APEx 2-hybrid, a quantitative protein–proteininteraction assay for antibody discoveryand engineeringKi Jun Jeong*†, Min Jeong Seo*†, Brent L. Iverson†‡, and George Georgiou*†§¶�

Departments of *Chemical Engineering, ‡Chemistry and Biochemistry, and §Biomedical Engineering, †Institute for Cellular and Molecular Biology,and ¶Section of Molecular Genetics and Microbiology, University of Texas, Austin, TX 78712

Communicated by Alan M. Lambowitz, University of Texas, Austin, TX, March 27, 2007 (received for review January 29, 2007)

We have developed a bacterial system for the discovery of inter-acting proteins that, unlike other two-hybrid technologies, allowsfor the selection of protein pairs on the basis of affinity orexpression. This technology relies on the anchored periplasmicexpression (APEx) of one protein (bait) on the periplasmic side ofthe inner membrane of Escherichia coli and its interacting partner(prey) as a soluble, epitope-tagged, periplasmic protein. Uponremoval of the outer membrane by spheroplasting, periplasmicproteins, including any unbound epitope-tagged prey, are releasedinto the extracellular fluid. However, if the epitope-tagged preycan bind to the membrane-anchored bait, it remains associatedwith the cell and can be detected quantitatively by using fluores-cent anti-epitope tag antibodies. Cells expressing prey:bait pairsexhibiting different affinities can be readily distinguished by flowcytometry. The utility of this technology, called APEx two-hybrid,was demonstrated in two demanding antibody engineering appli-cations: First, single-chain variable fragment (scFvs) with increasedaffinity to the protective antigen of Bacillus anthracis were isolatedfrom cells coexpressing libraries of scFv random mutants, togetherwith endogenously expressed antigen. Second, APEx two-hybridcoupled with multicolor FACS analysis to account for proteinexpression was used for the selection of mutant Fab antibodyfragments exhibiting improved expression in the bacterialperiplasm.

FACS � high-throughout screening � antibody engineering � Fab

The discovery and analysis of interacting proteins are essentialfor establishing biological networks, the understanding of

cellular function and for drug discovery. Beginning with thedevelopment of the yeast two-hybrid system more than 15 yearsago (1, 2), numerous methods for the in vivo identification ofpairs of interacting proteins from expression libraries have beendescribed. These methods include two-hybrid systems for organ-isms other than yeast, namely bacteria and mammalian cells, andprotein complementation assays (PCA) (3–7). In recent years,yeast two-hybrid and dihydrofolate reductase (DHFR) comple-mentation assays have been configured for robotic automationand used for the construction of large-scale protein networks (8,9). However, despite their extensive utility, existing methods forthe detection of protein:protein interactions in vivo suffer fromtwo shortcomings. First, they lack quantitation and therefore donot provide information on the affinity or the level of expressionof the interacting proteins that are being tested. Second, with afew recent exceptions, there has been little success in thedetection of interacting proteins within secretory compartments,such as proteins requiring disulfide bonds for folding (10–12).

The aforementioned shortcomings are of particular impor-tance in the application of in vivo protein interaction assays toantibody engineering. Coexpressing the antigen together with anantibody repertoire library eliminates the need for a source ofpurified target protein and thus could greatly expedite thehigh-throughput generation and affinity maturation of antibod-ies for proteomic purposes (13–15). The available protein inter-

action assays lack the quantitation needed for the selection ofhigh-affinity antibodies. For example, a recent study aimed at theselection of intracellular antibodies capable of binding antigenwithin the reducing environment of the cytoplasm by using thesplit murine enzyme dihydrofolate reductase (DHFR) proteincomplementation assay (PCA) resulted in the isolation of a fewbinders with equilibrium dissociation constants in the 30 �Mrange (16). In addition, with a few exceptions, antibody foldingdepends on disulfide bond formation and therefore has to takeplace in an oxidative cellular compartment such as the bacterialperiplasmic space.

The production yield of antibody fragments in Escherichia coli,especially Fabs, is highly variable. Mutations that enhanceantibody fragment expression have been identified in a fewinstances, but such studies are very time-consuming and labor-intensive (17–19). An alternative approach is to isolate Fabmutants exhibiting increased expression yields from combinato-rial libraries. However, isolation of well expressed mutant anti-bodies from libraries is predicated on the availability of appro-priate high-throughput screening strategies that allow thedetection of heavy and light chain pairing to form a completeFab, as well as antigen binding.

In this report, we describe a system for detecting and rankordering protein:protein interactions in the periplasmic space ofE. coli cells. The system is based on the anchored periplasmicexpression (APEx) (20) of one protein (bait) and the solubleexpression of the second protein fused to an epitope tag (prey)(Fig. 1A). The bait is tethered on the inner membrane by a 6-aa,N-terminal extension that becomes fatty acylated upon secre-tion, with the lipid moiety embedded in the lipid bilayer. Afterdisruption of the outer membrane by treatment with lysozyme/EDTA (spheroplast formation), the contents of the periplasmicspace, including the soluble prey protein, are released into theextracellular fluid. However, if the prey can bind to the innermembrane-tethered bait, then a complex is formed that remainsassociated with the spheroplasts. A fluorescently labeled anti-body specific for the tag on the prey is then used to quantitativelydetect the protein complex on the surface of spheroplasts. Thefluorescence of the spheroplasts is proportional to the numberof bait:prey complexes per cell and depends on the affinity andthe expression level of the two proteins. We demonstrate theutility of this protein:protein interaction assay, which we have

Author contributions: K.J.J. and G.G. designed research; K.J.J. and M.J.S. performed re-search; K.J.J., M.J.S., B.L.I., and G.G. analyzed data; and K.J.J., B.L.I., and G.G. wrote thepaper.

The authors declare no conflict of interest.

Abbreviations: APEx, anchored periplasmic expression; D4, domain 4; IPTG, isopropyl�-D-thiogalactoside; PA, protective antigen; PE, phycoerythrin; scAb, single-chain antibodyfragment; scFv, single-chain variable fragment.

�To whom correspondence should be addressed. E-mail: gg@che.utexas.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0702650104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0702650104 PNAS � May 15, 2007 � vol. 104 � no. 20 � 8247–8252

APP

LIED

BIO

LOG

ICA

LSC

IEN

CES

named APEx two-hybrid for (i) the isolation of single-chainvariable fragment (scFv) antibodies that bind to an endogenouslyexpressed protein antigen with enhanced affinity and (ii) thedirected evolution of Fab fragments exhibiting enhanced expres-sion yield.

ResultsThe APEx Two-Hybrid System. The 14B7 scFv binds the protectiveantigen (PA) component of the Bacillus anthracis toxin (20). Itrecognizes a conformational epitope located within the PAdomain 4 (PA-D4), a 139-aa fragment composed of amino acids596–735 (21). Affinity-matured variants of 14B7, such as the 1Hantibody (22) and M18, are clinically important for prophylaxisand postexposure treatment of inhalation anthrax (23). The14B7 and M18 scFv were used as the bait and were expressed asinner membrane lipoproteins by fusion to the leader peptide andthe first 6 aa of the mature sequence (CDQSSS) of the E. colilipoprotein NlpA [see supporting information (SI) Fig. 5 and SIMethods for details). As the prey, we used the PA-D4 fused toa C-terminal FLAG epitope tag and secreted into the periplas-mic space by using the pelB leader peptide. After induction ofthe scFv and PA-D4 proteins by isopropyl �-D-thiogalactoside(IPTG), the cells were converted to spheroplasts by treatment

with lysozyme and EDTA. The spheroplasts were washed and ahigh affinity anti-FLAG-FITC conjugated antibody was used tolabel any prey (PA-D4) that remained bound to the scFv bait(Fig. 1B). The signal obtained with the 14B7 scFv was �3-foldhigher than that observed with cells expressing PA-D4 alone.The higher affinity M18 scFv gave a signal that was an additional3-fold higher than that obtained with the 14B7 scFv (10-foldhigher than background). It should be noted that both the 14B7and the M18 scFvs are expressed well in the E. coli periplasm,and therefore the respective NlpA fusions accumulate at similaramounts, as determined by Western blotting (data not shown).Thus, the difference in the fluorescence signal is due to thehigher affinity of the M18 scFv:PA-D4 interaction and not todifferences in expression level.

Rosovitz et al. (24) reported that the Y688A mutation in PAinterferes with the binding of 14B7 antibody. Accordingly, whenPA-D4 with the Y688A mutation was used as the prey, thefluorescence signal obtained in cells expressing anchored M18scFv was marginally higher than background (Fig. 1B). Incontrast, PA-D4 Y681A a mutant that affects the ability of PAto bind to its receptor, but not its recognition by the 14B7antibody, resulted in high fluorescence in cells expressing an-chored M18 scFv. Collectively, the above results reveal that the

Fig. 1. Flow-cytometric schemes and assays by APEx two-hybrid system. (A) A schematic diagram showing the principle of the APEx two-hybrid system. (B)Fluorescence distribution of E. coli cells coexpressing NlpA-scFvs and PelB-[PA-D4wt-FLAG] or PA-D4 mutants (Y681A or Y688A). Spheroplasted cells were labeledwith anti-FLAG-FITC conjugates. Mn represents the mean fluorescence intensity of the spheroplast population as determined by forward scatter (FSC) vs. sidescatter (SSC).

8248 � www.pnas.org�cgi�doi�10.1073�pnas.0702650104 Jeong et al.

APEx two-hybrid system allows the specific detection of inter-acting proteins in a manner that reflects their binding specificityand affinity.

Affinity Maturation of a scFv Toward Endogenously Expressed Anti-gen. We examined whether the APEx two-hybrid system can beused for the isolation of affinity enhanced antibodies from largelibraries of random mutants. For this purpose, we sought toisolate variants of the 14B7 scFv (KD for PA � 3.3 nM) that bindto endogenously expressed PA-D4 antigen with increased affin-ity. The 14B7 scFv gene was subjected to random mutagenesis,ligated into a vector also containing the pelB-PA-D4-FLAGgene under the lac promoter, and a library of 2 � 106 indepen-dent transformants was obtained. Sequencing of 16 randomlyselected library clones revealed an average of 1.8% nucleotidesubstitutions per gene. Cells (2 � 108) were converted tospheroplasts, washed, labeled with the anti-FLAG-FITC conju-gate, and sorted on a MoFlo flow cytometer. Approximately0.8% of the highest f luorescence events were collected andresorted immediately as above. The resulting sort solutioncontaining 2.5 � 104 fluorescent events was isolated, and thescFv genes were rescued by PCR amplification. After ligationand retransformation, a second round of sorting and resortingwas performed, and 3.2 � 104 events were isolated. After PCRrescue of the scFv genes, cloning, and transformation in E. coli,the cells were plated. The fluorescence of individual clonesselected at random was analyzed by FACS. The four mostfluorescent clones were selected for further analysis (Fig. 2A).DNA sequencing revealed the presence of 4- to 6-aa substitu-tions (SI Fig. 6).

The 2–16 and 2–38 scFvs were converted to the betterexpressed single-chain antibody fragment (scAb) format byfusion to a human C� chain (25). The scAbs were expressed insoluble form in the E. coli periplasm purified by metal affinityand size-exclusion chromatography, and the kinetics of antigen

binding were determined by BIACore analysis (Fig. 2 B and C).As expected, both the 2–16 and 2–38 antibody fragments exhib-ited substantially lower KD values for PA relative to the parental14B7 scAb antibody. The lower KD values resulted primarilyfrom slower antigen dissociation (i.e., slower koff, Fig. 2C). Forexample, the 2–38 clone exhibited a KD of 277 pM (koff � 2.6 �10�4 sec�1), representing a �12-fold affinity improvement com-pared with the parental antibody 14B7 (KD � 3.3 nM, koff �3.2 � 10�3 sec�1). Overall, these results show that the APExtwo-hybrid system can be used to enrich and isolate cellsexpressing interacting proteins, in this case scFv variants and thePA-D4 antigen, exhibiting subnanomolar affinities. Moreover, areasonable correlation between the FACS signal and the in vitrodetermined equilibrium binding constant was observed forantibodies with KDs differing by two orders of magnitude (SIFig. 7).

Fab Expression Maturation. Although some Fabs can be expressedat the g/liter scale in fermenters, very often, aggregation orproteolysis lead to very low yields (26–28). Typically, the yield offunctional Fab protein is enhanced by cultivating E. coli at asuboptimal temperature (25°C), although for many antibodies,growth at a low temperature seems to have little effect. From amanufacturing standpoint, it is desirable to engineer Fab pro-teins that can be expressed with increased yields at 37°C, theoptimal growth temperature of E. coli. The APEx two-hybridsystem was adapted for monitoring the assembly of the heavy andlight chains into Fab antibody and subsequently, for the selectionof mutants exhibiting increased expression at 37°C. By anchoringthe light chain onto the inner membrane and by expressing theheavy chain with a FLAG epitope tag in soluble form in theperiplasm, the expression of Fab proteins could be monitored byflow cytometry with single-cell resolution (Fig. 3A). Addition-ally, incubation with antigen conjugated to a fluorescent dye(fluorescein) having a distinct emission wavelength was used forthe simultaneous detection of the amount of correctly folded,active Fab.

The Fab fragment of M18.1 hum, a humanized anti-PAantibody, was expressed from plasmid pAPEx-M18.1humFab asa dicistronic operon downstream from the lac promoter, withpelB-VH-CH1-FLAG being the first cistron and NlpA-VL-Ckbeing the second (see SI Methods for details). After inductionwith IPTG, the amount of FLAG-tagged heavy chain associatedwith the spheroplasts was determined by FACS by usingphycoerythrin (PE)-labeled anti-FLAG antibodies (570-nmemission), whereas PA conjugated to FITC (530-nm emission)was used to detect antigen binding.

When protein expression was carried out at 37°C, FACSanalysis indicated that the amount of FLAG-tagged heavy chainwas significantly lower than that observed in cells grown at 25°C(Fig. 3B). Similarly, the amount of functional Fab was alsodramatically reduced, as evidenced by the precipitous drop in thePA-FITC FACS signal in cells grown at 37°C (Fig. 3C). TheFACS signal obtained with membrane-tethered Fab was com-pared with the yield of protein expressed in soluble form underotherwise identical conditions. The NlpA leader and first 6 aafused to the light chain were replaced with the pelB leaderpeptide, and the yield of active Fab obtained per OD600 ofculture was quantified by sandwich ELISA. Consistent with theFACS data, the yield of Fab recovered from cells grown at 37°Cwas considerably lower than that observed at 25°C (Fig. 3D). Anexcellent correlation between FACS signals and yield of solubleprotein in cells grown at 37° or 25°C was also obtained with threeother Fab antibodies (M.J.S., B.L.I., and G.G., unpublisheddata).

To test the utility of the system shown in Fig. 3A for increasingthe expression of Fab antibodies (expression maturation), wesought to isolate mutants of the M18.1 hum Fab that are

Fig. 2. Analysis of anti-PA antibody fragments selected by APEx two-hybrid.(A) Flow cytometry histogram depicting the mean fluorescence (Mn) of E. coliexpressing anti-PA scFvs and labeled with anti-FLAG-FITC. (B) BIACore analysisof anti-PA scAb binding to PA. (C) Rate constants for antigen binding anddissociation acquired by surface plasmon resonance (SPR).

Jeong et al. PNAS � May 15, 2007 � vol. 104 � no. 20 � 8249

APP

LIED

BIO

LOG

ICA

LSC

IEN

CES

produced at an increased level when growth and expression arecarried out at 37°C. Under these conditions and in contrast to theexperiments that led to the isolation of higher-affinity mutantsof the 14B7 scFv, cell f luorescence is limited by the expressionof the Fab and not by antigen affinity. The VH domain of theM18.1 hum Fab was mutagenized by using error-prone PCR tocreate a library of 7 � 106 independent clones with an averageof 2% nucleotide substitutions per gene. A total of 2 � 108 cellswere sorted for both high level of correct heavy- and light-chainassembly at 37°C and high PA binding. After resort, PCR rescue,expression at 37°C, and a second round of sort–resort, sevenclones were isolated that displayed reproducibly stronger FACSsignals when expressed at 37°C compared with the parent M18.1hum (SI Fig. 8). The two clones with the highest f luorescence,referred to as Fab no. 3 and Fab no. 28, respectively, weresequenced and found to have two mutations each (SI Fig. 9). Fabno. 3 contained an L13V change in the pelB leader peptidesequence, and an N30D change in CDR1 of VH. Fab no. 28contained a Q13R change in FR1 and a Y102F change in CDR3of the VH chain.

The two Fab variants were expressed in soluble form at 37°C,and the expression yield was quantitated by sandwich ELISA.Consistent with the increased FACS signals (Fig. 4 A and B), thesoluble yields for Fab no. 3 and Fab no. 28 were 4.5- and 5.5-foldhigher, respectively, relative to the parental M18.1 hum antibody(Fig. 4C). The expression and assembly were analyzed withdenatured/nondenatured samples by Western blotting, and no. 3and no. 28 also showed increased level of assembled Fabscompared with original Fab (SI Fig. 10). BIACore analysisshowed that the koff values for Fab no. 3 and Fab no. 28 were4.5 � 10�5 sec�1 and 1.2 � 10�4 sec�1, respectively, similar tothat of the original M18.1 hum Fab (7.3 � 10�5 sec�1). Inter-estingly, several Fab libraries made from mutagenizing only theCH region failed to produce clones with enhanced expression at37°C (K.J.J., B.L.I., and G.G., unpublished data). This finding isconsistent with recent biophysical studies showing that VH:VL

pairing represents the limiting step in the folding and assemblyof many Fab antibodies (17).

Finally, we showed that Fab expressed in the periplasm canalso be used to bind endogenously expressed PA-D4 antigen (seeSI Fig. 11). Thus, the association of the heavy and light chain, aswell as the expression and binding of antigen, can all beaccomplished and monitored with single-cell resolution.

Fig. 3. Analysis of Fab assembly and expression of APEx two-hybrid system. (A) A schematic diagram of Fab assembly and expression analysis by APEx two-hybrid.The light chain is coexpressed with heavy chain fused to a FLAG tag. Antibody assembly occurs in the periplasm and, after spheroplasting, the cells are labeledwith anti-FLAG-PE and PA-FITC, which were used to detect Fab expression and antigen binding, respectively. (B and C) Cells expressing M18.1 hum Fab were grownat different temperatures [37°C (white) and 25°C (red)] and labeled with anti-FLAG-PE (B) or PA-FITC (C), which represent expression and antigen binding,respectively. (D) Fab was expressed in soluble form in cells grown at both temperatures, and the yields of antigen binding Fabs were analyzed by ELISA.

Fig. 4. Expression maturation of Fab antibodies by APEx two-hybrid. (A andB) Cells expressing Fab no. 3 or Fab no. 28 isolated by FACS were grown at 37°C,and labeled with anti-FLAG-FITC (for expression) (A) and PA-FITC (for activity)(B). (C) The yields of M18.1 hum Fab, clone no. 3, and no. 28 were determinedby ELISA.

8250 � www.pnas.org�cgi�doi�10.1073�pnas.0702650104 Jeong et al.

DiscussionAPEx two-hybrid is a simple yet powerful methodology for thedetection and isolation of interacting protein pairs in E. coli. Inthis system, the bacterial periplasm constitutes the milieu for theassociation of membrane-anchored bait and solubly expressed,epitope-tagged prey. Upon disruption of the outer membrane,only prey proteins that bind to the bait remain cell associated andare detected by flow cytometry by using fluorescently labeledanti-epitope antibodies.

In APEx two-hybrid, the fluorescence signal resulting from aspecific protein interaction is a function of both the affinity of theinteraction as well as the expression level of the interactingpartners. Using multicolor FACS, we have shown that it ispossible to carry out selections either for higher affinity or forimproved expression. When the bait and prey proteins are wellexpressed, as was the case for the 14B7 scFv and the PA-D4antigen, the fluorescence signal depends on the affinity (Fig.1B). Upon prolonged incubation, only cells expressing proteinsthat interact with a slow dissociation rate constant exhibit highfluorescence. Accordingly, isolation of highly fluorescent cellsfrom a library of random variants of the 14B7 scFv resulted inseveral antibody clones exhibiting higher affinity for the PA-D4antigen (Fig. 2). The isolation of a scFv having a KD in the 10�10

M range (clones 2–16 and 2–38) demonstrates that APExtwo-hybrid can readily distinguish and help isolate pairs ofinteracting proteins having subnanomolar affinities. To ourknowledge, APEx two-hybrid is a previously unreported pro-tein:protein interaction assay method that allows selections onthe basis of the affinity of the two partners. This feature is likelyto be useful not only for affinity maturation, as was demonstratedhere, but also for the de novo isolation of high-affinity antibodyfragments from repertoire libraries. The ability to isolate anti-bodies that bind to endogenously expressed antigens can be asignificant advantage, especially for proteins that are hard toexpress in a preparative fashion, e.g., membrane proteins. Ifnecessary, the level of the bait protein in the cell can be detectedby using a second epitope tag that is recognized by an appro-priate fluorescent antibody. In this manner the fluorescence dueto the formation of the bait:prey complex can be normalized onthe basis of the amount of bait present on each cell.

We also showed that the APEx two-hybrid system can be usedfor detecting the expression and assembly of heterodimeric Fabproteins at the single-cell level. As is the case with many antibodyfragments, the expression of the M18.1 hum Fab in cells grownat 37°C results in a low yield, which is manifest by a lower FACSsignals for both the capture of the heavy chain by the membraneanchored light chain and for antigen binding. Random mutagen-esis and FACS sorting was used to isolate Fab variants showingessentially identical antigen-binding kinetics but markedly im-proved expression and assembly that was confirmed by ELISA(Fig. 4C) and Western blotting (SI Fig. 10).

We note that the system does not allow for the absolutequantitation of protein affinity, because that would require theability to vary protein concentration in a precise manner,something that cannot be accomplished reliably in vivo. None-theless, we found that the fluorescence signal obtained by theformation of the prey:bait complex correlates reasonably wellwith the equilibrium dissociation constant as determined byBIACore (SI Fig. 7). This good correlation is because theequilibrium dissociation constant is dominated by the dissocia-tion rate. In APEx two-hybrid, the spheroplast f luorescence ata given time reflects the amount of undissociated antibody:an-tigen complex, which in turn depends on the dissociation kineticsand ultimately provides a relative measure of the KD.

The quantitative nature of flow cytometry and the ability touse multiparameter sorting constitute a distinct advantage ofAPEx two-hybrid relative to other systems in which the output

is transcriptional activation or complementation that confers cellgrowth under selective conditions. Because proteins are ex-pressed in secreted form, APEx two-hybrid is ideally suited fordetecting interactions among proteins that normally residewithin secretory compartments. Nonetheless, fusion to a bacte-rial leader peptide can also allow the translocation of cytoplas-mic proteins into the E. coli periplasm. Aberrant disulfide bondformation in cytoplasmic proteins exposed to the oxidativeenvironment of the periplasm can be abolished by using mutantstrains in which the machinery responsible for disulfide bondformation (the Dsb system) had been inactivated (29). There-fore, we expect APEx two-hybrid to prove useful for the analysisof interactions among proteins that fold in the cytoplasm, as wellas the periplasm. For example, in studies to be reported else-where, we have carried out an APEx two-hybrid selection in anE. coli dsbA� strain and isolated intrabody variants of the 14B7scFv that do not require disulfide bonds for folding and thereforecan be expressed in active form within the cytoplasm (our ownunpublished data). Although the work presented here wasfocused on the selection of antibodies in line with the interestsof our laboratory, the APEx two-hybrid system should also beuseful for proteomic studies.

Materials and MethodsStrains, Plasmids, and Growth Conditions. E. coli Jude-1[(DH10BF�::Tn10 (Tetr)] was used throughout this study (30). All plas-mids and primers used are summarized in SI Tables 1 and 2, andsimple schematic representations of expression system can befound in SI Fig. 5. Detailed methods for plasmid constructionsare described in SI Methods.

Cells were grown in Terrific Broth (TB) (Becton DickinsonDifco, Sparks, MD) at 25°C, unless specified otherwise. Proteinsynthesis was induced by adding 1 mM IPTG. Four hours later,the cells were converted to spheroplasts (20), resuspended inPBS, and labeled with 100 nM of the appropriate fluorescentprobes [anti-FLAG-FITC (Sigma–Aldrich, St. Louis, MO), PA-FITC (List Biological Laboratories, Campbell, CA), or anti-FLAG-PE (Prozyme, San Leandro, CA)] at room temperaturefor 30 min.

scFv Library Screening. Spheroplasts cells were sorted on a Moflo(DakoCytomation, Fort Collins, CO) droplet deflection flowcytometer by using a 488-nm Argon laser for excitation anddetection through a 530/40 band-pass filter. Sorted cells wereimmediately resorted. The scFv genes in the sort solution wereamplified by PCR, and the DNA was digested with SfiI andsubcloned into original vector (p14B7-D4). The resulting trans-formants were subjected to a second round of sorting andresorting at more stringent conditions but were otherwise es-sentially as above.

Fab Library Construction and Screening of Expression Maturated Fab.pAPEx-M18.1humFab, a plasmid for expression of Fab antibodyfragments in the APEx format was constructed (see SI Methods).The sequence of M18.1 hum Fab is shown in SI Fig. 9. For easycloning, the second XbaI site in pAPEx-M18.1hum Fab wasremoved by ligation with heavy chain containing complementarySpeI site at the 3� end (pAPEx-M18.1humFab2). A library wasconstructed by randomization of the pelB leader peptide and VHregion of the Fab gene by error-prone PCR with MoPac-F andMJ no. 14 primers, as above. The DNA was ligated intopAPEx-M18.1hum Fab, giving rise to 7 � 106 transformants.Cultures were grown at 37°C, protein synthesis was induced byadding 1 mM IPTG, and then the cells were grown for anadditional 4 h before harvesting and spheroplasting as above.The spheroplasted cells were labeled with PA-FITC and anti-FLAG-PE (100 nM), and the cells were sorted as above by using

Jeong et al. PNAS � May 15, 2007 � vol. 104 � no. 20 � 8251

APP

LIED

BIO

LOG

ICA

LSC

IEN

CES

a 530/40 (for FITC) and 570/40 (for PE) band-pass filters. Aftersorting and resorting, the gene fragment consisting of the pelBsignal sequence and the VH gene in the sort solution wasamplified by PCR and, after XbaI-HindIII digestion, was sub-cloned into original vector (pAPEx-M18.1humFab2). Aftertransformation into E. coli Jude-1, a second round of sorting wasperformed at more stringent conditions.

Protein Purification and BIACore Analysis. Gene encoding for af-finity improved scFvs were subcloned into SfiI-digestedpMoPac16 (31) in which the scFv gene is fused to human � lightchain to create a scAb. Cells from 1-liter cultures grown inTerrific Broth (TB) media were fractionated by osmotic shock,and then monomeric scAbs were purified by immobilized metalaffinity chromatography (IMAC) (Qiagen, Madison, WI) fol-lowed by gel filtration FPLC on a Superdex 200 HR10/30 column(Amersham Biosciences, Piscataway, NJ) as described in ref. 32.Similarly, Fab proteins were expressed from cells harboring Fab

expression plasmids and purified from 1-liter culture by anti-FLAG M2 affinity gel (Sigma–Aldrich).

ELISAs were performed on microtiter well plates coated with50 �l of 2 �g/ml recombinant PA63 (List Biological Laboratories,Campbell, CA) and blocked with 2% (wt/vol) milk–PBS buffer,as described in ref. 33. Antigen dissociation kinetics of purifiedantibodies were determined by surface plasmon resonance(SPR) analysis using a BIACore 3000 instrument (BIACore,Piscataway, NJ) as described in ref. 20. Protein concentrationswere quantified by using a microbicinchoninic acid quantifica-tion kit (Pierce, Rockford, IL).

We thank Clint Leysath and Christian Cobaugh for performing theBIACore measurements and Sang Taek Jung and Thomas Van Blarcomfor help with protein purification and for reading the manuscript. Thiswork was supported by National Institutes of Health Grant U01 AI56431(to B.L.I.), Department of Defense Grant DAAD17-01-D0001 (to B.L.I.and G.G.), and the Institute for Cell and Molecular Biologies Programof the University of Texas.

1. Fields S, Song O (1989) Nature 340:245–246.2. Brent R, Ptashne M (1985) Cell 43:729–736.3. Hu JC, Kornacker MG, Hochschild A (2000) Methods 20:80–94.4. Remy I, Michnick S (2004) Methods Mol Biol 261:411–426.5. Eyckerman S, Lemmens I, Lievens S, Van der Heyden J, Verhee A, Vandeker-

ckhove J, Tavernier J (2002) Sci STKE 2002:PL18.6. Eyckerman S, Verhee A, der Heyden JV, Lemmens I, Ostade XV, Vandeker-

ckhove J, Tavernier J (2001) Nat Cell Biol 3:1114–1119.7. Levin AM, Weiss GA (2006) Mol Biosyst 2:49–57.8. Fields S (2005) Febs J 272:5391–5399.9. MacDonald ML, Lamerdin J, Owens S, Keon BH, Bilter GK, Shang Z, Huang

Z, Yu H, Dias J, Minami T, et al. (2006) Nat Chem Biol 2:329–337.10. Nyfeler B, Michnick SW, Hauri HP (2005) Proc Natl Acad Sci USA 102:6350–

6355.11. Pollock S, Kozlov G, Pelletier MF, Trempe JF, Jansen G, Sitnikov D, Bergeron

JJ, Gehring K, Ekiel I, Thomas DY (2004) EMBO J 23:1020–1029.12. Hennecke F, Muller A, Meister R, Strelow A, Behrens S (2005) Protein Eng Des

Sel 18:477–486.13. Hayhurst A, Georgiou G (2001) Curr Opin Chem Biol 5:683–689.14. Bradbury A, Velappan N, Verzillo V, Ovecka M, Chasteen L, Sblattero D,

Marzari R, Lou J, Siegel R, Pavlik P (2003) Trends Biotechnol 21:312–317.15. Ohara R, Knappik A, Shimada K, Frisch C, Ylera F, Koga H (2006) Proteomics

6:2638–2646.16. Koch H, Grafe N, Schiess R, Pluckthun A (2006) J Mol Biol 357:427–441.17. Demarest SJ, Chen G, Kimmel BE, Gustafson D, Wu J, Salbato J, Poland J,

Elia M, Tan X, Wong K, et al. (2006) Protein Eng Des Sel 19:325–336.18. Rauchenberger R, Borges E, Thomassen-Wolf E, Rom E, Adar R, Yaniv Y,

Malka M, Chumakov I, Kotzer S, Resnitzky D, et al. (2003) J Biol Chem278:38194–38205.

19. Gill DS, Wong YW, Margolies MN (1997) Biotechnol Prog 13:692–694.20. Harvey BR, Georgiou G, Hayhurst A, Jeong KJ, Iverson BL, Rogers GK (2004)

Proc Natl Acad Sci USA 101:9193–9198.21. Krishnanchettiar S, Sen J, Caffrey M (2003) Protein Expr Purif 27:325–

330.22. Maynard JA, Maassen CB, Leppla SH, Brasky K, Patterson JL, Iverson BL,

Georgiou G (2002) Nat Biotechnol 20:597–601.23. Mohamed N, Clagett M, Li J, Jones S, Pincus S, D’Alia G, Nardone L, Babin

M, Spitalny G, Casey L (2005) Infect Immun 73:795–802.24. Rosovitz MJ, Schuck P, Varughese M, Chopra AP, Mehra V, Singh Y,

McGinnis LM, Leppla SH (2003) J Biol Chem 278:30936–30944.25. Hayhurst A (2000) Protein Expr Purif 18:1–10.26. Carter P, Kelley RF, Rodrigues ML, Snedecor B, Covarrubias M, Velligan MD,

Wong WL, Rowland AM, Kotts CE, Carver ME, et al. (1992) Biotechnology(NY) 10:163–167.

27. Humphreys DP (2003) Curr Opin Drug Discov Devel 6:188–196.28. Rothlisberger D, Honegger A, Pluckthun A (2005) J Mol Biol 347:773–

789.29. Segatori L, Paukstelis PJ, Gilbert HF, Georgiou G (2004) Proc Natl Acad Sci

USA 101:10018–10023.30. Griswold KE, Kawarasaki Y, Ghoneim N, Benkovic SJ, Iverson BL, Georgiou

G (2005) Proc Natl Acad Sci USA 102:10082–10087.31. Hayhurst A, Happe S, Mabry R, Koch Z, Iverson BL, Georgiou G (2003)

J Immunol Methods 276:185–196.32. Mabry R, Rani M, Geiger R, Hubbard GB, Carrion R, Jr., Brasky K, Patterson

JL, Georgiou G, Iverson BL (2005) Infect Immun 73:8362–8368.33. Mabry R, Brasky K, Geiger R, Carrion R, Jr., Hubbard GB, Leppla S, Patterson

JL, Georgiou G, Iverson BL (2006) Clin Vaccine Immunol 13:671–677.

8252 � www.pnas.org�cgi�doi�10.1073�pnas.0702650104 Jeong et al.