17
Human Immunodeficiency Virus Type 1-neutralizing Monoclonal Antibody 2F5 is Multispecific for Sequences Flanking the DKW Core Epitope Alfredo Menendez, Keith C. Chow, Oscar C. C. Pan and Jamie K. Scott * Department of Molecular Biology and Biochemistry Simon Fraser University, 8888 University Drive, Burnaby, BC Canada V5A 1S6 Human monoclonal antibody 2F5 is one of a few human antibodies that neutralize a broad range of HIV-1 primary isolates. The 2F5 epitope on gp41 includes the sequence ELDKWA, with the core residues, DKW, being critical for antibody binding. HIV-neutralizing antibodies have never been elicited by immunization with peptides bearing ELDKWA, suggesting that important part(s) of the 2F5 paratope remain unidentified. The use of longer peptides extending beyond ELDKWA has resulted in increased epitope anti- genicity, but neutralizing antibodies have not been generated. We sought to develop peptides that bind to 2F5, and that function as specific probes of the 2F5 paratope. Thus, we used 2F5 to screen a set of phage-displayed, random peptide libraries. Tight-binding clones from the random peptide libraries displayed sequence variability in the regions flanking the DKW motif. To further reveal flanking regions involved in 2F5 binding, two semi-defined libraries were constructed having 12 variegated residues either N-terminal or C-terminal to the DKW core (X 12 -AADKW and AADKW-X 12 , respectively). Three clones isolated from the AADKW-X 12 library had similar high affinities, despite a lack of sequence homology among them, or with gp41. The contribution of each residue of these clones to 2F5 binding was evaluated by Ala substitution and amino acid deletion studies, and revealed that each clone bound 2F5 by a different mechanism. These results suggest that the 2F5 paratope is formed by at least two functionally distinct regions: one that displays specificity for the DKW core epitope, and another that is multispecific for sequences C-terminal to the core epitope. The implications of this second, multispecific region of the 2F5 paratope for its unique biological function are discussed. q 2004 Elsevier Ltd. All rights reserved. Keywords: mAb 2F5; HIV-1; phage-displayed peptide library; epitope; neutralizing antibody *Corresponding author Introduction Human monoclonal antibody (mAb) 2F5 is one of the few human immunodeficiency type 1 (HIV-1) broadly neutralizing antibodies identified so far. 1,2 Passive transfer of 2F5, in combination with other neutralizing human mAbs, has conferred sterilizing protection to viral challenge in several animal models. 3–5 mAb 2F5 binds with high affinity to a con- served site on the extracellular region of the trans- membrane protein gp41 (residues 662–667, amino acid sequence ELDKWA). 6 Results from screening of phage-displayed peptide libraries and from 2F5 IgG binding studies to substituted synthetic peptides 0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. E-mail address of the corresponding author: [email protected] Abbreviations used: mAb, monoclonal antibody; HIV- 1, human immunodeficiency virus type 1; HRP, horseradish peroxidase; NZY/Tet/IPTG, NZY medium supplemented with 15 mg/ml of tetracycline and 1 mM isopropyl b-D-thiogalactopyranoside; MBP, E. coli maltose-binding protein; ABTS, 2,2 0 -azino-bis(3- ethylbenzthiazoline-6-sulfonic acid); TBS, Tris-buffered saline, pH 7.5; BSA, bovine serum albumin, fraction V; PEG/NaCl, 16.7% polyethylene glycol 8000, 3.3 M sodium chloride; K d , dissociation constant at equilibrium; (2 3), amino acid position three residues C- terminal to the Trp residue in the DKW core sequence; (þ 3), amino acid position three residues N-terminal to the Asp residue in the DKW core sequence. doi:10.1016/j.jmb.2004.02.051 J. Mol. Biol. (2004) 338, 311–327

Human Immunodeficiency Virus Type 1-neutralizing Monoclonal Antibody 2F5 is Multispecific for Sequences Flanking the DKW Core Epitope

Embed Size (px)

Citation preview

Human Immunodeficiency Virus Type 1-neutralizingMonoclonal Antibody 2F5 is Multispecific forSequences Flanking the DKW Core Epitope

Alfredo Menendez, Keith C. Chow, Oscar C. C. Pan andJamie K. Scott*

Department of MolecularBiology and BiochemistrySimon Fraser University, 8888University Drive, Burnaby, BCCanada V5A 1S6

Human monoclonal antibody 2F5 is one of a few human antibodies thatneutralize a broad range of HIV-1 primary isolates. The 2F5 epitope ongp41 includes the sequence ELDKWA, with the core residues, DKW, beingcritical for antibody binding. HIV-neutralizing antibodies have never beenelicited by immunization with peptides bearing ELDKWA, suggesting thatimportant part(s) of the 2F5 paratope remain unidentified. The use of longerpeptides extending beyond ELDKWA has resulted in increased epitope anti-genicity, but neutralizing antibodies have not been generated.

We sought to develop peptides that bind to 2F5, and that function asspecific probes of the 2F5 paratope. Thus, we used 2F5 to screen a set ofphage-displayed, random peptide libraries. Tight-binding clones fromthe random peptide libraries displayed sequence variability in the regionsflanking the DKW motif. To further reveal flanking regions involved in2F5 binding, two semi-defined libraries were constructed having 12variegated residues either N-terminal or C-terminal to the DKW core(X12-AADKW and AADKW-X12, respectively). Three clones isolated fromthe AADKW-X12 library had similar high affinities, despite a lack ofsequence homology among them, or with gp41. The contribution of eachresidue of these clones to 2F5 binding was evaluated by Ala substitutionand amino acid deletion studies, and revealed that each clone bound 2F5by a different mechanism. These results suggest that the 2F5 paratope isformed by at least two functionally distinct regions: one that displaysspecificity for the DKW core epitope, and another that is multispecific forsequences C-terminal to the core epitope. The implications of this second,multispecific region of the 2F5 paratope for its unique biological functionare discussed.

q 2004 Elsevier Ltd. All rights reserved.

Keywords: mAb 2F5; HIV-1; phage-displayed peptide library; epitope;neutralizing antibody*Corresponding author

Introduction

Human monoclonal antibody (mAb) 2F5 is oneof the few human immunodeficiency type 1 (HIV-1)broadly neutralizing antibodies identified so far.1,2

Passive transfer of 2F5, in combination with otherneutralizing human mAbs, has conferred sterilizingprotection to viral challenge in several animalmodels.3–5 mAb 2F5 binds with high affinity to a con-served site on the extracellular region of the trans-membrane protein gp41 (residues 662–667, aminoacid sequence ELDKWA).6 Results from screeningof phage-displayed peptide libraries and from 2F5IgG binding studies to substituted synthetic peptides

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

E-mail address of the corresponding author:[email protected]

Abbreviations used: mAb, monoclonal antibody; HIV-1, human immunodeficiency virus type 1; HRP,horseradish peroxidase; NZY/Tet/IPTG, NZY mediumsupplemented with 15 mg/ml of tetracycline and 1 mMisopropyl b-D-thiogalactopyranoside; MBP, E. colimaltose-binding protein; ABTS, 2,20-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid); TBS, Tris-bufferedsaline, pH 7.5; BSA, bovine serum albumin, fraction V;PEG/NaCl, 16.7% polyethylene glycol 8000, 3.3 Msodium chloride; Kd, dissociation constant atequilibrium; (23), amino acid position three residues C-terminal to the Trp residue in the DKW core sequence;(þ3), amino acid position three residues N-terminal tothe Asp residue in the DKW core sequence.

doi:10.1016/j.jmb.2004.02.051 J. Mol. Biol. (2004) 338, 311–327

have shown that the tri-peptide DKW is required forbinding, and thus constitutes the core of theepitope.6,7 The crystal structure of 2F5 in complexwith the short synthetic, peptide ELDKWA, revealeda type I b-turn formed by the residues DKW. It alsoshowed that extensive regions of the putative anti-gen-binding site on the antibody, more specificallythe apex of the long H3, did not contact the peptide,indicating sites of potential interaction with otherparts of the envelope spike and/or lipid on the sur-face of the virus.†

Direct binding of 2F5 to viral particles andinfected cells has been observed by Sattentau et al.8

and Zwick et al.,9 suggesting that the epitope isexpressed and accessible on the pre-fusogenicviral envelope. However, other studies indicatethat the 2F5 neutralization-sensitive epitope isexposed or formed only after the virus has con-tacted the cellular receptors, and triggered thecascade of events leading to fusion of the cellularviral membranes.10,11 This contradiction can beexplained by the work of Poignard et al.,12 whichindicates that the viral envelope spikes exist inboth infectious and non-infectious forms; veryprobably 2F5 binds to both forms. Thus, the uniqueneutralizing properties of 2F5 reside in itsspecificity for a cryptic site on the infectious spikethat is probably exposed during infection.

The potency and breadth of 2F5’s neutralizingcapacity suggest an important biological role forthe membrane-proximal ectodomain of gp41,making it an attractive candidate for vaccinepurposes. However, multiple efforts to elicit anti-bodies with the specificity and neutralizing activityof 2F5 by immunization with different forms of theELDKWA fragment have failed repeatedly.13– 16

Several studies have tested the antigenicity andimmunogenicity of larger peptide epitopes basedon the gp41 sequence, as well as peptides havingnovel structural constraints that optimizeantigenicity.17 – 20 However, even with theseimprovements, the antibodies generated bind tosynthetic peptides and/or to recombinantenvelope, but do not neutralize primary HIV-1isolates.17,18 These disappointing results suggestthat the ELDKWA fragment and extensions of itdo not represent the entire 2F5 epitope, or that theneutralizing epitope, if present, is not immuno-dominant. In addition, it is not clear whether thefailure to elicit 2F5-like antibodies results from thepresence of an incomplete epitope in the immuno-gens used so far, and/or because antibodies havingthe unusual structure of 2F5 (30 somatic mutationsand a very long, 22 residue, H3 loop)21 are difficultto elicit by conventional immunization strategies.Antibodies functionally similar to 2F5 are rarelyproduced in natural infection,21 – 23 even though

reactivity against ELDKWA-containing peptideshas been consistently observed in the serum ofHIV-1-infected subjects.24 – 26

Here, we screened several primary and semi-defined phage-displayed peptide libraries withmAb 2F5 with the objective of finding peptide anti-gens with more extensive coverage of the 2F5 para-tope than that of gp41-based peptides. Weanalyzed the sequence requirements of 2F5 byusing it to screen a panel of random peptidelibraries. Our data led us to postulate that if wereduced binding to the DKW core and extendedthe randomized regions flanking it, we could pro-duce libraries that would identify unique, nearbyregions in the antibody paratope that have thepotential to bind protein. We found that the resi-dues flanking the DKW core on the C-terminalside are required for high-affinity binding to 2F5,and that the antibody binds tightly to severalunrelated sequences in this region, making the 2F5paratope multispecific. Thus, our results indicatethat the 2F5 paratope comprises two regions: onethat is highly selective for the core sequence DKW,and another that binds C-terminal to the core andis multispecific. We discuss how these results per-tain to 2F5’s unique biological properties, and toepitope-targeted and antibody-targeted vaccinedevelopment.

Results

mAb 2F5 selects phage bearing peptides withmultiple DKW-flanking sequences from primaryphage-displayed peptide libraries

IgG 2F5 was used to screen 17 phage librariesdisplaying fully randomized peptides of varying

Table 1. Phage-displayed peptides selected by mAb 2F5contain the core sequence DKW and display sequencevariability in the regions flanking DKW

Clone ELISA Sequence

E6.5 0.98 MVLDKW(A)E4.6 0.96 LHEESMDKWSNLMQCCTE9.6 0.73 RANTCDKWALPCNTGNH2.5 0.68 SRKELDKWASTDCMQCQE11.6 0.67 YCTAPLNADKWSCTE10.10 0.58 LGTNCDKWACMHYE8.3 0.54 SCDKWVPYCAE6.7 0.54 NSLDKW(A)E11.9 0.42 DCLTPDRWVSFPSGCME8.1 0.40 TCDKWVARCEE11.1 0.36 VCPKSAQMDRWYCLE10.1 0.28 LDALCDKWSCAKMNE9.8 0.27 HYETCDKWHLCDSHSH2.6 0.25 RCAQVLACLPTHDLDKW(A)E9.9 0.23 AGLYCDKWQTNCKISEE7.3 0.15 SCPPESDRWSCFE7.6 0.14 KCDKWAYDEDCLF88-4 0.05 Parental vector

The core DKW sequence is in bold. The Ala residue corre-sponding to the first amino acid residue of the phage pVIIIprotein is indicated in parentheses. Values are expressed as theabsorbance at 405 nm minus that at 490 nm.

† Pai, E. F., Klein, M. H., Chong, P., Pedyczak, A.(2000). Fab’–epitope complex from the HIV-1 cross-neutralizing monoclonal antibody 2F5. Patent WO-00/61618. World Intellectual Property Organization.

312 Multispecificity of HIV-1-Neutralizing Antibody 2F5

length.27 The phage libraries are built in the f88-4vector, which allows a single phage particle to dis-play 30–300 copies of the same peptide fused tothe N terminus of the major coat protein pVIII.The displayed peptides in many of the librariesalso bear one or two fixed Cys residues. Sequenceanalysis of clones, shown in Table 1, revealed that2F5 mostly selected phage carrying a short coreepitope sequence homologous to that of HIV-1-gp41 (DKW, LDKW or DKWA). Some variationsas DRW and DPW were also observed. The clonesneither displayed significant homology to gp41,nor formed an identifiable consensus sequence inthe regions flanking the DKW, indicating that theselection was primarily based on the presence ofthe DKW motif. However, DKW-flankingsequences apparently determined overall bindingaffinity, since large differences in enzyme-linkedimmunosorbent assay (ELISA) signals wereobserved among clones sharing the DKW motif.Interestingly, phage-bearing peptides with poten-tially Cys-constrained loops of different lengthwere also selected, the shortest loop being fourresidues long. Clones E6.5 and E4.6 showed thestrongest 2F5 reactivity. The tighter binding of 2F5to these clones was not a result of increased pep-tide expression (nor of peptide dimerization forclone E4.6), as evaluated by reducing and non-reducing SDS-PAGE of phage clones, whichseparates recombinant pVIII proteins from wild-type ones (data not shown). Thus, 2F5 binds to alarge array of peptide sequences that contain theDKW core epitope, demonstrating both a strictrequirement for the DKW motif, and the toleranceof multiple sequences in the DKW flanking regionsthat contribute to affinity.

We focused on a group of clones bearing theDKW motif flanked by Cys residues, whosestructure is potentially constrained by disulfidebridging (Table 1, clones E7.3, E8.3, E9.6, E10.10and E11.6). The introduction of disulfide or

b-lactam bridges in 2F5-binding peptides has, inseveral instances, improved affinity for 2F5.19,20

Our previous experience has shown that, in mostcases, if two Cys residues are present in a phage-displayed peptide they usually form an intra-chain disulfide bond, and therefore the peptide isdisplayed as monomer on analysis of whole phageby non-reducing SDS-PAGE.28 In contrast, peptidesbearing a single Cys residue (and some containingtwo Cys) have a strong tendency to form inter-molecular bridges, and thus the peptides aretypically displayed as homo-dimers instead ofmonomers.28,29 SDS-PAGE performed in thepresence or absence of the reducing agent DTTconfirmed the monomeric status of the peptidesexpressed by the phage clones E7.3, E8.3, E9.6,E10.10 and E11.6. Phage ELISA under reducingconditions indicated that disulfide bridges wererequired for 2F5 binding to all of the clones (datanot shown).

A set of synthetic peptides was produced (H2.5,E4.6, E7.3, E8.3, E9.6, E10.10, E11.6 and H2.5) forthe experiments described below. The peptideswere designed so that their C-terminal sequenceswould include N-terminal residues (AEG orAAEG) of the phage major coat protein, pVIII.Also, a biotinylated Lys was introduced at the Cterminus to facilitate immunochemical assay.A peptide (ME2F5) containing the sequence of thenative, 6-mer epitope (ELDKWA) was also madeas a control. Binding of 2F5 to streptavidin-captured synthetic peptides was analyzed by directELISA, with results similar to those from phageELISAs (data not shown). Thus, the behavior ofthe synthetic peptides was similar to that of theirphage-displayed counterparts.

These ELISAs were again performed underreducing conditions (Table 2) to assess the role ofpotential disulfide bridges in the binding of 2F5 tothe cyclic peptides. Biotinylated peptides werecaptured on immobilized streptavidin and treated

Table 2. ELISA results

Peptide Sequence No DTT DTT

E9.6 ARANTCDKWALPCNTGN(AEGK)-bio 0.851 0.285E11.6 YCTAPLNADKWSCT(AEGK)-bio 0.570 0.205E8.3 SCDKWVPYCA(AEGK)-bio 0.742 0.118E10.10 ALGTNCDKWACMHY(AEGK)-bio 0.571 0.242E7.3 SCPPESDRWSCF(AEGK)-bio 0.106 0.064E4.6 LHEESMDKWSNLMQCCT(AEGK)-bio 0.893 0.907H2.5 SRKELDKWASTDCMQCQ(AEGK)-bio 0.796 0.865ME2F5 GGELDKWAGGGK(AEGK)-bio 0.772 0.799No peptide – 0.032 0.017

Phage clone Sequence ELISA 2F5 ELISA anti-phageE8.3 wild-type SCDKWVPYCA 1.00 0.76E8.3C50/S SSDKWVPYCA 0.08 0.77E8.3C30/S SCDKWVPYSA 0.07 0.79f88-4 – 0.01 0.00No phage – 0.01 0.01

Values are expressed as the absorbance at 405 nm minus that at 490 nm. A, ELISA on 2F5-binding synthetic peptides under non-reducing and reducing conditions. B, 2F5 ELISA on wild-type and Cys-substituted mutants of clone E8.3.

Multispecificity of HIV-1-Neutralizing Antibody 2F5 313

with 100 mM DTT. The peptides were then reactedwith 2F5 IgG in the presence of a low (5 mM) con-centration of DTT, which allowed binding of 2F5to the peptides, but prevented re-formation of dis-ulfide bridges in the peptides. Dependency of anintact disulfide bridge was observed for all pep-tides having the DKW sequence bordered by Cysresidues (E8.3, E9.6 and E11.6), and was notdetected in similarly treated clones bearing“linear” peptides (DKW outside the Cys loop orno Cys residues). Substitution of Cys to Ser inclone E8.3, by site-directed mutagenesis, confirmedthese findings and ruled out artifacts of the assayin the observed DTT sensitivity (Table 2). Theresults showed that, for most peptides containingtwo Cys, the constraint(s) imposed by a disulfidebridge are important for binding of 2F5 to phage-displayed peptides or to their synthetic peptidecounterparts.

To detect differences in binding affinity thatcould have been suppressed in the direct ELISAs(due to bivalent binding of the IgG to multivalentpeptide on phage), in solution titration assayswere performed, in which the 2F5 IgG wasimmobilized onto microtiter wells and syntheticpeptides were captured from solution in a mono-valent fashion. Bound peptides were then detectedvia their biotin group using a neutravidin-HRPconjugate. The results in Figure 1 show that pep-

tide E4.6 was the tightest binder, even thoughH2.5 carries the complete ELDKWAS sequencepresent in gp41. Different from gp41, in whichhydrophobic (Leu) and acid (Glu) residues pre-dominate proximal to the N-terminal side of theDKW core, peptide H2.5 has two basic amino acidresidues (Arg and Lys) that could negativelyinfluence its reactivity. The control peptide ME2F5showed poor reactivity as compared to E4.6 andH2.5.

Taken together, these results show that the DKWcore is essential for binding of mAb 2F5 tosynthetic and phage-displayed peptides. Moreover,2F5 has increased affinity for peptides bearing adiversity of sequences flanking the DKW motif,even though those sequences do not share hom-ology to HIV-1 gp41. Thus, for each clone, 2F5binding is determined by its flanking sequencesand/or by the presence of constraints imposed bydisulfide bridging.

The flanking region C-terminal to the DKWmotif is important for strong peptide reactivitywith mAb 2F5

We wanted to further explore the specificity ofthe 2F5 paratope with regard to the level ofsequence variation permitted around the DKWcore. Thus, we constructed two phage-displayed,

Figure 1. A synthetic peptide bearing the displayed sequence from the E4.6 phage clone binds 2F5 IgG tighter thanthe native hexa-peptide epitope (ME2F5) in titration ELISA. mAb 2F5 IgG was adsorbed to microwells and used tocapture biotinylated peptides from solution at the concentrations indicated; bound peptides were detected withneutravidin-HRP and ABTS. Values are expressed as the absorbance at 405 nm minus that at 490 nm.

314 Multispecificity of HIV-1-Neutralizing Antibody 2F5

semi-defined peptide libraries containing theminimal-binding sequence, DKW, flanked by 12variegated residues at either the N or C terminus,to investigate the extent to which the N andC-terminal regions are involved in binding to theantibody. The rationale of the design was to pro-duce two libraries with minimal binding to 2F5,and to select flanking residues that would increasethe affinity of interaction with the antibody. TwoAla residues were introduced at the N-terminalside of the DKW core to prevent selection of thenative Glu and Leu residues at those positions; aSer was fixed C-terminal to the DKW based on thesequence from clone E4.6. These three substitutionsdecreased binding to 2F5 to undetectable levels, asrevealed by ELISA on phage clones that werepicked from the libraries at random, and thewhole libraries (Table 3). Thus, the affinity oftight-binding clones selected from the libraries by2F5 should depend upon both the DKW core, andsequences selected from the variegated N orC-terminal flanking regions.

The libraries X12-AADKW and AADKW-X12

(comprising ,109 independent clones each) weresubjected to three rounds of high-stringencyaffinity selection using biotinylated 2F5 Fab insolution. The deduced amino acid sequences ofthe selected clones are shown in Table 3. Asobserved in clones selected from the primarylibraries, the sequences of the DKW-flankingregions shared little homology with the nativesequence of gp41, nor could they be formed intoconsensus groups. However, some general patternswere observed: a hydrophobic amino acid was

always selected at position (23) (i.e. three residuesN-terminal to the Asp in DKW) from the X12-AADKW library, and Leu was the most frequentresidue selected. Nevertheless, the selection of aLeu at this position was not sufficient to recoverhigh-affinity binding in the context of the clones.All clones carrying Leu(23) showed only marginalreactivity with 2F5 in ELISA; in fact, Tyr or Trpwere present at this position in the tightest-bindingclones (XD10.1, XD10.2 and XD1.2). Clonesselected from the AADKW-X12 library showedselectivity for a hydrophobic residue at theposition (þ3) (i.e. three residues C-terminal to theTrp in DKW), where a Leu or a Trp were preferen-tially selected. Again, the presence of a hydro-phobic residue at this position appeared to berequired, but was not sufficient to achieve high-affinity binding to mAb 2F5. We also noticed amarked prevalence of acidic residues (Asp andGlu) on both sides of the DKW core epitope.

Most of the clones isolated from the two librariesshowed relatively poor binding to 2F5 IgG inELISA, as compared with clone E4.6 (Table 3).Rare mutational events dominated selection fromthe X12-AADKW library, as the tight-binding clonesduplicated the epitope at a position closer to the Nterminus, and thus acquired a longer C-terminalflanking region (clone XD10.1, IEADMLDRWAYYAADKWSAAEG), or mutated the Alaresidues and created an expanded epitope (cloneXD10.2, TSPHASLHMWDLDKWSAAEG), orexpanded and moved the epitope toward the Nterminus and acquired a Trp in position (þ3)(clone XD1.2, EGPYVDYPLDKWAEWSAAEG).

Table 3. Sequences of peptides displayed by phage clones isolated by high-stringency screening of the X12-DKW andDKW-X12 libraries

Clone Binding Sequence

X12D sub 2 XXXXXXXXXXXXAADKWSAAEGDXD10.1 þþþþ IEADMLDRWAYYAADKWSAAEGDXD10.2 þþþþ TSPHASLHMWDLDKWSAAEGDXD1.2 þþþþ EGPYVDYPLDKWAEWSAAEGDXDR4.24 þ SNGAASLLNNYLAADKWSAAEGDXDR4.44 þ RERDKTSLAPPFAADKWSAAEGDXD10.3 2 DPWNDRGSELAIAADKWSAAEGDXD10.4 2 GIRTPMTEIEWLAADKWSAAEGDXD10.5 2 EYGESFDSYWPLAADKWSAAEGD

DX12 sub 2 AADKWSXXXXXXXXXXXXAAEGDDX1.1 þþþþ AADKWSHLSLTEFSPSHQAAEGDDX10.5 þþþþ AADKWSDWPNIAQYTTTLAAEGDDXR4.22 þþþþ AADKWSFWPWGDDSPLLTAAEGDDXR4.14 þþ AADKWSKLGYTPDLPTSPAAEGDDXR4.17 þ AADKWSDFIIYSTIPFRLAAEGDDX10.1 ^ AADKWSTEAHNLMAVDKWAAEGDDX10.3 ^ AADKWSLWDSQSEVRMLTAAEGDDX10.4 ^ AADKWSDWSSNGSILGTRAAEGDDX1.3 ^ AADKWSLWADAPTLGDAMAAEGDDX1.5 ^ AADKWSGWDHSPVTTVLTAAEGDE4.6 þþþþ LHEESMDKWSNLMQCCTAAEGDgp41 SQNQQEKNEQELLELDKWASLWNWFNITNWLWYI

Binding refers to phage reactivity with 2F5 IgG in ELISA, and is represented in arbitrary binding units based on two independentELISAs, normalized with the signal from clone E4.6. X12D sub and DX12 sub refer to the X12-DKW and DKW-X12 sublibraries beforethe screening, respectively. The sequence AAEGD corresponds to the N terminus of the phage major coat protein.

Multispecificity of HIV-1-Neutralizing Antibody 2F5 315

These results indicate that binding to 2F5 cannot beimproved significantly through this means of opti-mizing sequences flanking the N terminus ofAADKW.

In contrast, strong reactivity of 2F5 to severalclones from the AADKW-X12 library was basedsolely on differences in their C-terminal sequence,indicating that there are specific requirements forbinding that can be localized to the regionC-terminal to DKW. One clone with a secondDKW motif in the randomized region(DX10.1, AADKWSTEAHNLMAVDKWAAEG) wasselected, but it was a relatively weak binder.Clones DX1.1, DX10.5 and DXR4.22 were especiallytight binders, and did not display homology intheir flanking regions, indicating that tight bindingto 2F5 can be achieved by more than one particularsequence flanking the C-terminal side of the DKWcore. To confirm and further explore the bindingproperties of the selected peptides to mAb 2F5,the peptide-coding sequences of several cloneswere transferred to an Escherichia coli maltose-bind-ing protein (MBP) fusion expression system, andthe peptides were expressed as N-terminal fusionsto MBP. This allowed the peptides to be analyzedfor binding to 2F5 IgG in a monovalent fashion,out of the multivalent context of the phage coat.As shown in Figure 2, clone XD1.2 from the X12-AADKW library showed the strongest reactivitywith 2F5; this clone was heavily mutated withrespect to the original library design, and bearsthe native LDKWA sequence plus hydrophobicresidues at (23) and (þ3) positions. The reactivityof MBP fusions DX1.1, DX10.5 and DXR42.2, fromthe AADKW-X12 library was stronger than DX10.4,XD10.3 and XD10.4, in keeping with the resultsfrom the phage ELISA.

In summary, these results revealed that 2F5 dis-plays multispecificity for sequences flanking theDKW core, and suggest that binding can beachieved by a diversity of sequences, which likelyestablish different contacts with 2F5. Phage clonesbearing peptides with the DKW sequences wereselected from the libraries, regardless of the natureof their DKW flanking regions; but tight-bindingclones typically required a hydrophobic residue atpositions (23) and/or (þ3), and were more effec-tively selected on the basis of their C-terminalDKW-flanking region.

mAb 2F5 displays similar binding affinity forDX-MBP fusions with different DKWC-terminal sequences

With the objective of learning whether theobserved 2F5 multispecificity would significantlyinfluence the affinity of mAb–peptide interaction,we determined the intrinsic binding affinity of thetight-binding clones by surface plasmon resonance.We also wanted to compare the affinity of the MBPfusions with those of gp160 and gp41, as well assynthetic peptides carrying the full ELDKWAsequence. The initial experimental design involved

the immobilization of mAb 2F5 to the chip andflowing the analyte proteins samples over them.In this way, real-time kinetic data are reliablycollected. Several conditions were tested forimmobilizing 2F5 and regenerating the surfaceafter each use, but they all were unsatisfactory.The opposite arrangement (i.e. antigensimmobilized on the chip and injected antibody)was not explored, since it has been described asgenerating unreliable data, due to the avidity effectinherent in bivalent molecules and the re-bindingevents that take place during the dissociationphase.30,31 Thus, even though this experimentaldesign is widely used for the determination of onand off rates, kinetic data generated with thisprocedure may be questionable. We favored thealternative of determining the affinity at equi-librium in solution as described by Nieba et al.30 Inthis method, the antibody is at a fixed concen-tration and is reacted to equilibrium with increas-ing antigen concentrations. The concentration offree antibody is then determined by injecting thereactions over an antigen-coated sensor chip, andcomparing the data against an antibody calibrationcurve.

In control experiments with mAb b12 and B2.1peptide, we demonstrated that direct kineticanalysis under conditions in which the peptidewas immobilized to the chip and the antibodywas used as analyte, did not provide accurate Kd

values even if Fab was used (not shown). Signifi-cant differences in Kd between the two methodswere also observed for the E4.6 synthetic peptide(see Table 4, Kinetics versus Equil/Sln data forE4.6 pepitide-2F5-IgG). The affinity of 2F5 IgG forthe E4.6 synthetic peptide in solution was deter-mined using analyte concentrations ranging from5.4 mM to 27 nM, and the Kd from three indepen-dent experiments was 840–950 nM. The Kd for thecontrol peptide, ME2F5, was about one order ofmagnitude higher than that of E4.6, indicating astronger 2F5–peptide interaction for E4.6 than forME2F5. The coding sequence for clone E4.6 wasalso transferred to the pMALX vector andexpressed as an N-terminal fusion to MBP. Thepurified product was used to determine the Kd atequilibrium, and this Kd was compared to that ofthe free peptide. As shown in Table 4, the Kd ofthe E4.6-MBP fusion was threefold higher thanthat of the synthetic peptide; in agreement withour previous, unpublished observations and thoseof Jouault et al.,32 which have shown that peptidesfused to a scaffold protein often display a betteraffinity for their cognate antibody than their free,synthetic peptide counterparts.

Also shown in Table 4 are the equilibrium Kd

values for the MBP fusion proteins correspondingto the tight-binding clones DX1.1, DX10.5 andDXR4.22, along with the Kd of the MBP fusionfrom the weak-binding clone DX10.4. The Kd

values for IgG 2F5 and clones DX1.1, DX10.5 andDXR4.22 fusions are very similar, and fall withinthe range of 1.7–2.3 mM, whereas the Kd for the

316 Multispecificity of HIV-1-Neutralizing Antibody 2F5

Figure 2. Titration ELISA on MBP fusions produced from XD and DX clones. 2F5 IgG was adsorbed to microwellsand used to capture MBP fusion proteins from solution at the concentrations indicated. Bound proteins were detectedwith a biotinylated anti-MBP monoclonal antibody and neutravidin-HRP and ABTS. MBP-2* is a purified maltose-binding protein preparation purchased from New England Biolabs (NEB). Values are expressed as the absorbance at405 nm minus that at 490 nm.

Table 4. Kd values obtained for synthetic peptides and fusion proteins by surface plasmon resonance

Reactantsa Methodb Sequence Kd (M)

gp160-2F5 IgG Kinetics 1.8 £ 10212

gp160-2F5 IgG Equil/Sln na 7.2 £ 1027

gp41-2F5 IgG Equil/Sln na 2.3 £ 1028

ME2F5 peptide-2F5 IgG Equil/Sln GGELDKWAGGGK(bio) 7.1 £ 1026

E4.6 peptide-2F5 IgG Kinetics LHEESMDKWSNLMQCCTAAEGK(bio) 7.1 £ 10210

E4.6 peptide-2F5 IgG Equil/Sln LHEESMDKWSNLMQCCTAAEGK(bio) 9.5 £ 1027

E4.6MBP-2F5 IgG Equil/Sln LHEESMDKWSNLMQCCTAAEE-MBP 3.8 £ 1027

DXR4.22MBP-2F5 IgG Equil/Sln AADKWSFWPWGDDSPLLT-MBP 1.8 £ 1026

DXR4.22EL MBP-2F5 IIgG Equil/Sln ELDKWSFWPWGDDSPLLT-MBP 9.2 £ 1028

DX10.5MBP-2F5 IgG Equil/Sln AADKWSDWPNIAQYTTTL-MBP 2.3 £ 1026

DX10.5EL MBP-2F5 IgG Equil/Sln ELDKWSDWPNIAQYTTTL-MBP 7.3 £ 1028

DX1.1MBP-2F5 IgG Equil/Sln AADKWSHLSLTEFSPSHQ-MBP 1.7 £ 1026

DX1.1EL MBP-2F5 IgG Equil/Sln ELDKWSHLSLTEFSPSHQ-MBP 6.0 £ 1028

DX10.4MBP-2F5 IgG Equil/Sln AADKWSDWSSNGSILGTR-MBP 2.3 £ 1025

DX10.4EL MBP-2F5 IgG Equil/Sln ELDKWSDWSSNGSILGTR-MBP 1.9 £ 1027

a For the kinetic method, the first of the reactants in a pair is the one immobilized on the chip.b Method refers to the procedure use for the determination; kinetics involves the flowing of analyte over the immobilized ligand

and determination of association and dissociation rates; in this instance, the Kd at equilibrium is calculated from the rate values.Equil/Sln is based on the determination of free antibody at equilibrium, with one fixed antibody concentration and several concen-trations of peptide/protein analyte. In this case kinetic data are not obtained. na indicates not applicable.

Multispecificity of HIV-1-Neutralizing Antibody 2F5 317

DX10.4-MBP fusion was 23 mM, indicating aweaker interaction with mAb 2F5. These resultsdemonstrate that sequences flanking the C termi-nus of DKW can greatly influence the affinity of2F5 binding to peptide and protein antigens. Thedata also show that high-affinity binding to 2F5can be supported by diverse DKW-flankingsequences that are different from the sequence atthe corresponding location in HIV-1 gp41, thussuggesting that the 2F5 paratope is multispecificfor sequences outside the region binding the DKWcore.

To further optimize their antigenicity, thesequences of the three tight-binding DX cloneswere engineered to expand the native core epitopesequence to ELDKW, by substituting the twoN-terminal Ala residues preceding the DKW inthe DX clones to Glu and Leu. The coding regionsfor these sequences were transferred to pMALX,and the Kd values for their interaction with 2F5were determined. The data in Table 4 show thatthe re-introduction of Glu and Leu resulted in20–30-fold increases in affinity for the tight-bind-ing clones (DX1.1-EL, DX10.5-EL and R4.22-EL),and a 100-fold increase for the weak binder,DX10.4-EL. Thus, optimization decreased, but didnot eliminate, the difference in affinity betweenclone DX10.4-EL and the three other clones. Weobtained again very similar affinities for clonesDX1.1-EL, DX10.5-EL and R4.22-EL, which werealso stronger than the equilibrium Kd for gp160IIIB,and in the same range as gp41MN (see below).

The affinity at equilibrium of mAb 2F5 for HIV-1gp 160IIIB was also determined. Surprisingly, weobtained a much lower affinity value (720 nM)than those reported by Conley et al.6 for gp41 anda synthetic peptide (4.2 nM and 0.7 nM,respectively). We tried to replicate the results ofConley et al.,6 using their experimental conditions,and obtained low picomolar Kd values, based on atotal lack of dissociation of 2F5 IgG from gp160,over a dissociation time of 15 minutes. We believethat the basis for this discrepancy is that Conleyet al.6 used a kinetic assay with gp41 as ligand onthe chip and mAb 2F5 (presumably IgG) asanalyte. Thus the kinetic Kd may be an artifactproduced by the mechanisms mentioned above.Moreover, we used gp160 instead of gp41, thusbasic differences between 2F5 binding to gp160,gp41 and synthetic peptides may be at play. Also,variations in the protein samples (e.g. oligomeriza-tion state) may lead to differences in epitopeaccessibility, as shown by Zeder-Lutz et al.33

Experiments are currently underway to clarifythese observations.

Amino acid residues critical for binding tomAb 2F5 are found in regions flanking theDKW core

To investigate the contribution of particularresidues to 2F5 binding, we generated a set ofAla-substitutions and deletion mutations in clones

DX1.1, DX 10.5 and DXR4.22. Mutant phage cloneswere analyzed for their ability to bind 2F5 Fab(Figure 3 and Table 5) and to 2F5 IgG (Table 6) ina direct ELISA format. Similar results wereobtained in both assays; however, due to thepossibility of bivalent binding of IgG to multicopypeptide on the phage, moderate differences inbinding are typically masked in the IgG-basedassay. Moreover, IgG binding will be more sensi-tive to peptide copy number than Fab, becausepeptide spacing will affect the binding avidity.The data in Figure 3 and Tables 5 and 6 show thatmutation of any residue in the DKW core signifi-cantly decreased binding of all of the DX clones to2F5; this agrees with previous results obtainedwith a peptide with the sequence ELDKWA.7,17,19

In addition, Tables 5 and 6 show that Ala substi-tution of D and W in the DKW core of clone E4.6ablated binding completely. Also crucial to bindingin every case was the hydrophobic residue at the(þ3) position (Figure 3 and Tables 5 and 6). The(þ3) residue in the native gp41 (L669) is highly con-served, (present on 98% of the sequences in the LosAlamos HIV Database). Ala substitution at this sitedid not affect binding in previous work by Joyceet al.;17 however, Tian et al.19 showed by twoindependent means that Leu at this site affectedbinding significantly. Thus, our results, showingan absolute requirement for DKW and a hydro-phobic residue at the (þ3) position, agree withprevious studies involving gp41-derivedsequences. We conclude from this that all fourpeptides (E4.6, DX1.1, DX 10.5 and DXR4.22) aresimilar to the gp41 sequence in being “seated” inthe DKW-binding region of 2F5. Highly variablesequences outside of the DKW core contribute toaffinity; consequently, binding to 2F5 by theseregions appears to be multispecific.

The effect of Ala substitution in regions outsidethe DKW core varied significantly between clones.As summarized in Table 5, binding decreased byover 50% from Ala substitution at positions (þ2and þ7) of clone DX1.1, positions (þ2, þ4 andþ9) of clone 10.5, and positions (þ4, þ5, þ9 andþ11) of clone DXR4.22; there was no selection forshared residues at these sites. Interestingly, thechange of Ser(þ1) (which was a fixed residue inthe library) to Ala, the residue present in nativegp41, significantly increased binding (200–400%);this indicates a previously underestimated role ofthis Ala in binding affinity. Ala substitutions inclone E4.6 that decreased Fab binding by 50% ormore were at positions (25, 24, 23, 22, and 21)from the N-terminal side of the DKW core and(þ3) from the C-terminal side, indicating thatthose residues are important for binding. The IgGbinding data in Table 6 largely supported thesefindings, though, as noted above, differences inbinding (increases and decreases) seen in the Fabassay were muted in assays with IgG. In con-clusion, a number of different residues outside theDKW core were important for reactivity. With theexception of the (þ3) position, the location and

318 Multispecificity of HIV-1-Neutralizing Antibody 2F5

type of these residues varied greatly among theclones. Thus, in each clone studied, residues atdifferent positions C-terminal to the DKW core,and in several cases, C-terminal to the (þ3) site,were important for binding to 2F5. Moreover, forresidues outside the DKW core and the (þ3) pos-

ition, that appear to play a role in binding, therewas little similarity between clones in the type ofamino acid residues selected or their positionrelative to DKW.

We generated a set of deletion mutants for clonesDX1.1, DX10.5 and DXR4.22 by serially removing

Figure 3. Residues flanking the DKW core epitope are involved in the binding of 2F5 Fab to DX1.1, DX10.5 andDXR4.22 phage. ELISA data were converted to percentage binding relative to phage bearing non-mutated sequences.The sequences C-terminal to the DKW core are: DX1.1: SHLSLTEFSPSHQ; DX10.5: SDWPNIAQYTTTL; DXR4.22: SFW-PWGDDSPLLT. The percentage binding to the Ala substitution of the fixed Ser at position (þ1) of clone DX10.5 was469% and is indicated on the graph.

Table 5. Percentage binding of Ala-substituted and deletion mutants of E4.6 and DX phage clones to 2F5 Fab

HXB2gp41 E L D K W A S L W N W F N I T N W L

Position þ1 þ2 þ3 þ4 þ5 þ6 þ7 þ8 þ9 þ10 þ11 þ12 þ13DXR4.22 A A D K W S F W P W G D D S P L L TDXR4.22-Ala 3 0 6 182 74 3 15 11 82 64 73 5 108 6 80 124DXR4.22-Del 0 7 18 41

DX1.1 A A D K W S H L S L T E F S P S H QDX1.1-Ala 0 0 0 255 12 0 101 115 83 33 98 109 98 116 104 96DX1.1-Del 0 37 20 48

DX10.5 A A D K W S D W P N I A Q Y T T T LDX10.5-Ala 0 16 5 469 25 0 40 71 62 134 81 28 183 93 74 97DX10.5-Del 0 59 4 16

HXB2gp41 Q E L L E L D K W A S L W N W F NPosition 26 25 24 23 22 21 þ1 þ2 þ3 þ4 þ5 þ6 þ7 þ8E4.6 L H E E S M D K W S N L M Q C C TE4.6-Ala 99 44 34 15 40 39 0 nd 0 269 52 0 99 80 111 111 ndE4.6-Del 79 101 0 – – – 9 76 121 128

The sequence of the clones and gp41HXB2 are provided for reference. The core DKW sequence is in bold. Ala(þ7) in clone DX10.5was substituted by Gly. nd indicates not done.

Multispecificity of HIV-1-Neutralizing Antibody 2F5 319

three residues at a time from the C-terminal end,resulting in C(23), C(26) and C(29) mutants foreach clone (three, six or nine residues deleted,respectively). We also constructed a C(212) clonehaving only the sequence AADKWS, fixed in thelibraries. Binding to 2F5 Fab was assayed asdescribed above for the Ala substitutions, and thedata are shown in Tables 5 and 6. The resultsdemonstrate that binding of 2F5 to the DX phageis dramatically affected by deletion of theC-terminal residues, confirming their role forbinding in this context. The effect of deletion wasstrongest for clone DX10.5. Deletion of the threeC-terminal residues affected binding moderately(compare Fab to IgG data), even though Ala substi-tution of each of the residues in this region did notaffect binding significantly. Deletion of sixresidues, which included Tyr(þ9), dropped bind-ing to 4% of wild-type for both Fab and IgG. Alareplacement had only a moderate effect onTyr(þ9), as it decreased Fab binding significantly,but not IgG binding. Deletion of nine residues,including Tyr(þ9), did not. Thus, the deletion ofmultiple residues, which singly had small effectson binding, appeared to produce larger effectswhen deleted together. This may indicate thepresence at this site of a larger binding motif thatdepends on the combined effect of multipleresidues, including Tyr(þ9).

This trend was also observed with clone DX1.1.Ala substitution of Glu(þ7) affected bindingmoderately, whereas substitution of any of the lastsix C-terminal residues did not affect bindingsignificantly. Deletion of three residues from the Cterminus caused a moderate drop in binding (52%drop in Fab binding and a 19% drop in IgG bind-ing), even though Ala substitution of the residuesin this region had little effect. The removal of allsix terminal amino acid residues in this cloneresulted in a larger drop in binding, even thoughAla substitution, again, did not affect any singleresidue in this region.

In contrast to clones DX1.1 and DX10.5, the effectof Ala substitution was strongest for cloneDXR4.22, (especially for residues Trp(þ5), Ser(þ9)and Leu(þ12)), and deletion of regions coveringthese positions produced progressively largerdrops in binding (59%, 82% and 93% drops in Fabbinding to clones carrying the (23), (26) and(29) deletions, respectively). This indicates thatclone DXR4.22 contains residues in the C-terminalregion that most likely have stronger individualeffects on binding than the C-terminal residues inclones DX1.1 and DX10.5. Thus, binding to 2F5appears to depend on particular residues in theC-terminal region of clone DXR4.22, whereas bind-ing to clones DX1.1 and DX10.5 appears to dependon more global effects produced by multipleresidues within this region. These global effectscannot be provided by any random C-terminalsequence, but are rather specific, since 2F5 did notbind to clones chosen at random from the library,nor to the library as a whole.

The Ala substitution and deletion mutants ofclone E4.6 revealed that sequences N andC-terminal to the DKW core were important forbinding. Residues Glu(24), Glu(23) and Leu(þ3)were strongly affected by Ala substitution. How-ever, binding to E4.6 was ablated only by deletionof both Glu(24) and Glu(23) on the N-terminalside, or a deletion on the C-terminal side includingLeu(þ3). Thus, with this clone, but not clonesselected from the X12-AADKW library, residues N-terminal to the DKW core were important for bind-ing. Perhaps this is due to the presence of Leu inthe (þ3) position in the E4.6 clone, but not inclones in the X12-AADKW library.

Taken together, our results show that the DKWcore and a hydrophobic residue at the (þ3)position are absolutely required for binding bytight-binding clones from the AADKW-X12 library.Moreover, C-terminal sequences beyond (þ3) arealso involved in binding, but they fall into twogeneral categories: (i) those in which particular

Table 6. Percentage binding of Ala-substituted and deletion mutants of E4.6 and DX phage clones to 2F5 IgG

HXB2gp41 E L D K W A S L W N W F N I T N W L

Position þ1 þ2 þ3 þ4 þ5 þ6 þ7 þ8 þ9 þ10 þ11 þ12 þ13DXR4.22 A A D K W S F W P W G D D S P L L TDXR4.22-Ala 10 0 0 115 96 0 71 28 108 110 117 20 113 29 109 108DXR4.22-Del 0 5 67 85DX1.1 A A D K W S H L S L T E F S P S H QDX1.1-Ala 0 0 2 103 45 1 100 99 94 68 93 100 93 93 93 89DX1.1-Del 0 62 58 81DX10.5 A A D K W S D W P N I A Q Y T T T LDX10.5-Ala 0 23 15 132 18 6 26 59 84 99 73 71 100 70 82 81DX10.5-Del 0 2 4 81

HXB2gp41 Q E L L E L D K W A S L W N W F NPosition 26 25 24 23 22 21 þ1 þ2 þ3 þ4 þ5 þ6 þ7 þ8E4.6 L H E E S M D K W S N L M Q C C TE4.6-Ala 91 43 28 9 43 34 0 nd 0 106 49 3 85 93 100 100 ndE4.6-Del 76 95 0 – – – 1 67 86 100

The sequences of the clones and gp41HXB2 are provided for reference. The core DKW sequence is in bold. Ala(þ7) in clone DX10.5was substituted by Gly. nd indicates not done.

320 Multispecificity of HIV-1-Neutralizing Antibody 2F5

residues appear to play a significant role in binding(e.g. clone DXR4.22); and (ii) those in whichmultiple residues produce a more global effect,but do not individually affect binding significantly(e.g. clones DX1.1 and DX10.5). Besides this,residues N-terminal to the DKW core also canplay a significant role in binding (e.g. in cloneE4.6), but this may rely upon the presence of ahydrophobic residue in the (þ3) position.

Discussion

Here, we have shown that mAb 2F5 binds tomultiple peptides bearing the DKW motif, andthat sequences C-terminal to this motif arerequired for high-affinity binding in the context ofphage-displayed peptides. We documentdifferences in binding strength between phageexpressing distinct peptides, and show that somedegree of restriction is imposed by the nature ofthe sequences flanking the DKW core, as well assequence length and constraints imposed bydisulfide bridging (see Table 1). In agreement withprevious reports, phage selected by 2F5 containedthe DKW motif.6,34 However, we found that in thecontext of N-terminal fusion to phage major coatprotein or to the MBP of E. coli, the motif isnecessary but not sufficient for high-affinity bind-ing to mAb 2F5. Surface plasmon resonancestudies using MBP fusion showed that tight bind-ing in this context is recovered by extension of theC-terminal flanking region with sequences notnecessarily homologous to gp41. Peptides withdistinct sequence composition displayed similaraffinities, yet possessed different distributions ofcritical binding residues, indicating that therequirements for binding to 2F5 can be resolved ina variety of ways. This behavior illustrates asignificant degree of functional plasticity of the2F5 antibody, and indicates that its paratopecomprises at least two regions: one that is highlyselective for the core sequence DKW, and anotherthat binds C-terminal to the core and ismultispecific.

The epitope for mAb 2F5 was originally definedas the sequence ELDKWA on gp41.6,34 Usingsynthetic peptides, Purtscher et al.7 found that theN-terminal Glu and the C-terminal Ala residuescould be changed to any amino acid withoutmajor effect on 2F5 binding. In addition, theN-terminal Leu tolerated most substitutions,although the reactivity was reduced in most cases.Replacements in the Asp, Lys and Trp were muchmore restrictive, indicating that these threeresidues were absolutely required for binding.However, work by Tian et al.19 has reported that,to obtain maximal antigenicity, the epitope mustbe extended to a non-apeptide sequence runningfrom (and including) Leu661 to Leu669 of gp41(LELDKWASL), whereas Barbato et al.20 describedan even larger fragment of 13–15 residues (659–671/673). Furthermore, Parker et al.35 digested SOS

gp140-2F5 complexes with protease, and foundthat the sequence from Asn656 to Asn671 of thegp140 (NEQELLELDKWASLWN) was protectedby antibody binding. McGaughey et al.18 observedthat not only the extension of the epitope, but alsothe inclusion of a lactam-bridge constraint, werenecessary to achieve maximal binding of 2F5 tosynthetic peptides. In addition to these epitope-mapping studies, several immunizations withvariations of the 6-mer sequence (ELDKWA) failedto produce a 2F5-like, neutralizing antibodyresponse, even though high titers of peptide,gp160 or gp41-binding antibodies were routinelyproduced.13 – 16 Collectively, these results indicatethat the complete 2F5 epitope is formed by morethan the 6-mer sequence.

Using a protein-fusion system in which minimalbinding to the antibody was initially defined bythe presence of the core DKW tri-peptide, weshowed that 2F5 selected C-terminal sequencesthat significantly boost affinity. We found thattight-binding to 2F5 was associated with well-defined sequence patterns; hydrophobic residueswere selected at position (23) with respect to theDKW in clones from the X12-AADKW library, andat position (þ3) in clones from the AADKW-X12

library, indicating that hydrophobicity at thosepositions is required for high-affinity binding. Theimportance of hydrophobic residues at positions(23) and (þ3) was highlighted by clones E4.6(which follows the hydrophobic (þ3) rule) andXD1.2 (which follows the hydrophobic (23) and(þ3) rules). Fusions of the E4.6 and XD1.2sequences to MBP displayed very similar behaviorin titration assays (data not shown); for example,they were tighter binders than clone H2.5, whichcontains the full 7-mer ELDKWAS sequence foundon gp41, but does not follow the hydrophobic(23) or (þ3) rules. If the sequences of the XD andDX clones are superimposed onto the sequence ofHIV-1 gp41, the Leu(23) in the XD peptides alignswith Leu661 in gp41 and the (þ3) position in theDX clones aligns with Leu669. Those two Leu resi-dues are highly conserved in HIV-1 isolates, andare considered part of the high-affinity, extendedepitope reported by Parker et al.,35 Tian et al.19 andBarbato et al.20 Ala substitution studies on clonesE4.6, DX1.1, DX10.5 and DXR4.22 confirmed theimportance of the DKW motif as well as the (23)and (þ3) hydrophobic residues. Surprisingly, theintroduction of an Ala at position (þ1) increasedbinding significantly in all clones, suggesting animportant role of that residue, which had not beenappreciated by Purtscher et al.7 in the context ofsynthetic peptides.

We found that several of the peptides, includinga short loop of four residues, are dependent on anintact disulfide bridge for binding to 2F5. Con-straint of peptides containing the DKWA regionby disulfide or beta-lactam bridges converts therandom-coil structure of ELDKWASL peptides tobeta-turn structures, and in general, improves theaffinity of binding.17,18 Crystallographic data show

Multispecificity of HIV-1-Neutralizing Antibody 2F5 321

that the bound structure of the DKW sequencewithin the ELDKWA peptide forms a type Ib-turn,8 a common feature found on antibody-bound peptides.36 Apparently, this particularstructure is required for binding, which may inturn explain the strict requirement for the DKWmotif.

We confirmed previous findings by othersregarding the necessity of the core DKW sequencefor binding to 2F5, and the positive effect of anextended epitope, as well as hydrophobic residuesat positions (23) and (þ3) of the core. Neverthe-less, the screening of our X12-AADKW andAADKW-X12 libraries also showed that high-affinity binding is achieved via the selection ofsequences beyond the (þ3) residue on theC-terminal side of DKW. Antibody selection forresidues (þ4) to (þ14), C-terminal to the DKW,was evident by the Ala-substitution and thedeletion-mutagenesis results from clones DX1.1,DX10.5 and DXR4.22. Residues having a significanteffect on binding (as assessed by Fab and IgGbinding, Tables 5 and 6) were identified in theC-terminal flanking region of clone DXR4.22,demonstrating that tight-binding to 2F5 is sup-ported by residues downstream to the DKW core.Ala substitutions on clones DX1.1 and DX10.5 hada lesser effect on binding, and provided evidenceof an unequal number of influential residueswithin each clone, as well as differences in theirpositioning in the sequence.

In some instances the results of the deletionmutagenesis did not support those from the singleAla substitutions. Deletion of residues from the Cterminus of clone DXR4.22 decreased binding in amanner that followed loss of residues that weresensitive to Ala substitution. In contrast, similardeletions in clones DX1.1 and DX10.5 decreasedbinding significantly, even though the deletedresidues were not strongly affected by Ala substi-tution. The meaning of this latter discrepancy isdifficult to interpret but it indicates the necessityof a larger sequence for 2F5 binding. The existenceof a secondary binding site in the C-terminalregion of clones DX1.1 and DX10.5 may beinferred, but is not proven by our data. The appar-ent lack of critical binding residues in clonesDX1.1 and DX10.5, and their sensitivity to deletioncould also indicate that these residues are requiredto present the correct core to the antibody, ratherthan participate in specific side-chain contacts, ashas been observed by Barbato et al.20 In contrast,clone DXR4.22 behaves differently. In this case, thepresence of several critical binding residues mayreflect a stronger structural constraint of the pep-tide, or direct contact(s) with the 2F5 paratope.The latter alternative, the establishment of side-chain contacts, would imply an essentially differ-ent mechanism of binding between DXR4.22 andDX1.1/DX10.5. It is worth noting that in spite ofthe sequence diversity of the DX clones, their MBPfusions displayed very similar affinity for 2F5(Figure 2 and Table 4). These results indicate that

the requirements for high-affinity interaction with2F5 can be satisfied by DKW flanked by diverseC-terminal sequences.

There is no direct information on how the struc-tural features of the cognate epitope for 2F5 (i.e. inthe context of the native viral envelope) influencethe antigenicity of potential 2F5 ligands. Thenotion of a larger epitope is supported by bio-chemical evidences, and by the solved structure ofthe antibody in complex with the ELDKWApeptide from Pai et al.8. This peptide only covers afraction of the putative paratope, leaving open thepossibility that 2F5 makes additional contacts withother regions of the complete epitope. Based onthe structural and functional properties of theN-terminal37 – 39 and the C-terminal40,41 regionsflanking DKW, it can be assumed that in the viralspike, the 662–667 region of gp41 must have awell-defined structure. Whatever this structure is,it is likely to be important for both antigenicity(the binding of 2F5) and immunogenicity (theelicitation of neutralizing antibody), and it isunlikely to be replicated by peptide immunogensbearing the ELDKWA sequence alone.

Our results show that 2F5 is multispecific, in thatit can bind with high affinity to multiple non-homologous sequences. It is also functionally dis-tinct from other human and non-human antibodiescapable of binding to ELDKWA peptides, since itneutralizes a wide spectrum of primary HIV-1isolates in vitro. In general, virus neutralizationdoes not correlate with the binding of antibody torecombinant envelope,42 nor even with binding toinfectious viral particles.12 A recent study byPoinard et al.12 showed that non-neutralizing anti-bodies F105 and b6 bind to infectious viral par-ticles, but do not affect the ability of a neutralizingantibody b12 to block infection, even though F105and b6 compete with b12 for binding to gp120envelope protein in vitro. They suggest that theenvelope exists in two different forms on the viralsurface: a functional form involved in the infectionprocess (and hence, a target for neutralization byantibody) and a non-functional one that does notparticipate in infection (and is irrelevant to virusneutralization), and allows binding of bothneutralizing and non-neutralizing antibodies.

Recent results indicate that the neutralization-sensitive binding site for 2F5 is formed only afterthe fusion process is triggered.10,11 It is reasonableto assume that the ability of 2F5 to neutralize thevirus lies in a unique determinant that allows it tobind to its epitope on infectious viral spike,whereas other non-neutralizing ELDKWA-bindersdo not. This elusive feature of 2F5 has not yetbeen biochemically or structurally characterized; itprobably resides in regions of the paratope notinvolved in binding to the extended DKW core.The identification of such determinant(s) bearsrelevance to vaccine design. The binding studieswith our peptides and, our Ala substitution resultsdemonstrated that 2F5 is capable of binding to avariety of peptide sequences, suggesting multiple

322 Multispecificity of HIV-1-Neutralizing Antibody 2F5

mechanisms of binding by the C-terminal region ofour peptides. Such plasticity may be a uniquecharacteristic distinguishing 2F5 from otherELDKWA-binders, and could be involved in itsunique functional properties. We are currently con-ducting studies with sera from animals immunizedwith the ELDKWA region to assess whether thebinding to our DX peptides is an exclusive trait of2F5. Peptides having highly restricted specificityfor 2F5 have potential as structural probes ofunique regions of the 2F5 paratope.

mAb 2F5 has a high level of somatic mutation(16 mutations on VH and 14 on VL)21 and a verylong H3 of 22 residues, which very likely arosefrom sustained antibody evolution in the presenceof persistent antigen stimulation.43 The crystalstructure of 2F5 bound to ELDKWA peptideshows that most of the amino acid residues on theapex of the H3 are not contacted by peptide. How-ever, recent functional studies, in which residues ofthe 2F5 H3 were mutated to Ala, showed that theamino acid residues in the apex of H3 are import-ant for the binding of 2F5 both to gp41 and tosynthetic peptides, as well as being critical toneutralization.44 Elucidation of the structure ofR4.22 and 10.5 peptides bound to 2F5, coupledwith mutagenesis studies similar to those ofZwick et al.,44 may reveal site(s) on the antibodyinvolved in its unique neutralizing properties.

Peptide immunogens tested so far have failed toproduce a neutralizing response, very likely dueto their inability to present the appropriate,complete neutralizing epitope.13 – 16 The idealimmunogen for eliciting antibodies having 2F5-like neutralizing properties would comprise theneutralizing epitope on the infectious spike in theabsence of other immunodominant epitopes. Someapproaches to engineering such an immunogenmight include designed peptides in the context oflipid (e.g. Schibli et al.41), virus-like particles bear-ing such a structure, and/or larger fragments ofgp41 in which irrelevant, immunodominant siteshave been obscured (e.g. by glycosylation;Pantophlet et al.45). Peptide-immunogens contain-ing regions relevant to the neutralization-sensitivesite on 2F5 may serve to enhance the productionof 2F5-like antibodies in such a scenario, followinga prime-boost immunization strategy. In keepingwith the strategy of Beenhouwer et al.,46 neutraliz-ation may be enhanced by a priming immunizationusing an immunogen that contains the neutralizingepitope, followed by boosting immunization withan immunogen that selectively cross-reacts withthe targeted antibody. However, as 2F5 contains alarge number of somatic mutations, as well as anunusually long CDR-H3 loop, such an immuniz-ation approach may not elicit antibodies havingall of 2F5’s structural features. Nevertheless, pep-tides, such as those described here, may serve asprobes in revealing the site(s) on the virus that arerequired for neutralization, and perhaps, asimmunogens that will enhance the production ofneutralizing antibodies.

Materials and Methods

Materials

Seventeen “primary” phage libraries displaying fullyrandomized peptides (with some libraries containingone or two fixed Cys residues)27 were screened withmAb 2F5 IgG. Semi-defined peptide libraries wereconstructed using the vector f88-4,47 as described byBonnycastle et al.48 mAb 2F5 IgG and HIV-1 envelopeproteins were obtained from the NIH AIDS Researchand Reference Reagent Program and the MRCCentralised Facility for AIDS Reagents. mAb 2F5 wasalso donated by H. Katinger (University of AgriculturalSciences, Vienna). mAb 2F5 Fab was kindly donated byM. Zwick & D. Burton (The Scripps Research Institute,La Jolla, CA), it was produced in E. coli using apCOMB-3-based expression system developed by Zwicket al.44 Oligonucleotides were synthesized “in house”using an ABI 392 synthesizer, or by the NAPS Unit atthe University of British Columbia, Vancouver, Canada.Peptides E7.3, SCPPESDRWSCFAEGK(bio); E8.3,SCDKWVPYCAAEGK(bio); E9.6, ARANTCDKWALPCNTGNAEGK(bio); E10.10, ALGTNCDKWACMHYAEGK(bio); E11.6, YCTAPLNADKWSCTAEGK(bio);H2.5, SRK ELDKWASTDCMQCQAEGK(bio) andME2F5, GGELDKWAGGGK(bio) were synthesizedusing a multi-pin system from Chiron Mimotopes PTYLtd (Victoria, Australia) and Fmoc chemistry, accordingto the manufacturer’s instructions. Peptide E4.6,LHEESMDKWSNLMQCCTAAEGK(bio) was purchasedfrom Multiple Peptide Synthesis (San Diego, CA).Synthetic peptide concentrations were determined bymeasuring their absorbance at 280 nm, and calculatingconcentration using an extinction coefficient of 5560 foreach Trp residue in the peptide seqeunce. Protein con-centrations were determined with the Protein Assay kitfrom Biorad (Hercules, CA).

HRP-conjugated, goat anti-human Ig G (Fab-specific)and HRP-conjugated neutravidin were purchased fromPierce (Rockford, IL). A mAb against the maltose-bind-ing protein of E. coli (MBP) was from New EnglandBiolabs (Beverly, MA). Dialyzed BSA, dithiothreithol,H2O2 and ABTS were purchased from Sigma (St. Louis,MO).

Bacterial strains and culture media

E. coli strain K91 was used for phage production, fol-lowing the procedures described by Bonnycastle et al.49

NZY Tet/IPTG was routinely used for phage amplifica-tion. Electrocompetent E. coli strain MC1061 was used toproduce mutant phage in site-directed mutagenesisstudies, and for the production of peptide libraries.48

E. coli strain ER2507, a gift from New England Biolabs,was used for the production of the MBP fusion proteins.

DNA manipulations

The generation of deletion mutants was performedusing linear, single-stranded phage DNA as template,following the procedure described by Bonnycastle et al.27

Site-directed mutagenesis for the generation of aminoacid substitutions and deletions were made with closed,circular single-stranded phage DNA as described bySambrook et al.50 The transfer of peptide-codingsequences to the MBP fusion system pMALX and theconditions for culture and purification of the protein

Multispecificity of HIV-1-Neutralizing Antibody 2F5 323

have been described by Zwick et al.31 DNA sequencingfrom partially purified phage clones was performedwith the Thermo Sequenase II Dye Terminator Cycle kit(Amersham Biosciences, NJ) following the manufacture’sinstructions. DNA fragments from the sequencingreactions were resolved using an ABI 373 apparatus andanalyzed with EditView 1.0.1 software.

Screening of peptide libraries and characterizationof peptides

The biotinylation of antibodies was performed asdescribed by Menendez et al.51 The 17 primary phage-displayed peptide libraries were screened “in solution”with 100 nM 2F5 IgG, as described by Menendez et al.51

Phage–antibody complexes were captured out ofsolution with streptavidin-coated magnetic beads(Dynal, Lake Success, NY). The semi-defined librarieswere subjected to three separated screenings with threedifferent, low concentrations of biotinylated 2F5 Fab(0.1, 1 and 10 nM). Starved K91 cells were infected withpools of phage selected after the third rounds of screen-ing, and individual clones were isolated and amplifiedin 2 ml of NZY Tet/IPTG as described by Bonnycastleet al.49 Phage were partially purified and concentratedby precipitation in PEG/NaCl and re-suspension in TBScontaining 0.02% (w/v) sodium azide. Phage concen-trations were estimated by agarose gel electrophoresisand absorbance determination as described byBonnycastle et al.49 PEG-purified phage clones wereused for ELISAs and DNA sequencing.

Enzyme-linked immunosorbent assay (ELISA)

Phage ELISA were performed as described by Zwicket al.52 Briefly, 2 £ 1010 phage particles were adsorbed tomicrotiter wells at 4 8C overnight, followed by blockingwith 200 ml of TBS containing 2% (w/v) BSA for onehour at 37 8C. The plates were washed three times withTBS containing 0.1% (v/v) Tween, and incubated withthe target antibody for two to four hours at room tem-perature. The plates were washed six times, and 35 ml ofa 1:500 dilution of goat anti-human (Fab specific) anti-body conjugated to HRP were added for one hour atroom temperature. The plates were washed six times,and the reaction was developed with 0.03% (v/v) H2O2

and 400 mg/ml of ABTS in citrate/phosphate buffer.Absorbance was measured in a Biotek EL 312e platereader at 405 and 490 nm for a period of up to 45 minutes.Values of absorbance at 405 nm minus absorbance at490 nm are reported. Replica, phage-coated wells werereacted with a rabbit polyclonal anti-phage antibody toverify that similar amounts of phage particles wereadsorbed to the plate. Bound anti-phage antibody wasdetected with a protein A-HRP conjugate (Pierce) andthe H2O2/ABTS mixture described above.

For synthetic peptide ELISAs, 35 ml of TBS containing1 mg of streptavidin (Roche Diagnostics, QC, Canada)were added to the microtiter wells and incubated at4 8C overnight. Next day, after blocking with TBS/2%BSA at 37 8C for one hour, an excess of biotinylatedpeptide (100 ng in 35 ml of TBS) was added to each welland captured for 30 minutes at room temperature. Theaddition and detection of bound antibody wereperformed as described for the phage ELISA.

Reducing ELISA included an additional step of treat-ment with 100 nM DTT for one hour at 4 8C after theBSA-blocking step. The incubation with 2F5 was carried

out in the presence of 5 mM DTT to prevent re-formationof the disulfide-bridges in the peptides.

For peptides and MBP-titration ELISAs, we used aprocedure that captured the antigen out of solution in amonovalent fashion, thus minimizing avidity. mAb 2F5IgG (100 ng in 35 ml of TBS) was adsorbed to microtiterwells at 4 8C overnight. Next day, the antibody wasremoved, and the wells were blocked with 200 ml of 2%BSA in TBS at 37 8C for one hour. The wells were washedthree times with TBS/0.1% Tween, and 35 ml of serialdilutions of the synthetic peptides or the MBP fusionswere added and incubated for two hours at roomtemperature. The wells were washed six times withTBS/0.1% Tween. For detection in the peptide titrations,35 ml of 1 mg/ml of neutravidin-HRP were added andincubated for 30 minutes at room temperature. Fordetection in the MBP fusion titrations, 10 nMbiotinylated anti-MBP monoclonal antibody was addedand incubated for one hour at room temperature, theplates were washed six times, and 35 ml of 1 mg/mlneutravidin-HRP were added and incubated for30 minutes at room temperature. The reactions weredeveloped with H2O2 and ABTS as described above.

Affinity determinations

The affinity of 2F5-antigen interactions was studied atthe University of British Columbia Centre for BiologicalCalorimetry using a Biacore 3000 apparatus (BiacoreAB, Uppsala Sweden). The binding-inhibition proceduredescribed by Nieba et al.30 and Karlsson53 was followed.CM5 sensor chips were covalently coated at high densitywith synthetic E4.6 peptide using the Biacore amidecoupling kit and following the manufacture’s instruc-tions. Independent reactions of 2F5 Ig G and peptide orprotein analytes were incubated at 4 8C overnight inbuffer HBS-EP (Biacore) to allow them to reach equi-librium. The final 2F5 concentration used was 5 nM forall experimental points; six or seven analyte (syntheticpeptides or MBP fusion proteins) concentrations,ranging from 27 nM to 5.4 mM, were routinely exploredin a single experiment. Equilibrium reactions wereinjected in duplicate over a E4.6-coated chip at a flow of3 ml/minute. Under these conditions, binding of the anti-body to the chip is limited by mass transfer, thus thereaction is driven by free antibody concentration, andnot by the affinity for the immobilized ligand.54,55 Boundantibody was recorded as resonance units very early inthe association phase (ten seconds after the start of theinjection). The concentration of free antibody for eachreaction was estimated by comparison of the experimen-tal data with an antibody calibration curve generatedwith 2F5 concentrations of 1.25–20 nM, in the absenceof antigen. Free antibody concentrations were plottedagainst their respective analyte concentration, and thecurves were fitted to the “in solution” affinity model pro-vided with the BiaEvaluation 3.1 software (Biacore).

Acknowledgements

We thank A. Burguess, R. Astronomo, A. Johner,H. Sarling, B. Gangadhar, B. Vanderkist and S. Wufor their technical assistance. We are also gratefulto M. Zwick, D. Burton, H. Katinger, the MRCCentralised AIDS Facility (UK) and the NIH AIDS

324 Multispecificity of HIV-1-Neutralizing Antibody 2F5

Research and Reference Reagents Program for thedonation of materials. We thank L. Creagh (UBC)for help with the Biacore studies. This work wassupported by NIH/NIAID grants R21-AI49808and RO1-AI49111. A.M. was supported by aGraduate Scholarship from the Michael SmithFoundation for Health Research, BC, Canada.

References

1. Trkola, A., Pomales, A. B., Yuan, H., Korber, B.,Maddon, P. J., Allaway, G. P. et al. (1995). Cross-clade neutralization of primary isolates of humanimmunodeficiency virus type 1 by humanmonoclonal antibodies and tetrameric CD4-IgG.J. Virol. 69, 6609–6617.

2. D’Souza, M. P., Livnat, D., Bradac, J. A. & Bridges,S. H. (1997). Evaluation of monoclonal antibodies tohuman immunodeficiency virus type 1 primaryisolates by neutralization assays: performancecriteria for selecting candidate antibodies for clinicaltrials. AIDS Clinical Trials Group Antibody SelectionWorking Group. J. Infect. Dis. 175, 1056–1062.

3. Hofmann-Lehmann, R., Vlasak, J., Rasmussen, R. A.,Smith, B. A., Baba, T. W., Liska, V. et al. (2001). Post-natal passive immunization of neonatal macaqueswith a triple combination of human monoclonal anti-bodies against oral simian-human immunodeficiencyvirus challenge. J. Virol. 75, 7470–7480.

4. Mascola, J. R., Lewis, M. G., Stiegler, G., Harris, D.,VanCott, T. C., Hayes, D. et al. (1999). Protection ofMacaques against pathogenic simian/humanimmunodeficiency virus 89.6PD by passive transferof neutralizing antibodies. J. Virol. 73, 4009–4018.

5. Mascola, J. R., Stiegler, G., VanCott, T. C., Katinger,H., Carpenter, C. B., Hanson, C. E. et al. (2000).Protection of macaques against vaginal transmissionof a pathogenic HIV-1/SIV chimeric virus by passiveinfusion of neutralizing antibodies. Nature Med. 6,207–210.

6. Conley, A. J., Kessler, J. A., II, Boots, L. J., Tung, J. S.,Arnold, B. A., Keller, P. M. et al. (1994). Neutraliz-ation of divergent human immunodeficiency virustype 1 variants and primary isolates by IAM-41-2F5,an anti-gp41 human monoclonal antibody. Proc. NatlAcad. Sci. USA, 91, 3348–3352.

7. Purtscher, M., Trkola, A., Grassauer, A., Schulz, P. M.,Klima, A., Dopper, S. et al. (1996). Restricted anti-genic variability of the epitope recognized by theneutralizing gp41 antibody 2F5. Aids, 10, 587–593.

8. Sattentau, Q. J., Zolla-Pazner, S. & Poignard, P.(1995). Epitope exposure on functional, oligomericHIV-1 gp41 molecules. Virology, 206, 713–717.

9. Zwick, M. B., Labrijn, A. F., Wang, M., Spenlehauer,C., Saphire, E. O., Binley, J. M. et al. (2001). Broadlyneutralizing antibodies targeted to the membrane-proximal external region of human immuno-deficiency virus type 1 glycoprotein gp41. J. Virol.75, 10892–10905.

10. Gorny, M. K. & Zolla-Pazner, S. (2000). Recognitionby human monoclonal antibodies of free and com-plexed peptides representing the prefusogenic andfusogenic forms of human immunodeficiency virustype 1 gp41. J. Virol. 74, 6186–6192.

11. Binley, J. M., Cayanan, C. S., Wiley, C., Schulke, N.,Olson, W. C. & Burton, D. R. (2003). Redox-triggeredinfection by disulfide-shackled human immuno-

deficiency virus type 1 pseudovirions. J. Virol. 77,5678–5684.

12. Poignard, P., Moulard, M., Golez, E., Vivona, V.,Franti, M., Venturini, S. et al. (2003). Heterogeneityof envelope molecules expressed on primary humanimmunodeficiency virus type 1 particles as probedby the binding of neutralizing and nonneutralizingantibodies. J. Virol. 77, 353–365.

13. Coeffier, E., Clement, J. M., Cussac, V., Khodaei-Boorane, N., Jehanno, M., Rojas, M. et al. (2000).Antigenicity and immunogenicity of the HIV-1 gp41epitope ELDKWA inserted into permissive sites ofthe MalE protein. Vaccine, 19, 684–693.

14. Eckhart, L., Raffelsberger, W., Ferko, B., Klima, A.,Purtscher, M., Katinger, H. & Ruker, F. (1996).Immunogenic presentation of a conserved gp41epitope of human immunodeficiency virus type 1on recombinant surface antigen of hepatitis B virus.J. Gen. Virol. 77, 2001–2008.

15. Ferko, B., Katinger, D., Grassauer, A., Egorov, A.,Romanova, J., Niebler, B. et al. (1998). Chimericinfluenza virus replicating predominantly in themurine upper respiratory tract induces local immuneresponses against human immunodeficiency virustype 1 in the genital tract. J. Infect. Dis. 178,1359–1368.

16. Liang, X., Munshi, S., Shendure, J., Mark, G., III,Davies, M. E., Freed, D. C. et al. (1999). Epitope inser-tion into variable loops of HIV-1 gp120 as a potentialmeans to improve immunogenicity of viral envelopeprotein. Vaccine, 17, 2862–2872.

17. Joyce, J. G., Hurni, W. M., Bogusky, M. J., Garsky,V. M., Liang, X., Citron, M. P. et al. (2002). Enhance-ment of alpha -helicity in the HIV-1 inhibitory pep-tide DP178 leads to an increased affinity for humanmonoclonal antibody 2F5 but does not elicit neutra-lizing responses in vitro. Implications for vaccinedesign. J. Biol. Chem. 277, 45811–45820.

18. McGaughey, G. B., Citron, M., Danzeisen, R. C.,Freidinger, R. M., Garsky, V. M., Hurni, W. M. et al.(2003). HIV-1 vaccine development: constrainedpeptide immunogens show improved binding to theanti-HIV-1 gp41 MAb. Biochemistry, 42, 3214–3223.

19. Tian, Y., Ramesh, C. V., Ma, X., Naqvi, S., Patel, T.,Cenizal, T. et al. (2002). Structure-affinity relation-ships in the gp41 ELDKWA epitope for the HIV-1neutralizing monoclonal antibody 2F5: effects ofside-chain and backbone modifications and confor-mational constraints. J. Pept. Res. 59, 264–276.

20. Barbato, G., Bianchi, E., Ingallinella, P., Hurni, W. H.,Miller, M. D., Ciliberto, G. et al. (2003). Structuralanalysis of the epitope of the anti-HIV antibody 2F5sheds light into its mechanism of neutralization andHIV fusion. J. Mol. Biol. 330, 1101–1115.

21. Kunert, R., Ruker, F. & Katinger, H. (1998). Molecularcharacterization of five neutralizing anti-HIV type 1antibodies: identification of nonconventional D seg-ments in the human monoclonal antibodies 2G12and 2F5. AIDS Res. Hum. Retroviruses, 14, 1115–1128.

22. Mascola, J. R., Louder, M. K., VanCott, T. C., Sapan,C. V., Lambert, J. S., Muenz, L. R. et al. (1997). Potentand synergistic neutralization of human immuno-deficiency virus (HIV) type 1 primary isolates byhyperimmune anti-HIV immunoglobulin combinedwith monoclonal antibodies 2F5 and 2G12. J. Virol.71, 7198–7206.

23. Muster, T., Guinea, R., Trkola, A., Purtscher, M.,Klima, A., Steindl, F. et al. (1994). Cross-neutralizingactivity against divergent human immunodeficiency

Multispecificity of HIV-1-Neutralizing Antibody 2F5 325

virus type 1 isolates induced by the gp41 sequenceELDKWAS. J. Virol. 68, 4031–4034.

24. Broliden, P. A., von Gegerfelt, A., Clapham, P.,Rosen, J., Fenyo, E. M., Wahren, B. & Broliden, K.(1992). Identification of human neutralization-inducing regions of the human immunodeficiencyvirus type 1 envelope glycoproteins. Proc. Natl Acad.Sci. USA, 89, 461–465.

25. Calarota, S., Jansson, M., Levi, M., Broliden, K.,Libonatti, O., Wigzell, H. & Wahren, B. (1996).Immunodominant glycoprotein 41 epitope identifiedby seroreactivity in HIV type 1-infected individuals.AIDS Res. Hum. Retroviruses, 12, 705–713.

26. Geffin, R. B., Scott, G. B., Melenwick, M., Hutto, C.,Lai, S., Boots, L. J. et al. (1998). Association of anti-body reactivity to ELDKWA, a glycoprotein 41neutralization epitope, with disease progression inchildren perinatally infected with HIV type 1. AIDSRes. Hum. Retroviruses, 14, 579–590.

27. Bonnycastle, L. L., Mehroke, J. S., Rashed, M., Gong,X. & Scott, J. K. (1996). Probing the basis of antibodyreactivity with a panel of constrained peptidelibraries displayed by filamentous phage. J. Mol.Biol. 258, 747–762.

28. Zwick, M. B., Shen, J. & Scott, J. K. (2000). Homo-dimeric peptides displayed by the major coat proteinof filamentous phage. J. Mol. Biol. 300, 307–320.

29. Zwick, M. B., Bonnycastle, L. L., Menendez, A.,Irving, M. B., Barbas, C. F., III, Parren, P. W. et al.(2001). Identification and characterization of apeptide that specifically binds the human, broadlyneutralizing anti-human immunodeficiency virustype 1 antibody b12. J. Virol. 75, 6692–6699.

30. Nieba, L., Krebber, A. & Pluckthun, A. (1996). Com-petition BIAcore for measuring true affinities: largedifferences from values determined from bindingkinetics. Anal. Biochem. 234, 155–165.

31. Zwick, M. B., Bonnycastle, L. L., Noren, K. A.,Venturini, S., Leong, E., Barbas, C. F., III et al. (1998).The maltose-binding protein as a scaffold for mono-valent display of peptides derived from phagelibraries. Anal. Biochem. 264, 87–97.

32. Jouault, T., Fradin, C., Dzierszinski, F., Borg-Von-Zepelin, M., Tomavo, S., Corman, R. et al. (2001).Peptides that mimic Candida albicans-derived beta-1,2-linked mannosides. Glycobiology, 11, 693–701.

33. Zeder-Lutz, G., Hoebeke, J. & Van Regenmortel,M. H. (2001). Differential recognition of epitopespresent on monomeric and oligomeric forms ofgp160 glycoprotein of human immunodeficiencyvirus type 1 by human monoclonal antibodies. Eur.J. Biochem. 268, 2856–2866.

34. Muster, T., Steindl, F., Purtscher, M., Trkola, A.,Klima, A., Himmler, G. et al. (1993). A conservedneutralizing epitope on gp41 of human immuno-deficiency virus type 1. J. Virol. 67, 6642–6647.

35. Parker, C. E., Deterding, L. J., Hager-Braun, C.,Binley, J. M., Schulke, N., Katinger, H. et al. (2001).Fine definition of the epitope on the gp41 glyco-protein of human immunodeficiency virus type 1for the neutralizing monoclonal antibody 2F5.J. Virol. 75, 10906–10911.

36. Sundberg, E. J. & Mariuzza, R. A. (2003). Molecularrecognition in antibody–antigen complexes. InAdvances in Protein Chemistry (Janin, J. & Wodak, S. J.,eds), vol. 61, pp. 119–160, Academic Press, SanDiego.

37. Chan, D. C., Fass, D., Berger, J. M. & Kim, P. S. (1997).

Core structure of gp41 from the HIV envelope glyco-protein. Cell, 89, 263–273.

38. Weissenhorn, W., Dessen, A., Harrison, S. C., Skehel,J. J. & Wiley, D. C. (1997). Atomic structure of theectodomain from HIV-1 gp41. Nature, 387, 426–430.

39. Chan, D. C. & Kim, P. S. (1998). HIV entry and itsinhibition. Cell, 93, 681–684.

40. Salzwedel, K., West, J. T. & Hunter, E. (1999).A conserved tryptophan-rich motif in themembrane-proximal region of the human immuno-deficiency virus type 1 gp41 ectodomain is importantfor Env-mediated fusion and virus infectivity. J. Virol.73, 2469–2480.

41. Schibli, D. J., Montelaro, R. C. & Vogel, H. J. (2001).The membrane-proximal tryptophan-rich region ofthe HIV glycoprotein, gp41, forms a well-definedhelix in dodecylphosphocholine micelles.Biochemistry, 40, 9570–9578.

42. Herrera, C., Spenlehauer, C., Fung, M. S., Burton,D. R., Beddows, S. & Moore, J. P. (2003). Non-neutralizing antibodies to the CD4-binding site onthe gp120 subunit of human immunodeficiencyvirus type 1 do not interfere with the activity of aneutralizing antibody against the same site. J. Virol.77, 1084–1091.

43. Viau, M. & Zouali, M. (2001). Molecular determi-nants of the human antibody response to HIV-1:implications for disease control. J. Clin. Immunol. 21,410–419.

44. Zwick, M. B., Kiyomi Komori, H., Stanfield, R. L.,Wang, M., Parren, P. W., Kunert, R. et al. (2004). Thelong third complementarity-determining region ofthe heavy chain is important in the activity of thebroadly neutralizing anti-HIV-1 antibody 2F5.J. Virol. 78, 3155–3161.

45. Pantophlet, R., Wilson, I. A. & Burton, D. R. (2003).Hyperglycosylated mutants of human immuno-deficiency virus (HIV) type 1 monomeric gp120 asnovel antigens for HIV vaccine design. J. Virol. 77,5889–5901.

46. Beenhouwer, D. O., May, R. J., Valadon, P. & Scharff,M. D. (2002). High affinity mimotope of the poly-saccharide capsule of Cryptococcus neoformansidentified from an evolutionary phage peptidelibrary. J. Immunol. 169, 6992–6999.

47. Zhong, G., Smith, G. P., Berry, J. & Brunham, R. C.(1994). Conformational mimicry of a chlamydialneutralization epitope on filamentous phage. J. Biol.Chem. 269, 24183–24188.

48. Bonnycastle, L. L., Menendez, A. & Scott, J. K. (2001).Production of peptide libraries. In Phage Display: ALaboratory Manual (Barbas, C. F., Burton, D. R., Scott,J. K. & Silverman, G. J., eds), Cold Spring HarborLaboratory Press, New York.

49. Bonnycastle, L. L., Menendez, A. & Scott, J. K. (2001).General phage methods. In Phage Display: ALaboratory Manual (Barbas, C. F., Burton, D. R., Scott,J. K. & Silverman, G. J., eds), Cold Spring HarborLaboratory Press, New York.

50. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989).Molecular Cloning: A Laboratory Manual, Cold SpringHarbor Laboratory Press, New York.

51. Menendez, A., Bonnycastle, L. L., Pan, C. C. O. &Scott, J. K. (2001). Screening peptide libraries. InPhage Display: A Laboratory Manual (Barbas, C. F.,Burton, D. R., Scott, J. K. & Silverman, G. J., eds),Cold Spring Harbor Laboratory Press, New York.

52. Zwick, M. B., Menendez, A., Bonnycastle, L. L. &Scott, J. K. (2001). Analysis of phage-borne peptides.

326 Multispecificity of HIV-1-Neutralizing Antibody 2F5

In Phage Display: A Laboratory Manual (Barbas, C. F.,Burton, D. R., Scott, J. K. & Silverman, G. J., eds),Cold Spring Harbor Laboratory Press, New York.

53. Karlsson, R. (1994). Real-time competitive kineticanalysis of interactions between low-molecular-weight ligands in solution and surface-immobilizedreceptors. Anal. Biochem. 221, 142–151.

54. Richalet-Secordel, P. M., Rauffer-Bruyere, N.,Christensen, L. L., Ofenloch-Haehnle, B., Seidel, C.

& Van Regenmortel, M. H. (1997). Concentrationmeasurement of unpurified proteins using biosensortechnology under conditions of partial masstransport limitation. Anal. Biochem. 249, 165–173.

55. Zeder-Lutz, G., Benito, A. & Van Regenmortel, M. H.(1999). Active concentration measurements of recom-binant biomolecules using biosensor technology.J. Mol. Recogn. 12, 300–309.

Edited by I. Wilson

(Received 1 December 2003; received in revised form 13 February 2004; accepted 19 February 2004)

Multispecificity of HIV-1-Neutralizing Antibody 2F5 327