α-GlcNAc-1→2-α-Glc, the Salmonella Homologue of a Conserved Lipopolysaccharide Motif in the Enterobacteriaceae, Elicits Broadly Cross-Reactive Antibodies

Abstract

To define cross-reactive epitopes in Salmonella lipopolysaccharide (LPS), antisera designated anti-S, anti-Ra, and anti-Re were generated against smooth (S), complete-core (Ra), and deep-core mutant (Re) strains, respectively, and characterized immunochemically. The reactivities of anti-Ra and anti-S with rough LPS (rLPS) chemotypes in enzyme-linked immunosorbent assays (ELISA) decreased progressively with increasing truncation of the complete-core oligosaccharide (e.g., Ra > Rb₁ >…Re), while that of anti-Re increased (Ra < Rb₁ <…Re). Anti-Ra was relatively more reactive with nonhomologous smooth LPS (sLPS) than anti-S, which in turn was more reactive than anti-Re. This order reflected the relative reactivities of these sera with outer-core rLPS but not those with inner-core rLPS, which suggests that the cross-reactivities of all three sera with sLPS were mediated by antibodies which bind outer-core determinants. Anti-Ra, but not anti-S or anti-Re, reacted with molecules substituted by O chains in immunoblots and revealed ladder-like patterns in sLPSs of various serospecificities. Anti-Ra, however, did not react with O-antigen-specific neoglycoconjugates in ELISA, thus demonstrating specificity for core epitopes. Ra and Rb₁ but not other Salmonella core chemotypes inhibited the reactivity of anti-Ra with sLPS in ELISA, which showed that the terminal outer-core disaccharide, α-GlcNAc-1→2-α-Glc (GlcNAc→Glc), was the major epitope of cross-reactive antibodies in the serum. GlcNAc→Glc represents the conserved motif α-hexose-1→2-α-hexose in cores of the Enterobacteriaceae, other homologues of which should likewise be cross-reactive. These results demonstrate that S or Re strains do not elicit cross-reactive antibodies and indicate that immunization with Ra strains may represent a general strategy for eliciting cross-reactive antibodies against LPSs from enteric bacteria.

There are currently no effective preventive or therapeutic modalities for bacterial sepsis and septic shock. About 50% of patients who develop septic shock die despite advances in antimicrobial chemotherapy and in critical care (8). Many of the pathological consequences of gram-negative sepsis are attributable to lipopolysaccharide (LPS) (endotoxin), an amphipathic component of the outer membrane. LPS protects bacteria against host defenses and initiates the inflammatory cascade by eliciting the release of pharmacologically active mediators from monocytes (40). LPSs from enteric bacteria and other major causative agents of gram-negative sepsis have a common architecture comprising three structural domains (Fig. 1). The innermost domain, lipid A, is a highly conserved molecule invested with the biological properties of LPS, while the outermost domain, the O antigen, is structurally heterogeneous. O antigens demonstrate antiphagocytic properties but also elicit highly protective immune responses. The core, a short oligosaccharide which bridges the O antigen and lipid A, is also highly conserved among enteric organisms, with only a single type (occasionally incompletely expressed) found among Salmonella species and five types found among Escherichia coli. While the role of the core in pathogenesis is poorly understood, there is evidence that it modulates the biological activities of lipid A by mediating selective interaction with proteins in serum and by binding to sites on host cell membranes (5, 39).

FIG. 1.

Schematic structure of Salmonella typhimurium LPS showing core-defective chemotypes. Hep, heptose; KDO, 2-keto-3-deoxyoctonate.

Because the O antigen is structurally hypervariable, there has been long-standing interest about the identification of common core determinants as targets for broadly reactive immune responses (9). Rough mutants, which expose conserved inner-core determinants, have been extensively investigated in this regard; however, contradictory findings have emerged (1). While some studies showed that immunization with rough mutants protects experimental animals against bacterial infections and LPS-mediated lethal toxicity (18, 30, 37, 49), others failed to show similar protective effects (2, 19). In controlled human studies, antibodies directed against conserved deep-core epitopes reduced neither mortality from sepsis (7, 23) nor the incidence of postsurgical gram-negative bacteremia (11). Two major clinical trials have likewise shown that monoclonal antibodies (MAbs) directed against lipid A, the toxic center of LPS, are quite ineffective as antisepsis agents (4, 31).

Since virulent enteric bacteria normally make smooth LPS (sLPS), the accessibility of the inner-core and lipid A regions and, thus, their appropriateness as targets for immunotherapy remain major issues. The theoretical expectation that determinants in these regions are masked by both the outer core and the sterically bulky O chains is supported by a large body of work demonstrating that most antibodies directed against the inner core or lipid A do not bind sLPS (12, 13, 20, 29, 32, 41, 46). However, that some core epitopes are accessible in sLPS molecules has been demonstrated by two broadly reactive core-specific MAbs (14, 34), one of which recognizes the inner-core disaccharide l-α-d-heptose-1→7-l-α-d-heptose-1→ (34). Given the high mortality that results from septic shock, the lack of effective therapies, and the increasing number of debilitated and immunocompromised individuals, there is a critical need for cross-protective vaccines and other approaches to reduce the incidence of sepsis.

The main purpose of this study was to examine whether immunization with an enteric organism which has the complete LPS core, and thus all core epitopes, would generate cross-reactive antibodies which bind long-chain sLPS molecules of different serospecificities and, if so, to map the epitopes against which such antibodies are directed. This strategy was based on two surmises: that epitopes in the complete core resemble their native conformations in sLPS and elicit antibody responses according to the extents of their accessibility in this moiety and that the major cross-reactive epitopes can be identified by reactivity inhibition patterns generated by use of a series of truncated-core chemotypes as ligands. The Salmonella type Ra complete core was chosen for investigation because it is the only complete core for which a complete set of truncated forms is available. Two other strains were also investigated. One of these was a smooth strain in order to gain insight on the steric influence of the O antigen on the development of cross-reactive anti-LPS responses. The other was an Re strain that is of comparative interest as a core mutant and that has been extensively tested as a cross-protective immunogen but whose cross-reactive properties remains quite controversial.

MATERIALS AND METHODS

Bacterial strains.

Salmonella strains (serotype or chemotype) IS2 (AO), SL3201 (BO), SL3622 (BO), SL2824 (CO), SL4388 (CO), SH1262 (DO), IS78 (EO), TV119 (Ra), SN55 (Ra), SN57 (Ra), SL733 (Rb₁), TV161 (Rb₂), TV148 (Rb₃), SL805 (Rc), SL1032 (Rd₁), SL1181 (Rd₂), SL1102 (Re), and R595 (Re) have all been described previously (34–36, 46, 47). E. coli strains of defined core types—HF4704 (R1), EH100 (R2), F653 (R3), F2513 (R4), and W3110 (K-12)—have also been described before (22, 24). Two other strains, 16 and 25, were urine isolates determined to have core type R2 on the basis of full and partial sensitivities to phage FO, respectively.

LPSs, polysaccharides, and glycoconjugates.

The procedures used for large-scale extraction of rough LPS (rLPS) and sLPS and for their subsequent purification to eliminate proteins and reduce nucleic acid contamination to <5% have been described (34, 48). Whole-cell lysates were prepared by the method of Hitchcock and Brown (21). Polysaccharides (PSs) were prepared by hydrolysis of LPSs as described previously (32). Synthetic Salmonella O-factor-specific neoglycoconjugates prepared by copolymerization of haptenic glycosides and acrylamide were kindly provided by Anatoly Chernyak (10). These are PM-PAA (O:2 specific), MRG-PAA (O:3 specific), AM-PAA (O:4 specific), and TM-PAA (O:9 specific). Another glycoconjugate, CO-bovine serum albumin (BSA) (O:7 specific), comprised a dodecasaccharide from Salmonella O:6,7 PS covalently coupled to BSA (15).

Immunogens.

Strains kept at −80°C were streaked out directly on nutrient agar plates. Following overnight incubation, several colonies were typed for phenotypic characteristics by serological, phage-sensitivity, and bile-sensitivity tests. Appropriate colonies were pooled, resuspended in saline, and spread on several agar plates. After overnight incubation, the resulting confluent layers were scraped off with cotton swabs, resuspended in 5 ml of saline, and washed twice by centrifugation (10,000 × g, 10 min). Final suspensions were plated for viable counts and kept overnight at 4°C; they were then adjusted to contain ca. 2 × 10¹⁰ CFU/ml and heated (80°C, 3 h) to kill the bacteria. Aliquots (50 μl) of heat-treated suspensions were transferred into 5 ml of fresh nutrient broth and incubated overnight to check sterility.

Antisera.

TO mice and New Zealand White rabbits were bred in the facilities at the Desert and Marine Environment Research Center, United Arab Emirates University. Preimmune sera were obtained from these animals 1 week before immunization began. The murine antisera anti-Ra/1, anti-Ra/2, anti-S/1, anti-S/2, anti-S/3, anti-Re/1, and anti-Re/2 were prepared by intraperitoneal inoculation of groups of 10 mice. Each group was inoculated with one of the heat-killed strains SN57, SN55, SL4388, SL3201, SL2824, SL1102, and R595, which are the respective sources of the above-listed antisera. The regimen comprised six weekly injections starting with 10⁸ bacterial bodies (BB) and doubling each subsequent time, so that the final dose was ca. 3 × 10⁹ BB. Mice were bled 4 days after the fourth, fifth, and sixth injections; sera from each group were pooled and labeled as described above. Anti-Ra/4 was prepared by immunizing three female rabbits with an initial course of four progressively doubling doses of heat-killed SN57 (2 × 10⁸ to 2 × 10⁹ BB) at 5-day intervals via the marginal ear vein. After a 3-week rest, the rabbits were given another course of three weekly injections (2 × 10⁹ BB) and bled 5 days after the last dose. The Salmonella polyvalent O antiserum was purchased from Murex Diagnostics, Dartford, United Kingdom.

Determination of RR by ELISA.

Maxisorp enzyme-linked immunosorbent assay (ELISA) plates (NUNC, Roskilde, Denmark) were coated with glycoconjugates (1 μg/ml) in 0.05 M carbonate buffer (pH 9.6) or with LPS by chloroform-ethanol evaporation (16). The plates were blocked (1 h at 37°C with a solution containing 0.5% BSA and 0.025% gelatin in 0.05 M carbonate buffer, pH 9.6) and washed three times (0.15 M NaCl, 0.05% Tween 20). ELISA was then continued as described before (34) with either peroxidase-labeled rabbit anti-mouse polyvalent immunoglobulins (Dakopattis, Glostrup, Denmark) or goat anti-rabbit immunoglobulin G (Sigma, St. Louis, Mo.) as the conjugate and o-phenylenediamine HCl as the substrate. The end point titer (EPT) is the dilution of serum that gives an A₄₉₀ of 0.2. The relative reactivity (RR) of each serum with each antigen was calculated as follows. Data from each assay were plotted on double logarithmic axes; then a region encompassing three successive dilutions where titration curves were linear or approximately so was delineated by visual inspection. The geometric mean absorbance (Gm_ab) for the three dilutions in this region was determined for each antigen. The Gm_ab for the homologous LPS was assigned an RR value of 100%. Those for other antigens were calculated by expressing their Gm_abs as percentages of the homologous Gm_ab.

ELISAs with inhibition.

ELISAs with inhibition were done to test the effectiveness of ligands as inhibitors. The protocol for these assays has been described previously (34). The 50% inhibitory concentration (IC₅₀) is the concentration of inhibitor needed to obtain a 50% lowering of the optical density at 490 nm compared to that of control wells with no inhibitor added.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting.

LPS samples (sLPS [7.5 μg] or rLPS [2.5 μg]) or proteinase K-treated lysates were resolved in 15% polyacrylamide gels by incorporating SDS and urea as described before (44). They were electrophoretically transferred (120 mA, 12 h) to nitrocellulose membranes and tested for reactivity with sera as described before (34) with the same conjugate as that used in the ELISA but with diaminobenzidine-H₂O₂ (Sigma) as the substrate.

Stereoplots of core oligosaccharides.

Stereoplots of core oligosaccharides in their minimum energy conformations as predicted by the hard sphere exo anomeric method (48) were provided by Per-Erik Jansson of the Clinical Research Center, Karolinska Institute, Stockholm, Sweden.

RESULTS

Reactivity patterns of murine sera with LPS.

On initial titration, immune sera were found to have high end point titers against their homologous LPSs (72,000 to 218,000) while preimmune sera were found to be nonreactive (EPT, <300). The preimmune sera remained nonreactive when they were subsequently titrated against sLPSs and rLPSs of various serogroups (A, B, C₁, D, and E) and chemotypes (Ra to Re), while immune sera reacted to various degrees. The reactivities of both anti-Ra/1 and anti-S/1 with rLPS progressively decreased as the core oligosaccharide was increasingly truncated (Fig. 2A). However, anti-Ra/1 was more reactive with each core chemotype than anti-S/1, indicating that the presence of O chains diminished the immune response to core determinants. Unlike the reactivities of both anti-S/1 and anti-Ra/1, the reactivity of anti-Re/1 with rLPS increased, with some exceptions, as the core oligosaccharide was progressively truncated. Thus, antibodies in both anti-Ra/1 and anti-S/1 were more reactive with outer- than inner-core epitopes while those in anti-Re/1 were the opposite. The degrees to which the sera reacted with sLPS (Fig. 2B) may be summarized as follows: anti-Ra/1 (23 to 45%) > anti-Re/1 (12 to 30%) > anti-S/1 (<10%), with anti-S/1 reacting poorly with all but its homologous sLPS. Another set of anti-LPS sera (anti-Ra/2, anti-S/2, and anti-Re/2) was similarly characterized with results practically identical to those described above (data not shown). Thus, vaccination of mice with S or Ra strains elicited stronger antibody response against outer- than against inner-core epitopes while vaccination with Re strains did the opposite. Despite this dichotomy, however, the reactivities of all three sera with sLPS (anti-Ra > anti-Re > anti-S) reflected their RRs with outer-core chemotypes (anti-Ra > anti-Re > anti-S) and contrasted with those of inner-core chemotypes (anti-Re > anti-Ra > anti-S). These results suggested that the cross-reactivities of all three sera with sLPS were mediated mainly by antibodies which bind outer-core determinants.

FIG. 2.

Reactivities of anti-S/1 (□), anti-Ra/1 (■), and anti-Re/1 (▨) with rLPSs (A) and sLPSs (B) of different chemotypes and serospecificities.

Immunoblotting of murine sera against electrophoretically resolved sLPS.

After resolution by SDS-PAGE, sLPSs displayed molecular heterogeneity visualized as ladder-like patterns of bands (Fig. 3A). The fastest band in each lane represents the complete core, while each subsequent band of higher molecular weight (M_r) represents the core plus increasing numbers of O repeat units. When immunoblotted against these LPSs, both anti-Ra/1 (Fig. 3B) and anti-Ra/2 (not shown) bound the fastest band in each lane as well as bands of higher M_rs to generate ladder-like patterns. By contrast, anti-S/1 (Fig. 3C), anti-S/2, and anti-S/3 (data not shown) did not generate ladder-like patterns with nonhomologous sLPS. Anti-Re/1 (Fig. 3D) and anti-Re/2 (data not shown) likewise reacted only with Re LPS and failed to recognize even the complete-core moiety. When similarly immunoblotted, preimmune sera failed to recognize any bands (data not shown). Thus, core-specific cross-reactive antibodies which bind sLPS molecules are normally generated by immunization of mice with Ra but not S or Re strains.

FIG. 3.

Analysis of cross-reactivities of murine antisera by immunoblotting against electrophoretically resolved LPSs. Gels were silver stained (A) or blotted against anti-Ra/1 (B), anti-S/1 (C), or anti-Re/1 (D).

Reactivity of anti-Ra with serogroup-specific glycoconjugates.

Since LPSs are potent B-cell mitogens, it was possible that the cross-reactivity of anti-Ra/1 with sLPS was mediated by O-specific antibodies generated by polyclonal B-cell activation rather than by antibodies directed against core determinants. The relative involvement of anti-O antigen versus anti-core antibodies in the cross-reactivity of anti-Ra/1 was therefore tested by comparison of its reactivity with sLPSs in ELISA with those of the corresponding synthetic O-antigen-specific neoglycoconjugates. While both the glycoconjugates and the sLPSs reacted equally well with a commercial polyvalent O-antigen-specific antiserum (Fig. 4A), the glycoconjugates, unlike the sLPSs, were poorly reactive with anti-Ra/1 (Fig. 4B). Thus, anti-Ra lacked antibodies recognizing epitopes in O antigen and could have reacted with sLPS only via the binding of antibodies to determinants in the core or possibly in lipid A. Compared to anti-Ra, anti-Re (Fig. 4C) as well as anti-S (Fig. 4D) reacted more strongly with the glycoconjugates while being less reactive with the corresponding sLPS. The cross-reactivities of anti-Re and anti-S with nonhomologous sLPSs thus seems to involve significant binding of antibodies to O antigen.

FIG. 4.

Titration of polyvalent Salmonella O antiserum (A), anti-Ra/1 (B), anti-Re/1 (C), and anti-S/1 (D) against sLPS and O-specific glycoconjugates. Symbols indicate results with LPS and the cognate glycoconjugates as follows: B-LPS (○) and AM-PAA (•), A-LPS (▿) and PM-PAA (▾), and D-LPS (□) and TM-PAA (■). Data from experiments with C₁-LPS–CO-BSA and E-LPS–MRG-PAA (not shown) are consistent with those presented for the above LPS and glycoconjugate pairs.

Cross-reactivity of rabbit anti-Ra serum (anti-Ra/4).

The reactivity patterns of anti-Ra/4 with rLPS, sLPS, and neoglycoconjugates in ELISA were practically identical to those of anti-Ra/1 (data not shown). Anti-Ra/4 also generated ladder-like patterns when it was immunoblotted against sLPSs resolved by SDS-PAGE (Fig. 5A). Therefore, like its murine counterparts, anti-Ra/4 cross-reacted with sLPSs via the binding of antibodies to core epitopes. Pooled preimmune serum from the same rabbits was nonreactive with LPS both in ELISA (EPT, <300) and in immunoblots (data not shown). To gain insight into the ability of anti-Ra/4 to recognize other core types, it was blotted against electrophoretically resolved E. coli LPSs of various O specificities. The serum was nonreactive with O:26, O:28, O:90, O:111, and O:143 LPSs (data not shown) but revealed ladder-like patterns in proteinase K-treated lysates from two phage FO-sensitive clinical isolates (Fig. 5B). Therefore, the cross-reactive antibodies in anti-Ra also recognized determinants in the R2 core even in sLPSs from E. coli.

FIG. 5.

Immunoblotting of anti-Ra/4 against Salmonella and E. coli LPSs.

Molecular specificities of cross-reactive antibodies.

The ability of core chemotypes to inhibit the cross-reactivities of anti-Ra with sLPSs was investigated in order to delineate the epitopes involved. Initial studies showed that both Ra and Rb₁ PSs inhibited the reactivity of anti-Ra/1 with five different Salmonella sLPSs (serogroups A to E), with Ra PS (IC₅₀, 1 to 11 μg/ml) being much more effective than Rb₁ PS (IC₅₀, ca. 8 to 200 μg/ml). By contrast, Rb₂ PS did not inhibit (IC₅₀, >500 μg/ml). With only serogroup B LPS for subsequent more detailed studies, it was found that Ra, Rb₁, and SL5007 LPSs (or PSs) all effectively inhibited the reactions of both anti-Ra/1 and anti-Ra/4 with sLPSs (Fig. 6 and Table 1), while all other core chemotypes did not. Furthermore, while Ra LPS and PS were of nearly equal efficacies as inhibitors of anti-Ra/1 (IC₅₀, ca. 1 to 4 μg/ml), the former (IC₅₀, ca. 0.03 μg/ml) was at least 1,000-fold more effective than the latter (IC₅₀, ca. 38 μg/ml) in inhibiting anti-Ra/4. This result suggests that the cross-reactive antibodies in anti-Ra/4 were much more avid than those in anti-Ra/1 and that they require multivalent presentation of epitopes by native micellar LPS in solution for their inhibition. Since SL5007 LPS was an effective inhibitor, it may be deduced that the α-Gal-1→6 branch (Fig. 1) it lacks is not essential for the binding of any major cross-reactive antibody species. The monosaccharides GlcNAc, Glc, and Gal were also tested and found to be ineffective as inhibitors (data not shown). These findings are consistent with a deduction that the cross-reactive antibodies in both anti-Ra/1 and anti-Ra/4 were mainly directed against the GlcNAc→Glc disaccharide of the outer-core domain.

FIG. 6.

Inhibition of the reactivities of anti-Ra/1 (A) and anti-Ra/4 (B) with serogroup B sLPS. Symbols represent results with Ra LPS (■), Ra PS (□), Rb1 LPS (•), Rb1 PS (○), Rd₂ LPS (▵), Re LPS (×), and SL5007 LPS (⊞).

TABLE 1.

Inhibition of the reactivity of anti-Ra/1 with BO sLPS

Chemotype(s) of inhibitors	IC50 (μg/ml)
	LPS	PS
Ra	1	4
Rb1	90	50
Rb2, Rb3, Rc	>500	>500
Rd1, Rd2, Re	>500	NDa

^aND, not determined. 

Analyses of separate stereoplots of the binding region.

Molecular models of the Ra core capped with two O repeat units as well as those of five other complete cores were analyzed in order to visualize the conformational bases for the binding of cross-reactive antibodies to GlcNAc→Glc and to determine what these imply for other core types. The models designated smooth-1 and smooth-2 (Fig. 7) depict the capped Ra core with the O repeat units in two possible orientations. In both these orientations the terminal, α-1,2-linked GlcNAc of the Ra core clearly protrudes from the oligosaccharide backbone and evidently is readily accessible to cross-reactive antibodies. Circles highlight the corresponding terminal regions of the R2, R3, R1, and K-12 cores. While the terminal sugar residues differ among these cores, they are all α-1,2 linked to the penultimate sugar residue, which also provides the capping site for O antigen. Because of this similarity of linkages, the terminal regions of the R1, R2, and R3 cores display conformational motifs nearly identical to the binding site of cross-reactive antibodies in the Ra core. These motifs should, therefore, be similarly accessible to antibodies in sLPS though not necessarily mutually cross-reactive. This analysis extends to the R4 core (not shown), which also terminates in α-hexose-1,2→α-hexose, but not necessarily to the K-12 core. Two versions of the K-12 core oligosaccharide, one of which terminates in l-α-d-heptose-1→-α-d-Glc (22) and the other of which terminates in β-GlcNAc-1,2→α-Glc, have been described (24). The stereoplot of the latter version (Fig. 6) depicts an outer-core region clearly different in shape from those of other cores of Enterobacteriaceae.

FIG. 7.

Molecular models of different core types depicting the Ra core capped with two O repeat units, which are in two different orientations (Smooth-1 and Smooth-2). Circles indicate the terminal regions of complete cores which appose the O antigen in sLPS. I, terminal sugar residue; II, penultimate sugar residue.

DISCUSSION

The need for preventive approaches to the sepsis syndrome is highlighted by recent failures of several prospective antisepsis therapies in clinical trials (1, 38, 42). That LPS is an appropriate target for antisepsis prophylaxis is indicated by studies which have correlated natural levels of antiendotoxin core antibodies with reduced incidence of postsurgical complications and better outcomes during illness (3, 17). The development of vaccines to engender or augment natural levels of appropriate antiendotoxin core antibodies, therefore, seems a reasonable preventive approach to gram-negative sepsis. Ideally, such vaccines should comprise epitopes which are not only shared by various endotoxins but are also accessible to the immune system even in the presence of the bulky, hypervariable O antigen.

One of the major problems associated with the concept that antiendotoxin antibodies are protective in sepsis has been the failure of most previous studies to demonstrate that antibodies directed against deep-core epitopes bind sLPS molecules. Many laboratories have characterized anti-LPS MAbs with a view to identifying cross-reactive epitopes in the core domain. In this regard Nnalue et al. have reported that the disaccharide l-α-d-heptose-1→7-l-α-d-heptose-1→ of the inner core is accessible in sLPS and mediates broad reactivity between Salmonella and E. coli strains (34). Another MAb apparently recognizing a common, but as yet unidentified, epitope in LPS preparations from members of the family Enterobacteriaceae has more recently been described (14). However, the question of whether LPS core structures can be used as immunogens to elicit broadly cross-reactive responses in mammalian hosts has remained unanswered.

This study has shown that immunization with Ra strains elicits antibodies that are directed against the terminal core disaccharide, GlcNAc→Glc, and that bind long-chain LPS molecules of various serospecificities. These results, some of which have appeared in abstract form (33), demonstrate for the first time the possibility of eliciting broadly cross-reactive polyclonal anti-LPS responses by immunization with rough bacterial strains. It is of interest that none of the many core-specific MAbs that have been characterized (26, 28, 34, 41, 45) display broad reactivity directed against the GlcNAc→Glc determinant. That cross-reactive anti-GlcNAc→Glc antibodies were readily detected in immune sera is evidence that the characterization of total polyclonal responses may, in some situations, have advantages over the use of MAbs for detection of cross-reactive epitopes.

The potential for smooth bacteria to elicit cross-reactive antibodies is of interest because they are often encountered by the immune system as transient invaders from the gut or as vaccines. In remarkable contrast to anti-Ra sera, several anti-S sera of high titer failed to react with heterologous sLPS molecules in immunoblots. This result indicates that vaccination with smooth organisms or natural contact with them would probably not elicit strongly cross-reactive anti-LPS responses, in agreement with the finding that most patients who have had a bacteremic episode with smooth strains lack anticore antibodies (12).

One aim of this work was to compare the cross-reactive responses elicited by determinants in Ra and Re LPSs. The finding that anti-Re sera lacked cross-reactive antibodies agrees with the findings of several previous reports. In a study of the binding specificities of five Re-specific MAbs, it was found that the accessibility of their epitopes were markedly reduced following the attachment of a single l-glycero-d-manno-heptose residue to the terminal 2-keto-3-deoxyoctonate of the inner-core domain (27). Another study reported that repeated immunization of rabbits with Re strains generated only modest levels of cross-reactive antibodies (25); these antibodies were apparently nonspecific because they were reduced or eliminated by absorption with heterologous but not by Re LPS. Other investigations have likewise failed to demonstrate the presence of cross-reactive antibodies in sera generated by immunization with Re strains (32). However, reports that active or passive immunization with Re strains protect against experimental gram-negative infections (13) have sustained interest in the inner-core domain as a cross-protective immunogen.

Because LPSs from members of the Enterobacteriaceae have a common architecture, the findings of this study may have broad implications. For example, the outer hexose regions of core oligosaccharides from Salmonella and E. coli are known to comprise a common sequence of residues of the general structure α-hexose-1→2-α-hexose-1→2-α-hexose-1→3-α-Glc-1→3α (24, 43). Semiempirical calculations of the minimum energy conformations of these oligosaccharides have shown that they exhibit two sides: a front side of similar overall shapes in all five cores (Ra, R1, R2, R3, and R4) and a back side from which different groups protrude to determine core type specificity (6). The conformational similarities between these cores are evidently dictated by similar linkages, and they in turn dictate similarity of function, such as the common ability of these regions to satisfy the binding requirements of certain core-specific phages. As depicted in stereoplots, the terminal residues in all five core types protrude from the main axis in similar orientations and, therefore, should demonstrate similar degrees of accessibility to antibody in the complete O-antigen-substituted LPS molecule. It appears that, regardless of O-antigen or core-type specificity, antibodies directed against the terminal disaccharides in all core types which share the α-hexose-1→2-α-hexose-1→2-α-hexose-1→3-α-Glc-1→3-α backbone (represented by α-d-Gal-1→2-α-d-Gal [R1], α-d-GlcNAc-1→2-α-d-Gal [R2], α-d-Glc-1→2-α-d-Gal [R3], and α-d-Gal-1→2-α-d-Glc [R4]) cross-react with appropriate sLPSs. The cross-reactivity of anti-Ra with long-chain molecules in E. coli sLPSs which have core type R2 would support this hypothesis.

Some perspective on cross-reactive core epitopes may be gained from comparison of the binding characteristics of core-specific antibodies with those of bacteriophages. The inability of smooth strains to adsorb rough-strain-specific phages such as BR2, BR60, 6SR, Ffm, and the φX-like phages, all of which act on Ra strains, demonstrates the effectiveness of the O antigen as a steric barrier to core determinants. It is noteworthy, therefore, that FO, the only rough-strain-specific phage to attack smooth strains, recognizes a receptor involving the same terminal disaccharide recognized by the cross-reactive antibodies in anti-Ra. A consideration of the binding characteristics of G13, a φX-like phage, is also informative. The binding domain for the phage is the trisaccharide moiety, α-Gal-1→2-α-Gal-1→3-α-Glc, which is terminal in Rb₂ LPS but also quite accessible to the phage in Rb₁ as well as Ra LPS. The α-Gal-1→3-α-Glc (α-hexose-1→3-α-Glc) segment of this trisaccharide is part of the conserved surface of the Ra-R4 core types. While the strong reactivities of anti-Ra sera with Rb₂ and Rb₃ LPSs (unpublished data) demonstrate the accessibility of this region to immune recognition in Ra LPS, inhibition data show its lack of participation in cross-reactivity with sLPS. Phage G13 and other φX-like phages likewise fail to bind their receptor once the core is substituted with at least one O repeat unit. Therefore, data from phage and antibody binding studies of the outer-core region fully concur as regards the accessibility and inaccessibility in sLPSs of the GlcNAc→Glc and α-Gal-1→2-α-Gal-1→3-α-Glc moieties, respectively.