Conformational Studies of the O-specific Polysaccharide of Shigella flexneri 5a and of Four Related Synthetic Pentasaccharide Fragments Using NMR and Molecular Modeling*

As part of a program for the development of synthetic vaccines against the pathogen Shigella flexneri, we used a combination of NMR and molecular modeling methods to study the conformations of the O-specific polysaccharide (O-SP) of S. flexneri 5a and of four related synthetic pentasaccharide fragments. The NMR study, based on the analysis of 1H and 13C chemical shifts, the evaluation of inter-residue distances, and the measurement of one- and three-bond heteronuclear coupling constants, showed that the conformation of one of the four pentasaccharides is similar to that of the native O-SP in solution. Interestingly, inhibition enzyme-linked immunosorbent assay demonstrated that a protective monoclonal antibody specific for S. flexneri 5a has a greater affinity for this pentasaccharide than for the others. We carried out a complete conformational search on the pentasaccharides using the CICADA algorithm interfaced with MM3 force field. We calculated Boltzmann-averaged inter-residue distances and 3JC,H coupling constants for the different conformational families and compared the results with NMR data for all pentasaccharides. Our experimental data are consistent with only one conformational family. We also used molecular modeling data to build models of the O-SP with the molecular builder program POLYS. The models that are in agreement with NMR data adopt right-handed 3-fold helical structures in which the branched glucosyl residue points outwards.

Capsular polysaccharides (CPS) 1 and lipopolysaccharides (LPS) of Gram-negative bacteria are important virulence fac tors and major targets of the host's immune response upon infection (1). The potential of CPSs as vaccine candidates against bacterial infections was demonstrated in the early 1930s, and several polysaccharide vaccines such as those targeting Streptococcus pneumoniae, Neisseria meningitidis, and Salmonella typhi are now commercially available (2). However, these vaccines are ineffective in infants, thus polysaccharide:protein conjugate vaccines were developed, as illustrated by the Haemophilus influenzae b and N. meningitidis Group C vaccines that became available recently (3,4). Several laboratories are studying glycoconjugate vaccines derived from detoxified LPSs as protein conjugates involving the polysaccharide moiety of LPSs have been shown to be safe and immunogenic in humans (5). However, it would still be an advantage if we could improve the immunogenicity of such conjugates, although this may be difficult as we do not know much about the critical parameters. Possible alternatives to polysaccharide conjugate vaccines include synthetic molecules that accurately mimic the bacterial polysaccharide.
We are currently trying to develop such an alternative, targeting a synthetic vaccine against Shigella flexneri infections. S. flexneri is a Gram-negative bacillus that is responsible for the endemic form of shigellosis, a dysenteric syndrome. The disease, characterized by bacterial invasion of the human colonic mucosa (6), causes a high rate of mortality among infants, particularly in developing countries (7). It has been demonstrated that the Ospecific polysaccharide moiety (O-SP) of the LPS is the major target of the protective immune response. Indeed, protein conjugates of the polysaccharide moiety of S. flexneri serotype 2a appear to be promising vaccine candidates (8).
There is a close relationship between the shape and biological function of polysaccharides, and it is assumed that knowledge of the conformational behavior of the bacterial surface polysaccharide in solution may help the mimic approach. Indeed, the pioneering work by Lemieux et al. (9), and the development of powerful new NMR and molecular modeling methods for conformational analysis, have made it possible to study the conformation of bacterial polysaccharides in detail (10). In this study we concentrated on the model bacterium S. flexneri 5a, the specificity of which is defined by the structure of the repeating unit II, shown below, of its O-SP (11). , and by the Région Ile-de-France for the grant that allowed the Institut Pasteur to purchase the Varian 600-MHz spectrometer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Frame-shifted di-, tri-, tetra-and pentasaccharide fragments of the O-SP 5a, all bearing the characteristic EB segment, have been synthesized recently as their methyl glycosides with the natural anomeric configuration at their reducing end (13)(14)(15)(16)(17). To determine whether short fragments have some of the conformational features of the O-SP 5a, we compared the conformations of the four possible pentasaccharides 1, 2, 3, and 4 specific for the O-SP 5a ( Fig. 1) with that of the native O-SP. Given the putative flexible nature of oligosaccharides and polysaccharides, we used a combination of NMR and molecular modeling to study the conformational properties of these molecules. We also used inhibition enzyme-linked immunosorbent assay (ELISA) to investigate the recognition of a S. flexneri 5a-specific monoclonal antibody for these pentasaccharides.
The NMR conformational study was based on the analysis of 1 H and 13 C chemical shifts, the calculation of inter-residue distances, and the measurement of heteronuclear 3 J C,H coupling constants across the glycosidic linkages that are related to ⌽ and ⌿ torsion angles in a Karplus-type relationship (18). The 3 J C,H coupling constants were determined by two different techniques, EXSIDE (excitation-sculptured indirect-detection experiments) (19) and gradient-selected J-HMBC experiments (gs-J-HMBC) (20). The molecular modeling of the pentasaccharides involved a complete conformational search with the CI-CADA algorithm (21) interfaced with MM3 force field (22,23), which is very efficient for exploring the conformational space of flexible molecules (21), including oligosaccharides (24 -26). Boltzmann-averaged inter-residue distances and 3 J C,H coupling constants were calculated for the conformers generated and compared with NMR data for all frame-shifted pentasaccharides. As the regular helical conformation of polysaccharides in their solid state (27) can be used as starting point for studying their behavior in solution, the possible helical shapes of the O-SP 5a were determined by means of a molecular builder, POLYS (28), which combines a data base of monosaccharide structures with conformational information on disaccharide fragments. Based on the comparison of the NMR and molecular modeling data described herein with those obtained for the S. flexneri variant Y polysaccharide (29) (O-SP Y), the repeating unit of which is the tetrasaccharide I, we paid particular attention to the effect of the branched glucosyl residue E on the conformational properties of O-SP 5a.

EXPERIMENTAL PROCEDURES
Material-Pentasaccharides 1, 2, 3, and 4 were dissolved in 600 l of deuterium oxide to final concentrations of 28, 30, 24, and 12.7 mM, respectively. The O-SP 5a was prepared by acid hydrolysis of the LPS as described previously (30). Mass spectrometry analysis showed that the purified O-SP was composed of an average of 15 repeating units. About 10 mg of O-SP 5a were dissolved in 400 l of deuterium oxide.
Nomenclature-The two torsion angles describing a glycosidic linkage are defined as ⌽ ϭ O 5 -C 1 -O 1 -CЈ X and ⌿ ϭ C 1 -O 1 -CЈ X -CЈ X ϩ 1 where the CЈ atoms belong to the reducing side, and the sign is in agreement with the IUPAC nomenclature (31).
NMR Spectroscopy-NMR experiments were recorded on Varian Inova spectrometers, operating at 1 H frequencies of 500 and 600 MHz, equipped with a triple resonance pulsed field gradient probe with an actively shielded z gradient. Chemical shifts are given relative to an external standard of sodium 2,2-dimethyl-2-silapentane-5 sulfonate (DSS) at 0 ppm for both 13 C and 1 H chemical shifts. DQF-COSY (32), TOCSY (33), off-resonance ROESY (34), gHSQC (35), gHSQC-TOCSY (35,36), and gHMBC (35) experiments were performed as described previously (16,17), at 35 or 38°C for the pentasaccharides and at 50°C for the O-SP 5a. The 3 J H,H coupling constants were obtained from a one-dimensional spectrum with a digital resolution of 0.1 Hz/point or from the DQF-COSY experiment with a digital resolution of 0.5 Hz/ point. The 1 J C1,H1 coupling constants were measured from the gHMBC spectrum with a digital resolution of 0.5 Hz/point.
The NOESY experiments were carried out at 10°C with mixing periods of 100, 200, 400, and 600 ms for the pentasaccharides and at 50°C with mixing periods of 80, 100, 200, and 300 ms for the O-SP to obtain build-up curves.
Two different methods were used to measure long range heteronuclear 3 J C,H coupling constants, EXSIDE (19) and gs-J-HMBC (20). With EXSIDE, all proton band-selective pulses were Gaussian cascade Q3 pulses (37) with phase modulation to achieve off-resonance inversion. They were generated using the Pandora's Box pulse-shaping program available in the Varian software. Sixteen scans of 2048 complex points were collected for each of the 512 t 1 increments. A J-scaling factor (N) of 15 was applied. A recovery delay of 2.5 s was used prior to each scan. The spectra were transformed after zero filling to 2048 ϫ 4096 complex points using the unshifted Gaussian window function along the F 1 and F 2 dimensions. In EXSIDE spectra, the 3 J C,H coupling constants were measured in the F 1 dimension with a digital resolution of 0.3 Hz/point. For the gs-J-HMBC experiments, nine two-dimensional spectra were acquired with 16 and 96 scans per increment (356 t 1 increments) for the pentasaccharides and the O-SP 5a, respectively. A constant time delay max of 230 ms and values from 50 to 220 ms were used. The 3 J C,H coupling constant values were obtained by measuring cross-peak intensities as a function of and fitting them to the equation, y ϭ A sin ( n J C,H ).
Indeed, the intensity, s(t 1 ,t 2 ), of a particular correlation in a HMBC spectrum depends on the amplitude of the long range J C,H and can be described by the following equation (20,38): The 13 C spin-lattice relaxation times of O-SP 5a were measured at 125 MHz by means of two-dimensional double-INEPT-type inversedetected experiments with suppression of cross-correlations (39). Eight experiments were performed with relaxation delays, , ranging from 0.005 to 1 s. T 1 values were deduced by measuring the cross-peak intensities as a function of T and fitting the volumes to the equation, y ϭ A exp(Ϫ/T 1 ). The estimated error on the data points was 5 noise root mean square deviation. All two-dimensional data, except for EXSIDE and HMBC data, were collected in the phase-sensitive mode using the States-Haberkorn method (40).
Calculation of Inter-proton Distances from Cross-peak Volumes-The cross-peak volumes from off-resonance ROESY (400-ms mixing time) and NOESY experiments were measured with the VNMR software. The distances between neighboring protons were calculated by use of the usual 1/r 6 NOE/distance relationship (41). NOE-derived distances were obtained from initial NOE build-up rates, which were calculated by fitting NOE volumes at the different mixing times. The intra-residue distance of 2.52 Å between the H-1 and H-2 protons of the ␣-rhamnose unit B was used as a reference for distance calibration.
Energy Calculations-All geometry optimization steps were performed using the molecular mechanics program, MM3 (22,23). The block-diagonal minimization method was used for geometry optimization, with the default energy-convergence criterion (0.00008*n kcal/mol per five iterations, n ϭ number of atoms). The dielectric constant was set at 78 to attenuate the effect of hydrogen bonding on the potential energy surface.
Starting Models-All the disaccharides and oligosaccharides were built using MONOBANK, a data base of the three-dimensional structures of monosaccharides (42).
Relaxed Energy Maps of the Disaccharides-The five disaccharides constituting the O-SP 5a were the subject of a systematic grid search study for the conformation of the glycosidic linkage. The calculations were performed on disaccharide methyl glycosides. Starting from minimized disaccharides, the ⌽ and ⌿ torsion angles were increased by 20°s teps over the whole range, whereas the molecular mechanics program, MM3, provided full geometry relaxation. Several maps were calculated for each disaccharide to take into account several possible orientations of the primary and secondary hydroxyl groups. A maximum of 24 starting geometries were needed to take into account the two most stable staggered orientations of the hydroxymethyl groups, referred to as gg and gt, and the two possible networks of secondary hydroxyl groups around each ring referred to as clockwise and counterclockwise. For each disaccharide, the results of these calculations were projected on a so-called adiabatic map where only the conformer with the lowest energy for each (⌽,⌿) value is considered. Iso-energy contours were then plotted by interpolation of 1 kcal/mol within an 8 kcal/mol window.
CICADA Calculations-The potential energy surface of the pentasaccharides was explored with the CICADA program (43). Input for the CICADA program, which is an interface to the MM3 force field, consists mainly of one or a few conformers in MM3 format and a file containing the list of torsion angles to be driven and/or monitored. During the CICADA calculations, each selected torsion angle is driven one after the other in each direction from the initial conformation at a given increment. For the pentasaccharides, the driven torsion angles were ⌽ and ⌿ at each linkage and the torsion angle of each hydroxymethyl group (O 5 -C 5 -C 6 -O 6 ), leading to a dimensionality of 10 for the potential energy surface to be explored. The torsion angles of the N-acetyl group (C 1 -C 2 -N-C 8 ) and of all the secondary hydroxyl groups were monitored but not driven. The increment step was set at 20°, and two conformations were considered to be different when one of the driven or monitored angles differed by at least 30°. A relative energy cut-off of 50 kcal/mol was applied for exploring the potential energy surface. The search was stopped when no new conformers with energy lower than 5 kcal/mol could be detected.
Analysis of the Potential Energy Surface-The conformations and transition states found by CICADA were analyzed by the PANIC program (44), which explores the paths along the potential energy surface. Conformations were clustered into families within an energy window of 5 kcal/mol with the FAMILY program (45). In our study, a conformer was considered to belong to a conformational family if at least one of its torsion angles differed by less than 10°from at least one of the conformers of the family. Their population was calculated as for the relative importance of each family. The relative population of the ith conformational state, Pi, with energy, Ei, is dictated by the Boltzmann distribution, Pi ϭ exp(ϪEi/kT)/ exp(ϪEi/kT).
Calculations of Theoretical Distances and Coupling Constants-For each pentasaccharide, Boltzmann-averaged interproton distance ͗r Ϫ6 ͘ matrices and ͗ 3 J C,H ͘ coupling constants were calculated for the two lowest energy conformational families and for the average of all conformational families with a population of more than 1% at 298 K.
The 3 J C,H coupling constants were calculated by use of a Karplustype equation, 3 J C,H ϭ 5.7 cos 2 H Ϫ 0.6 cos H ϩ 0.5, for C-O-C-H segments (18). We then used the fractional population (Pi) for each conformational microstate, to calculate the average interproton distances and coupling constants as follows: ͗r͘ ϭ (⌺ Pi.r Ϫ6 ) Ϫ1 ⁄6 and Construction of the O-SP 5a-Possible conformations of the O-SP 5a were modeled using the molecular builder, POLYS (28). Different polysaccharide fragments containing six repeating units were constructed using all the combinations of the energy-minima obtained from adiabatic energy maps calculated on the disaccharides constituting it. The models that did not present serious steric conflicts were submitted to a POLYS procedure that optimized the values of ⌽ and ⌿ at each glycosidic linkage to give the closest regular fold helical symmetry.
Inhibition ELISA-A standard curve was established with IgGC20, a protective monoclonal antibody specific for S. flexneri 5a. 2 Different concentrations of the antibody were maintained at 4°C overnight and then placed in microtiter plates coated with purified S. flexneri 5a LPS at a concentration of 5 g/ml in carbonate buffer at pH 9.6 and previously incubated with 1% PBS/bovine serum albumin for 30 min at 4°C. After washing with PBS-Tween 20 (0.05%), alkaline phosphatase-conjugated anti-mouse IgG was added at a dilution of 1:5000 (Sigma) for 1 h at 37°C. After washing with PBS-Tween 20 (0.05%), the substrate was added (12 mg of p-nitrophenylphosphate in 1.2 ml of Tris, HCl buffer, pH 8.8, and 10.8 ml of 5 M NaCl). Once the color had developed, the plate was read at 405 nm (Dinatech MR 4000 microplate reader). A standard curve OD ϭ f(antibody concentration) was fitted to the quadratic equation Y ϭ aX 2 ϩ bX ϩ c, where Y is the OD, and X is the antibody concentration. Correlation factors (r 2 ) of 0.99 were routinely obtained.
The amount of oligosaccharides that inhibited the binding of IgGC20 to LPS by 50% (IC 50 ) was then determined as follows. A given concentration of IgGC20 (the minimal concentration of antibody that gave the maximal OD on the standard curve) was incubated overnight at 4°C with various concentrations of each of the oligosaccharides to be tested, in 1% PBS/bovine serum albumin. Unbound IgGC20 was measured as described above using microtiter plates coated with purified LPS from S. flexneri 5a, and the antibody concentration was deduced from the standard curve. IC 50 was then determined.

RESULTS AND DISCUSSION
NMR Spectroscopy-The protons and carbons of the methyl glycosides 3 and 4 and those of the O-SP 5a were assigned as described previously for pentasaccharides 1 (16) and 2 (17) using one-and two-dimensional NMR spectra such as DQF-COSY, TOCSY, gHSQC, gHMBC and gHSQC-TOCSY, which are the only means of overcoming all ambiguities. Given that different references were used, the 1 H and 13 C chemical shifts of the O-SP 5a determined here are in good agreement with those published previously (46,47).
Comparison of the 1 H and 13 C chemical shifts of the pentasaccharides with those of the O-SP 5a showed that the lack of a third branching sugar on residue B of pentasaccharides 1 and 3 affects not only residue B but also internal residues (see Supplemental Tables S1 and S2). For example, the chemical shifts of the H-2 C , H-3 C , and C-3 C of 1 and of the H-4 A , H-5 A , and C-5 A of 3 did not coincide with those of the corresponding atoms in the native O-SP. Moreover, all the 1 H and 13 C chemical shifts of internal residues of pentasaccharide 4 were similar to those of the O-SP 5a, with the exception of protons H-1 B and H-5 B (see Tables I and II). For pentasaccharide 2, many more proton and carbon differences were observed (see Supplementary Tables S1 and S2).
As chemical shifts are extremely sensitive to conformation, data show that the conformation of pentasaccharide 4 mimics that of the native O-SP most closely. The large differences observed for residues at the reducing and non-reducing ends were because of glycosylation effects or due to the presence of the methyl aglycone. Comparison of the 1 H and 13 C chemical shifts of O-SP 5a with those of O-SP Y (29) (see Tables I and  II) showed that the presence of the glucosyl residue E at position 3 of rhamnosyl residue B in O-SP 5a significantly affected the chemical shifts of the backbone residues. Major differences in chemical shifts were clearly visible for residue B. Additionally, minor differences were observed for residues A and C, the former being involved in a 2,3-cis vicinal pattern with residue E. The chemical shifts of residue D were also affected by residue E. These observations demonstrate the critical effects of the branched glucopyranoses E on the overall conformation of the O-SP 5a backbone in solution.
The vicinal coupling constants 3 J H,H of 3 and 4 are fully consistent with a 1 C 4 conformation for the L-rhamnopyranoses (A, B, and C) and a 4 C 1 conformation for the D-glucose and the N-acetyl D-glucosamine residues (E and D) (see Supplemental  Table S3), as was observed previously for the frame-shifted pentasaccharides 1 (16) and 2 (17).
As expected, the heteronuclear one-bound 1 J C-1,H-1 coupling constants measured in the gHMBC experiment for each pentasaccharide and O-SP 5a are in agreement with an ␣ configuration for the L-rhamnopyranose and D-glucose residues and with a ␤-configuration for N-acetyl D-glucosamine residue (see Table II and Supplemental Table S4) (48). The 1 J C-1,H-1 coupling constants of the rhamnose residues A and B in the native O-SP are slightly higher than expected for such residues (173 versus 170 Hz). Interestingly, among the synthetic pentasaccharides, only compound 4 presented this increase for both residues A and B. The higher 1 J C-1,H-1 coupling constants may be because of an increase in steric constraints surrounding C-1 and C-2 of these residues (49). This argument is supported by the 13 C spin-lattice relaxation times (T 1 ) (Table II) of O-SP 5a. Indeed, the 13 C T 1 values of C-1 A and C-2 A show that these carbon atoms are in more rigid environment than the other of its cycle (300 versus 350 ms). C-1 B and C-2 B presented the lowest 13 C T 1 values (Ϸ270 ms). On the contrary, the high 13 C T 1 values of the branched glucosyl residue E are suggestive of greater mobility.
We examined the conformations of the four pentasaccharides and of O-SP 5a in more detail by use of the inter-residue 1 H-1 H distances obtained from NOESY and off-resonance ROESY experiments. The ROESY experiment avoids the spin diffusion effect and Hartmann-Hahn artifacts (34). The averaged interresidue 1 H-1 H distances deduced from ROE and NOE volumes were similar (see Table III and Supplemental Table S5). Moreover, no significant difference was observed between the interresidue 1 H-1 H distances of the four pentasaccharides and those of the native O-SP. Generally, NOE and ROE cross-peaks across glycosidic linkages can fit many different conformations. Here, the large number of connectivities observed in addition to NOEs and ROEs across glycosidic linkages between anomeric and aglyconic protons are useful for the definition of a single conformation. Furthermore, the presence of long range NOE and ROE connectivities, such as H-6 A /N-Ac D in 2 and H-6 B /N-Ac D in 1, 2 (see Supplemental Table S5), and O-SP 5a, suggests that these molecules are folded. To complete this conformational analysis, long range 3 J C,H coupling constants across the glycosidic linkages, which are related to ⌽ and ⌿ torsion angles in a Karplus relationship, were determined from two different  NMR experiments, EXSIDE and gradient-selected J-HMBC (Table IV). EXSIDE is a semiselective version of the gradient-selected HSQC sequence (50, 51) that provides cross-peaks J-scaled in the carbon dimension. The 3 J C,H coupling constants are measured directly from the spectrum (Fig. 2) with no interference from any homonuclear couplings because of a proton band selection based on the excitation-sculpting technique (52). The unique chemical-shift region of anomeric protons in oligosaccharides means that this method can easily be used to rapidly obtain coupling constants involving H-1 protons.
Nevertheless, it is difficult to measure some 3 J C-1,H-X coupling constants because of the overlap of proton resonances, which makes it impossible to avoid interference from homonuclear couplings. To obtain the missing 3 J C-1,H-X coupling constants, gs-J-HMBC experiments were performed, although this method of obtaining 3 J C,H coupling constants is much more time consuming than the EXSIDE method. Indeed, several two-dimensional spectra must be acquired with different evolution time, . As the amplitude of signals are modulated by sin ( J C,H ), the heteronuclear long range 3 J C,H coupling constants are obtained by fitting a sine curve to the experimental data. According to the 3 J C,H values measured (see Table IV and  Supplemental Table S6    We investigated the conformational behavior of the four pentasaccharides (1, 2, 3, and 4) with the CICADA algorithm (21) interfaced with MM3 force field, starting from the geometries of energy minima of each glycosidic linkage. After about 10,000 energy minimizations, the CICADA calculations led to a total of about 3000 or 4000 energy minima on the potential energy surface of each pentasaccharide. As the conformational analysis was performed in ten-dimensional conformational spaces, it is not easy to describe the results. Projection of the calculated conformations of each glycosidic linkage on the energy map confirmed that the conformational space had been well explored (see Supplemental figures). All of the conformations resulting from the CICADA analysis were clustered into different conformational families within an energy window of 5 kcal/ mol above the global minimum (see Supplemental Tables S7  and S8).
The ⌽ torsion angles for all conformational families of all pentasaccharides are always similar. The definition of the different families depends essentially on the ⌿ C-D and ⌿ D-A torsion angles. Indeed, the conformers of the major families can adopt one of the low energy conformation I or II for both C-D and D-A glycosidic linkages, whereas they always adopt the energy minimum conformation I for A-B, B-C, and E-B glycosidic linkages. For pentasaccharide 4, the only difference between the lowest energy conformations of the two major conformational families is the ⌿ D-A torsion angle, which adopts one of the low energy conformations I or II (Table V).
These lowest energy conformations are energetically equivalent. They belong to a very flat plateau that makes interconversions easy. Fig. 4 shows the two lowest energy families of pentasaccharide 4.
Combination of NMR and Modeling Data-Averageweighted inter-proton distances and heteronuclear 3 J C,H coupling constants were calculated for the two lowest energy families, Fam. 1 and Fam. 2, and for the average of all families with an energy-weighted population of more than 1% at 298 K (see Supplemental Tables S9 and S10). The experimental values were in agreement with the conformations adopted in Fam.1 for the pentasaccharide 1 and with those adopted in Fam.2 for pentasaccharides 2, 3, and 4 (see Supplemental Table S9). Indeed, the distances measured between residues A and D for pentasaccharide 4 are almost in agreement with the conformations of Fam.2, in which the ⌿ D-A torsion angle adopts the low energy conformation II (Table III). This is substantiated by the comparison of experimental and calculated 3 J C,H coupling constants and particularly those of C-1 D /H-2 A atom pairs (Table  IV). Thus, the NMR data of all pentasaccharides showed that the A-B, B-C, C-D, and E-B glycosidic linkages adopted the low energy conformation I and that the D-A glycosidic linkage adopted the low energy conformation II.
Immunochemical Properties of Oligosaccharide Fragments of O-SP 5a-Inhibition ELISA was used to evaluate the affinity of a protective monoclonal antibody specific for S. flexneri 5a, IgGC20, for different frame-shifted di-, tri-, tetra-, and pentasaccharide fragments of O-SP 5a, with or without residue E. We found that residue E is essential for recognition. Moreover, the presence and positions of residues C and D in the different fragments seem to be important (Table VI). Indeed, the affinity was increased if residue C was added to the reducing end of the trisaccharide A(E)B-OMe. Conversely, the addition of residue D at the reducing end of the tetrasaccharide A(E)BC-OMe did not change the affinity. Similarly, the addition of residue D to the non-reducing end of the trisaccharide A(E)B-OMe improved the affinity whereas the addition of residue C at non-reducing end of the tetrasaccharide DA(E)B-OMe did not. The affinity of IgGC20 was highest when residues C and D were placed on either side of the trisaccharide A(E)B-OMe i.e. for the pentasaccharide DA(E)BC-OMe. DA(E)BC-OMe corresponds precisely to the pentasaccharide fragment that most closely mimics the conformational features of the O-SP 5a according the structural analysis.
Molecular Modeling of the O-SP-We used the POLYS program to build possible models of O-SP 5a using an approximation of independent neighboring glycosidic linkages (28). Different fragments composed of six repeating units were constructed using all combinations of the energy minima of disaccharide energy maps. Interestingly, only combination that gave a structure devoid of steric clashes corresponded to the energy minima adopted by all pentasaccharide fragments i.e. energy minimum I for A-B, B-C, C-D, and E-B glycosidic linkages and energy minimum II for D-A glycosidic linkages. This combination resulted in a conformation that was close to a right-handed helical structure with 3-fold symmetry. The helical parameters were then refined as a function of small variations of the ⌽ and ⌿ torsion angles at each glycosidic linkage around the different energy minima. Several 3-fold righthanded regular helices with similar overall shape could be built using conformations in this particular low energy region. Fig. 5 shows two possible helical structures of the O-SP 5a, O-SP 5a (1) and O-SP 5a (2), characterized by the torsion angles listed in Table VII.
The two models differ mainly by the extension of the helices. The O-SP 5a (1) structure has a pitch of 19.4 Å and a diameter of about 15 Å, whereas the O-SP 5a (2) structure has a pitch of 23.2 Å and a diameter of about 14 Å. The O-SP 5a (1) structure presents hydrogen bonds between HN-D and HO-3 A that help to stabilize the helical structure. Interestingly, for both models, the glucosyl residue E protrudes at the helix surface and is well exposed to the solvent. This is consistent with the high T 1 values measured for the carbons of these residues. The ability of O-SP 5a to form different helices with the same shape but different extensions, based on minor fluctuations in the ⌽/⌿  torsion angles, has been observed for other bacterial polysaccharides and may be correlated to their biological properties as was hypothesized in the case of Type III Group B Streptococcus CPS (GBSP III) (54). If O-SP 5a presents different conformations in solution, the NOEs will be averaged over all the conformations. The inter-residue distances measured for the two possible structures are almost in agreement with the NMR data, suggesting that the predominant conformation of O-SP 5a in solution is close to the regular 3-fold shape of O-SP 5a (Table III). The few small differences observed in relation to NMR data do not make it possible to discriminate between the two models. The role of residue glucosyl E in the helix formation was deduced by comparing our model of O-SP 5a with that of O-SP Y constructed by the hard-sphere exo-anomeric approach (29). The model of O-SP Y (Fig. 6), which is in agreement with NMR data, is characterized by the ⌽ and ⌿ torsion angles given in Table VII. According to the ⌽/⌿ values, residue E only seems to affect the ⌿ A-B and ⌿ B-C torsion angles. Indeed, these angles adopt the low energy conformation I in O-SP 5a and the low energy conformation II in O-SP Y. The O-SP Y model consists of a randomly coiled compact chain. Thus, the helical shape found for the O-SP 5a structure is probably because of the presence of its branched residues E. The conformational features of bacterial polysaccharides are often discussed in terms of the size and shape of antigenic determinants. In relation to vaccine development, the immunological functions of these polysaccharides can be considered to be the specific interaction of carbohydrate epitopes and antibody binding sites. Carbohydrate epitopes often contain two to four sugar residues effectively interacting with the paratope and may be compact with a relatively rigid conformation even though they belong to a rather flexible polysaccharide (55). Alternatively, they may be flexible, thus allowing antibodies to bind to selected conformations. Extensive analysis of the meningococcal Group B CPS (56) and of the type III Group B Streptococcal CPS (54) has led to the concept of conformational epitope, thus outlining a key feature in the design of functional molecules mimicking such bacterial polysaccharide. Overall, the three-dimensional conformation of biologically active epitopes, isolated either as short oligosaccharides or as related mimics thereof, may not be identical to that of the corresponding fragment in the native polymer, emphasizing the need for comparative conformational studies.
Our study, involving a combination of NMR and molecular modeling analysis, showed that the conformations of all pentasaccharide fragments of the O-SP of S. flexneri 5a are rather similar to that of the native O-SP. Indeed, the pentasaccharides and O-SP 5a adopt the same low energy conformations. However, according to 1 H and 13 C chemical shifts analysis and 1 J C-1,H-1 coupling constants the pentasaccharide 4, DA(E)BC-OMe, appears to mimic the conformation of the O-SP 5a highly   accurately. Inhibition ELISA showed that the protective monoclonal antibody specific for O-SP 5a, IgGC20, has a higher affinity for this pentasaccharide than for the others. Thus, NMR and antigenicity data seem to suggest that the pentasaccharide 4 might be likely to induce polysaccharide-specific antibodies that are effective against S. flexneri 5a and could therefore be used in conjugated vaccines against this pathogen.
Both NMR and molecular modeling data imply that O-SP 5a adopts a right-handed 3-fold helical structure with the glucosyl residue E pointing outwards. Comparison of this structure with a model of O-SP Y revealed that the branched glucosyl residue E, which constitutes the structural specificity of O-SP 5a, is responsible for the helical shape of the latter. Inhibition ELISA using oligosaccharide fragments of O-SP 5a showed that residue E is essential for antibody recognition. Interestingly, according to our models of O-SP 5a, this residue is ideally located to interact with antibodies. As found for S. dysenteriae type 1 (57) or Group A Streptococcus (58), this residue is likely to be one of the epitopes recognized by protective monoclonal antibodies targeting S. flexneri 5a infection. To test this hypothesis, we are currently carrying out structural studies of the interaction of the pentasaccharide fragments of the O-SP 5a with such antibodies (59).