curve (see Table 1). Since the kappa Fab fragments used in this study share identical constant regions, the affinity of the secondary antibodies used to detect the Fabs should be very similar. Cytoplasmic expression of cytFkpA resulted in 5.3 to 7.6-fold and 5.5 to 26-fold increases in production of XPA23 and ING1 (kappa) Fabs, respectively, in the periplasm compared to the same antibodies without cytFkpA. As observed by ELISA (Fig. 4), expression of the CF1 kappa Fab benefited to a lesser extent (1.7 to 2-fold) from expression of cytFkpA.

3.3. Co-expression of cytFkpA and Fab on a single plasmid

A tricistronic vector (Fig. 1b) was developed for co-expressing the ING1 Fd and light chains in the periplasm along with cytFkpA under control of the lac promoter. Western blot analysis confirmed that most of the cytFkpA was expressed in the cytoplasm (data not shown). Accumulation of total and functional Fabs in the periplasm, assessed by expression and target ELISAs, was improved when co-expressed with cytFkpA (Fig. 6a), thus establishing the usefulness of incorporating cytFkpA along with Fd and light chains as a tricistronic unit in the expression vector. We also confirmed by SPR that total periplasmic ING1 Fab was increased by co-expressing with cytFkpA from a single vector in the E. coli cytoplasm (Fig. 6b). Yields of periplasmic soluble Fab ranged from 0.4 to 2.45 ^g/ml without cytFkpA and 3.514.2 |ag/ml in the presence of cytFkpA.

3.4. Expression of cytFkp increases hit diversity and expression from panning phage display libraries

Since co-expression of cytFkpA enhances expression in the E. coli periplasm of functional Fabs with kappa (and some lambda) light chains, we examined the effects of co-expressing cytFkpA on selection of antigen-specific Fab or scFv fragments from naive phage display libraries. Three rounds of phage panning were performed with biotinylated target (kinase) using a large kappa scFv library (Schwimmer et al., 2013). Following the third round of panning, clones were picked for evaluation of scFv expression in the periplasm. Periplasmic extracts were also tested for binding to kinase. Panning was performed with or without expression of cytFkpA from a separate arabinose-inducible vector (pAR3) containing a p15A origin of replication which is compatible with the library

CD != 3.00 с P

Ö со 2.00

-Q О CO — LL 3

g 10.00 m

с ® 8.00

ING1 ING1+cytFkpA

8 ® ф i^

со о

ING1 ING1+cytFkpA

Fig. 6. Effect of co-expression of cytFkpA in the cytoplasm of TG1 cells on the accumulation of ING1 Fab in the periplasm when both proteins are expressed from a single tricistronic vector under the influence of the lac promoter. (a) The concentrations of the total amount of soluble periplasmic Fab (light gray columns) and functional (dark gray columns) amounts of periplasmic Fab were monitored by ELISA in the presence or absence of cytFkpA. Binding is represented as fold over absorbance (450 nm) of Fab alone, in the absence of chaperones. Growth and induction conditions of TGlcells were described in the Materials and methods section. (b) Relative amounts of total ING1 Fab were estimated by SPR using the Biacore BIAevaluation software and expression is represented as fold over concentration of ING1 Fab in the absence of chaperones. Periplasmic extracts of TG1 cells served as a negative control. Biacore was performed on periplasmic extracts of ING1 Fab captured by F(ab')2-specific anti-human IgG immobilized on the surface of a CM5 sensor chip. Error bars represent SEM values.

phagemid vector that carries the lac promoter and harbors the ColE1 origin of replication.

Ninety three output clones were selected after the third round of phage panning performed with or without cytFkpA

Table 1

Estimate of periplasmic Fab protein yields measured by Surface Plasmon Resonance (SPR).a

Fab Experiment # Light chain Chaperone co-expressed Periplasmic yield (no chaperone) (|g/ml) Periplasmic yield (with chaperone) ( | g/ml) Improvement (%)

XPA23 1 Kappa cytFkpA 0.74 5.52 646

XPA23 2 Kappa cytFkpA 0.30 2.03 577

ING1 1 Kappa cytFkpA 0.02 0.10 400

ING1 2 Kappa cytFkpA 0.13 0.55 323

CF1 1 Kappa cytFkpA 1.44 2.39 66

CF1 2 Kappa cytFkpA 0.81 1.50 85

a Calculation of Fab yields was based on Fab capture by anti-human IgG (Fab specific) immobilized on a high density Biacore A4000 or 2000 chip, as described in the Materials and methods section. Concentration measurements were based on standard curves made with serial dilutions of a human Fab. Experiment 1 was performed in singlicate using Biacore A4000. Experiment 2 represents two independent measurements (A and B) performed in triplicate using three flow cells on Biacore 2000. Experiments 1 and 2 used two different periplasmic extract preparations (details in Supplementary figures and tables). Variations of expression improvements between different experiments maybe attributed to the low accuracy of Biacore measurements when Fab concentrations are very low or to different efficiencies of periplasmic extract preparations.

expression. While scFv clones selected from panning campaigns without cytFkpA were induced only with IPTG, clones selected from panning with cytFkpA also were induced with L-arabinose to allow cytFkpA expression. The amount of functional scFv in the bacterial periplasmic extracts in the absence and presence of cytFkpA was assessed by ELISA. Overexpression of cytFkpA significantly increased both the frequency and expression levels of sequence-unique clones obtained by panning with a scFv phage display library containing kappa light chains (Table 2). Only 10% of the output clones selected from panning without cytFkpA were sequence-unique and antigen-specific, with an ELISA signal (OD450) greater than 3-fold over the background, compared to 48% of clones selected when cytFkpA was co-expressed. Thus, the diversity of the selected kinase-binding clones, as defined by the number of sequence-unique clones and their expression levels, was greatly improved in the presence of cytFkpA.

To confirm the cytFkpA expression benefit for phage hit diversity and expression, two arms of panning with and without cytFkpA were performed against the Tie-2 antigen using a Fab phage display library containing similar kappa and lambda light chain representation (Schwimmer et al., 2013). Periplasmic extracts of 93 selected clones from each of the panning arms were screened by ELISA for binding to Tie-2. Hits from panning with cytFkpA appeared to express much better, as 43% of the output clones were sequence unique and generated an ELISA signal greater than 3-fold above background (including 16% at more than 12-fold over background); without cytFkpA expression, only 16% of the output clones bound to Tie-2 with a signal greater than 3-fold over the background (including 5% more than 12-fold above background) (Table 2). Thus, panning with cytFkpA also enhanced the diversity of the Tie-2-specific selected Fab clones, generating a higher number of sequence-unique and better expressing clones. In the presence or absence of cytFkpA, the percentage of kappa vs. lambda Tie-2 binding clones remained virtually unchanged (30% kappa and 70% lambda). However, there was a 2.5-3.3-fold increase of the number of both kappa and lambda-containing sequence-unique Tie-2 binding clones in the presence of cytFkpA (4 kappa and 11 lambda clones without expression of cytFkpA, as opposed to 13 kappa and 27 lambda clones in the presence of cytFkpA).

3.4.1. Effect of cytoplasmic cytFkpA expression on Fab fragment display on phage

We compared the Fab fragment display levels on M13 phage in the presence or absence of cytFkpA expression. E. coli TG1 cell cultures expressing a Fab phagemid library with lambda or kappa light chains were allowed to express with or without cytFkpA. Following growth and induction, phage were rescued and precipitated after 25 h to assess Fab display levels. The relative amount of phage was estimated through phage capture of serial dilutions on ELISA plates coated with M13-specific polyclonal antibodies and detection with anti-M13 antibodies conjugated with HRP. Fab display levels of the M13-captured dilutions were determined using anti-V5 antibodies that recognize the C-terminal V5 tag on the displayed Fab fragments. The ratio of the inverse EC50 of the two ELISAs provided a direct indication of the number of phage displaying Fab fragments. Comparing the ratios of phage rescue (Fig. 7) clearly showed that rescues performed with both kappa and lambda libraries in the presence of cytFkpA had greater than 3.5-fold increase in display than rescues performed in the absence of cytFkpA.

3.4.2. Effect of cytFkpA expression on dissociation rates of selected clones

Since co-expression with cytFkpA facilitates improved selection of unique functional clones from phage display libraries, we evaluated the dissociation kinetics of scFv or Fab clones selected by phage display against the kinase (Fig. 8a) and Tie-2 targets (Fig. 8b) using SPR. Periplasmic extracts of anti-kinase scFv or anti-Tie-2 Fab fragments were allowed to bind to Biacore chips coated with anti-V5 antibodies (for kinase detection) or Tie-2 ligand, respectively. The results of this experiment (Fig. 8) demonstrate that co-expression of cytFkpA during phage panning allowed selection of a greater number of sequence-unique clones with slower dissociation rates (kd values lower than 10-3 1/s are represented by columns higher than the horizontal dashed line in Fig. 8). This effect was most pronounced for anti-kinase candidates from the scFv phage display kappa-only library where 8 of 26 clones selected via panning in the presence of cytFkpA had kd values lower than 10-3 1/s, as compared to none panned without cytFkpA.

Table 2

Selection of scFv or Fab antibody candidates with the presence or absence of cytFkpA.a

Antibody Library cytFkpA Unique binding clones Total unique

clones panned (+/-) over background (%) binding clones

3-fold 6-fold 12-fold

Kinase scFv - 9 1 0 10

scFv + 26 22 0 48

Tie-2 Fab - 8 3 5 16

Fab + 9 18 16 43

a Binding of periplasmic extracts to kinase or biotin-Tie-2 antigens was measured by ELISA, as described in the Materials and methods section. Clones were selected following panning against kinase or biotinylated-Tie-2 targets using a scFv phage library containing only kappa light chains or a Fab phage library with both kappa and lambda light chains. Panning and screening were performed in the absence or presence of cytFkpA.

■ I ■ I

K25 TG1 K25 L25 TG1 L25

TG1+cytFkpA TG1+cytFkpA

Fig. 7. Relative display of kappa (K) and lambda (L) Fabs on phage produced in TG1 cells with or without cytFkpA. Samples were taken 25 (K25, L25) hours after helper phage infection. Values are displayed relative to kappa or lambda library rescues in TG1 cells, without cytFkpA, at 25 h after infection.

1.0E-04

1.0E-03

1.0E-02

1.0E-04

1.0E-03

1.0E-02

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

1.0E-01

No cytFkpA

With cytFkpA

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51

Fig. 8. Dissociation rates (kd) of sequence-unique scFv or Fab library clones selected by phage display against kinase using a scFv phage library (a) or Tie-2 using a Fab library (b) in the presence (dark gray columns) or absence (light gray columns) of cytFkpA. Periplasmic extracts of anti-kinase scFv or anti-Tie-2 Fab clones were allowed to bind to Biacore sensor chips capturing anti-V5 antibodies or Tie-2 ligand, respectively. A horizontal dashed line emphasizes clones with kd values lower than 10-3 s-1. Clones with too-fast-to-measure kd values are not shown.

4. Discussion

In studies reported here, we demonstrate that cytoplasmic expression of the periplasmic chaperone, peptidyl prolyl cis-trans isomerase, FkpA, improved the secretion of functional antibody fragments into the bacterial periplasm. The benefit of the cytoplasmically-expressed chaperone exceeded that of native FkpA expressed in the periplasm. We also report an improvement in periplasmic expression of functional Fab for all the tested antibody fragments containing kappa light chains co-expressed with cytoplasmic FkpA. By comparison, cytoplas-mic expression of FkpA did not benefit secretion of antibody fragments containing lambda light chains to the same degree. The results observed with antibody fragments containing kappa versus lambda light chains suggest that the PPIase activity of cytFkpA may play a major role in the enhancement of functional antibody fragment secretion into the periplasm since only kappa light chains carry cis-prolines.

Pluckthun et al. have shown that co-expression of full-length FkpA in the periplasm can also improve the yields of scFv fragments without cis-prolines (i.e. lambda light chains), establishing the usefulness of the FkpA molecular chaperone

domain (Ramm and Pluckthun, 2000). Our results show that the molecular chaperone activity of cytSkp by itself is unable to improve secretion of functional ING1 and XPA23 Fab fragments, suggesting that the cis-trans peptidyl prolyl isomerase activity of cytFkpA in the E. coli cytoplasm is instrumental for these kappa antibody fragments prior to their transport into the oxidizing environment of the periplasm where formation of intra-chain and interchain disulphide bonds takes place. Thus, the cytoplasmic localization of the PPIase activity of cytFkpA could explain our results. In our studies, we employed leader peptides that enable periplasmic localization of antibody fragments through the Sec secretion pathway. The Sec route allows transmembrane transport of proteins into the periplasm in a loosely folded or unfolded, translocationally competent state or insertion into the cytoplasmic membrane through the SecYEG apparatus (Natale et al., 2008; Dalal and Duong, 2011; Zalucki et al., 2011). We can speculate that cytFkpA isomerizes key prolines of kappa light chains prior to the periplasmic export, and by doing so removes a folding bottleneck following the translocation of the heterologously expressed polypeptides into the periplasm. To further validate our hypothesis, additional studies should be performed utilizing a variant of

cytFkpA that lacks the chaperone activity but retains the domain with peptidyl prolyl isomerase activity. Such a variant would have to be tested to determine whether cytoplasmic expression still confers the beneficial secretion-enhancing effects of full-length cytFkpA.

As a consequence of the inability of overexpressed heterol-ogous proteins to fold properly in a timely fashion, misfolded proteins can be deposited in the form of cytoplasmic or periplasmic inclusion bodies or they can be driven towards degradation (Georgiou et al., 1986; Betton et al., 1998; Baneyx and Mujacic, 2004) Therefore, we isolated insoluble fractions of E. coli cells expressing XPA23 or ING1 Fabs, in the absence of cytFkpA, but we were unable to detect any Fab species by Western blot analysis (unpublished data), suggesting that no Fab was localized in inclusion bodies. Thus, we cannot support the notion that co-expression of cytFkpA increases the amount of functional Fab by means of improving its solubility. We hypothesize that misfolded or unfolded antibody fragment species serve as substrates for proteolytic degradation, instead ofassociating into inclusion bodies.

We also demonstrate that co-expression ofcytFkpA together with the kappa light chain-containing ING1 Fab expressed on a single tricistronic vector results in an improvement of functional Fab secretion relative to expression in the absence of cytFkpA. Similarly, it previously was shown that the amounts of single chain antibodies expressed in the periplasm of E. coli upon the co-expression of Skp were also increased when expression of both proteins was driven from a dicistronic vector (Hayhurst and Harris, 1999).

After observing the benefit of cytFkpA co-expression on Fab secretion, we evaluated its contribution to the antibody discovery process by incorporating the same expression platform with cytFkpA into phage panning selection and screening assays. The isolation of ideal lead candidates requires the design of methodologies allowing efficient screening of the libraries and exploitation of the vast repertoire of different library members. The choice of antibody formats (mostly scFv and Fab), the protein expression yields, the sequence diversity, the levels of display (i.e. on phage or yeast), and the ease and quality of in vitro screening are just a few of the factors that can impact the quality of antibody libraries (Mondon et al., 2008). In fact, it can be increasingly challenging to design screening assays that allow the identification of high-affinity library members and distinguish them from high-expressing clones since they are both able to display efficiently. Poorly expressed, functional library members are underrepresented and as a consequence, fail to be selected during screening. Thus, it is of paramount importance to maximize the solubility and functional expression ofantibody library members.

Our studies revealed that co-expression of cytFkpA in the bacterial cytosol improves the level of display of Fab or scFv phage fragments, thus promoting discovery of antibody candidates that otherwise would not be selected. The enhancement of phage display likely drives selection of a more diverse repertoire of target-binding clones, as we observed experimentally, which may lead to the discovery of higher affinity clones with the desired functional properties. The distribution of off-rate values did not differ following selections in the presence ofcytFkpA. However, the larger number ofsequence-unique antibody clones that we discovered in the presence of cytFkpA could increase the probability of selecting clones with

higher affinity. In contrast to our observations with expression of individual Fab fragments, phage panning in the presence of cytFkpA improved the diversity of both lambda and kappa scFv and Fab libraries, resulting in a higher number of antigen-binding clones with unique sequences and/or improved dissociation constants (Table 2 and Fig. 8). This improvement can be attributed to the elevated numbers of phage displaying antibody fragments that we observed in the presence of cytFkpA expression (Fig. 7). Since the improvement in diversity was observed for selection ofboth lambda and kappa antibody fragments, we conclude that both the peptidyl prolyl isomerase and molecular chaperone activities of cytFkpA are important contributors to selection of antibodies from phage libraries.

In our work, we obviate the use of mutant bacterial strains by expressing chaperones in the bacterial cytoplasm while we continue to express antibody fragments in the periplasm, which has frequently served as the standard milieu for heterologous protein expression in E. coli. Previous studies have co-expressed chaperones in the E. coli cytoplasm (e.g. the trigger factor which is a PPIase, like FkpA, that possesses distinct molecular chap-erone and enzymatic activities) (Hesterkamp and Bukau, 1996; Lee and Bernstein, 2002) and improved the production of antibody fragments in the cytoplasm of E. coli. However, in these cases, expression had to be limited to the oxidative cytoplasm of trxBgor mutant E. coli strains to allow the formation of the disulphide bonds of the antibody fragments (Levy et al., 2001; Heo et al., 2006). Overexpression of cytoplasmic chaperones (i.e. the GroES/L chaperonin system, DnaKJE) (Duenas et al., 1994; Hu et al., 2007) or periplasmic chaperones (i.e. FkpA) (Bothmann and Pluckthun, 2000; Ramm and Pluckthun, 2000) in their natural E. coli environment (cytoplasm or periplasm) has been employed successfully to enhance the production yields of functional scFv fragments and has been extensively reviewed in the past (Wall and Pluckthun, 1995; Kolaj et al., 2009). In contrast to these studies, our work reports the use of cytFkpA and cytSkp in the E. coli cytoplasm, while their native counterparts, FkpA and Skp, are normally localized in the bacterial periplasmic space.

We also demonstrate that co-expression of cytFkpA in the cytoplasm improves the functional protein yields of the anti-EpCAM ING1 and anti-IL1(3 XPA23 Fab fragments in the periplasm. When expressed alone, these Fabs express poorly (Table 1). Low periplasmic expression can be attributed to cell toxicity issues often resulting from poor translocation across the inner E. coli membrane and/or aggregation in the peri-plasm. Therefore, our results are consistent with previous studies that showed more apparent beneficial effects of FkpA on the functional expression of toxic scFv antibody fragments (Bothmann and Pluckthun, 2000). Interestingly, a recent study suggested that overproduction of FkpA, and to a lesser extent Skp, in E. coli enhances the viability of cells by elevating the expression of genes encoding heat-shock proteins or proteins leading to responses to misfolded protein stress (Ow et al., 2010). It remains to be seen if the cell viability is also improved when cytFkpA is co-expressed in the bacterial cytoplasm. The same group reported that co-production of FkpA together with Skp in the periplasm not only increases the solubility and secretion of a scFv to the extracellular medium, but also improves the cell viability.

A major advantage of our approach is that the native sequences of Fabs or scFvs do not have to be altered. This

approach is in contrast to previous efforts that employ protein engineering techniques to optimize the sequence of antibody fragments by either introducing mutations to increase their solubility (i.e. by generating cysteine-free mutants allowing expression in the cytoplasm without the requirement for refolding) (Proba et al., 1998; Worn and Pluckthun, 1998), or by using fusion proteins (Bach et al., 2001; Jurado et al., 2006).

In conclusion, co-expression of the chaperone variant, cytFkpA, offers multiple benefits over alternative approaches for the selection of novel antibody candidates or the optimization of production of existing antibody fragments. Based on the results reported here, the novel expression platform we describe in this work is a useful tool for phage display and recombinant antibody manufacturing.

Acknowledgments

We would like to thank Diane Wilcock for her critical reading of this manuscript.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/jjim.2013.04.010.

References

Arie, J.P., Sassoon, N., et al., 2001. Chaperone function of FkpA, a heat shock prolyl

isomerase, in the periplasm of Escherichia coli. Mol. Microbiol. 39 (1), 199. Bach, H., Mazor, Y., et al., 2001. Escherichia coli maltose-binding protein as a molecular chaperone for recombinant intracellular cytoplasmic single-chain antibodies. J. Mol. Biol. 312 (1), 79. Baneyx, F., Mujacic, M., 2004. Recombinant protein folding and misfolding in

Escherichia coli. Nat. Biotechnol. 22 (11), 1399. Betton, J.M., Sassoon, N., et al., 1998. Degradation versus aggregation of misfolded maltose-binding protein in the periplasm of Escherichia coli. J. Biol. Chem. 273 (15), 8897. Bothmann, H., Pluckthun, A., 1998. Selection for a periplasmic factor improving phage display and functional periplasmic expression. Nat. Biotechnol. 16 (4), 376.

Bothmann, H., Pluckthun, A., 2000. The periplasmic Escherichia coli peptidylprolyl cis, trans-isomerase FkpA. I. Increased functional expression of antibody fragments with and without cis-prolines. J. Biol. Chem. 275 (22), 17100. Dalal, K., Duong, F., 2011. The SecY complex: conducting the orchestra of

protein translocation. Trends Cell Biol. 21 (9), 506. Duenas, M., Vazquez, J., et al., 1994. Intra- and extracellular expression of an scFv antibody fragment in E. coli: effect of bacterial strains and pathway engineering using GroES/L chaperonins. Biotechniques 16 (3), 476 (480-473).

Feige, M.J., Hendershot, L.M., et al., 2010. How antibodies fold. Trends Biochem. Sci.35 (4), 189.

Forsberg, G., Forsgren, M., et al., 1997. Identification of framework residues in a secreted recombinant antibody fragment that control production level and localization in Escherichia coli. J. Biol. Chem. 272 (19), 12430. Georgiou, G., Telford, J.N., et al., 1986. Localization of inclusion bodies in Escherichia coli overproducing beta-lactamase or alkaline phosphatase. Appl. Environ. Microbiol. 52 (5), 1157. Hayhurst, A., Harris, W.J., 1999. Escherichia coli skp chaperone coexpression improves solubility and phage display of single-chain antibody fragments. Protein Expr. Purif. 15 (3), 336. Heo, M.A., Kim, S.H., et al., 2006. Functional expression of single-chain variable fragment antibody against c-Met in the cytoplasm of Escherichia coli. Protein Expr. Purif. 47 (1), 203. Hesterkamp, T., Bukau, B., 1996. The Escherichia coli trigger factor. FEBS Lett. 389 (1), 32.

Horne, S.M., Young, K.D., 1995. Escherichia coli and other species of the Enterobacteriaceae encode a protein similar to the family of Mip-like FK506-binding proteins. Arch. Microbiol. 163 (5), 357. Hu, X., O'Hara, L., et al., 2007. Optimisation of production of a domoic acid-binding scFv antibody fragment in Escherichia coli using molecular

chaperones and functional immobilisation on a mesoporous silicate support. Protein Expr. Purif. 52 (1), 194.

Jurado, P., de Lorenzo, V., et al., 2006. Thioredoxin fusions increase folding of single chain Fv antibodies in the cytoplasm of Escherichia coli: evidence that chaperone activity is the prime effect of thioredoxin. J. Mol. Biol. 357 (1), 49.

Knappik A., Pluckthun, A., 1995. Engineered turns of a recombinant antibody improve its in vivo folding. Protein Eng. 8 (1), 81.

Kolaj, O., Spada, S., et al., 2009. Use of folding modulators to improve heterologous protein production in Escherichia coli. Microb. Cell Fact. 8,9.

Lee, H.C., Bernstein, H.D., 2002. Trigger factor retards protein export in Escherichia coli. J. Biol. Chem. 277 (45), 43527.

Levy, R., Weiss, R., et al., 2001. Production of correctly folded Fab antibody fragment in the cytoplasm of Escherichia coli trxB gor mutants via the coexpression of molecular chaperones. Protein Expr. Purif. 23 (2), 338.

Mavrangelos, C., Thiel, M., et al., 2001. Increased yield and activity of soluble single-chain antibody fragments by combining high-level expression and the Skp periplasmic chaperonin. Protein Expr. Purif. 23 (2), 289.

Maynard, J., Adams, E.J., et al., 2005. High-level bacterial secretion of single-chain alphabeta T-cell receptors. J. Immunol. Methods 306 (1-2), 51.

Missiakas, D., Betton, J.M., et al., 1996. New components of protein folding in extracytoplasmic compartments of Escherichia coli SurA FkpA and Skp/ OmpH. Mol. Microbiol. 21 (4), 871.

Mondon, P., Dubreuil, O., et al., 2008. Human antibody libraries: a race to engineer and explore a larger diversity. Front. Biosci. 13,1117.

Natale, P., Bruser, T., et al., 2008. Sec- and Tat-mediated protein secretion across the bacterial cytoplasmic membrane—distinct translocases and mechanisms. Biochim. Biophys. Acta 1778 (9), 1735.

Ow, D.S., Lim, D.Y., et al., 2010. Co-expression of Skp and FkpA chaperones improves cell viability and alters the global expression of stress response genes during scFvD1.3 production. Microb. Cell Fact. 9, 22.

Padiolleau-Lefevre, S., Debat, H., et al., 2006. Expression of a functional scFv fragment of an anti-idiotypic antibody with a beta-lactam hydrolytic activity. Immunol. Lett. 103 (1), 39.

Perez-Perez, J., Gutierrez, J., 1995. An arabinose-inducible expression vector, pAR3, compatible with ColE1-derived plasmids. Gene 158 (1), 141.

Proba, K., Worn, A., et al., 1998. Antibody scFv fragments without disulfide bonds made by molecular evolution. J. Mol. Biol. 275 (2), 245.

Ramm, K., Pluckthun, A., 2000. The periplasmic Escherichia coli peptidylprolyl cis, trans-isomerase FkpA. II. Isomerase-independent chaperone activity in vitro. J. Biol. Chem. 275 (22), 17106.

Saul, F.A., Arie, J.P., et al., 2004. Structural and functional studies of FkpA from Escherichia coli, a cis/trans peptidyl-prolyl isomerase with chaperone activity. J. Mol. Biol. 335 (2), 595.

Schwimmer, L.J., Huang, B., et al., 2013. Discovery of diverse and functional antibodies from large human repertoire antibody libraries. J. Immunol. Methods 391 (1-2), 60.

Soltes, G., Barker, H., et al., 2003. A new helper phage and phagemid vector system improves viral display of antibody Fab fragments and avoids propagation of insert-less virions. J. Immunol. Methods 274 (1-2), 233.

Sonoda, H., Kumada, Y., et al., 2010. Functional expression of single-chain Fv antibody in the cytoplasm of Escherichia coli by thioredoxin fusion and co-expression of molecular chaperones. Protein Expr. Purif. 70 (2), 248.

Soos, M.A., Siddle, K., et al., 1986. Monoclonal antibodies reacting with multiple epitopes on the human insulin receptor. Biochem. J. 235 (1), 199.

Strachan, G., Williams, S., et al., 1999. Reduced toxicity of expression, in Escherichia coli, of antipollutant antibody fragments and their use as sensitive diagnostic molecules. J. Appl. Microbiol. 87 (3), 410.

Suominen, I., Karp, M., et al., 1987. Extracellular production of cloned alpha-amylase by Escherichia coli. Gene 61 (2), 165.

Volokhina, E.B., Grijpstra, J., et al., 2011. Role of the periplasmic chaperones Skp, SurA, and DegQ in outer membrane protein biogenesis in Neisseria meningitidis. J. Bacteriol. 193 (7), 1612.

Wall, J.G., Pluckthun, A., 1995. Effects of overexpressing folding modulators on the in vivo folding of heterologous proteins in Escherichia coli. Curr. Opin. Biotechnol. 6 (5), 507.

Walton, T.A., Sandoval, C.M., et al., 2009. The cavity-chaperone Skp protects its substrate from aggregation but allows independent folding of substrate domains. Proc. Natl. Acad. Sci. U. S. A. 106 (6), 1772.

Worn, A., Pluckthun, A., 1998. Mutual stabilization of VL and VH in single-chain antibody fragments, investigated with mutants engineered for stability. Biochemistry 37 (38), 13120.

Zalucki, Y.M., Beacham, I.R., et al., 2011. Coupling between codon usage, translation and protein export in Escherichia coli. Biotechnol. J. 6 (6), 660.

Zhang, Z., Song, L.P., et al., 2003. Production of soluble and functional engineered antibodies in Escherichia coli improved by FkpA. Biotechniques 35(5), 1032 (1041-1032).