NMR Studies of Carbohydrates and Carbohydrate-mimetic Peptides Recognized by an Anti-Group B Streptococcus Antibody* □ S

As part of a program to investigate the origins of pep-tide-carbohydrate mimicry, the conformational preferences of peptides that mimic the group B streptococcal type III capsular polysaccharide have been investigated by NMR spectroscopy. Detailed studies of a dodecapeptide, FDTGAFDPDWPA, a molecular mimic of the polysaccharide antigen, and two new analogs, indicated a propensity for (cid:1) -turn formation. Different (cid:1) -turn types were found to be present in the trans and cis (Trp-10– Pro-11) isomers of the peptide: the trans isomer favored a type I (cid:1) -turn from residues Asp-7–Trp-10, whereas the cis isomer exhibited a type VI (cid:1) -turn from residues Asp-9–Ala-12. The interaction of one-dimensional STD-NMR (19–23) was recorded at K and selective saturation of protein resonances at (cid:3) ppm (30 ppm for using a series of 40 Gaussian shaped pulses (50 ms, 1-ms delay between pulses, (cid:5) B /2 (cid:7) (cid:1) 110 for a total saturation time of 2.04 s. The one-dimensional STD-NMR of were recorded at 800 MHz and 298 K with 8192 scans, selective satu- ration of protein resonances at 10 ppm, and a total saturation time of 2.04 s. Subtraction of saturated spectra from reference spectra was performed by phase Transferred nuclear Overhauser effect (trNOESY) (24–28) spectra of mAb S9/peptide were recorded at 800 MHz and 298 K using the States (hypercomplex) method, with 4096 points and 1024 t 1 increments, and a relaxation delay of 1 s. The observed trNOE correlations were large and negative; a sample of the peptide in the presence of a different antibody showed only minimal NOE intensities at 800 MHz. Structure calculations were based on a spectrum recorded with 64 scans/incre- ment and a mixing time of 200 ms. Data processing was performed by zero filling to 2048 points in F 1 to give a final data matrix of 4096 (cid:4) 2048 followed by multiplication with a squared cosine function and Fourier transformation. Automatic base-line correction was performed prior to integration of cross-peak volumes. Cross-peak were

major cause of neonatal sepsis and meningitis (1,2). These bacteria are encapsulated with type-specific polysaccharides (3). The immunogenicity of these polysaccharides is unexpected because they all contain terminal sialic acid residues, which are not usually immunogenic because of their presence in human tissues (4). For example, the NeuAc-␣-(233)-Gal glycosidic linkage is found in human blood group substances (4).
Strikingly, all bacteria with polysaccharide capsules containing terminal sialic acid are human pathogens (4,5). The presence of sialic acid is important in pathogenesis in that it confers resistance to the immune system by inhibition of activation of the alternative pathway of complement (4). Several studies have indicated that sialic acid may also have important effects on the conformational properties of polysaccharides. A study by Jennings and co-workers (6) showed an unusual length dependence of the binding affinity of oligosaccharide portions of the GBS type III capsular polysaccharide (GBSPIII; see Scheme 1) to antibodies. At least two repeating units were required for even suboptimal binding, and binding affinity increased with chain length. Furthermore, the terminal sialic acid residues were shown not to be immunodominant, yet were essential for the formation of GBSPIII-specific antibodies. In addition, NMR studies indicated that GBSPIII was capable of forming extended helices in solution (7). Therefore, the presence of a conformational epitope, that is, a certain conformation or secondary structure present within an ensemble of conformations adopted by the polysaccharide and being the only one recognized by antibodies, was hypothesized (6,7). Antibodies directed against such an epitope, only formed by extended segments of polysaccharide, would avoid binding to the short oligosaccharides found in human tissue antigens and allow defense against invading bacteria without harmful effects to the human body. Evidence for such length-dependent conformational epitopes has also been reported for the group B Neisseria meningitidis (8) and Streptococcus pneumoniae type 14 (9) polysaccharides, and the x-ray crystal structure of an antibody directed against the N. meningitidis polysaccharide showed an unusually long, groove-shaped binding site, also suggesting that the antibody recognized a long, helical structure (10). The evidence showing that sialic acid is essential for formation of the epitope, yet is not itself involved in the epitope, raises the question of exactly how the polysaccharide-antibody recognition occurs.
Vaccines based on GBSPIII or GBSPIII-protein conjugates are expected to be successful (11,12). More extensive knowledge of the modes of interaction of GBSPIII oligosaccharides with antibodies would be helpful in the design of targeted vaccines. Furthermore, the use of carbohydrate-mimetic peptides (13) may be valuable in this system. Carbohydrate-mimetic peptides are a promising strategy to improve the immunological characteristics of polysaccharides. Thus, although some polysaccharides have been developed into effective vaccines, some undesirable characteristics remain; T-independent responses lacking memory and lack of response in infants are examples. In many cases, conjugation of the polysaccharide to a protein carrier overcomes these problems. However, in certain cases, when it may be desirable to select and amplify only one component of the immune response, the combination of peptides with polysaccharides may be very useful.
The feasibility of an alternative approach to GBS vaccination, based on carbohydrate-mimetic peptides, has been demonstrated (14). Thus, a protective monoclonal antibody against GBS, directed against GBSPIII, was developed (mAb S9; Refs. 14 and 15). Peptides acting as specific ligands for mAb S9 were isolated by screening a phage-displayed peptide library with this mAb. A high affinity (IC 50 ϭ 23 M) peptide ligand, FDTGAFDPDWPAC, competitively inhibited binding of GBS not only to mAb S9 but also to polyclonal anti-GBS antibodies. Immunization of mice with peptide-protein conjugates produced an anti-GBS and anti-GBSPIII immune response (14).
In an effort to understand the molecular basis of GBSPIII mimicry and to design better immunogens, we have undertaken detailed conformational studies of the immunogenic peptide FDTGAFDPDWPA and two of its analogs in solution, using NMR spectroscopy. In addition, we have investigated the interactions of the parent peptide and of oligosaccharides corresponding to the type-specific polysaccharide with mAb S9. We report herein the conformational preferences of the three peptides in solution and of one in complex with the antibody. We show that the biologically active peptide mimics exhibit the propensity to form type I ␤-turn structures that are highly dynamic in the free peptide systems. Further, we identify important determinants of oligosaccharide recognition by the antibody. This study serves as a prelude to the detailed comparison of the topographies of a carbohydrate and of its peptide mimetic in recognition by antibodies.

EXPERIMENTAL PROCEDURES
Peptides-The dodecapeptide FDTGAFDPDWPA (peptide 1) was purchased from Cell Essentials (Boston, MA). The peptides FDTGAFD-PDWPYD (peptide 1A) and FDTGAFDPDWPY (peptide 1B) were purchased from LSU Health Sciences Center Core Laboratories (New Orleans, LA). All three peptides were N-acylated and C-amidated at their respective termini.
Competitive Inhibition Enzyme-linked Immunosorbent Assay (ELISA)-The ability of synthetic peptides to inhibit the binding of mAb S9 to intact GBS was measured by ELISA, as described previously (14). MAb S9 is a murine IgM directed against GBSPIII (14) and is highly protective against infection by type III GBS in an animal model. Purified S9 was obtained by distilled water dialysis of ascites fluid. Antibody was titrated for binding to GBS, and two dilutions on the midpoint of the binding curve were selected, 0.5 and 0.2 g/ml. Peptide concentrations were determined carefully using the BCA protein determination kit (Pierce). Serial dilutions of peptides were mixed with mAb S9 diluted in PBS and 1% bovine serum albumin and incubated for 1 h at room temperature. Microtiter plates (Immulon 2HB, Dynatech, Chantilly, VA) were coated with intact GBS by first coating the wells with poly-L-lysine, then adding the GBS, and cross-linking the GBS with glutaraldehyde, as described previously (14,15). Plates were blocked with PBS and bovine serum albumin at 4°C for 18 h. The antibody-peptide mixture (100 l) was plated into the microtiter wells and incubated overnight at 4°C. The plates were washed with PBS and 0.1% Tween 20 six times; alkaline phosphatase-conjugated goat antimouse IgG (heavy plus light chain-specific, Zymed Laboratories Inc., South San Francisco, CA) was added. After a 6-h incubation at room temperature, the secondary antibody was washed off, and the chromogenic substrate p-nitrophenyl phosphate (Sigma) was added (0.5 mg/ml in diethanolamine buffer, pH 9.8). The absorbance at 405 nm was monitored at various time points thereafter. The data presented are the means of multiple determinations, with error bars indicating the S.E. When no error bars are present, the S.E. was smaller than the symbol used in the figure.
NMR Spectroscopy of Free Peptides-1 H NMR spectra were recorded at Montana State University on Bruker DRX 500-and 600-MHz spectrometers using sample tubes with a 5-mm outer diameter. The spectra of the three peptides at 2 mM sample concentration were recorded at 280 K in 20 mM sodium phosphate buffer at pH 4.2 and pH 4.8, prepared in 90% H 2 O and 10% D 2 O. Spectra were also recorded for samples prepared at 2 mM concentration in water:trifluoroethanol, 7:3. Identical NOE patterns were observed in these spectra, indicating that trifluoroethanol did not alter the conformational preferences of the peptides. Two-dimensional NMR spectra were recorded in the phase-sensitive mode using the States-TPPI method (16) or TPPI method (17) for quadrature detection in the t 1 dimension. All experiments were carried out using the WATERGATE pulse sequence for water suppression (18). Two-dimensional 1 H, 1 H NOESY spectra were recorded with mixing times of 220 and 280 ms. Two-dimensional 1 H, 1 H TOCSY spectra were recorded using an MLEV-17 spin-lock sequence with a 10-kHz rf field strength and a mixing time of 60 ms. For two-dimensional 1 H, 1 H ROESY, mixing pulses of 220-and 280-ms duration were applied at a 3-kHz rf field strength. Typically, spectra were acquired with 448 -512 t 1 increments (except 1024 t 1 increments for DQF-COSY), 2048 data points, and a relaxation delay of 1.5 s, spanning a spectral width of 11.0 ppm. Spectra were acquired with 96 scans/t 1 increment for NOESY, 96 scans/t 1 increment for ROESY, 16 scans/t 1 increment for TOCSY, and SCHEME 1. Repeating unit of the group B streptococcal type III capsular polysaccharide.

Group B Streptococcus Antigen-Antibody Recognition by NMR
40 scans/t 1 increment for DQF-COSY. Spectra were processed using XWIN-NMR, Version 2.6 (Bruker). All spectra were zero-filled once in both spectral dimensions, and a sine-bell window function, phaseshifted by 45-90°, was applied in both dimensions prior to Fourier transformation.
NMR Spectroscopy of Ligand-Antibody Interactions-NMR spectra were recorded on a Bruker AMX 600-(Simon Fraser University) and on a Varian 800-MHz NMR spectrometer (NANUC, Edmonton, Alberta). Chemical shifts were referred to DSS (internal standard). MAb S9 was prepared by Ligocyte, Inc. (Bozeman, MT) by growth in serum-free hybridoma growth medium and precipitation with an equal volume of saturated ammonium sulfate. The precipitate was washed with 50% saturated ammonium sulfate, resuspended in PBS, and dialyzed against several changes of PBS to remove the remaining ammonium sulfate.
For preparation of NMR samples, antibody solutions were concentrated using a stirred cell ultrafiltration unit (Amicon, model 8010) with a regenerated cellulose membrane (nominal molecular weight cutoff 10,000). A molecular mass of 975,000 Da for the IgM antibody was assumed.
For the preparation of an NMR sample of the decasaccharides in the presence of the antibody, mAb S9 was exchanged into D 2 O buffer (10 mM (KH 2 PO 4 /K 2 HPO 4 ), 150 mM NaCl, 0.02% NaN 3 , prepared in 99.9% D 2 O) as follows. MAb S9 (6.2 mg, 4.6 mg/ml in PBS, pH 6.5) was concentrated from 1.35 ml to 0.5 ml, then diluted to 15 ml with D 2 O buffer, and concentrated to 0.6 ml by ultrafiltration as described above. Two more exchanges were performed (15 ml 3 0.7 ml, 5 ml 3 0.5 ml), for a total of three exchanges; the sample was then filtered through a 0.2 M membrane. To the concentrated, filtered antibody solution was added 0.9 mg of decasaccharide 1, and precipitation was observed (precipitin reaction). The precipitate was collected by centrifugation, and the supernatant was filtered into the NMR tube. The antibody concentration remaining in the supernatant was estimated by A 280 (⑀ ϭ 1.2) to be 0.95 M; at a nominal 1 concentration of 0.75 mM, this would mean a ligand:binding site ratio of 79:1; however, the actual ratio may be somewhat less. The final sample contained 40 M DSS. 1 mg of the N-propionylated decasaccharide 2 was later added to the sample for a nominal concentration of 0.82 mM; very little precipitation was observed at this time.

TABLE II Temperature coefficients (⌬␦/⌬T, ppb/K) for backbone amide protons and three-bond, amide proton to H ␣ coupling constants ( 3 J HN,H␣ ), for the trans and cis isomers of peptide 1 and for the trans isomer of peptide 1A (major species, 1Ai)
Ϫ4 (Ϫ3) 7.6 (7.8) a Values in parentheses are for the cis isomer. b ND, not determined (due to overlap of resonances).
The one-dimensional STD-NMR (19 -23) spectrum of mAb S9/peptide was recorded at 600 MHz and 310 K with 2048 scans and selective saturation of protein resonances at Ϫ1 ppm (30 ppm for reference spectra) using a series of 40 Gaussian shaped pulses (50 ms, 1-ms delay between pulses, ␥B 1 /2 ϭ 110 Hz), for a total saturation time of 2.04 s. The one-dimensional STD-NMR spectra of mAb S9/decasaccharides were recorded at 800 MHz and 298 K with 8192 scans, selective saturation of protein resonances at 10 ppm, and a total saturation time of 2.04 s. Subtraction of saturated spectra from reference spectra was performed by phase cycling (19 -21).
Transferred nuclear Overhauser effect (trNOESY) (24 -28) spectra of mAb S9/peptide were recorded at 800 MHz and 298 K using the States (hypercomplex) method, with 4096 points and 1024 t 1 increments, and a relaxation delay of 1 s. The observed trNOE correlations were large and negative; a sample of the peptide in the presence of a different antibody showed only minimal NOE intensities at 800 MHz. Structure calculations were based on a spectrum recorded with 64 scans/increment and a mixing time of 200 ms. Data processing was performed by zero filling to 2048 points in F 1 to give a final data matrix of 4096 ϫ 2048 followed by multiplication with a squared cosine function and Fourier transformation. Automatic base-line correction was performed prior to integration of cross-peak volumes.
Cross-peak intensities were converted to distances by first correcting for peak multiplicity by dividing the intensity by the factor (n i *n j ), where n i and n j are the numbers of protons contributing to the crosspeak in F 2 and F 1 (29). Corrected intensities were then converted to distances using the relationship I ij /I ref ϭ ((r ref /r ij ) 6 ) (30) using the distance between the Pro-11 H␦ protons (1.8 Å) as a reference. The calculated distances were incorporated into a simulated annealing protocol within the program XPLOR (31). Only unambiguously assigned crosspeaks were included, leading to a set of 47 distance restraints (33 inter-residue; Table S1). Distance restraints consisted of a minimum distance of 1.8 Å and a maximum distance of (r ij ϩ 0.75 Å). Distances involving unresolved or unassigned diastereotopic protons were treated using center averaging (31).
A structure of the peptide in an extended conformation was used as a starting point for molecular dynamics calculations. Simulations consisted of 30 ps of molecular dynamics at 1000 K, using the parallhdg force field, with increasing force constants for geometry and a force constant for NOE-derived distances of 50 kcal/mol/Å 2 , followed by cooling to 100 K for 20 ps. 40 structures produced by this method were refined by another cooling cycle from 1000 to 100 K over 45 ps, followed by 200 cycles of energy minimization. During the calculations, nonglycine residues were restrained to negative values. The criteria for acceptance of structures were correct geometry and no distance restraint violations of Ͼ 0.5 Å; 34 of 40 structures satisfied these criteria. Analysis of the structures was performed within Insight II 2000 (Accelrys, Inc.).

RESULTS
NMR Resonance Assignments-1 H chemical shifts and resonance assignments were established using two-dimensional 1 H, 1 H TOCSY, NOESY, and ROESY experiments (32) and are reported in Table I. Sequential assignments of 1 H resonances were based on characteristic sequential NOE connectivities observed between the ␣-proton of residue i and the amide proton of residue iϩ1, i.e. d ␣N (i,iϩ1), in both NOESY and ROESY data sets. Three-bond 3 J HN,H␣ coupling constants were extracted from either one-dimensional 1 H spectra or two-dimensional DQF-COSY spectra. Coupling constants and temperature coefficients describing the changes in amide proton chemical shifts as a function of temperature are reported in Table II. A summary of sequential (d(i,iϩ1)) and short range (d(i,iϩ2)) and (d(i,iϩ3)) 1 H-1 H NOE connectivities is presented in Fig. 1.
Conformation of FDTGAFDPDWPA (Peptide 1)-The NMR spectra of the dodecapeptide FDTGAFDPDWPA revealed that the peptide exists in two isoforms due to cis-trans isomerization about the Trp-10 -Pro-11 amide bond (Figs. 2 and 3). The trans isoform was identified by the presence of characteristic Trp-10 H ␣ to Pro-11 H ␦ NOE cross-peaks. In addition, an NOE between Trp-10 H ␣ and Pro-11 H ␣ , characteristic of cis-proline, was apparent and confirmed the presence of a cis/trans isomer mixture (Fig. 3B). NMR signals originating from both isomers are identified in the NOESY and TOCSY spectra shown in Figs. 2 and 3A, respectively. The ratio of trans to cis isomer was estimated to be 7:3, based on the differential signal intensity of the resolved indole NH resonance of each isomer in the one-dimensional 1 H spectrum. No cis-trans isomerization was detected across the Asp-7-Pro-8 amide bond, but this does not exclude the possibility that a small amount of the peptide, too small to be detected by NMR, exhibits the cis configuration.
The NOE connectivities for peptide 1 in aqueous solution at 280 K are summarized in Fig. 1A. Interestingly, a continuous stretch of d NN (i,iϩ1) and d ␣N (i,iϩ2) NOEs was observed from residues Phe-1 to Asp-7, indicative of turn-like structures often referred to as "nascent helices" (33) and typically dynamic and transient in nature. Consistent with this observation, small but significant upfield shifts of the H ␣ resonances (ϳ 0.1-0.2-ppm deviations from random coil values) were observed for Phe-1, Asp-2, Thr-3, Ala-5, Phe-6, Pro-8, and Asp-9 in the trans isomer. With the exception of the H ␣ of Asp-9, a similar chemical shift trend was observed for the cis isomer (Table I).
For the trans isomer, a type I ␤-turn was detected from residues Asp-7 through Trp-10. Supporting the presence of a type I ␤-turn, weak d ␣N NOEs were observed, between Asp-7 and Trp-10, and between Pro-8 and Trp-10 (Fig. 3). Observa- tion of a weak d ␦N NOE between Pro-8 and Asp-9, as well as a strong d NN NOE between Asp-9 and Trp-10, also confirmed the presence of a type I ␤-turn. These data were corroborated by the small temperature coefficient (i.e. change in chemical shift as a function of temperature, ⌬␦/⌬T) of Ϫ2 ppb/K measured for the backbone amide proton of Trp-10, which indicated that the Trp-10 amide hydrogen participates in a hydrogen bond (Table II).
For the cis isomer, analogous NOE patterns were observed for residues Asp-7 to Trp-10, except that the d ␣N NOE between residues Asp-7 and Trp-10 present in the trans isomer was not present in the cis-isomer species (Fig. 1B). This suggests that the type I ␤-turn observed in a large portion of the trans conformers is not represented in the population of structures of the cis species in solution. Rather, the cis isomer may populate interconverting conformations, possibly involving type VI ␤-turns spanning residues Asp-9 -Ala-12, as suggested by the observation of a d ␣N NOE between Trp-10 and Ala-12 (Fig. 1B). Type VI ␤-turns have often been observed in peptides containing cis-proline residues, especially when the cis-Pro is preceded by an aromatic residue, and are typically characterized by upfield shifted H ␣ and H ␤ proton chemical shifts for the proline residue (34). In this case, the H ␣ and H ␤ proton chemical shifts of Pro-11 are observed at 3.38 ppm, and 0.90, 1.60 ppm respectively, significantly upfield-shifted from canonical random coil chemical shift values for cis-proline residues (4.60 (H ␣ ), and 2.39 and 2.18 (H ␤ ) (35)). The presence of a d ␣N NOE between residues Trp-10 and Ala-12 is in agreement with the chemical shift data, which suggest the presence of significant populations of type VI ␤-turns in the cis isomer of FDTGAFDPDWPA. For type VI ␤-turns, 3 J HN,H␣ coupling constant values of 3.5 Hz are expected for residue iϩ1, corresponding to a torsional angle of Ϫ60°. In this case, the 3 J HN,H␣ coupling constant for Trp-10 was measured to be 6 Hz, reflecting conformational averaging, and cannot be used to identify specific structural elements or different types of ␤-turns which may be interconverting rapidly in the cis isomer in solution.
Thus, the NMR data reveal that the major species of peptide 1 is the trans isomer, which favors a type I ␤-turn conformation in solution. However, this structural motif is believed to be flexible and dynamic in the free peptide. The less populated cis isomer appears to favor a type VI ␤-turn motif, although rapid interconversion between different conformations and turn motifs is also likely to take place.
As peptide 1 exhibited a 7:3 mixture of trans and cis isomers in solution, the design of two new analogs was performed, with the goals of altering the peptide trans and cis content and investigating the resulting effect on the biological activity. Studies have shown that the cis content is increased significantly in peptides having the following X-Ar-Pro-Ar-Hy arrangement, where Ar and Hy refer to aromatic and hydrophilic residues, respectively, and X is any amino acid (36). For this purpose, two new peptides were synthesized. The Ala-12 of peptide 1, FDTGAFDPDWPA, was replaced by a tyrosine residue in both analogs, peptide 1A (FDTGAFDPDWPYD), and peptide 1B (FDTGAFDPDWPY). An extra Asp residue was incorporated at the C terminus of the 1A sequence to test the possibility that the negative charge imparted by the Asp residue might mimic the negative charge of the sialic acid component of the original GBSPIII pentasaccharide antigen. These peptide analogs were tested for in vitro biological activity and investigated by NMR to determine whether differences in con-  formational preferences could account for the differential biological activity observed in ELISAs.
ELISA-The ability of synthetic peptides to inhibit the bind-ing of mAb S9 to intact bacteria is shown in Fig. 4. Because the antibody is IgM, with 10 binding sites, and the capsular polysaccharide on GBS is a repeating oligomer, it is not possible  to determine directly the affinity of the antibody for the antigens by ELISA. Rather, this experiment demonstrates the relative avidity of the antibody for the different peptides compared with the intact polysaccharide antigen. The peptide FDTGAFDPDWPYD (1A) was shown to be able to compete with the capsular polysaccharide approximately twice as effectively as either the originally selected peptide FDTGAFDPD-WPA (1) or the peptide FDTGAFDPDWPY (1B). Thus, the addition of an Asp residue at the C terminus (1A) appeared to increase the relative avidity of the antibody for this peptide by a factor of 2.
Conformation of FDTGAFDPDW*PYD (Peptide 1A)-In water, the 1A peptide comprised two chemical species in a ratio of 3:2, designated herein 1Ai and 1Aii. Inspection of both ROESY and NOESY spectra revealed that both species were present only as trans isomers across the Asp-7-Pro-8 and Trp-10 -Pro-11 peptide bonds. Examination of the TOCSY and DQF-COSY spectra (Fig. 5A) revealed that in both species, the ␤-proton chemical shifts of the tryptophan residue (Trp*-10) were significantly shifted upfield and resonated at 2.12 and 1.95 ppm for the 1Ai peptide and 2.23 and 1.95 ppm for the 1Aii peptide (Table III). These spectra also showed that the ␤-protons were coupled to a proton resonance at ϳ3.56 ppm, assigned to the 3H proton of a modified tryptophan indole ring. This resonance was in turn connected to two geminal protons at 3.98 and 3.62 ppm, assigned to the 2H protons (Table III). Observation of upfield shifted ␤-proton chemical shifts and the scalar connectivity patterns described above led to the unambiguous conclusion that the indole ring of the tryptophan 10 residue was reduced in both species of the 1A peptide, and it is therefore denoted as Trp*-10 (Scheme 2). Reduction of Trp-10 must have occurred during peptide synthesis, as this process often requires harsh conditions for deprotection of functional groups and may lead to chemical modification of indole side chains. The integrity of the 1A peptide was investigated further by mass spectrometry. Mass spectra of 1A revealed that this peptide contained two mass species, one corresponding to Mϩ for 1Ai with a reduced Trp*-10 indole ring, and a second peak at Mϩ40 for a second species attributed to 1Aii (data not shown). The major 1Ai SCHEME 2. Reduction of a tryptophan side chain, as may occur during peptide synthesis. 6. Sections of the two-dimensional 1 H, 1 H TOCSY spectrum (A) and the two-dimensional DQF-COSY spectrum (B) of peptide 1A (FDTGAFDPDWPYD) depicting connectivities between side chain protons of the reduced 2,3-dihydrotryptophan residue at position 10, also referred to as Trp*-10. species was identified by NMR to be FDTGAFDPDW*PYD, i.e. the 2,3-dihydrotryptophan analog of Trp-10, as the complete spin system of the reduced indole ring of the Trp-10 residue could be mapped out in the NMR spectra (Fig. 6, A  and B). The 1Aii species could not be fully characterized by NMR, although the spectra suggested that the 6-position of Trp-10 may have been derivatized/alkylated in this second species, because cross-peaks between the 4H and 5H protons were identifiable, but no cross-peaks were observed between the 5H and 6H protons of the indole ring.

FIG.
Surprisingly, peptide 1A with a reduced tryptophan indole ring was found to be more active in the competition ELISA than the peptide 1 originally selected by phage display (Fig. 4). The conformation of peptide 1A was therefore investigated further. The NOE patterns for the major 1Ai species are summarized in Fig. 1C. NOE connectivities, ␣-proton chemical shifts, and backbone amide temperature coefficient data (Table II) all support the conclusion that the majority of the 1A conformers exhibits the trans configuration across the Asp-Pro and Trp-Pro peptide bonds. The NOE patterns are also very similar to those identified for the trans isomer of peptide 1, suggesting that peptide 1A, like peptide 1, has the propensity to form type I ␤-turns in solution.
Conformation of FDTGAFDPDW*PY (Peptide 1B)-As for peptide 1A, peptide 1B existed as two chemical species in solution, with both adopting only the trans configuration across the X-Pro peptide bonds. The mass spectrum of peptide 1B showed the presence of two major species at Mϩ and Mϩ40 mass units, leading us to conclude that as for peptide 1A, the indole ring of Trp-10 had been modified during synthesis of peptide 1B. Because of considerable spectral overlap of amide proton and ␣-proton resonances (Fig. 6), NOE connectivities could not be identified unambiguously and thus prevented detailed structural characterization of peptide 1B in solution.
Interactions of Ligands with mAb S9 -Having investigated the conformational preferences of peptide 1 and two new analogs, 1A and 1B, when free in solution, it was also of interest to investigate the interaction with the anti-GBSPIII monoclonal antibody, S9, that was used to select peptide 1. Therefore, we performed STD (19 -23) and transferred NOE (24 -28) NMR experiments.
The STD-NMR technique (19 -23) is a method of epitope mapping by NMR spectroscopy. During the experiment, resonances of the protein are selectively saturated, and intramolecular NOEs develop within the protein and in addition, give rise to intermolecular NOE effects in a binding ligand. These negative NOE effects may be observed as enhancements in the difference (STD-NMR) spectrum resulting from subtraction of this spectrum from a reference spectrum in which the protein is not saturated (19 -23). Enhancements are observed only for the resonances of protons in close contact with the protein, and this allows direct observation of areas of the ligand that comprise the epitope (19 -23).
We investigated the interaction of the mimetic peptide, FDT-GAFDPDWPA (peptide 1), with mAb S9. In this case, the observed STD-NMR effects were weak; this is likely due to unfavorable binding kinetics. However, enhancements of the Trp-10 H-7 resonance, and other aromatic resonances belonging to either Trp-10 or the Phe-1 or Phe-6 residues, were observed (Fig. 7). In addition, enhancements of the side chain methyl resonances of Thr-3, Ala-5, and Ala-12 were apparent, along with weak enhancements of Pro-8, Pro-11, and Asp-7, indicating that these side chains are involved in the epitope. No enhancements of resonances of the minor cis isomer were observed; however, because this may be the result of its lower concentration, this possibility is not excluded.
The combination of STD-NMR epitope mapping data with knowledge of the bound conformations of ligands, which may be obtained by trNOESY experiments, is a powerful method to build up models of antibody-ligand interaction (37)(38)(39)(40)(41)(42)(43). Therefore, trNOESY experiments (24 -28) were used to investigate the bound conformation of the peptide. A trNOESY spectrum is shown in Fig. 8. The structure was not well defined except in the region of Asp-7 to Pro-11, where a turn was apparent. A strong contact between Asp-9 H N and Trp-10 H N , and other inter-residue contacts such as Trp-10 H 2 -Asp-9 H ␤ , Trp-10 H ␣ -Pro-11 H ␦ , and Asp-9 H N -Pro-8 H ␤/␥/␦ , helped to define this turn. The calculated structures ( Fig. 9 and Table IV) comprise a type I ␤-turn (44) from Asp-7 to Trp-10. This structure would expose the Pro and Trp side chains for interactions with the antibody (Fig. 9C), a hypothesis consistent with the STD-NMR data.
The average root mean square difference for superimposition of the 10 structures with lowest NOE restraint energy, with the lowest energy structure as a template, was 0.46 Å for the backbone atoms (N, C ␣ , C, O) of residues 7-10.
The question of the mechanism of peptide-carbohydrate mimicry requires an appreciation of the modes of interaction of the natural ligands with the antibody as well as those of the mimetic peptide ligands. Therefore, we also investigated the recognition of oligosaccharide fragments of GBSPIII with mAb S9.
The oligosaccharide ligands employed in this study were decasaccharides (Scheme 3) corresponding to two pentasaccharide repeating units (Scheme 1) of GBSPIII. Thus, 1 is a "natural" decasaccharide, whereas 2 is modified in that the terminal sialic acid residue bears an N-propionyl substituent on C-5, replacing the natural N-acetyl substituent group (45). It is noteworthy that the N-propionylated polysaccharide of group B N. meningitidis is immunogenic and has led to the development of a specific vaccine (46). The syntheses of 1 and 2 are described in Ref. 45.
The STD-NMR spectra of a mixture of decasaccharides 1 and  9. Views of the calculated bound structure of the peptide FDTGAFDPDWPA (peptide 1). A, backbone (N, C ␣ , C, O) atoms of the 10 structures with lowest NOE restraint energy produced by simulated annealing refinement, superimposed using backbone atoms of residues 7-10. The lowest energy structure is colored by atom (C, green; N, blue; O, red), and a ribbon representation is shown in green. The other structures are shown in light blue. Residues 7-10 comprise a type I ␤-turn and are labeled. B, close-up views of residues 6 -12 of the lowest energy calculated structure (colored by atom). Backbone atoms only are shown for residues 6 and 12. In the left panel, the turn is shown in an orientation similar to the structures in A; in the right panel, it is shown in an orthogonal orientation, so that the exposed position of the Trp-10 side chain, on one face of the turn, is visible. C, close-up views of residues 6 -12, as in B, but in a typical type I ␤-turn conformation, as observed for the free peptide in solution. The conformation is very similar to the calculated bound conformation (B).
2 in the presence of mAb S9 revealed some interesting patterns. Strong enhancements of the N-acetyl methyl groups of 1 were observed. After addition of 2, strong enhancements of the N-propionyl methyl and methylene groups were apparent (Fig.  10), indicating that the antibody does in fact recognize the N-propionyl groups on the sialic acid residues. Importantly, enhancements of the H-3 protons of sialic acid were not observed, and therefore, it is clear that the entire sialic acid residue does not contribute to the epitope. Small enhancements of the H-3c/cЈ, H-4a/aЈ, and (H-1a and/or H-6b/bЈ) proton resonances were also observed, indicating that certain parts of the c and a (Gal) and b (GlcNAc) residues contribute to the epitope. Enhancements of several other resonances in the 3.5-4.0 ppm region were apparent (Fig. 10); the assignment of the rest of these enhancements will require further experiments, because this is a region of significant resonance overlap. DISCUSSION Antibodies to the capsular polysaccharide of GBS play a critical role in host defense, and significant progress toward polysaccharide and polysaccharide-protein conjugate vaccines has been made (11,12,14). However, anti-GBS vaccine development is not yet complete because sufficient protective efficacy has not yet been demonstrated in human studies. The use of carbohydrate-mimetic peptides is another promising strategy to improve the efficacy of vaccines directed against carbohydrate targets. The ability of the carbohydrate-mimetic peptide FDTGAFDPDWPA (peptide 1) to elicit an anti-GBS SCHEME 3. Decasaccharides representing two repeating units of the GBS type III capsular polysaccharide. immune response has been demonstrated (14).
We describe here the conformational preferences of peptide 1, and two new analogs, as investigated by NMR spectroscopy. Peptide 1 exhibited a significant population of a type I ␤-turn conformation from residues Asp-7 to Trp-10. This conformation would be stabilized by an intramolecular Asp-7 O-Trp-10 HN hydrogen bond, as indicated by the reduced temperature coefficient of the Trp-10 amide proton. In contrast, the cis isomer of peptide 1 exhibited a preference for a type VI ␤-turn from residues Asp-9 -Ala-12, a conformation often observed in peptides containing cis-proline residues. As is the case for other short peptides, these conformations are not likely to be permanent, but rather are significantly represented in an ensemble of many rapidly interconverting conformers. TrNOE NMR studies of peptide 1 in the presence of the protective mAb S9 showed that the bound conformation of the trans isomer is also a type I ␤-turn, although in the calculated conformation, no intramolecular hydrogen bond was present. This type of turn would expose the side chains of residues 7-10 for specific interactions with the antibody combining site (Fig. 9B); this is consistent with the STD-NMR data, which showed enhancements of residues in the same region (Fig. 7). The turns present in the bound and free peptide are very similar, as shown in B and C of Fig. 9. Therefore, the bound conformation is also significantly represented in the free peptide, and this may be an important reason for the effective immunogenicity of peptide 1. The ␤-turn motif has also been observed in other cases, for many short, antigenic peptides free in solution (33,(47)(48)(49) and bound to antibodies (50 -53).
Peptide 1 competitively inhibited the binding of mAb S9 to GBS cells, and effective competition was also demonstrated for two new analogs (peptides 1A and 1B). It is interesting to note that rather than the predicted increased preference for the cis isomer in the two new analogs, these peptides showed a strong preference for the trans isomer. The peptide 1A showed 2-fold greater avidity for the antibody; this may result from the introduction of the terminal Asp residue or possibly from the modification of the Trp-10 indole ring.
The STD-NMR studies of the oligosaccharides 1 and 2 in the presence of mAb S9 showed that the N-acetyl group of either or both of the b (GlcNAc) and e (NeuAc) sialic acid residues is recognized strongly by the antibody. In addition, the N-propionyl group of the modified NeuAc residue is recognized, whereas the H-3 protons of this residue are not. These observations are consistent with the conformational epitope hypothesis (6,7). Therefore, the position of sialic acid within a larger, extended helical epitope may be such as to present the N-alkyl group to antibodies, while not exposing the rest of the monosaccharide. The charged carboxylate group is likely required for the formation of such a secondary structure, as shown by NMR spectroscopic (7) and immunochemical data (4). The STD-NMR data described above also show that the epitope comprises a surface extending over several residues of GBSPIII, with contributions being made by branching residues (c, Gal; e, NeuAc) and by backbone residues (a, Gal; b, GlcNAc). Several additional proton resonances giving rise to enhancements in the STD-NMR spectrum remain to be identified.
The ability of the carbohydrate-mimetic peptide isolated by phage display library screening, FDTGAFDPDWPAC (14), to inhibit competitively binding to GBS by both a mAb and polyclonal anti-GBS Abs could indicate the possibility of a significant level of true structural mimicry of GBSPIII, rather than functional mimicry, as has been demonstrated in several other cases (13). However, an alternative possibility is that the polyclonal Ab response in this case, as for other carbohydrates, is structurally restricted (54). This possibility remains to be in-vestigated, and structural investigations of this system are particularly interesting. Further, we note that the level of structural mimicry required for effective immunogenicity is still unknown. The success of a cross-priming strategy in vaccine development against tumor-associated antigens (55,56) and against Cryptococcus neoformans (57) has been demonstrated. The level of structural mimicry in these systems is unknown; however, the crystal structure of an anti-Cryptococcus mAb-peptide complex showed less than perfect complementarity, leaving open a reasonable possibility that the peptide is not a structural mimic (58). Therefore, structural mimicry is probably not a prerequisite for the effective use of peptide mimotopes in vaccination strategies.
Supporting Information Available-A table of restraints derived from the trNOESY data and used in the calculation of the bound conformation of peptide 1 is available in the supplemental material.