The role of structure in antibody cross-reactivity between peptides and folded proteins

Article No. mb981907 J. Mol. Biol. (1998) 281, 183±201

The Role of Structure in Antibody Cross-reactivityBetween Peptides and Folded Proteins

Lisa Craig1, Paul C. Sanschagrin2, Annett Rozek1, Steve Lackie3

Leslie A. Kuhn2* and Jamie K. Scott4*

1Department of ChemistryInstitute of Molecular Biologyand Biochemistry, Simon FraserUniversity, Burnaby, BCCanada V5A 1S62Protein Structural Analysisand Design LaboratoryDepartment of BiochemistryMichigan State UniversityEast Lansing, MI 48824, USA3Sapidyne Instruments Inc.967 E. ParkCenter Blvd. 445Boise, ID 83706-6700, USA4Biochemistry Program andDepartment of BiologicalSciences, Institute of MolecularBiology and BiochemistrySimon Fraser UniversityBurnaby, BC, Canada V5A 1S6

E-mail addresses of the correspon

Abbreviations used: Ab, antibodybovine serum albumin; TBS, Tris-buStructural Assignment; ELISA, enzyELISA signal to 50%; RMSD, root-mHIV-1, human immunode®ciency vrotating-frame Overhauser enhanceproportional phase incrementation;acid; F, phenylalanine; G, glycine; Hproline; Q, glutamine; R, arginine; Sacid or a non-critical amino acid.

0022±2836/98/310183±19 $30.00/0

Peptides have the potential for targeting vaccines against pre-speci®edepitopes on folded proteins. When polyclonal antibodies against nativeproteins are used to screen peptide libraries, most of the peptides isolatedalign to linear epitopes on the proteins. The mechanism of cross-reactivityis unclear; both structural mimicry by the peptide and induced ®t of theepitope may occur. The most effective peptide mimics of protein epitopesare likely to be those that best mimic both the chemistry and the struc-ture of epitopes. Our goal in this work has been to establish a strategyfor characterizing epitopes on a folded protein that are candidates forstructural mimicry by peptides. We investigated the chemical and struc-tural bases of peptide-protein cross-reactivity using phage-displayed pep-tide libraries in combination with computational structural analysis.Polyclonal antibodies against the well-characterized antigens, hen egg-white lysozyme and worm myohemerythrin, were used to screen a panelof phage-displayed peptide libraries. Most of the selected peptidesequences aligned to linear epitopes on the corresponding protein; thecritical binding sequence of each epitope was revealed from these align-ments. The structures of the critical sequences as they occur in othernon-homologous proteins were analyzed using the Sequery andSuperpositional Structural Assignment computer programs. Theseallowed us to evaluate the extent of conformational preferenceinherent in each sequence independent of its protein context, andthus to predict the peptides most likely to have structural preferencesthat match their protein epitopes. Evidence for sequences having aclear structural bias emerged for several epitopes, and synthetic pep-tides representing three of these epitopes bound antibody with sub-micromolar af®nities. The strong preference for a type II b-turn pre-dicted for one peptide was con®rmed by NMR and circular dichroismanalyses. Our strategy for identifying conformationally biased epitopesequences provides a new approach to the design of epitope-targeted,peptide-based vaccines.

# 1998 Academic Press

Keywords: antibody cross-reactivity; epitope mimicry; phage-displayedpeptide libraries; peptide structural analysis; peptide vaccines
*Corresponding authors
ding authors: [email protected]; [email protected]

; HEL, hen eggwhite lysozyme; MHr, myohemerythrin; DTT, dithiothreitol; BSA,ffered saline; PDB, Brookhaven Protein Data Bank; SSA, Superpositionalme-linked immunosorbent assay; IC50, inhibitory concentration required to reduceean-square deviation; NMR, nuclear magnetic resonance; CD, circular dichroism;

irus type 1; DQF-COSY, double-quantum-®ltered correlated spectroscopy; ROESY,ment spectroscopy; NOE, nuclear Overhauser enhancement; TPPI, time-single-letter amino acid code: A, alanine; C, cysteine; D, aspartic acid; E, glutamic, histidine; I, isoleucine; K, lysine; L, leucine; M, methionine; N, asparagine; P,, serine; T, threonine; V, valine; W, tryptophan; Y, tyrosine; X, any natural amino

# 1998 Academic Press

mailto:[email protected]; [email protected]

184 Cross-reactivity Between Peptides and Proteins

Introduction

Antibody (Ab) responses against folded proteinsinclude reactivities against linear and discontinu-ous epitopes. Alanine mutagenesis studies (Jin et al.,1992) and other work have shown that the reactiv-ities against linear epitopes represent a small frac-tion of the Ab response compared to those againstdiscontinuous ones. Thus, the linear epitopes ident-i®ed by mapping studies using synthetic peptides(Milton & Van Regenmortel, 1979; Atassi 1984;Geysen et al., 1987) and phage-displayed peptidelibraries (Folgori et al., 1994; Yao et al., 1995;Bonnycastle et al., 1996), though minor, are signi®-cant in that they are the sites that typically cross-react well with peptides. Since the idea of peptide-based vaccines was ®rst proposed (Lerner, 1982),much effort has focused on identifying peptidesthat will elicit Abs that cross-react with pre-speci-®ed sites on a target protein. Although a numberof such peptides have been reported (Arnon et al.,1971; Francis et al., 1987; Hastings et al., 1990;Javaherian et al., 1990; Keller et al., 1993; White-Scharf et al., 1993; van der Werf et al., 1994; Mottiet al., 1994; Robinson et al., 1995; Lundkvist et al.,1996), it remains dif®cult to predict the epitopes onfolded proteins that will make effective targets forpeptide-based vaccines. More often than not, Absraised against peptides bind with high af®nity tothe peptide, but only weakly to the folded proteinfrom which they were derived (Jemmerson, 1987).

The linear epitopes identi®ed by cross-reactivitywith short peptides are characterized by de®nedstructures of relatively high ¯exibility (Tainer et al.,1984; Westhof et al., 1984), surface accessibility(Atassi & Lee, 1978; Milton & Van Regenmortel,1979) and hydrophilicity (Hopp & Woods, 1981);these features usually correspond to turns andloops in folded proteins (Westhof et al., 1984).These parameters have been used with some suc-cess to predict the regions of proteins that willcross-react with peptides (Thornton et al., 1986;Pellequer et al., 1991). Presumably, shape comple-mentarity between an epitope and the antigen-binding site of an Ab is more readily induced inmobile accessible regions of a protein than in morerigid ones; however, as yet there are no structuresof Abs in complex with linear epitopes available tosupport this. Evidence for an induced-®t bindingmechanism for linear epitopes has been obtainedby Getzoff et al. (1987), who identi®ed the criticalbinding residues of several cross-reactive epitopesusing peptides bearing single amino acid replace-ments. Several epitopes contained critical residuesthat were buried in the native protein, suggestingthat the structure of these epitopes was alteredupon binding Ab. Thus, epitope ¯exibility hasfunctional signi®cance in peptide cross-reactivity.

While epitope ¯exibility is clearly involved inpeptide cross-reactivity, the role of conformationalbias has also been recognized (Dyson et al., 1988a).Although most short peptides possess considerableconformational ¯exibility in aqueous solution,

sequences have been identi®ed that confer structur-al stability to a peptide, as assessed by nuclearmagnetic resonance (NMR) and circular dichroism(CD) analyses (Dyson et al., 1988b). Such sequencesare expected to form largely the same structures inthe context of a folded protein or a peptide, there-by allowing antigen cross-reactivity. In support ofthis, conformational preferences for turns andhelices in solution have been demonstrated in anumber of antigenic peptides (Dyson & Wright,1995). Structural elements that are present in pep-tides in solution have also been demonstrated inthe corresponding folded protein (Wien et al.,1995), and in peptide-Ab complexes (Tsang et al.,1992; Rini et al., 1993; Ghiara et al., 1994; Campbellet al., 1997). The role of structural stability in pep-tide mimicry of protein epitopes is further illus-trated by studies demonstrating improved cross-reactivity (Luzzago et al., 1993; Hoess et al., 1994)and the production of higher-af®nity, protein-cross-reactive Abs (Arnon et al., 1971; Satterthwaitet al., 1989; Zhong et al., 1994; Cuniasse et al., 1995)when disul®de and other covalent constraints areincorporated into peptides. Thus, both structuralmimicry by a peptide and ¯exibility of the corre-sponding epitope contribute to cross-reactivity andthe production of cross-reactive Abs.

The computational strategy, Sequery, was devel-oped as a statistical screen of the Protein DataBank (PDB; Bernstein et al., 1977; Abola et al.,1987), to determine whether an amino acidsequence, or sequence pattern, possesses a localconformational preference independent of its pro-tein context. Sequery has been successfully appliedto model protein-targeting motifs (Collawn et al.,1991; Chang et al., 1993) as con®rmed by NMR(Bansal & Gierasch, 1991; Eberle et al., 1991), andto model loop structures for crystallographic ®tting(Parge et al., 1993) and homology modeling (Fisheret al., 1994). We employed Sequery, and an acces-sory program, Superpositional Structural Assign-ment (SSA), to identify conformational preferencesin linear epitope sequences that are known tocross-react with peptides. Peptides bearing theseconformationally preferred epitope sequences areexpected to mimic the epitope structure, thus facili-tating recognition by anti-protein Abs. The cross-reactive epitopes used for this analysis were ident-i®ed by screening a panel of phage-displayed pep-tide libraries with polyclonal Abs against themodel proteins hen eggwhite lysozyme (HEL) andworm myohemerythrin (MHr). For several of theepitope sequences, structural preferences emergedthat matched the structures of the correspondingepitopes on HEL or MHr. An epitope sequencebearing a strong Sequery-predicted bias for a typeII b-turn was shown to have the same well-de®nedturn preference when analyzed as a free peptide insolution by NMR and CD spectroscopy. Moreover,synthetic peptides predicted to bear structurallybiased sequences were shown to bind Ab withsub-micromolar af®nities. Thus, we have shownthat short, cross-reactive sequences on linear epi-

Cross-reactivity Between Peptides and Proteins 185

topes have conformational preferences that allowpeptides to functionally mimic epitope structures.The ability to identify, from a set of cross-reactiveepitopes, those that have the highest structural pro-pensities, should prove valuable in the design ofpeptide-based vaccines.

Results

Cross-reactive peptides reveal linear epitopeson native HEL and MHr

To identify peptides that bind Abs againstfolded proteins, we screened a panel of linear andconformationally biased peptide libraries withpolyclonal anti-HEL and anti-MHr Abs. The ran-dom peptides in the majority of the libraries werefused to the N terminus of the major coat proteinpVIII of ®lamentous phage, whereas the peptidesin two of the libraries were fused to the minor coatprotein pIII. Three to four rounds of screeningwere performed on the library panel, and ®ve to20 clones were analyzed from each of the phagepools showing high enrichment and/or strongELISA signals. As shown in Table 1A and B, con-sensus sequences were revealed by aligning similarpeptide sequences into consensus groups. Linearepitopes on HEL and MHr were subsequentlyidenti®ed by aligning consensus sequences to theamino acid sequence of the corresponding antigen.From this, ®ve epitopes comprising the followingsequences were identi®ed on each antigen: DVQ,DITASV, NAWV, RTPGS, and TQATNR on HEL,and KYSEV, PEPYV, WDESF, GLSAPVDA, andAPNLA on MHr. Two ``sub-epitopes'' were ident-i®ed for the KYSEV epitope; KYSE from the pVIIIlibrary screening and YSEV from the pIII libraryscreening. This probably re¯ects small differencesin the anti-MHr response, as the pIII and pVIIIlibraries were screened with anti-MHr Abs fromdifferent rabbits (see Materials and Methods).

In a few cases, the alignment of a peptide to asingle consensus group was not clear; either resi-due patterns common to two consensus groupswere present, or the peptide had only two residuesin common with a consensus group. Clone-against-clone competition ELISAs were performed todetermine correct alignments; a peptide wasassigned to a given consensus group if its corre-sponding phage clone could inhibit Ab binding toan immobilized clone whose displayed peptidealigned well to this group. The degree to whichselected phage clones inhibited Ab-binding torepresentative immobilized phage is shown inTable 1A and B. This provided a functional assayfor assigning peptides to a consensus group.

For peptides containing multiple cysteine resi-dues, we investigated whether constraints imposedby disul®de bridging were required for Ab-bind-ing. Reducing ELISAs were performed, in whichbinding to immobilized clones was assayed in thepresence of 5 mM dithiothreitol (DTT), afterreduction of disul®de bridges with 15 mM DTT.

The level of inhibition by DTT is shown inTable 1A and B. Peptides in the DVQ and KYSEVconsensus groups were most dramatically affectedby DTT reduction; binding of Ab was inhibitedconsiderably for clones containing multiplecysteine residues. In both these groups, however,at least one strong-binding clone was derived froman unconstrained library, suggesting that therequirement for disul®de-bridging depends on thesequence of the peptide. Only the weak-bindingclones in the DITASV group were affected, and lit-tle or no effect was seen on peptides in the PEPYVgroup. Thus, the presence of disul®de bonds cancontribute to binding for many of the peptides, butis not an absolute requirement for cross-reactivitywith any of the epitopes identi®ed.

To ensure that the cross-reactive peptides werenot isolated by minor Ab reactivities againstdenatured or proteolyzed forms of the HEL andMHr antigens, competition ELISAs were per-formed using HEL and MHr in solution. Both anti-gens were in their native, folded form, as each oftheir electrophoretic mobilities corresponded tosingle, discrete bands on a non-denaturing poly-acrylamide gel. Table 1A and B shows that bindingto the majority of clones was strongly inhibited(575% inhibition) by their cognate antigen. Weakinhibition (less than 25%) was observed for a fewclones, particularly those having relatively highaf®nities. In these cases, the Abs may have boundmore tightly to the phage clone than to cognateantigen. The alternative, that these clones wereselected by Abs speci®c for the unfolded protein, isunlikely, since other clones in the same consensusgroup (and in the same assays) were inhibited bynative antigen.

Binding of synthetic peptides to anti-HEL andanti-MHr Abs

To determine if free peptides behave similarlyto the phage-displayed peptides, several peptideswere synthesized and tested for their ability tobind Ab. Four biotinylated peptides, each derivedfrom a tight-binding clone in a major consensusgroup, were synthesized as follows: Cys6-7 fromthe DVQ group; Cys5-1 from the DITASV group,X6-1 from the RTPGS group and X6-14 from thePEPYV group (Table 1A and B; see Materials andMethods for peptide sequences). All of the pep-tides were shown to bind strongly to Ab bydirect ELISA and to inhibit binding of Ab to thecorresponding clones in a dose-dependent man-ner (data not shown). Average equilibrium dis-sociation constants (Kd values) were obtained foreach of the four peptides using the KinExA auto-mated immunoassay instrument (Sapidyne Instru-ments, Boise ID). The KinExA instrument is acomputer-controlled ¯ow spectro¯uorimeterdesigned to achieve the rapid separation andquanti®cation of uncomplexed Ab present in reac-tion mixtures of Ab, antigen, and Ab-antigencomplexes (Blake et al., 1997). As shown in


Table 2, all four peptides bound Ab with sub-micromolar Kd values. Binding curves for each ofthe four peptides are shown in Figure S1 ofSupplementary Material. Taken together, the dataindicate that the Abs bind with moderate af®nitesto these peptides, whether they are displayed aspVIII-fusions on phage, or are free in solution.

Assignment of critical epitopes on HELand MHr

Each epitope on HEL and MHr was mapped bymatching conserved residues in each consensussequence group to a sequence within the cognateprotein antigen. The relative importance of eachresidue for cross-reactivity was further deducedfrom its pattern of amino acid replaceability withinthe consensus sequence group. Residues that wereirreplaceable, or could only be conservativelyreplaced, were considered ``critical binding resi-

Table 1. Consensus groups and alignments

HEL epitopeand aligning Peptide sequence andphage clonesa alignment with HEL

A. Consensus groups, sequence alignment with HEL, direct ELISA and inHEL (113-127) ****NRCKGTDVQAWIRGC****e

X15-5 HDVQRELIQLQRYMPCys6-7 VNIACNPTDIQCLIRLCys6-6 GPYTCTPNDIQCRVRICys6-5 NDVQCRLRSHQCETMRCys4-5 ERMSCDSQRCQKAMCys6-3 GNSHCDANDVQCNTHNLX6-5 DCFDDVQRCVX30-6 QPQSSSGRPQDVQSIARE(Z13)

f

X30-1 ERDVQAHMRPH(Z19)Critical sequence DVQ

HEL (83-95) ****LLSSDITASVNC****Cys5-1 NQPSCQDITACLTPQLX6-2 QCHADITKCLLX6-3 ACNADITSCLLX8-2 QCAPDVTNSACFLX6-1 TCPDITRCLX8CX8-6 TDCNDITACFGDHFQIMCys6-1 MNDNCLGGDITCTFSQCys5-2 VEPTCADATACLSRRX30-2 (Z17)DVADITTFLHPGSCritical sequence DITXXV

HEL (104-111) ****GMNAWVAW****X30-7 NAWAELTFGPYHIPSDCX30-3 NSWA(Z13)HAPGSFF(Z6)X6-5 NAWYLRX6-2 NAWATHX8CX8-1 NSWHDHVSCVAMCLRPPCritical sequence NAW

HEL (65-74) ****NDGRTPGSRN****X6-1 SKTPGAX6-4 GRFPGSCritical sequence RXPGS

HEL (38-50) ****FNTQATNRNTDGS****LX8-1 SCVTQASNRGCLCys4-1 MTEPCMGQYCNRFRX8CX8-2 MATNRVYGCQHSMYARSCys3-4 SRNDRCTDLCSAVK

dues''. These residues are likely to contribute sig-ni®cantly to binding energy by interacting directlywith the Ab combining site, and/or by contribut-ing to the overall structure of the epitope (Getzoffet al., 1987). Based on this analysis, the criticalbinding sequences, or ``critical epitopes'', for HELand MHr are DVQ, DITXXV, NAW, RXPGS, KYSE(pVIII), YSEV (pIII), EPYV and WDXSF, in whichX represents a non-critical residue in the sequence.Table 1A and B shows the critical sequences for theepitopes whose consensus groups contain at leastfour peptides.

Figure 1(a) and (b) shows the locations of the lin-ear epitopes on the HEL and MHr structures,respectively. Each of the epitopes is well exposedon the surface of the protein, with the exception ofNAW, which lies on the surface of HEL, but at theopening of the substrate-binding cleft. Most of theepitopes comprise turns or loops between regionsof regular secondary structure. In HEL, the RTPGS

Inhibition ELISAs-level of inhibitionc

in the presence of:Direct In-solution 5 mM 5 mM

ELISAb phage cloned DTT HEL

hibition ELISA dataCys6-5

0.649 � ÿ ÿ0.602 �� 0.543 �� 0.274 �� 0.225 �� 0.179 �� 0.174 �� 0.186 ��0.131 ��

Cys6-10.530 �� 0.479 �� 0.447 � ��0.377 �� 0.349 ÿ ��0.292 �� 0.264 �� 0.216 �� 0.102 ��

X8CX8-10.337 �� 0.254 �� 0.203 ��0.200 ��0.111 �� ÿ ��

X6-40.254 �� 0.197 ��

X8CX8-20.318 �� 0.208 ÿ �� 0.133 �� ÿ ��0.100 ��

Table 1 (continued)

Inhibition ELISAs-level of inhibitionc

MHr epitope in the presence of:and aligning Peptide sequence and Direct In-solution 5 mM 5 mMphage clonesa alignment with MHr ELISAb phage cloned DTT MHr

B. Consensus groups, sequence alignment with MHr, direct ELISA and inhibition ELISA dataMHr (63-74) ****DAAKYSEVVPHK****e X15-9X15-5 CPTSSLWMDAKSSDL 0.602 � ��X8CX8-1 QYCKYSETCIQSATAAH 0.317 �� ÿX15-9 TSGMAHDKYSDSPIW 0.270 �� ÿ ��X8CX8-13 TPAKYSETCVPVTMSTL 0.236 � ��LX6-12 DCKYSGSPCV 0.175 �� X15-3 ETSKYSTPIVRBBBTg 0.134 ��X8CX8-9 APSGDSDSCGPSGKWSC 0.131 �� Cys4-4 EKNNCLKWSCAAPM 0.101 ��Critical sequence KYSE

pIII-X6:1 FRYSEV 0.404pIII-X6:8 WKYSEV 0.378pIII-X15:28 GARIMYSEVPGYLAV 0.342pIII-X6:3 FMYSEV 0.247pIII-X6:16 PFYSEV 0.181pIII-X15:26 PGFAYSEAHPFASLP 0.135pIII-X6:4 YSEFMY 0.107Critical sequence YSEV

MHr (2-11) ****WEIPEPYVWD**** X6-8X6-14 QEPYVL 0.906 �� ÿ ��LX8-2 DCPVGAEPYVCL 0.739 ÿ ��X6-8 SDVEPY 0.714 �� ÿ ��X15-11 FENYTQLGIPEPYVL 0.706 �� ÿ ��X15-13 SGQSSPELYVARAEY 0.681 �� ÿ ��LX6-7 TCEPYIIKCT 0.661 ÿ ��LX8-1 HCLPSQEMYTCM 0.629 � ��LX6-1 FCEPYVGNCL 0.618 ÿ ��Cys6-1 MKTACTGEKYTCVPTI 0.522 �� aCT-2 SCCTQPCEAYVTC 0.490 � ��X6-3 VPEPYA 0.447 ÿ ��X6-10 VPEKYV 0.446 ��pIII-X6:7 PEAYVI 0.148Critical sequence EPYV

MHr (8-18) ****YVWDESFRVFY****pIII-X15:30 GSSTFDASFFWPCCD 0.564 ÿ ��pIII-X15:36 GKYDLSFTMVVSHS 0.422pIII-X15:38 LAFDASFSFTVSS 0.347pIII-X1523 PPWDISFSVLGSDLL 0.111Critical sequence WDXSF

MHr (84-94) ****IGGLSAPVDAKN****pIII-X15:27 DLSGLWAPGGTDHLN 0.408 ÿ ��pIII-X6:4 PGLLSALCAPGLA 0.204

MHr (35-47) ****CIRDNSAPNLATL****pIII-X15:24 PGLLSALCAPGLA 0.204

a Phage clones from the pIII-displayed peptide libraries are indicated; all others are from the pVIII libraries. The pIII sequences arepreceded by the N-terminal sequence ADGA.

b ELISA signals are reported in absorbance units, A405-A490 minus the background signal from f88 or fd-tet control phage.c Level of inhibition of Ab binding to immobilized clones produced by in-solution phage clones, free HEL or MHr, or 5 mM DTT:� � � , 75 to 100% inhibition; � � , 50 to 74%; �, 25 to 49%; ÿ, <24%.

d Level of inhibition of binding by 2 � 1010 phage clones in solution.e Underlined and bold-faced amino acids, critical binding residues; boldfaced (but not underlined) amino acids align with the cor-

responding HEL or MHr sequence or with other peptides in the consensus group, but are not critical, as determined by incompleteconservation within the consensus group.

f Z, Non-aligning residues in peptides from the X30 libraries.g B, Residues not determined.


epitope forms a type II b-turn, DVQ and DITASVboth occur at the beginning of a-helices, and NAWforms part of a loop. In MHr, EPYV and WDESFcomprise type VI and type I b-turns, respectively,in the N-terminal loop; KYSEV and GLSAPVDA

are located in loops connecting helices, whereGLSAPVDA contains a type VI b-turn, andAPNLA is at the beginning of a helix. Of all ofthese epitopes, only the TQATNR epitope in HELis extended, forming a b-strand. Most of the epi-

Table 2. Average binding af®nities and error limits for synthetic peptides

Error boundsb

Antibodya Peptide Kd (nM) Kd low (nM) Kd high (nM)

Anti-HEL Cys5-1 38.1 23.7 47.0Anti-HEL Cys6-7 64.5 36.4 104Anti-HEL X6-1 234 120 455Anti-MHr X6-14 780 586 1038

a Initial Ab concentrations were as follows: 20 nM anti-HEL for Cys5, 5 nM anti-HEL forCys6-7, 50 nM anti-HEL for X6-1, and 100 nM anti-MHr for X6-14.

b The error bounds represent a 95% con®dence interval (similar to �2 standard devia-tions) in the Kd estimate, and are a function of best ®t error of the theoretical binding curveto the data.


topes reside in regions of relatively high mobilityand surface accessibility (data not shown; seeThornton et al., 1986). Flexibility and accessibilityare features typical of protein epitopes that cross-react with peptides (Westhof et al., 1984; Taineret al., 1984; Thornton et al., 1986; Geysen et al.,1987).

Evidence for structural mimicry of proteinepitopes by peptides

The sequences of critical epitopes were analyzedin three phases to assess the structural basis formimicry by the peptides. In the ®rst phase, theamino acid sequences of 493 non-redundant pro-tein structures in the PDB (Hobohm et al., 1992; seeMaterials and Methods) were scanned for occur-rences of all or part of each critical epitopesequence. Four-residue sequence patterns were

Figure 1. Ribbon structures of HEL and MHr showing(a) HEL, PDB code 193l (M. C. Vaney, S. Maignan, M. Riecode 2mhr (Sheriff et al., 1987). Epitopes are colored yellow

used, each comprising three critical binding resi-dues and one variable residue, since exact matchesto more than three residues are rarely found in thePDB. Thus, for example, the critical epitope RXPGSwas analyzed using the sequence patterns[R,K,H]XPG, XPGS and PGSX. Sequery provided alist of all sequence matches, or analogs, for eachsequence pattern. The epitopes TQATNR on HELand GLSAPVDA and APNLA on MHr were notincluded in the Sequery analysis, since only a fewpeptides aligned to each of these. At least onesequence pattern for each of the other epitopesyielded a substantial number of sequence analogsfrom the Sequery search. Sequence patterns yield-ing less than ten analogs were not considered tohave given a suf®cient statistical sample, and thuswere not analyzed further.

In the second phase of the analysis, the structureof each sequence analog was compared with that

the linear epitopes identi®ed by peptide cross-reactivity.s-Kautt & A. Ducruix, unpublished results). (b) MHr, PDB.

Table 3. Summary of structural features for the HEL and MHr epitopes and the corresponding sequence patterns

Critical epitopeon HEL or MHr

Frequency ofanalogs matching

the epitopea

Overallconformational

preferencebEpitope

structure

Hydrogen-bondingpattern within the

critical epitopeSurface accessibility of

critical residuesa

DVQ DVQX DVQX: helix Beginning of None Exposed: D119, Q121HEL: 119-121 45% (9/20) 45% (9/20) helix (irreg.) Buried: V120

RXPGS XPGS XPGS: type II TPGS: type II S72:Og! T69:O Exposed: R68, P70, G71HEL: 68-72 39% (13/33) b-turn b-turn Buried: S72

39% (13/33)

DITXXV DITX DITX: irreg. Beginning of T89:N! D87:Od1 Exposed: D87HEL: 87-92 25% (5/20) 70% (14/20) helix (irreg.) T89:Og! D87:Od1 Intermediate: T89

Buried: I88, V91

WDXSF DXSF DXSF: helix DESF: type I S13:N! D11:O Intermediate: D11, S13MHr: 10-14 29% (5/17) 29% (5/17) b-turn S13:N! D11:Od1 Buried: W10, F14

S13:Og! D11:Od1

F14:N! D11:O

NAW NA/SWX NA/SWX: helix NAWA: irreg. None Intermediate: N106, A107HEL: 106-108 0/20 50% (10/20) loop Buried: W108

EPYV EXYV EXYV: helix EPY: positions V9:N! P7:O Exposed: E6MHr: 6-9 0/12 58% (7/12) 2, 3 and 4 of a Intermediate: P7, V9

type VI b-turn Buried: Y8

a Number of sequence analogs with structures that superimpose on the corresponding HEL or MHr epitope with an RMSD 40.75 AÊ

divided by the total number of sequence analogs for a given sequence pattern. Sequence analogs from proteins whose crystal struc-tures were not well-resolved (52.6 AÊ resolution) or whose backbones could not be superimposed due to missing atoms in the crys-tal structures, were not included in the analysis.

b An overall conformational preference was assigned to the sequence pattern if 525% of the analogs shared the same secondarystructure. When this did not occur, the percentage of irregular structure is indicated.

c Surface accessibility (Kabsch & Sander, 1983): buried, 0 to 40 AÊ 2; intermediate, 41 to 80 AÊ 2; exposed, >80 AÊ 2.


of the corresponding four-residue sequence withinthe appropriate epitope on HEL or MHr. The pro-gram SSA compared these structures by superim-posing their backbone atoms and calculating theroot-mean-square deviation (RMSD) between thepositions of corresponding atoms in the two back-bones. The structure of a given analog was con-sidered to closely resemble an epitope if thesuperpositional RMSD between the two backboneswas 40.75 AÊ . A sequence pattern was consideredto have a conformational preference for the epitopestructure if at least 25% of the analogs superim-posed well on the epitope.

Based on this analysis, conformational prefer-ences matching the corresponding epitope struc-tures were detected for the sequence patterns,DVQX, DITX, XPGS, and DXSF as shown inTable S1 in Supplementary Material. The frequencyof analogs matching the structure of the corre-sponding epitopes is indicated in Table 3; thosewith the highest frequency of superposition areassumed to have the strongest preference for theepitope conformation. These analogs are shownsuperimposed on their corresponding epitopes inFigure 2(a) to (d). In each case, conformational pre-ference was highly dependent on the position ofthe variable residue, X, within the sequence pat-tern. For instance, nine of the 20 DVQX analogssuperimposed well on the DVQA epitope in HEL,yet none of the XDVQ analogs superimposed wellon the TDVQ site. This indicates that the confor-mation imposed by the critical binding sequenceDVQ involves a fourth residue at its C terminusand not its N terminus. The fact that each critical

sequence occurs in the same structure in a numberof non-homologous proteins suggests that localconformation is de®ned by local interactionswithin these sequences, rather than by tertiaryinteractions with the rest of the protein. Cross-reac-tive peptides bearing these sequences are likely tomimic the structures of the corresponding proteinepitopes. Sequence patterns for the remaining epi-topes did not yield analogs that superimposed wellon the epitope structures.

In the third phase of the structural analysis, SSAwas used to classify the structures of each epitopesegment and the corresponding sequence analogsas turn, helix, extended (b-strand) or irregular,based on their superposition with model tetrapep-tides having ideal secondary structures. An overallconformational preference for each sequence pat-tern was assigned if at least 25% of the analogspossessed the same structure. These data are sum-marized in Table 3 for the DVQ, RXPGS, WDXSFand DITXXV epitopes. Both the DVQX and DXSFpatterns have helical conformational preferences,and the XPGS pattern has a type II b-turn prefer-ence. The conformational preference of the DITXpattern was de®ned as irregular, since neither theDITASV epitope, nor any of its analogs, superim-posed well on any of the model tetrapeptides. Yet,®ve of the 20 DITX analogs superimpose well onDITA in HEL, and in all of these structures, thebackbone of the aspartic acid is in an extendedconformation, with the remainder of the sequenceforming part of a turn (Figure 2(b)). Hence, theconformational preference of all four sequence pat-terns is considered turn-like. Helical conformation-

Figure 2. Superposition of sequence analogs on the cor-responding epitopes. Only those analogs that superim-pose on the epitope with an RMSD 40.75 AÊ are shown.


al preferences were also observed for the sequencepatterns NAWX and EXYV (see Table 3). Yet theseanalogs did not superimpose well on their corre-sponding epitopes, and hence, were not investi-gated further as candidates for structural mimicry.

All of the epitope sequences that we analyzedwere identi®ed by cross-reactivity with peptides; anumber of these showed a structural propensity.As a comparison, we wanted to determine theextent to which randomly chosen sequences wouldalso show a structural propensity. A set ofsequence patterns was randomly generated, andSequery was used to identify sequence analogsfrom the list of non-redundant protein structures inthe PDB. Ten of the sequence patterns yielded ana-log sets containing ten or more matches, as shownin Table S2 of Supplementary Material. Next, apseudo-epitope was randomly chosen from eachanalog set, with the only requirement being that itbe surface-exposed within its protein structure.SSA was used to compare the structure of eachpseudo-epitope with its corresponding sequenceanalogs, and to assign secondary structures, as wasdone with the Ab-selected epitope sequencesshown in Table 3. As with the Ab-selectedsequence patterns, several of the randomly gener-ated sequence patterns showed a clear propensityfor a regular secondary structure. Furthermore, ifthe structure of a pseudo-epitope matched that ofthe overall structural propensity of the sequencepattern, a high proportion of the analogs alsosuperimposed well on the pseudo-epitope. Thus,this analysis showed that the structural propensi-ties of the epitopes we studied are similar to thoseof randomly chosen surface-exposed sites on pro-teins in general, and are not unique to peptide-cross-reactive sequences. Importantly, analysis bySequery in combination with SSA has allowed usto target, from a set of peptide-cross-reactive epi-topes, those whose structures are more likely to bemimicked by peptides.

The molecular basis ofconformational preferences

To understand the basis of locally induced struc-tural preferences, we analyzed the inter-residue

The epitope tetramers are shown in yellow, thesequence analogs are shown in magenta for the HELepitopes, and in red for the MHr epitope. Hydrogenbonds are indicated in white. (a) DVQX analogs super-imposed on the DVQA epitope; PDB codes for the pro-teins bearing the sequence analogs are 1pox, 1pea, 2dnj,7rsa, 1cns, 1smn, 1ash, 1ukz and 1qpg. (b) DITX analogssuperimposed on the DITA epitope; PDB codes, 3cla,1dpb, 8cat, 1rcb and 1cpt. (c) XPGS analogs superim-posed on the TPGS epitope; PDB codes, 1slt, 1arv, 1mhl,1cdo, 1fuj, 1pya, 1qpg, 1hxn, 1abr, 1eft, 2dnj, 8acn and1arb. (d) DXSF analogs superimposed on the DESF epi-tope; PDB codes, 1pvd, 1gsa, 1gpc, 2ctc and 3pmg.

Figure 3. The HN-Ha,b region of the 1H ROESY spectrumof Ac-TPGS-NH2 at 283 K showing intra- and inter-resi-due cross-peaks. The labels at the top of the Figure referto the HN resonances, whereas the labels near the cross-peaks indicate the Ha or Hb resonances.


interactions within each epitope and within thecorresponding structural matches. These inter-actions are summarized in Table 3, along withother relevant structural features. The epitopes andtheir analogs comprise turn-like conformations,which allow more stabilizing interactions withinthe tetramer than would occur in extended confor-mations. Speci®c hydrogen bonds appear to beimportant in stabilizing the structures of theDITASV, RTPGS, and WDESF epitopes and theiranalogs (Table 3 and Figure 2(b), (c) and (d)). TheDITASV epitope and all ®ve of its structuralmatches have a hydrogen bond between the Og ofthreonine and the Od1 of aspartic acid, as well asother hydrogen bonds; these interactions probablystabilize the helical-turn structure. Several of thehydrogen bonds present in the WDESF epitope arealso seen in its structural matches. Finally, thehydrogen bond between the Og of serine andthe carbonyl oxygen of threonine is present in theRTPGS epitope and in ten of the 13 XPGS analogs.Commonly, b-turns are stabilized by a hydrogenbond between the carbonyl oxygen at position 1and the amide at position 4 of the turn(Venkatachalam, 1968). Atypical hydrogen-bond-ing in the XPGS analogs resulted in several ofthem being classi®ed as irregular turns, since theirstructures are more open than classical type IIb-turns. In addition to stabilization from hydrogenbonds, the type II b-turn preference suggested forthe XPGS sequence pattern is promoted by prolineand glycine, which often occur at positions 2 and3, respectively, of type II turns (Wilmot &Thornton, 1988; Richardson & Richardson, 1990).Proline produces a kink in the polypeptide chain atposition 2, due to its ring closure, and glycine isfavored in position 3 because it lacks a b-carbonand therefore does not sterically interfere with thecarbonyl oxygen of proline. These interactionslikely contribute to the stability of the protein epi-topes and to preferred structures for the cross-reac-tive peptides.

TPGS has a type II turn conformationalpreference in solution

We predicted that peptides bearing confor-mationally biased sequences would have pre-ferred structures in solution. To test this, thepeptide Ac-TPGS-NH2 was synthesized, and itsstructure was analyzed using NMR and CDspectroscopy. The TPGS sequence was selectedfor structural analysis over DVQA, DITA andDESF for several reasons: (i) the type II b-turnappeared to be a strong conformational propen-sity, with 39% of the XPGS sequence analogspossessing this structure; (ii) the TPGS turn inHEL and all of the XPGS analogs are stabilizedby a hydrogen bond; and (iii) the peptides inthe RXPGS consensus group were short and didnot contain cysteine residues, suggesting that¯anking residues were not necessary for stabiliz-ation of structure. The TPGS sequence was syn-

thesized as a blocked peptide in order toneutralize the termini and better mimic the natu-ral state of this sequence within a protein or alonger peptide.

The structure of Ac-TPGS-NH2 in aqueous sol-ution was analyzed by 1H NMR spectroscopy.Since the amino acid residues were unique, andsequential assignments followed unambiguouslyfrom spin system assignments, the proton reson-ances of Ac-TPGS-NH2 were assigned directlyfrom the DQF-COSY spectrum. Proton resonancesare summarized in Table S3 of SupplementaryMaterial. The trans conformation of the peptidewas identi®ed by the presence of strong cross-peaks between Thr1 Ha and Pro2 Hd in the ROESYspectrum; no evidence for the cis conformation wasfound. The ROESY spectrum in Figure 3 shows astrong ROE between Pro2 Ha and Gly3 HN. Inaddition, a medium ROE was observed betweenGly3 HN and Ser4 HN (not shown), indicating asigni®cant population of type II b-turn (Wagneret al., 1986). The weak connectivity between Pro2Ha and Ser4 HN, which is also expected for type IIb-turns, was not observed; this may be due tooverlap of the Ha resonances of Pro2 and Ser4. Thetemperature coef®cient of Ser4 HN is ÿ6.5 ppb/K,which is smaller than the temperature coef®cientobserved for Thr1 HN and Gly3 HN, ÿ9.7 and ÿ9.3ppb/K, respectively; this suggests that the Ser4backbone amide proton participates in an intra-molecular hydrogen bond. Structural calculationswere performed on the unblocked TPGS peptide.An ensemble of 50 peptide structures was calcu-

Figure 4. Calculated average structure of the TPGS pep-tide. Interproton distances (arrows) and the hydrogenbond distance (wavy line) between the serine amide andthe threonine carbonyl oxygen are shown in AÊ . Thebackbone torsion angles of Pro2 and Gly3 are indicatedin degrees.

Figure 5. CD spectrum and pure component curves ofAc-TPGS-NH2 in aqueous solution at 278 K. ObservedCD spectrum (triangles); pure component curve 1 (cir-cles) re¯ecting a type II b-turn; and pure componentcurve 2 (squares) re¯ecting an atypical or random sec-ondary structural element.


lated using 17 ROE-based distance restraints andone hydrogen-bond distance restraint between Ser4and Thr1, which was derived from the loweredSer4 HN temperature coef®cient (see Table S4, Sup-plementary Material). The distance restraint viola-tions in the calculated structures were below 0.1 AÊ .The RMSD of all calculated structures to the aver-age conformation was 0.61 AÊ for backbone atomsand 1.12 AÊ for all atoms. The backbone torsionangles of the average structure, f2 � ÿ66�,c2 � �138� and f3 � �74�, c3 � �3� correlate wellwith those of an ideal type II b-turn, which has tor-sion angles of f2 � ÿ60�, c2 � �120� andf3 � �90�, c3 � 0 (Richardson, 1981). The back-bone RMSD for the superposition of the calculatedTPGS average structure and the TPGS epitope inHEL (residues 69 to 72) was 0.44 AÊ , indicatingclose structural similarity for the two backbones.A ball-and-stick model of the calculated averageconformation is shown in Figure 4, showing theinterproton distances and the hydrogen bondbetween the serine amide and the threonine carbo-nyl oxygen. A hydrogen bond between the serinehydroxyl and the threonine carbonyl, as is seen inthe TPGS epitope (Figure 2(c)), was not detected inthe NMR spectra.

CD analysis of Ac-TPGS-NH2 was performedto corroborate the NMR ®ndings. The CD spec-trum of Ac-TPGS-NH2 in water, shown inFigure 5, is highlighted by two minima at200 nm and 223 nm. Deconvolution of the CDspectrum by convex constraint analysis (Perczelet al., 1991) yielded two pure component curvesrepresenting 55% (curve 1) and 35% (curve 2) ofthe peptide conformation. Curve 1 has a maxi-mum at 203 nm characteristic of a type II b-turnconformation, whereas curve 2 has a minimum at199 nm and was assigned as atypical or randomsecondary structure (Hollosi et al., 1987; Perczelet al., 1993). Thus, the NMR and CD results forthe Ac-TPGS-NH2 peptide are consistent with atype II b-turn, as predicted by Sequery/SSA andobserved in the HEL epitope.

Discussion

Peptide library screening in combination withcomputational and structural analysesidentifies epitope sequences having inherentstructural propensities

The goal of this work was to establish a meansof identifying immunogenic sites on proteins thatcan be structurally mimicked by peptides. Wecharacterized a set of linear epitopes on HEL andMHr based on their cross-reactivity with phage-displayed peptides. The critical binding sequenceswithin the epitopes were de®ned from the patternsof conserved residues within the selected peptides,and were assessed for their degree of intrinsicstructural bias using the Sequery and SSA compu-ter algorithms. Turn-like structures for thesequences DVQX, XPGS, DITX, and DXSF wereconserved between each epitope and a number ofnon-homologous proteins, suggesting that theselinear epitopes have inherent conformational pre-ferences. Analysis of the structures of the epitopesand the matching sequence analogs showed similarhydrogen-bonding patterns that could potentiallystabilize these conformations. A type-II b-turn pre-ference was con®rmed for the RTPGS epitope byNMR and CD analysis of the synthetic peptide Ac-TPGS-NH2. Furthermore, synthetic peptides weremade from selected clones from three of the con-sensus sequence groups predicted to have confor-mational preferences; these bound to Ab with sub-micromolar af®nities, and inhibited Ab binding toimmobilized phage bearing the same sequence.Thus, we have identi®ed, from a set of linear epi-


topes, those whose structures and binding activi-ties are mimicked by peptides.

Sequery, in combination with SSA, provides ameans of identifying short sequences that havestructural propensities, based on the structuralsimilarity of a given sequence in the context of anumber of non-homologous proteins. The molecu-lar basis of these propensities could be understoodby analyzing the interactions within the sequencesin each protein structure. Thus, in our analysis ofthe peptide-cross-reactive epitopes in HEL andMHr, the presence of stabilizing features, such asintra-epitope hydrogen bonds that were also pre-sent in the analog set, further strengthened ourcon®dence in a given epitope sequence having astructural propensity. Thus, the results of the Sequ-ery, SSA and structural analyses were viewedtogether in determining epitope sequences that arelikely to adopt de®ned conformations in the con-text of a peptide. We expect that these peptideswill be more effective at eliciting Abs capable ofrecognizing the cognate antigen than peptides thatdo not mimic their epitope structures, and are inthe process of testing this prediction.

Critical binding residues play a dual role incross-reactivity; they impart structural stability,which allows recognition by Abs, and they provideenergetically favorable binding contacts. Mostlikely, all peptides that cross-react with linear epi-topes mimic the corresponding epitope structuresto some degree. The structural biases we identi®edappear to be conferred mainly by hydrogen bond-ing and, in one case, a proline-induced backboneconstraint (Table 3). Yet, the residues involved inpromoting a given structure are not necessarilysuf®cient for recognition by Abs. In the case of theRTPGS epitope, although the TPGS sequence wasimportant for structural mimicry, it was insuf®-cient for Ab recognition, as the synthetic peptideAc-TPGS-NH2 did not inhibit Ab binding to phagefrom the TPGS consensus group, even at a concen-tration of 50 mM free peptide (L.C. & J.K.S.,unpublished data). Thus, a basic residue N-term-inal to the TPGS sequence appears to be requiredfor binding (as seen by the phage ELISAs inTable 1A), but not for structure. In some cases,structural bias may require, or bene®t from,further stabilization by other residues, such ascysteine. This was apparent from the variationin requirements for disul®de bridging within thedifferent consensus groups (Table 1A and B).Additionally, residues other than critical bindingresidues can contribute to af®nity throughincreased contacts with Abs.

Polyclonal anti-protein Abs favor peptide-mimics of linear epitopes

All of the peptides selected in our polyclonal Abscreenings aligned to linear epitopes on HEL andMHr. This was surprising, since antibodies againstdiscontinuous epitopes predominate in anti-proteinresponses (Benjamin et al., 1984; Jin et al., 1992),

and peptide-mimics of discontinuous epitopes, or``mimotopes'', have been selected by others usingpolyclonal serum Abs (Folgori et al., 1994; Meolaet al., 1995), and monoclonal Abs against discon-tinuous epitopes (Balass et al., 1993; Felici et al.,1993; Luzzago et al., 1993; Stephen et al., 1995;Bonnycastle et al., 1996; L.C., J. Shen & J.K.S.,unpublished data). The critical residues in mimo-topes appear to be larger in number and more dis-persed than in linear epitope mimics, and hence,they are likely to be present in smaller numbers inthe libraries. This may explain why the clones thatwe select with monoclonal Abs against discontinu-ous epitopes are often weak binders (Bonnycastleet al., 1996), which require optimization of residues¯anking the consensus sequences to obtain moder-ate binding af®nities (K. L. Brown, E. Leong &J.K.S., unpublished data). Our ultimate goal is toselect peptides that not only cross-react with anti-protein Abs, but that also elicit protein-cross-reac-tive Abs; this is most likely to be successful if theaf®nities of the cross-reactive peptides are compar-able to those of the proteins. Our results indicatethat peptide-mimics of linear epitopes dominate inaf®nity-selections; thus, they may prove viable tar-gets for peptide-based vaccine strategies.

Critical binding residues provide insight intothe mechanism of Ab binding

Four of the ®ve MHr epitopes found in ourstudy were identi®ed by Geysen et al. (1987), whotested the reactivity of polyclonal anti-MHr seraagainst a panel of short, overlapping synthetic pep-tides representing the entire sequence of the anti-gen. Several of the most reactive peptides wereremarkably similar to the linear epitopes found inour peptide-library screenings. Although the criti-cal residues de®ned by Geysen et al. (1987) differedsomewhat from our assignments, both studiesrevealed a requirement for aromatic residues in thecross-reactive peptides. Aromatic residues thatcould not be replaced by chemically similar aminoacids were assumed to be directly involved in Abbinding (Getzoff et al., 1987). The side-chains ofthese critical aromatic residues are buried in theprotein epitope and thus are locally inaccessible(see Table 3 and Getzoff et al., 1987). To explainhow Ab might interact with these inaccessible resi-dues in the protein epitope, Getzoff et al. (1987)proposed an induced-®t binding mechanism, inwhich Ab initially contacts the solvent-accessibleresidues on the epitope, then induces a confor-mational change exposing the hydrophobic side-chain for binding. Such changes may be limited tolocal movements exposing a buried side-chain, ormay involve more signi®cant backbone rearrange-ments.

Conformational changes of epitopes inducedupon Ab binding are likely facilitated by theirlocations in ¯exible regions of the protein. How-ever, even those epitopes with high mobility valueshave well-de®ned structures, since they can be


resolved crystallographically. Structural stabilityhas been demonstrated for peptides that cross-react with linear epitopes, as shown here and byothers (Dyson & Wright, 1995). These ®ndings areconsistent with a two-step binding mechanism, inwhich Abs initially recognize gross structuralelements, followed by conformational adjustmentsthat improve complementarity at the Ab/antigeninterface (Tainer et al., 1984; Friedman et al., 1994).Thus, structural bias can play a role in cross-reac-tivity with peptides, even when an induced-®tbinding mechanism is involved.

Cross-reactivity can occur with a minimalnumber of shared residues

We and others have shown that peptides havingonly a few residues in common with a protein epi-tope can cross-react with that epitope. The ideathat short peptides can cross-react with folded pro-teins has been challenged on the basis of crystallo-graphic data (Laver et al., 1990). X-ray crystalstructures of Ab-protein complexes show 15 to 22protein residues in contact with the combining siteof an Ab (Davies et al., 1988; Wilson & Stan®eld,1993; Braden & Poljak, 1995), and thus short pep-tides seem unlikely to provide suf®cient contactsfor binding. However, thermodynamic studies ofAb-protein interactions indicate that only ®ve orsix of the large number of protein residues contact-ing the antigen combining site actually contributeto the energetics of binding (Novotny, 1991). More-over, the only Ab-protein complexes whose crystalstructures have been solved involve discontinuousepitopes, in which the critical binding residues arespread over two or more segments of the protein.Peptides generally cross-react with linear epitopeson folded proteins, where the critical residues arelocated on a continuous segment of the protein(although additional residues on neighboring seg-ments may make non-critical contacts). Thus,although the critical binding residues identi®ed forthe HEL and MHr epitopes probably representonly a subset of the contact residues in the proteinepitope, they make major contributions to thebinding energetics.

Cross-reactivity requires that the Abcombining site has shape complementaritywith both the linear epitope and the peptide

It has been suggested that all epitopes are dis-continuous to some extent, and that short pep-tides that cross-react with native antigen onlymimic part of the epitope (Barlow et al., 1986).For this to occur, the critical binding residuesmust be located on a single segment of the pro-tein that protrudes somewhat, so that these resi-dues will make extensive contacts with the Abcombining site. The locations of most of the lin-ear epitopes identi®ed here correspond to regionsof high mobility and accessibility. Such regionstend to have high protrusion indices (Barlow

et al., 1986; Thornton et al., 1986), which allowAb to make central, critical contacts with a con-tinuous segment of the protein. Abs that bind tosuch sites, as well as to cross-reactive peptides,must have unique binding sites that will accomo-date the epitopes on both antigens.

The structure of the Ab combining site is deter-mined to a large degree by the surface formedfrom its six hypervariable loops (L1-L3, H1-H3).Chothia & Lesk (1987) identi®ed a set of back-bone frameworks, or canonical structures, whichdescribe much of the structural variation in ®veof the six hypervariable loops (all except H3).The combinations of canonical structures used byAbs against a variety of antigens including pro-teins, peptides, carbohydrates and haptens hasbeen analyzed (Vargas-Madrazo et al., 1995). Thisstudy showed that anti-peptide Abs use arestricted set of canonical structure combinationsas compared to anti-protein Abs; this restrictionis re¯ected in variable-gene usage (Lara-Ochoaet al., 1996). This group and others (Wilson &Stan®eld, 1993; Wilson et al., 1995; MacCallumet al., 1996) have characterized the combiningsites of anti-peptide Abs as grooved; this shapeallows extensive interaction with a small numberof peptide residues. In contrast, the antigen bind-ing site of Abs speci®c for discontinuous epitopesis ¯attened, to accommodate a larger surfacearea, and the residues of these epitopes are, onaverage, less buried within the Ab combining sitethan those of bound peptides.

The cross-reactivity observed here betweenpeptides and linear epitopes is likely due to Abswhose combining sites resemble those of anti-peptide Abs. Evidence supporting this comesfrom crystallographic studies of two monoclonalAbs in complex with peptides (Tormo et al., 1994;Wien et al., 1995). Both Abs were raised againstnative or heat-inactivated virions, and recognizeintact viral particles; they also cross-react withpeptides, and their combining sites resemblethose of anti-peptide Abs. Moreover, the linearepitopes they recognize protrude (Wien et al.,1995), or are predicted to protrude (Tormo et al.,1994) from the viral coat, which would allowthem to be buried in the Ab combining sitemuch like the peptides. Thus, in these cases,cross-reactivity is associated with a grooved,``anti-peptide±Ab-like'' combining site.

Future studies

An understanding of the molecular basis of pro-tein-peptide cross-reactivity is critical in the designof vaccines that use peptides to target epitopes onfolded proteins. We describe a means of identify-ing immunodominant, linear epitopes on proteinantigens that are likely to be structurally mimickedby peptides. These peptides, when used as immu-nogens in conventional immunization strategies,may not elicit strong reactivity against a targetedsite, since the combining-site structure of the anti-


peptide Abs elicited will typically be groove-like.As mentioned, even though the combining site ofsuch Abs may be centrally con®gured to makefavorable contacts with a protruding linear epi-tope, their ridged periphery may not ®t withregions surrounding the epitope. Abs having adifferent combining-site structure from those eli-cited against the peptide may be required forcross-reactivity with the targeted site. Severalprime-and-boost strategies, involving immuniz-ations with whole protein and peptide, have suc-cessfully enhanced protein cross-reactivity (Eminiet al., 1983; Girard et al., 1995; Davis et al., 1997).These strategies involve priming with one antigenand boosting with the other, so as to enhance theproduction of cross-reactive Abs. These immuniz-ation strategies, using peptides identi®ed in thisstudy, should provide a substantial test of the via-bility of epitope-targeted vaccines. Ultimately, ourability to target Ab responses to pre-selected epi-topes will depend on our understanding of allaspects of cross-reactivity, including the structuresof the peptide, the targeted site and the Absinvolved in the desired cross-reaction. This knowl-edge should allow us to develop a rational meansfor producing in vivo Abs having predeterminedantigen speci®city.

Materials and Methods

Materials

All of the IgGs and antisera used were derived fromrabbits. MHr protein and anti-MHr Abs that had beenaf®nity puri®ed on MHr (Getzoff et al., 1987) were pro-vided by H. Alexander (Univerisity of Missouri-Colum-bia); anti-MHr serum (Getzoff et al., 1987) was a giftfrom S. J. Rodda (Chiron Corporation, Melbourne). Anti-HEL immune sera and IgG that had been af®nity puri-®ed on HEL, as well as puri®ed HEL were provided byI. Kumagai (Tohoku University, Sendai). The pIII-dis-played X6 library (Scott & Smith, 1990) and the Cys2-6pVIII libraries were provided by G.P. Smith (Universityof Missouri-Columbia); the 15mer pVIII library (Nishiet al., 1993) was provided by H. Saya (University ofKumamoto, School of Medicine). The remaining pVIIIlibraries, with the exception of AH (see below), aredescribed by Bonnycastle et al. (1996). Biotinylation ofAbs, isolation of IgG from immune sera, preparation ofthe anti-phage antisera, and the sources of other reagentsare as described (Bonnycastle et al., 1996).

The TPGS peptide (Ac-TPGS-NH2) was synthesizedby UBC Biotechnology, NAPS Unit (Vancouver). Thebiotinylated peptides Cys6-7, Cys5-1, X6-1 and X6-14were synthesized and puri®ed by reversed phase HPLCby The Alberta Peptide Institute (Edmonton). The Cys6-7and Cys5-1 peptides were subsequently oxidized and re-puri®ed to isolate only disul®de-bridged peptides. Thesequences of the peptides are as follows: Cys6-7, NH2-VNIACNPTDIQCLIRLPAE-bioK-norL-K-CONH2; Cys5-1, NH2-NQPSCQDITACLTPQ-bioK-norL-K-CONH2; X6-1, NH2-SKTPGAAAE-bioK-norL-C-CONH2; and X6-14,NH2-QEPYVLAAEGD-bioK-norL-K-CONH2, in whichbioK is biotinylated lysine, and norL is norleucine. Cy3-conjugated Af®Pure goat-anti-rabbit IgG was purchasedfrom Jackson ImmunoResearch (West Grove, PA). Bio-

tin-labeled bovine serum albumin (BSA) was purchasedfrom Sigma Chemical Co. (St. Louis, MO). Polymethyl-methacrylate (PMMA) beads (100 mm) were fromSaÅpidyne Instruments Inc. (Boise, ID).

The panel of epitope libraries

The peptides in phage-display libraries are fused tothe N terminus of either the minor coat protein pIII orthe major coat protein pVIII. The panel of peptidelibraries comprised two pIII-displayed libraries, X6 andX15, and 13 pVIII-displayed libraries, X5CX2CX5 (Cys2),X5CX3CX4 (Cys3), X4CX4CX4 (Cys4), X4CX5CX4(Cys5), X4CX6CX4 (Cys6), XCX6CX (LX6), XCX8CX(LX8), X8CX8, XCCX3CX5C (aCT), X6, X15, X30 and AH(see below); Xn represents n randomized residues (X),and C represents a ®xed cysteine. The random peptidesof the pIII libraries are preceeded by the N-terminalsequence ADGA and followed by GGAGAE. The ran-dom peptides in all of the pVIII-display libraries begin atthe N terminus and are followed by a variety ofsequences: PAEGDD (Cys libraries); GGIEGRGP (aCT),GGPAEG (one of the two LX6 libraries), AAEGDD (theremaining libraries). The AH library was designed byT. J. Mason (Melbourne), J. A. Tainer (The ScrippsResearch Institute, La Jolla) and H. M. Geysen (Glaxo-Wellcome, Research Triangle Park), and constructed byL. L. C. Bonnycastle (Simon Fraser University) asdescribed (Bonnycastle et al., 1996). The peptide inserthas the sequence EQLAKSEEQLAKXXEQXXKXXXK-XEQLAKSEEQLAX, which is predicted to have a-helicalpropensity.

Screening the panel of libraries with anti-HEL andanti-MHr Abs

The screening of pVIII libraries with anti-HEL andanti-MHR Abs (screening 1) has been described(Bonnycastle et al., 1996). Caprylic acid-puri®ed Abswere used throughout four rounds of anti-MHr selection,and for rounds 1, 3 and 4 of the anti-HEL selection, withaf®nity-puri®ed anti-HEL Abs being used in round 2.The pIII and the AH libraries were screened with anti-MHr in two separate experiments; af®nity-puri®ed anti-MHr Abs were used to screen the two pIII libraries(screening 2), and caprylic acid-puri®ed, anti-MHr Abswere used to screen the AH library (screening 3) essen-tially as described (Bonnycastle et al., 1996). Brie¯y,�1010 to 1012 virions from each library were af®nity-selected over three or four rounds of panning on biotiny-lated Abs that had been immobilized on avidin or strep-tavidin-coated microwells or Petri plates; enrichment forAb-binding phage was assessed by titering the phageoutput after each round of panning. Phage were ampli-®ed in Escherichia coli cells, and ampli®ed phage poolsfrom the third and fourth rounds were tested for Abbinding by ELISA. Individual clones were isolated frompools showing positive enrichment and/or ELISA sig-nals. These clones were sequenced in the region of thepeptide insert, and tested for binding by ELISA. Cloneswere sequenced using the method of either Haas &Smith (1993) or Sanger et al. (1977). Pooled phage fromthe Cys2, Cys3 and AH libraries showed only weakenrichment and ELISA signals after three rounds, andwere not studied further.


ELISAs

Following a previously described protocol(Bonnycastle et al., 1996), �2 � 1010 virions were immobi-lized either by direct adsorption to microwells by over-night incubation at 4�C, or by incubation for one hour atroom temperature in wells that had been coated over-night with 1 mg of anti-phage Ab. Wells containingbound phage were blocked with BSA or non-fat milk,then 60 nM biotinylated anti-HEL Ab or 100 nM biotiny-lated anti-MHr Ab was added for 3.5 hours at room tem-perature. The plates were washed after each step usingan automated microplate washer. Bound Ab wasdetected using horseradish peroxidase-avidin complexesand 2,20-azino-bis(3-ethylbenzthiazoline-6) sulfonic acidas substrate. Absorbance values were measured at405 nm minus 490 nm. The ELISA signal for fd-tet (pIII)or f88.4 (pVIII) phage, without a peptide insert was sub-tracted from those of the isolated phage clones (f88,0.017 to 0.042 absorbance unit; fd-tet, 0.046 to 0.087absorbance unit). The results are reported only for cloneshaving an A405-490 5 0.100 absorbance unit above that ofthe fd-tet or f88 phage. For competition ELISAs, biotiny-lated Abs (100 nM) were pre-incubated with 5 mM nativeHEL or MHr, or with �2 � 1010 virions, for two hours atroom temperature prior to incubation with immobilizedphage. Only a few of the clones from the pIII librarieswere tested for competition with MHr, because the Absused to screen the pIII libraries had been af®nity-puri®edon native MHr.

KinExA immunoassay

PMMA beads were coated with peptide as follows:200 mg of beads were suspended in Tris-buffered saline(TBS, 50 mM Tris-HCl (pH 7.4), 150 mM NaCl) contain-ing 0.1 mg of biotinylated BSA and gently rotated forone hour at 37�C. The beads were washed ®ve timeswith TBS, then rotated with 0.1 mg of streptavidin inTBS for one hour at 37�C. The beads were washed tentimes with TBS, then rotated with 10 mg of biotinylatedpeptide in TBS for one hour at 37�C, and stored at 4�C.Prior to use, 10 mg of biotin was added to the coatedbeads, which were then rotated for one hour at 37�C andsuspended in 30 ml of TBS/10 mg/ml BSA.

The KinExA assay for Kd determination is describedby detail by Blake et al. (1997), and is described herebrie¯y. For each peptide, the optimal Ab concentrationwas chosen to give a ¯uorescent signal of approximatelyone volt when 100 ml was passed through a KinExA ¯owcell containing peptide-coated beads. Ab at this concen-tration was incubated with varying concentrations ofbiotinylated peptide in TBS/10 mg/ml BSA/0 1% (v/v)Tween and allowed to come to equilibrium at room tem-perature for a minimum of two hours. The amount ofuncomplexed Ab in the equilibrium reaction was quanti-®ed by capture on peptide-coated beads, followed bydetection using ¯uorescently labeled anti-rabbit Abs.A fraction of the uncomplexed Ab in each equilibriumreaction was captured by ¯owing 100 ml of the reactionover peptide-coated beads, followed by a wash with350 ml of TBS/0.1% Tween. The amount of primary Abbound to the beads was quanti®ed by ¯owing 1 ml ofCy3-conjugated goat anti-rabbit Ab (1 mg/ml) over thebeads, followed by a wash (1.625 ml). All reagents werepassed over the beads in the ¯ow cell at a rate of0.250 ml/min. Fluorescence of the bead-bound Cy3-con-jugate in the ¯ow cell was measured over the course of

the reaction, and the difference in the signal before andafter addition of Cy3 was taken as a measure of bead-bound primary Ab. This signal should be directly pro-portional to the amount of uncomplexed Ab in the equi-librium reaction mixture. Reactions were run induplicate, and ¯uorescent signals were plotted as a func-tion of peptide concentration.

Data analysis was performed using the software pro-vided by the manufacturer (SaÅpidyne Instruments, Inc.).The ¯uorescent signals were entered along with the cor-responding antigen concentrations, and the best ®tvalues for the Kd and the concentration of antibodyactive sites were calculated, as well as the 95% con®-dence interval for these values. The software performedthe calculations in the following manner: binding wasassumed to follow the equation Kd � [Ab][Ag]/[AbAg],in which [Ab] and [Ag] represent the concentrations offree Ab and antigen, respectively. This equation can berearranged to yield an equation for the free Ab at equili-brium: [Ab] � Kd[AbAg]/[Ag]. Equations stating that thetotal Ab concentration (Abt) is the sum of the free andbound components (Abt � [Ab] � [AbAg]), and likewise,for the total antigen (Agt � [Ag] � [AbAg]), were substi-tuted into the free Ab equation to get free Ab as a func-tion of Kd, Abt, and Agt.

The ¯uorescent signal in the KinExA instrument islinear with [Ab] (Blake et al., 1997), and is de®ned bythe signals in the absence of free Ab (Sig0%), andwhen 100% of the Ab is free (Sig100%). Thus, the cal-culation of intermediate signals is a simple linearinterpolation:

Signal � �Sig100%ÿ Sig0%��Ab�=Abt� � Sig0%

Combining this equation with the equation for freeantibody at equilibrium resulted in a theory of signalas a function of Kd, Abt, Agt, Sig100% and Sig0%. Thesoftware used this theory to ®nd the values of Kd, Abt,Sig100%, and Sig0% which minimized the squarederror between the theory and the measured signals.The 95% con®dence interval was found by varying theKd parameter and re-optimizing the other variables ateach point. Thus, the analysis ®nds the range overwhich the Kd or Abt maintain ®t to the data points.

Sequery

A set of amino acid sequence patterns was determinedfor each epitope, based on the critical binding residuesidenti®ed by alignment of peptides with the HEL andMHr epitopes. The computer program Sequery wasdeveloped by L.A.K., M. E. Pique, E. D. Getzoff & J. A.Tainer at The Scripps Research Institute (as a successorto Searchwild, which is described by Collawn et al.,1990). Sequery was used to search the largest non-redun-dant set of PDB sequences (493 with no more than 25%sequence identity; Hobohm et al., 1992; PDB_Selectdatabase: http://swift.embl-heidelberg.de/pdbsel) forsequences (analogs) matching a given pattern. Eachsequence pattern comprised four residues, with onedesignated as fully variable (indicated by an X). Thesequence patterns used in the Sequery searches wereXDVQ, DVQX, DXTA, XDIT, DITX, XN[A,S]W,N[A,S]WX, [R,K,H]XPG, XPGS and PGSX for HEL; andXKYS, K[W,Y]SX, XYSE, YSEX, PEXY, EPYX, EXYV,DXSF and WDXF for MHr. Sequery provided a list of allanalogs for each sequence pattern, including the namesand PDB codes of the proteins in which the analogs

http://swift.embl-heidelberg.de/pdbsel)


were found, and the sequence and position of the ana-logs and residues ¯anking the analogs. The structure ofeach sequence analog was then assessed using SSA(developed by P.C.S. & L.A.K.).

SSA

The SSA algorithm was used to (i) assess the degreeof structural similarity between epitopes on HEL andMHr and the corresponding sequence analogs foundby Sequery, and (ii) assign secondary structures to thesequence analogs. SSA uses a least-squares ®ttingalgorithm to superimpose the backbone atoms of ananalog in question onto a template, then calculates thebackbone RMSD as a measure of structural similarityof the two. For the analysis of the HEL and MHr epi-topes, tetrapeptide templates were derived from thestructural coordinates of each epitope identi®ed bypeptide mapping. The sequence analogs were thensuperimposed on the corresponding epitope template;those having a RMSD of 40.75 AÊ were considered tomatch the epitope structure.

For the assignment of secondary structures to thesequence analogs, each analog was compared with a setof tetrapeptides comprising ideal structures of an a-helix,a b-strand, and the most prevalent b-turns (type I, typeII, type II0, type II0, and type VIII). These ideal structureswere created using the Biopolymer module in InsightII(Molecular Simulations, Inc., San Diego). InsightII valuesfor the backbone f and c angles were used for thea-helix and b-strand constructs, and the average f and cvalues from a statistical analysis of b-turns, as reportedby Hutchinson & Thornton (1994), were used to con-struct each turn segment. Sequence analogs that super-imposed onto a model secondary structure with a RMSDof 40.75 AÊ were assigned to that structure; however,some analogs matched well with both the helical andb-turn models. To distinguish between these, the RMSDbetween the helical model and the analog was alsomeasured by shifting one residue forward or backwardon the protein structure from which the analog wasderived; if either of these was also 40.75 AÊ , the analogwas assigned as a helix. Analogs were assigned as ``irre-gular'' if none of the model peptides superimposed withan RMSD 40.75 AÊ .

Four-residue peptides were used in the secondary-structure assignments for several reasons: (i) b-turns areinherently four residues long; (ii) the periodicity ofa-helices is approximately four residues; (iii) b-strandsalternate twice within four residues; and (iv) most of thework with Sequery uses tetrapeptides, since exactmatches to longer sequences are not likely to be found inthe PDB. Prior to this study, SSA was validated by com-paring its automated assignments to those made bydetailed visual molecular graphics analyses for 158 tetra-peptide segments from a broad range of proteins (P.C.S.& L.A.K., unpublished data). This test resulted in a four-state (a-helix, b-strand, b-turn and irregular) agreementbetween SSA and visual assignments of 81.6%. A secondcomparison was performed on 533 tetrapeptide segmentsfrom two proteins representing all b, and a and b struc-tural folds, and resulted in a four-state agreement of85.9% over 533 segments. An upper bound of 0.75 AÊ forbackbone atom RMSD was found to be optimal for auto-mated assignment of secondary structure.

The Sequery and SSA software were written in theC programming language, and will be provided toother research groups upon request to L.A.K.

NMR spectroscopy and the structure calculation ofthe TPGS peptide

The sample of Ac-TPGS-NH2 (7 mM in 0.5 ml) wasprepared by diluting a stock solution of the peptidewith aqueous buffer containing 0.1 M sodium chloride,0.04 M sodium phosphate, 1% (w/v) sodium azide and5% 2H2O (pH 5.5). NMR spectra were acquired on aBruker AMX600 spectrometer (Bruker, Karlsruhe,Germany). Two-dimensional phase-sensitive DQF-COSY (Rance et al., 1983) and ROESY (Bothner-By et al.,1984; Bax & Davis, 1985) spectra were acquired at283 K using TPPI (Red®eld & Kunz, 1975). The sweepwidths were 6666 Hz in both dimensions, and 512 t1

increments were collected in 2048 complex data pointswith 32 transients. The spin-lock pulse for the ROESYspectrum had a strength of 6024 Hz and a duration of300 ms. Water suppression was achieved using theWATERGATE technique (Piotto et al., 1992; SklenaÂret al., 1993) in the ROESY spectrum, and selectiveirradiation of the water frequency during the recyclingdelay in the DQF-COSY spectrum. All NMR data wereprocessed with the XWIN-NMR 1.2 program (Bruker).The data were apodized by 90� shifted quadratic sine-bell windows functions, zero-®lled in both dimensionsand transformed to 2000 � 2000 matrices with a post-acquisition treatment to suppress the residual waterpeak in the ROESY spectrum. An automatic base-linecorrection was applied to the ROESY spectrum beforeintegrating the cross-peaks with XWIN-NMR. Allchemical shifts were referenced to internal 4,4-dimethyl-4-silapentane-1-sulfonate.

Temperature coef®cients for the amide resonanceswere obtained by recording one-dimensional spectrabetween 283 and 303 K in steps of 5 K. The amidechemical shift was plotted as a function of temperatureand the plot was analyzed by linear regression to extractthe slope.

The ROESY cross-peak volumes were converted todistance restraints and calibrated to Pro2 Hb1-Hb2 (1.8 AÊ )using the procedure described by Hyberts et al. (1992).The upper distance bounds were adjusted for pseudo-atoms by adding 1.5 AÊ for any methyl group. Forrestraints involving only one of the non-stereospeci®callyassigned methylene protons, 1.7 AÊ were added to theupper distance limit. The structure calculations were per-formed with the unblocked TPGS peptide (i.e. the N-terminal acetyl group and the C-terminal amide werereplaced with amide and carbonyl groups, respectively).The restraints to Thr1 HN were imposed on the pseudo-atom of the N-terminal amino group, HN*, with no pseu-do-atom correction. The structural ensemble of TPGSwas calculated with the DGII program of Insight II 95.0(Biosym/MSI, San Diego, CA).

CD spectroscopy of the TPGS peptide

The CD spectrum of Ac-TPGS-NH2 was obtained at278 K on a Jasco J710 spectropolarimeter (Japan Spec-troscopic Co., Ltd, Tokyo). The peptide was dissolvedin water and diluted to a concentration of 1 mM. Themeasurements were performed in a quartz cell of0.1 cm pathlength. The CD spectrum is the average of30 consecutive scans from 180 to 260 nm, recordedwith a bandwidth of 0.5 nm, a time constant of 0.25second and a scan rate of 20 nm/min. Following base-line correction, the observed ellipticities were convertedto molar ellipticities. The CD spectrum was deconvo-


luted using the convex constraint algorithm of Perczelet al. (1992).

Acknowledgments

We appreciate generous gifts from H. Alexander, S. J.Rodda, I. Kumagai, H. Saya and G. P. Smith, and thehelical library design of J. A. Tainer, T. J. Mason andH. M. Geysen. We thank M. Rashed, J. S. Mehroke andC. Hook for excellent technical assistance, and B. M.Pinto, M. B. Zwick and P. Blancafort for comments onthe manuscript. We are grateful to B. M. Pinto and H. J.Dyson for their advice on the structural analysis of thesynthetic peptide, and to R. J. Cushley and M. Tracy fortheir assistance and use of the Simon Fraser UniversityNMR Facility. This work was supported by grants fromthe US Army Research Of®ce (DAA L03-92-G-0178 toG. P. Smith, University of Missouri-Columbia), the Presi-dent's Research Fund at Simon Fraser University (J.K.S.),the Natural Sciences and Engineering Research Council(J.K.S.), and the Michigan Research Excellence Fund Cen-ters for Protein Structure, Function, and Design and forAcademic Computing (L.A.K.). J.K.S. was supported, inpart, by a fellowship from Terrapin Technologies and ascholarship from the B.C. Health Research Foundation.L.C. was supported by a GREAT award from the B.C.Science Council and a Graduate Research Fellowshipfrom the Institute of Molecular Biology and Biochemistryat SFU. The development of Sequery by L.A.K., M. E.Pique, E. D. Getzoff and J. A. Tainer at The ScrippsResearch Institute was supported by National ScienceFoundation grant BIR 9631436 (M.E.P., E.D.G. andJ.A.T.).

References

Abola, E. E., Bernstein, F. C., Bryant, S. H., Koetzle, T. F.& Weng, J. (1987). Protein Data Bank. In Crystallo-graphic Databases ± Information Content, Software Sys-tems, Scienti®c Applications (Allen, F. H., Bergerhoff,G. & Sievers, R., eds), pp. 107±132, DataCommission of the International Union ofCrystallography, Bonn, Cambridge, Chester.

Arnon, R., Maron, E., Sela, M. & An®nsen, C. B. (1971).Antibodies reactive with native lysozyme elicitedby a completely synthetic antigen. Proc. Natl Acad.Sci. USA, 68, 1450±1455.

Atassi, M. Z. (1984). Antigenic structures of proteins.Eur. J. Biochem, 145, 1±20.

Atassi, M. Z. & Lee, C. L. (1978). The precise and entireantigenic structure of native lysozyme. Biochem. J.171, 429±434.

Balass, M., Heldman, Y., Cabilly, S., Givol, D.,Katchalski-Katzir, E. & Fuchs, S. (1993). Identi®-cation of a hexapeptide that mimics a confor-mation-dependent binding site of acetylcholinereceptor by use of a phage-epitope library. Proc.Natl Acad. Sci. USA, 90, 10638±10642.

Bansal, A. & Gierasch, L. M. (1991). The NPXY internal-ization signal of the LDL receptor adopts a reverse-turn conformation. Cell, 67, 1195±1201.

Barlow, D. J., Edwards, M. S. & Thornton, J. M. (1986).Continuous and discontinuous protein antigenicdeterminants. Nature, 322, 747±748.

Bax, A. & Davis, D. G. (1985). Practical aspects of two-dimensional transverse NOE-spectroscopy. J. Magn.Res. 63, 207±213.

Benjamin, D. C., Berzofsky, J. A., East, I. J., Gurd, F. R.,Hannum, C., Leach, S. J., Margoliash, E., Michael,J. G., Miller, A., Prager, E. M. et al. (1984). The anti-genic structure of proteins: a reappraisal. Annu.Rev. Immunol. 2, 67±101.

Bernstein, F. C., Koetzle, T. F., Williams, G. J., Meyer,E. E., Jr, Brice, M. D., Rodgers, J. R., Kennard, O.,Shimanouchi, T. & Tasumi, M. (1977). The ProteinData Bank: a computer-based archival ®le formacromolecular structures. J. Mol. Biol. 112, 535±542.

Blake, D. A., Khosraviani, M., Pavlov, Ar. R. & Blake,R. C. (1997). Characterization of a metal-speci®cmonoclonal antibody. Immunochem. Technol. Environ.Applic. 657, 49±60.

Bonnycastle, L. L. C., Mehroke, J. S., Rashed, M., Gong,X. & Scott, J. K. (1996). Probing the basis of anti-body reactivity with a panel of constrained peptidelibraries displayed by ®lamentous phage. J. Mol.Biol. 258, 747±762.

Bothner-By, A. A., Stephens, R. L., Lee, J.-M., Warren,C. D. & Jeanloz, R. W. (1984). Structure determi-nation of a tetrasaccharide: transient nuclear over-hauser effects in the rotating frame. J. Am. Chem.Soc. 106, 811±813.

Braden, B. C. & Poljak, R. J. (1995). Structural features ofthe reactions between antibodies and protein anti-gens. FASEB J. 9, 9±16.

Campbell, A. P., Wong, W. Y., Houston, M., Jr,Schweizer, F., Cachia, P. J., Irvin, R. T., Hindsgaul,O., Hodges, R. S. & Sykes, B. D. (1997). Interactionof the receptor binding domains of Pseudomonas aer-uginosa pili strains PAK, PAO, KB7 and P1 to across-reactive antibody and receptor analog: impli-cations for synthetic vaccine design. J. Mol. Biol.267, 382±402.

Chang, C.-P., Lazar, C. S., Walsh, B. J., Komuro, M.,Collawn, J. F., Kuhn, L. A., Tainer, J. A.,Trowbridge, I. S., Farquhar, M. G., Rosenfeld, M. G.,Wiley, H. S. & Bill, G. N. (1993). Ligand-inducedinternalization of the epidermal growth factorreceptor is mediated by multiple endocytic codesanalogous to the tyrosine motif found in constitu-tively internalized receptors. J. Biol. Chem. 268,19312±19320.

Chothia, C. & Lesk, A. M. (1987). Canonical structuresfor the hypervariable regions of immunoglobulins.J. Mol. Biol. 196, 901±917.

Collawn, J. F., Stangel, M., Kuhn, L. A., Esekogwu, V.,Jing, S., Trowbridge, I. S. & Tainer, J. A. (1990).Transferrin receptor internalization sequence YXRFimplicates a tight turn as the structural recognitionmotif for endocytosis. Cell, 63, 1061±1072.

Collawn, J. F., Kuhn, L. A., Sue, Liu L.-F., Tainer, J. A.& Trowbridge, I. S. (1991). Transplanted LDL andmannose-6-phosphate receptor internalization sig-nals promote high-ef®ciency endocytosis of thetransferrin receptor. EMBO J. 10, 3247±3253.

Cuniasse, P., Thomas, A., Smith, J. C., Thanh, H. L.,Leonetti, M. & Menez, A. (1995). Structural basis ofantibody cross-reactivity: solution conformation ofan immunogenic peptide fragment containing bothT and B epitopes. Biochemistry, 34, 12782±12789.

Davies, D. R., Sheriff, S. & Padlan, E. A. (1988). Anti-body-antigen complexes. J. Biol. Chem. 263, 10541±10544.


Davis, D., Morien, B., Akerblom, L., van Gils, M. E.,Bogers, W., Teeuwsen, V. & Heeney, J. L. (1997).A recombinant prime, peptide boost vaccinationstrategy can focus the immune response on to morethan one epitope even thought these may not beimmunodominant in the complex immunogen.Vaccine, 15, 1661±1669.

Dyson, H. J. & Wright, P. E. (1995). Antigenic peptides.FASEB J. 9, 37±42.

Dyson, H. J., Lerner, R. A. & Wright, P. E. (1988a). Thephysical basis for induction of protein-reactive anti-peptide antibodies. Annu. Rev. Biophys. Biophys.Chem. 17, 305±324.

Dyson, H. J., Rance, M., Houghten, R. A., Lerner, R. A.& Wright, P. E. (1988b). Folding of immunogenicpeptide fragments of proteins in water solution. I.Sequence requirements for the formation of areverse turn. J. Mol. Biol. 201, 161±200.

Eberle, W., Sander, C., Klaus, W., Schmidt, B., vonFigura, K. & Peters, C. (1991). The essential tyrosineof the internalization signal in lysosomal acid phos-phatase is part of a beta turn. Cell, 67, 1203±1209.

Emini, E. A., Jameson, B. A. & Wimmer, E. (1983). Prim-ing for and induction of anti-polio virus neutraliz-ing antibodies by synthetic peptides. Science, 304,699.

Felici, F., Luzzago, A., Folgori, A. & Cortese, R. (1993).Mimicking of discontinuous epitopes by phage-dis-played peptides, II. Selection of clones recognizedby a protective monoclonal antibody against theBordetella pertussis toxin from phage peptidelibraries. Gene, 128, 21±27.

Fisher, C. L., Greengard, J. S. & Grif®n, J. H. (1994).Models of the serine protease domain of the humanantithrombotic plasma factor activated protein Cand its zymogen. Protein Sci. 3, 588±599.

Folgori, A., Ta®, R., Meola, A., Felici, F., Galfre, G.,Cortese, R., Monaci, P. & Nicosia, A. (1994). A gen-eral strategy to identify mimotopes of pathologicalantigens using only random peptide libraries andhuman sera. EMBO J. 13, 2236±2243.

Francis, M. J., Hastings, G. Z., Sangar, D. V., Clark, R. P.,Syred, A., Clarke, B. E., Rowlands, D. J. & Brown,F. (1987). A synthetic peptide which elicits neutra-lizing antibody against human rhinovirus type 2.J. Gen. Virol. 68, 2687±2691.

Friedman, A. R., Roberts, V. A. & Tainer, J. A.(1994). Predicting molecular interactions andinducible complementarity: fragment docking ofFab-peptide complexes. Proteins: Struct. Funct.Genet. 20, 15±24.

Getzoff, E. D., Geysen, H. M., Rodda, S. J., Alexander,H., Tainer, J. A. & Lerner, R. A. (1987). Mechanismsof antibody binding to a protein. Science, 235, 1191±1196.

Geysen, H. M., Tainer, J. A., Rodda, S. J., Mason, T. J.,Alexander, H., Getzoff, E. D. & Lerner, R. A. (1987).Chemistry of antibody binding to a protein. Science,235, 1184±1190.

Ghiara, J. B., Stura, E. A., Stan®eld, R. L., Profy, A. T.& Wilson, I. A. (1994). Crystal structure of theprincipal neutralization site of HIV-1. Science, 264,82±85.

Girard, M., Meignier, B., Barre-Sinoussi, F., Kieny, M. P.,Matthews, T., Muchmore, E., Nara, P. L., Wei, Q.,Rimsky, L. & Weinhold, K., et al. (1995). Vaccine-induced protection of chimpanzees against infectionby a heterologous human immunode®ciency virustype 1. J. Virol. 69, 6239±6248.

Haas, S. J. & Smith, G. P. (1993). Rapid sequencing ofviral DNA from ®lamentous bacteriophage. Biotech-niques, 15, 422±431.

Hastings, G. Z., Speller, S. A. & Francis, M. J. (1990).Neutralizing antibodies to human rhinovirus pro-duced in laboratory animals and humans thatrecognize a linear sequence from VP2. J. Gen. Virol.71, 3055±3059.

Hobohm, U., Scharf, M., Schneider, R. & Sander, C.(1992). Selection of representative protein data sets.Protein Sci. 1, 409±417.

Hoess, R. H., Mack, A. J., Walton, H. & Reilly, T. M.(1994). Identi®cation of a structural epitope byusing a peptide library displayed on ®lamentousbacteriophage. J. Immunol. 153, 724±729.

Hollosi, M., Kover, K. E., Holly, S., Radics, L. &Fasman, G. D. (1987). Beta-turns in bridged proline-containing cyclic peptide models. Biopolymers, 26,1555±1572.

Hopp, T. P. & Woods, K. R. (1981). Prediction of proteinantigenic determinants from amino acid sequences.Proc. Natl Acad. Sci. USA, 78, 3824±3828.

Hutchinson, E. G. & Thornton, J. M. (1994). A revisedset of potentials for beta-turn formation in proteins.Protein Sci. 3, 2207±2216.

Hyberts, S. G., Goldberg, M. S., Havel, T. F. & Wagner,G. (1992). The solution structure of eglin c based onmeasurements of many NOEs and coupling con-stants and its comparison with X-ray structures.Protein Sci. 1, 736±751.

Javaherian, K., Langlois, A. J., LaRosa, G. J., Profy, A. T.,Bolognesi, D. P., Herlihy, W. C., Putney, S. D. &Matthews, T. J. (1990). Broadly neutralizing anti-bodies elicited by the hypervariable neutralizingdeterminant of HIV-1. Science, 250, 1590±1593.

Jemmerson, R. (1987). Antigenicity and native structureof globular proteins: low frequency of peptide reac-tive antibodies. Proc. Natl Acad. Sci. USA, 84, 9180±9184.

Jin, L., Fendly, B. M. & Wells, J. A. (1992). High resol-ution functional analysis of antibody-antigen inter-actions. J. Mol. Biol. 226, 851±865.

Kabsch, W. & Sander, C. (1983). Dictionary of proteinsecondary structure: pattern recognition of hydro-gen-bonded and geometrical features. Biopolymers,22, 2577±2637.

Keller, P. M., Arnold, B. A., Shaw, A. R., Tolman, R. L.,Van M., F., Bondy, S., Rusiecki, V. K., Koenig, S.,Zolla-Pazner, S. & Conard, P., et al. (1993). Identi®-cation of HIV vaccine candidate peptides by screen-ing random phage epitope libraries. Virology, 193,709±716.

Lara-Ochoa, F., Almagro, J. C., Vargas-Madrazo, E. &Conrad, M. (1996). Antibody-antigen recognition: acanonical structure paradigm. J. Mol. Evol. 43, 678±684.

Laver, W. G., Air, G. M., Webster, R. G. & Smith-Gill,S. J. (1990). Epitopes on protein antigens: miscon-ceptions and realities. Cell, 61, 553±556.

Lerner, R. A. (1982). Tapping the immunological reper-toire to produce antibodies of predetermined speci-®city. Nature, 299, 593±596.

Lundkvist, A., Kallio-Kokko, H., Sjolander, K. B.,Lankinen, H., Niklasson, B., Vaheri, A. & Vapalahti,O. (1996). Characterization of Puumala virusnucleocapsid protein: identi®cation of B-cell epi-topes and domains involved in protective immu-nity. Virology, 216, 397±406.


Luzzago, A., Felici, F., Tramontano, A., Pessi, A. &Cortese, R. (1993). Mimicking of discontinuous epi-topes by phage-displayed peptides, I. Epitope map-ping of human H ferritin using a phage library ofconstrained peptides. Gene, 128, 51±57.

MacCallum, R. M., Martin, A. C. & Thornton, J. M.(1996). Antibody-antigen interactions: contact anal-ysis and binding site topography. J. Mol. Biol. 262,732±745.

Meola, A., Delmastro, P., Monaci, P., Luzzago, A.,Nicosia, A., Felici, F., Cortese, R. & Galfre, G.(1995). Derivation of vaccines from mimotopes:immunological properties of HBsAg mimotopes dis-played on ®lamentous phage. J. Immunol. 146,3162±3172.

Milton, L. R. C. & Van Regenmortel, M. H. V. (1979).Immunochemical studies of tobacco mosaic virus-III. Mol. Immunol. 16, 179±184.

Motti, C., Nuzzo, M., Meola, A., Galfre, G., Felici, F.,Cortese, R., Nicosia, A. & Monaci, P. (1994). Recog-nition by human sera and immunogenicity ofHBsAg mimotopes selected from an M13 phage dis-play library. Gene, 146, 191±198.

Nishi, T., Tsuri, H. & Saya, H. (1993). Construction of aphage random peptide library and its applications.Exp. Med. 11, 1759±1764.

Novotny, J. (1991). Protein antigenicity: a thermodyn-amic approach. Mol. Immunol. 28, 201±207.

Parge, H. E., Arvai, A. S., Murtari, D. J., Reed, S. I. &Tainer, J. A. (1993). Human CksHs2 atomic struc-ture: a role for its hexameric assembly in cell cyclecontrol. Science, 262, 387±395.

Pellequer, J. L., Westhof, E. & Van, R. M. (1991). Predict-ing location of continuous epitopes in proteins fromtheir primary structures. Methods Enzymol. 203,176±201.

Perczel, A., Hollosi, M., Tusnady, G. & Fasman, G. D.(1991). Convex constraint analysis: a natural decon-volution of circular dichroism curves of proteins.Protein Eng. 4, 669±679.

Perczel, A., Park, K. & Fasman, G. D. (1992). Analysis ofthe circular dichroism spectrum of proteins usingthe convex constraint algorithm: a practical guide.Anal. Biochem. 203, 83±93.

Perczel, A., Kollat, E., Hollosi, M. & Fasman, G. D.(1993). Synthesis and conformational analysis of N-glycopeptides. II. CD, molecular dynamics, andNMR spectroscopic studies on linear N-glycopep-tides. Biopolymers, 33, 665±685.

Piotto, M., Saudek, V. & Sklenar, V. (1992). Gradient-tai-lored excitation for single-quantum NMR spec-troscopy of aqueous solutions. J. Biomol. NMR, 2,661±665.

Rance, M., Sorensen, O. W., Bodenhausen, G., Wagner,G., Ernst, R. R. & Wuthrich, K. (1983). Improvedspectral resolution in COSY 1H NMR spectra ofproteins via double quantum ®ltering. Biochem. Bio-phys. Res. Commun. 117, 479±485.

Red®eld, A. G. & Kunz, S. D. (1975). Quadrature FourierNMR detection: simple multiplex for dual detectionand discussion. J. Magn. Reson. 19, 250±254.

Richardson, J. S. (1981). The anatomy and taxonomy ofprotein structure. Advan. Protein Chem. 34, 167±339.

Richardson, J. S. & Richardson, D. C. (1990). Principlesand patterns of protein conformation. In Predictionof Protein Structure and the Principles of Protein Con-formation (Fasman, E. G., ed.), pp. 2±98, PlenumPress, New York.

Rini, J. M., Stan®eld, R. L., Stura, E. A., Salinas, P. A.,Profy, A. T. & Wilson, I. A. (1993). Crystal structureof a human immunode®ciency virus type 1 neutra-lizing antibody, 50.1, in complex with its V3 looppeptide antigen. Proc. Natl Acad. Sci. USA, 90,6325±6329.

Robinson, K., Mostratos, A. & Grencis, R. K. (1995).Generation of rubella virus-neutralising antibodiesby vaccination with synthetic peptides. FEMSImmunol. Med. Microbiol. 10, 191±198.

Sanger, F., Nicklen, S. & Coulson, A. R. (1977). DNAsequencing with chain-terminating inhibitors. Proc.Natl Acad. Sci. USA, 74, 5463±5467.

Satterthwait, A. C., Arrhenius, T., Hagopian, R. A.,Zavala, F., Nussenzweig, V. & Lerner, R. A. (1989).The conformational restriction of synthetic peptides,including a malaria peptide, for use as immuno-gens. Phil. Trans. Roy. Soc. Lond. Biol. 323, 565±572.

Scott, J. K. & Smith, G. P. (1990). Searching for peptideligands with an epitope library. Science, 249, 386±390.

Sheriff, S., Silverton, E. W., Padlan, E. A., Cohen, G. H.,Smith-Gill, S. J., Finzel, B. C. & Davies, D. R. (1987).Three-dimensional structure of an antibody-antigencomplex. Proc. Natl Acad. Sci. USA, 84, 8075±8079.

SklenaÂr, V., Piotto, M., Leppik, R. & Saudek, V. (1993).Gradient-tailored water suppression for 1H-15NHSQC experiments optimized to retain full sensi-tivity. J. Magn. Reson. A102, 241±245.

Stephen, C. W., Helminen, P. & Lane, D. P. (1995).Characterisation of epitopes on human p53 usingphage-displayed peptide libraries: insights intoantibody-peptide interactions. J. Mol. Biol. 248,58±78.

Tainer, J. A., Getzoff, E. D., Alexander, H., Houghten,R. A., Olson, A. J., Lerner, R. A. & Hendrickson,W. A. (1984). The reactivity of anti-peptide anti-bodies is a function of the atomic mobility of sitesin a protein. Nature, 312, 127±134.

Thornton, J. M., Edwards, M. S., Taylor, W. R. &Barlow, D. J. (1986). Location of `continuous' anti-genic determinants in the protruding regions ofproteins. EMBO J. 5, 409±413.

Tormo, J., Blaas, D., Parry, N. R., Rowlands, D., Stuart,D. & Fita, I. (1994). Crystal structure of a humanrhinovirus neutralizing antibody complexed with apeptide derived from viral capsid protein VP2.EMBO J. 13, 2247±2256.

Tsang, P., Rance, M., Fieser, T. M., Ostresh, J. M.,Houghten, R. A., Lerner, R. A. & Wright, P. E.(1992). Conformation and dynamics of an Fab'-bound peptide by isotope-edited NMR spec-troscopy. Biochemistry, 31, 3862±3871.

van der Werf, S., Briand, J. P., Plaue, S., Burckard, J.,Girard, M. & Van Regenmortel, M. H. (1994). Abil-ity of linear and cyclic peptides of neutralizationantigenic site 1 of poliovirus type 1 to induce viruscross-reactive and neutralizing antibodies. Res.Virol. 145, 349±359.

Vargas-Madrazo, E., Lara-Ochoa, F. & Almagro, J. C.(1995). Canonical structure repertoire of the anti-gen-binding site of immunoglobulins suggestsstrong geometrical restrictions associated to themechanism of immune recognition. J. Mol. Biol. 254,497±504.

Venkatachalam, M. (1968). Stereochemical criteria forpolypeptides and proteins: conformation of a sys-tem of three linked peptide units. Biopolymers, 6,1425±1436.

http://www.hbuk.co.uk/jmb


Wagner, G., Neuhaus, D., Worgotter, E., Vasak, M.,Kagi, J. H. & Wuthrich, K. (1986). Nuclear magneticresonance identi®cation of ``half-turn'' and 3(10)-helix secondary structure in rabbit liver metallothio-nein-2. J. Mol. Biol. 187, 131±135.

Westhof, E., Altschuh, D., Moras, D., Bloomer, A. C.,Mondragon, A., Klug, A. & Van Regenmortel, M. H.(1984). Correlation between segmental mobility andthe location of antigenic determinants in proteins.Nature, 311, 123±126.

White-Scharf, M. E., Potts, B. J., Smith, L. M.,Sokolowski, K. A., Rusche, J. R. & Silver, S. (1993).Broadly neutralizing monoclonal antibodies to theV3 region of HIV-1 can be elicited by peptideimmunization. Virology, 192, 197±206.

Wien, M. W., Filman, D. J., Stura, E. A., Guillot, S.,Delpeyroux, F., Crainic, R. & Hogle, J. M. (1995).Structure of the complex between the Fab frag-ment of a neutralizing antibody for type 1 polio-virus and its viral epitope. Nature Struct. Biol. 2,232±243.

Wilmot, C. M. & Thornton, J. M. (1988). Analysis andprediction of the different types of beta-turn in pro-teins. J. Mol. Biol. 203, 221±232.

Wilson, I. A. & Stan®eld, R. L. (1993). Antibody-antigeninteractions. Curr. Opin. Struct. Biol. 3, 113±118.

Wilson, I. A., Ghiara, J. B. & Stan®eld, R. L. (1995).Structure of anti-peptide antibody complexes. Res.Immun. 145, 73±78.

Yao, Z. J., Kao, M. C. & Chung, M. C. (1995). Epitopeidenti®cation by polyclonal antibody from phage-displayed random peptide library. J. Protein Chem.14, 161±166.

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

Edited by J. A. Wells

(Received 29 May 1997; received in revised form20 April 1998; accepted 28 April 1998)

Supplementary material comprising four Tablesand one Figure is available from JMB Online.

http://www.hbuk.co.uk/jmb

The role of structure in antibody cross-reactivity between peptides and folded proteins

Documents

Transcript of The role of structure in antibody cross-reactivity between peptides and folded proteins