In Vivo Phase Variation and Serologic Response to Lipooligosaccharide of Campylobacter jejuni in Experimental Human Infection

Abstract

Some Campylobacter jejuni strains which exhibit mimicry of gangliosides in their lipooligosaccharides (LOSs) are associated with development of Guillain-Barré syndrome, which complicates the selection of a suitable C. jejuni strain in a live-attenuated vaccine. C. jejuni 81-176 is the most well characterized strain available, but structurally, LOS of C. jejuni 81-176 exhibits mimicry of predominantly GM₂ and GM₃ gangliosides. We compared the antiganglioside human serologic responses of 22 volunteers post-oral vaccination (two-dose series, 14 days apart) with a killed whole-cell C. jejuni vaccine, those of volunteers (22 following initial challenge and 5 upon rechallenge) experimentally infected with the homologous C. jejuni vaccine strain 81-176, and those of 12 volunteers used as controls (placebo recipients). All volunteers were evaluated using thin-layer chromatography immuno-overlay and a panel of nine gangliosides at days 0, 21, and 28 either postvaccination or postinoculation. Antiganglioside antibodies were identified at baseline in 6 of the 61 volunteers (9.8%). There were no antiganglioside antibodies observed following vaccination or experimental infection rechallenge. Evidence of seroconversion was observed in 2 of 22 (9.1%) in the initial infection challenge group, comparable to 1 of 12 (8.3%) in the placebo recipients. Additional testing of seven selected volunteers in the initial challenge group at days 0, 3, 7, 10, 21, 28, and 60 showed that when antiganglioside antibodies occurred (mostly anti-GM₁ and -GM₂), responses were weak and transient. Furthermore, evidence from serologic probing of LOSs of isolates recovered from stools of six volunteers indicated that the isolates had undergone antigenic phase variation in ganglioside mimicry during passage in vivo. Collectively, with the exception of one volunteer with anti-GM₂ antibodies at day 60, the results show an absence of persistent antiganglioside antibodies after experimental infection with C. jejuni or following administration of a killed C. jejuni whole-cell oral vaccine, although LOS phase variation occurred.

Campylobacter jejuni is the most commonly reported bacterial cause of human gastroenteritis worldwide (6) and is a major antecedent of traveler's diarrhea (25). Due to the lack of a suitable animal model of virulence, the pathophysiology of C. jejuni infection is poorly understood, which in turn has delayed vaccine development efforts. In addition, the link between C. jejuni infection and development of reactive arthritis, as well as that between C. jejuni infection and the postinfectious neurological complications Guillain-Barré syndrome (GBS) and Miller Fisher syndrome, further increases the public health and economic burdens of C. jejuni-associated disease (2). Importantly, C. jejuni is antigenically complex, as demonstrated by the presence of more than 70 heat-stable (HS) serotypes (14).

Lipopolysaccharide and lipooligosaccharides (LOSs) are HS antigens on the surface of gram-negative bacteria, and many C. jejuni HS serotypes have been shown to bear LOS structures that mimic human gangliosides, components of nerve tissue (21). It is thought that this molecular mimicry is responsible for the induction of antiganglioside antibodies, which are considered to play a role in GBS and/or Miller Fisher syndrome development (15, 21). Also, it is now evident that C. jejuni LOS undergoes high-frequency phase variation (7, 8, 10, 12, 13), and one strain of C. jejuni can produce a repertoire of LOS molecules. This complicates the selection of a suitable C. jejuni strain as a candidate for a live-attenuated vaccine. Thus, the development of a vaccine to control rising C. jejuni incidences must take into account the phenomenon of molecular mimicry of gangliosides by C. jejuni LOS, phase variation, and the lack of a conserved protective antigen.

In the present study, using thin-layer chromatography (TLC) with immunostaining, we serologically characterized LOS from C. jejuni 81-176, a reference strain currently being developed as a killed whole-cell vaccine, using a panel of ganglioside-specific and ligand-binding reagents. Also, a unique opportunity to define the antiganglioside human serologic responses resulted upon completion of a volunteer experimental oral infection study with the C. jejuni 81-176 reference strain (D. Tribble, S. Baqar, M. Oplinger, D. Scott, D. Rollins, S. Walz, E. Burg, A. Moran, and A. Bourgeois, Abstr. 10th Int. Congr. Immunol., abstr., p. 424, 1998). Our aim was to investigate the potential of the strain to trigger an antiganglioside response and, thus, contribute to neurological symptoms. We monitored the serologic response to gangliosides at three time points (0, 21, and 28 days) during the trial. Furthermore, we serologically characterized the LOS from six C. jejuni isolates recovered from volunteers after passage and compared it to the LOS from the reference strain.

MATERIALS AND METHODS

Administration of CWC vaccine.

C. jejuni 81-176 (Penner serotype HS:23/36; Lior serogroup 5) was used following inactivation in the Campylobacter whole-cell (CWC) vaccine (25). The strain was previously isolated from the feces of a 9-year-old girl with diarrhea during a milk-borne outbreak in Minnesota (11). The strain has subsequently been studied in an experimental infection study of volunteers at the Center for Vaccine Development, University of Maryland (1). The inactivated whole-cell vaccine was produced using good manufacturing practices and good laboratory practices at the Walter Reed Army Institute of Research Forest Glen Annex facility through a Cooperative Research and Development Agreement with Antex Biologics, Gaithersburg, Md. The vaccine dose was delivered as a suspension in a bicarbonate buffer solution (4.0 g of USP-grade sodium bicarbonate dissolved in 150 ml of sterile water) containing 2.5 × 10⁹ CFU coadministered with LT(R192G) as mucosal adjuvant, a genetically modified heat-labile toxin of enterotoxigenic Escherichia coli. LT(R192G) was constructed as described by Dickinson and Clements (3) with purification using agarose affinity chromatography, and the Swiss Serum Vaccine Institute, Bern, Switzerland, under contract from the Naval Medical Research Center, produced the good-manufacturing-practices lot. The adjuvant was added as a separate suspension in a total volume of 5 ml (25 μg) with the CWC vaccine just prior to ingestion. The placebo consisted of only bicarbonate buffer. The vaccine was delivered as a two-dose regimen at a 14-day interval. The vaccination procedure was as follows: 90-min fast, delivery of assigned treatment in buffer drink (ingested as a combined solution), verification of full vaccination, and postdose 90-min fast. All vaccinations were done at the U.S. Army Medical Research Institute of Infectious Diseases, Fort Detrick, in Frederick, Md.

Following informed consent, civilian and military volunteers underwent evaluations to determine eligibility, including medical history, physical examination, and screening laboratory and preenrollment testing to assess their understanding of the study. Volunteers were excluded from enrollment due to chronic gastrointestinal disorders, personal or family history of an inflammatory arthritis, autoimmune or neurologic disease, pregnancy, HLA B27 positivity, human immunodeficiency virus seropositivity, or any disease or condition as determined by a study physician that would place the subject at increased risk for adverse effects. Additional exclusions not related to volunteer safety included previous vaccination with experimental Campylobacter vaccines, traveler's diarrhea in the past year, and history of microbiologically confirmed Campylobacter infection or serologic evidence of infection (defined as Campylobacter-specific antiglycine extract immunoglobulin A [IgA]; optical density [OD] at 650 nm > 0.5 at 1:1,000 dilution). The same eligibility criteria were used for all volunteers in either the vaccination or infection groups.

Administration of C. jejuni 81-176 reference strain to volunteers.

A C. jejuni strain 81-176 experimental infection challenge study was undertaken to assess preliminary efficacy of the CWC vaccine. The organism, prepared from the master seed stock stored at −85°C in glycerol thioglycolate soy broth, was administered at an inoculum of 10⁹ CFU. C. jejuni confirmation used standard microbiologic methods, and viable colony counts correlated with OD readings. Inocula were prepared from individual confirmed colonies. These colonies were suspended in thioglycolate soy broth and then plated onto Brucella agar (Difco, Detroit, Mich.) overnight at 42°C in a microaerobic atmosphere (5% O₂, 10% CO₂, 85% N₂) for confluent growth. Growth was harvested in 5 ml of phosphate-buffered saline (PBS) and adjusted to a target OD with PBS for standardizing a subsequent infectious dose. Pre- and postdose colony counts were performed. The appropriate inoculum (10⁹ CFU) was suspended in a solution containing 2 g of sodium bicarbonate in 150 ml of sterile water and then ingested by the volunteer. Volunteers were admitted to an inpatient research ward for daily monitoring. All volunteers were treated with antibiotics (no later than 1 week after inoculation) and had documented resolution of symptoms and eradication of infection prior to discharge.

Blood and stool specimen collection schedule.

For specimen collection, blood was collected throughout the course of the study to measure the immune response to vaccination or experimental infection based on group assignment. Comparable time points available for study for all groups were selected for measurement of antiganglioside antibody determination. Blood samples were processed to remove mononuclear cells for immunologic testing in a related study (Tribble et al., Abstr. 10th Int. Congr. Immunol.), and to avoid additional blood collection, plasma samples were used for serologic testing. Importantly, in other studies, we did not find a difference in antibody levels between plasma and serum samples (S. Baqar and A. P. Moran, unpublished data). Plasma samples were stored at −70°C at the time of initial processing and shipped on dry ice for immediate analysis. Study days 0 (baseline prior to intervention), 21, and 28 were selected for comparison among 61 volunteers (n = 183 samples). More-frequent blood specimen time points were available following experimental infection. Subsequent to the initial screening, four volunteers who showed antiganglioside antibodies at one or more of these test time points and three others who did not exhibit antibodies at these time points were selected. Additional testing was performed on plasma samples from these seven volunteers taken at 0, 3, 7, 10, 21, 28, and 60 days (n = 49 samples). All stool specimens obtained during inpatient observation were cultured for the presence of C. jejuni.

Growth conditions of C. jejuni 81-176.

For analysis, the reference strain C. jejuni 81-176 was cultured separately on blood agar (Columbia agar base [CBA; Oxoid Ltd., London, England], or Brucella agar base with 10% (vol/vol) unlysed horse blood (BBA) or on brain heart infusion (BHI) agar (Oxoid) under microaerobic conditions (16). Isolates were recovered from the stools of six volunteers, and bacterial identification was accomplished by established procedures (26). Recovered isolates were stored in glycerol solution at −70°C. Isolates were cultured from the frozen stocks onto the three media described above and grown under the same conditions.

Extraction of LOSs.

Bacterial biomass was harvested and bulk extraction of C. jejuni 81-176 LOS was performed by the hot phenol-water extraction procedure (16). For the C. jejuni isolates recovered from experimentally infected volunteers, the LOSs were extracted by a mini-phenol-water extraction procedure (19).

Electrophoresis.

The purified LOS preparations from the C. jejuni reference strain and from the six recovered isolates, each grown in the three culture media (CBA, BBA, and BHI), were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, using a stacking gel of 5% (wt/vol) acrylamide and a separation gel of 15% (wt/vol) acrylamide containing 3.2 M urea (BDH Laboratory Supplies, Poole, England), followed by silver staining (20).

Sialic acid determination.

Concentrations of sialic acid (Neu5Ac, N-acetylneuraminic acid) were determined in LOSs by the modified Ehrlich reaction assay (4).

TLC and immunostaining.

For characterization of the ganglioside mimicry expressed by strains, gangliosides (1-μg aliquots; Sigma Chemical Co., St. Louis, Mo.) and LOS extracts (1 μg or 5-μl aliquots) were analyzed by TLC on precoated silica gel 60 glass plates (Merck, Darmstadt, Germany). Solvent systems consisting of chloroform-methanol-0.22% CaCl₂ · 2H₂O (50:45:10 [vol/vol/vol]) (23), and n-propanol-water-25% NH₄OH (60:30:10 [vol/vol/vol]) (24) were used as developers for gangliosides and LOSs, respectively. Separated gangliosides and LOSs were visualized by spraying plates with resorcinol-HCl reagent (24). TLC with immunostaining was performed using the immuno-overlay procedure of Saito et al. (23), as modified by Schwerer et al. (24). LOS preparations were tested for reaction with the antiganglioside antibodies and other antibodies listed in Table 1, with the appropriate peroxidase-conjugated secondary antibody (Sigma) diluted 1:500 in 0.3% gelatin-PBS solution (19). Immunoreactants were visualized by use of a horseradish-peroxidase development system (Bio-Rad Laboratories, Hercules, Calif.). Alternatively, to assay the ganglioside-reactivity of human plasma samples, lanes were overlaid with sera diluted 1:100 in gelatin-PBS. This procedure was chosen, rather than enzyme-linked immunosorbent assay (ELISA), since the former has been applied and validated in previous serologic studies on C. jejuni LOS and antiganglioside antibodies (17-20, 24) and optimal presentation of the relevant ganglioside epitopes for detection by anti-C. jejuni antibodies occurs in TLC immuno-overlay (A. P. Moran, unpublished data). Positive and negative control experiments for antibody binding were performed whereby serum from a patient with a high titer of anti-GM₁ IgG antibodies was used (a generous gift from B. Schwerer, Vienna, Austria), or normal human serum without antiganglioside antibodies was used instead of plasma from volunteers. Confidence intervals for antiganglioside antibody responses were generated using a normal approximation to the binomial distribution. Binding studies with cholera toxin (CT)-peroxidase conjugate (1 μg/ml; Sigma) and peanut agglutinin-peroxidase conjugate (20 μg/ml; Kem-En-Tec, Copenhagen, Denmark) were performed under the same conditions as those described for immunostaining (19).

TABLE 1.

Antibodies used in characterization of C. jejuni LOS preparations

Antibody specificity	Host	Antibody class	Clonality	Source
Anti-GM1	Rabbit	IgG	Polyclonal	MIa
Anti-asialo-GM1	Rabbit	IgG	Polyclonal	MI
Anti-GM2	Rabbit	IgG	Polyclonal	MI
Anti-asialo-GM2	Rabbit	IgG	Polyclonal	MI
Anti-GD2	Mouse	IgG	Monoclonal	MI
Anti-GD3	Mouse	IgG	Monoclonal	MI
Anti-disialosyl	Human	IgM	Monoclonal	INSb
Anti-C. jejuni HS:10 serostrain	Rabbit	IgG	Polyclonal	NMRCc

^aAntibody was obtained from Matreya Inc., Pleasant Gap, Pa.

^bAntibody was provided by H. Willison, University Department of Neurology, Institute of Neurological Sciences, Southern General Hospital, Glasgow, Scotland.

^cAntibody was provided by P. Guerry, Naval Medical Research Center.

RESULTS

Electrophoretic characterization of LOSs from the reference strain and recovered isolates.

All the hot phenol-water-extracted LOSs from the C. jejuni 81-176 reference strain, independent of growth medium, had the same electrophoretic profile of a low-M_r LOS (data not shown), identical to that observed previously (10). Furthermore, LOSs extracted from C. jejuni 81-176 using the mini-phenol-water procedure (19) gave a profile identical to those obtained using LOSs extracted by the lengthier phenol-water extraction procedure (16). We have previously validated the mini-phenol-water technique and demonstrated that it is a suitable extraction procedure for LOS and yields material sufficiently pure for subsequent analysis by TLC immunostaining (19). As with the reference strain, the LOSs from each of the six recovered isolates, grown in all three media (18 LOS preparations [Table 2]), produced low-M_r bands in silver-stained gels but, compared with the reference strain, differences in LOS banding pattern of the isolates were observed similar to those seen previously in phase-variable LOSs (10).

TABLE 2.

Reaction of antibodies and ligands with LOSs of C. jejuni reference strain 81-176 and six isolates from inoculated volunteers when grown on three media^a

Antibody or ligand	Positive control	Strength of reaction with LOS of C. jejunib:
		Reference strain 81-176	05600			05601			06052			06053			06072			06518
			CBA	BBA	BHI	CBA	BBA	BHI	CBA	BBA	BHI	CBA	BBA	BHI	CBA	BBA	BHI	CBA	BBA	BHI
Antibodies
Anti-GM1	Strain HS:41	−	+++	++	+++	+++	++	+++	+++	++	+++	(+)	+++	++	+++	+++	+++	++	+++	++
Anti-asialo-GM1	Strain HS:2c	−	(+)	++	+++	(+)	++	+++	++	++	+++	(+)	+++	+++	++	+++	+++	++	+++	+++
Anti-asialo-GM2	Asialo-GM2	−	(+)	ND	ND	(+)	ND	ND	(+)	ND	ND	−	ND	ND	−	ND	ND	−	ND	ND
Anti-GM2	GM2	−	+	++	++	+	++	++	+	+++	++	(+)	+++	++	(+)	++	++	(+)	+++	++
Anti-GD2	GD2	−	(+)	ND	ND	(+)	ND	ND	(+)	ND	ND	−	ND	ND	−	ND	ND	−	ND	ND
Anti-disialosyl	Strain PG836	−	+++	ND	ND	++	ND	ND	++	ND	ND	(+)	ND	ND	++	ND	ND	+	ND	ND
HS:10 antiserum	Strain HS:10	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−
Ligandsd
CT-HRP	Strain HS:41	−	+	++	+	(+)	(+)	+	(+)	(+)	+	−	−	−	−	−	−	−	−	−
PNA-HRP	Strain HS:2c	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−

^aCulture growth media: CBA, Columbia blood agar; BBA, Brucella blood agar; BHI, brain heart infusion agar (no blood). The reference strain, 81-176, was grown on all three media like the other C. jejuni isolates.

^b+++, strong reaction; ++, moderate reaction; +, weak reaction; (+), very weak reaction; −, no reaction; ND, not determined.

^cDesialylated C. jejuni NCTC 11168 LOS was used.

^dLigands: CT-HRP, cholera toxin-peroxidase conjugate; PNA-HRP, peanut agglutinin-peroxidase conjugate.

Neu5Ac determination.

The LOS of the reference strain that was grown on CBA had the highest concentration of Neu5Ac (119.1 nmol/mg), whereas Neu5Ac was detected at concentrations of 92.1 and 87.4 nmol/mg in the LOSs of the BBA- and BHI-grown strains, respectively. Thus, medium composition influenced the amount of Neu5Ac expressed in the LOSs of C. jejuni 81-176 and is reflective of phase variation in ganglioside mimicry by LOS of the reference strain in vitro (10). Thus, in subsequent serologic probing of the recovered isolates for ganglioside mimicry (see below), analysis was performed on each strain grown on these three respective media (Table 2). Nevertheless, in vitro selection of specific types of ganglioside mimicry, not reflective of what may take place in vivo, could occur in the recovered isolates when grown in vitro, and thus isolates were subjected to minimal subculturing before testing. This led to limitations in available amounts of biomass from the recovered isolates for purification of LOSs, and since the requirement of material for serologic characterization was paramount, the Neu5Ac contents of these LOSs were not determined.

Serologic characterization of LOSs from the reference strain and recovered isolates.

To determine the ganglioside mimicry occurring in C. jejuni LOSs, serologic probing was used since small amounts of purified LOSs were available, particularly from recovered isolates, which were sufficient for serologic probing but not complete structural analysis. In addition to analysis of the recovered isolates, the reference strain was examined for comparison. Our previous structural studies have established that C. jejuni 81-176 LOS expresses mimicry of predominantly GM₂ and GM₃ gangliosides but also expresses, to a minor extent, mimicry of GD₂ and GD_1b gangliosides (10; A. P. Moran et al., unpublished data). As shown in Table 2, tests for reactivity of reference strain LOSs with anti-GM₁ antibodies were negative, consistent with the absence of GM₁ mimicry. Antibodies to gangliosides asialo-GM₁, GM₂, asialo-GM₂, GD₂, and GD₃ and to disialosyl disaccharides did not react with the LOSs but reacted with a panel of control gangliosides and C. jejuni LOSs which are known to bear these ganglioside structures. Also, results of testing with CT (a GM₁ ligand) and peanut agglutinin (a Gal→GalNAc ligand) indicated the absence of mimicry of the major gangliosides implicated in GBS development. Compared with the structural data, the lack of reaction of C. jejuni 81-176 LOSs with anti-GD₂ antibodies is likely attributable to the low proportion of molecules with this type of mimicry in the LOSs of the reference strain. Although the anti-GM₂ antibodies reacted with control GM₂, the lack of reaction of these antibodies with reference strain LOSs contrasts with the structural findings of predominant expression of GM₂ and GM₃ mimicry. Nevertheless, these antibodies did not react with another control C. jejuni LOS bearing GM₂ ganglioside mimicry (serostrain HS:23), identical to the observations with reference strain LOSs. These results reflect the importance of epitope presentation for detection of ganglioside-like epitopes in C. jejuni LOSs by antiganglioside antibodies.

Remarkably, as shown in Table 2, LOS from each passaged isolate reacted with a range of antiganglioside antibodies and ganglioside-specific ligands and, thus, bore recognizable ganglioside epitopes. For each isolate, the spectrum of reactions of their LOSs with ganglioside-detecting reagents were independent of growth medium, although sometimes differing in strength. In general, C. jejuni LOSs from all these isolates reacted very strongly with anti-GM₁ and -GM₂ antibodies, suggesting the occurrence of such mimicry. However, the lack of binding of CT (a GM₁ ligand) in three of the six isolates did not confirm GM₁ mimicry in these isolates. Due to reaction of LOSs from all the six isolates with antidisialosyl antibodies, a neuroaminobiose unit on a GM₁/GM₂-like OS backbone is proposed. In addition, for some of the isolates weak reactions with anti-GD₂ antibodies were observed. Thus, in contrast to the absence of serologically detectable mimicry by the major gangliosides in the reference strain, after in vivo passage the isolates possess a Neu5Acα2-3Gal or Neu-5Acα2-8-Neu5Aca2-3Gal epitope in the form of GM₂/GD₂/GD_1b mimicry. Hence, the C. jejuni LOS underwent antigenic phase variation during passage in vivo. Consistent with these findings, a previous in vitro study has shown that C. jejuni 81-176 possesses the genetic mechanisms for its LOS to undergo phase variation resulting in not only expression of GM₂ and GM₃ ganglioside-like structures but also GD_1b and GD₂ ganglioside mimicry (10).

Testing of human sera for ganglioside reactivity.

The serologic response to a panel of gangliosides in plasma samples (n = 183) obtained at the day of intervention (day 0), vaccination or infection, and 21 and 28 days postintervention was studied. At baseline (prior to intervention) 6 of the 61 volunteers, or 9.8% (95% confidence interval [CI], 2.4 to 17%), exhibited a serological response to at least one of the gangliosides in the panel. There were no seroconversions among the 22 volunteers in the CWC vaccine group or the 5 volunteers following their rechallenge to the experimental infection. All of the 22 volunteers in the experimental infection group became infected, and 17 (77%) became ill, with clinical outcomes ranging from mild to severe diarrhea. Two of the 22 volunteers, 9.1% (95% CI, 0 to 21%), seroconverted to one or more of the gangliosides after their initial infection. This rate was similar to that observed in the placebo recipients, 8.3% (95% CI, 0 to 24%), following no intervention. All volunteers that exhibited a serological response at any time point (day 0, 21, and/or 28) are included in Table 3. In general, the samples that tested positive for the presence of antiganglioside antibodies had only weak responses and were mostly directed against GM₁/asialo-GM₁ and GM₂ gangliosides, although some volunteers exhibited serological responses against a wider variety of gangliosides. One positive sample, that from volunteer 3 taken at day 0, had weak antibody responses to eight of nine gangliosides tested. Similarly, another five of the eight volunteers who tested positive (volunteers 1, 2, 4, 7, and 8) had positive samples at day 0 prior to any intervention. Hence, these responses were not induced by an exposure to either vaccine or infection. Furthermore, the occurrence of an antiganglioside response at the subsequent time points of 21 or 28 days in volunteers 7 and 8 was consistent with the presence of such antibodies on day 0. Nevertheless, for the three remaining plasma samples (volunteer 5, day 28; volunteer 6, days 21 and 28) antiganglioside serological responses (anti-GM₁, -asialo-GM_1, -GD_1b, and -GM₂) were observed which were absent on day 0 (one in the placebo group and two from the initial infection group, respectively). Importantly, none of the volunteers developed GBS.

TABLE 3.

Seropositive reactions among plasma samples (n = 183) of the 61 volunteers to gangliosides^a

Volunteer	Group	Day of sampling	Strength of reactionc with ganglioside:
			Asialo-GM1	GM1	GD1a	GD1b	GM2	GD2	GD3	GT1b	GQ1b
GBSb			++++	++	−	++	(+)	−	−	−	−
1	Vaccine	0	−	−	−	−	+	−	−	−	−
2	Experimental infection (rechallenge)	0	−	+	−	−	(+)	−	−	+	−
3	Placebo	0	+	+	+	+	+	+	+	−	+
4	Placebo	0	−	(+)	−	−	+	−	−	−	−
5	Placebo	28	−	+	−	−	−	−	−	−	−
6	Experimental infection (initial)	21	−	−	−	−	+	−	−	−	−
		28	(+)	−	−	(+)	+	−	−	−	−
7	Experimental infection (initial)	0	−	+	−	−	+	−	−	−	−
		28	−	+	−	−	+	−	−	+	−
8	Experimental infection (initial)	0	+	++	+	+	++	+	−	++	−
		21	+	+	+	+	++	−	−	+	−
		28	−	−	−	−	+	(+)	−	(+)	−

^aPlasma samples (n = 183) were screened using TLC immunostaining to detect antiganglioside antibodies at day of infection (day 0) and postinoculation (days 21 and 28) in all volunteers. Only volunteers with positive serologic reactions after screening are shown.

^bControl serum containing anti-GM₁ antibodies from a patient with Guillain-Barré syndrome.

^c++++, very strong reaction; +++, strong reaction; ++, moderate reaction; +, weak reaction; (+), very weak reaction; −, no reaction.

Therefore, to better assess these postinfection observations, four volunteers who showed antiganglioside antibodies at one or more of these test time points (volunteers 2, 6, 7, and 8) and three others who did not exhibit antibodies at these time points (volunteers 9, 10, and 11) were selected for additional testing. Plasma samples from these seven volunteers taken at days 0, 3, 7, 10, 21, 28, and 60 (n = 49 samples) were assayed. Table 4, which summarizes the results of this testing, shows that the most frequent antibody responses were observed against GM₂, GM₁, GD_1a, GD_1b, asialo-GM₁, and GT_1b. Volunteers 2, 7, and 8 had weak or moderate responses to GM₁ and GM₂ on day 0, but this response disappeared in volunteer 2 by day 21. In these three volunteers, antibodies to these and other gangliosides were weak and transient or, if persistently present at later time points (e.g., 21 to 60 days), had been previously detectable at day 0, indicating the serological responses were not induced by the infection. Although plasma samples from volunteers 6, 9, 10, and 11 had no antiganglioside antibodies on day 0, a weak transient response to GM₁ and GM₂ gangliosides developed in volunteers 9, 10, and 11, which was absent by day 21. In volunteer 6, although a weak response to GM₂ was observed even at day 60, only a weak transient response to GM₁ occurred. No further responses to the remaining gangliosides of the test panel were observed in volunteers 10 and 11, and the limited responses to other gangliosides in volunteers 6 and 9 were transient. Furthermore, using a procedure for optimal coating of gangliosides on ELISA plates (5), we formulated an antiganglioside ELISA which gave comparable results to those obtained with the TLC-immuno-overlay technique for these seven volunteers (data not shown). Collectively, with the exception of one volunteer (volunteer 6, anti-GM₂ antibodies) the results demonstrate an absence of induction of persistent antiganglioside antibodies after experimental infection with C. jejuni 81-176.

TABLE 4.

Serologic reactions of plasma samples (n = 49) from selected C. jejuni experimentally infected volunteers (n = 7) at additional time points^a

Volunteer	Day of sampling	Strength of reaction with gangliosideb:
		Asialo-GM1	GM1	GD1a	GD1b	GM2	GD2	GD3	GT1b
2	0	−	+	−	−	(+)	−	−	+
	3	−	+	−	+	+	−	−	+
	7	+	++	+	+	+	−	+	+
	10	+	++	+	+	+	+	+	+
	21	−	−	−	−	−	−	−	−
	28	−	−	−	−	−	−	−	+
	60	−	−	−	−	−	−	−	−
6	0	−	−	−	−	−	−	−	−
	3	+	++	+	+	+	−	−	−
	7	−	−	−	−	−	−	−	−
	10	−	+	−	−	++	−	−	−
	21	−	−	−	−	+	−	−	−
	28	(+)	−	−	(+)	+	−	−	−
	60	−	−	−	−	+	−	−	−
7	0	−	+	−	−	+	−	−	−
	3	−	+	−	−	+		−	−
	7	−	−	−	−	−	−	−	−
	10	−	+	−	−	+	−	−	+
	21	−	−	−	−	−	−	−	−
	28	−	+	−	−	+	−	−	+
	60	−	+	−	−	−	−	−	−
8	0	+	++	+	+	++	+	−	++
	3	+	−	−	−	+	−	−	(+)
	7	(+)	+	(+)	−	++	−	−	+
	10	−	+	−	−	+	−	−	(+)
	21	+	+	+	+	++	−	−	+
	28	−	−	−	−	+	(+)	−	(+)
	60	−	+	−	−	++	(+)	−	(+)
9	0	−	−	−	−	−	−	−	−
	3	−	++	+	+	+	+	−	−
	7	−	−	−	−	−	−	−	−
	10	+	+	(+)	−	+	−	+	−
	21	−	−	−	−	−	−	−	−
	28	−	−	−	−	−	−	−	−
	60	−	−	−	−	−	−	−	−
10	0	−	−	−	−	−	−	−	−
	3	−	−	−	−	−	−	−	−
	7	−	+	−	−	+	−	−	−
	10	−	−	−	−	−	−	−	−
	21	−	−	−	−	−	−	−	−
	28	−	−	−	−	−	−	−	−
	60	−	−	−	−	−	−	−	−
11	0	−	−	−	−	−	−	−	−
	3	−	+	−	−	+	−	−	−
	7	−	+	−	−	+	−	−	−
	10	−	−	−	−	+	−	−	−
	21	−	−	−	−	−	−	−	−
	28	−	−	−	−	−	−	−	−
	60	−	−	−	−	−	−	−	−

^aPlasma samples (n = 49) were screened using TLC immunostaining to detect antiganglioside antibodies at the day of infection (day 0) and postinoculation (days 3, 7, 10, 21, 28, and 60) in the seven selected volunteers; four volunteers (2, 6, 7, and 8) had shown antiganglioside antibodies in initial screening at day 0, 21, or 28 (Table 3), and three other volunteers (9, 10, and 11) did not. Additional testing was performed on plasma samples from these seven volunteers taken at days 0, 3, 7, 10, 21, 28, and 60 (n = 49 samples).

^bIn addition to testing for reaction with the gangliosides listed, plasma samples were screened for reaction with GQ_1b but no reactions were observed. Symbols: ++, moderate reaction; +, weak reaction; (+), very weak reaction; −, no reaction.

DISCUSSION

C. jejuni 81-176 is a very well characterized strain, was originally isolated from the feces of a 9-year-old girl with diarrhea (11), and has been shown to induce diarrhea in human volunteers (1). In the present work, studies were undertaken arising from the completion of a volunteer experimental oral infection study with the C. jejuni 81-176 reference strain (Tribble et al., Abstr. 10th Int. Congr. Immunol.).

Initially, LOSs from the C. jejuni 81-176 reference strain and six recovered isolates from volunteers were characterized. To determine the ganglioside mimicry occurring in C. jejuni LOSs, serologic probing was used since small amounts of purified LOSs were available, particularly from recovered isolates, which were sufficient for serologic probing but not complete structural analysis. Structural studies have established that C. jejuni 81-176 LOS expresses mimicry of predominantly GM₂ and GM₃ gangliosides, but also to a minor extent of GD₂ and GD_1b gangliosides (10; A. P. Moran et al., unpublished data). Antiganglioside antibodies and ganglioside-specific ligands did not react with the purified LOS of the reference strain, and collectively, the results indicated the absence of mimicry by the major gangliosides implicated in GBS development. Compared with the structural data, the lack of reaction of C. jejuni 81-176 LOSs with anti-GD₂ antibodies is likely attributable to the low proportion of molecules with this type of mimicry in the LOSs of the reference strain. Although the anti-GM₂ antibodies reacted with control GM₂ ganglioside, the lack of reaction of these antibodies with reference strain LOSs contrasts with the structural findings of predominant expression of GM₂ and GM₃ mimicry. However, these antibodies did not react with another control C. jejuni LOS bearing GM₂ ganglioside mimicry (serostrain HS:23), identical to the observations with reference strain LOSs. These results reflect the importance of epitope presentation for detection of ganglioside-like epitopes in C. jejuni LOSs by antiganglioside antibodies and are consistent with previous findings that the core sugars in C. jejuni LOS, on which the ganglioside-mimicking structures are carried, can influence epitope recognition (19, 20).

Remarkably, in contrast to the absence of serologically detectable mimicry by the major gangliosides in the reference strain, after in vivo passage the isolates possess a Neu5Acα2-3Gal or Neu-5Acα2-8-Neu5Aca2-3Gal epitope in the form of GM₂/GD₂/GD_1b mimicry. Hence, the C. jejuni LOS underwent antigenic phase variation during passage in vivo. Importantly, this is the first report of in vivo phase variation by C. jejuni LOS in humans. Consistent with these findings, a previous in vitro study has shown that C. jejuni 81-176 possesses the genetic mechanisms for its LOS to undergo phase variation, resulting in expression of GM₂ and GM₃ ganglioside-like structures but also in GD_1b and GD₂ ganglioside mimicry (10). Furthermore, comparative sequence analysis of the cgtA gene (which encodes a phase-variable N-acetylgalactoaminyltransferase involved in synthesis of GM₂/GD₂/GD_1b structures) from individual colonies of the recovered isolates and the reference strain indicated the occurrence of homopolymeric tracts whose length is consistent with an actively expressed enzyme (10) in all colonies of the recovered isolates examined (A. P. Moran, unpublished results). Also, our comparative electrophoretic analyses of the reference strain and isolates showed differences in LOS banding patterns similar to those observed previously in phase-variable LOSs (10).

Subsequently, the serologic response to a panel of gangliosides in plasma samples (n = 183) obtained at the day of intervention (day 0), vaccination or infection, and 21 and 28 days postintervention was assessed with the TLC-immuno-overlay technique. Overall, only 12 of 183 (6.6%) of the plasma samples tested from eight volunteers were positive for the presence of antiganglioside antibodies, irrespective of intervention group. Of the subjects testing positive, six volunteers had antibodies on day 0 of the trial, and thus the serologic responses in these volunteers were not induced by the intervention. Evidence of seroconversion was observed in 2 of 22 (9.1%) volunteers in the initial infection challenge group, comparable to 1 of 12 (8.3%) in placebo recipients, following no intervention.

Further testing of seven selected volunteers at additional time points of 0, 3, 7, 10, 21, 28, and 60 days (n = 49 samples) showed that when antiganglioside antibodies occurred (mostly anti-GM₁ and -GM₂), responses were weak and transient. Only one volunteer showed a weak response to GM₂ even at day 60, and with the exception of this volunteer, the results showed an absence of persistent antiganglioside antibodies after CWC vaccination or experimental infection with C. jejuni 81-176. Furthermore, testing of plasma samples from these seven selected volunteers in an antiganglioside ELISA confirmed these findings. Nevertheless, we cannot exclude the possibility that because of the rare incidence of GBS a much larger group of volunteers would be required for antiganglioside antibodies to be detected. Also, knowledge of the kinetics and the quantity of the antiganglioside response to induce GBS is lacking from the scientific literature, and thus, at present, it is not possible to define what quantity of a response would induce GBS in a susceptible host. Furthermore, although immunogenetic factors may play a role in the pathogenesis of GBS, efforts to demonstrate an association between human leukocyte antigen (HLA) types and the development of C. jejuni-associated GBS have been inconclusive (21). Thus, the requirements for an individual to be susceptible to GBS development after C. jejuni infection remain to be identified since HLA type has not been consistently correlated with C. jejuni-associated GBS nor other host factors identified.

Phase variation of LOS structures has been shown in mucosal pathogens such as Neisseria meningitidis, Neisseria gonorrhoeae, and Haemophilus species (22) and results in the heterogeneity of core oligosaccharide length, variation in sugar composition, and variability in linkages and sugar substitutions (27). The ability of C. jejuni to undergo phase variation in LOS biosynthesis genes has been demonstrated in vitro (7, 8, 10, 13). It is evident that one C. jejuni strain has the ability to produce a repertoire of LOS molecules, and this variability of cell glycolipids may provide an advantage during the course of C. jejuni infection. In fact, insertional inactivation of cgtA in C. jejuni 81-176, which results in loss of an N-acetylgalactosamine residue, thereby producing a truncated LOS core structure resembling GM₃ ganglioside, resulted in a significant increase in invasion of intestinal epithelial cells in vitro (10). Also, variation in sialylation of C. jejuni LOS affects the immunogenicity and serum resistance of C. jejuni (9). Hence, current data suggest that many strains of C. jejuni can synthesize ganglioside mimics (15, 21) and have the ability to convert between different LOS structures (7, 8, 10, 12, 13). Although this would appear to have serious consequences for the development of a successful vaccine, particularly using an attenuated approach, in our study the development of ganglioside mimicry in the reference strain did not necessarily lead to the induction of antiganglioside antibodies in the experimentally infected volunteers. The use of inactivated C. jejuni strains lacking evidence of ganglioside mimicry or subunits of Campylobacter not associated with GBS pathogenesis are vaccine development concepts that circumvent the potential of in vivo phase variation concerns and potential GBS risk. On the other hand, an alternative would be to inactivate the genes involved in LOS core extension in C. jejuni 81-176 to yield a strain in which ganglioside mimicry is no longer expressed. Finally, in addition to the presence of ganglioside mimicry by C. jejuni, it is evident that other contributions of the host are necessary to trigger pathogenic antiganglioside antibody production, as suggested previously (21).