Directing the Immune Response to Carbohydrate Antigens*

Peptide mimetics may substitute for carbohydrate antigens in vaccine design applications. At present, the structural and immunological aspects of antigenic mimicry, which translate into immunologic mimicry, as well as the functional correlates of each, are unknown. In contrast to screening peptide display libraries, we demonstrate the feasibility of a structure-assisted vaccine design approach to identify functional mimeotopes. By using concanavalin A (ConA), as a recognition template, peptide mimetics reactive with ConA were identified. Designed peptides were observed to compete with synthetic carbohydrate probes for ConA binding, as demonstrated by enzyme-linked immunosorbent assay and isothermal titration calorimetry (ITC) analysis. ITC measurements indicate that a multivalent form of one particular mimetic binds to ConA with similar affinity as does trimannoside. Splenocytes from mimeotope-immunized mice display a peptide-specific cellular response, confirming a T-cell-dependent nature for the mimetic. As ConA binds to the Envelope protein of the human immunodeficiency virus, type 1 (HIV-1), we observed that mimeotope-induced serum also binds to HIV-1-infected cells, as assessed by m M ) and concentration of ConA ranging from 0.025 to 0.2 m M were added from the computer-controlled microsy- ringe, at an interval of 4 min, into the lectin solution (cell volume (cid:2) 1.3582 ml), while stirring at 350 rpm, at 27 °C. Both lectin and peptide were dissolved in 100 m M HEPES buffer (pH 7.2) or 100 m M sodium acetate buffer (pH 5.2), containing 5 m M CaCl 2 and MnCl 2 . Control experiments were performed by making identical injections of peptide into a cell-containing buffer, where protein showed insignificant heats of dilution. The experimental data were fitted to a theoretical titration curve using software supplied by Microcal Software. The quantity c (cid:2) K a M t (0), where M t (0) is the initial macromolecule concentration, is of importance in titration calorimetry. All experiments were performed with c values 1 (cid:5) c (cid:5) 200. The instrument was calibrated using the calibration kit containing RNase A and 2 (cid:6) -CMP, supplied by the man-ufacturer. Thermodynamic parameters were calculated from the equa- tion, (cid:7) G (cid:2) (cid:3) RT ln K a , where K a and (cid:7) G are the association constant and changes in free energy, respectively; T is the absolute temperature, and R (cid:2) 1.98 cal mol (cid:3) 1 K (cid:3) 1 .

conjugation strategies that elicit carrier-specific T-and B-cell responses do not necessarily enhance PS immunogenicity (7) nor do PS conjugates elicit responses in immunodeficient mice. Furthermore, in cases where a large number of carbohydrate antigens are required to afford protection, much like that representative of the large number of pneumococcal carbohydrate serotypes, PS conjugates will be far more complicated to produce (6).
Immunization with peptide mimetics of carbohydrate antigens can overcome the T-cell-independent nature of the immune response (8 -12). Peptide antigens have an absolute requirement for T cells that can mediate memory responses upon carbohydrate boosting (11,13). In contrast to carbohydrate conjugates, peptide mimetic conjugates can facilitate cognate interactions between B and T cells after immunization of immunodeficient mice that lack Bruton's tyrosine kinase (10). Peptide mimetics therefore afford a vaccine approach to break tolerance to carbohydrate self-antigens (13).
Whereas peptide library screening has led to the identification of a variety of peptide mimetics of carbohydrate antigens (14,15), concepts described for the design of small molecules may apply equally well to the design of mimetics of carbohydrate antigens (16,17). To further facilitate concepts for a structure-assisted vaccine design, we considered, as a model system, small molecule interactions with concanavalin A (ConA). The mannose/glucose-specific lectin ConA is the most extensively studied plant lectin, known for its application as a biochemical tool and as a model protein to gain further knowledge about lectin-ligand interactions (18 -20). Lectins are also particularly relevant to human immunodeficiency virus (HIV) pathogenesis. Lectin-induced inhibition of syncytium formation and infection of cells by both T-cell line adapted and primary isolates (21)(22)(23)(24)(25)(26)(27) focuses attention on oligomannosidic glycans, such as those characterized by interaction with ConA ( Fig. 1). More recently the lectin DC-SIGN, expressed on dendritic cells, has been recognized to participate in facilitating HIV transmission (24,28,29). Consequently, defining peptide mimetics reactive with ConA may facilitate the development of immunogens to augment carbohydrate responses to HIV in future vaccine applications.
Structural studies of peptidyl-ConA complexes suggest that the carbohydrate-binding site on ConA can accommodate an extended array of carbohydrate antigens that might lend to its biological properties (30,31). We have further defined a peptide that binds at or near the carbohydratebinding site of ConA that displays a free energy of association comparable to those reported for core trimannoside-ConA and pentasaccharide-ConA interactions. The designed peptide elicits a robust thymus-dependent response, stimulating splenocytes from peptide-immunized mice. We observe that the peptide, rendered as a multiple antigenic peptide (MAP), used to emulate the clustered array of Envelope protein of HIV (Env)-associated carbohydrates (32) induced an antibody response reactive with cell-bound Env protein. We observe that serum induced to the peptide mimetic paralleled neutralization results obtained by using a mannan preparation from Saccharomyces cerevisiae or from Candida albicans (33,34), which further suggests that carbohydrate cross-reactive responses induced by peptide mimetics might be rendered even more effective immunogens.

EXPERIMENTAL PROCEDURES
Epitope Mapping of ConA Ligand-binding Site-By using the crystallographically positioned pentasaccharide structure within the ConAcombining site, we implemented the program Ligand-Design (LUDI (35), Micron Separations/Biosym Technologies), as described previously (36 -39), to search a fragment library and identify amino acid residue types able to interact with ConA. This program identifies small molecular fragments in a data base and then docks them into the proteinbinding site in such a way that hydrogen bonds and ionic interactions can be formed between the protein and the molecular fragments. The positioning of the small fragments is based upon rules about energetically favorable non-bonded contacts and on geometry between functional groups of the protein and the ligand. The center of search was defined using the crystallographic position of the central mannose residue and searching 15 Å surrounding the centroid of the sugar for potential contact sites on ConA.
The search was performed using standard default values and a fragment library supplied with the program. Peptides were built using INSIGHTII (Micron Separations/Biosym Technologies) and accommodated in relation to the docked LUDI fragments. The peptide backbone and side chain torsion angles were rotated using a fixed docking algorithm (affinity program) within INSIGHTII, until the side chains of the peptide were approximated to the corresponding LUDI fragments. The peptide-ConA complex was subjected to energy optimization and molecular dynamic simulations, as described previously (36 -39).
BALB/c mice (n ϭ 4 per group), 4 -6 weeks of age, were immunized intraperitoneally three times, at intervals of 2 weeks, with 100 g of a respective peptide or 50 g of LeY, each combined with 20 g of the adjuvant QS21 (Aquila Biopharmaceuticals Inc., Framingham, MA). A FIG. 1. Representative carbohydrate antigen structures expressed on gp120, as defined by specific molecular reagents (antibodies and lectins). Monoclonal antibodies reactive with sialyl-Tn, LeY, and A1 neutralize HIV infection in vitro (65). Lectins reactive with mannose-containing carbohydrates also the display ability to block infection.
control group was immunized with QS21 alone. The LeY-expressing cell line MCF7 (ATCC, Manassas, VA) (40), without adjuvant, was also used to immunize groups of mice four times. Serum was collected at days 7 and 14 after the last immunization and stored at Ϫ80°C until use.
ELISA-ELISAs were performed as described (41). Immulon-2 plates were coated overnight, at 4°C, with 100 mM of a selected peptide or carbohydrate probe to assess the binding of sera to these antigens. After blocking the plates (PBS, 0.5% FCS, and 0.2% Tween 20), serial dilutions of sera were added and resolved with anti-mouse isotypematched HRP (Sigma). To assess the binding of ConA to carbohydrates and peptides, serial concentrations (10 to 0.6 ng/ml) of ConA biotinlabeled (Sigma) were added to pre-coated plates and reacted with streptavidin-HRP (Sigma). All results were calculated from triplicate measurements.
Peptide-Carbohydrate Competition Assay-Immulon-2 plates precoated with 20 g/ml peptide were blocked. The binding of ConA (0.4 g/ml) biotin-labeled (Sigma) to peptides was assessed in the presence of serial concentrations of Me␣Man (0.8 -50 mM). Control wells with ConA, but not Me␣Man, were also run. Plates were reacted with streptavidin-HRP (Sigma) and results calculated from triplicate measurements. Percentage of inhibition of ConA binding to peptides was calculates as 1 Ϫ (mean of test well/mean of control wells) ϫ 100.
Isothermal Titration Calorimetry-Isothermal titration calorimetry (ITC) was performed using an MCS isothermal titration calorimeter (Microcal Software, Northampton, MA). In individual titrations, injections of 5 l of peptides (0.4 -5.2 mM) and concentration of ConA ranging from 0.025 to 0.2 mM were added from the computer-controlled microsyringe, at an interval of 4 min, into the lectin solution (cell volume ϭ 1.3582 ml), while stirring at 350 rpm, at 27°C. Both lectin and peptide were dissolved in 100 mM HEPES buffer (pH 7.2) or 100 mM sodium acetate buffer (pH 5.2), containing 5 mM CaCl 2 and MnCl 2 . Control experiments were performed by making identical injections of peptide into a cell-containing buffer, where protein showed insignificant heats of dilution. The experimental data were fitted to a theoretical titration curve using software supplied by Microcal Software. The quantity c ϭ K a M t (0), where M t (0) is the initial macromolecule concentration, is of importance in titration calorimetry. All experiments were performed with c values 1 Ͻ c Ͻ 200. The instrument was calibrated using the calibration kit containing RNase A and 2Ј-CMP, supplied by the manufacturer. Thermodynamic parameters were calculated from the equation, ⌬G ϭ ϪRT ln K a , where K a and ⌬G are the association constant and changes in free energy, respectively; T is the absolute temperature, and R ϭ 1.98 cal mol Ϫ1 K Ϫ1 .
Precipitation Study-Measured volumes of known concentrations of lectin and peptide solutions (in 100 mM HEPES buffer (pH 7.2) containing 150 mM NaCl, 5 mM CaCl 2 , and MnCl 2 ) were mixed in a quartz cuvette, at room temperature, and the time-dependent development of turbidity was measured at 420 nm (42). Absorbances were monitored continuously, until they remained constant. A portion of the precipitate was treated with 400 mM Me␣Man, to check whether or not the precipitation was due to the binding of the peptide at the carbohydratebinding sites of the lectin. Absorbancy of the solution was recorded at 420 nm, before and after the addition of Me␣Man.
Cell Proliferation Assay-Spleens were aseptically removed and splenocytes, as the responder cells, isolated by lysis of erythrocytes. Responder cells were used for detection of cell proliferation using Cell-Titer 96 R Aqueous One Solution (Promega, Madison, WI), based on the manufacturer's instructions. Briefly, cells (2.5 ϫ 10 5 /well) were cultured in a flat-bottomed 96-well plate with 106-MAP, 911-MAP, or only medium RPMI 1640 (Life Technologies, Inc.) supplemented with 5% heat-inactivated fetal calf serum (FCS), 1% L-glutamine, 100 IU/ml penicillin, and 100 g/ml streptomycin. At the 3rd day of incubation, the provided solution was added to each well, and plates were incubated for an additional 1-2 h in a humidified 5% CO 2 incubator at 37°C. As an indicator of cell proliferation, absorbancy was measured at 490 nm, using a 96-well plate reader (Spectra Fluor, Tecan, Research Triangle Park, NC), Cells and Antibodies for Fluorescence-activated Cell Sorter-Sup-T1, a non-Hodgkin's T-cell lymphoma cell line (43), and the same cells stably infected with HIV type 1, III-B (A1953), were kindly provided by Dr. J. Hoxie. The mouse monoclonal antibody 902, specific for gp120 of HIV-1 III-B (44,45), was used to differentiate infected versus noninfected cells. Mouse sera were tested with dilutions ranging between 1:10 and 1:100. The secondary antibody used was anti-mouse IgG (␥-specific), fluorescein-conjugated isothiocyanate (Sigma). Cells were fixed for 30 min with 4% paraformaldehyde diluted in PBS. Acquisition of data was performed by using the FACSCAN flow cytometer and histogram analysis by using the CELLQuest software (Becton Dickin-son Immunocytometry Systems, Mansfield, MA).
Determination of the TCID 50 -The procedure was performed as reported (52). Briefly, 200 l/well of a viral isolate, diluted 1:3 in R-20, were added in sextuplicates in flat-bottomed 96-well plates. 50 l from these wells were admixed with 150 l of R-20 in wells of the next row, and so, in successive rows (total of 20 serial dilutions). 2 ϫ 10 4 MT2 cells (53,54) per well, in 50 l of R-20, were added and incubated at 37°C in a humidified atmosphere with 5% CO 2 . Cells were fed as necessary, all wells at the same time, and observed daily for presence of CPE. When no further migration of CPE was evident, wells were scored either positive (presence of CPE) or negative (absence of CPE), and TCID 50 values were calculated by using the method of Reed and Muench (52).
Neutralization Assays-Test sera were obtained from immunized mice and control sera from pre-immune mice (normal mouse serum) and from normal and HIV-infected human individuals (IHS, respectively). Sera were inactivated at 56°C for 1 h and sterilized by exposure to UV light. Determined dilutions of sera and viral outputs were admixed and incubated for 1 h at 37°C and then added (25 l) in triplicated or sextuplicated wells (round-bottomed 96-well plates) containing 10 4 CEMx174 cells resuspended in 175 l of R-20. Plates were incubated for 24 -40 h. Control wells without virus or serum (uninfected wells) or without serum but with a selected viral isolate (infected wells) were assayed. After incubation, cells were washed, resuspended in 200 l of R-20, and transferred to homologous flat-bottomed 96-well plates. Media were replaced at the same time in all plates, as necessary. Cultures were maintained until no further progression of CPE was observed in infected control wells, and at this time 25 l of supernatant per well were admixed with 225 l of 0.5% Triton X-100 (lysing solution) for p24 antigen detection by ELISA. Samples were stored at Ϫ80°C, until use. Percentage of neutralization was calculated as 1 Ϫ (mean absorbancy of test wells/ mean absorbancy of infected control well) ϫ 100.
ELISAs to Determine HIV-1 p24 -The assay was performed by using the HIV-1 p24 Antigen Capture Assay Kit from the AIDS Vaccine Program of the NCI-Frederick Cancer Research and Development Center (Frederick, MD). Briefly, plates pre-coated with a monoclonal anti-HIV-1 p24 antigen were washed and blocked with PBS, 0.5% FCS and 0.2% Tween 20. Supernatant lysates were added in duplicates and incubated at 37°C for 2 h. A rabbit anti-HIV-1 p24 serum and a goat anti-rabbit IgG (H & L)-HRP-labeled antibody were used in successive steps. 3,3Ј,5,5Ј-Tetramethylbenzidine-peroxidase substrate (0.1 mg/ml) (Sigma), in 0.05 M phosphate-citrate buffer and 0.03% sodium perborate buffer (Sigma), was allowed to react for 20 min. Reaction was stopped with 4 N H 2 SO 4 , and plates were read as described (41). (56) provides a template to compare peptide-ConA complexes. Prototypic peptides that have been defined to bind ConA include 908 and 712 (Table I) (30). CD analyses of these binding analogs indicate that they share a similar CD profile (30). Secondary structure comparison of these two peptide sequences indicates similarities in tertiary class type, except in the all-␤ prediction, in which an extended structure spans the WYPY sequence tract of MYWYPYASGS (Table I) (57). Although the YPY motif can be viewed as adopting a ␤-turn conformation that might emulate the spatial position of the trimannoside configuration (31), an extended conformation might also be plausible for mimetics to interact with ConA. An extended structure conformation is observed in Sesbania mosaic virus coat protein (deposited in the Protein Data Bank at Rutgers University, New Brunswick, NJ) (58) for the homologue sequence WYPY. The extended structure can overlap with the pentasaccharide within the ConA-binding site ( Fig. 2A).

Peptidyl Ligands That Bind to ConA-Crystallographic
We attempted to identify amino acid sequences that could adopt the extended secondary profile and display an adequate interaction with the ConA site. To identify likely residue types that can interact with ConA, we used the program LUDI. We have shown previously that LUDI could be used to structurally map the binding of peptide mimetics to the combining site of anti-carbohydrate antibodies (36,39). By using this approach, LUDI identified 153 interacting ligands for ConA, with some contacting the same sites as the pentasaccharide. In the search procedure, we identified moieties with Tyr-and Trp-like side changes and guanidinium groups that fit within the ConA site, but not always in the same fashion as the pentasaccharide (Fig.  2B). Substitution of Arg for Pro within the prototypic peptide 908 conserves the extended structure (peptide 909 in Table I), as does a concomitant substitution of the first Tyr in the 909 peptide with a Trp residue forming the 910 sequence (Table I).
To test the ability of the peptide analogs to bind to ConA, MAP forms of the peptides 908, 909, and 910 were synthesized. The MAP forms all bound to ConA in a concentration-dependent manner, with parallel activity (data not shown). Competition analysis with solid phase 908-MAP indicated that Me␣Man inhibits ConA peptide reactivity in a concentrationdependent manner, reaching a plateau of about 60% inhibition at a 1.6 mM concentration of Me␣Man, a 100-fold less concentration effect than reported previously (30) (Fig. 3). As expected, lactose, as control inhibitor, did not affect ConA peptide binding (data not shown). The variant MAPs 909 and 910 displayed some differences in the Me␣Man inhibition profile compared with 908-MAP. At a 1.6 mM concentration, Me␣Man inhibited about 20 and 45% of ConA binding to 910-MAP and 909-MAP, respectively.
We further defined a putative peptide sequence, RYGRY, in which the Pro residue in the WYPY motif was replaced by Gly, with the first and fourth Tyr replaced by Arg, and the addition of a Tyr at the fifth position. This putative sequence was chosen because of the identification of these residues and their ConAreactive positions by LUDI analysis. The peptide motif represented in Fig. 2C, involving the putative RYGRY tract of peptide 912 (Table I), maintains an extended secondary structure profile spanning these residues (Table I). Relative to the other class types, this peptide sequence is the same as that for peptides 908 and 712 and is perhaps more like 712, as represented in the all-␤ class ( Table I). The putative RYGRY tract makes contact with 3 residues within ConA, as does the central mannose residue. A bifurcated hydrogen bond between the guanidinium group of Arg, at the 4th position, and Ser-21 and Tyr-12 side chains of ConA, and a bifurcated hydrogen bond between the guanidinium group of Arg, at the 1st position, and Ser-223 and Ser-168, are observed (Table II). The root mean square deviation of the ConA-peptide complex, after minimization and dynamics calculations, was found to be 1.2 Å, compared with the ConA-pentasaccharide complex, indicating that the extended structure is readily accommodated within the ConA site.  Table  II. The calculated location for the peptide mimetic is, however, not an optimal binding mode in terms of mimicking the conformational properties of the pentasaccharide and in contacting the same ConA residues, as does the pentasaccharide (Table  II). Nevertheless, the interaction energy for this ConA/peptidebinding mode was found to be Ϫ75.9 kcal/mol, falling within the range of interaction energies calculated for the trimannoside constituent (Table II).
Peptide Mimetic Competes for Carbohydrate Binding to ConA-Clustering or repeating the 912 sequence, forming the peptide 911, manifests an extended secondary structure (Table  I). ELISAs carried out using various concentrations of ConA showed a concentration-dependent binding to the clustered form of the RYGRY containing peptide 911, which was inhibited by Me␣Man (Fig. 4A). We observed binding of ConA to 911-MAP, at ConA concentrations lower than those required for binding to ligands tested, which included extended structure peptides, and better than that for binding to 908-MAP. This result suggests a high avidity interaction of ConA with the multivalent 911 peptide and therefore requires a higher concentration of Me␣Man for inhibition of 911-MAP binding to ConA than that for the 908-MAP (Fig. 4B).
The putative monovalent peptide 912 and 911-MAP binding to ConA was studied further by ITC, to determine the binding parameters such as K a and ⌬G. Experimental conditions were previously standardized for studies of multivalent interaction by ITC (9). The association constant (K a ) (1.1 ϫ 10 4 M Ϫ1 ) and ⌬G (5.6 kcal mol Ϫ1 ) values of ConA for the monovalent 912 peptide were observed to be comparable to that observed for Me␣Man (Fig. 5). Earlier, microcalorimetric studies showed that the K a and ⌬G values of ConA for carbohydrate ligands, such as Me␣Man, trimannoside, and pentasaccharide, were 0.82 ϫ 10 4 , 49 ϫ 10 4 , and 92 ϫ 10 4 M Ϫ1 and 5.3, 7.8, and 8.1 kcal mol Ϫ 1, respectively (20). In contrast, K a of ConA for 911-MAP was found to be 26 ϫ 10 4 M Ϫ1 , with a value for ⌬G of 7.4 kcal mol Ϫ1 . These results indicate that the multivalent 911 peptide displays a comparable association constant and free energy of binding, as do oligosaccharide ligands.
To verify further the valence of the peptides for ConA, a precipitation study was carried out (Fig. 6). The number of binding site per monomer (n), as determined by ITC, suggests that peptide 912 is monovalent, whereas 911-MAP is multivalent for ConA. Multivalent lectin-ligand interactions often lead to the formation of insoluble cross-linked complexes, which can easily be monitored spectrophotometrically by measuring the absorbancy at 420 nm. The efficient precipitation by the 911-MAP form confirms that it possesses multiple binding sites for ConA. About 70% of the cross-linked complexes are re-dissolved when treated with Me␣Man (400 mM). This observation clearly

Binding of Serum from Immunized Mice to HIV-1 III-Binfected Cells-
The 911 peptide is predicted to have a major histocompatibility complex class II motif spanning the RYRYGRYRS sequence. Immunization with peptide 911 indicated a robust cellular response specific for peptide 911 (Fig. 7). To determine if serum antibodies react with membrane-ex- pressed gp120/gp41, we examined serum IgG binding to constitutively infected cells compared with the binding to the same non-infected cells. Results in Fig. 8 show IgG antibody binding to chronically infected cells (A1953 cells). We observe that the monoclonal antibody 902 differentiates infected from non-infected cells (Fig. 8A). Immunization with control MCF7 cells induce serum reactive with the neolactoseries antigen LeY (13) and also antibodies that are potentially reactive with major histocompatibility complex class I, which shares some homology with gp120, as anti-class I antibodies bind to Env protein (Fig. 8B) (59,60). Serum from 911-immunized mice reacted stronger with infected cells than non-infected cells (Fig. 8C). IgG from mice immunized with other formulations (LeY or QS21) did not show any significant increased binding to A1953 cells compared with their binding to Sup-T1 cells (data not shown).
Immune Serum Affects in Vitro Neutralization-The neutralizing activity of the serum was assessed by p24 ELISA. In Fig.  9A, we observe that serum from MCF7-and 911-MAP-immunized mice were able to neutralize a viral input of 100 TCID 50 of HIV-1 III-B, up to 1:128 dilutions, with a calculated percentage of neutralization of about 80% for each, reflecting a reduction of p24 antigen in supernatants. In contrast, serum to 910 peptide did not neutralize this viral load at 1:64 dilutions. Positive control (IHS) displayed a neutralizing capability beyond 1:512 dilution.
By increasing the viral load to 200 TCID 50 (Fig. 9B), IHS still displayed neutralizing activity, whereas MCF7 serum retained some neutralizing capacity, better than that for serum raised to 911-MAP. However, in a representative neutralization assay of 50 TCID 50 of HIV-1 III-B, the 911-MAP serum displayed neutralization capability with dilutions up to 1:256 (data not shown). As with III-B, 911-MAP serum was able to neutralize 100 TCID 50 of MN, at 1:128 dilutions, with partial neutralization activity at 1:256 dilution (data not shown). Negative controls, normal human serum and normal mouse serum, did not show neutralizing capabilities to any viral isolate, even at low dilutions. tural concepts to develop novel carbohydrate forms (61). The clustering and multivalent presentation of carbohydrate antigens appears relevant to induce antibody responses to natively expressed carbohydrate antigens on cell surfaces (13,62,63). We have shown that peptide mimeotopes can elicit carbohy-drate cross-reactive immune responses to natively expressed bacterial and tumoral antigens related to those expressed on HIV-1 Env glycoprotein (9,13,41). To explore further the utility of targeting the Env glycoprotein and generalizing peptide design strategies (39, 64), we are optimizing peptides re- active with ConA seen as a template. The mimetics described here represent rationally designed candidate immunogens despite significant gaps in our knowledge regarding the molecular and functional characteristics of PS mimeotopes (12).

DISCUSSION
We identified a peptide that, rendered as a clustered and multivalent form, was reactive with ConA at lower concentrations than those required for reaction of some native oligosaccharide ligands of ConA. The 911-MAP displayed competitive inhibition with carbohydrate ligands of ConA, indicating that it binds at an overlapping carbohydrate-binding site on ConA. ITC and precipitation experiments suggest that the putative peptide 912 is monovalent for ConA, and its affinity is comparatively weak (similar to that of Me␣Man). The MAP format of peptide 911 resulted in a higher association constant and free energy of association with ConA compared with that found upon binding of the putative 912 peptide. Detailed thermodynamic analysis of binding of trimannoside to ConA has been performed by ITC (18,20), and the K a and ⌬G values of 911-MAP are comparable to those of ConA-reactive trimannoside and pentasaccharide.
Whereas the enhancement in K a of 911-MAP (relative to 912) is due to multivalent presentation, the increased affinity of the two carbohydrate ligands (compared with the monosaccharide) is an outcome of an extended site interaction. Initial analysis of binding raw data obtained from ConA-911-MAP titration, and its comparison with the data of multivalent sugars, points to certain fundamental differences in the overall binding mechanism. Differences between two multivalent binding systems are primarily attributed to the structural dissimilarities of the peptide and carbohydrate ligands, a conclusion indicated from the molecular modeling study. It has previously been observed that even a minor structural alteration in lectin structure profoundly affects the binding thermodynamics. ConA and the homologous lectin from Dioclea grandiflora possess conserved binding sites for the pentasaccharide, yet minor differences in the lectin structure result in a totally different binding energetic for the oligosaccharide. D. grandiflora binds the pentasaccharide with much lower K a values than does ConA (20). Therefore, it is possible that structural differences between the peptide and carbohydrate ligands contribute to the lower K a value of the 912 peptide (compared with the pentasaccharide). However, the affinity is significantly boosted in the MAP format, where the peptide is functionally multivalent.
Many naturally occurring carbohydrates, including glycoproteins, are multivalent, which results in increased avidity for lectins and antibodies. This characteristic must be emulated to affect functional immune responses (13). Although the benefits of multivalency are well established for both antibody and lectin binding to carbohydrates, the molecular mechanisms underlying these phenomena are poorly understood. Presumably, the effect is not attributable to the recognition of a combined epitope encompassing three or more sugar chains, as such a structure would exceed the size of an antibody-combining site (39). It is probable that multivalency contributes both to structural properties and entropy involved in binding (19). It is likely that the density of antigen expressed on cell or pathogen surfaces can play a significant role in mediating avidity interactions.
We have shown that the 911 peptide mimetic, formulated as a MAP form, can induce functional carbohydrate cross-reactive antibodies, in concordance with our other studies (13,41). 911-MAP induces serum that neutralizes HIV-1 III-B at levels comparable to serum induced by MCF7 cells, using a viral input of 100 TCID 50 . For lower viral inputs (50TCID 50 ), anti-911 serum neutralizes the same isolate up to 90% at 1:256 dilution (data not shown). The presentation of the MAP form of the mimeotopes, while emulating the multivalent carbohydrate surface, may still not effectively cope with micro-heterogeneity in carbohydrate structures. However, this same problem exists when immunizing with carbohydrate immunogens, since, many times, synthetic carbohydrate forms do not induce responses cross-reactive with native carbohydrate forms, requiring modifications in synthetic strategies. Likewise, cyclization of peptide mimetic immunogens may further restrict carbohydrate cross-reactive responses much as it does in inducing responses to protein antigens, by limiting the polyclonal response.
In summary, these results indicate that designed peptide mimetics of carbohydrate antigens can induce functional responses that may find utility in priming strategies to further augment carbohydrate immune responses against pathogens or tumor cells (11). Although modeling does not account for multivalent interactions, as represented by MAP forms, modeling can define potential binding site constituents. Consequently, strategies that evaluate potential mimetic binding modes and thermodynamics of binding should further facilitate structure-assisted design of surrogates for vaccine applications. Likewise, this study encourages further investigation to ascertain the mechanism(s) by which certain peptide mimics of PS antigens play their role as mimeotopes, in that they can stimulate immunity that targets PS (12).