Báo cáo khoa học: Selection of full-length IgGs by tandem display on filamentous phage particles and Escherichia coli fluorescence-activated cell sorting screening doc

We report here a system that takes advan-tage of display of full-length IgGs on filamentous phage particles as a prescreening step to reduce library size and enable subsequent rounds of F

filamentous phage particles and Escherichia coli

fluorescence-activated cell sorting screening

Yariv Mazor1,2, Thomas Van Blarcom1,2, Sean Carroll1,2and George Georgiou1,2

1 Institute for Cellular and Molecular Biology, University of Texas at Austin, TX, USA

2 Department of Chemical Engineering, University of Texas at Austin, TX, USA

Introduction

Recombinant antibodies have made a tremendous

impact on biomedical research, and are increasingly

being used as clinical diagnostic and therapeutic

reagents [1,2] Consequently, the demand for new

tech-nologies that aid in the discovery and selection of

novel therapeutic antibodies has never been greater

During the past two decades, several display

technolo-gies and other library screening techniques have been

developed for the isolation of antigen-specific

antibod-ies from large collections of recombinant antibody

genes [3] Phage display is the most prevalent method for the display of large ensembles of antibody frag-ments, and is currently considered to be the standard procedure in many molecular biology laboratories for antibody discovery and evolution [3] Unique antibod-ies are isolated from immune [4–7], naı¨ve [8–13] or synthetic [14–21] repertoires, and are further engi-neered for improved affinities for their antigens by using the selected antibody gene as the basis for subse-quent libraries and screening [22–26] Humira [27,28]

Keywords

fluorescence-activated cell sorting (FACS);

full-length IgG; fUSE5–ZZ phage; protective

antigen (PA); spheroplasts

Correspondence

G Georgiou, Department of Chemical

Engineering, University of Texas at Austin,

Austin, TX 78712, USA

Fax: +1 512 471 7963

Tel: +1 512 471 6975

E-mail: gg@mail.che.utexas.edu

(Received 7 February 2009, Revised 4

March 2010, accepted 9 March 2010)

doi:10.1111/j.1742-4658.2010.07645.x

Phage display of antibody libraries is a powerful tool for antibody discov-ery and evolution Recombinant antibodies have been displayed on phage particles as scFvs or Fabs, and more recently as bivalent F(ab¢)2 We recently developed a technology (E-clonal) for screening of combinatorial IgG libraries using bacterial periplasmic display and selection by fluores-cence-activated cell sorting (FACS) [Mazor Y et al (2007) Nat Biotechnol

25, 563–565] Although, as a single-cell analysis technique, FACS is very powerful, especially for the isolation of high-affinity binders, even with state of the art instrumentation the screening of libraries with diversity

> 108is technically challenging We report here a system that takes advan-tage of display of full-length IgGs on filamentous phage particles as a prescreening step to reduce library size and enable subsequent rounds of FACS screening in Escherichia coli For the establishment of an IgG phage display system, we utilized phagemid-encoded IgG with the fUSE5–ZZ phage as a helper phage These phage particles display the Fc-binding ZZ protein on all copies of the phage p3 coat protein, and are exploited as both helper phages and anchoring surfaces for the soluble IgG We demon-strate that tandem phage selection followed by FACS allows the selection

of a highly diversified profile of binders from antibody libraries without un-dersampling, and at the same time capitalizes on the advantages of FACS for real-time monitoring and optimization of the screening process

Abbreviations

CFU, colony-forming units; FACS, fluorescence-activated cell sorting; FITC, fluorescein isothiocyanate; HRP, horseradish peroxidase; IPTG, isopropyl thio-b- D -galactoside; PA, protective antigen; RU, response units; VH, variable heavy; VL, variable light.

was the first fully human mAb discovered using phage

display to receive FDA approval, and at least 16

human mAbs derived from phage display are currently

in advanced clinical trials for a wide range of human

diseases [29,30]

Phage display is based on encoding the gene of

interest in-frame with one of the phage coat proteins

(phenotype), and encapsulates the fusion gene within

the phage particle (genotype) Recombinant antibodies

have been displayed on phage particles as scFv [31] or

Fab [7,32] fragments These monovalent proteins,

although relatively easy to produce in Escherichia coli,

are typically devoid of avidity effects that allow the

recovery of low-affinity binders [33–36] Polyvalent

display of scFv and Fab can be readily achieved,

par-ticularly by using phage-based vectors [37] The

resul-tant avidity effects allow for the recovery of low

affinity-binders, but these same avidity effects make it

difficult to select stringently on the basis of intrinsic

affinity

More recently, bivalent Fab [F(ab¢)2] has been

dis-played on phage particles in a manner that

effec-tively resembles the binding behavior of natural IgGs

[36] Nevertheless, for the vast majority of diagnostic

and therapeutic applications, antibody fragments

iso-lated from most existing display technologies must

be converted to full-length IgG, the format of choice

in the clinics This process requires additional cloning

steps and the expression of the reformatted antibody

gene in mammalian cells A conspicuous drawback

of the scFv format is that reformatting to IgG can

result in loss of activity (TVB, YM, SAC, SK, BLI

and GG, unpublished observations) Yet another

dis-advantage of most existing phage display systems is

that the antibody gene is expressed as a fusion

pro-tein with one of the phage coat propro-teins As a result,

many of the antibodies isolated through library

screening can only fold in the context of a fusion

protein and cannot be expressed independently, a

phenomenon that many laboratories do not report

[38]

We recently reported the development of an E

coli-based technology, termed E-clonal, for the successful

production of soluble full-length IgGs in bacteria and

for screening of combinatorial IgG libraries using

bac-terial periplasmic display [39,40] Library cells

express-ing intact IgGs specifically labeled with fluorescently

conjugated antigen are readily distinguished and

iso-lated by fluorescence-activated cell sorting (FACS)

Unlike phage display, FACS has the distinct advantage

of relying on real-time quantitative multiparameter

analysis of individual cells, allowing single-cell

resolu-tion for selecresolu-tion Although FACS is a very powerful

high-throughput screening methodology, sorting a library > 109cells using FACS is time-consuming and challenging [40–43]

To reduce the initial library to a size that is manage-able by FACS and to demonstrate that full-length IgG libraries can be displayed on phage particles and undergo selection, we sought to develop a display sys-tem that will effectively display intact IgGs on filamen-tous phage particles The system can efficiently downsize very large libraries by employing an initial round of phage biopanning that specifically pre-enriches target cells from the library prior to subse-quent rounds of FACS Specifcally, E coli⁄ F¢ cells expressing soluble IgGs in the periplasm (E-clonal cells) are simultaneously infected with the fUSE5–ZZ phage [44] These phage particles allow polyvalent dis-play of the Fc-binding ZZ protein [45] on all five cop-ies of the gene-3 minor coat protein of filamentous bacteriophages [44] The fUSE5–ZZ phage in this sys-tem serves not only as a helper phage, but also as the IgG-capturing surface via the surface-displayed ZZ protein (Fig 1) Rescued fUSE5–ZZ–IgG phage parti-cles harboring the IgG phagemid are then selected for antigen binding by standard phage biopanning We describe here the feasibility of propagating phage parti-cles that stably display functional full-length IgGs, and demonstrate that an initial round of phage biopanning followed by FACS facilitates the isolation of a diversi-fied collection of antigen-specific binders from very large antibody libraries

Results

Model system validation

As a model for our studies and for validation of the display of full-length IgGs on phage particles, we chose two well-characterized antibodies, M18 and 26.10, which are specific for the protective antigen (PA) from Bacillus anthracis (Kd= 30 pm) and digoxin (Kd= 1.7 nm), respectively [39] The full-length heavy and light chain genes were expressed from a dicistronic operon and secreted into the periplasm, where they assembled into aglycosylated IgGs that were fully func-tional for antigen binding [39] To display the full-length IgG noncovalently on phage particles, we uti-lized fUSE5–ZZ, which displays the Fc-binding ZZ protein [45] on all five copies of the gene-3 minor coat protein, but maintains its ability to infect and propa-gate in E coli⁄ F¢ cells [44] (Fig 1) As the pMAZ360– IgG expression vector contains the packaging signal of f1 bacteriophage that enables the packaging of the plasmid as ssDNA in the presence of a helper phage,

rescued fUSE5–ZZ–IgG particles harbor the high copy

number pMAZ360–IgG phagemid preferentially over

the very low copy number and replication-defective

fUSE5–ZZ genome [44]

Initially, fUSE5–ZZ particles were evaluated for

their ability to capture purified IgGs in solution The

phage particles were mixed with purified M18 IgG or

26.10 IgG, washed to remove any unbound antibodies,

and analyzed by ELISA Phage that had captured the

purified IgG via the ZZ protein showed strong ELISA

signals with the respective antigens but not with

unre-lated antigens (data not shown) We then evaluated

whether fUSE5–ZZ is able to capture IgG within the

periplasm and form a noncovalent complex that is

sta-ble upon extrusion of the phage from the bacteria

E coliK91K⁄ F¢ cells transformed with phagemid

pMAZ360–M18–IgG or pMAZ360–26.10–IgG were

grown under conditions permissive for phage infection

Following infection with fUSE5–ZZ, the cultures were

allowed to grow overnight under conditions favorable

for phage production On the following day, phage

particles were precipitated and evaluated for specific

binding by direct ELISA (Fig 2) As expected,

fUSE5–ZZ propagated in cells expressing M18 IgG

bound specifically to PA (Fig 2A), whereas phage

par-ticles produced in cells expressing 26.10 IgG bound

specifically to digoxin (Fig 2B) Competition of the

bound IgG by an excess of standard human IgGs

resulted in a small reduction ( 15%) of ELISA signal

(Fig 2A,B), indicating that the phage ZZ–IgG

com-plex is kinetically very stable, presumably owing to the

polyvalent display of the ZZ protein on all copies of the phage p3 coat protein

To determine the number of IgG molecules dis-played on fUSE5-ZZ–IgG particles, we employed the technique described by Junutula et al [46] Purified fUSE5-ZZ–M18 IgG phage was applied at varying concentrations to ELISA plates coated with anti-M13 IgG, anti-Fc IgG, or PA Following incubation, the ELISA was developed with horseradish peroxidase (HRP)-conjugated anti-M13 IgG The number of IgG molecules displayed on each phage particle was deter-mined by the ratio of the linear range of the ELISA signals obtained with anti-Fc IgG⁄ anti-M13 IgG or

PA⁄ anti-M13 IgG (Fig 2C) Analysis of the results indicated that there is an average of 0.6–0.7 IgG mole-cules per phage particle

To assess the efficacy of the IgG phage display sys-tem for selections, we tested the ability to enrich fUSE5–ZZ–M18 IgG phage particles from a

1 000 000-fold excess of phage particles displaying the control 26.10 IgG The mixture was subjected to three rounds of phage biopanning against PA, and the enrichment after each round of selection was moni-tored by ELISA (Fig 2D) The increase in ELISA sig-nal for PA in parallel with the decrease in ELISA signal for digoxin clearly indicated a significant enrich-ment of the fUSE5–ZZ–M18 IgG phage population at the expense of a reduction in the number of phage par-ticles displaying the control 26.10 IgG Sequence anal-ysis revealed that, following the third round of phage panning, seven of 20 randomly selected clones carried

g3 ZZ

P

pMAZ360-IgG

pelB pelB

PLAC

Fd ori Tet R

Amp R F1ori

Fig 1 Schematic diagram of the IgG phage display format Left: map of phagemid pMAZ360–IgG for expression of soluble intact IgGs in the E coli periplasm This vector facilitates convenient cloning of VHand Vj domains linked to human c1 and j constant domains, respec-tively, as a bicistronic operon downstream of the lac promoter Center: map of fUSE5–ZZ phage for polyvalent display of the Fc-binding ZZ protein on all copies of the gene-3 minor coat protein Right: infection of E coli cells carrying phagemid pMAZ360–IgG with fUSE5–ZZ leads

to the production of fUSE5–ZZ–IgG phage particles that stably display functional full-length IgGs.

the M18 mAb sequence, corresponding to over

300 000-fold enrichment

Library selection by tandem phage biopanning and FACS

To evaluate the utility of tandem phage panning– FACS for library screening, we used an anti-PA mouse immune IgG library [39] This library was constructed

by cloning pools of variable heavy (VH) and variable light (VL) genes from spleens of mice that were immu-nized with PA from B anthracis into the pMAZ360– IgG vector Library DNA was transformed into

E coliK91K⁄ F¢ to generate a total of 107independent transformants Cells carrying phagemid pMAZ360– IgG were infected with fUSE5–ZZ, and the culture was allowed to grow under conditions favorable for phage production Phage particles were purified and subjected to an initial cycle of panning by incubation with soluble biotinylated PA in solution, before being captured on streptavidin-coated magnetic beads Unbound phage particles were removed by washing, and bound phage particles were eluted, neutralized, and used for infection of E coli Jude-1⁄ F¢ cells har-boring plasmid pBAD33–NlpA–ZZ for subsequent rounds of FACS screening in cells expressing an NlpA–ZZ fusion that allows capture of the IgG on the inner membrane Cells were converted to spheroplasts

by disruption of the outer membrane with Tris⁄ EDTA and lysozyme treatment, to allow exposure of the membrane-bound IgG to the extracellular fluid Two color flow cytometry steps, using PA63–fluorescein iso-thiocyanate (FITC) and Alexa Fluor 647–anti-(human IgG) to monitor for affinity and expression of full-length IgG, were employed, and fluorescent clones

100

25

50

75

0

fUSE5-ZZ-26.10 IgG

fUSE5-ZZ-26.10 IgG

50

75

100

0

25

fUSE5-ZZ-M18 IgG

fUSE5-ZZ-M18 IgG

fUSE5-ZZ-M18 IgG + hIgG competitor

fUSE5-ZZ-26.10 IgG + hIgG competitor

2

2.5

PA

anti-M13

1

1.5

2

A450 nm

0

0.5

Phage concentration (CFU·mL –1 )

1.5

2

2.5

fUSE5-ZZ-M18 IgG fUSE5-ZZ-26.10

0.5

1

1.5

A450 nm

0

1E+07 3E+07 1E+08 3E+08 1E+09 3E+09 1E+10 3E+10

Library Cycle 1 Cycle 2 Cycle 3

A

B

C

D

Fig 2 Characterization of fUSE5–ZZ–IgG phage Binding analysis

of fUSE5–ZZ–IgG phage in ELISA was tested with plates coated with PA (A) or digoxin ⁄ BSA (B) For analysis of the stability of the ZZ–IgG complex on phage particles displaying either M18 or 26.16 mAb, fUSE5–ZZ–IgG was incubated with 1 l M standard human IgG

as a competitor before being applied to the ELISA plates (C) Deter-mination of IgG molecules per phage fUSE5–ZZ–M18 IgG phage particles were serially diluted and applied to ELISA plates coated with either anti-M13 IgG or PA The number of IgG molecules per phage particle was determined by the phage concentration derived from PA ⁄ anti-M13 IgG in the linear range of the ELISA signals (D) Enrichment of fUSE5–ZZ–M18 IgG from a 1 000 000-fold excess of fUSE5–ZZ–26.10 IgG Phage biopanning against PA was performed

as described in Experimental procedures Evaluation of the enrich-ment following each round of phage selection was monitored by phage ELISA on plates coated with PA or digoxin ⁄ BSA ELISA plates were developed with HRP-conjugated anti-M13 IgG; values

at 450 nm represent three independent experiments.

falling into the double-positive quadrant were sorted

(Fig 3) For better selectivity and for the isolation of

clones exhibiting improved dissociation rates, cells

col-lected after the first sort were immediately resorted on

the flow cytometer As no additional probe was

pro-vided, only clones exhibiting low dissociation rates

sur-vive the second sorting cycle IgG genes in the sort

mixture were rescued by PCR amplification, recloned

into pMAZ360–IgG, transformed into fresh NlpA–ZZ

cells, and induced for IgG expression Following two

rounds of FACS selection, it was clear that IgGs

spe-cific for PA had been enriched (Fig 3), and IgG genes

from the second round of FACS selection were ligated

into the pMAZ360–IgG expression vector and

trans-formed into fresh E coli Jude-1 cells not carrying

plas-mid pBAD33–NlpA–ZZ for expression of soluble

antibodies Individual clones were grown in 96-well

plates and induced for expression of soluble IgG, and

PA-specific clones were identified by ELISA

Forty-nine of 192 of the screened clones gave a PA-specific

signal Sequence analysis of 25 clones revealed the

iso-lation of six unique clones that were subjected to

addi-tional characterization The selected clones were

expressed and purified by protein A affinity

chroma-tography, and IgG in yields of 0.5–3 mgÆL)1 were

obtained Biacore analysis of binding kinetics revealed

that the affinities of the IgGs derived from the

immu-nized library ranged from the low to high nanomolar (Table 1) The highest-affinity and lowest-affinity IgGs were determined to have KD values of 1 and 440 nm, respectively

Enrichment of high-affinity and moderate-affinity IgGs

To evaluate the utility of tandem phage biopanning followed by FACS for the isolation of IgGs with dif-ferent affinities from very large libraries, we used fUSE5–ZZ–IgG phage displaying either M18, YMF10

or VA IgG, displaying high-affinity binding (30 pm), moderate-affinity binding (30 nm) and no binding to the PA antigen The high-affinity and moderate-affinity IgGs displayed on the fUSE5-ZZ–IgG phage were each diluted 1 : 108in 1010copies of VA IgG displayed

on the fUSE5-ZZ–IgG phage The phage mixture was subjected to one round of phage biopanning, using sol-uble biotinylated PA antigen, and, following neutral-ization and infection of E coli Jude-1⁄ F¢ cells carrying plasmid pBAD33–ZZ, yielded an output of 1.5·

106cells harboring phagemid pMAZ360–IgG The cells were induced for the expression and display of IgG, and subjected to three rounds of FACS following labeling with fluorescently conjugated PA63 protein After the third round of FACS, plasmids encoding the isolated IgG gene inserts were transformed into fresh

E coli cells not carrying plasmid pBAD33–NlpA–ZZ for expression of soluble IgG and antigen specificity screening by ELISA Twenty-six of 186 of the screened clones were PA-specific binders Sequence analysis of the positive clones confirmed the selection of the high-affinity M18 and the moderate-high-affinity YMF10 To assess the enrichment of M18 and YMF10, phagemid was isolated from 2.5· 109cells from the pre-sort (post-phage biopanning) and after three rounds of

Table 1 Binding kinetics of isolated IgG determined by Biacore.

Antibody kon( M )1Æs)1) k

off (s)1) KD(n M )

PA-63-FITC (antigen binding)

10 2 10 3 10 4 10 5

Q4 Q3

Q2 Q1

10 2 10 3 10 4 10 5

Q4 Q3

Q2 Q1

10 2 10 3 10 4 10 5

Q4 Q3

Q2 Q1

10 2 10 3 10 4 10 5

Q4 Q3

Q2 Q1

Fig 3 Library selection using sequential phage biopanning and FACS Cells were labeled with PA–FITC and Alexa Fluor 647–anti-(human IgG) probes, and flow cytometry was used to monitor the progress of library selection by quantifying the percentage of fluorescent cells fall-ing into the double-positive quadrant, indicatfall-ing both expression and affinity The values in quadrant 2 refer to the percentage of double-posi-tive cells as a proportion of cells that express IgG (total cells minus quadrant 3).

sorting Five nanograms of purified phagemid from

each population was subjected to PCR amplification

with gene-specific primers, and the intensities of the

PCR products were determined by DNA

electrophore-sis (Fig 4) The densities of the respective bands

indi-cated significant enrichment of both clones Notably,

following one round of phage biopanning, a noticeable

PCR band was identified for clone YMF10 but not for

clone M18, despite there being equal numbers of

cop-ies of each clone in the initial mixture and in spite of

the much higher affinity of M18 This further

empha-sizes that library selection based on bivalent IgG

dis-play is not dictated by intrinsic affinity

Discussion

Recombinant antibodies are routinely displayed on

phages as scFv or Fab fragments The scFv is a

mono-mer consisting of the VH and VL gene fragments

con-nected by a peptide linker [47] This small protein of

25 kDa is displayed very efficiently on phages, both in

monovalent (single-copy) format when fused to the

gene-3 minor coat protein in a phagemid-based system

[31], and in multivalent (multiple-copy) format when

the scFv gene is fused to all five copies of gene-3 in a

phage-based system [37] Monovalent display systems

are more popular, as they allow the selection of

anti-bodies of higher affinity, and because it is far easier to

create large libraries in phagemids than in phages [48]

Nevertheless, scFvs often oligomerize, both when

dis-played on phages and as soluble proteins in solution,

thus making it difficult to select stringently on the

basis of intrinsic affinity; furthermore, they can be

dif-ficult to express and purify in soluble form [36,49–51]

Fab is a heterodimer consisting of the entire light

chain (VL–CL) paired with the variable and first

con-stant domain of the heavy chain (VH–CH1) [52,53] As

opposed to scFv, the Fab molecule with a total size of

50 kDa is displayed on phages in a monovalent format [54,55] However, Fab display is not suitable for anti-gens for which high-affinity binders cannot be obtained, either because of limitations in the library diversity, or because of the physicochemical properties

of the target (e.g carbohydrates) Furthermore, the display of Fab on phages is far less efficient than that

of scFv [31,36] To address some of the limitations associated with scFv and Fab phage display, Lee et al [36] recently reported the development of a system for the display of bivalent Fab [F(ab¢)2] on the gene-3 coat protein of a single phage particle in a manner that effectively resembles the binding behavior of natural IgGs This display system was successfully employed for the isolation of specific F(ab¢)2 from synthetic libraries [56,57] Bivalent display results in an avidity effect that reduces the off-rates of phage bound to immobilized antigen or to cell surface antigens Yet, at the same time, the display valency is not high enough

to influence binding to soluble antigen, and thus biva-lent phage bind to solution-phase antigen with appar-ent affinities close to intrinsic monovalappar-ent affinities [36] Consequently, bivalent display systems can aid in the recovery of antibodies with moderate affinities, and also in selections that require dimerization for activity

In recent years, the significance of bivalent display for the selection of a broader spectrum of antibodies has led to the development of several display systems that display dimers of scFvs or Fabs to effectively mimic the natural IgG [36,58]

Display of IgGs in their natural conformation expands the sequence diversity that can be encoded, and therefore increases the functional library size for screening [38] For this reason, we recently developed

an E coli-based technology for the isolation of full-length IgGs from combinatorial libraries using FACS [39,40] FACS is a very powerful and reliable high-throughput screening methodology However, sorting

of a library greater than 109clones using only FACS

is time-consuming and challenging Commercially available flow cytometers are capable of sorting rates

of up to 40 000 s)1, permitting the screening of approximately 108cellsÆh)1 [40] Therefore, the screen-ing of very large naı¨ve⁄ synthetic libraries comprising more than 109clones would require 1 day of continu-ous operation of the instrument, which is clearly very challenging This is particularly impractical if one con-siders that at least 10 times the initial library size should be screened to obtain efficient coverage of the library diversity

To reduce the initial library to a size manageable by FACS, we describe here a method that capitalizes on the display and selection of full-length IgGs on

fila-Pre-sort

M18

Pre-sort YMF10

R3 YMF10

R3 M18

Fig 4 Enrichment of high-affinity and moderate-affinity IgGs

through FACS, determined using PCR with antibody-specific

prim-ers The amounts of M18 (30 p M ) and YMF10 (30 n M ) DNA present

on this agarose gel increase following three rounds (R3) of FACS

on the phage output (Pre-sort) This indicates FACS-dependent

enrichment of these antibodies.

mentous phage particles For the establishment of an

IgG phage display system, we took advantage of the

fUSE5–ZZ phage These phage particles display the

Fc-binding ZZ protein on all five copies of the phage

p3 coat protein, and are exploited as both the helper

phage and for capturing the soluble IgG via the ZZ

protein Even though the ZZ domain has a relatively

moderate affinity for Fc IgG in solution (10 nm) [45],

its display on all copies of p3 gives rise to multivalent

display of the ZZ domain that sufficiently diminishes

the functional dissociation of the IgG We showed that

phage particles displaying M18 IgG were efficiently

enriched from a 1 000 000-fold excess of phage

dis-playing the control 26.10 IgG This significant

enrich-ment of specific binders from an excess of nonbinders

validated the competency of our display system, and

also provided the fundamental basis for selection of

fUSE5–ZZ–IgG from combinatorial libraries

We took advantage of the IgG phage display system

as a prescreening step prior to selection by FACS The

phage system provides an elegant means for the

effi-cient downsizing of very large libraries by employing

an initial round of phage selection that specifically

pre-enriches target cells from the library The downsized

library can subsequently be subjected to rounds of

FACS, a technique that enables very precise control of

the selection process as compared with phage display

and, importantly, enables the isolation of clones

exhib-iting high affinity and selectivity

Using an anti-PA mouse immune IgG library with

an estimated size of 2· 107 as a model library, we

showed that, following an initial round of phage

selec-tion and two tandem rounds of FACS, specific

anti-PA IgGs were isolated with affinities ranging from the

low-nanmolar to mid-nanomolar range This spectrum

of affinities emphasizes the advantages of employing

bivalent IgG display, as it facilitates the selection of

both moderate-affinity clones (R17 and R12) with

affinities in the single-digit nanomolar range and very

rare binders with modest binding kinetics (R15 and

R42) that otherwise could probably not be detected by

monovalent display When desired, low-affinity

anti-bodies can be further engineered for improved

affini-ties by using the selected antibody gene as the basis

for subsequent mutagenesis libraries

To demonstrate the utility of this technology with

respect to the screening of large libraries, we

demon-strated the enrichment of both a high-affinity IgG and

a moderate-affinity IgG that recognize the protective

antigen of B anthracis from a 1 : 108 dilution of

con-trol antibody Such a dilution represents only 100

mem-bers of each specific clone in a library of 1010clones,

which is near the upper limit of the diversity currently

available in synthetic libraries [17,20] We showed that the use of sequential IgG–phage panning followed by FACS allowed the simultaneous selection of both anti-bodies despite the fact that they display a 1000-fold dif-ference in affinity and each was present at a frequency

of only 1 : 100 million in the initial population

To conclude, the significance of the methodology described here is illustrated by the fact that it provides the first demonstrated approach permitting the selection

of full-length IgGs from libraries displayed on phage

We believe that sequential selection by phage display and then FACS enables the efficient screening of very large IgG libraries by sufficiently oversampling to cover diversity and by simultaneously utilizing the superior technique of FACS for final enrichment Furthermore, with the bivalent IgG format, it should be possible to both select moderate-affinity antibodies from large naı¨ve⁄ synthetic repertoires, and also to affinity mature low-affinity binders using stringent solution-phase selec-tions that discriminate on the basis of intrinsic affinity

Experimental procedures

Cell lines and plasmids

Phagemid pMAZ360–IgG for production of full-length IgG has been described previously [39] Phage fUSE5–ZZ [44] was kindly provided by I Benhar (Tel Aviv University, Israel) E coli K91K⁄ F¢ [59] cells were used for propaga-tion of fUSE5–ZZ and producpropaga-tion of fUSE5–ZZ–IgG

E coli JUDE-1⁄ F¢ cells [39] (DH10B harboring the F¢ factor derived from XL1-blue) were used for expression and purification of soluble IgG molecules

Production of fUSE5–ZZ phage

Escherichia coli K91K⁄ F¢ cells carrying fUSE5–ZZ DNA were inoculated overnight at 30 C and 250 r.p.m in 5 mL

of 2· YT medium supplemented with 20 lgÆmL)1 tetra-cycline and 50 lgÆmL)1 kanamycin On the following day, the culture was diluted into 500 mL of 2· YT medium supplemented with 20 lgÆmL)1 tetracycline and grown overnight at 30C and 250 r.p.m Cells were pelleted at

4500 g for 15 min at 4C, and the supernatant was filtered through a 0.22 lm filter The phages were precipitated by addition of 20% (w⁄ v) poly(ethylene glycol) 6000 and 2.5 m NaCl, and this was followed by centrifugation at 8000 g for

30 min at 4C The phages were suspended in sterile and filtered NaCl⁄ Piat a concentration of 1013colony-forming units (CFU)⁄ mL and stored at 4 C To titer the phage stock, 10-fold serial dilutions of the phages were made in sterile 2· YT medium A logarithmic E coli K91K ⁄ F¢ cul-ture was infected with the diluted phages, and the mixed culture was incubated for 60 min at 37C without shaking

and then for 30 min with gentle shaking at 110 r.p.m.

Infected cells were plated on 2· YT plates supplemented

with 20 lgÆmL)1 tetracycline and 50 lgÆmL)1 kanamycin,

and grown overnight at 37C

Preparation of fUSE5–ZZ–IgG

For preparation of the fUSE5–ZZ–IgG phage particles,

cul-tures of E coli K91K⁄ F¢ cells transformed with phagemid

pMAZ360–IgG were grown overnight at 30C and

250 r.p.m in 5 mL of 2· YT medium supplemented with

100 lgÆmL)1ampicillin, 50 lgÆmL)1kanamycin, and 2%

glu-cose On the following day, the cultures were diluted 1 : 100

into 10 mL of 2· YT medium supplemented with

100 lg⁄ mL ampicillin and 2% glucose, and grown at 37 C

and 250 r.p.m until 0.6£ A600 nm£ 0.8 Cultures were

infected with fUSE5–ZZ helper phage at a ratio of 1 : 20

(number of bacterial cells⁄ number of helper phage particles,

assuming that A600 nm= 1.0–5· 108bacteriaÆmL)1) The

cultures were incubated at 37C for 60 min without shaking,

and then with gentle shaking at 110 r.p.m for an additional

30 min The infected cells were collected by centrifugation at

4000 g for 10 min at 4C and suspended in 40 mL of

2· YT medium supplemented with 100 lgÆmL)1ampicillin

and 20 lgÆmL)1tetracycline, and grown overnight at 30C

and 250 r.p.m On the following day, the resulting fUSE5–

ZZ–IgG phage particles were precipitated with poly(ethylene

glycol)⁄ NaCl as described above; the supernatant was

care-fully aspirated off, and the phage particles were suspended in

sterile and filtered NaCl⁄ Pi at a concentration of

1011CFUÆmL)1, and kept at 4C

Binding analysis of fUSE5–ZZ–IgG in ELISA

ELISA plates were coated with 5 lgÆmL)1 PA from

B anthracis(List Biological Labs, Campbell, CA, USA) or

5 lgÆmL)1digoxin⁄ BSA [60] in NaCl ⁄ Piat 4C overnight,

and then blocked with 2% (v⁄ v) nonfat milk in NaCl ⁄ Pi

(NaCl⁄ Pi-M) for 2 h at room temperature Next, 50 lL of

1011CFUÆmL)1 fUSE5–ZZ–IgG phage particles was added

in a three-fold dilution series to plates already containing

100 lL of NaCl⁄ Pi-M and incubated for 1 h at room

tem-perature The plates were washed three times with NaCl⁄ Pi

containing 0.05% (v⁄ v) Tween-20 (NaCl ⁄ Pi-T), and bound

phage particles were detected with HRP-conjugated goat

anti-M13 IgG (antibody against pVIII)

(Amersham-Pharma-cia Biosciences, Piscataway, NJ, USA) The ELISA was

developed using the chromogenic HRP substrate TMB+

(DAKO, Glostrup, Denmark), and color development was

terminated with 1 m H2SO4 The plates were read at 450 nm

To determine the number of IgG molecules per phage,

ELISA plates were coated with either 5 lgÆmL)1goat

anti-M13 IgG (Amersham-Pharmacia Biosciences), 5 lgÆmL)1

chicken anti-(human IgG) Fc-specific (GeneTex, Irvine, CA,

USA), or 5 lgÆmL)1 PA in NaCl⁄ Piat 4C for 20 h, and

blocked with 2% NaCl⁄ Pi-M for 2 h at room temperature Then, 50 lL of 1011CFUÆmL)1fUSE5–ZZ–M18 IgG phage particles were added in a three-fold dilution series to plates already containing 100 lL of NaCl⁄ Pi-M and incubated for

1 h at room temperature The plates were washed three times with NaCl⁄ Pi-T, and the ELISA was developed as above The A450 nm values were plotted against the phage concentration and used as a standard curve fUSE5– ZZ–M18 IgG was tested for binding to plates coated with anti-M13 IgG, anti-Fc IgG, and PA The number of IgG molecules per phage was determined by calculating the ratio

of the phage concentration derived from anti-Fc IgG⁄ anti-M13 IgG or from PA⁄ anti-M13 IgG in the linear range To evaluate the stability of the ZZ–IgG complex, freshly pro-duced 1011CFUÆmL)1 fUSE5–ZZ–IgG in NaCl⁄ Pi was incubated for 1 h at room temperature with 1 lm standard human IgG (Jackson Immunolaboratories, West Grove, PA, USA) as a competitor prior to being applied to the ELISA plates The plates were washed three times with NaCl⁄ Pi-T, and bound phage particles were detected with HRP-conju-gated goat anti-M13 IgG (Amersham-Pharmacia Bioscienc-es) The ELISA was developed using the chromogenic HRP substrate TMB+ (DAKO), and color development was ter-minated with 1 m H2SO4 The plates were read at 450 nm

Enrichment of fUSE5–ZZ–M18 IgG by phage biopanning

Anti-PA IgG-displaying fUSE5–ZZ–M18 phage particles were enriched from a 1 000 000-fold excess of phage display-ing the control 26.10 IgG A 35 mm tissue culture six-well plate was coated overnight at 4C with 75 lgÆmL)1PA in NaCl⁄ Piin a total volume 0.7 mL After the excess solution had been discarded, the wells were washed once with NaCl⁄ Pi and blocked with 3 mL of NaCl⁄ Pi containing 0.25% (w⁄ v) gelatin (NaCl ⁄ Pi-G) for 2 h at room tempera-ture Then, the wells were washed five times with NaCl⁄ Pi, incubated with a phage mixture consisting of a 1 000 000-fold excess (1010fUSE5–ZZ–26.10 IgG and 104fUSE5–ZZ– M18 IgG) in a total volume of 0.7 mL, and rocked gently at room temperature for 2 h Unbound phage particles were rinsed away, and the plate was washed extensively 10 times with NaCl⁄ Pi-T, and then five times with NaCl⁄ Pi Bound phages were eluted by the addition of 400 lL of 0.1 m HCl adjusted to pH 2.2 with glycine and 1 mgÆmL)1 BSA for

10 min at room temperature with gentle agitation The eluted phage particles were transferred into a 1.5 mL microfuge tube, and immediately neutralized with 75 lL of 1 m Tris⁄ HCl (pH 9) The selected phage particles were used for reinfection of E coli K91K⁄ F¢ cells for subsequent rounds of phage selection The neutralized eluted phage particles (0.5 mL) were mixed with 4.5 mL of 2· YT medium and

5 mL of logarithmic K91K⁄ F¢ cells, and infection was per-formed as described above The infected cells were spread on

2· YT plates supplemented with 100 lgÆmL)1 ampicillin,

50 lgÆmL)1 kanamycin, and 2% glucose, and grown

over-night at 30C On the following day, the plates were scraped

and subcultured into 50 mL of 2· YT medium

supple-mented with 100 lgÆmL)1ampicillin, 50 lgÆmL)1kanamycin

and 2% glucose to give a starting A600 nmof 0.1 The culture

was grown at 37C and 250 r.p.m until 0.6 £ A600 nm£ 0.8,

and 10 mL of the culture was used for infection with fUSE5–

ZZ, as described above Evaluation of the enrichment in each

round of selection was performed by phage ELISA as

described above

Library selection by phage biopanning

Electrocompetent E coli K91K⁄ F¢ cells were transformed

with DNA phagemid of the anti-PA mouse immune library

[39] to generate a final library of 107independent

transfor-mants Library cells carrying phagemid pMAZ360–IgG

were inoculated in 500 mL of 2· YT medium

supple-mented with 100 lgÆmL)1 ampicillin and 2% glucose to

give a starting A600 nm of 0.1, and grown at 37C and

250 r.p.m until 0.6£ A600 nm£ 0.8 Then, the culture

was infected with fUSE5–ZZ helper phage at a ratio of

1 : 20, as described above The infected cells were collected

by centrifugation at 4500 g for 10 min at 4C, suspended

in 2000 mL of 2· YT medium supplemented with

100 lgÆmL)1 ampicillin and 20 lgÆmL)1 tetracycline, and

grown overnight at 30C and 250 r.p.m On the following

day, fUSE5–ZZ–IgG library phages were precipitated with

poly(ethylene glycol)⁄ NaCl as described above; the

super-natant was carefully aspirated off, and the phages were

sus-pended in sterile and filtered NaCl⁄ Piat a concentration of

1012CFUÆmL)1and kept at 4C

An initial cycle of phage panning was performed in

solu-tion with biotinylated PA, using streptavidin-coated

para-magnetic beads (Invitrogen, Carlsbad, CA, USA), essentially

as previously described [61] For negative selection, the

library phages were incubated with 150 lL of

streptavidin-coated beads for 30 min at room temperature for depletion

of nonspecific fUSE5–ZZ–IgG phage particles The phage–

bead mixture was then applied to a magnet apparatus, and

the unbound library phage particles in the supernatant were

removed to a new tube For positive selection, depleted

library phage particles were incubated with 500 nm

biotiny-lated PA in 1 mL of NaCl⁄ Pi-M for 1 h at room

tempera-ture Then, the phage–antigen complex was incubated with

150 lL of streptavidin magnetic beads for 30 min at room

temperature, and the phage–antigen–bead complex was

applied to the magnet The beads were washed vigorously,

and bound phage particles were eluted with 1 mL of 100 mm

triethylamine (pH 13) for 10 min while being rotated The

eluted phage particles were separated from the beads and

immediately neutralized with 200 lL of 1 m Tris⁄ HCl

(pH 7.4) Library-selected phages were used for infection of

E coli Jude-1⁄ F¢ cells harboring plasmid pBAD33–NlpA–

ZZ [39] for subsequent rounds of FACS selection

Neutral-ized eluted phage particles were mixed with 4.5 mL of

2· YT medium and 5 mL of logarithmic Jude-1–NlpA–ZZ cells, and infection was carried out as above The infected cells were plated on 2· YT plates supplemented with

100 lgÆmL)1 ampicillin, 30 lgÆmL)1 chloramphenicol, and 2% glucose, and grown overnight at 30C

Library selection by FACS

Selection by FACS was performed essentially as previously described [40] Briefly, E coli Jude-1⁄ F¢ cells carrying plas-mids pBAD33–NlpA–ZZ and pMAZ360–IgG were inocu-lated in TB medium supplemented with 100 lgÆmL)1 ampicillin, 30 lgÆmL)1 chloramphenicol, and 2% glucose, and grown at 30C to an A600 nmof 1.0 Then, cells were collected by centrifugation at 4500 g for 10 min at 4C, induced for IgG expression by resuspension in TB medium supplemented with 100 lgÆmL)1ampicillin, 30 lgÆmL)1 chl-oramphenicol, and 1 mm isopropyl thio-b-d-galactoside (IPTG), and grown overnight at 25C and 250 r.p.m On the following day, cells were induced with 0.2% arabinose for an additional 3 h at 25C and 250 r.p.m for NlpA–ZZ expression Then, the cellular outer membrane of

A600 nm5.0 freshly induced library cells was permeabilized

by Tris⁄ EDTA and lysozyme treatment For two-color FACS based on antigen binding (affinity) and expression of the displayed antibody, cells (10-fold excess of phage-bio-panning-output) were labeled with 500 nm PA63–FITC (List Biological Labs) and 100 nm Alexa Fluor 647-conju-gated chicken anti-(human IgG) Fc-specific (GeneTex) for

2 h at 4C Highly fluorescent cells falling into the double-positive quadrant, indicating both expression and affinity, were sorted on a FACSAria droplet deflection flow cytome-ter (Becton Dickinson, Franklin Lakes, NJ, USA) equipped with both a 488 and a 633 nm laser For better selectivity, cells captured after the first sort were immediately resorted

on the flow cytometer, using the same collection gate as used for the initial sort Subsequently, a DNA fragment corresponding to the VL–CK–VHsequence of the IgG gene was amplified from DNA plasmid pMAZ360–IgG of sorted cells using the following primers: VLlibrary amplifier,

PCR product was recloned into the pMAZ360–IgG vector, retransformed into fresh Jude-1–NlpA–ZZ cells, and grown overnight on agar plates at 30C The resulting clones were grown, induced for expression of IgG, and subjected to an additional round of sorting

Screening of selected clones by ELISA

Following two rounds of FACS selection, single selected cells were screened for antigen binding in ELISA, essen-tially as previously described [40] Briefly, PCR-recovered IgG genes were ligated into the pMAZ360–IgG expression

vector and transformed into fresh E coli Jude-1 cells (not

carrying plasmid pBAD33–NlpA–ZZ) for expression of

sol-uble, noncaptured IgGs Randomly selected colonies were

inoculated into round-bottomed 96-well plates containing

200 lL of LB medium supplemented with 100 lgÆmL)1

ampicillin and 2% glucose, and grown overnight at 30C

and 150 r.p.m on a shaker platform On the following day,

the cultures were diluted 1 : 20 for inoculation on fresh

round-bottomed 96-well plates containing 200 lL of TB

medium supplemented with 100 lgÆmL)1ampicillin and 2%

glucose, and grown for 3 h at 30C and 150 r.p.m Then,

the plates were centrifuged for 10 min at 4500 g, and pellets

were resuspended in 200 lL of TB medium supplemented

with 100 lgÆmL)1 ampicillin and 1 mm IPTG Cells were

induced overnight at 25C and 150 r.p.m for expression of

soluble IgG antibodies On the following day, the plates

were centrifuged, and the cells were lysed in 200 lL of 20%

BugBuster HT Protein Extraction Reagent (Novagen,

Gibbstown, NJ, USA) in NaCl⁄ Pifor 1 h at room

tempera-ture The plates were centrifuged as above, and soluble cell

extracts were tested for direct binding to the PA antigen as

follows ELISA plates were coated with 5 lgÆmL)1 PA in

NaCl⁄ Piat 4C overnight, and blocked with 2% NaCl ⁄ Pi

-M for 2 h at room temperature Then, 25 lL volumes of

the cell extracts were applied to plates containing 75 lL of

NaCl⁄ Pi-M, and incubated for 1 h at room temperature

The plates were washed three times with NaCl⁄ Pi-T, and

bound IgG was detected with HRP-conjugated goat

anti-(human IgG) (Jackson Immunolaboratories) ELISA plates

were developed as above

Expression and purification of soluble IgGs

Expression and purification of soluble full-length IgG was

performed essentially as previously described [40] Briefly,

E coli Jude-1 cells carrying plasmid pMAZ360–IgG were

grown at 30C in 200 mL of TB medium supplemented with

100 lgÆmL)1 ampicillin and 2% glucose until the A600 nm

was 1.0 Subsequently, cells were collected by centrifugation

at 4500 g, for 10 min at 4C and IgG expression was

induced by resuspension in TB medium supplemented with

100 lgÆmL)1 ampicillin and 1 mm IPTG; cells were then

grown for 16 h at 25C Induced cultures were lysed as

above, and IgGs were purified from the soluble fraction of

total cell extracts using protein A Sepharose (Amersham

Biosciences, Sweden) chromatography with final yields of

0.2–1 mgÆL)1of cells Bound antibody was eluted with 0.1 m

citric acid (pH 3) and neutralized with 1 m Tris⁄ HCl (pH 9)

Protein-containing fractions were combined, dialyzed

against 5 L of NaCl⁄ Pi, sterile filtered, and stored at 4C

Biacore analysis

The antigen-binding kinetics of purified IgGs were

deter-mined by surface plasmon resonance analysis, using a

Biacore 3000 (Biacore-GE Healthcare, NJ, USA) instru-ment, essentially as previously described [39] Both a direct binding method and a capture method were utilized to determine kinetic parameters Briefly, PA63 was coupled to

a CM5 chip to an equivalent of 750 response units (RU)

by using 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide⁄ N-hydroxysuccinimide chemistry as recommended by the manufacturer Human transferrin (Jackson Immunolabora-tories) was similarly coupled, and used for in-line subtrac-tion Various concentrations of purified IgG in NaCl⁄ Pi were injected in duplicate at a flow rate of 50 lLÆmin)1 for

1 min at 25C, and the surface was regenerated using one pulse of 50 mm NaOH and 1 m NaCl The data were ana-lyzed using biaevaluation software with appropriate sub-traction methods, and the bivalent analyte model was used

to account for avidity effects associated with the IgG (Biacore-GE Healthcare) For the capture method, goat anti-(human IgG1) Fc-specific (Jackson Immunolaborato-ries) was coupled to a CM5 chip to an equivalent of

10 000 RU by using 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide⁄ N-hydroxysuccinimide chemistry as recom-mended by the manufacturer Various concentrations of purified IgG in Hepes-buffered saline⁄ EP buffer (Biacore, Pittsburgh, PA, USA) were injected at a flow rate of 5 lLÆ min)1 at 25C to achieve  100 RU of captured IgGs Buffer and antigen were then injected serially through in-line flow cells at a flow rate of 50 lLÆmin)1(5 min of sta-bilization, 1 min of association, and 5 min of dissociation), and the surface was regenerated using two pulses of 100 mm

H3PO4 A three-fold dilution series of PA-63, starting at

15 nm, was analyzed in duplicate using biaevaluation soft-ware (Biacore) with appropriate subtraction methods

Enrichment of high-affinity and moderate-affinity IgG by tandem phage FACS

Three fUSE5–ZZ–IgG phages displaying M18, YMF10 and

VA IgGs were produced and purified as described above The M18, YMF10 and VA IgGs display high affinity (30 pm) or moderate affinity (30 nm) for B anthacis PA, and the VA IgG binds to an unrelated antigen, the V pro-tein of Yersinia pestis The high-affinity and moderate-affin-ity IgGs displayed on fUSE5–ZZ–IgG phage particles were each diluted 1 : 108in 1011copies of the VA IgG displayed

on fUSE5–ZZ–IgG phage particles Phage biopanning using biotinylated PA antigen was performed as described above The library was further subjected to three rounds of FACS screening, performed essentially as described above, using

500 nm PA63–FITC and 100 nm Alexa Fluor 647-conju-gated chicken anti-(human IgG) Fc-specific The top 3%, 2% and 1% of fluorescent cells were collected, respectively,

in rounds 1, 2, and 3 After each round of sorting, the insert DNA was rescued by PCR amplification, recloned into vector pMAZ360–IgG as described above, and trans-formed into Jude-1 cells harboring the NlpA–ZZ plasmid