Effect of the nonreducing end of Shigella dysenteriae type 1 O-specific oligosaccharides on their immunogenicity as conjugates in mice

Abstract

Endemic and epidemic shigellosis, an acute invasive disease of the lower intestines, afflicts millions of people worldwide with an estimated one million fatalities per annum at a low infectious dose. Our approach to vaccine development against Shigella is based on the hypothesis that serum IgG antibodies to the O-specific polysaccharide (O-SP) domains of the LPS of these organisms confer protection to infection. The synthetic oligosaccharides corresponding to the tetrasaccharide repeating unit of the O-SP of Shigella dysenteriae type 1 covalently linked to human serum albumin elicited O-SP-specific IgG in mice. The antibody levels were a function of both the saccharide chain length and their loading on the protein. These synthetic saccharide conjugates elicited significantly higher levels of IgG anti O-SP than conjugates prepared with the O-SP from the bacteria. Here, we evaluated the influence of the nonreducing terminal monosaccharide on the serum antibody response. To this end, we prepared synthetic oligosaccharides comprising hexa- to tridecasaccharide fragments of the native O-SP, having one of the four monosaccharide residues that constitute the repeating unit at their termini and bound them to BSA by a single-point attachment. The conjugates contained an average of 19 saccharide chains per BSA. The synthetic oligosaccharides inhibited the binding of serum raised against whole bacteria to its LPS to a similar extent but lower than the native O-SP. The highest anti-LPS levels were elicited by conjugates having N-acetylglucosamine (10-mer) or galactose residues (7- and 11-mers) at their nonreducing termini.

Bacillary dysentery continues to be a major human disease despite the fact that Shiga's bacillus (Shigella dysenteriae type 1), the first recognized species of the genus of Shigella, has been known for more than a century (1). S. dysenteriae type 1 causes endemic and epidemic shigellosis, characterized by high fever, cramps, seizures, bloody diarrhea and dysentery, hemolytic uremic syndrome, and death. Although shigellosis caused by S. dysenteriae type 1 is rare in developed countries, it is a frequent cause of disease in the developing world (2–4). It is estimated that from 1966 to 1997, ≈160 million cases of shigellosis occurred worldwide, with more than one million fatalities (5). Because of resistance of Shigella to most antibiotics, treatment of shigellosis is becoming increasingly difficult (2). Although the spread of this disease could be controlled by improved hygienic conditions in the affected areas, this is not likely to occur in the near future. Vaccination would control and potentially eradicate shigellosis; however, there is yet no licensed Shigella vaccine.

Our approach to vaccine development is based on the hypothesis that serum antibodies to the O-specific polysaccharide (O-SP) domain of the LPS confer immunity by killing the inoculum of homologous bacteria on the epithelial surface of the small intestine (6). Although the O-SPs are nonimmunogenic, presumably because of their low molecular weights, they can be converted to immunogens by their covalent attachment to immunogenic proteins (7). Such conjugates of S. dysenteriae type 1, Shigella flexneri type 2a, and Shigella sonnei elicited O-SP-specific antibodies in mice and in humans and were effective in preventing S. sonnei infection in Israeli soldiers (7, 8). Based on our experience, as well as on published data (9), we hypothesized that protein conjugates of oligosaccharides, smaller than the native O-SPs, may also elicit O-SP-specific antibodies. Recent advances in carbohydrate chemistry have enabled the synthesis of extended oligosaccharide chains (10). The use of protein conjugates of such oligosaccharides may have advantages over conjugates prepared with high-molecular-weight polysaccharides. Structurally well defined oligosaccharides may lead to a better understanding of the molecular requirements for their immunogenicity.

Several factors are related to the immunogenicity of the polysaccharide component. This paper is concerned with the relation between the nonreducing terminal monosaccharide of synthetic O-SP of S. dysenteriae type 1 and their immunogenicity as conjugates with BSA. The repeating unit of the O-SP is a tetrasaccharide of the structure: [→3)-α-l-Rhap-(1→2)-α-d-Galp-(1→3)-α-d-GlcpNAc-(1→3)-α-l-Rhap-(1]_n (11).

Antiserum to S. dysenteriae type 1 O-SP may be raised by injection of inactivated bacteria into experimental animals, by disease, or by asymptomatic infection with cross-reacting organisms (“natural” antibodies). We have mapped the effects of the oligosaccharide length and the number of saccharide chains per protein (density) on the immunogenicity of conjugates. A maximal O-SP antibody response was observed with conjugates of four repeating units and ≈10 chains per human serum albumin (12).

Another factor affecting the serum antibody response is the nonreducing terminal residue of oligosaccharides. Goebel et al. (13) first showed that the specificity of antibodies induced by synthetic disaccharides bound to horse globulins was related to the structure of their nonreducing ends (13). Later, Karush (14) showed that the major portion of binding energies of rabbit antibodies elicited by lactoside-protein conjugates was directed to the terminal β-linked galactose. We evaluate the relation between the levels of IgG antibodies elicited by conjugates of this antigen differing in their nonreducing termini.

Results and Discussion

We reported the immunogenicity of conjugates of spacer-linked synthetic oligosaccharides, containing the Rha-Gal-GlcNAc-Rha repeating unit bound at their reducing end to a carrier protein by single point attachment (12). We showed that the synthetic oligosaccharide-protein conjugates induced higher antibody levels than the conjugates of the native O-SP. The immunogenicity of the synthetic product was related to the saccharides' chain length (i.e., the number of repeating units in the chain) as well as to the density of the saccharide chains on the protein. A chain length of one repeating unit (four monosaccharides) was nonimmunogenic. The optimal number of oligosaccharide chains was ≈10 for the hexadecasaccharide, both lower and higher incorporations resulted in a lesser antibody response.

We propose that another variable related to the anti-O-SP immunogenicity of the synthetic oligosaccharide–protein conjugates is the identity of the monosaccharide at the nonreducing end. To test the validity of this hypothesis, we synthesized a panel of oligosaccharides 1–10 ranging from hexa- to tridecasaccharides differing by the chain length and by the residue at the nonreducing terminus (Table 1). To minimize the uncertainty caused by the addition of a saccharide to its progenitor, thereby increasing the chain length, we also prepared three decasaccharides (5–7) differing from each other by their termini.

Table 1.

Oligosaccharides synthesized in this study

α-d-GlcpNAc-1,3-α-l-Rhap-1,3-α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap-R	1A R = A, 1B R = B, 1C R = C
α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap-1,3-α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap-R	2A R = A, 2B R = B, 2C R = C
{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-R	3A R = A, 3B R = B, 3C R = C
α-l-Rhap-1,3-{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-R	4A R = A, 4B R = B, 4C R = C
α-d-GlcpNAc-1,3-α-l-Rhap-1,3-{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-R	5A R = A, 5B R = B, 5C R = C
α-l-Rhap-1,3-{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-α-l-Rhap-R	6A R = A, 6B R = B, 6C R = C
{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-α-l-Rhap−1,2-α-d-Galp-R	7A R = A, 7B R = B, 7C R = C
α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap-1,3-{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}2-R	8A R = A, 8B R = B, 8C R = C
{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}3-R	9A R = A, 9B R = B, 9C R = C
α-l-Rhap-1,3-{α-l-Rhap−1,2-α-d-Galp-1,3-α-d-GlcpNAc-1,3-α-l-Rhap}3-R	10A R = A, 10B R = B, 10C R = C

Competitive Inhibition of Binding.

The synthetic oligosaccharides inhibited the binding of the serum raised against whole bacteria to the homologous LPS within a 20–39% range. There was little difference between the inhibition exerted by 50–100 μg of any inhibitor. The native O-SP was a better inhibitor than any of the synthetic oligosaccharides (Table 2). In the hexa- to tridecasaccharide range, there was no correlation between the inhibition of the binding and the structural characteristics of the oligosaccharide inhibitors. These experiments suggest that the polyclonal sera elicited by the whole bacteria contain specificities to a variety of antigenic sites. Therefore, binding studies do not predict the optimal immunogen, notwithstanding the finding that monoclonal antibodies show distinct affinity specificities to oligosaccharides, as demonstrated e.g., by Glaudemans and coworkers (15) and Phalipon et al. (16).

Table 2.

Competitive inhibition of the hyperimmune anti-S. dysenteriae type 1 serum binding to the homologous LPS by different synthetic oligosaccharides 1A-5A and 8A-10A and by the S. dysenteriae type 1 O-SP

Inhibitor	Nonreducing end group of oligosaccharide	Inhibition, %Inhibitor per well, μg
		100	75	50	25
1A {S.dys., 6-mer}	GlcNAc	17	20	20	7
2A {S.dys., 7-mer}	Gal	24	14	10	9
3A {S.dys., 8-mer}	Rha	28	22	24	8
4A {S.dys., 9-mer}	Rha	17	25	17	5
5A {S.dys., 10-mer}	GlcNAc	15	22	23	12
8A {S.dys., 11-mer}	Gal	20	21	23	11
9A {S.dys., 12-mer}	Rha	17	19	25	9
10A {S.dys., 13-mer}	Rha	35	39	30	9
S. dys., O-SP	Unknown	71	75	78	50

S.dys., S. dysenteriae.

Serum Antibody Responses.

The compositions of the experimental conjugates and IgG anti-LPS levels elicited after three injections are shown in Table 3. The highest geometric means (GM) anti-LPS level [20.7 ELISA units (EU)] was elicited by the conjugate of the decasaccharide 5 having GlcNAc at the nonreducing end. The conjugate of hexasaccharide 1, also having a terminal GlcNAc residue but containing one less tetrasaccharide repeating unit than compound 5, elicited a much lower response (2.8 EU) that was still higher than any of the responses to the saccharides having a terminal Rha. A high antibody level (10.3 EU) was also obtained with the conjugate of the heptamer 2 having a terminal Gal. The response to the conjugate of undecasaccharide 8, also having a terminal Gal, was somewhat higher (12.1 EU), but the difference is not significant. In contrast, low responses (0.2–2.4 EU) were found for all oligosaccharide conjugates with either rhamnose residue at their nonreducing end. In summary, these data indicate that oligosaccharides with a Gal or a GlcNAc residue at their nonreducing terminus were optimal immunogens for this construct.

Table 3.

Characterization of the S. dysenteriae type 1 oligosaccharide/BSA conjugates and their effect upon anti-S. dysenteriae LPS antibody levels

Hapten	Nonreducing end group of oligosaccharide	Molecular mass of conjugate,* kDa	Average number of saccharide chains per BSA	Protein-sugar ratio, wt:wt	GM of IgG anti-LPS, EU
1, 6-mer†	GlcNAc	99.0	19	2.70	2.8
2, 7-mer	Gal	102.8	19	2.38	10.3
3, 8-mer	Rha	105.4	19	2.17	2.4
4, 9-mer	Rha	110.4	20	1.85	1.4
5, 10-mer	GlcNAc	109.6	18	1.89	20.7
6, 10-mer	Rha	108.1	17	2.00	1.1
7, 10-mer	Rha	110.4	18	1.89	0.7
8, 11-mer	Gal	115.2	19	1.64	12.1
9, 12-mer	Rha	117.3	19	1.56	0.2
10, 13-mer	Rha	117.5	18	1.54	0.2

2.8 vs. 20.7, P = 0.04; 2.8 vs. 12.1, P = 0.03; 10.3 vs. 1.1, P = 0.05; 10.3 vs. 0.7, P = 0.06; 10.3 vs. 0.2, P = 0.05; 2.4 vs. 20.7, P = 0.05; 2.4 vs. 12.1, P = 0.05; 20.7 vs. 1.1, P = 0.02; 20.7 vs. 0.7, P = 0.02; 20.7 vs. 0.2, P = 0.02; 1.1 vs. 12.1, P = 0.007; 0.7 vs. 12.1, P = 0.01; 12.1 vs. 0.2, P = 0.007; 2.8, 10.3, 2.4, 20.7, 12.1 vs. control group, P < 0.05.

*Molecular mass of aminooxy-derivatized BSA of 74.5 kDa was used for all conjugates.

^†mer, number of monosaccharides in the synthetic oligosaccharide; control group GM = 0.3.

The structure of the biosynthetic repeating unit is controversial, but all of the investigators agree about the monosaccharide sequence and their linkages. Most studies were done by transferring the rfb gene cluster from S. dysenteriae type 1 to Escherichia coli K-12. Using this approach, Sturm et al. (17, 18) showed that the order of the monosaccharides in the biological repeating unit is GlcNAc→Rha→Rha→Gal. In contrast, Fält et al. (19) found that the terminal saccharide of the core region was GlcpNAc, indicating that the structure of the biological repeating unit is Rha→Rha→Gal→GlcNAc. The same sequence was proposed recently by Feng et al. (20).

Using a monoclonal murine IgM antibody raised against the O-SP of S. dysenteriae type 1 and a number of synthetic oligosaccharides in inhibition experiments, Glaudemans and coworkers (15) proposed that the Rha-Gal disaccharide is the basic determinant of the O-SP. Mapping of the carbohydrate-binding specificity of four murine monoclonal antibodies showed that two recognized the Rha-Gal portion of the O-SP, whereas the other two recognized only the longer Rha-Rha-Gal-GlcNAc sequence (21). From the binding studies, it was concluded that a protein conjugate vaccine should include the entire Rha-Rha-Gal-GlcNAc sequence (21).

Conformational studies of the O-SP by Nyholm et al. (22) revealed a helical structure of the O-SP with pronounced exposure of the Rha-Gal disaccharide (22). Molecular modeling indicated a hairpin conformation for the Gal-GlcNAc disaccharide, with the individual units pointing in opposite directions. For a sequence comprised of eight repeating units, it was shown that the Gal residues are preferentially exposed. In contrast, the Rha and the GlcNAc units are only partially accessible. These studies led to the proposal that the conformational epitope of the O-SP is fully attained in Rha-Rha-Gal-GlcNAc-Rha pentasaccharide. Our finding that the conjugates of the Gal-terminated oligosaccharides elicit high levels of antibodies is in agreement both with antibody binding experiments (15) and with molecular modeling studies (22).

We found that the nonreducing terminus of synthetic O-SPs is an important determinant for the immunogenicity of these conjugates, possibly because conjugates of synthetic oligosaccharides contain a large number of termini compared with those prepared with the native polysaccharide. Despite the identical or very similar size of their saccharide components, the conjugates of 10-mers (5–7) and that of the 11-mer (8), all having similar loadings per carrier molecule, markedly differed as immunogens. The highest antibody level was elicited by conjugate 5 with GlcNAc terminus (20.7 EU), the next highest by conjugate 8 with Gal terminus (12.1 EU). These levels are not statistically different. Both these levels are significantly higher that those elicited by conjugates 6 and 7 terminating with Rha. The results of this study will aid the design of experimental vaccines against S. dysenteriae type 1 comprised of synthetic oligosaccharide-protein conjugates.

Materials and Methods

Chemical Synthesis of Oligosaccharide Glycosides 1C–10C.

We synthesized tetra-, octa-, dodeca-, and hexadecasaccharides comprised of one to four repeating units, as shown above (23). The choice of this frame was based on the high stereoselectivity of α-rhamnosylation compared with that of the formation of either α-galactopyranosyl or α-glucosaminyl linkages (23–25). In this study, we prepared the oligosaccharide glycosides 1C–10C (Table 1) with an oxo-group for their covalent attachment to aminooxy-functionalized proteins by using oxime chemistry (26, 27) from the amino group-containing precursors 1B–10B obtained from the corresponding methyl esters 1A–10A. The synthesis of oligosaccharide methyl esters 1A–5A, and 8A–10A has been described (23, 25, 28). The synthesis of methyl esters 6A and 7A is described herein.

The synthesis of decasaccharide 6A started with the attachment of the linker to the reducing end-terminal rhamnose moiety. Treatment of rhamnose trichloroacetimidate (23) 11 with alcohol 12 by using trimethylsilyl trifluoromethanesulfonate (TMSOTf) as the activator afforded the intermediate 13 in 82% yield from which the acceptor 14 was prepared by treatment with thiourea (Fig. 1). Next, alcohol 14 was condensed with tetrasaccharide trichloroacetimidate (23) 15 to afford the pentasaccharide 16 in 41% yield. Exposure of compound 16 to thiourea in the presence of pyridine cleaved selectively the monochloroacetyl protecting group to yield pentasaccharide alcohol 17. Iterative condensation of alcohol 17 with the donor 15 (→18) followed by removal of the monochloroacetyl group by thiourea provided the nonasaccharide alcohol 19 in 90% combined yield. The assembly of the decasaccharide 20 was completed by condensation of the alcohol 19 with the rhamnosyl donor 11 by using TMSOTf as the activator. Because rhamnosyl donors having a participating group at O-2 form 1,2-trans (α) linkages exclusively, we did not verify the configuration of the newly formed interglycosidic linkages. Verification was carried out, however, for the deprotected oligosaccharides. Conventional removal of the protecting groups by using Zemplén-type deacylation followed by hydrogenolysis afforded the unprotected decamer 6A that was purified by gel filtration through a BioGel P-4 column by using pyridine-acetate buffer as the eluant. The structure of 6A was confirmed by its NMR spectra and its high-resolution mass spectrum catalogued in SI Appendix. Proof for the all-trans interglycosidic linkages came from the one-bond, C-1, H-1 heteronuclear coupling constants, which were in the 168- to 174-Hz range, diagnostic for such linkages (29).

Fig. 1.

Reagents and conditions: (a) 1.1 equiv of 12, TMSOTf (cat), CH₂Cl₂, 0 → 22°C, 1 h, 82%; (b) 4.4 equiv of CS(NH₂)₂, DMF, C₅H₅N, 23°C, 24 h, 95%; (c) 2 equiv of 14, TMSOTf (cat), CH₂Cl₂, 0 → 23°C, 30 min, 41%; (d) 4.4 equiv of CS(NH₂)₂, DMF, C₅H₅N, 23°C, 24 h, 91%); (e) 2.3 equiv of 15, TMSOTf (cat), CH₂Cl₂, 0 → 22°C, 4 h, 94%; (f)) 4.4 equiv of CS(NH₂)₂, DMF, C₅H₅N, 23°C, 24 h, 90%; (g) 3 equiv of 11, TMSOTf (cat), CH₂Cl₂, 0 → 22°C, 4 h, 77%; (h) (1) NaOMe, MeOH, 24 h, (2) CH₂N₂; and (i) H₂, Pd/C, EtOH, 24 h.

The first step of the synthesis of the alternate sequence 7A was condensation of the highly reactive spacer 12 with the galactose donor (25) 21 in the presence of TMSOTf to afford an inseparable mixture of diastereomers 22α,β (β:α 4:1) (Fig. 2). The separation problem was solved later in the synthetic sequence as described below. Exposure of 22 to ceric ammonium nitrate removed selectively the p-methoxybenzyl group. The anomeric mixture so formed (23α,β) still could not be separated and was used directly in condensation with the rhamnose imidate (23) 11 in the presence of TMSOTf to afford the diastereomeric mixture of disaccharides 24α,β in 92% yield. After the removal of the monochloroacetyl group from the mixture 24α,β, the alcohols 25 and 26 could be separated, providing the required disaccharide block 26 in a stereochemically pure form. Next, acceptor 26 was condensed with the tetrasaccharide imidate (23) 15 by activation of the latter with TMSOTf to yield hexasaccharide 27 (32% yield) from which removal of the monochloroacetyl group (→28) followed by a second chain extension with the tetrasaccharide donor 15 afforded the fully protected decasaccharide 29. Saponification of the acyl groups followed by hydrogenolytic cleavage of the O-benzyl protecting groups afforded the unprotected tetrasaccharide construct 7A. The one-bond heteronuclear H-1, C-1 coupling constants in 7A confirmed the all-trans interglycosidic configurations.

Fig. 2.

Reagents and conditions: (a) 1.5 equiv of 12, TMSOTf (cat), CH₂Cl₂, −55 → −30°C, 1 h, 58%; (b) 2.7 equiv of (NH₄)₂Ce(NO₃)₆, THF, H₂O, 23°C, 1 h, 76%); (c) 2 equiv of 11, TMSOTf (cat), CH₂Cl₂ 0°C, 92%; (d) 4.4 equiv of CS(NH₂)₂, DMF, C₅H₅N, 23°C, 24 h, 95%; (e) 1.2 equiv of 15, TMSOTf (cat), CH₂Cl₂, 0 → 23°C, 30 min, 32%; (f) 4.4 equiv of CS(NH₂)₂, DMF, C₅H₅N, 23°C, 24 h, 75%; (g) 6 equiv of 15, TMSOTf (cat), CH₂Cl₂, 0 → 23°C, 30 min; (h) (1) NaOMe, MeOH, 24 h, (2) CH₂N₂; (i) H₂, Pd/C, EtOH, 24 h.

With the hexa- (1A) to tridecasaccharide (10A) methyl esters in hand, we incorporated the linking moiety in the oligosaccharide constructs. First, methyl esters 1A–10A were subjected to ethylenediamine to introduce free amino groups in the spacer moiety, resulting in constructs 1B–10B. These conversions were confirmed by the disappearance of the methoxy signals in the NMR spectra (3.7 ppm for CH₃ and ≈52.8 ppm for CH₃) and by the appearance of signals attributable to the NHCH₂CH₂NH₂ moiety (e.g., a pseudotriplet at ≈3.1–3.2 ppm for CH₂N and ≈38 and 40 ppm for NHCH₂CH₂NH₂. Next, amino-oligosaccharides 1B–10B were N-acylated with 5-ketohexanoic anhydride in MeOH to afford keto-derivatives 1C–10C. Proof of this conversion was given by the appearance of a four-proton pseudosinglet at ≈3.3 ppm (NHCH₂CH₂NH), a two-proton pseudotriplet at 2.6 ppm (CH₂COCH₃), a three-proton singlet for the CH₃CO moiety at 2.2 ppm in the ¹H-NMR spectra, and the signals at 20.4 ppm (CH₃CO), 30.1 (CH₂), 42.9 (CH₂), and 177.2 (COCH₃) in the ¹³C-NMR spectra, characteristic of the ketohexanoyl group. Further confirmation for the structures of oligosaccharides 1–10 was provided by their high-resolution mass spectra (see SI Appendix).

Conjugation of Oligosaccharides 1C–10C to BSA.

Covalent attachment of the keto-derivatized oligosaccharides 1C–10C to BSA was achieved by oxime formation (26, 27, 30–33) between the saccharides' keto group and aminooxypropyl groups bound to the protein through a thioether linkage (26, 27, 34–37). For this purpose, BSA was reacted with succinimidyl 3-(bromoacetamido)propionate (36) 30, as described (26, 27, 38) (Fig. 3). MALDI-TOF MS indicated the incorporation of ≈35 bromoacyl groups per BSA. Next, the bromoacyl-derivatized BSA 31 was treated with the aminooxypropylthiol reagent (39) 32 to provide aminooxylated BSA 33. Treatment of the keto-derivatized oligosaccharide glycosides 1C–10C with the aminooxypropylated BSA yielded the oligosaccharide–protein conjugates (Table 3).

Fig. 3.

Reagents and conditions: (a) BSA (90 mg, 1.4 μmol) in PBS (1.5 ml), 38 (20 mg, 65 μmol) in DMSO (100 μl), 22°C, pH 7.4, 2 h; (b) Sephadex G-50; (c) 40 (15 mg, 105 μmol), 22°C, pH 7.4.

Immunization.

All animal experiments were approved by the National Institute of Child Health and Human Development Animal Care and Use Committee. Five to six-week-old female National Institutes of Health general-purpose mice were immunized s.c., two or three times at 2-week intervals, with 2.5 or 10 μg of oligosaccharide as a conjugate in 0.1 ml of PBS. Groups of 10 mice were exsanguinated 7 days after the second or the third injections (40). Controls received PBS.

Serology.

Serum IgG antibodies were measured by ELISA by using human serum albumin for blocking (40). Antibody levels were calculated relative to a pool of highest-titer sera obtained from mice immunized three times and assigned a value of 100 EU. Results were computed with an ELISA data-processing program provided by the Biostatistics and Information Management Branch, Centers for Disease Control and Prevention (41). Competitive inhibition ELISA was done by incubating a pool of mice sera injected with whole killed bacteria, diluted 1/100 to give an A₄₀₅ of 1.0, with 25, 50, 75, or 100 μg per well of O-SP or oligosaccharide, incubated for 1 h at 37°C and overnight at 4°C (Table 3). The assay was then continued as above. Sera with inhibitor were compared with the serum pool without inhibitor at the same dilution. Percent inhibition was defined as (1-A₄₀₅ adsorbed serum/A₄₀₅ nonadsorbed serum) × 100%. Double immunodiffusion was performed as described (42), by using rabbit anti-BSA (Sigma, St. Louis, MO) and rabbit anti-S. dysenteriae type 1 whole killed bacteria.

Statistics.

ELISA values are expressed as the GM. Unpaired Student's t tests were used to compare GMs of different groups.

Abbreviations

O-SP

O-specific polysaccharide

EU

ELISA units

GM

geometric means

TMSOTf

trimethylsilyl trifluoromethanesulfonate.