Characterization of Oligopeptides That Cross-react with Carbohydrate-specific Antibodies by Real Time Kinetics, In-solution Competition Enzyme-linked Immunosorbent Assay, and Immunological Analyses*

Phage displaying random cyclic 7-mer, and linear 7-mer and 12-mer peptides at the N terminus of the coat protein, pIII, were panned with the murine monoclonal antibody, 9-2-L379 specific for meningococcal lipo-oligosaccharide. Five cyclic peptides with two sequence motifs, six linear 7-mers, and five linear 12-mers with different sequence motifs were identified. Only phage displaying cyclic peptides were specifically captured by and were antigenic for 9-2-L379. Monoclonal antibody 9-2-L379 exhibited “apparent” binding affinities to the cyclic peptides between 11 and 184 nm, comparable with lipo-oligosaccharide. All cyclic peptides competed with the binding of 9-2-L379 to lipo-oligosaccharide with EC50 values in the range 10–105 μm, which correlated with their apparent binding affinities. Structural modifications of the cyclic peptides eliminated their ability to bind and compete with monoclonal antibody 9-2-L379. Mice (C3H/HeN) immunized with the cyclic peptide with optimal apparent binding affinity and EC50 of competition elicited cross-reactive antibodies to meningococcal lipo-oligosaccharide with end point dilution serum antibody titers of 3200. Cyclic peptides were converted to T-cell-dependent immunogens without disrupting these properties by C-terminal biotinylation and complexing with NeutrAvidin®. The data indicate that constrained peptides can cross-react with a carbohydrate-specific antibody with greater specificity than linear peptides, and critical to this specificity is their structural conformation.

Numerous studies have demonstrated that oligopeptides, identified from antigenic regions of antibody idiotype sequences or by direct binding to anti-carbohydrate antibodies, can elicit cross-reactive antibody responses to bacterial polysaccharides (1). Such peptides have been termed conformational mimics of bacterial carbohydrates. Conformational peptide mimics have been identified for a number of pathogenic bacterial capsular polysaccharides, Neisseria meningitidis groups A (2), B (3), and C (4), group B Streptococcus type III (5), as well as for lipo-oligosaccharides from Brucella melitensis Rev1, Brucella abortus W99 (6), and N. meningitidis (7).
Conformational peptide mimics were first identified from peptide sequences within the antigen-binding sites of antiidiotypic antibodies (8) and subsequently by interaction with carbohydrate-specific mAbs (9 -11). By using peptide phage display technology, we previously identified two linear peptide mimics of meningococcal LOS 1 by binding to the bactericidal murine antibody, mAb 9-2-379, that elicited immunogenic responses in mice to LOS (7). However, many candidate conformational peptide mimics identified by these methods fail to elicit the required immunological cross-reactive responses (3). In this study more detailed characterization was undertaken of the specificity, binding affinity, and ability to compete with the binding of antibody to its nominal carbohydrate antigen to determine how well the peptides could mimic the LOS antigen prior to immunization experiments.
Antigen-antibody interactions depend not only on chemical interactions between the antibody epitope and antigen but also on the conformation of the antigen, because it has to complement surface topology present on the complementarity-determining region of the antibody. Peptides with a constrained structure would therefore be more likely to depend on conformation as well as chemical interactions to bind to the antibody paratope and act as better conformational mimics that elicit a cross-reactive antibody response to the bacterial glycan structure, whereas the more flexible linear peptides may only adopt a shape on docking to the antibody. To test this hypothesis phage libraries displaying structurally constrained peptides were panned in addition to linear peptides by direct interaction with an antibody. The libraries express random peptides as N-terminal fusions to phage coat protein pIII, which favors the selection of high affinity peptides, as only one to five copies of the pIII protein are expressed on the surface. As an appropriate model to test our hypothesis, we chose the outer membrane lipo-oligosaccharide present on several serogroups, including group B, of the opportunistic pathogen N. meningitidis. In the absence of a comprehensive vaccine meningococcal group B disease remains a major global health problem, and any conformational mimics identified in this study may be potential vaccine candidates.
Meningococcal LOS has been proposed as a vaccine candidate (12), but its mimicry to human glycosphingolipids (13) and the presence of endotoxin raises concerns about its safety as a vaccine component (14,15). Nevertheless, LOS has epitopes within the proximal oligosaccharide region that are conserved among several immunotypes (16) and are immunogenic in infants and children (17,18). Of the 12 immunotypes described in the literature (19), L3, L7, and L9 have almost identical structures (20,21) and are frequently associated with disease (22,23).
Panning phage libraries with mAb 9-2-L379 specific for the LOS3,7,9 immunotype identified six linear 7-mers and 12-mers and five cyclic 7-mer peptides. Phage capture assays revealed that only the cyclic peptides were specific for mAb 9-2-L379 and were immunoreactive with this mAb. Data are presented that indicate the apparent binding affinities of mAb 9-2-L379 to the cyclic peptides are comparable with the apparent binding affinity of the nominal antigen, LOS, and that their ability to compete for the binding of LOS is related to their binding affinity. Data are also presented that suggest the structure of the cyclic peptides is critical to their ability to cross-react with mAb 9-2-L379 and inhibit its binding to LOS. Initial immunization data show that the cyclic peptide with optimal apparent binding and EC 50 of competition elicited cross-reactive antibody responses to meningococcal LOS.

EXPERIMENTAL PROCEDURES
Unless otherwise stated, all chemicals were purchased from Sigma or BDH. Tryptone, yeast extract, and bacteriological agar were purchased from Oxoid.

Coliphage Libraries
The coliphage libraries Ph.D 7 (linear 7-mer random peptides), Ph.D 12 (linear 12 mer peptides), and Ph.D C7C (cyclic 7-mer peptides) were purchased from New England Biolabs. All three libraries express random peptides as an N-terminal fusion to coliphage coat protein pIII.

Purification of Monoclonal Antibodies
The monoclonal antibodies used in this study are 9-2-L379 (IgG2a) specific for lipo-oligosaccharide immunotype L3,7,9 and MN14C11.6 (IgG2a) specific for serosubtype P1.7 and MN4A8-B2 (immunotype L379). The ascites or hybridoma culture supernatants (supplied where necessary by the Large Scale Laboratory, National Institute of Medical Research, Mill Hill, UK) were dialyzed against three changes of 4 liters of 20 mM sodium phosphate buffer, pH 7.0 (phosphate buffer). mAbs were purified by protein G chromatography. The protein concentration of the dialyzed material was determined, after which it was stored in aliquots at Ϫ20°C until required. The purity of the mAbs was confirmed by SDS-PAGE (24), and the final antibody titer against LOS L3,7,9 was determined by ELISA (28).

Panning the Peptide Phage Display Libraries
One hundred microliters of mAb 9-2-L379 diluted to a concentration of 50 g ml Ϫ1 in coating buffer (100 mM NaHCO 3 , pH 8.6) was dispensed into the wells of 96-well microtiter plates (Nunc-Immuno TM Plate Maxi-Sorp TM Surface). The plates were covered to prevent drying and incubated overnight at 4°C. Wells were washed by three changes of 50 mM Tris-HCl, 150 mM NaCl, 0.1% v/v Tween 20, pH 7.5 (TBST). Two hundred microliters of BSA blocking buffer (coating buffer with 5 mg ml Ϫ1 bovine serum albumin (BSA; Sigma), pH 8.6) was added to each well and incubated for 60 min at room temperature. Wells were then washed 5 times with TBST. The coliphage library (10 12 pfu ml Ϫ1 ) was suspended in 100 l of TBST and dispensed into the wells coated with the panning antibody, mAb 9-2-L379, and incubated for 60 min at room temperature. Wells were either left uncoated or coated with an unrelated mAb as a negative control. They were washed 10 times with TBST prior to the addition of 100 l of elution buffer (200 mM glycine HCl, pH 2.2) to each well and further incubation for 10 min at room temperature. The eluted coliphages were pipetted into sterile tubes containing 15 l of neutralization buffer (1 M Tris-HCl, pH 9.0). Five microliters of each eluate was titrated to determine the number of pfu recovered after each round of panning. The phage eluate was amplified by transfecting Escherichia coli cells according to the manufacturer's instructions, and the amplified phage suspension was stored at 4°C.
Two further rounds of panning were undertaken as before, except 10 12 pfu ml Ϫ1 of the amplified phage stock prepared from the previous round was used instead of the initial coliphage library. The concentration of Tween 20 in the TBST was increased to 0.5% v/v for the subse-quent washing steps. The third round phage eluate was titrated, mixed 1:1 with 40% (v/v) glycerol, and stored at Ϫ20°C until required. The fourth round of panning was performed at 50, 5.0, and 0.5 g ml Ϫ1 of the target antibody 9-2-L379 coating the microtiter plate wells. The plates used for titrating the phage eluates following the third and fourth rounds of panning were also used as a source of phage for nucleic acid sequencing. Plates used for picking individual phage plaques for sequencing were not incubated for longer than 18 h. Coliphage display libraries and phage captured by the mAbs were titrated according to the methods described by the manufacturer, New England Biolabs. The phage DNA was extracted following a protocol recommended by the manufacturer, and the nucleotide sequence was determined from a cycle sequence reaction using the Ϫ96 universal primer provided with the library and Big Dye terminators (PerkinElmer Life Sciences) resolved on an Applied Biosystems Prism 377 automated sequencer.

Phage Capture Assay
Microtiter plate wells were coated overnight at 4°C with 100 l of mAb 9-2-L379, the unrelated anti-porin P1.7 mAb, MN14C11.6, or with BSA all diluted to 50 g ml Ϫ1 in 100 mM NaHCO 3 , pH 8.6. Wells were washed 5 times with TBST buffer and then blocked with 3% (w/v) skimmed milk in TBST for 2 h at room temperature. Wells were washed 5 times with TBST, and 10 10 phage particles of a specific test phage (10 11 pfu ml Ϫ1 in TBST) were dispensed into wells containing the target antigens. The individual phage clones were amplified to Ͼ10 12 pfu ml Ϫ1 and diluted to 10 11 pfu ml Ϫ1 in TBST. Microtiter plates were incubated for 1 h at room temperature with gentle rocking. Unbound phage were washed off by 10 changes of TBST. The bound phage were incubated for 10 min at room temperature with 100 l of 200 mM glycine HCl, pH 2.2. The phage were pipetted into sterile tubes containing 15 l of neutralization buffer (1 M Tris-HCl, pH 9.1). Five microliters of each phage eluent was titrated to determine the number of pfu recovered after capture. Each phage titration was performed in triplicate and expressed as the mean pfu Ϯ S.D. For negative controls phage capture was performed on random clones picked from the relevant peptide phage display libraries (Table III).

Peptide Synthesis
Peptides were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry, biotinylated by incorporation of biocytin amide, and cyclized by oxidation under controlled conditions to form an intramolecular disulfide bridge between the cysteine residues located at Cys 2 and Cys 10 (MedProbe, Oslo, Norway). The cyclic peptides were supplied after purification on high pressure liquid chromatography and determined to be Ͼ95% pure by mass spectrometry. The peptides were resuspended in PBS to a stock concentration of 10 mg ml Ϫ1 and stored at Ϫ20°C. For the determination of their binding affinity, peptides were immobilized to a biosensor cuvette coated with biotin and activated by addition of streptavidin. For peptide ELISAs microtiter plates were coated with avidin, streptavidin, or NeutrAvidin ® prior to addition of the peptide.

Structural Modification of the Consensus Cyclic Peptides
Reduction of Cyclic Peptides-Twenty microliters of each peptide stock (10 mM) was mixed with an equal volume of 500 mM dithiothreitol (DTT) and allowed to incubate overnight at room temperature. The reduced peptides were then stored at -20°C until required for use.
Acetylation of Cyclic Peptides-Twenty microliters of each peptide stock (10 mM) was mixed with an equal volume of 200 mM sodium acetate and then 20 l of acetic anhydride. The reaction mixture was incubated at room temperature for 1 h and then stored at -20°C until required for use.
Removal of the DTT or acetic anhydride from the peptides prior to storage was not necessary, because any traces of either reagent were removed during the thorough wash procedures after the peptides were immobilized to either the microtiter well or biosensor cuvette before the interaction with the mAb.
Factor Xa Digests-Reacti-Bind Maleic Anhydride plates (Pierce) were coated with 10 g ml Ϫ1 of the test peptides in carbonate buffer, pH 9.6, at 37°C overnight. The plates were washed with 1ϫ factor Xa cleavage buffer (Novagen, Inc) prior to the addition of 1 g ml Ϫ1 factor Xa protease in 1ϫ cleavage buffer. Samples were incubated at room temperature for 2 h; control samples had no protease added, but otherwise were treated identically. Samples were washed with PBS plus 0.05% (v/v) Tween 20, pH 7.2 (PBST), and then blocked with 3% (w/v) skimmed milk in PBST for 2 h at 37°C. Samples were washed with PBST prior to ELISA as described below.

ELISA
Peptide ELISA-96-Well microtiter plates (Nunc-Immuno TM Plate MaxiSorp TM Surface) were coated with NeutrAvidin ® (Pierce) by adding to each well 100 l of 5 g ml Ϫ1 NeutrAvidin ® solution in coating buffer (15 mM Na 2 CO 3 and 35 mM NaHCO 3 at pH 9.6) and incubating overnight at 4°C covered or placed inside a humidified chamber to prevent drying. The wells were washed (all washes unless otherwise stated were performed 5 times with PBST) and then blocked with 200 l per well of PBS with 3% (w/v) skimmed milk powder and 0.05% (v/v) Tween 20, pH 7.2 (blocking buffer). After 30 min incubation at 37°C, the wells were washed, and 100 l of the biotinylated peptides diluted to 2 g ml Ϫ1 in PBS was dispensed into each well and incubated for 30 min at 37°C. After washing 100 l of the primary antibody, mAb 9-2-L379 diluted (1:2000 -1:4000 according to the titer) in blocking buffer was added to each well and incubated for 60 min at 37°C. Wells were washed and 100 l of the secondary goat anti-mouse IgG conjugated to horseradish peroxidase, appropriately diluted in blocking buffer, was added and incubated for 60 min at 37°C. After washing 100 l of TM Blue (Intergen Co.) was added to each well and allowed to develop for 5-10 min at room temperature. The reaction was terminated by adding 100 l per well of 1 M H 2 SO 4 , and absorbance was measured at 450 nm.
Meningococcal LOS ELISA-This assay was performed according to the protocol of Charalambous and Feavers (7). Peptide-elicited crossreactive antibody response against LOS was expressed as end point dilution antibody titers.
Competitive Inhibition ELISA-mAb 9-2-L379 was diluted in blocking buffer (PBS with 3% (w/v) skimmed milk powder and 0.05% (v/v) Tween 20, pH7.2) so that when reacted against L3,7,9 LOS in an ELISA the absorbance values at 450 nm were between 0.75 and 1.0. The diluted mAb was incubated overnight at 4°C with 0 -62.5 M biotinylated peptides. As a negative control an irrelevant peptide was picked at random (Table III) and incubated with the mAb. Microtiter wells were coated with 0.3 g of LOS L3,7,9 conjugated to BSA diluted in coating buffer and incubated overnight at 4°C without drying. Wells were washed, and 200 l of the blocking buffer was added and incubated for 30 min at 37°C. After washing 100 l of mAb 9-2-L379 incubated with various concentrations of the peptides was added per well. Following incubation for 60 min at 37°C, the wells were washed, and 100-l aliquots of the secondary antibody appropriately diluted in the blocking buffer were added. The amount of secondary antibody bound was determined as described previously.
Inhibition data were fitted by non-linear regression analysis to the one-site competition equation: Y ϭ bottom ϩ (top Ϫ bottom)/(1 ϩ 10 (XϪLogEC50) ); where Y ϭ percent maximum binding and X ϭ log (peptide concentration) using software from GraphPad Prism ® (GraphPad Software Inc.). Each determination was the mean Ϯ S.D. of triplicate determinations.

Resonant Mirror Biosensor Analysis
The real time binding kinetics of the anti-LOS L3,7,9 mAb, 9-2-L379, and other control mAbs to the consensus peptides were determined at 25°C in PBST with a resonant mirror biosensor according to the manufacturer's instructions (Thermo Labsystems, formerly Affinity Sensors, Saxon Way, Barhill, Cambridge, UK). Biotinylated peptides were immobilized to biotin-coated biosensor cuvettes via a streptavidin bridge. Peptides that were previously treated with DTT or acetic anhydride were immobilized to the biosensor cuvette surface and then washed 10 times to remove all traces of these reactants. To ensure that similar levels of native and structurally modified peptides were immobilized to the biosensor cuvette, their interaction profiles as arc second response were monitored and compared. Because antibodies have two possible binding sites, only the initial monophasic part of the interaction profile was used to determine the on rate at each concentration of antibody from which the K a was determined. The dissociation rate constant (K d ) was determined from the mean of the off rates at various concentrations of the interacting mAb diluted to zero concentration (25). To ensure that rebinding effects were negligible, only the off rates in the presence of excess bound mAb were included (26). Kinetic data were analyzed by curve fitting software (FASTfit version 2.01) and interaction profiles overlaid with the FASTplot TM software provided by Affinity Sensors. The binding affinity is expressed as the equilibrium dissociation constant (K D ), calculated from the ratio of K d /K a .
The validity of the biosensor data was tested to determine whether monophasic or biphasic kinetics were observed. A meaningful K D was obtained when monophasic or first-order kinetics are observed, but if bi-phasic or second-order kinetics are observed then an apparent K D is derived that represents the avidity of the bivalent antibody to the polyvalent surface. The total amount of antibody bound (arc second response) to the biosensor cuvette was plotted as a function of antibody concentration, with no presumptions on whether it is first-or secondorder kinetics. From non-linear regression analysis of the saturation curves, we derived approximate K D values. The K D values from this plot were consistent with the K D values derived from the K a and K d values, indicating that the data are internally consistent with "pseudo" firstorder kinetics (27).

Mouse Immunization Studies with the C22 Cyclic Peptide
The C22 peptide (C22-bio, Table III) was complexed to NeutrAvidin ® in a molar ratio of 10:1 (peptide:Neutravidin ® ) by prior mixing at 37°C for 2 h. C3H/NeH mice were immunized with 20 g of the complex in 50 l of PBS mixed with an equal volume of either complete Freund's adjuvant, for the primary immunization, or incomplete Freund's adjuvant for the booster immunization 2 weeks after the primary. The control group of mice were immunized with 20 g of NeutrAvidin ® in 50 l of PBS and mixed with equal volume complete Freund's adjuvant or incomplete Freund's adjuvant or immunized with 50 l of PBS in either complete Freund's adjuvant or incomplete Freund's adjuvant. Mice were terminally bled 2 weeks after the booster injection, and the serum was aspirated from coagulated blood and frozen at Ϫ20°C until assayed individually by ELISA for total IgG antibodies with meningococcal LOS as target antigen.

Identification of Consensus
Peptides-Phage libraries displaying random linear 7-mer and 12-mer peptides and cyclic 7-mer peptides were panned by direct interaction with the anti-LOS L3,7,9 mAb 9-2-L379, and individual phage clones were picked following three or four rounds (Table I). To identify tight binders the fourth round of panning was performed with decreasing amounts of target antibody from 50 to 0.5 g ml Ϫ1 to increase the competition for binding. Higher consensus frequencies were observed with cyclic peptides compared with linear peptides. After four rounds of panning two cyclic peptide motifs were identified as follows: motif 1, SWXH(M/Q)PY, and motif 2, XT(L/I)GGYE. Motif 1 was represented by five peptides in the 3rd round and six peptides in the 4th round and was 59 and 97% of the total consensus peptides, respectively. Motif 2 was seen in three peptides (9%) in the third round and one peptide in the 4th round (6%). In contrast the linear libraries yielded different sequence motifs. The cyclic peptides have aromatic residues; motif 1 has a tryptophan at position 2 and a tyrosine at position 7, and motif 2 has only a tyrosine at position 6. Some of the linear peptides also contained aromatic residues (Table I). Because phage displaying the same cyclic peptide sequences could have arisen from either independent clones or from a single clone in earlier rounds and undergone mutation during subsequent rounds of panning, we compared the DNA sequences of the individual clones. Table II shows the amino acid and DNA sequences of individual phage clones encoding the two cyclic peptide motifs identified. SWFHMPY is derived from three different phage clones distinguished by synonymous base changes. All the other peptides in motif 1 had single DNA sequences. NTIGGYE and TTLG-GYE peptide sequences of motif 2 were evident in the fourth round, and two further peptide sequences containing the XTXGGYE motif were present in the third round, which were not identified in the 4th round. DNA sequence data from clones encoding the same linear consensus sequences were all identical (data not shown).
Phage Capture Assay-Phage particles displaying linear 7-mer and 12-mer and cyclic 7-mer peptides were enriched from ϳ10 5 ml Ϫ1 in the first round of panning to 10 9 ml Ϫ1 in the fourth round against the anti-LOS mAb 9-2-L379 (data not shown). To determine the specificity of the phage clones expressing the identified peptide sequences (Table I), the titer of phage clones captured by mAb 9-2-L379 was compared with the titer of phage captured by an irrelevant anti-meningococcal porin mAb MN411C.6 or a nonspecific protein BSA (Fig. 1). Phage expressing linear 7-mer (A) and 12-mer peptides (B) showed no significant differences between the number of phage captured by interaction with mAb 9-2-L379, MN411C.6, or BSA controls. In contrast, between 10 3 and 10 4 more phage expressing cyclic peptides were captured by mAb 9-2-L379 compared with the controls or when phage expressing a random cyclic peptide was used (C). These cyclic peptides were chemically synthesized for further detailed characterization (Table III).
Immunoreactivity of the Anti-LOS mAb 9-2-L379 to Phage Clones Displaying the Consensus Peptides and Synthetic Peptides-The antigenicity of the enriched phage clones against mAb 9-2-L379 was tested by ELISA. Phage clones expressing the identified linear 7-mer and 12-mer consensus peptides (Table III) showed no reactivity with mAb 9-2-L379, whereas phage clones expressing all the cyclic peptides (Table III) reacted strongly with mAb 9-2-L379 (data not shown). The reactivity of the synthetic cyclic peptides with mAb 9-2-L379 was then determined. Fig. 2 shows that the relative reactivities of each peptide to the mAb were as follows: C10 Ͼ C22 Ͼ C19 Ͼ B05 (B12 was not tested). The specificity of the anti-LOS mAb 9-2-L379 to the cyclic peptides was determined by reacting mAbs specific for other antigens including meningococcal capsular polysaccharide serogroups A-C and W135; no reactivity was evident (data not shown). In addition, no reactivity was seen with a another LOS L3,7,9 immunotyping mAb, MN4A8-B2, that recognizes a different LOS L3,7,9 epitope (data not shown).
Competitive Inhibition ELISA-To determine whether the anti-LOS mAb 9-2-L379 interacted with the consensus cyclic peptides by the same paratope as it interacts with LOS, insolution competitive inhibition ELISAs against meningococcal LOS L3,7,9 were performed. The reactivity data of the mAb preincubated with the peptides fitted to a one-site competition equation by a least squares fit method, converging for all data sets with R 2 values between 0.9973 and 0.9999. No inhibition was observed in the presence of a random peptide (P-CRC). Table IV shows the calculated EC 50 values, which ranged from 9.6 M for C22-bio to 105.2 M for C19-bio.
Real Time Binding Kinetics of mAb 9-2-L379 to the Consensus Cyclic Peptides-Increasing concentrations of mAb 9-2-L379 were added to the resonant mirror biosensor cuvette, and the binding interaction between the mAb and the immobilized peptides was monitored in real time as a continuous change in arc second response. Table V shows the association and dissociation rate constants and the K D for each of the cyclic peptides. The binding affinities ranged from 10.9 nM for peptide C10-bio TABLE II Amino acid and DNA sequences of phage clones identified by panning against the anti-LOS L3,7,9 mAb 9 -2-L379 The DNA sequences and the frequency of each sequence were determined after each round of panning. The fourth round of panning was performed with different concentrations of the target antibody, mAb 9 -2-L379, to identify high avidity binders. to 184 nM for peptide C19-bio. If mAb 9-2-L379 interacts with the cyclic peptides via the same paratope as it binds to meningococcal LOS, the affinity of binding to each peptide should correlate with their ability to compete with the binding of mAb 9-2-L379 to LOS. Linear regression analysis on the K D and the EC 50 values was performed and revealed a correlation coefficient (R 2 ) value of 0.9277 (p Ͻ 0.0084).
Structural Modifications of the Consensus Cyclic Peptides-The importance of conformation on the ability of the cyclic peptides to interact with mAb 9-2-L379 was studied by a series of structural modifications. Dissociation of the disulfide bond constraining the cyclic structure of the peptides by reduction with DTT led to a time-dependent loss in mAb 9-2-L379 binding, with no further decrease seen after 30 min at room temperature (data not shown). Fig. 3A shows the reactivity of mAb 9-2-L379 to the native and DTT-reduced biotinylated cyclic peptides. Any trace of DTT remaining when the peptides were stored was completely removed by thorough washing after the peptides were coated to the microtiter plates. There was a loss of reactivity with mAb 9-2-L379 to B05-bio of 64%, to B12-bio of 94%, to C10-bio of 70%, to C19-bio of 90%, and to C22-bio of 98%.
In addition to the constraints imposed by the disulfide bond, modifications to the N-terminal end of the cyclic peptides also affected antibody binding. Substitution of the terminal amino group with an acetyl group resulted in complete loss of reactivity of mAb 9-2-L379 to the cyclic peptides (Fig. 3B). Any trace of acetic anhydride was completely removed by washing after the peptides were coated to the microtiter plate well.
To determine whether additional amino acids at the N terminus of the cyclic peptides could also influence the interaction of mAb 9-2-L379, peptides C10 and C22, representing motifs 1 and 2, respectively, were synthesized with the factor Xa cleavage site IIEGR (single letter amino acid code) at the N terminus (Table III). This would allow the mAb reactivity to be tested on the same peptides with and without an N-terminal modification. Fig. 3C illustrates that the addition of these amino acid residues resulted in complete loss of reactivity to mAb 9-2-L379 and that the reactivity of mAb 9-2-L379 to both cyclic peptides was restored by the removal of the additional amino acids by factor Xa protease digestion.
To establish whether acetylated and DTT-reduced peptides were able to inhibit the binding of mAb 9-2-L379 to meningococcal LOS, competitive ELISAs were performed, making certain to remove all traces of these reagents prior to interaction with the antibody. The modified cyclic peptides were no longer able to inhibit antibody binding to LOS up to a concentration of 160 M (Table IV).
The effects of these structural modifications on the real time interaction of mAb 9-2-L379 were studied on a resonant mirror biosensor. Our previous study (7) had shown that a measurable binding affinity between mAb 9-2-L379 and linear peptides could be observed in the absence of reactivity by ELISA. Again all traces of the DTT or acetic anhydride were removed by the wash procedures following immobilization of the peptides to the biosensor cuvette. In addition, to ensure that the modified peptides were immobilized to the biosensor at a similar level to the native peptides, the arc second responses were monitored during the immobilization procedure (data not shown). Fig. 4 shows the association profiles of the C10-bio, C19-bio, and C22-bio peptides. Acetylation of the cyclic peptides resulted in complete loss of mAb 9-2-L379 reactivity, as assessed by ELISA, to all the peptides, but some association of mAb 9-2-L379 was observed by resonant mirror analysis to the peptides, in particular to C19-bio (Fig. 4B). The DTT-reduced peptides had weak association profiles observed by biosensor analysis of mAb 9-2-L379 with C10-bio and C19-bio peptides, and the C22-peptide maintained about 50% of its association with mAb 9-2-L379.
Mouse Immunization Studies with the C22 Cyclic Peptide-Although peptides C10 and C22 had comparable reactivities in ELISA as well as similar apparent binding affinities based on real time kinetic measurements, C22 was chosen for initial immunogenicity studies because data from competition ELISA revealed that it was the better inhibitor in-solution with an EC 50 value of 9.64 M, 2.4-fold better than C10 (Table IV). Five C3H/HeN mice were immunized with C22 peptide complexed to NeutrAvidin ® , and five C3H/HeN mice were immunized with  (Table I) were aliquoted to the coated microtiter wells and incubated for 60 min at room temperature. After washing the wells 10 times to remove unbound phage, the bound phage were eluted by acidification. Phage suspensions were neutralized and titrated to determine the concentration of pfu ml Ϫ1 . Each determination was performed in triplicate, and the mean Ϯ S.E. is shown. RC is a random clone picked from each peptide phage display library as a negative control.
NeutrAvidin ® alone. Fig. 5 shows the cross-reactive antibodies elicited to meningococcal LOS with an end point antibody of 3200.

DISCUSSION
In this study we characterized linear and cyclic peptides that cross-react with an antibody specific for a bacterial carbohydrate by phage capture assays, ELISA, measurement of apparent binding affinities, and in-solution competition to assess their specificity. The importance of conformation to the binding of antibody was confirmed by structural modifications of the cyclic peptides. Initial immunization studies with one of the cyclic peptides identified in this study revealed that it induced cross-reactive antibodies to a bacterial glycan antigen and is defined as a conformational mimic of a carbohydrate epitope.  (Table III) were immobilized to NeutrAvidin ® -coated microtiter plate wells, which oriented the peptides in the same way as they were displayed at the N terminus of the coliphage minor coat protein pIII. Positive control wells were coated with meningococcal LOS L3,7,9 conjugated to BSA by standard ethyl dimethylaminopropyl carbodiimide chemistry. Each determination was performed in triplicate and the mean Ϯ S.E. shown. Key: C10-bio, OE; C22-bio, q; C19-bio, ࡗ; B05-bio, ; NeutrAvidin ® control, Ⅺ.
Acetyl-IIEGRACNTIGGYECGGGSK-acid a Peptides are displayed at the exposed N terminus of the coliphage minor coat protein glycoprotein III. b Random clone (RC) negative control. c Biocytin amide. d Acetylated N terminus. e Unmodified C terminus.

TABLE IV
The inhibitory potency (EC 50 ) of cyclic, reduced, and acetylated peptides on the binding of mAb 9 -2-L379 to meningococcal LOS L3,7,9 Cyclic, DTT-reduced, and acetylated peptides were preincubated with the anti-LOS mAb 9 -2-L379 prior to reacting with meningococcal LOS conjugated to BSA and coated onto microtiter plate wells. No inhibition (NI) was observed up to a concentration of 125 M peptide. P-CRC is a negative clone picked at random.

TABLE V Real time binding kinetics of monoclonal antibody 9 -2-L379
Real time binding kinetics between mAb 9 -2-L379 and cyclic peptides were measured with a resonant mirror biosensor. C-terminal biotinylated peptides were immobilized via a streptavidin bridge to a resonant mirror biosensor cuvette coated with biotin. The K a was derived from the slope of the correlation line of the on rates and antibody concentration. The K d was determined from the mean of the off rates from several concentrations of the antibody bound to the peptides and diluted to zero concentration. To reduce rebinding effects, only the off rates in the presence of excess concentrations of bound antibody were included. The K D value was derived from K d /K a . mAb 9 -2-L379 concentrations from 0.78 to 400 nM were interacted with the immobilized cyclic peptides; n ϭ 12. As group B meningococcal disease is a public health priority, an antibody specific for and with high apparent affinity to an immunogenic region of the meningococcal outer membrane LOS, mAb 9-2-L379, was chosen for these studies. Following three or four rounds of panning by direct interaction with mAb 9-2-L379 six 7-mer, five 12-mer linear peptides with different sequences from each other and from sequences identified in our previous study (7), and five cyclic peptides with two sequence motifs were identified. Although linear and cyclic consensus sequences were identified, phage capture assays and ELISA revealed that only cyclic peptides interacted specifically with mAb 9-2-L379. As we postulated that peptides with a constrained structure might elicit greater antibody responses than linear (flexible) peptides, the cyclic peptides were further characterized.
The nucleotide sequences of phage clones encoding linear and cyclic consensus peptide sequences revealed that only clones encoding the predominant cyclic peptide motif, ACSWL-HQPYC, had synonymous base changes. This indicated that enrichment of this cyclic peptide sequence was more likely to have occurred from independent clones, although enrichment of different clones that arose from mutations of a single clone and expanded in an earlier round cannot be excluded. Whichever is the case, this implies that phage expressing this peptide have been positively selected in the panning process. mAb 9-2-L379 had the greatest reactivity to the cyclic peptides C10 and C22 with antibody titers Ͼ80,000. Peptides C22 and C10 were the most potent inhibitors of mAb 9-2-L379 binding to its nominal antigen, LOS. The apparent binding affinities of mAb 9-2-L379 to both these cyclic peptides were comparable with its apparent binding affinity to LOS (25). The cyclic peptides inhibited the binding of mAb 9-2-L379 to LOS with single-site inhibition kinetics, indicating that the peptides interact very close to or at the same paratope as LOS.
The apparent K D values for the peptides were between 500and 1000-fold lower than their inhibitory EC 50 values, which suggests that even with the precautions outlined under "Experimental Procedures" to measure only the monophasic binding interaction and to limit rebinding effects, the biosensor is measuring avidity which overestimates the affinity of mAb 9-2-L379 to both the peptides and LOS. As peptide or LOS molecules are likely to be immobilized in close proximity on the biosensor surface, and antibodies are bivalent, they are likely to bind to more than one peptide or LOS molecule. The overestimation of binding affinity compared with competition studies performed in solution agrees with the findings of Nieba et al. (26). The "true" monovalent interaction between the antibody paratope and peptide or LOS is likely to be closer to the micromolar EC 50 values determined by solution competition ELISA. However, the relative binding affinities of each cyclic peptide is valid as the apparent K D and the EC 50 values were positively correlated and established that C10 and C22 peptides were the optimal antigenic mimics identified in this study.
Structural modifications of the cyclic peptides revealed that the N terminus and tertiary structure are critical to their antigenicity and ability to inhibit binding of mAb 9-2-L379 to meningococcal LOS. Weak binding interactions of the reduced and acetylated peptides were observed by resonant mirror biosensor that may be due to the sensitivity of analysis compared with a fixed end point in ELISA assays. This binding did not directly correlate to the reactivity of mAb 9-2-L379 observed by ELISA and may be due to the detection of either a small subset of peptides that were not modified or the slow on rate produced when the reduced, unrestrained peptides adopt the correct structure on association with the antibody paratope. Similarly, the biosensor may be detecting either a small population of non-acetylated peptides or the weak interaction of acetylated peptides to the antibody. Further analysis would be required to determine the binding affinities of the modified peptides. Preliminary, structural studies by solution NMR show that the C22 peptide (motif 2) has a ␤-hairpin loop structure (45), with the N-terminal alanine forming an intramolecular interaction with the glycine adjacent to the cysteine residue that is critical to the peptide's mimicry of LOS and binding to the antibody paratope.
Immunization studies were initiated with the cyclic peptide C22 in C3H/HeN mice. Chemical conjugation to a carrier protein, as in our earlier immunization studies (7) with linear peptides, eliminated the antigenicity of the cyclic peptide. So a method was developed for complexing the C22 peptide to NeutrAvidin ® as carrier protein. This method maintained the FIG. 3. The reactivity of the anti-LOS mAb 9-2-L379 to native and structurally modified consensus peptides. The interaction of the anti-LOS mAb 9-2-L379 to native and structurally modified consensus peptides was determined by ELISA. Biotinylated peptides were immobilized to NeutrAvidin ® -coated microtiter wells, which oriented the peptides in the same way as they were displayed at the N terminus of the coliphage minor coat protein pIII. Prior to the interaction with the monoclonal antibody, the peptide-coated wells were washed thoroughly to remove any trace of the structure-modifying agents. A shows the interaction of the anti-LOS mAb 9-2-L379 to native (Ⅺ) and DTT reduced (f) consensus peptides. B shows the interaction of the anti-LOS mAb 9-2-L379 to native (Ⅺ) and N-terminal acetylated (f) consensus peptides. C shows the interaction of the anti-LOS mAb 9-2-L379 to consensus peptides with an N-terminal extension containing a factor Xa cleavage site (Table III), before (Ⅺ) and after factor Xa protease digestion (f). The data are the mean Ϯ S.E. of triplicate determinations. reactivity of mAb 9-2-L379 to the cyclic peptide prior to vaccination. The end point serum antibody titer of the cross-reactive response elicited by this peptide to meningococcal LOS was 3200.
The linear peptide consensus sequences identified in this study were different from those characterized in our previous investigation (7). There are two possible explanations for this: first, the poor binding specificity of mAb 9-2-L379 to linear peptides revealed in the present study, and second, an alternative panning protocol was employed. mAb 9-2-L379 interacted with the two linear peptides identified in our previous study with lower apparent binding affinities (419 and 488 nM) compared with the LOS antigen (Table V) and the cyclic pep-tides identified in this study (7,25). In addition, competition of mAb 9-2-L379 binding to LOS by these two linear peptides was only observed by real time kinetic analysis (7). Nevertheless, when these linear peptides were conjugated to diphtheria toxoid CRM 197 they elicited cross-reactive antibody responses against meningococcal LOS with end point dilution antibody titer of 800 (7), 4-fold lower than the C22 cyclic peptide identified and characterized in this investigation. One possible reason why these linear peptides elicited a smaller antibody response against LOS than C22 may be due to their low apparent binding affinity to mAb 9-2-L379, resulting from their slow on rate for binding. This would be expected if the linear peptides had to adopt the correct conformation on interaction with the antibody. Moreover, a flexible peptide structure could adopt a number of possible conformations, not all of which would elicit a cross-reactive antibody response. We postulated therefore that screening constrained peptides with a defined structure might identify epitope mimics with better binding affinities that could be developed into more effective immunogens for eliciting a response against LOS.
The screening of random peptides displayed on coliphage with mono-specific antibodies is an effective method for identifying conformational mimics of pathogenic bacterial carbohydrate structures that can elicit cross-reactive antibodies to the carbohydrate antigen (6, 7, 29 -36). To favor the binding of high affinity peptides they can be displayed at the N terminus of the phage minor coat protein, pIII, at 1-5 molecules per phage. Alternatively, for efficient immunogenicity to elicit specific antibodies or for presenting a T-helper epitope, peptides can be displayed as an N-terminal fusion of the major coat protein, pVIII, which can display from 270 to 2,700 copies (37,38). The 7-mer linear and cyclic phage libraries used in this investigation consist of 2.0 ϫ 10 9 and 3.7 ϫ 10 9 independent clones, respectively. They are sufficiently complex to contain most of the possible heptameric peptide sequences. In contrast the 12-mer peptide phage library used has only 1.9 ϫ 10 9 independent clones, which represents only a very small number of the potential peptide sequences (20 12 ϭ 4.1 ϫ 10 15 ). However, the increased length may allow peptides to fold into small structural elements that may be necessary for binding to the target and identify peptides that require more than seven residues for tight binding. FIG. 4. Real time interaction profiles of the anti-LOS mAb 9-2-L379 to native and structurally modified cyclic peptides. The effects of structural modifications on the interaction profiles of the cyclic peptides to mAb 9-2-L379 were observed with a resonant mirror biosensor. Biotinylated peptides were immobilized via a NeutrAvidin ® bridge to biotin-coated biosensor cuvettes. To the cuvette 100 nM mAb was added, and the binding was monitored as an arc second response. The unmodified native peptide interaction profiles (N) were compared with the N-terminal acetylated peptides (Ac), and the DTT-reduced peptides (R). A-C show the C10, C19, and C22 peptide interaction profiles, respectively.
FIG. 5. Cross-reactive serum antibody responses to meningococcal LOS. C3H/NeH mice were immunized with 20 g of C22 peptide complexed to NeutrAvidin ® in 50 l of PBS and mixed with an equal volume of Freund's Complete Adjuvant and then boosted 2 weeks later with the same immunogen mixed with Freund's Incomplete Adjuvant (q). Mice were immunized with 20 g of NeutrAvidin ® in 50 l of PBS and mixed with an equal volume of Freund's Complete Adjuvant and then boosted 2 weeks later with the same immunogen mixed with Freund's Incomplete Adjuvant (E). Mice were terminally bled 2 weeks after the booster immunization, and serum was carefully aspirated from the coagulated blood. Each point represents the mean Ϯ S.E. of five C22 immunized mice and five NeutrAvidin ® control mice. S.E. values Յ5% of the mean are not shown.
The presence of a constrained structure in peptide mimics of carbohydrates has been demonstrated indirectly, where two linear peptide mimics of the LPS of Shigella flexneri serotype 5a, flanked by two cysteine residues, were identified from linear libraries (36). The two cysteines could form a disulfide bridge and constrain the peptides into specific conformations. Peptide mimics of Brucella LPS were identified that contained a pair of cysteines, which produced the greatest reactivity by ELISA to the anti-Brucella LPS mAb B66-268 (6). In general, when cyclic peptides have been used to mimic antigens, it has been shown that their binding affinities are greater than linear peptides (39,40). Furthermore, cyclic peptides may also elicit better immune responses compared with linear peptides (41). These observations are in agreement with our findings that show that the constrained peptide mimics identified in this study had apparent binding affinities up to 46-fold better than the previously identified linear peptides, and effectively inhibited binding of mAb 9-2-L379 to LOS.
Peptide mimics of bacterial carbohydrates have been shown to contain aromatic residues in specific motifs (8,32,42,43), but the cyclic peptides identified in this study did not contain the same motifs, although motif 1 has two aromatic residues and motif 2 has one. Motif 1 in this study has a proline adjacent to the tyrosine (ACSWXH(M/Q)PYC), similar to the DRPVPY peptide that binds to group A Streptococcus cell wall polysaccharide-binding site (44). The presence of synonymous DNA changes present in different phage clones expressing motif 1 observed in this study indicate that the aromatic residues in addition to other residues are important for interacting with mAb 9-2-L379. Our findings appear to confirm other published studies (8,32,42,43) that conformational mimics of carbohydrate antigens may interact with the antibody complementarity-determining regions partly via aromatic residues.
Further studies are underway to compare the immune responses of the other four cyclic peptides and to study different methods of presenting these peptides to further enhance and modify the immune responses elicited.