Tolerance to Self Gangliosides Is the Major Factor Restricting the Antibody Response to Lipopolysaccharide Core Oligosaccharides in Campylobacter jejuni Strains Associated with Guillain-Barré Syndrome

Abstract

Guillain-Barré syndrome following Campylobacter jejuni infection is frequently associated with anti-ganglioside autoantibodies mediated by molecular mimicry with ganglioside-like oligosaccharides on bacterial lipopolysaccharide (LPS). The regulation of antibody responses to these T-cell-independent antigens is poorly understood, and only a minority of Campylobacter-infected individuals develop anti-ganglioside antibodies. This study investigates the response to gangliosides and LPS in strains of mice by using a range of immunization strategies. In normal mice following intraperitoneal immunization, antibody responses to gangliosides and LPS are low level but can be enhanced by the antigen format or coadministration of protein to recruit T-cell help. Class switching from the predominant immunoglobulin M (IgM) response to IgG3 occurs at low levels, suggesting B1-cell involvement. Systemic immunization results in poor responses. In GalNAc transferase knockout mice that lack all complex gangliosides and instead express high levels of GM3 and GD3, generation of anti-ganglioside antibodies upon immunization with either complex gangliosides or ganglioside-mimicking LPS is greatly enhanced and exhibits class switching to T-cell-dependent IgG isotypes and immunological memory, indicating that tolerance to self gangliosides is a major regulatory factor. Responses to GD3 are suppressed in knockout mice compared with wild-type mice, in which responses to GD3 are induced specifically by GD3 and as a result of polyclonal B-cell activation by LPS. The anti-ganglioside response generated in response to LPS is also dependent on the epitope density of the ganglioside mimicked and can be further manipulated by providing secondary signals via lipid A and CD40 ligation.

Guillain-Barré syndrome (GBS) is an acute, postinfectious polyneuropathy that is believed to have an autoimmune basis (21-23, 27, 29, 64, 66). It is associated with a wide range of precipitating bacterial and viral infections, including Campylobacter jejuni enteritis in 10 to 50% of cases depending on geographical region (5, 30, 46). In addition to immune responses specific to the preceding infection, 40% of post-Campylobacter infection GBS sera contain transient immunoglobulin M (IgM), IgA, and IgG antibodies to a variety of self gangliosides, including GM1, GM2, GD1a, GalNAc-GD1a, GD1b, GD3, GT1a, and GQ1b, which are believed to be among the principal pathogenic factors (7, 8, 28, 37, 65). Gangliosides are a family of sialic-acid-containing glycosphingolipids distributed throughout the body but highly enriched in the nervous system, where they are capable of acting as targets for anti-ganglioside autoantibodies (22, 34, 35, 60).

One of the mechanisms by which anti-ganglioside antibodies arise in GBS is through molecular mimicry with microbial oligosaccharides, including those borne by Campylobacter species (43, 56). Chemical and structural analysis of lipopolysaccharide (LPS) and lipooligosaccharide (LOS) outer core oligosaccharide (core OS) structures from C. jejuni serotypes isolated from GBS and non-GBS patients have identified sialylated moieties with configurations identical to those of several gangliosides (9, 11, 42, 44, 45, 50). For example, LPSs from C. jejuni HS:19, a serotype commonly associated with GBS, have been shown to contain GM1-, GD1a-, GD3-, and GT1a-like motifs, and antibody mimicry is supported by the finding that immunization of experimental animals with these LPSs produces the corresponding anti-ganglioside antibody response (4, 18). Serotyping studies have determined that certain C. jejuni serotypes, including HS:19, have greater potential for triggering GBS, and this may be due to quantitative differences in ganglioside-like LPS and LOS epitopes compared with non-GBS-associated strains (9, 44, 50).

Whereas C. jejuni is one of the commonest causes of acute diarrhea worldwide, affecting approximately 1% of the U.S. population per annum, GBS has a much lower incidence of 1.5/100,000 population, and thus it is estimated that only 0.01% of C. jejuni infections trigger GBS (2, 30). Although the absence of ganglioside mimics on some C. jejuni LPSs may be part of the explanation for this, clinical studies have demonstrated that even when humans are exposed to C. jejuni strains possessing ganglioside-like epitopes, their presence is not sufficient in itself to trigger the production of anti-ganglioside antibodies.

The host and microbial factors that determine whether any individual will mount an immune response to core OS structures that mimic self gangliosides are likely to be multifactorial. One confounding microbial factor is the presence of high levels of phase variation in C. jejuni LOS that may alter the level and nature of the mimic in any one strain (19, 36). Antibody responses to carbohydrate structures, including LPS, are T cell independent (TI) and arise early in ontogeny from B1 B cells, which produce a large pool of IgM class natural antibodies acting as an early defense against invading microorganisms (17, 41, 57). B1 B cells do not switch class to T-cell-dependent (TD) isotypes, form memory cells, or affinity mature (39).

In GBS, anti-ganglioside antibodies do switch class to the TD IgG1 and IgG3 isotypes, suggesting they may have arisen from conventional B2 cells and were able to recruit T-cell help or other accessory signals (55, 62). Whether the help comes from intermolecular cooperativity (uptake of carbohydrate-protein complexes by carbohydrate-specific B-cell receptors [BCR] and subsequent presentation of peptides to conventional T helper cells), presentation via CD1 and LPS signaling via Toll receptors, or other noncognate pathways is unknown.

A limitation of pathophysiological studies of anti-ganglioside antibody-mediated neuropathy has been the inability to generate high-titer IgG antibody responses in mice. Many studies have shown that mice immunized with gangliosides using a variety of immunization strategies generate poor antibody responses. This unresponsiveness has been attributed to poor immunogenicity, T-cell independence, and tolerance (32, 38, 49). The extent to which tolerance for self gangliosides is responsible for limiting the antibody response to core OS structures in C. jejuni has not been explored. We have previously shown that mice immunized with O:3 LPS, which does not contain a self ganglioside core OS structure, produce a vigorous antibody response to O:3 LPS compared with the poor response to self ganglioside-mimicking LPSs (18).

The red blood cell glycolipid antigens that define the ABO blood group system are also examples of carbohydrate antigens under strict tolerance control, which when disrupted can lead to severe antibody-mediated disease (61). In humans, natural anti-Gal antibodies, reactive with alpha-Gal epitopes that are absent in humans, comprise 1% of total human immunoglobulins and have a major role in mediating nonprimate xenograft rejection (14). In the alpha-1,3-galactosyltransferase knockout mouse, which lacks the ability to synthesize alpha-Gal epitopes, high levels of antibody to alpha-Gal can be induced by immunization with alpha-Gal antigen, in contrast with the absence of responses in wild-type mice (14). Similar studies of GalNAc transferase knockout (GalNAcT^−/−) mice, which contain high levels of GM3 and GD3 but lack all complex gangliosides, have demonstrated vigorous antibody responses in ganglioside-deficient mice immunized with ganglioside-protein conjugates; this also suggests that self gangliosides play an important role in inducing tolerance (38).

In this study, we have examined the roles of T-cell help, noncognate accessory signals, and tolerance in enhancing or restricting the antibody response in the mouse to ganglioside-mimicking LPS core OS structures. We have approached this by using GalNAcT knockout mice, in which one would predict that autoreactive B cells specific for complex gangliosides would not be eliminated but that the mice would be highly tolerant to LPS bearing GD3-like ganglioside mimics. In order to determine whether tolerance could be overcome by immunological manipulations, we also investigated the roles of different adjuvants and accessory stimuli, including anti-CD40 antibodies, and the role of lipid A signaling mediated through CD14 in C3H/HeN and C3H/HeJ mice.

MATERIALS AND METHODS

Animals.

Female BALB/c, C3H/HeN, and LPS-hyporesponsive C3H/HeJ mice were supplied by Harlan, UK Ltd., Bicester, United Kingdom. Mice lacking a functional GalNAcT gene (GalNAc^+/+) and wild-type homozygous GalNAc^−/− mice were generated and genotyped as described previously (59). The animals were housed and handled under University of Glasgow institutional guidelines and United Kingdom Home Office licensing. Food and water were supplied ad libitum.

C. jejuni.

C. jejuni serostrains and isolates with structurally defined core OSs that were used in this study are shown in Table 1, and the oligosaccharide structures are shown in Fig. 1. The HS:19 (formerly O:19) and HS:4 serostrains, the OH4384 isolate of the HS:19 serostrain (11), and the PG836 isolate of the HS:10 serostrain (48, 54) were provided by J. Penner and D. Woodward, Centre for Disease Control, Ottawa, Ontario, Canada (42, 50, 51). The HS:3 serostrain was provided by A. Moran, Galway, Eire (10). For clarity, HS:19 is referred to as HS:19(GM1⁺ GD1a⁺), the isolate OH4384 is referred to as HS:19(GM1⁺ GT1a⁺), and the isolate PG836 is referred to as HS:10. The bacteria were grown on blood agar plates in a microaerobic atmosphere and harvested in distilled water after 48 h of growth. The bacteria were killed by heating them at 60°C for 1 h. LPS was isolated by hot phenol-water extraction, quantitated, analyzed for purity by silver staining after sodium dodecyl sulfate-polyacrylamide gel electrophoresis and thin-layer chromatography, resuspended in distilled water, and stored at −20°C (18).

TABLE 1.

Ganglioside-like structures present on LPSs from C. jejuni strains

Ganglioside mimicked	Presencea of structure on LPS:
	HS:3	HS:4	HS:19 (GM1+ GD1a+)	HS:10d	HS:19 (GM1+ GT1a+)e
GM1	−	+b	+c	−	+
GD1a	−	+	+	−	−
GD3	−	−	−	+	−
GT1a	−	−	−	−	+

^a+, present; −, absent.

^bHS:4 LPS bears GD1a and GM1 at a ratio of 9:1 (50).

^cHS:19 LPS bears GD1a and GM1 at a ratio of 1:1 (50, 51) and is referred to here as HS:19(GM1⁺ GD1a⁺).

^dHS:10 was previously referred to as PG836 (48).

^eHS:19(GM1⁺ GT1a⁺) was previously referred to as OH4384 (11).

FIG. 1.

Oligosaccharide structures. The whole structure of GQ1b is shown, but this has not been found in C. jejuni. NeuAc, N-acetyl neuraminic acid; X, Glc (1→1) ceramide (gangliosides) or the remaining core OS-lipid A (LPSs).

Liposomes.

Extruded liposomes were made according to protocols provided by Lipex Biomembranes Inc., Vancouver, Canada (40). Cholesterol, sphingomyelin, dicetylphosphate, and ganglioside GQ1b, GD1a, GM1, or GD3, supplied by Sigma Chemicals (Poole, Dorset, United Kingdom), were dissolved in chloroform-methanol (1:1) to yield stock solutions at 10 μg/ml and used to create liposomes with a molar ratio of 5:4:1:1, respectively. In order to generate a final volume of 1 ml of liposomes (sufficient for immunization of 10 mice with 100 μl of resuspended liposomes each), the lipid mixture was dried under nitrogen and resuspended to 5 ml in phosphate-buffered saline (PBS), pH 7.4. To create ovalbumin (ova)-containing liposomes, the PBS contained ova at 5 mg/ml. The mixture was alternately vortexed (Rotamixer; Fisher Scientific, Loughborough, United Kingdom) and sonicated (Nusonics; Quayle Dental, Worthing, United Kingdom) at room temperature for up to 15 min until the dried lipid was released from the walls of a 15-ml tube and the suspension was uniformly milky. The suspension was frozen and thawed five times by plunging it alternately into liquid nitrogen and a 37°C water bath to create multilamellar vesicles. This mixture was clarified by centrifugation at 600 × g (2,000 rpm in a Beckman GS6R benchtop centrifuge) for 15 min, and the pellet was discarded. Unilammelar liposomes were then created by repeated extrusion (10 times) through an N₂-driven extruder (Lipex Biomembranes Inc.) using a double 0.4-μm-pore-size polycarbonate fiber filter (no. 110605; Nuclepore). The extruded liposomes were then ultracentrifuged at 110,000 × g (35,000 rpm; Beckman LS50B centrifuge) for 1 h and resuspended in 1 ml of PBS for immunization in a final volume of 100 μl/mouse. The final product is referred to according to its constituents, e.g., ova-GD1a liposomes. In some immunization experiments in which liposomes were compared with simpler lipid preparations, gangliosides were dried on glass vessels and then resuspended by being vortexed and sonicated for 15 to 30 min in PBS containing 5 mg of ova/ml. In the series of experiments with anti-CD40 antibodies, LPS liposomes, without any protein adjuvant, were created by sonication of cholesterol, phosphatidylcholine, and HS:4 or HS:19(GM1⁺ GT1a⁺) in a ratio of 5:4:1 in liposomes, which were then admixed by gentle vortexing with 0.5 mg of monoclonal anti-CD40 antibody or the control MAb, GL117 (25)/mouse.

Immunizations.

The mice were first immunized at 6 to 8 weeks of age via the footpad (FP), subcutaneous (s.c.), or intraperitoneal (i.p.) route and subsequently on up to three further occasions at 2- to 4-week intervals. Immunizations comprised final volumes of 100 μl per mouse. The animals were studied in groups of three to eight either in single or in duplicated experiments. In experiments using protein adjuvant, the mice were primed with ova (60 μg/mouse) in alum (Alhydrogel 2%; Superfos Biosector, Frederikssund, Denmark) or in complete Freund's adjuvant (CFA; Sigma Chemicals). For ova-liposome immunizations, the mice were initially primed with ova-alum and then immunized 7 to 10 days later via the i.p. route with liposomes containing 100 μg of ganglioside per mouse. Subsequent immunizations were administered at 2-week intervals. Mice immunized with LPS were given 100 μg of LPS per mouse in a 100-μl final volume of CFA (1:1, admixed by vortexing to emulsion for 30 min). For subsequent immunizations, LPS was delivered in incomplete Freund's adjuvant (IFA) at 2- to 3-week intervals. Serial blood samples were collected at multiple time points via the tail vein and stored at −20°C for measuring antibody responses. Prebleeds were performed in selected experiments, and significant antibody responses to gangliosides were never seen in these samples. All immunization and venipuncture procedures conformed to United Kingdom Home Office and University of Glasgow institutional guidelines.

Detection of antibodies to gangliosides, LPS, and ova.

All gangliosides (GM1, GD1a, GD3, and GQ1b) supplied by Sigma Chemicals have an estimated purity of >95%. GT1a is a rare ganglioside species and is not available in sufficient quantities in purified form for these studies; thus, sera were screened against GQ1b, because it has a structure almost identical to that of GT1a, with which it invariably has immunological cross-reactivity (Fig. 1). Mouse sera were tested for IgG and IgM responses to gangliosides by enzyme-linked immunosorbent assay (ELISA) as previously described (63). Immunolon 2 microtiter plates (Dynatech, Chantilly, Va.) were coated with 200 ng of ganglioside per well in methanol. Each sample was also screened against a methanol-treated, ganglioside-free control well from which background OD readings were obtained. Some serum samples were assayed to determine the isotype (subclass and light-chain type) of the anti-ganglioside antibody response using anti-mouse IgG1, -2a, -2b, and -3 and kappa- and lambda-specific antibodies (diluted 1/3,000). All secondary antibodies were supplied by Sigma Chemicals or by Southern Biotechnology Associates Inc., Birmingham, Ala. The ELISA results for ganglioside-free control wells are shown alongside those for ganglioside-coated wells in all figures in order to demonstrate the nonspecific background signals seen in this glycolipid ELISA, particularly for IgM class antibodies. These nonspecific signals are believed to be increased as a result of polyclonal activation of predominantly IgM-secreting B cells producing natural antibodies that often yield high background levels. Such responses were seen particularly with LPS immunizations. Background OD levels are especially prominent in these assays because of the absence of detergent in the washing and blocking buffers: glycolipids bind polystyrene poorly and are further stripped from ELISA wells by detergents, rendering the assay uselessly insensitive. Antibodies to LPS were detected by coating Immunolon 2 plates with 10 μg of LPS/μl (100 μl per well) in PBS (pH 7.4) overnight at 4°C and then discarding the unbound supernatant. Subsequent steps were performed as for anti-ganglioside antibodies. Antibodies to ova were detected by coating ELISA wells with 100 μl of 10-μg/ml ova per well in bicarbonate coating buffer, pH 9.6, and the same procedure described above was followed except for the addition of 0.05% Tween 20 to the PBS washing buffer.

Statistical analysis.

Antibody responses in groups of animals were analyzed for significant intragroup differences using Student's t test.

RESULTS

Immunization of BALB/c mice with ganglioside liposomes with and without adjuvants through FP, s.c., and i.p. routes.

In order to determine the level of immunological tolerance to gangliosides in mice with a normal complement of self gangliosides, groups of BALB/c mice (n = 3) were immunized with the self ganglioside GM1 or GQ1b incorporated into liposomes via the FP, s.c., or i.p. route using a range of immunization strategies. The effect of providing either a TH2- or TH1-type helper environment with ova-alum or ova-CFA priming and subsequent delivery of ova encapsulated in liposomes was examined. Ova-liposomes were also compared with delivery of gangliosides and ova simply admixed by vortexing.

Priming with ova-alum and subsequent immunization with GM1-ova liposomes via the i.p. route yielded significantly elevated anti-GM1 IgM titers compared with undetectable titers following the same regime via FP immunization (mean OD, 0.32 compared with 0.02; P = 0.004; three mice in each group; data from bleeds collected 10 days after the second immunization). The anti-GM1 IgM titers obtained via the i.p. route appeared after the first immunization, decayed rapidly, and reappeared at a similar titer on subsequent immunizations (data not shown). These data indicated that the peritoneal compartment, which principally contains B1 cells, could be more readily stimulated to produce anti-GM1 IgM antibody-secreting B cells than the systemic B2 compartment, which would be stimulated by FP immunization. Sequential immunizations, via either the FP or i.p. route, with GM1 liposomes in the absence of any ova priming or subsequent ova administration yielded mean OD values for IgG and IgM antibodies to GM1 at or close to zero. These data indicate that exposure of any naturally occurring GM1-specific B cells to GM1 liposomes alone, in either the conventional B2- or i.p. B-cell compartments, is not in itself a sufficient stimulus to induce secretion of anti-GM1 IgM antibodies.

In order to examine the class-switching characteristics of ganglioside-ova liposome administration in the peritoneal compartment, groups of mice were primed with ova-alum and then immunized with GQ1b-ova liposomes via the i.p. route. As described above for GM1-ova liposomes and anti-GM1 antibodies, immunization via the i.p. route with GQ1b-ova liposomes induced anti-GQ1b IgM antibodies. Class switching to IgG clearly occurred, albeit at a low level, after the second and third immunizations in a pattern suggestive of a secondary immune response (Fig. 2). No response to GM1 was seen, indicating the specificity of the response for GQ1b. Subclass analysis of the IgG response demonstrated all the anti-GQ1b IgG activity to be IgG3, the IgG subclass most usually seen when B1 cells switch class. These data indicate that i.p. immunization with ganglioside-ova liposomes stimulates B1 B cells to secrete anti-ganglioside antibodies that have the capacity to switch to IgG3.

FIG. 2.

Anti-ganglioside IgM (A) and IgG (B) responses in BALB/c mice (n = 3) following i.p. immunization with GQ1b-ova liposomes. The error bars indicate SD. The mice were primed with ova-alum and then immunized on day 7 with GQ1b-ova liposomes and bled 4 and 14 days later. The mice were reimmunized 2 and 4 weeks following the initial immunization and rebled 4 and 11 days later. A diminishing IgM response and rising IgG response to GQ1b can be clearly seen upon repeated immunization. The IgG subclass of the anti-GQ1b response was exclusively IgG3 (mean OD at 1:200 serum dilution, 0.61 ± 0.16 SD).

In control experiments, anti-ova IgM and IgG antibodies were highly elevated in all groups of mice primed via either the FP or i.p. route with ova-alum or ova-CFA and subsequently boosted with ganglioside-ova liposomes. The mice showed a primary and then a secondary immune response to ova, with diminution of IgM titers and increase of IgG titers over time, indicating that ova is being efficiently delivered and processed (data not shown).

In order to establish the effectiveness of extruded liposomes compared with simpler lipid preparations, immunizations of groups of mice (n = 3) with extruded GM1-ova liposomes or GM1 vortexed with ova in PBS were compared. The former was efficient at inducing anti-GM1 antibodies, whereas the latter failed to induce any anti-GM1 antibody responses (data not shown). This indicates that the preparation of uniform extruded liposomes with encapsulated ova is a necessary part of the immunization protocol. In experiments comparing the effects of CFA and alum as priming agents in conjunction with ova, no significant differences were observed in total IgM or IgG anti-ganglioside antibody titers when immunizations were performed via either the s.c. or i.p. route (data not shown).

These data indicate that in normal mice, the B-cell pool that can be expanded to secrete antibodies to self gangliosides requires an appropriate antigen configuration (in this case extruded liposomes) in the presence of accessory signals (provided by ova as an adjuvant) and is largely limited to an IgM and IgG3 response in the peritoneal compartment, suggestive of involvement principally of B1 B cells. The conventional B2 B-cell pool, stimulated by s.c. or FP immunization, is highly tolerant of self gangliosides.

Immunization of GalNAcT^−/−.and GalNAcT^+/+ mice with ganglioside liposomes.

In order to study the effect of tolerance to self gangliosides on antibody responses, mice with or without complex gangliosides were immunized with GD1a, GQ1b, or GD3. GalNAcT^−/− mice possess only GM3 and GD3, whereas wild-type GalNAcT^+/+ mice possess a normal complement of simple and complex gangliosides. Groups of GalNAcT^−/− or GalNAcT^+/+ mice (three to five mice per group) were first primed i.p. with ova-alum and then immunized i.p. with GQ1b-ova or GD1a-ova liposomes. All immunized GalNAcT^−/− mice developed highly elevated anti-GQ1b or anti-GD1a antibody responses that switched to IgG after the second and third immunizations, consistent with secondary immune responses (Fig. 3). As observed for BALB/c mice, GQ1b-ova liposomes also induced very modest IgM and IgG antibody responses in wild-type GalNAcT^+/+ mice; however, they were especially small in the IgG class compared to IgG responses in GalNAcT^−/− mice. The IgG subclass response in GQ1b-ova liposome-immunized GalNAcT^−/− mice was restricted to IgG3 and IgG1, suggesting possible involvement of peritoneal B1 B cells to account for the IgG3 and a TH2-driven B2 response accounting for the IgG1 that would be expected following priming with the TH2-type adjuvant, ova-alum (data not shown).

FIG. 3.

Anti-ganglioside antibody responses in GalNAcT^−/− and GalNAcT^+/+ mice following i.p. immunization with GD1a-ova liposomes (A) and GQ1b-ova liposomes (B). The error bars indicate SD. Group sizes were as follows: GD1a-ova liposomes, n = 8 (GalNAcT^−/−) and n = 5 (GalNAcT^+/+); GQ1b-ova liposomes, n = 5 (GalNAcT^−/−) and n = 3 (GalNAcT^+/+). The mice were primed with ova-alum and immunized on day 7 with GQ1b-ova or GD1a-ova liposomes. The mice were further immunized at 2-week intervals and bled 4 days after each immunization. Elevated IgG anti-GD1a antibody titers are evident in GalNAcT^−/− mice compared with GalNAcT^+/+ mice, and they increase upon subsequent immunization (*, P < 0.005 after the second immunization). Elevated IgG anti-GQ1b antibody titers are evident in GalNAcT^−/− mice compared with GalNAcT^+/+ mice (**, P < 0.001 after the second immunization).

Despite the development of a high anti-GQ1b IgG response in the GalNAcT^−/− mice, the IgM response to GQ1b in the GalNAcT^−/− mice was insignificantly elevated compared with either the anti-GD1a IgM response in GalNAcT^−/− mice or the anti-GQ1b IgM response in wild-type mice (Fig. 3). Furthermore, GalNAcT^+/+ mice immunized with GQ1b also developed a greater anti-GD3 response than GalNAcT^−/− mice. GQ1b shares a disialosyl epitope with GD3 (Fig. 1), sited on the terminal galactose of the ganglioside core. IgM antibodies frequently react with this epitope, irrespective of any adjacent oligosaccharide structure, and many human and mouse IgM MAbs cross-react with GQ1b, GD3, and other disialylated structures, including bacterial LPS. These data indicate that the anti-disialosyl IgM response, which most likely arises from the i.p. B1 pool, is suppressed in GalNAcT^−/− mice (which contain large amounts of GD3 compared with the wild type) but that the matured IgG anti-GQ1b response, which is specific for a unique GQ1b epitope(s) not shared by GD3, is suppressed in the wild type (which contains GQ1b) but not suppressed in the GQ1b-deficient, GD3-rich GalNAcT^−/− mouse.

In order to directly address the antibody response to GD3, we immunized GalNAcT^−/− (GD3-high) and GalNAcT^+/+ (GD3-normal) mice (n = 5) with GD3-ova liposomes and found the GalNAcT^−/− mice to be completely unresponsive to GD3. In comparison, anti-GD3 responses were seen in four of five wild-type mice in the IgM class, which switched in three of four mice to produce low levels of anti-GD3 IgG (Fig. 4). These data indicate that normal mice possess a pool of GD3-reactive B cells (most likely i.p. B1 cells) that can be expanded by i.p. immunization with GD3 to secrete predominantly IgM antibodies and can be suppressed by high levels of endogenous GD3.

FIG. 4.

Anti-ganglioside antibody responses in GalNAcT^+/+ (A) and GalNAcT^−/− (B) mice (two groups; n = 5) following repeated i.p. immunizations with GD3-ova liposomes. The error bars indicate SD. The mice were primed with ova-alum, immunized on day 7 with GD3-ova liposomes, and bled 4 days later. The mice were reimmunized at 2-week intervals and bled 4 days after each immunization. Anti-GD3 IgM and IgG antibody responses are depressed in the GD3-rich GalNAcT^−/− mice compared with GalNAcT^+/+ mice, but this did not reach statistical significance (P = 0.08 post-third immunization for IgM, and P = 0.3 for IgG).

Immunization of GalNAcT^−/− and GalNAcT^+/+ mice with C. jejuni LPSs bearing ganglioside-like core OS structures.

Our previous studies have shown that immunization of mice with C. jejuni LPSs containing ganglioside-mimicking core OS structures, using a variety of immunization protocols and mouse strains, leads to the generation of anti-ganglioside antibodies. However, the responses are generally poor, highly variable between animals, and principally IgM class. In order to determine the degree to which tolerance to self gangliosides might be limiting this response, we immunized GalNAcT^−/− and GalNAcT^+/+ mice (three to five mice per group) with HS:4 and HS:19(GM1⁺ GT1a⁺) LPSs emulsified in CFA (a TH1-driving adjuvant) followed by serial immunizations with LPS in IFA (Fig. 5). LPS from HS:4, which predominantly bears a GD1a-like epitope, induced high levels of IgG class antibody to GD1a in GalNAcT^−/− mice by the third immunization, in contrast with no response in wild-type GD1a-containing GalNAcT^+/+ mice and no response to the structurally dissimilar ganglioside GM1 (Fig. 5A). In the IgM class, it is clearly evident that nonspecific background signals are greatly increased upon immunization with LPS in both GalNAcT^−/− and GalNAcT^+/+ mice, as manifested by equivalently high OD readings from non-ganglioside-coated blank wells and from the ganglioside-coated wells; this is interpreted as due to nonspecific polyclonal activation of B1 cells and underlines the importance of including these control data. LPS from HS:19(GM1⁺ GT1a⁺)-induced high levels of IgG class antibody to both GQ1b and GM1 in GalNAcT^−/− mice (which lack GT1a, GQ1b, and GM1) compared with those in wild-type GalNAcT^+/+ mice (Fig. 5B). Sera were screened against GQ1b because it has a structure almost identical to that of GT1a, with which it invariably has immunological cross-reactivity (Fig. 1; also see Materials and Methods). IgG subclass analysis revealed class switching to IgG3 > IgG2b = IgG2a > IgG1, suggesting B1 switching to IgG3 and B2 switching to the TH1-preferred IgG2a/b and -3 subclasses had occurred (data not shown). In a larger group of 11 GalNAcT^−/− mice immunized with HS:19(GM1⁺ GD1a⁺) LPS in CFA followed by IFA, a similar subclass pattern was observed.

FIG. 5.

Anti-ganglioside antibody responses in GalNAcT^−/− and GalNAcT^+/+ mice after i.p. immunization with HS:4 LPS (A) or HS:19(GM1⁺ GT1a⁺) LPS (B) in CFA and subsequently at 2-week intervals with LPS in IFA. The error bars indicate SD. The group sizes were as follows: HS:4 group, n = 9 (GalNAcT^−/− mice) and n = 7 (GalNAcT^+/+ mice); HS:19(GM1⁺ GT1a⁺) group, n = 5 (GalNAcT^−/− mice) and n = 3 (GalNAcT^+/+ mice). All sera were screened 4 days after each immunization. IgG anti-ganglioside antibodies, corresponding to the core OS structure on the immunizing LPS [GD1a for HS:4 LPS; GT1a and GM1 for HS:19(GM1⁺ GT1a⁺) LPS], appear after the second and third immunizations in significantly greater amounts in GalNAcT^−/− than in GalNAcT^+/+ mice (*, P < 0.05). In the ganglioside ELISA, GQ1b is substituted for GT1a (see Materials and Methods for details).

In order to more formally demonstrate the presence of B-cell memory for gangliosides, GalNAcT^−/− and GalNAcT^+/+ mice (n = 4) received a primary immunization with HS:19(GM1⁺ GD1a⁺) LPS in CFA, were screened for IgG anti-GD1a on day 7, and then were boosted on day 20 and rescreened on day 24. The anti-GD1a IgG response in GalNAcT^−/− mice showed typical characteristics of an antigen-specific memory response, being absent on day 7 (mean OD, 0.009 ± 0.007 standard deviation [SD]) and present on day 25 (mean OD, 0.41 ± 0.23 SD). No anti-GD1a IgG responses were observed in GalNAcT^+/+ mice (mean OD on day 7, 0.002 ± 0.001 SD; day 25, 0.015 ± 0.015 SD). Subsequent immunizations further magnified the IgG response (data not shown). These results indicate that tolerance to self gangliosides is the major factor limiting a class-switched memory response to ganglioside-mimicking core OS structures in C. jejuni LPS.

Analysis of individual GalNAcT^−/− mice immunized with HS:19(GM1⁺ GT1a⁺) LPS showed that two of four generated both anti-GQ1b and GM1 responses, whereas two of four developed anti-GQ1b responses alone. The ratio of GT1a to GM1 on HS:19(GM1⁺ GT1a⁺) is unknown, but the quantitatively major structure might be predicted to induce antibodies more favorably than the less abundant structure. In order to address this directly, we immunized groups of GalNAcT^−/− and GalNAcT^+/+ mice (four to five mice per group) with HS:4 and HS:19(GM1⁺ GD1a⁺) LPSs that are known to contain GD1a/GM1 ratios of 9:1 and 1:1, respectively. In the group of eight GalNAcT^−/− mice immunized with HS:4, GD1a was clearly immunodominant over GM1, with only anti-GD1a responses being seen, whereas in the group immunized with HS:19(GM1⁺ GD1a⁺) LPS, antibody responses to both GD1a and GM1 were present (Table 2). Among HS:19(GM1⁺ GD1a⁺) LPS-immunized GalNAcT^−/− mice, four developed an antibody response to GD1a alone and one developed an antibody response to GM1 alone, which persisted on secondary immunization. However, of the three mice that responded to both GD1a and GM1 on primary immunization, two of three failed to mount a secondary response to GM1 but did respond to GD1a. Two of eight GalNAcT^+/+ mice immunized with HS:19(GM1⁺ GD1a⁺) LPS developed low-magnitude but nevertheless significant IgG responses to GD1a (data not shown), and none of eight GalNAcT^+/+ mice immunized with HS:4 LPS developed IgG responses to either GM1 or GD1a.

TABLE 2.

Immune responses to GM1 and GD1a in GalNAcT^−/− mice immunized with HS:4 and HS:19(GM1⁺ GD1a⁺) LPSs

Characteristic	Value
	HS:4	HS:19(GM1+ GD1a+)
GM1/GD1a ratio	1:9	1:1
No. of animals immunized	8	8
No. of responders
GM1 alone	0	1
GD1a alone	8	4
GM1 and GD1a	0	3a

^aIn the three mice that developed both anti-GD1a and -GM1 IgG responses after two immunizations, no response to GM1 could be detected in two of three mice on subsequent boosts, although a secondary response to GD1a was observed.

In control immunization experiments with C. jejuni serostrain HS:3 LPS, which is not associated with GBS and lacks any ganglioside-like epitopes, neither GalNAcT^−/− nor GalNAcT^+/+ mice developed anti-ganglioside antibodies, but they did possess strong antibody responses to HS:3 LPS epitopes when screened in ELISA against HS:3, confirming that the immunization was effectively delivered (data not shown).

Immunization of GalNAcT^−/− mice with C. jejuni HS:10.

Experiments with GD3-ova liposomes showed complete unresponsiveness of GalNAcT^−/− mice to immunization with GD3, compared with responding wild-type mice, suggesting a high level of B-cell tolerance induced by overexpression of GD3 in GalNAcT^−/− mice. However, among GalNAcT^−/− mice immunized with HS:19(GM1⁺ GT1a⁺), two of four animals with anti-GQ1b antibodies also developed anti-GD3 responses, and two of four GalNAcT^+/+ mice developed anti-GD3 antibodies without anti-GQ1b antibodies. This suggests that there is an intrinsic pool of GD3-specific B cells that do not respond to immunization with our schedule of GD3-ova liposomes in GalNAcT^−/− mice but are not entirely suppressed following immunization with GD3-like oligosaccharides associated with a highly potent immunomodulator like LPS. To explore this further, GalNAcT^−/− mice (n = 6) were immunized with LPS from the C. jejuni strain HS:10, which contains GD3-like structures in its LPS. No significant anti-GD3 IgM responses were seen. However, three of six mice developed low-level IgG antibodies to GD3 after the third immunization (mean OD, 0.18 ± 0.08 SD; n = 6 mice), indicating that tolerance to GD3 can be overcome when the GD3-like epitope is delivered on LPS, even in GD3-ova liposome-unresponsive mice. GalNAcT^−/− mice immunized with other LPS species which do not contain GD3 or GD3-like epitopes did not generate anti-GD3 responses, indicating that the anti-GD3 response is not simply due to a polyclonal activating effect of LPS in this mouse strain.

Anti-ganglioside responses in C3H/HeN and C3H/HeJ mice immunized with ganglioside-bearing LPS.

Our experiments comparing anti-ganglioside antibody responses following immunization with ganglioside-mimicking LPS species or native gangliosides suggest that in some situations, such as that seen with HS:10 LPS and GD3, the LPS provides a stronger antigenic stimulus than the ganglioside. In order to determine whether the lipid A component of LPS is providing any accessory help to B cells in generating antibody responses to LPS core OS structures, we examined antibody responses in C3H/HeJ mice. Lipid A associates with lipid binding protein, binds to CD14 on B cells, and signals via the transmembrane Toll receptor, Tlr4. C3H/HeJ mice harbor a mutation in the Toll receptor which renders them unresponsive to lipid A signaling via CD14, but BCR and other accessory signaling pathways remain intact. HS:4 LPS, which bears the immunodominant GD1a structure, was used to examine core OS responses in LPS-responsive C3H/HeN and LPS-unresponsive C3H/HeJ mice (Fig. 6). All C3H/HeN mice showed some anti-GD1a antibody response, whereas responses were significantly attenuated in C3H/HeJ mice, with only two of five mice developing very low-level anti-GD1a IgG antibodies. The mean anti-GD1a responses were greater in magnitude and developed earlier in C3H/HeN than in C3H/HeJ mice. The profile of anti-GD1a antibodies over time did not show any clear secondary response, remaining at constant levels upon subsequent immunizations (data not shown). This is consistent with the tolerance to GD1a one would predict in a GD1a-bearing mouse and from the lack of a substantial GD1a-specific B-cell memory compartment. Interestingly, C3H/HeN mice also mounted a modest anti-GD3 response, despite the absence of GD3 on HS:4 LPS, suggesting that the anti-GD3 antibodies have arisen through polyclonal B-cell activation by LPS, which included a pool of GD3-specific B1 cells. C3H/HeJ mice did not exhibit this anti-GD3 response, again suggesting it is due to Lipid A-mediated polyclonal activation. Thus, the data indicate that the lipid A component of LPS mediates an important signaling pathway that helps B cells respond to ganglioside-like core OS structures.

FIG. 6.

Anti-ganglioside IgM (A) and IgG (B) antibody responses following HS:4 LPS immunization of C3H/HeN and C3H/HeJ (LPS low responder) mice. The error bars indicate SD. HS:4 LPS bears a GD1a-like epitope. Groups of mice (n = 5) received serial immunizations with HS:4 LPS in CFA-IFA at 2-week intervals, and sera were assayed on days 4 and 10 following each immunization. C3H/HeN mice showed significantly greater IgG responses than C3H/HeJ mice to GD3 (*, P = 0.002), and elevated IgG responses to GD1a, compared with C3H/HeJ mice (**, P = 0.066).

Effect of CD40 activation in enhancing anti-LPS core OS responses in BALB/c mice.

B cells proliferate and differentiate into antibody-forming cells in response to CD40 ligation, either provided naturally by cognate TH cell interactions or via experimental mechanisms, such as anti-CD40-stimulating antibodies. Since LPS lacks protein and thus has no clearly identified mechanism for recruiting T-cell-dependent help in the mouse, we examined the possibility that coadministration of anti-CD40 antibodies with ganglioside-mimicking LPS might provide the accessory help that would also result in breaking tolerance to self gangliosides. Groups of BALB/c mice (n = 4) were immunized with HS:4 or HS:19(GM1⁺ GT1a⁺) LPS containing liposomes admixed with anti-CD40 antibody or the control antibody GL117 (0.5 mg/mouse) in the absence of any protein adjuvant. Sera studied by ELISA for anti-ganglioside and anti-LPS antibodies on day 7 following HS:4 immunization with CD40 or GL117 exhibited an increase in the nonspecific background signal, more marked with the CD40 antibody (IgM mean OD on day 0, 0.15 ± 0.04 SD; day 7, 0.87 ± 0.08 SD) than control antibody (mean OD on day 0, 0.10 ± 0.005 SD; day 7, 0.19 ± 0.04 SD). This elevation in background, attributed to elevated serum IgM and IgG as a result of nonspecific polyclonal B-cell activation, partially obscured any underlying ganglioside- or LPS-specific responses, making interpretation difficult. Although anti-GD3 IgM and IgG responses were elevated above background level, they did not reach significance compared with the control antibody GL117. However, the absence of GD3 on HS:4 LPS again indicates that anti-GD3 antibodies form part of the natural memory repertoire.

On day 28 following immunization with HS:19(GM1⁺ GT1a⁺) LPS, background IgM levels had fallen compared with day 7 (data not shown), and enhancement of the ganglioside-specific signal, expected to comprise GQ1b, GM1, and GD3 for HS:19(GM1⁺ GT1a⁺) LPS, could be seen, although this did not achieve statistical significance compared with the control MAb, GL117 (Fig. 7A). However, in the IgG class on day 28 (Fig. 7B), a significant increase in GQ1b and GD3 titers was seen in the anti-CD40-treated mice compared with the control antibody (P < 0.005 and P < 0.05, respectively), indicating that anti-CD40 antibody had a modest effect on expanding the IgG anti-ganglioside B-cell pool and thus could partially overcome tolerance to self gangliosides.

FIG. 7.

. Anti-ganglioside IgM (A) and IgG (B) antibody responses in BALB/c mice immunized with HS:19(GM1⁺ GT1a⁺) LPS coadministered with anti-CD40 antibody or control antibody, GL117. Groups of BALB/c mice (n = 4) were immunized i.p. with HS:19(GM1⁺ GT1a⁺) LPS and anti-CD40 antibody or GL117 and then reimmunized on day 21 and bled on day 28 (7 days post-second immunization). Mice treated with anti-CD40 antibody showed slightly greater anti-GQ1b and anti-GD3 responses than control antibody-treated mice (*, P < 0.005; **, P < 0.05), but there was no significant difference in anti-GM1 antibody responses.

DISCUSSION

Gangliosides are self glycolipid structures widely distributed on extracellular surfaces throughout the body and thus subject to normal mechanisms regulating B-cell tolerance to TI antigens, as also seen, for example, in the ABO blood group system (61). In attempts to enhance their immunogenicity in mice, many studies have converted gangliosides to TD antigens by conjugation to proteins or incorporation into protein-enriched liposomes (1, 26, 31, 38). Indeed, we show here that BALB/c mice immunized with GQ1b-ova liposomes do generate low-level anti-GQ1b responses that show characteristics of a memory response. Priming with ova is a necessary step and likely acts through the generation of an anti-ova-specific TH cell pool that then helps the B-cell pool recognizing the ganglioside to switch class and affinity mature. This model relies on uptake of liposome by B cells and subsequent presentation of ova peptides in association with major histocompatibility complex class II to the primed ova-specific TH cells. Only mice immunized i.p. with ganglioside-ova liposomes responded in this manner, whereas immunization via the FP or s.c. route, although generating strong anti-ova responses, failed to generate anti-ganglioside antibodies. These data strongly suggest that the i.p. B1-cell pool can be more readily stimulated to secrete anti-ganglioside antibodies but that the conventional B2-cell pool is highly tolerant to self gangliosides.

I.p. immunization of normal mice with C. jejuni LPS containing ganglioside-like epitopes typically results in transient IgM anti-ganglioside responses that lack a highly developed secondary immune response. Since the gut is the principal site of infection with LPS-bearing enteric bacteria, the production of low-affinity IgM anti-ganglioside antibodies upon exposure to C. jejuni LPS is consistent with models of a natural protective repertoire in early host defense (15, 53). Provided the maturation of this response is regulated, the rapid production of low-affinity antibody to host-mimicking carbohydrate structures on LPS-bearing bacteria outweighs the danger of developing autoimmune disease mediated through such antibodies.

GalNAcT^+/+ mice, which bear a normal complement of gangliosides, developed similar responses to BALB/c mice when immunized with GQ1b liposomes. In contrast, GalNAcT^−/− mice, which express only GM3 and GD3, developed highly elevated anti-ganglioside-specific secondary responses when immunized with liposomes containing non-self GD1a or GQ1b but no response to self GD3, a quantitatively dominant tissue ganglioside in the GalNAcT^−/− mouse. Thus, tolerance to self gangliosides is clearly the major factor that prevents the maturation of antibody responses, as previously observed in studies of GD1a ganglioside (38).

This study also clearly shows that tolerance to self gangliosides is the major factor restricting the antibody response to ganglioside-mimicking LPS core OSs. Thus, GalNAcT^−/− mice are an important tool for studying antibody responses to C. jejuni LPSs that bear ganglioside-like structures. These responses in the GalNAcT^−/− mouse also switch class to IgG and exhibit memory. It is also evident that the antibody response to ganglioside-like structures on LPS is dependent on both the ganglioside mimicked and its level of expression. For example, the GM1/GD1a ratio of 1:9 on HS:4 LPS favors the production of anti-GD1a antibodies, whereas immunization with HS:19(GM1⁺ GD1a⁺), with a GM1/GD1a ratio of 1:1, produces both anti-GM1 and anti-GD1a antibodies. This differential effect between HS:4 and HS:19(GM1⁺ GD1a⁺) LPSs may be related to the differences in efficiency of density-dependent cross-linking of the BCR specific for GM1 and GD1a; thus, the environment created by a 9:1 ratio of GD1a over GM1 favors activation of GD1a-specific B cells.

In mice immunized with ganglioside liposomes containing ova, or with ganglioside-protein conjugates, the signals necessary for generating class-switching and memory responses are dependent upon the presence of protein. BCR clustering mediated by repeating carbohydrate epitopes, as found in TI-2 antigens, can often lead to B-cell proliferation with antibody production in the absence of additional stimuli. However, our ganglioside liposomes, designed to mimic TI-2 antigens, clearly required protein (in the form of ova) to be effective immunogens.

It is unclear from our experiments how class switching and memory responses were generated in response to the ganglioside-mimicking core OS structures in ganglioside-deficient mice. The LPS was delivered in CFA at the first immunization but subsequently in protein-free IFA. One possibility is that the lipid associated with bacterial polysaccharide antigens could be presented via CD1d to generate ganglioside-specific T cells; however, there is no previously defined example by which such T cells could provide cognate help to polysaccharide-specific B cells (20, 58). LPS is a TI-1 antigen that acts as a polyclonal B-cell activator; such responses are principally IgM and do not exhibit memory. We saw some evidence that such polyclonal activation occurred and that this was in part directed against GD3 (see below). Other than GD3 responses, the anti-ganglioside antibody response to LPS was critically dependent upon the core OS structure, exhibiting clear evidence of mimicry.

As an alternative mechanism, the lipid A moiety of LPS might be providing a secondary signal to drive the maturation of anti-core OS responses. We found the responses of C3H/HeJ mice, hyposensitive to lipid A due to a mutation in Tlr4 (3, 13, 33, 52, 53), were diminished in comparison to those of normally responding C3H/HeN mice, bearing in mind that the responses to core OS structures in ganglioside-tolerant C3H/HeJ and C3H/HeN mice would not be expected to be as high as in GalNAcT^−/− mice. This provides in vivo evidence that lipid A signaling via Tlr4 is an important secondary signaling pathway contributing to the B-cell response to TI epitopes within LPS core OS structures.

The terminal disialosyl structure of ganglioside GD3, also present on HS:10 LPS, appears to be a dominant antigen in the mouse. Even GalNAcT^−/− mice, which contain large amounts of GD3, generated low but significant anti-GD3 responses when immunized with HS:10, although they did not respond to GD3 liposomes. This further emphasizes the potent stimulatory properties of LPS in comparison to liposomes. GQ1b liposomes that also contain this terminal disialosyl epitope generated anti-GQ1b responses in BALB/c and GalNAcT^+/+ mice, whereas in contrast, GD1a-immunized GalNAcT^+/+ mice did not develop anti-GD1a responses. In addition to anti-GD3 responses generated through GD3-specific BCR on B cells, it is also evident that anti-GD3 antibodies can arise through polyclonal B-cell activation, mediated by lipid A. Thus, C3H/HeN mice immunized with HS:4 LPS frequently developed anti-GD3 antibody responses, although no GD3-mimicking structures are present on this LPS. The predominance of anti-GD3 antibodies arising from both antigen-specific and nonspecific B-cell activation indicate that this specificity forms an important part of the natural antibody repertoire in mice, most likely produced by peritoneal B1 cells, involved in early host defense against enteric bacteria.

The role of CD40 ligation, using stimulating antibodies, in substituting for cognate T-cell help has been previously considered and tested with respect to polysaccharide antigens (12, 16). We thus considered this a possible mechanism for driving anti-ganglioside antibody responses following immunization with LPS in the absence of any protein adjuvant. We observed a very modest increment in ganglioside responses to LPS, but this was not observed in parallel experiments with ganglioside liposomes (data not shown). Thus, although CD40 is clearly critical to accessory signaling in cognate T-cell help, we were unable to reproduce the effect to any great extent using this experimental paradigm.

In human subjects with uncomplicated enteritis caused by C. jejuni strains containing self-ganglioside mimics in the core OS structures, one would predict that antibody responses to the self carbohydrate structures should be suppressed. Indeed, acute-phase anti-ganglioside antibodies do not arise in this clinical group (6, 47). However, in the small proportion of subjects with Campylobacter enteritis who develop GBS, tolerance is clearly broken in that high-titer class-switched anti-ganglioside antibodies arise in response to ganglioside mimics in the LPS. Why such individuals are unable to maintain tolerance in the setting of this infection is unknown. There is no evidence that such individuals express smaller amounts of tissue gangliosides than normal, as is the case with GalNAcT^−/− mice. It seems more likely that any negative selection is overcome by powerful TD help provided by elements of the infection. The class switching to the TD IgG1 and IgG3 subclasses that occurs in GBS supports this view and also suggests that such B cells originate in the B2 compartment. A T helper repertoire could arise de novo as part of the infectious episode triggering GBS or could be stimulated from the memory compartment, the latter possibly providing a more potent TD environment. In either of these situations, the time course of B-cell antibody production would be that of a primary response, as is the pattern seen in GBS. Positive selection by self antigen, LPS, or other microbial components of B cells in the B1 compartment that accounts for low-affinity predominantly IgM anti-carbohydrate antibodies would be unlikely to result in the high-titer, class-switched anti-ganglioside antibodies seen in GBS (24). Irrespective of the mechanisms by which anti-ganglioside antibodies arise in GBS, these experiments show that ganglioside-mimicking C. jejuni LPS, when combined with appropriate TD help, provided here by protein adjuvants, can induce a high-titer class-switched antibody response exhibiting memory and that the major negative regulator of this response is tolerance to self gangliosides.