Introduction

A phage-displayed random peptide library is created whenrandom oligonucleotides are inserted at the 5′ end of one ofthe genes for the coat proteins of filamentous bacteriophage,gIII or gVIII, thus allowing for expression of foreign peptidesas pIII or pVIII fusion proteins on the phage surface. Accord-ingly, screening of phage-displayed random peptide librarieswith antisera to a given antigen can reveal mimotopes of Bcell epitopes of that antigen. A phage that displays a particu-lar peptide selected by an antibody is known as a phagotope.Currently efforts are being made to optimize both experimen-tal protocols and the interpretation of data from phage-displayed library screening.1–4 Although phagotopes selectedby a given antiserum should react in immunoassays with thatselecting antiserum, such phagotopes usually exhibit a diversearray of peptide sequences and their immunoreactivity mayvary. Furthermore, this reactivity also can vary according tothe assay format used.5 There is a need to investigate the

possible sources of variation in immunoreactivity with theselecting antibody of phagotopes derived from randompeptide phage-displayed libraries.

We have previously isolated phagotopes by screeningphage-displayed random peptide libraries with a monoclonalantibody (mAb), CII-C1, to type II collagen.6–8 Both the anti-body and the antigen are very well characterized. The con-formational epitope for CII-C1 on native type II collagen, C1,has been deduced by use of a series of chimeric collagens inwhich overlapping fragments of type II collagen wereinserted into type X collagen; the minimum primarysequence required for CII-CI reactivity was ARGLT.8 In addi-tion, the complementarity determining regions (CDR) of CII-C1 have been sequenced.9,10 Our results from screeningphage-displayed random peptide libraries with CII-C1 havebeen consistent with knowledge of the CII-C1 epitope, in thatmost of the selected phagotopes have at least one basic andone hydrophobic residue within the expressed peptide.7 Themost frequently isolated phagotope expressed the peptideRRLPFGSQM and many, but not all, expressed peptides withvariations on both the motifs RRL and FGxQ. We have pro-posed that the glutamine, which is present in most of thephage-displayed peptides, but does not occur in the C1epitope, may bind to the CDR3 of CII-C1. Because the reac-tivity of CII-C1 with the conformational C1 epitope on

Immunology and Cell Biology (1999) 77, 483–490

Research Article

Phagotopes derived by antibody screening of phage-displayedrandom peptide libraries vary in immunoreactivity: Studies using anexemplary monoclonal antibody, CII-C1, to type II collagen

JANET M DAVIES, MERRILL J ROWLEY and IAN R MACKAY

Department of Biochemistry and Molecular Biology, Monash University, Clayton, Victoria, Australia

Summary Antibody screening of phage-displayed random peptide libraries to identify mimotopes of conforma-tional epitopes is promising. However, because interpretations can be difficult, an exemplary system has been usedin the present study to investigate whether variation in the peptide sequences of selected phagotopes correspondedwith variation in immunoreactivity. The phagotopes, derived using a well-characterized monoclonal antibody, CII-C1, to a known conformational epitope on type II collagen, C1, were tested by direct and inhibition ELISA forreactivity with CII-C1. A multiple sequence alignment algorithm, PILEUP, was used to sort the peptides expressedby the phagotopes into clusters. A model was prepared of the C1 epitope on type II collagen. The 12 selected phago-topes reacted with CII-C1 by both direct ELISA (titres from < 100–11 200) and inhibition ELISA (20–100% inhibition); the reactivity varied according to the peptide sequence and assay format. The differences in reactivitybetween the phagotopes were mostly in accord with the alignment, by PILEUP, of the peptide sequences. Thefinding that the phagotopes functionally mimicked the C1 epitope on collagen was validated in that amino acidsRRL at the amino terminal of many of the peptides were topographically demonstrable on the model of the C1epitope. Notably, one phagotope that expressed the widely divergent peptide C-IAPKRHNSA-C also mimicked theC1 epitope, as judged by reactivity in each of the assays used: these included cross-inhibition of CII-C1 reactivitywith each of the other phagotopes and inhibition by a synthetic peptide corresponding to that expressed by the mostfrequently selected phagotope, RRLPFGSQM. Thus, it has been demonstrated that multiple phage-displayed pep-tides can mimic the same epitope and that observed immunoreactivity of selected phagotopes with the selectingmAb can depend on the primary sequence of the expressed peptide and also on the assay format used.

Key words: monoclonal antibody, phage display, random peptide libraries, type II collagen.

Correspondence: Ian R Mackay, Department of Biochemistry and Molecular Biology, Monash University, Clayton, Vic. 3168,Australia. Email: <Ian.Mackay@med.monash.edu.au>

Received 23 February 1999; accepted 7 June 1999.

type II collagen is so well defined, we chose this system toexamine the degree to which immunoreactivity of the select-ing antibody, CII-C1, with the selected phagotopes would bein agreement with variation in peptide sequence. Accord-ingly, we have used direct, inhibition and cross-inhibitionELISA formats to assess immunoreactivity and an objectivecomputer-based method to align the sequences of the pep-tides displayed by the phagotopes. To aid the interpretation ofthe data, we have prepared a model of the known C1 epitopeof type II collagen for CII-C1.

Materials and Methods

Derivation of the phagotopes

The mAb CII-C1 (400 µg/mL) to type II collagen was prepared fromhybridoma cultured in serum-free medium.7 The variable regiongenes of CII-C1 have been sequenced; these include the use of V

Hb

J558 and VL

Vk21C, which are frequently used by other mAb thatreact with the same C1 epitope.9,10 Fifteen phagotopes with uncon-strained peptide inserts (1–15) and two phagotopes with peptideinserts constrained by flanking cysteines (16, 17) were derived froma pool of two random nonapeptide phagemid libraries11,12 by bio-panning with CII-C1 as described by Cook et al.7 We studied 10 ofthe 15 unconstrained phagotopes and the two constrained phagotopesand used one of each phagotope expressing a peptide that wasselected more than once. Each phagotope was propagated inEscherichia coli K91 in 60 mL Lauria Bertoni (LB) broth in thepresence of M13 K07 helper phage, ampicillin and kanomycin.11

Wild type f1 (f1) phage was cultured in LB in the absence of helperphage without antibiotics. Phage were purified from clarified culturemedium13 and suspended in 3 mL of 10 mmol/L Tris-HCl, 1 mmol/LEDTA pH 8.0 (TE). There were approximately 1012 ampicillin-trans-ducing units per mL (equivalent to 1013 plaque-forming units) ineach of the phage suspensions and wild-type phage was at 5 × 1013

plaque-forming units per mL.

Direct ELISA

Bovine type I and II collagens were prepared14 and dissolved at1 mg/mL in 0.05 mol/L acetic acid. Nunc Maxisorp (Roskilde,Denmark) microtitre plate wells were coated with 2 µL phage sus-pension or 0.5 µg collagen in 100 µL sterile PBS. The method ofCook et al.7 was followed except that plates were incubated withprimary antibody for 16 h at 4°C and with secondary antibody for4 h. The titration of CII-C1 commenced at a dilution of 1:100 forwells coated with the phagotopes and 1:10 000 for wells coated withcollagen. The titre of CII-C1 reactivity with each phagotope wasextrapolated from the curve as the dilution resulting in an opticaldensity (OD) of 0.5, which was in the linear range. To compare theamount of phage bound to the plates, each phagotope was tested withpolyclonal sheep antibodies to the filamentous phage M13 (anti-M13, Pharmacia, Uppsala, Sweden) diluted 1:10 000. The reactivityof anti-M13 with each of the phagotopes bound to the plate wassimilar, with a mean OD ± SD at 415 nm of 1.672 ± 0.289.

Inhibition ELISA

The ability of collagens, peptides and phagotopes to inhibit the reac-tivity of CII-C1 was tested by an inhibition ELISA. Plates werecoated with collagen or the phagotopes as for the direct ELISA. Theantibody CII-C1 was used at a concentration that yielded an OD at415 nm of approximately 1 after 30 min incubation with substrate;

thus CII-C1 was diluted 1:30 000 for collagen coated plates, 1:400for plates coated with phagotopes 11, 13, 14, 15 and 16 and at 1:100for plates coated with the other phagotopes. The CII-C1 was pre-pared at 1.25 times the chosen concentration in 0.8 times the finalvolume of diluent and suspensions of the phagotopes, equivalent to20, 10 and 5 µL per 100 µL well, were added in 0.2 times the finalvolume. Types I and II collagen were used as inhibitors at concen-trations of 10, 2.5 and 0.5 µg/well. The peptide RRLPFGSQM was synthesized, purified by high performance liquid chromatography to single peak purity (> 96%) and analysed by mass spectrometry by Chiron Mimotopes (Melbourne, Vic., Australia). A peptideLKIGDFPAG, derived from the same library by an irrelevant anti-body, was synthesized in-house by F-moc chemistry on the PS3Protein Technologies Automated Peptide Synthesiser (Rainin Indus-tries Co., Woburn, MA, USA) and purified as above to the same levelof purity. These peptides were dissolved at 1 mg/mL in PBS andtested as inhibitors at 200 µg/mL. Each inhibitor was tested in quad-ruplicate. Tubes with antibody and inhibitor were mixed by inversionand incubated at room temperature for 4 h before being transferredto antigen-coated microtitre plates and incubated overnight. Subse-quent steps in the ELISA were performed as for the direct ELISA.For each plate, 20 wells in which diluent was added in place ofinhibitor were distributed throughout the 96-well plate in sets ofquadruplicates and served as uninhibited controls. The mean reac-tivity of each set of quadruplicate uninhibited wells was calculatedand the percentage inhibition of each well with inhibitor relative tothe mean reactivity of the closest set of uninhibited wells was deter-mined. The mean percentage inhibition of the reactivity of CII-C1 in100 uninhibited wells coated with collagen was 0.03% ± 0.7. A valueof 20% was arbitrarily assigned as a cut-off for significant inhibitionfor all cross-inhibition assays; this was based approximately on themean + 3SD of the inhibition of the reactivity of CII-C1 in 60 un-inhibited wells coated with each phagotope. If the reaction with CII-C1 was weak, for example with phagotopes 5, 9 and 12, platescoated were developed for 2 h.

Alignment of peptide sequences

The phage-expressed peptides were grouped into clusters of similarpeptides by a computer-based algorithm.15 The peptide sequenceswere aligned on the basis of similarity of sequence using PILEUP16

in conjunction with the Tudos matrix,17 which is based on isomorphicproperties of amino acids and local neighbourhood effects of aminoacids within proteins. A low penalty of 1 was set for the introductionof gaps between aligned sequences.

Modelling of the C1 epitope

The antibody CII-C1 recognizes the conformational epitope C1 oftype II collagen, which requires amino acids 359–364 and the nativehelical structure of type II collagen.8 A model of type II collagen wasgenerated using the algorithm MODELLER,18 with a template beingthe theoretical model of the triple-helical region of type II collagen(1bbf) deposited by Nemethy et al.19 to the protein data bank.20 Theimage was prepared using Q UA N TA (Microsoft International Inc.)and S H OW C A S E (SGI Inc.).

Results

Direct ELISA: Reactivity of CII-C1 with type II collagenand the selected phagotopes

The CII-C1 reacted strongly by ELISA with type II collagento a titre of 72 000 and not with type I collagen. The CII-C1

JM Davies et al.484

antibody reacted with each of the 12 phagotopes selected byCII-C1, but in each case to a lower titre than with type II collagen (Table 1). There were striking differences in thelevel of reactivity among the phagotopes from titres from< 100–11 200. Of note, phagotope 14 (HRLAFGQNT, titre1330), which differed from phagotope 15 by only one aminoacid N→Y, was less reactive than phagotope 15 (titre11 200). The two phagotopes, 16 and 17, derived from thecysteine-constrained library were reactive with CII-C1 (titres3020 and 175). There were three phagotopes selected by CII-C1, 9, 12 and 10, that were only weakly reactive withCII-C1 by direct ELISA.

Inhibition ELISA: Inhibitory effect of phagotopes on CII-C1 reactivity with type II collagen

The reactivity of CII-C1 with type II collagen was inhibitedby 100 µg/mL of type II collagen (80%), while type I colla-gen had no effect (Table 1). Representative curves for theinhibition by various phagotopes of reactivity of CII-C1 withtype II collagen are shown in Fig. 1a; phagotopes 11, 15, 16,14 and 13 had the highest inhibition values (Table 1). Theslopes of the inhibition curves as well as their percentageinhibition vary considerably, which, in addition to data fromthe direct ELISA, indicates differences in binding strengthbetween CII-C1 and the phagotopes. However, the differencebetween phagotopes 14 and 15 was less marked in the inhi-bition assay than in the direct ELISA. The cysteine-constrained phagotope 16 (C-IAPKRHNSA-C) differed insequence from all others, but notwithstanding was stronglyinhibitory, suggesting this phagotope also functionallymimics the C1 epitope.

The reactivity of CII-C1 with the phagotopes in theELISA and their ability to inhibit the reactivity of CII-C1with type II collagen was correlated, albeit weakly (r2 = 0.6)

(Fig. 2), with some phagotopes (9 and 12) showing muchgreater capacity to inhibit the reactivity of CII-C1 with typeII collagen than to react with CII-C1 by direct ELISA. Wenote that the peptides displayed by these phagotopes, TRS-FGIQAT and TRAFGNEAT, were similar to each other butdiffered from all other peptides by the presence of threonineat the amino-terminal end of the peptide insert.

Cross-inhibition ELISA: Mutually inhibitory effect ofphagotopes on CII-C1 reactivity

Each phagotope was tested for its ability to compete withother phagotopes for the binding of CII-C1 and exemplarycases are shown (Fig. 3). As expected, the inhibitory capacityof the phagotopes was dose-dependent. The capacity of eachphagotope to inhibit reactivity of CII-C1 with other phago-topes was proportional to their inhibitory capacity on CII-C1reactivity with type II collagen; phagotopes 9 and 12 wereinhibitory of the reactivity of CII-C1 with other phagotopes.Of note, the cysteine-constrained phagotope 16, for which thepeptide sequence (C-IAPKRHNSA-C) was different fromother phagotopes, inhibited the reactivity of CII-C1 with allthe other phagotopes and each of the phagotopes 11, 13, 14and 15 cross-inhibited CII-C1 reactivity with phagotope 16.Moreover, the synthetic peptide RRLPFGSQM correspondingto the sequence expressed by phagotope 1, which was selectedfour times by CII-C1, was used to inhibit the reactivity of CII-C1 with phagotope 1 and four other phagotopes. The reactivity of CII-C1 with phagotopes numbered 1, 11 (HEHT-FGRQW), 13 (RAAPFGNQW), 15 (HRLAFGQYT) and 16(C-IAPKRHNSA-C) was in each instance inhibited by thesynthetic peptide, respectively, by 68, 87, 90, 84 and 80%(range of SEM 0.3–4.8%), but not by an irrelevant syntheticcontrol peptide LKIGDFPAG. These data further support theconclusion that phagotope 16, which expressed the divergentpeptide, is indeed a mimic of the C1 epitope.

Interpretation of phage-display data 485

Table 1 Reactivity of CII-C1 with phagotopes and type II collagen and type I collagen by direct ELISA and inhibition ELISA

Phagotope No. times selected Peptide insert Titre* (OD 415 nm) Percentage inhibition†Mean SEM

1 4 RRLPFGSQM 725 21 2.05 2 RYAFGSQIA 200 45 2.07 1 RRLPFGSSL 100 7 1.39 1 TRSFGIQAT <100 (0.23) 36 2.510 1 SRLAFGDQL <100 (0.05) –2 2.111 1 HEHTFGRQW 800 90 0.812 1 TRAFGNEAT <100 (0.26) 58 1.413 1 RAAPFGNQW 5970 81 0.914 1 HRLAFGQNT 1330 72 1.615 1 HRLAFGQYT 11200 82 0.616 1 C–IAPKRHNSA–C 3020 73 1.117 1 C–ESAQRPFGC–C 175 20 1.6CII NA NA 72000 80 0.6CI NA NA Nil 6 2.5

*The titre by direct ELISA represented by the inverse of the CII-C1 dilution was extrapolated from the curve of CII-C1 reactivity with eachphagotope at an OD 415 nm of 0.5. For those phagotopes only weakly reactive by direct ELISA, the OD 415 nm at 1:100 is shown. No reactiv-ity was observed with wild-type f1, phage lacking peptide inserts or phagotopes expressing three unrelated peptides SRSKSALS, QRKKGAPYand YRDLLYSPI (see Fig. 1). No reactivity was observed with type I collagen at the highest dilution tested; at 1:5000 the OD 415 nm was 0.007.†The means and SEM are listed for the percentage inhibition of the reactivity of CII-C1 with type II collagen by each phagotope tested with10 µL of inhibitor per well. NA, not applicable.

Analysis of the peptides displayed by the selected phagotopes and the C1 epitope

To see whether variation in immunoreactivity was in accordwith variation in peptide sequence, the peptides expressed by

the phagotopes were aligned on the basis of similarity in theiramino acid sequence (Fig. 4). A main cluster of 10 phago-topes aligned according to the presence of the amino acids R,L, A, F, G and Q. The two phagotopes from the cysteine-constrained library numbered 16 and 17 were aligned to separate clusters. The inner clusters contained those phago-topes, 1, 7, 5 and 10, that were most frequently selected byCII-C1 and those that were least reactive with CII-C1. Therewas a general concordance between the alignment of thepeptide sequences and the observed immunoreactivity of the phagotopes. Thus, the phagotopes 9 and 12, which wereoutliers on Fig. 2, which correlates reactivity by directELISA and competitive inhibition ELISA, were aligned toone cluster and two pairs of phagotopes, 14 and 15, 11 and13, which were similarly reactive, were aligned as clusters.Nonetheless, the concordance between the sequence align-ment and the observed reactivity was incomplete in that thepair of phagotopes 1, 7 differed by only two amino acids, yettheir reactivity differed greatly in potency. The Q→S changemost likely accounts for this difference in reactivity, becausemost of the other reactive phagotopes possess a Q andbecause the second difference, M→L, is unlikely to influencereactivity because both are hydrophobic amino acids.

A model of the C1 epitope on type II collagen

The structure of collagen in the region of the C1 epitope thatrequires the minimum sequence of ARGLT was modelled inorder to understand how the phage-displayed peptides mimicthe epitope recognised by CII-C1. The arginine and leucineresidues of adjacent chains are juxtaposed on the surface ofthe triple helix (Fig. 5) and there are no glutamine residuespresent in the region of the C1 epitope of collagen. On one

JM Davies et al.486

Figure 1 Reactivity of exemplary phagotopes with CII-C1 antibody. (a) Titration of CII-C1 with phagotopes 1 (s), 12 (n), 14 (✱), 15 (m), 16 (h) and 17 (d) by direct ELISA. Differences in the level and slope of the titration curves with different phagotopes are evident.There was no reactivity of CII-C1 with type I collagen, wild-type f1 bacteriophage, phagotopes lacking inserts or phagotopes expressingthree unrelated peptides, SRSKSALS, QRKKGAPYH and YRDLLYSPI, that had been selected with other irrelevant mAbs. At a dilutionof 1/100, the reactivity of CII-C1 with these negative-control antigens had a combined mean of 0.025 and a standard deviation of 0.013.(b) Inhibition of the reactivity of CII-C1 with type II collagen by phagotopes 1 (s), 11 (j), 12 (n), 15 (m) and 16 (h). Phagotope 12(TRAFGNEAT), which was weakly reactive in (a), was inhibitory and phagotope 16 (C-IAPKRHNSA-C) was reactive by both assays.Type I collagen, wild-type f1, buffer (TE), phagotopes with no insert or an insert encoding a stop codon had no effect on CII-C1 reactivity with type II collagen (mean ± SEM = 5 ± 2.5, –3 ± 3.4, 2 ± 2.7 and –6 ± 5.1% at 10 µL/mL). The data points are the mean ofquadruplicate wells and standard errors are marked.

Figure 2 Correlation between reactivity by direct and inhibitionELISA. The values for the reactivity of CII-C1 antibody at 1:100are plotted against the percentage inhibition of CII-C1 reactivitywith type II collagen by each phagotope tested at 20 µL/well.Reactivity of each phagotope with CII-C1 by ELISA correlated(r 2 = 0.6) with the inhibitory capacity of that phagotope on CII-C1 reactivity with type II collagen, with the exception of phago-topes 9 and 12, which had a greater inhibitory effect on CII-C1reactivity with type II collagen than reactivity by direct ELISA.

plane of the collagen molecule, two arginine residues and aleucine residue are less than 2.5 nm apart and thus are withinthe size of an antibody paratope/epitope contact site. Manyof the peptides in the main cluster shown in Fig. 4 containedone or more basic residues (R or H) and a hydrophobicamino acid (L, Y or A) in the first part of the peptide, whichis in accord with the model of the C1 epitope. Theimmunoassay data, taken together with the sequence align-ment and the model, suggest that the amino-terminals of thepeptides in the main cluster, frequently RRL, mimic the C1epitope. It is likely that the carboxy-terminal of the peptide,frequently FGxQ, contributes to binding strength through itsinteraction with CDR3. It is noteworthy that while phago-tope 16 (C-IAPKRHNSA-C) is different in sequence fromthe other peptides, it does contain a pair of basic residues K,R as well as hydrophobic residues I, A, albeit in a differentorder to that in the other peptides. The peptide expressed byphagotope 16 is structurally constrained by two flanking cys-teine residues and its conformation also could mimic that ofthe C1 epitope.

Discussion

The conformational structure of an epitope to which an anti-body binds is difficult to define. In particular, there is a needto characterize the specificity of the interaction betweenepitope and paratope in the case of autoantibodies, because offrequent suggestions that this interaction can occur as a resultof mimicry between a host structure and an extrinsicpathogen.21 Antibody screening of phage-displayed randompeptide libraries provides very useful information on struc-tural features of epitopes, but accurate analysis of phagotopesderived by screening of phage libraries can be a problematic.Accordingly, we examined sources of variation in immuno-

Interpretation of phage-display data 487

Figure 3 Cross - inh ib i to r ycapacity between phagotopes forreactivity with CII-C1. The fivenumbered phagotopes and theirpeptide inserts used to coat themicrotitre plate are shown on thex axis, with the percent inhibitionby phagotopes 16 (j), 15 (r), 12(d), 7 (❉) and 1 (m) shown on they axis. Each inhibitor was testedin quadruplicate using 20, 10 and5 µL of the phagotope suspension and the average per-centage inhibition for the 5 µLpoints are plotted. The inhibitionwas dose dependent. Phagotopes15 and 16 strongly inhibited thereactivity of CII-C1 with all otherphagotopes, whereas phagotopes1, 12 and 7 gave only partialcross-inhibition with phagotopes15 and 16. Buffer (TE), f1, andphage clones lacking inserts werenon-inhibitory.

Figure 4 Alignment of the peptides displayed by phagotopesselected by CII-C1 antibody. The guide tree produced by thePILEUP algorithm is shown adjacent to the number andsequence of each peptide. Phagotopes that gave similar reactivi-ties to each other in the immunoassays were paired together onthe guide tree (e.g. 14 and 15, 9 and 12). The branch lengths areinversely proportional to the degree of similarity betweensequences. Amino acids that are common to the main motif arein bold and conservative substitutions are in italics. Most pep-tides in the main cluster c–g contained one or more basicresidues followed by a hydrophobic residue. In addition, the peptides in the main cluster contained variations of the motifFGxQ, which contributes to the binding of CII-C1 at the CDR3region.7 The two peptides 16 and 17, which were derived from acysteine-constrained library marked (C), did not cluster with theother peptides. Phagotopes expressing the same peptides as 1and 5 were isolated more than once from the library by CII-C1,as indicated.

reactivity of selected phagotopes in an exemplary system, theCII-C1 mAb and its defined epitope on native type II colla-gen, using multiple assay formats. Twelve phagotopesselected by CII-C1, which each displayed peptides that dif-fered to a varying extent, were assessed for their capacity toreact with CII-C1 by direct ELISA and to inhibit the reactiv-ity of CII-C1 with the cognate antigen type II collagen. Inaddition, a comparison was made of the immunoreactivity ofthe phagotopes with the relatedness of their peptidesequences as discerned by a multiple sequence alignmentalgorithm,15,16 and also with a model that was created of theC1 epitope on type II collagen.

The supposition may be made that the sequence of the phage-displayed peptide will dictate the capacity of thephagotope to react with the selecting antibody and that vari-ations in the sequences of the expressed peptides will explainthe differences in immunoreactivity of the derived phago-topes. However, phage display technology involves a bio-logical system that has intrinsic sources of variation. We haveused a library in which two copies of gVIII are present ineach bacterium that produces phage particles: a wild-typecopy on the helper-phage DNA and a recombinant gVIII onthe phagemid DNA.11 Thus, variation in the level of expres-sion of particular peptides may occur; this is estimated to bebetween 10 and 30% of the 2700 copies of pVIII display pep-tides.22,23 In the present study, the highly reactive phagotopesreached a similar level of reactivity at high concentrations ofCII-C1 (Fig. 1a), suggesting that a comparable level ofpeptide was in fact displayed on each of these phagotopes.We can also conclude that the low level of reactivity of CII-C1 with phagotopes 9 and 12 was not due to low peptideexpression, because the same phagotopes were highly reac-tive by other assays. Furthermore, we have previously shownthat different phage isolates expressing the same peptide andselected by CII-C1 more than once, phagotopes 1 and 7, had

the same reactivity with CII-C1 by direct ELISA.7 Thus, itcan be assumed that a major source of variation in immunore-activity observed in our studies is variation in binding capac-ity of different peptides with the selecting antibody.

As may be expected then, all 12 phagotopes selected byCII-C1 reacted with CII-C1 to varying degrees. The ability ofthe selected phagotopes to inhibit antibody reactivity with thecognate antigen is a measure of this. Our data using an inhi-bition ELISA indicated that binding of the phagotopes withCII-C1 was in every case at, or very near, the paratope site ofCII-C1. The likelihood is remote that inhibition by the phago-topes of the reactivity of CII-C1 with the C1 epitope was due to steric hindrance, because a synthetic peptide,RRLPFGSQM, which corresponds to the most frequentlyexpressed peptide (phagotope 1), completely inhibited CII-C1 reactivity with both type II collagen7 and other phago-topes. We therefore conclude that the phagotopes werecapable of functionally mimicking the C1 epitope for CII-C1,albeit to different extents.

The cross-inhibition ELISA tested whether phagotopes ofdifferent sequence compete with each other for binding CII-C1. It is interesting that the cysteine-constrained phago-tope 16, with the divergent peptide sequence C-IAP-KHRNSA-C, inhibited the reactivity of CII-C1 with all of theother phagotopes, each containing a variation of the RRL andFGxQ motifs. These data attest that phagotope 16, like theothers, functions as a mimic of the C1 epitope despite itsdivergent peptide primary structure. Furthermore, the reac-tivity of phagotope 16 with CII-C1 was inhibited by the synthetic peptide RRLPFGSQM.

We sought an objective procedure for analysing similarityof sequences of the expressed peptides to examine therelationship between immunoreactivity and amino acidsequence. The multiple sequence alignment algorithmPILEUP clusters sequences, either DNA or protein, into

JM Davies et al.488

Figure 5 The triple-helicalstructure of type II collagen isshown as a ribbon and stick model covering residues 359–367(ARGLTGRP), which encompassthe C1 epitope for CII-C1,ARGLT. Arginine (R) and leucine(L) of the C1 epitope for CII-C1are labelled. Two arginines fromseparate α chains and one leucineare juxtaposed on the surface ofthe type II collagen molecule inthe same plane. Many of the reac-tive phagotopes selected by CII-C1expressed peptides starting withthe sequence RRL or variationsthereof (see text).

groups based on evolutionary relatedness and aligns onesequence to the subsequent sequence in the cluster.16 PILEUPproved to be effective for clustering short peptides displayedby phagotopes selected by polyclonal antibodies from sub-jects with autoimmune diseases, when used with an aminoacid substitution matrix based on physicochemical similari-ties and a low penalty for gaps introduced to align the peptidesequences.15 In the present study, PILEUP aligned most ofthe sequences according to the presence of two motifs in thepeptide in the main cluster, these being the basic andhydrophobic residues at the amino terminus and the FG andQ at the carboxyl terminus. When the sequences of thephage-displayed peptides selected by CII-C1 are comparedwith the known structure of the C1 epitope, originallydeduced by Schulte et al.8 and modelled here in Fig. 5, itbecomes apparent that the residues of the peptide that mimicthe C1 epitope of collagen were RRL or similar residues atthe amino terminus. The carboxy terminus of the peptideFGxQ is likely to contribute specifically to binding to theCDR3 region of the CII-C1 paratope, because the major difference in the gene usage between the variable region ofCII-C1 and several other mAb that bind the same C1 epitopeoccurs in the CDR3 region.9,10 The importance of the FGxQ,and in particular of the glutamine, for binding CII-C1 isexemplified by the greater reactivity with CII-C1 of phago-tope 1 than phagotope 5, which lacks the glutamine.

We observed that pairs of sequences that were clusteredaccording to PILEUP mostly displayed comparableimmunoreactivity in our assay formats. However, there wasno absolute concordance between the clustering of theexpressed peptides according to their sequence and theobserved reactivity of the phagotopes with CII-C1. Onereason for this could be that only the amino-terminus RRL ofthe displayed peptide actually mimics the C1 epitope on collagen. The multiple sequence alignment approach ishighly useful for objectively grouping peptides into clustersof similar sequences. It must be seen as complementary to theuse of functional assays for interpretation of data for the fol-lowing reasons: (i) phagotopes that express peptides with adifferent primary sequence can also function as mimotopes ofthe epitope for a given antibody; (ii) the peptides themselvesmay be multifaceted; and (iii) the choice of assay format canaffect the immunoreactivity readout of phagotopes.

It is noteworthy that variation in reactivity occurredbetween different assays. There was an incomplete correla-tion of reactivity of CII-C1 with the phagotopes by directELISA and their ability to inhibit the binding of CII-C1 withtype II collagen. If several of these phagotopes had beentested only by direct ELISA, they would have been over-looked as candidate mimotopes. Petersen et al. have constructed a human p53 gene-specific phage-display libraryand similarly obtained phagotopes that were not reactive bya dot-blot assay, yet were reactive with the selecting antibodyby direct ELISA.5 Hence, we recommend the use of morethan one assay format for analysis of phagotopes derived byantibody screening of phage libraries.

We also noted that a difference in the spacing of aminoacids in some of the phage-displayed peptides did not appearto affect the reactivity with CII-C1, because phagotopes thatexpressed peptides with either a motif of FGQ (phagotopes14 and 15) or FGxQ (phagotopes 13 and 11) were among the

highly reactive phagotopes selected by CII-C1. This suggeststhat contiguity of the spacing of the amino acids in phage-displayed peptides need not be critical for antibody binding,such that some degree of flexibility of structure of expressedpeptides, and/or the variable region of the antibody, can existfor phage-displayed systems. The peptides expressed by thephage are inserted near the amino terminal of the pVIII coatprotein11 and so are free to adopt their own secondary struc-ture.24,25

In conclusion, in an exemplary system, we have found thatthe phagotopes selected by the mAb CII-C1 react with theselecting mAb, inhibit its reactivity with the cognate antigenand fit with a model created of the relevant epitope C1 ontype II collagen. Thus, the phagotopes selected by CII-C1,even one that expressed a widely divergent peptide sequence,were mimics of the C1 epitope. However, the phagotopesselected by CII-C1 did differ in their immunoreactivity withCII-C1, explicable by both variation in the peptide sequenceexpressed by the phagotopes and also the type of immuno-assay format used.

Acknowledgements

We thank Dr Andrew Cook for initially screening the librarywith CII-C1, Dr James Whisstock for preparation of thefigure of the collagen model and Dr Mark Myers for valuablediscussions. The CII-C1 hybridoma developed by Professor RHolmdahl and Dr K Rubin was obtained from the Develop-mental Studies Hybridoma Bank maintained by The Univer-sity of Iowa, Department of Biological Sciences, Iowa City,IA 52242, USA.

References

1 Valadon P, Scharff MD. Enhancement of ELISAs for screeningpeptides in epitope phage display libraries. J. Immunol. Meth.1996; 197: 171–9.

2 Levitan B. Stochastic modeling and optimization of phagedisplay. J. Mol. Biol. 1998; 277: 893–916.

3 Wilson DR, Finlay BB. Phage display: Applications, innova-tions, and issues in phage and host biology. Can. J. Microbiol.1998; 44: 313–29.

4 Rodi DJ, Janes RW, Sanganee HJ, Holton RA, Wallace BA,Makowski L. Screening of a library of phage-displayed peptidesidentifies human Bcl-2 as a taxol-binding protein. J. Mol. Biol.1999; 285: 197–203.

5 Petersen G, Song D, Hugle-Dorr B, Oldenburg I, Bautz EKF.Mapping of linear epitopes recognized by monoclonal anti-bodies with gene-fragment phage display libraries. Mol. Gen.Genet. 1995; 249: 425–31.

6 Burkhardt H, Holmdahl R, Deutzmann R et al. Identification ofa major antigenic epitope on CNBr-fragment 11 of type II col-lagen recognized by murine autoreactive B cells. Eur. J.Immunol. 1991; 21: 49–54.

7 Cook AD, Davies JM, Myers M, Mackay IR, Rowley MJ. Iden-tification of a mimotope to the type II collagen monoclonalantibody CII-C1 using phage display technology. J. Autoimmun.1998; 11: 205–11.

8 Schulte S, Unger C, Mo JA et al. Arthritis-related B cell epi-topes in collagen II are conformation-dependent and stericallyprivileged in accessible sites of cartilage collagen fibrils. J. Biol.Chem. 1998; 273: 1551–61.

Interpretation of phage-display data 489

9 Mo JA, Bona CA, Holmdahl R. Variable region gene selectionof immunoglobulin G-expressing B cells with specificity for adefined epitope on type II collagen. Eur. J. Immunol. 1993; 23:2503–15.

10 Mo JA, Holmdahl R. The B cell response to autologous type IIcollagen: Biased V gene repertoire with V gene sharing andepitope shift. J. Immunol. 1996; 157: 2440–8.

11 Felici F, Castagnoli L, Musacchio A, Jappelli R, Cesareni G.Selection of antibody ligands from a large library of oligo-peptides expressed on a multivalent exposition vector. J. Mol.Biol. 1991; 222: 301–10.

12 Luzzago A, Felici F, Tramontano A, Pessi A, Cortese R.Mimicking of discontinuous epitopes by phage-displayed peptides, I. Epitope mapping of human H ferritin using a phagelibrary of constrained peptides. Gene 1993; 128: 51–8.

13 Parmley SF, Smith GP. Antibody-selectable filamentous fdphage vectors, affinity purification of target genes. Gene 1988;73: 305–18.

14 Rowley MJ, Tait B, Mackay IR, Cunningham T, Phillips B. Col-lagen antibodies in rheumatoid arthritis, significance of anti-bodies to denatured collagen and their association with HLA-DR4. Arth. Rheum. 1986; 29: 174–84.

15 Davies JM, Scealy M, Cai Y-P, Whisstock J, Mackay IR, RowleyMJ. Multiple alignment and sorting of peptides derived fromphage-displayed random peptide libraries with polyclonal seraallows discrimination of relevant phagotopes. Mol. Immunol. inpress.

16 Feng DF, Doolittle RF. Progressive sequence alignment as a

prerequisite to correct phylogenetic trees. J. Mol. Evol. 1987;25: 351–60.

17 Tudos E, Cserzo M, Simon I. Predicting isomorphic residuereplacements for protein design. Int. J. Peptide Res. 1990; 36:236–9.

18 Sali A, Blundell TL. Comparative protein modelling by satis-faction of spatial restraints. J. Mol. Biol. 1993; 234: 779–815.

19 Nemethy G, Gibson KD, Palmer KA et al. Energy parameters inpolypeptides. 10 Improved geometrical parameters and non-bonded interactions for use in the ECEPP/3 algorithm, withapplication to proline-containing peptides. J. Phys. Chem. 1992;96: 6472–84.

20 Bernstein FC, Koetzle TF, Williams GJB et al. The protein data-bank: A computer-based archival file for macromolecular structure. J. Mol. Biol. 1977; 112: 535–42.

21 Davies JM. Molecular mimicry: Can epitope mimicry induceautoimmune disease? Immunol. Cell Biol. 1997; 75: 113–26.

22 Greenwood J, Willis AE, Perham RN. Multiple display offoreign peptides on a filamentous bacteriophage. J. Mol. Biol.1991; 220: 821–7.

23 Malik P, Terry TD, Gowda LR et al. Role of capsid structure andmembrane protein processing in determining the size and copynumber of peptides displayed on the major coat protein of fila-mentous bacteriophage. J. Mol. Biol. 1996; 260: 9–21.

24 Makowski L. Structural constraints on the display of foreignpeptides on filamentous bacteriophages. Gene 1993; 128: 5–11.

25 Smith GP, Petrenko VA. Phage display. Chem. Rev. 1997; 97:391–410.

JM Davies et al.490