Proc. Natl. Acad. Sci. USAVol. 90, pp. 10444-10448, November 1993Biochemistry

Production and fluorescence-activated cell sorting of Escherichiacoli expressing a functional antibody fragment on theexternal surfaceJOSEPH A. FRANCISCO*, ROB CAMPBELLt, BRENT L. IVERSONt§, AND GEORGE GEORGIOU*§Departments of *Chemical Engineering and tChemistry and Biochemistry, The University of Texas, Austin, TX 78712; and tBecton Dickinson ResearchCenter, P.O. Box 12016, Research Triangle Park, NC 27709-2076

Communicated by Lester J. Reed, August 9, 1993

ABSTRACT We have expressed a single chain Fv (scFv)antibody fragment, consisting of the variable heavy and vari-able light domains from two separate anti-digoxin monoclonalantibodies, on the external surface ofEscherichia coli by fusingit to an Lpp-OmpA hybrid previously shown to direct heter-ologous proteins to the cell surface. This scFv fusion wasexpressed at a high level and was shown to bind the hapten withhigh affinity and specificity. Whole cell ELISAs, fluorescencemicroscopy, protease sensitivity, and flow cytometry all con-firmed that the scFv was anchored on the outer membrane andwas accessible on the surface. Utilizing fluorescence-activatedcell sorting, we were able to specifically enrich scFv-producingcells from a 105-fold excess ofcontrol cells in only two steps. Theexpression of antibody fragments on the surface of E. coli isbeing evaluated as an attractive method for the in vitro pro-duction and selection of useful antibody fragments.

A general method for the identification of protein sequencespossessing a high affinity for a particular ligand can greatlyfacilitate a better understanding of biomolecular recognitionand will have important applications in fields such as drugdesign and antibody engineering. One of the most successfulapproaches for identifying proteins having a desired affinityis by screening large libraries generated by genetic means(1-4). A widely used technique for screening such libraries isbased on the display of proteins or peptide sequences on thesurface of flamentous phage (1, 2). Briefly, recombinantproteins are displayed on the surface of phage particles asfusions to the N terminus ofeither gIII or gVIII coat proteins.Subsequently, the library of phage clones is screened forbinding to the target molecule attached to a solid phase.Successive rounds of binding and elution result in selectiveenrichment of phage possessing proteins or peptide se-quences that are specific for the target molecule. However,the binding to the target immobilized on a solid phasedepends not only on molecular affinity but also on otherfactors such as surface characteristics, number ofattachmentpoints (avidity effects), and hydrodynamic conditions.

Libraries of antibody molecules expressed as single chainFv (scFv) or Fab fragments have been displayed on thesurface of phage and screened for binding toward immobi-lized antigen (5). Several recent studies with phage havedemonstrated that the in vitro screening of antibody librariesrepresents a promising method for the isolation of interestingantibody fragments specific for a variety of different mole-cules. Libraries have been constructed from the PCR prod-ucts derived from antibody-producing cells isolated fromanimals or through a semisynthetic strategy involving therandomization of antibody complementarity-determining re-gions using synthetic oligonucleotides (6-12). This latter

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

approach is especially interesting as a way to expand theimmunological repertoire beyond what is found in nature.

In theory, the cell-associated fluorescence due to thebinding of a fluorescent hapten could be employed to selectEscherichia coli displaying surface-anchored antibody frag-ments from a large excess of negative cells (13). Along theselines, a previous attempt at placing an active antibodyfragment on the surface of E. coli has been reported (14, 15).However, this construct was not fully characterized or usedfor any selection experiments.

Herein, we report the production and characterization of afunctional scFv antibody fragment attached to the outersurface of E. coli. Using a fluorescently labeled hapten, wehave found that these constructs can be efficiently selectedusing flow cytometry. A 105-fold enrichment was reproduc-ibly obtained after only two rounds of sorting followed bygrowth of the sorted cell population. Our E. coli surfaceexpression and flow cytometry selection system is beinginvestigated as a "user friendly" technology for the produc-tion and isolation of interesting antibody fragments in vitro.

MATERIALS AND METHODSBacterial Strains, Plasmids, and Growth Conditions. E. coli

strain JM109 (endAl recAl gyrA thi-l hsdR17 (ri, mk ) relAlsupE44 A(lac-proAB)/F' traD36proAB lacIq lacZAM15) wasused for all experiments. pTX101 codes for an Lpp-OmpA-3-lactamase fusion (16). pTX152 codes for an Lpp-OmpA-scFv(digoxin) fusion, where the scFv(digoxin) is an anti-digoxin scFv consisting ofthe heavy- and light-chain variableregions (VH and VL). The VH and VL, joined by a 15-aminoacid [(Gly)4Ser]3 linker (17), were amplified from mRNAisolated from two separate anti-digoxin hybridomas. An11-amino acid peptide from the herpes simplex virus (HSV)glycoprotein (Novagen) was introduced at the C terminus ofthe scFv for analytical purposes. To construct pTX152 thebla from pTX101 was first removed by digestion with EcoRIand BamHI. Subsequently, the amplified gene coding for theanti-digoxin scFv was digested with EcoRI and BamHI andligated into pTX101. Both pTX101 and pTX152 carry thechloramphenicol-resistance gene.

Cultures were grown in LB medium (Difco) supplementedwith 0.2% glucose and chloramphenicol (50 p,g/ml).ELISA. Overnight cultures grown at 24°C were harvested,

resuspended in phosphate-buffered saline (PBS) at OD600 =2.0, and lysed by passage through a French pressure cell at20,000 psi (1 psi = 6.89 kPa). The lysates were then dilutedwith 1 volume ofPBS containing 2.0% bovine serum albumin(PBS/2% BSA) and 5 mM of the protease inhibitor phenyl-

Abbreviations: scFv, single chain Fv; VH, heavy-chain variableregion; VL, light-chain variable region; BSA, bovine serum albumin;FACS, fluorescence-activated cell sorting; FITC, fluorescein iso-thiocyanate; HSV, herpes simplex virus.§To whom reprint requests should be addressed.

10444

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

Proc. Natl. Acad. Sci. USA 90 (1993) 10445

methylsulfonyl fluoride. Microtiter plates (96 wells) wereincubated overnight at 37°C with 100 ,ul of 100 p,g/ml of eitherBSA or digoxin-conjugated BSA (digoxin-BSA) in 0.1 Msodium carbonate buffer (pH 9.2). All subsequent steps werecarried out at room temperature. The wells were fixed for 5min with 100 ,ul of methanol and were then blocked for 45 minwith 200 Al of PBS/1% BSA. After removing the blockingsolution, the wells were incubated for 2 hr with 100 ul oflysates. Subsequently, they were washed three times with200 ,l ofPBS/0.1% Tween 20 and incubated for 1 hr with 100,ul per well of monoclonal antibodies against the HSV peptideor antiserum against ,-lactamase. The wells were againwashed three times with PBS/0.1% Tween, incubated for 1hr with 100 ,ul of the appropriate secondary antibodiesconjugated with horseradish peroxidase, and finally washedfive times with PBS/0.1% Tween and two times with PBS.After addition of the substrate 2,2'-azinobis(3-ethylbenzthi-azoline-6-sulfonic acid) (Pierce) the absorbance of each wellwas measured at 410 nm.Whole cell ELISAs were performed as described above

except that 100-,l samples of overnight cultures that hadbeen resuspended in PBS/1% BSA at OD600 = 1.0 were usedinstead of cell lysates.

Fluorescence Microscopy and Fluorescence-Activated CellSorting (FACS). For fluorescence microscopy, overnightcultures grown at 24°C were harvested, resuspended at OD600= 0.5 in PBS containing 10-7 M fluorescein-conjugateddigoxin (digoxin-FITC; ILS LTD, London), and incubated atroom temperature for 1 hr. Prior to microscopy the cells werewashed once with PBS and resuspended in equal volumes ofPBS and Vectashield mounting medium (Vector Laborato-ries) at 0)D600 of 2.0.For the FACS experiments, overnight cultures grown at

24°C were harvested and resuspended in PBS at OD600 = 0.5.Positive (JM109/pTX152) and control (JM109/pTX101) cellswere mixed at the desired ratios and the mixture was incu-bated at room temperature for 1 hr with 10-7 M digoxin-FITC. The cells were then pelleted by centrifugation, resus-pended in PBS at 3 x 107 cells per ml, and either counted orsorted on the basis of fluorescence intensity using a FACSortfluorescence-activated cell sorter (Becton Dickinson). Flowrates ofbetween 12 and 60 ,ul of sample per minute were used,with 488-nm excitation (argon-ion laser) and emission detec-tion between 515 and 545 nm. Counting runs involved at least10,000 events per experiment. The sorting experiments in-volved 100,000-500,000 cells per run. After sorting, theselected cells were added to 50 ml of LB broth containing 50jig of chloramphenicol per ml and 0.2% glucose and thengrown overnight at 37°C with shaking. Subsequently, thecells were subcultured into fresh medium and grown over-night at 24°C with shaking. Finally, the culture was incubatedfor 1 hr with 10-7 M digoxin-FITC, resuspended in PBS at 3X 107 cells per ml, and run through another round of FACS.General Procedures. The protein composition of whole cell

membrane fractions isolated from overnight cultures wasanalyzed by SDS/PAGE on 12% acrylamide gels and byWestern blotting (18) using anti-HSV monoclonal antibodies(Novagen) and anti-OmpA antiserum. Whole membrane frac-tions were prepared as described (16).

RESULTS AND DISCUSSIONExpression of Lpp-OmpA(46-159)-scFv(digoxin). The cell

envelope ofE. coli and other Gram-negative bacteria consistsof the inner membrane (cytoplasmic membrane), the pepti-doglycan cell wall, and the outer membrane. Although thelatter normally serves as a barrier to protein secretion, wehave constructed a chimeric targeting sequence that, whenfused to normally soluble proteins, can direct them to the cellsurface (16). The chimeric targeting sequence consists of the

leader peptide and first 9 amino acids of the E. coli majorouter membrane lipoprotein (Lpp) and amino acids 46-159 ofthe outer membrane protein OmpA. The function of the Lppis to direct the chimera to the outer membrane. Laukkanenet al. (19) have recently demonstrated that an active scFvantibody fragment can be anchored to the internal face of theE. coli outer membrane when fused to this region ofLpp. TheOmpA region in our chimeric targeting sequence then trans-verses the membrane leaving its C terminus exposed on thecell's exterior. We have used Lpp-OmpA(46-159) fusions toanchor a variety of proteins such as ,B-lactamase, a cellulosebinding protein, and alkaline phosphatase on the E. colisurface (refs. 16 and 20; C. Stathopoulos, C. F. Earhart andG.G., unpublished data).For the display of antibodies on the surface ofE. coli a scFv

specific for digoxin was constructed as described in Materialsand Methods. For detection purposes, a 33-bp fragmentencoding an epitope from the HSV glycoprotein was added tothe 3' end of the VL to facilitate detection by immunochem-ical techniques. The nucleotide sequence of the antibodyfragment is shown in Fig. 1. The Lpp-OmpA(46-159)-scFv(digoxin) construct is encoded by the plasmid pTX152.E. coli JM109 was transformed with pTX152, grown at 37°C,and used to isolate total membranes. In Western blots a bandof the expected size (42 kDa) was recognized by antibodiesspecific for both OmpA and the HSV peptide (Fig. 2A). Thenear absence of lower molecular mass bands crossreactingwith either the anti-HSV or the anti-OmpA antibodies indi-cates that the scFv(digoxin) was not subjected to proteolysis,probably because it is anchored on the cell surface (seebelow), and consequently it is physically separated fromintracellular proteases. The intensity of the Lpp-OmpA(46-159)-scFv(digoxin) band in Fig. 2A is comparable to that ofthe native OmpA band. The latter is a highly expressedprotein that is present in the E. coli outer membrane at about100,000 copies per cell (21). Thus the level of expression ofLpp-OmpA(46-159)-scFv(digoxin) appears to be on the or-der of 50,000-100,000 copies per cell.

Display of scFv(digoxin) on the Cell Surface. The ability ofthe scFv domain of the fusion protein to bind the hapten wasdetermined by ELISA. Whole cell lysates from JM109/pTX152 and JM109/pTX101 were incubated on microtiterwells that had been coated with either digoxin-conjugatedBSA (digoxin-BSA) or unconjugated BSA. Subsequently, thewells were treated with antibodies against the HSV peptideor ,B-lactamase as necessary to detect the respective fusionproteins. JM109/pTX152 lysates bound specifically to wellscoated with digoxin-BSA but not to unconjugated BSA,whereas the lysates from the control strain, JM109/pTX101,did not give a signal with either. Thus, Lpp-OmpA(46-159)-scFv(digoxin) is active and can bind to the hapten specifi-cally.

Fig. 2B shows the results of ELISAs using intact cells.Samples containing the same number of cells were used in allexperiments described here. Cells containing pTX101 gavethe same low signal when incubated on microtiter wellscoated either with unconjugated BSA or with digoxin-BSA.A similar weak signal was detected with JM109/pTX152incubated on BSA-coated wells and is presumably due tononspecific binding. In contrast, a much higher absorbancewas evident in wells coated with the digoxin-BSA, indicatingthat there are active fusion protein molecules on the cellsurface.The display of the active scFv antibody on the cell surface

was confirmed by fluorescence microscopy (Fig. 3). JM109/pTX152 cells were grown overnight at 24°C, incubated with10-7 M digoxin-FITC for 1 hr, and washed. As shown in Fig.3, all of the cells visible with phase-contrast microscopy gavea strong fluorescence signal. In control experiments, whenJM109/pTX101 cells were incubated with the same concen-

Biochemistry: Francisco et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

10446 Biochemistry: Francisco et al.

Variable Heavy ChainGlu Val Gln Leu Gln Gln Ser Gly Pro Glu Leu Val Lys Pro Gly Ala Ser Val Arg Met Ser Cys Lys Ser SerGAA GTT CAA CTG CAA CAG TCT GGT CCT GAA TTG GTT AAA CCT GGC GCC TCT GTG CGC ATG TCC TGC AAA TCC TCA

Gly Tyr Ile Phe Thr Asp Phe Tyr Met Asn Trp Val Arg Gln Ser His Gly Lys Ser Leu Asp Tyr Ile Gly TyrGGG TAC ATT TTC ACC GAC TTC TAC ATG AAT TGG GTT CGC CAG TCT CAT GGT AAG TCT CTA GAC TAC ATC GGG TAC

Ile Ser Pro Tyr Ser Gly Val Thr Gly Tyr Asn Gln Lys Phe Lys Gly Lys Ala Thr Leu Thr Val Asp Lys SerATT TCC CCA TAC TCT GGG GTT ACC GGC TAC AAC CAG AAG TTT AAA GGT AAG GCC ACC CTT ACT GTC GAC AAA TCT

Ser Ser Thr Ala Tyr Met Glu Leu Arg Ser Leu Thr Ser Glu Asp Ser Ala Val Tyr Tyr Cys Ala Gly Ser SerTCC TCA ACT GCT TAC ATG GAG CTG CGT TCT TTG ACC TCT GAG GAC TCC GCG GTA TAC TAT TGC GCC GGC TCC TCT

LinkerGly Asn Lys Trp Ala Met Asp Tyr Trp Gly His Gly Ala Ser Val Thr Val Ser Ser Gly Gly Gly Gly Ser GlyGGT AAC AAA TGG GCC ATG GAT TAT TGG GGT CAT GGT GCT AGC GTT ACT GTG AGC TCT GGT GGC GGT GGC TCG GGC

Variable Light ChainGly Gly Gly Ser Gly Gly Gly Gly Ser Asp Val Val Met Thr Gln Thr Pro Leu Ser Leu Pro Val Ser Leu GlyGGT GGT GGG TCG GGT GGC GGC GGA TCA GAC ATA GTA CTG ACC CAG TCT CCA GCT TCT TTG GCT GTG TCT CTA GGA

Asp Gln Ala Ser Ile Ser Cys Arg Ser Ser Gln Ser Leu Val His Ser Asn Gly Asn Thr Tyr Leu Asn Trp TyrCAA AGG GCC ACG ATA TCC TGC CGA TCC AGC CAA AGT CTC GTA CAT TCT AAT GGT AAT ACT TAT CTG AAC TGG TAC

Leu Gln Lys Ala Gly Gln Ser Pro Lys Leu Leu Ile Tyr Lys Val Ser Asn Arg Phe Ser Gly Val Pro Asp ArgCAA CAG AAA CCA GGA CAG CCA CCC AAG CTT CTC ATC TAT AAG GTA TCC AAC CGA TTC TCT GGA GTC CCT GCC AGG

Phe Ser Gly Ser Gly Ser Gly Ser Asp Phe Thr Leu Thr Ile Asp Arg Val Glu Glu Asp Asp Ala Ala Ile TyrTTC AGT GGC AGT GGG TCT GAG TCA GAC TTC ACC CTC ACC ATC GAT CCT GTG GAG GAA GAT GAT GCT GCA ATA TAT

RSHTyr Cys Ser Gln Thr Thr His Val Pro Pro Thr Phe Gly Ser Gly Thr Lys Leu Glu Leu Lys Pro Ala Ser GlnTAC TGT AGC CAA ACT ACG CAT GTT CCA CCC ACG TTC GGC TCG GGG ACC AAG CTG GAG CTG AAA CGT GCT AGC CAG

Pro Glu Leu Ala Pro Glu Asp Pro Glu AspCCA GAA CTC GCC CCG GAA GAC CCC GAG GAC

FIG. 1. Nucleotide and amino acid sequences of the anti-digoxin scFv antibody fragment.

tration of digoxin-FITC and then washed, none of the cellsbecame fluorescently labeled (data not shown). Furthermore,protease treatment drastically reduced the ability of the cellsto bind the fluorescently labeled hapten, as evidenced bothby fluorescence microscopy and by FACS (Fig. 4C).The intensity of the fluorescence signal from JM109/

pTX152 was dependent on the cell growth temperature andwas much higher for cultures grown at 24°C instead of 37°C.This is consistent with previous results that show that theamount of proteins expressed on the surface of E. coli byfusion to Lpp-OmpA(46-159) increases as the temperature isdecreased (16, 20). If the efficiency of surface display in thiscase is similar to that of 3-lactamase (16), then at 24°Cvirtually all the scFv antibody chains should be accessible onthe cell surface.

Detection by FACS. Before FACS could be performed itwas first necessary to determine whether E. coli displayingthe scFv(digoxin) antibody and labeled with digoxin-FITCcould be discriminated from the background. Samples of 108cells per ml from cultures grown at 24°C were incubated withdigoxin-FITC at 10-7 M, washed in buffer, and diluted to 3x 106 cells per ml prior to sorting. The samples were thenanalyzed using a FACSort flow cytometer. Fig. 4 A and Bshow that the fluorescence intensity of JM101/pTX152 wasclearly distinguishable from the intrinsic background signal ofcontrol E. coli (in this case, JM101/pTX101). When JM109/pTX152 cells were preincubated with an excess of freedigoxin prior to incubation with the digoxin-FITC, the fluo-rescence intensity of the cells was the same as for thebackground (Fig. 4D). This specific inhibition was also seenusing fluorescence microscopy and demonstrates that thesurface-expressed scFv(digoxin) specifically binds the fluo-rescently labeled hapten in the binding site and is not theresult of nonspecific interactions.Treatment of intact cells with trypsin prior to incubation

with digoxin-FITC almost completely eliminated the popu-lation of fluorescently labeled cells detected by flow cytom-etry (Fig. 4C). In Gram-negative bacteria, the outer mem-brane serves as a barrier to preclude the diffusion of largeextracellular molecules such as proteins. The action of tryp-

sin is assumed to be limited to the proteolysis of proteinsexposed on the external surface of E. coli (22). As such, theabove result provides further evidence that the active scFv-(digoxin) is indeed accessible on the outer surface, free tointeract with molecules in solution.The high background due to light scattering precluded a

detailed quantitative evaluation of the K. for the surface-expressed scFv(digoxin) using cell sorting of samples incu-bated with different concentrations of digoxin-FITC. How-ever, the scFv(digoxin) binding sites appeared to be fullysaturated at concentrations of digoxin-FITC above 10-7 M.Appreciable fluorescent signal was clearly detected atdigoxin-FITC concentrations of 10-9 M. These results areconsistent with a binding constant that is at least within anorder of magnitude of the value of 1 x 10-9 M-1 determinedfor the scFv(digoxin) in solution (R.C. and P. Hamilton,unpublished data).Enrichment of Cells Displaying scFv(digoxin) by FACS. We

found that antibody-expressing cells can be sorted essentiallyquantitatively from a moderate excess of control E. coli in asingle step. Specifically, in mixtures containing JM109/pTX101 at an excess, over JM109/pTX152, of either 100:1 or1000:1, the fraction of the total population that was sorted inthe high fluorescence intensity window was 1.1% and 0.1%,respectively, as expected from the ratio of input cells (datanot shown). To evaluate the potential use of FACS forisolating rare clones from a very large excess of background,JM109/pTX101 and JM101/pTX152 were mixed at a ratio of100,000:1 and labeled with digoxin-FITC; 500,000 cells fromthe input mixture were run through the FACSort flow cy-tometer. In this experiment a wide sorting gate-i.e., theminimum fluorescence required for acceptance of an indi-vidual cell-was selected such that up to 0.2% of the controlcells fell within the sorting window. This ensured that allscFv(digoxin)-expressing cells would be recovered. The cellshaving an allowable fluorescence signal were collected andgrown in fresh medium. Following growth, the cells wereagain run through the FACSort, grown in fresh medium asbefore, and resorted. To ensure the complete absence ofartifacts due to nonspecific cell adhesion in the flow path of

Proc. Natl. Acad Sci. USA 90 (1993)

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

Proc. Natl. Acad. Sci. USA 90 (1993) 10447

AkDa

10680

49.5..... ... .

32.5

27.5

18.5

1 2 3 4 5

B0.8r

0.6 F

04

0.2 F

0.0 -Lysate Whole cell

FIG. 2. (A) Western blot of total membrane fractions fromJM109/pTX101 (lanes 2 and 4) and JM109/pTX152 (lanes 3 and 5).Lanes 2 and 3 were probed with anti-OmpA antiserum at 1:5000dilution. Lanes 4 and 5 were probed with monoclonal anti-HSVantibodies at 1:5000 dilution. Arrowheads indicate the Lpp-OmpA-P-lactamase fusion (lane 2) and the Lpp-OmpA-scFv(digoxin) fusion(lane 3). The 32-kDa band in lanes 2 and 3 corresponds to OmpA.Lane 1, molecular mass markers (in kDa). (B) Lysate and whole cellELISAs of JM109 cells containing plasmid pTX101 (n) or pTX152(o). Samples were incubated on microtiter wells coated with digoxin-conjugated BSA and probed with anti-3-lactamase (pTX101) oranti-HSV (pTX152) antibodies. Absorbance readings were refer-enced to wells that were untreated with either lysates or whole cells.

the FACSort, each run was followed by extensive washingwith bleach. Fig. 4 E-G show the cell fluorescence distributionfor the sorting runs. After only two rounds of growth andsorting, the fluorescence intensity of79%o ofthe cell populationfell within the positive window. A similar enrichment wasreproducibly obtained in three independent experiments. Im-portantly, these results are not due to a growth advantage ofthe cells expressing Lpp-OmpA(46-159)-scFv(digoxin), sincesuccessive regrowth ofthe input cell mixture in the absence ofsorting did not result in any detectable enrichment.To verify that the cells with the increased fluorescent signal

after the final sorting step were indeed JM109/pTX152, wetook advantage of the fact that pTX152 confers resistanceonly to chloramphenicol, whereas the plasmid present in thecontrol cells also confers resistance to ampicillin. A sampleof cells from the final round of FACS was plated on chlor-

FIG. 3. Phase-contrast (A) and fluorescence (B) micrographs ofthe same field of JM109/pTX152 cells after 1 hr of incubation with10-7 M digoxin-FITC.

amphenicol plates and then replica plated on plates contain-ing 100 ,ug of ampicillin per ml. Over 95% of all the coloniesexamined were chloramphenicol resistant/ampicillin sensi-tive (cm+/amp-), consistent with the phenotype expected forJM101/pTX152. As an additional test, plasmid DNA wasisolated from eight cm+/amp- colonies and the presence ofpTX152 was confirmed by restriction analysis.

Condusions. We have demonstrated that a scFv antibodyfragment specific for digoxin can be expressed on the E. coli

Biochemistry: Francisco et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

10448 Biochemistry: Francisco et al.

Ri

0 100 10'0i

400B 400' C 400 D

0.2% 1 81.5i R 5.7% 4 , 1.9%

0.2R, R, R .O -w

102 10310: 10w 101 1o2 103104 °l Ip lol l2 102 104 1 P 101 10231041 100 1'102 103 1o4 0 01100 0010113102i0 10

40(

z400- F 400- G

4 ^ Rl1 0.2% FZ.6%itnt

10° 101 102 103 104 10° 101 102 1(3 1(4 10 101Fluorescence intensity

FIG. 4. Histogram data from FACS. The bar in each graph represents the sorting gate or the fluorescence intensity defined as a positive event.The sorting gate was chosen to maximize the number of positive events while minimizing the number of negative events within the window.All samples were labeled with 10-7 M digoxin-FITC. (A) JM109/pTX101 sample used as a negative control. (B) JM109/pTX152 sample usedas a positive control. (C) JM109/pTX152 pretreated with 0.2 mg of trypsin per ml. (D) JM109/pTX152 pretreated with free digoxin. (E) A100,000:1 mixture of JM109/pTX101:JM109/pTX152 prior to the first cell sorting run. (F) A 100,000:1 mixture after growing cells recoveredfrom the first cell sorting run. (G) A 100,000:1 mixture after growing cells recovered from the second cell sorting run.

surface and binds to the hapten with high affinity. UsingFACS, the antibody-expressing cells were recovered from a105-fold excess ofcontrol E. coli in only two rounds of sortingand regrowth of the sorted cell population (Fig. 4). Sorting isboth rapid and efficient. Using a low-end flow cytometer it ispossible to sort 1 x 106 cells per hour when operated so thatthe cells pass through the laser beam single file. Highersorting rates can be obtained with a larger model (up to 5 x107 cells per hour) and even higher still by sorting multiplecells at a time (at the expense of increased background thatcan presumably be eliminated by resorting the selectedpopulation). In addition, the sorting of positive clones isessentially quantitative and is limited only by the accuracy ofthe flow cytometer, which is on the order of 95% (23).

Preliminary experiments have been undertaken using ex-pression on the E. coli surface and flow cytometry to selectfor high-affinity antibodies from libraries derived from im-munized animals as well as from a semisynthetic library andalso screening peptide libraries for sequences with highaffinity for various fluorescently labeled receptors. It isanticipated that by using very low concentrations of fluores-cently labeled hapten or receptor (i.e., <10-8 M), it should bepossible to select directly for antibodies or peptides with veryhigh affinity (i.e., Ka > 108 M-1). In our system avidity effectsdue to multivalent binding do not come into play in theselection of high-affinity clones. In contrast, the high numberof scFv (104 molecules or higher) on the cell surface shouldfacilitate fine affinity discriminations. These characteristics,combined with the simple nature of the techniques involved,could make our E. coli-based system an attractive alternativeto phage display for the isolation of high-affinity moleculesfrom large libraries.

We thank Paul Hamilton for synthesis of the hybrid scFv, Kath-leen Harrison for excellent technical assistance, and Ulf Henning forproviding us with anti-OmpA antibodies. This work was supportedby the Whitaker Foundation for Biomedical Engineering Researchand National Science Foundation Grants BCS-9013007 and BCS-9212305 to G.G. and Texas Higher Education Coordinating BoardGrant 3658-06 to B.L.I.

1. Scott, J. K. & Smith, G. P. (1990) Science 249, 386-390.

2. Devlin, J. J., Panganiban, L. C. & Devlin, P. E. (1990) Science249, 404-406.

3. Cull, M. G., Miller, J. F. & Schatz, P. J. (1992) Proc. Natl.Acad. Sci. USA 89, 1865-1869.

4. Bass, S., Greene, R. & Wells, J. A. (1990) Protein Struct.Funct. Genet. 8, 309-314.

5. Chiswell, D. J. & McCafferty, J. (1992) Trends Biotechnol. 10,80-84.

6. Barbas, C. F., III, Kang, A. K., Lemer, R. A. & Benkovic,S. J. (1991) Proc. Natl. Acad. Sci. USA 88, 7978-7982.

7. Barbas, C. F., III, Bain, J. D., Hoekstra, D. M. & Lerner,R. A. (1992) Proc. Natl. Acad. Sci. USA 89, 4457-4461.

8. Garrard, L. J., Yang, M., O'Connell, M. P., Kelley, R. F. &Henner, D. J. (1991) BiolTechnology 9, 1373-1377.

9. Collet, T. A., Roben, P., O'Kennedy, R., Barbas, C. F., HI,Burton, D. R. & Lerner, R. A. (1992) Proc. Natl. Acad. Sci.USA 89, 10026-10030.

10. McCafferty, J., Griffiths, A. D., Winter, G. & Chiswell, D. J.(1990) Nature (London) 348, 552-554.

11. Marks, J. D., Hoogenboom, H. R., Bonnert, T. P., McCaf-ferty, J., Griffiths, A. D. & Winter, G. (1991) J. Mol. Biol. 222,581-597.

12. Clackson, T., Hoogenboom, H. R., Griffiths, A. D. & Winter,G. (1991) Nature (London) 352, 624-628.

13. Geisow, M. J. (1992) Trends Biotechnol. 10, 75-76.14. Fuchs, P., Breitling, F., Dubel, S., Seehaus, T. & Little, M.

(1991) BiolTechnology 9, 1369-1372.15. Little, M., Fuchs, P., Breitling, F. & Dubel, S. (1993) Trends

Biotechnol. 11, 3-5.16. Francisco, J. A., Earhart, C. F. & Georgiou, G. (1992) Proc.

Natl. Acad. Sci. USA 89, 2713-2717.17. Huston, J. S., Levinson, D., Mudgett-Hunter, M., Tai, M.-S.,

Novotny, J., Margolies, M. N., Ridge, R. J., Bruccoleri, R. E.,Haber, E., Crea, R. & Oppermann, H. (1988) Proc. Natl. Acad.Sci. USA 85, 5879-5883.

18. Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D.,Seidman, J. G., Smith, J. A. & Struhl, K. (1989) CurrentProtocols in Molecular Biology (Wiley, New York).

19. Laukkanen, M. L., Teeri, T. T. & Keinanen, K. (1993) ProteinEng. 6, 449-454.

20. Francisco, J. A., Stathopoulos, C., Warren, R. A. J., Kilburn,D. G. & Georgiou, G. (1993) BiolTechnology 11, 491-495.

21. Lugtenberg, B. & van Alphen, L. (1983) Biochim. Biophys.Acta 737, 51-115.

22. Kornacker, M. G. & Pugsley, A. P. (1990) Mol. Microbiol. 4,1101-1109.

23. Tanke, H. J. & van der Keur, M. (1992) Trends Biotechnol. 11,55-62.

Ab- lr

Proc. Natl. Acad. Sci. USA 90 (1993)

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021