efficient recovery of high-affinity antibodies from a single-chain fab yeast display library

doi:10.1016/j.jmb.2009.04.019 J. Mol. Biol. (2009) 389, 365–375

Available online at www.sciencedirect.com

Efficient Recovery of High-Affinity Antibodies from aSingle-Chain Fab Yeast Display Library

Laura M. Walker, Diana R. Bowley and Dennis R. Burton⁎

Department of Immunology andMicrobial Science and IAVINeutralizing Antibody Center,The Scripps Research Institute,10550 North Torrey Pines Road,La Jolla, CA 92037, USA

Received 19 January 2009;received in revised form3 April 2009;accepted 9 April 2009Available online16 April 2009

*Corresponding author. E-mail [email protected] used: scFv, single-c

fragment; Fab, antigen binding fragsingle-chain Fab; MFI, mean fluoresCDRH3, heavy-chain third complemregion; mAb, monoclonal antibody;immunodeficiency virus; IgG, immustreptavidin–phycoerythrin; PBS, phsaline; BSA, bovine serum albumin.

0022-2836/$ - see front matter © 2009 E

Yeast display is a powerful technology for the isolation of monoclonalantibodies (mAbs) against a target antigen. Antibody libraries have beendisplayed on the surface of yeast as both single-chain variable fragment(scFv) and antigen binding fragment (Fab). Here, we combine these twoformats to display well-characterizedmAbs as single-chain Fabs (scFabs) onthe surface of yeast and construct the first scFab yeast display antibodylibrary.When expressed on the surface of yeast, two out of three anti-humanimmunodeficiency virus (HIV)-1 mAbs boundwith higher affinity as scFabsthan scFvs. Also, the soluble scFab preparations exhibited binding andneutralization profiles comparable to that of the corresponding Fabfragments. Display of an immune HIV-1 scFab library on the surface ofyeast, followed by rounds of sorting against HIV-1 gp120, allowed for theselection of 13 antigen-specific clones. When the same cDNA was used toconstruct the library in an scFv format, a similar number but a lower affinityset of clones were selected. Based on these results, yeast-displayed scFablibraries can be constructed and selected with high efficiency, characterizedwithout the need for a reformatting step, and used to isolate higher-affinityantibodies than scFv libraries.

© 2009 Elsevier Ltd. All rights reserved.

Keywords: immune repertoire; antibody display format; HIV-1; flowcytometry; antibody library
Edited by J. Karn
Introduction

Over the past decade, yeast display has emerged asa leading technology for the selection and character-ization of monoclonal antibodies (mAbs) from bothimmune and nonimmune libraries.1–5 The majorityof yeast display antibody libraries that have beenconstructed to date utilize the single-chain variablefragment (scFv) format,1–3 and there are only threeexamples in the literature of combinatorial antigenbinding fragment (Fab) yeast display libraries.6–8

Because both scFv and Fab libraries have uniqueadvantages and disadvantages, there is controversy

ess:

hain variablement; scFab,cence intensity;entary determiningHIV, humannoglobulin G; SA–PE,osphate-buffered

lsevier Ltd. All rights reserve

over which format is better suited for antibodydisplay. ScFvs are generally less stable and moreprone to aggregation than Fab fragments,9,10 anddue to their decreased stability, scFvs generally bindwith lower affinity than Fabs.11 Furthermore, it hasbeen shown that certain scFvs lose antigen bindingand/or neutralization activity after conversion intoFabs or immunoglobulin Gs (IgGs).7,12,13 However,because Fab fragments are larger in size and requirethe assembly of two separate polypeptide chainswith a disulfide bond, they generally fold lessefficiently than scFvs, which leads to decreased solu-bility and lower levels of expression in Escherichiacoli.14 The most commonly used method to displayFab libraries on the surface of yeast, which employsyeast mating, allows for the display of unpairedheavy chains.7 As a result, nonspecific binding ofheavy-chain-only clones could dominate libraryselection and result in the loss of specific binders.Recently, a single-chain Fab (scFab) was cons-

tructed and shown to be compatible with phagedisplay technology and suitable for soluble expres-sion in E. coli.15,16 By combining the high levels ofexpression of the scFv with the stability of the Fabfragment, the scFab might represent a useful anti-

d.

366 scFab Yeast Display Antibody Library

body display format for yeast display technology.Here, we have displayed three well-characterizedanti-human immunodeficiency virus (HIV)-1 anti-bodies as scFabs on the surface of yeast andconstructed the first scFab yeast display library. Inorder to evaluate the scFab as a selection platform,we used the same cDNA as starting material for theconstruction of libraries in the scFv and scFabformats, and the libraries were subsequently selec-ted against the same antigens. Our results suggestthat scFab libraries displayed on the surface of yeastcan be constructed and selected with higherefficiency than Fab libraries and yield higher-affinityantibodies than scFv libraries.

Results

Generation of scFab and scFv yeast displayvectors

The scFab yeast display vector was generatedfrom the vector pPNL200.7 Using overlap PCR, a

Fig. 1. scFab yeast display platform. (a) scFab expression isconnects the heavy and light chains, and the Aga2 protein is ethat form the interchain disulfide bond were removed. (b) Mdetected with SA–PE. Using an anti-c-myc antibody and goattag allows for visualization of yeast expression.

34-amino-acid linker containing the sequence(SGGG)2(SEGGG)4(SGGG)SG was inserted betweenthe heavy and light chains. Primers were designedto remove the two cysteines involved in the disul-fide bond connecting the two chains (Fig. 1a).Removal of this disulfide bond has been shown toincrease the solubility of scFabs when expressed inE. coli.15 Next, BssHII and SalI restriction sites wereinserted for light-chain gene cloning and NheI andNcoI sites were introduced for heavy-chain genecloning. These restriction enzyme sites were chosenbecause they are rarely found in germ line antibodygenes. Stuffer sequences containing only the light-chain or heavy-chain constant region were insertedbetween restriction sites to aid in cloning.It has been demonstrated that linear plasmid

vectors that circularize by homologous recombina-tion transform yeast with higher efficiency thansupercoiled plasmids.17 The traditional method foryeast homologous recombination, which involveselectroporation of PCR products with 5′ and 3′homology to a linearized vector,18 was not usedbecause it requires ∼100 μg of PCR product togenerate large (109) libraries. This quantity of PCR

under control of the Gal1 promoter. A 34-amino-acid linkerxpressed as a C-terminal fusion to the scFab. The cysteinesonomeric gp120 is labeled with biotinylated HIVIG andanti-mouse Alexa647 (GaM A647) dye, a C-terminal c-myc

Fig. 2. (a) Binding curves for yeast-displayed 4E10,b12, and b6 scFvs and scFabs are shown in (A)–(C). TheMFI, determined by flow cytometry, is plotted againstincreasing concentrations of antigen to determine the Kdvalues listed in (b). Data are mean±SD from threeindependent experiments.

367scFab Yeast Display Antibody Library

product was not produced from the limited amountof starting material; therefore, we inserted a 51-base-pair linker with homology to the adjacentvector sequence directly after the heavy chain to beused for homologous recombination (Fig. 1a). Bydigesting the library at a site located between thetwo homologous sequences, we found that the yeasttransformation efficiency was increased by at least10-fold.

Validation of the new display format using mAbs

The new display systemwas validated using threewell-characterized anti-HIV-1 antibodies: 4E10, b12,and b6. 4E10 targets the membrane proximalexternal region on HIV-1 gp41 and both b12 andb6 bind to epitopes overlapping the CD4 bindingsite on HIV-1 gp120.19 Both 4E10 and b12 exhibitbroad neutralization of HIV-1, whereas b6 onlyneutralizes a few very sensitive viral isolates.19 ThemAbs were expressed as both scFabs and scFvs onthe surface of yeast, and the binding affinities werecalculated by titrating the amount of antigen (gp120or gp41) and measuring the mean fluorescenceintensity (MFI) of antigen binding (Fig. 2a and b).Antibodies b6 and 4E10 bound antigen with 12-foldand 4-fold higher affinity, respectively, in the scFabformat. Antibody b12 bound to monomeric gp120with approximately 15 nM affinity in both formats,although the MFI plateau was higher in the scFvformat. This is likely a result of increased levels ofb12 surface expression in the scFv format. All threescFab clones were expressed on 50–80% of yeastcells (data not shown), which is comparable to pub-lished observations regarding scFv and Fab expres-sion of single clones.Because it has been shown previously that soluble

scFabs have a tendency to dimerize and multime-rize,15 we sought to determine whether the solublescFabswould exhibit enhanced levels of binding andneutralization due to avidity effects. b12, b6, and4E10 were cloned into the vector pComb3X andexpressed in E. coli in order to obtain soluble scFabs.Crude scFab supernatants were assayed for bindingand neutralization activity and compared to thecorresponding crude Fab fragments (Fig. 3a). Allthree scFabs exhibited similar binding profiles as thecorresponding Fab fragments in an enzyme-linkedimmunosorbent assay (ELISA), although we cannotexclude the possibility that multimerizationincreases the apparent affinity of soluble scFabs. Itshould also be noted that b12 and 4E10 scFabssaturate at a higher signal than their Fab counterpart,which also may be due to scFab multimerization.The three scFabs exhibited similar neutralizationprofiles as the corresponding Fab fragments(Fig. 3b), indicating that soluble scFabs can reliablybe used for this functional assay. It is possible that theslight decrease in neutralization for the scFab couldbe caused by multimerization, but this decrease isnot statistically significant and does not interferewith the general trend of neutralizing versus non-neutralizing antibodies.

Construction and selection of scFv and scFabimmune HIV-1 libraries

ScFv and scFab immune libraries were clonedfrom the bone marrow RNA of an HIV-1-infecteddonor who exhibits broad serum neutralization.Both the scFv and the scFab libraries were cloned

Fig. 3. Binding and neutralization profiles of soluble scFabs. b12, b6, and 4E10 were expressed solubly in E. coli, andcrude supernatants were assayed for binding and neutralization activity. (a) (A)–(C) represent ELISA binding curves, and(D)–(F) represent neutralization activity of soluble Fabs and scFabs. (b) EC50 and IC50 values derived from the data shownin (a). Data are mean±SD from two experiments.


from the same pool of cDNA, and the number ofindependent transformants after electroporationinto E. coli was on the order of 5×109 for bothlibraries. The scFab and scFv library DNA werepropagated in E. coli, linearized with NotI, and usedto transform EBY100 yeast cells using homologousrecombination and the lithium acetate method.20

Dilution plates indicated that the size of both libra-ries in yeast were ∼109. Based on restriction enzymedigestion and sequencing of unselected library

clones, 90% of clones from both libraries containedfull-length and in-frame inserts.The scFab and scFv libraries were both subject to

one round of magnetic bead selection21 followedby two to three rounds of sorting by flow cytome-try against monomeric gp120JR-FL and gp120JR-CSF(Fig. 4). No-antigen controls were used to determinesort gates that would only allow for selection ofantigen binding clones. Pooled IgG derived fromHIV-1-infected donors (HIVIG) was biotinylated

Fig. 4. scFv and scFab library sorting by flow cytometry. Yeast cells are double labeled so that antibody expression and antigen binding can be monitored simultaneously.(a)–(c) represent scFv library sorts (rounds 1–3), and (d) and (e) represent scFab library sorts (rounds 1 and 2). The circled populations represent double-positive yeast cellsthat were sorted, grown, and induced for the next round of selection. The percentages of antigen-positive cells are indicated in the top right corner of each plot. (f) shows thepercentage of antigen-positive induced cells for each round of selection.

369scF

abYeast

Display

Antibody

Library


and used for gp120 visualization (Fig. 1b). A poly-clonal secondary reagent was chosen because it isunlikely to compete with library clones for gp120binding. Using an anti-c-myc antibody and goat anti-mouse Alexa 647 dye, we monitored antibodyexpression using a C-terminal c-myc epitope tag(Fig. 1b). After the third round of sorting, 70 clonesfrom each library were selected for sequenceanalysis and initial antibody characterization.

Comparison of selected scFab and scFv clones

Selected scFab and scFv clones were groupedbased on examination of the amino acid sequencesof the third complementary determining region ofthe heavy chain (CDRH3) (Table 1). Clones thatappeared to represent somatic variants of the sameB-cell were grouped together. Based on this analysis,13 unique clones were selected from the scFablibrary and 11 unique clones were selected from thescFv library, with only 1 clone represented in bothlibraries. Of the selected scFv and scFab clones, 80%and 100% contained full-length inserts, respectively.Because the constant regions of the heavy and lightchains were used as stuffer sequences, all of thetruncated scFvs contained one full-length chain andonly the constant region of the other chain. Selectedclones from both libraries exhibited a bias for theVH1–69 family, but it has been previously demon-strated that anti-HIV-1 antibodies commonly usethis germ line gene family.23,24

Table 1. Sequence analysis of selected scFv and scFab clones

scFab clone No.a Germ line IGVLb Germ line IGVH

1B3 13 L1–51⁎02 3–23⁎01 E3C1 4 L2–14⁎01 1–69⁎011A1 2 L2–23⁎03 3–43⁎011A4 4 K3–20⁎01 1–69⁎061A3 6 L2–11⁎01 1–69⁎06J3 20 K1–39⁎01 1–69⁎12Q10 1 L1–44⁎01 1–2⁎01 AQ3 7 L3–19⁎01 1–24⁎01Q7 4 L2–8⁎01 3–33⁎03J5 1 K3–11⁎01 3–21⁎013J17 4 K3–20⁎01 3–30⁎01Q16 1 L3–21⁎02 5–51⁎01 Q2Q10 1 L1–51⁎01 1–3⁎01 G

scFv clone No.a Germ line IGVLb Germ line IGVH

1R2-17 15 L1–44⁎01 5–51⁎01 A1R2–48 6 L2–23⁎02 1–3⁎011R2–20 1 L2–14⁎01 1–2⁎021R2–18 6 L1–47⁎01 3–64⁎02 A2R2–12 1 L1–40⁎01 1–69⁎01 Q2R2–14 5 K3–20⁎01 1–2⁎022R2–21 1 L1–44⁎01 3–7⁎01 A2R2–25 1 K3D–15⁎01 1–69⁎082R2–50 1 L3–19⁎01 1–69⁎01 Y1R2–46 5 L3–21⁎03 3–30⁎04 QT13 4 K3–20⁎01 3–43⁎01

a Number of times an individual sequence was observed from seleb Germ line gene sequences were determined using the IMGT dat

kappa, respectively.c Clones were grouped based on analysis of the CDRH3 region.

The affinities of single clones were calculated byincubating each yeast-displayed scFv or scFab withvarying amounts of monomeric gp120 (Table 2).The binding constants for the scFab clones rangedfrom 0.5 to 36 nM, with an average affinity of10 nM. In contrast, the affinities of selected scFvclones ranged from 0.3 to 100 nM, with an averageaffinity of 43 nM.Three antibodies from each library were cloned in

the alternate display format and assayed for surfaceexpression and antigen binding in order to deter-mine whether the display format had an intrinsiceffect on the types of clones selected. We found thatthe three selected scFab clones exhibited high levelsof surface expression as scFvs but bound antigenwith lower affinity. Similarly, the three selectedscFvs could be displayed on the surface of yeast inthe scFab format and bound antigen with 2- to 4-foldhigher affinity as scFabs (Table 2). Based on thisanalysis, the difference in selection of clones is pro-bably not a direct result of the display format, asdiscussed below.

Discussion

Here, we have constructed the first scFab yeastdisplay library and shown that this format offersdistinct advantages over the existing scFv and Fabformats.We found that the majority of antibodies wetested bound with higher affinity as scFabs than as

CDRL3 CDRH3c

TWDSRLSAGVF GPLGRFLDFWSSYTTRGTRIF EGGRSYGTQVLDPWCAYASKSLSFVF ERHIAGNGGIDYWQQYVSSPYTF SRIVPVAAPQFTYWCSYAGSRILWIF GSDSRSYYHYMDVWQQSYSTPLTF SRTTIFGVAQDNWFDPW

AWDGNLDGVVF DPRPTIAVLPPGMSVWFDPWNSRDITYNSVVF AGFDGLKGYYKGFDYWSSYGGSNNLIF DPFFKNSPTYYLDFYMDVWQQRSNWPITF SSGWFFVAKYYFDSWQQYSTSITF DAWGDNYFDYW

VWDSTFDPWVF MGPYNDLWSGLPRGGVDPWTWDSSLGAWVF YGDYGRAFAIW

CDRL3 CDRH3

AWDDTLHAYVF PVDYFESSGYYWFDFWCSDAGIWVF IGRYFDLWFSFRSSSIWVF VSYYDSSGADWEWDDLLNGWVF DDWVGSNHTFSFDVWSYDHRVNGYVF ATWNDDCGTTNCYDWFDPWQQYDSSGSF ERRGGFDEYW

AWDDTLDAYVF RPCDGAKCDDPPWQQYYSSPLTF GDSGSSSPFDYIYMDVW

SRDNSGSHNYVF AVLPFLEWFAAHQYFYMDVWVWDGSHDPREVF DAWGDNYFDYWQQYGSSEGSF DGASIGKYYDFWSGSSRRGKYYMDVW

ction. Somatic variants (not shown) were grouped together.abase (http://imgt.cines.fr).22 “L” and “K” refer to lambda and

Table2.Binding affinities of selected scFv and scFab clones

Clone Kd scFab (nM)a Kd scFv (nM)

1B3 1 343C1 6 —1A1 14 —1A4 3 71A3 8.5 —J3 2 —Q10 0.5 —Q3 17 —Q7 36 103J5 22 —3J17 21 —Q16 1 —2Q10 6 —1R2–17 — 611R2–48 — 9.51R2–20 — 461R2–18 — 52R2–12 10 422R2–14 — 0.32R2–21 30 682R2–25 — 522R2–50 16 601R2–46 — 29T13 — 100

a Kd values were calculated by incubating single clones withvarying amounts of antigen and measuring the MFI of antigenbinding.


scFvs. This was demonstrated for well-characterizedanti-HIV-1 mAbs as well as for selected clones froman immune HIV-1 library. It has been shown that thefolding and stability of the Fab fragment aredependent on cooperation between the VH/VL andCH1/CL interfaces25 and stabilization of thesedomain interactions improves antigen binding.26

The lack of a stabilizing CH1/CL interface in scFvfragments likely increases the flexibility of thebinding pocket, which results in an increased entro-pic cost to binding.27

Surprisingly, when the same cDNA was used toconstruct an immune library in both the scFab andscFv format and selected against the same antigens,two almost completely different sets of clones wereselected. We reasoned that this observation might bea result of the display format and that certain clonesonly display or bind antigen in one of the formatstested. Six antibodies were cloned in the alternatelibrary format, displayed on the surface of yeast, andassayed for antigen binding in order to evaluate thispossibility. We found that all of the antibodies testedcould be displayed on the surface of yeast and bindantigen in both formats. Consistent with ourprevious results, all six antibodies bound antigenwith higher affinity in the scFab format. Because sortgates were drawn to select all c-myc-positive antigenbinding clones, neither differences in antibody dis-play nor those in antigen binding appear to accountfor differences in selected antibodies. Remainingexplanations for the differences include samplingsize and library construction. An explanation basedon sampling size would require that the libraries

contain an unusually large number of gp120-specificclones for only one clone to be shared among 70from each library. We therefore favor an explanationbased on differences in library construction (PCRpriming, chain shuffling, internal restriction sites,etc.).One observation worth noting concerns the nature

of the antibodies selected from each library. Wefound that the scFab library yielded a much higherproportion of antigen-specific clones over back-ground binders than the scFv library. About 20%of selected scFv clones contained the vector stuffersequence, and 18% of the full-length clones boundnonspecifically to fluorescence reagents. In contrast,all of the selected scFab clones were full length andonly 4% bound to secondary reagents. Because bothof the unselected libraries contained 90% full-lengthclones with similar germ line gene representation,this discrepancy is probably due to the unstablenature of scFvs and not to differences in libraryconstruction.ScFab yeast display libraries also provide advan-

tages over current formats for display of yeast Fablibraries. The most commonly used method requiresyeast mating and a two-vector system. This systemallows for the display of heavy-chain-only clones,which could potentially interfere with library selec-tion. It has previously been reported that 15% ofantigen-specific yeast express only a heavy chain.7

Therefore, although yeast mating allows for theconstruction of large libraries, heavy-chain-onlydisplay results in some loss of selection efficiency.In contrast, we found that our scFab library could beconstructed and selected with very high efficiency.Using our homologous recombination system, wewere easily able to generate a yeast display librarywith ∼109 members, which is comparable to the sizeof existing Fab libraries. Also, the scFab linkerprevents the display of unpaired heavy or lightchains. When expressed in E. coli, we found that thesoluble scFabs exhibited similar binding and neu-tralization profiles as the corresponding Fab frag-ments, although one caveat in the binding andneutralization assays is the possibility of solublescFab multimerization. This result suggests that thescFab selection format can also be used for subse-quent antibody characterization. Based on theseobservations, yeast display of scFab libraries is likelya more robust system than Fab display.In conclusion, we have generated a new yeast

display format that offers clear advantages over theexisting platforms. Yeast-displayed scFab librariesmay allow for the increased recovery of novel anti-bodies with diagnostic or therapeutic applications.

Materials and Methods

Yeast cell lines and media

Yeast strain EBY100 (GAL1-AGA1∷URA3 ura3–52 trp1leu21 his3200 pep4∷HIS2 prb11.6R can1 GAL) wasmaintained in YPD broth (Difco). Transformation of


EBY100 with the vector pYDscFab and pYDscFV wascompleted using the lithium acetate method20 andmaintained in SD-CAA medium, pH 4.5 (6.7 g/L yeastnitrogen base, 5 g/L casamino acids, 20 g/L dextrose,4.29 g/l citric acid, and 14.7 g/L citric acid monohydrate),and on SD-CAA plates (SD-CAA+17 g/L agar). Yeastsurface expression of scFv was induced by transferring toSG/R-CAA medium (6.7 g/L yeast nitrogen base, 5 g/Lcasamino acids, 20 g/L each galactose and raffinose, 1 g/Ldextrose, 9.67 g/L NaH2PO4·

2H2O and 10.19 g/L Na2-HPO4·

7H2O). After yeast sorting, 0.25 μg/mL ketocona-zole (Sigma-Aldrich) was added to the SD-CAA media inorder to inhibit growth of Candida parapsilosis.28

Antigens and antibodies

Monomeric gp120JR-FL and soluble CD4were purchasedfrom Progenics (Tarrytown, NY), and gp120JR-CSF wasprocured under contract from Advanced Products Enter-prises (MD). The human anti-gp120 mAbs used in thisstudy are IgGs b12,29 b6,29 and 4E10.30 Polyclonal humanIgG against HIV-1 (HIVIG) was obtained from the AIDSResearch Reference Reagent Program, provided by JohnMascola. Antibodies were biotinylated using the EZ-LinkSulfo-NHS-Biotinylation Kit fromPierce.MousemAb anti-c-myc (9E10) was obtained from AbD Serotec. Fluorescentreagents goat anti-mouse-Alexa 647 and streptavidin–phycoerythrin (SA–PE) were obtained from MolecularProbes.

Yeast display vector construction

The scFab yeast display vector was generated from thevector pPNL200 (M. Feldhaus, Pacific Northwest NationalLaboratory). A BssHII cloning site was inserted at the 5′end of the light chain and a NheI site was inserted at the 3′end of the heavy chain using the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene). Overlap PCR wasused to insert a 34-amino-acid linker between the heavyand light chains, add anNcoI site at the 5′ end of the heavychain and a SalI site at the 3′ end of the light chain, andremove the two cysteines involved in the disulfide bondconnecting the two chains. Two stuffer sequences, com-posed of either the constant heavy genes or the constantlight genes, were inserted on both sides of the linker. Ahomologous recombination linker containing the phos-phorylated sequence 5′-CTAGTCAGGAACTGACA-ACTATATGCGAGCAAATCCCCTCACCAACTCGC-3′was ligated in after the heavy chain using the restrictionenzyme sites NotI and SpeI. The sequence of the modifiedvector, pYDscFab, was verified by DNA sequencing.The yeast display scFv vector, pYDscFv, was genera-

ted by modification of pYDscFab. The 34-amino-acidscFab linker was digested out of pYDscFab using therestriction enzymes NcoI and NheI and the digestedvector was gel purified using the QIAquick Gel Purifica-tion Kit (Qiagen). A 15-amino-acid 5′ phosphorylatedoligo containing the scFv linker sequence 5′-TCGA-CAGGGGAGAATTCGGTGGTGGTGGTTCTGGTGGTG-GTGGTTCTGGTGGTGGTGGTTCTCTC-3′ was ligatedinto the digested vector. The sequence of pYDscFv wasverified by DNA sequencing.

scFab and scFv library construction

Total RNA was isolated from the bone marrow mono-nuclear cell RNA from an HIV-infected donor who

exhibits broad serum neutralization of HIV-1. Bonemarrow was collected, homogenized, and stored in TRIreagent (received from the International AIDS VaccineInitiative) and placed at −80 °C for long-term storage.Homogenized bone marrow (5 mL) was centrifuged at5000g for 10 min at 4 °C, supernatant was removed, andthe pellet was resuspended in a 10× volume of TRI reagentto isolate total RNA. 1-Bromo-3-chloropropane (3 mL;Sigma) was added to each supernatant, vortexed for 15 s,and incubated for 15 min at room temperature. Aftercentrifugation at 17,000g for 15 min at 4 °C, the upperphase was transferred to a fresh, 50-mL tube. Isopropanol(15 mL) was added to the tube, vortexed for 15 s, andincubated for 10 min at room temperature. Centrifugationwas again performed at 17,000g at 4 °C and the super-natant was discarded. The pellet was washed with 30 mLof 75% ethanol, air-dried at room temperature for 5–10 min, and resuspended in 500 μL of RNase-free water.Total RNA was reverse transcribed into cDNA usingtranscriptor first-strand cDNA synthesis kit (Roche) andan oligo(dT) primer. Heavy- and light-chain antibodygenes were PCR amplified from cDNA using a primer setbased on that of Sblattero and Bradbury.31 The followingoligonucleotides were used as heavy- and light-chainscFab reverse primers to remove the cysteines that formthe interchain disulfide bond: scFabHrev: 5′-GACTTGGC-TAGCAGATTTGGGCTCAACTTTC-3′, scFablamdarev:5′-GCATGT GTCGACTTCYGYAGGGGCMACTGTC-3′,and scFabKaprev: 5′-GCATGTGTCGACCTCTCCCCT-GTTGAAGCTC-3′. Kappa and lambda light-chain geneswere gel purified with the QIAquick Gel Extraction Kit(Qiagen), pooled in an equal ratio, digested with therestriction enzymes BssHII and SalI, and ligated into thevector pYDscFab or pYDscFv. After ethanol precipitation,the ligation mixture was used to transform E. coli strainElectro-Ten Blue (Strategene). pYDscFab plasmid DNAcontaining the light-chain genes was isolated using aMaxiprep kit (Qiagen) and digested with the restrictionenzymes NheI and NcoI for ligation of the heavy-chaingenes (performed the same as light-chain ligation). Thefinal libraries contained approximately 5×109 indepen-dent transformants. The libraries were then linearizedwith NotI, ethanol precipitated, and used to transform theyeast strain EBY100 using the high-efficiency lithiumacetate method,20 and grown in SD-CAA until saturation(about 24 h). The size of both the scFab and the scFvlibraries were estimated at 1×109 members.

Library selection, flow cytometric analysis, and singleclone characterization

Libraries were grown as described previously.2 Briefly,yeast were grown in SD-CAA media for 12–18 h at30 °C, pelleted at 3000g for 10 min, and resuspended inSG/R-CAA for 16–18 h at 20 °C in volumes appropriatefor the size of the library. For the first round of selection,a magnetic bead sorting technique utilizing the MiltenyiMacs system was used as described previously.32 Yeastcells (1×1010) were incubated with 10 μg/mL gp120 for1 h at room temperature in FACS wash buffer[phosphate-buffered saline (PBS)/0.5% bovine serumalbumin (BSA)/2 mM ethylenediaminetetraacetic acid]followed by 10 min on ice and then washed 3 times withwash buffer. Yeast cells were pelleted and incubatedwith 15 μg/mL biotinylated HIVIG for 30 min at 4 °C.After washing 3 times with ice-cold wash buffer, the cellpellet was resuspended in 4 mL wash buffer. SA micro-beads (200 μL) were added to the yeast and incubated


for 10 min, followed by two 50-mLwashes. The yeast wereresuspended in 50 mL and loaded onto a Miltenyi LScolumn at 4 °C in 7-mL aliquots. After each 7 mL, thecolumn was removed from the magnetic field and imme-diately replaced. The column was then washed with 1 mLof wash buffer before another 7 mL was loaded onto thecolumn. After the entire 50 mL was loaded, the columnwas washed 3 times with 2 mL of wash buffer withremoval and replacement in between. The column wasthen removed from the magnetic field, and the yeast wereeluted with 12 mL of SD-CAA media and grown over-night. The yeast recovered from the magnetic bead sortwere induced in 100 mL SG/R-CAA for 20 h at 20 °C. Thefollowing two to three rounds of sorting were done usingflow cytometry. Approximately 1×108 yeast were pelleted,washed once with wash buffer, and resuspended in 1 mLwash buffer with 10 μg/mL gp120 and a 1:200 dilution ofanti-c-myc antibody. After a 1-h incubation at roomtemperature, yeast were washed 3 times and thenresuspended in 1 mL wash buffer. Biotinylated HIVIG(15 μg/mL) was added to the yeast, incubated at 4 °C for30 min, and washed 3 times with wash buffer. Fivemicrograms permilliliter of both SA–PE andGaM-647wasadded to the cells and incubated for 30 min on ice in thedark, washed 3 times again, and then resuspended in 6mLwash buffer for flow cytometric sorting. Sorting wasperformed using a FACS ARIA sorter and sort gates weredetermined to select only antigen binding clones. Collectedcells were grown overnight in SD-CAAmedia at 30 °C andinduced in SG/R-CAA for the next round of sorting. Afterthe final round, yeast were plated on SD-CAA plates andindividual colonies were picked for characterization.Plasmid DNAwas isolated from individual yeast clones

using the Zymoprep Yeast Plasmid Miniprep Kit (ZymoResearch). The recovered plasmids were electroporated inthe XL-1 Blue E. coli for amplification of plasmid DNA andminiprepped using the QIAprep Spin Miniprep Kit(Qiagen). After sequencing, clones were grouped basedon analysis of the CDRH3 region and a representativeclone was chosen from each group for subsequentcharacterization. To determine approximate binding con-stants, we incubated ∼106 induced yeast cells with adilution series of gp120 (0–200 nM) and anti-c-myc anti-body for 1 h at room temperature. Cells were then washed3 times and incubated with biotinylated HIVIG at 4 °C for30 min. After washing, yeast were incubated withfluorescent reagents for 30 min at 4 °C. Cells werepelleted, washed, and resuspended in approximately100 μL FACS wash buffer. A BD Bioscience FACSArrayplate reader was used for flow cytometric analysis, andFlowJo software was used for analysis. Quantitativeequilibrium binding constants (Kd values) were calculatedby plotting the c-myc-normalized MFI against antigenconcentration and fitting the curve using nonlinearregression analysis.Variable region genes were amplified by PCR, digested

with SalI and BssHII (light chains) or NheI and NcoI(heavy chains), gel purified with the Qiagen QIAquick GelPurification Kit, and ligated into a similarly digested andpurified pYDscFv in order to clone scFabs as scFvs. Theligation mixture was heat shocked into Top10 cells andindividual colonies were picked for plasmid isolation.Positive clones were verified by DNA sequencing.An scFab clone that contained ApaI and AvrII restric-

tion sites at the variable–constant region interface of theheavy and light chains, respectively, was used for cloningscFvs as scFabs. Variable region genes were amplified byPCR, digested with AvrII and BssHII (light chains) or ApaIand NheI (heavy chains), gel purified with the QIAquick

Gel Extraction Kit (Qiagen), and ligated into a similarlydigested vector. The ligation mixture was used to trans-form Top10 cells, and colonies were picked for plasmidisolation. Positive transformants were verified by DNAsequencing.

Preparation of soluble scFabs and Fabs

Individual scFab clones were digested out of pYDscFabwith the restriction enzymes BssHII and NheI, gelextracted and purified with the QIAquick gel extractionkit, and ligated into the similarly digested and gel purifiedpComb3X vector for soluble expression in E. coli. Theligation reaction was used to transform Top10 cells, andindividual colonies were picked for plasmid isolation.Positive inserts were verified by DNA sequencing. ScFabor Fab clones in pComb3X were then used to transformXL-1 Blue cells and grown at 37 °C in 10 mL Superbrothmedium (30 g/L bactotryptone, 20 g/L yeast extract, and10 g/L Mops) supplemented with 25 μg/mL carbenicillinand 10 μg/mL tetracycline until an OD600 (optical densityat 600 nm) of 0.6–0.8 was reached. IPTG (1 mM) was thenadded and cells were grown for an additional 16–18 h at30 °C. The cells were centrifuged at 3000 RPM for 10 min,and the pellet was resuspended in 0.5 mL 1× PBS. Afterfour rounds of freeze–thawing between 37 and −80 °C, thecells were centrifuged at full speed in a tabletop centri-fuge. Supernatants were collected and sterile-filtered foruse in ELISAs and neutralization assays.

ELISA binding assays

Ninety-six-well ELISA plates were coated overnight at4 °C with 50 μL PBS containing 50 ng of goat anti-humanIgG F(ab′)2 (Pierce) or 75 ng gp120JR-FL. The wells werewashed 4 times with PBS containing 0.05% Tween 20 andblocked with 3% BSA at room temperature for 1 h. Thewells were then washed once, and 50 μL of the bacterialsupernatants (2× diluted) containing Fab or scFab wasadded in a 2-fold dilution series. The plates were incu-bated at room temperature for 1 h and washed 4 times,and goat anti-human IgG F(ab′)2 conjugated to alkalinephosphatase (Pierce), diluted 1:1000 in PBS containing 1%BSA and 0.025% Tween 20, was added to the wells. Theplate was incubated at room temperature for 1 h andwashed 4 times, and the plate was developed by adding50 μL of alkaline phosphatase substrate (Sigma) to 5 mLalkaline phosphatase staining buffer (pH 9.8), according tothe manufacturer's instructions. The optical density at405 nm was read on a microplate reader (MolecularDevices). The concentration of Fab was determined withan anti-Fab ELISA (2-fold dilution series) using simplenonlinear regression. Half maximal effective concentra-tions (EC50 values) were determined using fixed concen-trations of monomeric gp120 or gp41 (20 nM) and titratingthe amount of crude Fab or scFab. In general, 12 differentscFab or Fab concentrations were used, which covered arange of 0.002–2.0 μg/mL. The concentration of Fab orscFab corresponding to half maximal signal (EC50) wasdetermined by interpolation of the resulting binding curveusing nonlinear regression.

HIV-1 neutralization assays

HIV-1 pseudovirus, competent for a single round ofreplication, was generated by cotransfecting 293T cellswith the luciferase reporter plasmid pNL4-3.Luc.R-E,33


an Env-complementation vector, and the polyethyle-neimine transfection reagent34 as described previously.35

Neutralization was measured using U87.CD4.CCR5target cells. Antibodies were diluted in Dulbecco'smodified Eagle's medium (Gibco) containing 10% fetalbovine serum and mixed 1:1 with pseudovirus (100 μLtotal volume). The antibody–virus mixture was incubatedfor 1 h at 37 °C and then transferred (1:1 by volume) tothe target cells. After a 72-h incubation at 37 °C, the cellswere lysed, luciferase reagent (Promega) was added, andthe luminescence in relative light units was measuredusing an Orion microplate luminometer (Berthold Detec-tion Systems). Virus neutralization was measured as areduction in viral infectivity compared to an antibody-free control. Neutralization curves were generated byplotting the percentage of virus neutralization againstvarious scFab or Fab concentrations. The concentration ofFab or scFab, which yielded a 50% reduction in viralinfectivity (IC50), was determined by interpolation of theresulting neutralization curve using nonlinear regression.

Acknowledgements

We thank Mansun Law, Lars Hangartner, andMichael Huber for helpful discussion; CodyWilliams and Christina Corbaci for artwork inFig. 1; and the members of The Scripps ResearchInstitute Flow Cytometry Facility for assistance.This work was supported by National Institutes ofHealth Grant RO1 AI33292 and the InternationalAIDS Vaccine Initiative Neutralizing AntibodyConsortium.

References

1. Boder, E. T. & Wittrup, K. D. (1997). Yeast surfacedisplay for screening combinatorial polypeptidelibraries. Nat. Biotechnol. 15, 553–557.

2. Feldhaus, M. J., Siegel, R. W., Opresko, L. K.,Coleman, J. R., Feldhaus, J. M., Yeung, Y. A. et al.(2003). Flow-cytometric isolation of human antibo-dies from a nonimmune Saccharomyces cerevisiaesurface display library. Nat. Biotechnol. 21, 163–170.

3. Bowley, D. R., Labrijn, A. F., Zwick, M. B. & Burton,D. R. (2007). Antigen selection from an HIV-1immune antibody library displayed on yeast yieldsmany novel antibodies compared to selection fromthe same library displayed on phage. Protein Eng.Des. Sel. 20, 81–90.

4. Lee, H. W., Lee, S. H., Park, K. J., Kim, J. S., Kwon,M. H. & Kim, Y. S. (2006). Construction and cha-racterization of a pseudo-immune human antibodylibrary using yeast surface display. Biochem. Biophys.Res. Commun. 346, 896–903.

5. Wang, X. X., Cho, Y. K. & Shusta, E. V. (2007).Mining a yeast library for brain endothelial cell-binding antibodies. Nat. Methods, 4, 143–145.

6. van den Beucken, T., Pieters, H., Steukers, M., vander Vaart, M., Ladner, R. C., Hoogenboom, H. R. &Hufton, S. E. (2003). Affinity maturation of Fabantibody fragments by fluorescent-activated cellsorting of yeast-displayed libraries. FEBS Lett. 546,288–294.

7. Weaver-Feldhaus, J. M., Lou, J., Coleman, J. R., Siegel,R. W., Marks, J. D. & Feldhaus, M. J. (2004). Yeastmating for combinatorial Fab library generation andsurface display. FEBS Lett. 564, 24–34.

8. Blaise, L., Wehnert, A., Steukers, M. P., van denBeucken, T., Hoogenboom, H. R. & Hufton, S. E.(2004). Construction and diversification of yeast cellsurface displayed libraries by yeast mating: applica-tion to the affinity maturation of Fab antibodyfragments. Gene, 342, 211–218.

9. Rippmann, J. F., Klein, M., Hoischen, C., Brocks, B.,Rettig, W. J., Gumpert, J. et al. (1998). Procaryoticexpression of single-chain variable-fragment (scFv)antibodies: secretion in L-form cells of Proteus mirabilisleads to active product and overcomes the limitationsof periplasmic expression in Escherichia coli. Appl.Environ. Microbiol. 64, 4862–4869.

10. Marks, J. D., Hoogenboom, H. R., Griffiths, A. D. &Winter, G. (1992). Molecular evolution of proteins onfilamentous phage. Mimicking the strategy of theimmune system. J. Biol. Chem. 267, 16007–16010.

11. Bird, R. E. & Walker, B. W. (1991). Single chain anti-body variable regions. Trends Biotechnol. 9, 132–137.

12. Labrijn, A. F., Poignard, P., Raja, A., Zwick, M. B.,Delgado, K., Franti, M. et al. (2003). Access of antibodymolecules to the conserved coreceptor binding site onglycoprotein gp120 is sterically restricted on primaryhuman immunodeficiency virus type 1. J. Virol. 77,10557–10565.

13. Quintero-Hernandez, V., Juarez-Gonzalez, V. R.,Ortiz-Leon, M., Sanchez, R., Possani, L. D. &Becerril, B. (2007). The change of the scFv into theFab format improves the stability and in vivo toxinneutralization capacity of recombinant antibodies.Mol. Immunol. 44, 1307–1315.

14. Skerra, A., Pfitzinger, I. & Plückthun, A. (1991). Thefunctional expression of antibody Fv fragments inEscherichia coli: improved vectors and a generallyapplicable purification technique. Biotechnology (N.Y.),9, 273–278.

15. Hust, M., Jostock, T., Menzel, C., Voedisch, B., Mohr,A., Brenneis, M. et al. (2007). Single chain Fab (scFab)fragment. BMC Biotechnol. 7, 14.

16. Jordan, E., Al-Halabi, L., Schirrmann, T., Hust, M. &Dübel, S. (2007). Production of single chain Fab (scFab)fragments in Bacillus megaterium.Microb. Cell Fact. 6, 38.

17. Raymond, C. K., Mugford, V. R. & Sexson, S. L. (1999).Plasmid topologies that enhance the transformationefficiency of yeast. BioTechniques, 27, 892–894, 896.

18. Swers, J. S., Kellogg, B. A. & Wittrup, K. D.(2004). Shuffled antibody libraries created by invivo homologous recombination and yeast surfacedisplay. Nucleic Acids Res. 32, e36.

19. Binley, J. M., Wrin, T., Korber, B., Zwick, M. B.,Wang, M., Chappey, C. et al. (2004). Comprehensivecross-clade neutralization analysis of a panel of anti-human immunodeficiency virus type 1 monoclonalantibodies. J. Virol. 78, 13232–13252.

20. Gietz, R. D. & Schiestl, R. H. (2007). Large-scale high-efficiency yeast transformation using the LiAc/SScarrier DNA/PEG method. Nat. Protoc. 2, 38–41.

21. Yeung, Y. A. & Wittrup, K. D. (2002). Quantitativescreening of yeast surface-displayed polypeptidelibraries by magnetic bead capture. Biotechnol. Prog.18, 212–220.

22. Brochet, X., Lefranc, M. P. & Giudicelli, V. (2008).IMGT/V-QUEST: the highly customized and integratedsystem for IG and TR standardized V–J and V–D–Jsequence analysis. Nucleic Acids Res. 36, W503–W508.


23. Gorny, M. K., Wang, X. H., Williams, C., Volsky, B.,Revesz, K., Witover, B. et al. (2009). Preferential use ofthe VH5–51 gene segment by the human immuneresponse to code for antibodies against the V3 domainof HIV-1. Mol. Immunol. 46, 917–926.

24. Choe, H., Li, W., Wright, P. L., Vasilieva, N., Venturi,M., Huang, C. C. et al. (2003). Tyrosine sulfation ofhuman antibodies contributes to recognition of theCCR5 binding region of HIV-1 gp120. Cell, 114,161–170.

25. Rothlisberger, D., Honegger, A. & Plückthun, A.(2005). Domain interactions in the Fab fragment: acomparative evaluation of the single-chain Fv and Fabformat engineered with variable domains of differentstability. J. Mol. Biol. 347, 773–789.

26. Acierno, J. P., Braden, B. C., Klinke, S., Goldbaum,F. A. & Cauerhff, A. (2007). Affinity maturationincreases the stability and plasticity of the Fvdomain of anti-protein antibodies. J. Mol. Biol. 374,130–146.

27. Thorpe, I. F. & Brooks, C. L., III (2007). Molecularevolution of affinity and flexibility in the immunesystem. Proc. Natl Acad. Sci. USA, 104, 8821–8826.

28. Scholler, N., Garvik, B., Quarles, T., Jiang, S. &Urban, N. (2006). Method for generation of invivo biotinylated recombinant antibodies by yeastmating. J. Immunol. Methods, 317, 132–143.

29. Burton, D. R. & Barbas, C. F., III (1994). Humanantibodies from combinatorial libraries. Adv. Immunol.57, 191–280.

30. Buchacher, A., Predl, R., Strutzenberger, K.,Steinfellner, W., Trkola, A., Purtscher, M. et al.(1994). Generation of human monoclonal antibo-dies against HIV-1 proteins; electrofusion andEpstein–Barr virus transformation for peripheralblood lymphocyte immortalization. AIDS Res.Hum. Retroviruses, 10, 359–369.

31. Sblattero, D. & Bradbury, A. (1998). A definitive set ofoligonucleotide primers for amplifying human Vregions. Immunotechnology, 3, 271–278.

32. Feldhaus, M. J. & Siegel, R. W. (2004). Yeast display ofantibody fragments: a discovery and characterizationplatform. J. Immunol. Methods, 290, 69–80.

33. Connor, R. I., Chen, B. K., Choe, S. & Landau, N. R.(1995). Vpr is required for efficient replication ofhuman immunodeficiency virus type-1 in mono-nuclear phagocytes. Virology, 206, 935–944.

34. Kirschner, M., Monrose, V., Paluch, M., Techodam-rongsin, N., Rethwilm, A. & Moore, J. P. (2006). Theproduction of cleaved, trimeric human immunodefi-ciency virus type 1 (HIV-1) envelope glycoproteinvaccine antigens and infectious pseudoviruses usinglinear polyethylenimine as a transfection reagent.Protein Expr. Purif. 48, 61–68.

35. Nelson, J. D., Kinkead, H., Brunel, F. M., Leaman,D., Jensen, R., Louis, J. M. et al. (2008). Antibodyelicited against the gp41 N-heptad repeat (NHR)coiled-coil can neutralize HIV-1 with modestpotency but non-neutralizing antibodies also bindto NHR mimetics. Virology, 377, 170–183.

efficient recovery of high-affinity antibodies from a single-chain fab yeast display library

Documents