Serotyping of Campylobacter jejuni based on heat stable antigens: relevance, molecular basis and implications in pathogenesis

Authors


Dr A. P. Moran, Department of Microbiology, National University of Ireland, University Road, Galway, Ireland (e–mail: anthony.moran@nuigalway.i.e).

1. Introduction

It has been recognized since the 1970s that Campylobacter jejuni, previously named Vibrio fetus and Camp. fetus subsp. jejuni, is a major cause of enteric infections in humans and animals ( Skirrow 1977; Penner 1988). Campylobacter jejuni is the most common bacterial cause of diarrhoea in many industrialized countries and is consequently responsible for a major public health and economic burden ( Ackerley & Jones 1985; Anon 1995; Buzby et al. 1997 ). Also, some strains of the species are associated with the development of diseases of the peripheral nervous system in humans ( Kuroki et al. 1991 ; Mishu & Blaser 1993; Rees et al. 1995 ) which further adds to economic burden ( Buzby et al. 1997 ). Studies on the pathogenesis of Camp. jejuni have yielded insights into the molecular mechanisms by which this bacterium causes disease ( Walker et al. 1986 ; Penner 1988; Ketley 1997), and may lead to the development of new methods for detecting and controlling the spread of this bacterium. Numerous studies of Camp. jejuni infections have been carried out and the epidemiology is now well understood, due, in large measure, to the application of one or both of two systems for serotyping clinical isolates ( Penner & Hennessy 1980; Lior et al. 1982 ). Other strain typing methods have been developed and have met with more limited success. Phage typing has been developed, but used in only a limited number of studies on Camp. jejuni ( Grajewski et al. 1985 ; Salama et al. 1990 ; Patton et al. 1991 ). Similarly, DNA typing, based on total genomic digest patterns from conventional ( Owen & Beck 1987; Bruce et al. 1988 ; Owen et al. 1989 ; Hernandez et al. 1991b ; Patton et al. 1991 ) and pulsed-field gel electrophoresis ( Yan et al. 1991 ; Suzuki et al. 1994 ), ribosomal RNA gene restriction digest patterns ( Geilhausen et al. 1990 ; Owen et al. 1990 ; Hernandez et al. 1991a , b; Patton et al. 1991 ; Fayos et al. 1992, 1993 ; Tee et al. 1992 ) and plasmid profiling ( Tenover et al. 1984 ; Fayos et al. 1993 ) have been applied to a limited extent. The polymerase chain reaction has been used to obtain randomly amplified polymorphic DNA (RAPD) profiles, which provide an additional and apparently discriminatory method of typing Camp. jejuni strains ( Mazurier et al. 1992 ; Fayos et al. 1993 ; Giesendorf et al. 1994 ; Fujimoto et al. 1997 ), and to perform restriction fragment length polymorphism analysis ( Nishimura et al. 1996 ; Fujimoto et al. 1997 ).  The serotyping system which differentiates strains on the basis of heat-labile (HL) antigens was developed according to classical methods used for identifying flagellar antigens of Salmonella, Escherichia coli and other Gram-negative species ( Lior et al. 1982 ). This system depends on the absorption of antisera with heated bacterial suspensions of the homologous strains and the heterologous cross-reacting strains to remove antibodies directed against HS antigens. The traditional methods of bacterial cell agglutination are used to identify the specificities of the HL antigens. Progress in elucidating the biochemical characteristics of these antigens has been stimulated by interest in mechanisms of pathogenesis and host cell invasion. The antigens have been asserted to be flagellar proteins and studies on the genetics of flagella are underway to determine if this is the case ( Guerry et al. 1992 ; Ketley 1997).  In a departure from the traditional approach, the passive haemagglutination (PHA) technique, formerly used primarily for research on endotoxins ( Neter et al. 1956 ), was introduced in 1980 as the procedure for routine serotyping ( Penner & Hennessy 1980). The system differentiates strains on the basis of soluble heat-stable (HS) antigens and has been used successfully in numerous laboratories for epidemiological studies. In some investigations it was the only system used and in other studies it was used in conjunction with the HL typing system (e.g. Penner et al. 1983 ; Jones et al. 1985 ; Kaijser & Sjögren 1985; Kuroki et al. 1993 ; Patton et al. 1993 ; Lastovica et al. 1997 ). A renewed interest in the use of the HS antigen typing system has arisen since HS antigens of particular serotypes have been implicated as possible pathogenic factors in the development of Guillain–Barré syndrome (GBS) and Miller–Fisher syndrome (MFS) ( Kuroki et al. 1993 ; Yuki et al. 1994 ; Moran et al. 1996a ; Salloway et al. 1996 ; Lastovica et al. 1997 ). The HS antigens have been referred to by such names as ‘Pen’ (for Penner) and ‘O’ to reflect the assumption that they are lipopolysaccharides (LPSs), but in this brief review, the original designation of ‘HS’ will be used.  The discovery of neuropathic strains of Camp. jejuni has given urgency to research into determining the molecular structures of the HS antigens, particularly those of strains implicated in neuropathogenesis. Progress to date indicates that the HS antigen is of a novel type, incorporating features of both enterobacterial high molecular weight (-Mr) LPSs and the non-enterobacterial low-Mr lipooligosaccharides (LOSs) as shown in Fig. 1 ( Rietschel et al. 1990 ; Moran 1995a; Preston et al. 1996 ). Rapid progress has been made in gaining new insights into the nature of the unique HS antigen and in this review, attention is focused on the present state of knowledge of the biochemistry and structure of the HS antigens and their possible role in pathogenesis.

Figure 1.

Schematic representation of the general structure of (a) high-M▵ LPS and (b) low-M▵ LOS

2. The HS antigen serotyping system

2.1 Brief history of early work

In the first attempts to classify strains of Campylobacter species using both slide and tube agglutination tests, three antigens, believed to be O antigens (designated O serotypes A, B, and C), were recognized ( Berg et al. 1971 ). With these procedures, 10 isolates that would currently be classified as either Camp. jejuni or Camp. coli were all found to belong to serotype C. Other investigators of Campylobacter serotyping ( Ristic & Brandly 1959; Winter 1966; Newsam & St. George 1967; Bokkenheuser 1972) used the PHA technique which is known to be effective for identifying enterobacterial LPS antigens ( Neter et al. 1956 ).

In an initial trial to define HS antigens of the bacteria then known as Camp. fetus subsp. jejuni, only weak cross-reactive and non-reproducible agglutinations were observed in bacterial agglutination tests and, therefore, the PHA technique was evaluated for serotyping ( Penner & Hennessy 1980). As in the earlier studies ( Ristic & Brandly 1959; Winter 1966; Newsam & St. George 1967; Bokkenheuser 1972), the use of the PHA technique and typing antisera (produced in rabbits against bacterial cell suspensions), enabled the differentiation of strains on the basis of HS antigens.

In the initial study, 23 HS antigens were identified in Camp. fetus subsp. jejuni (Penner and Hennessy l980). With the advantage of the hippurate hydrolysis test ( Harvey 1980) for distinguishing Camp. jejuni from Camp. coli, the serotype reference strains (serostrains) were separated to produce a system for each species. The systems have been extended to include 42 serostrains for Camp. jejuni and 18 for Camp. coli (Penner et al. l983; Mills et al. l991, 1992). Tests using hybridization of DNA confirmed the separation of the species on the basis of the hippurate hydrolysis test and other bacteriological tests ( Leaper & Owen 1982). These serotyping systems have been used in reference laboratories to investigate the epidemiology of enteric infections and, in particular, the system for Camp. jejuni has also become useful in studying strains associated with the development of GBS and MFS ( Kuroki et al. 1993 ; Yuki et al. 1994 ; Lastovica et al. 1997 ; Nachamkin 1997; Penner & Aspinall 1997), diseases of the peripheral nervous system ( Fisher 1956; Hughes 1990).

2.2 The PHA technique

Passive haemagglutination (PHA) has been known for decades as a highly sensitive technique for examining the specificities of LPS ( Neter et al. 1956 ). Briefly, extracted LPS molecules adhere to the surface of mammalian erythrocytes, a process referred to as sensitization, and antibodies specific for LPS cause agglutination of the sensitized erythrocytes. As the erythrocyte is much larger than the bacterial cell, the detection of an antigen-antibody reaction is more sensitive with the PHA technique than with techniques employing bacterial cell agglutination. The basis for the sensitization is the presence in the erythrocyte membrane of a 228 kDa lipoglycoprotein receptor that binds LPS but not homopolysaccharides, heteropolysaccharides, monosaccharides or protein antigens of either Gram-negative or Gram-positive bacteria ( Springer et al. 1973 ).

Like LPS molecules, the HS antigens of Camp. jejuni and Camp. coli that sensitize erythrocytes may be extracted by the hot phenol-water method ( Westphal & Jann 1965) by treating bacterial suspensions with ethylenediaminetetraacetate (EDTA) according to the method of Leive et al. (1968) , or by heating saline suspensions of the bacteria ( Neter et al. 1956 ). Although the titres of the PHA reactions observed with products of the various extraction procedures vary from one product to the next, the antigenic specificities are virtually identical, indicating that the different extraction procedures yield essentially the same antigenic material ( Penner & Hennessy 1980; Preston & Penner 1987). For routine serotyping, the method of heating cell suspensions is used primarily because it is the least labour-intensive method of extracting antigenic material and is more practical in the general microbiology laboratory. The sensitized erythrocytes are washed three times to remove any less firmly bound heat-labile bacterial debris. Thus, the washings serve as a purification step to ensure that the serological reaction is restricted to that between the HS antigen and the antibodies against the HS antigen. Conversely, as only the antibodies directed against the HS antigen participate in the reaction, there is no requirement for costly and labour-intensive absorptions of antisera to remove antibodies against other bacterial components.

There continues to be interest in simplifying the serotyping system by eliminating the PHA step. In adopting alternative techniques, such as coagglutination ( Kosunen et al. 1982 ; Wong et al. 1985 ; Fricker et al. 1986 ; Illingsworth & Fricker 1987) or agglutination in microtitration plates ( Frost et al. 1998 ), it is necessary to emphasize that in these procedures there is no equivalent of the erythrocyte LPS receptor to select HS antigen preferentially from the antigenic preparation. In these alternative systems, measures must be taken to provide either purified HS antigens or purified immune reagents, such as monoclonal antibodies, or multiply absorbed polyclonal antisera to achieve reaction specificity ( Moran & Kosunen 1989).

2.3 The serotyping procedure

As routinely performed in most laboratories, the serotyping procedure consists of dispensing suspensions of heat-extracted, antigen-sensitized sheep or human erythrocytes into wells of microtitration plates containing antiserum in twofold dilutions. The plates are incubated overnight at 37 °C and the highest dilution of the antiserum in which haemagglutination occurs is recorded as the titre ( Penner & Hennessy 1980; Penner et al. 1983 ). Dilution of the antisera may be performed manually or with automated titration equipment. The titration step is omitted in the rapid slide agglutination technique; in this technique, the antiserum is added directly to sensitized and thrice-washed erythrocytes on a slide, which is subsequently gently rocked for 5 min, to observe the occurrence of haemagglutination ( Mills et al. 1991 ). A major advantage of these procedures is that only the heat-extracted antigenic material is required for serotyping by the reference laboratory. The shipment of extracts eliminates the problem of viability loss during transport and reduces the hazard associated with shipment of live cultures ( Lastovica et al. 1986 ).

3. Biochemical characteristics of the HS antigens

3.1 General properties suggestive of a glycolipid nature

The observation that the HS antigens of Camp. jejuni adhere to erythrocyte membranes and can be extracted by procedures specific for LPS ( Penner & Hennessy 1980) indicates that they are similar to the LPSs of Enterobacteriaceae or the LOSs of Neisseria and Haemophilus ( Rietschel et al. 1990 ; Moran 1995a; Preston et al. 1996 ). The HS antigens are unlikely to be proteins as these are generally thermolabile and would not withstand the heat treatment of the HS antigen extraction procedure. Moreover, the inability to stain phenol-water extracts with Coomassie blue (a protein stain) indicates a lack of protein in the HS preparations ( Preston & Penner 1987). Also, proteins require coupling reagents such as tannic acid, glutaraldehyde or bis-diazotized benzidine to sensitize erythrocytes ( Boyden 1951; Onkelinx et al. 1969 ). Nor are the HS antigens likely to be typical lipid-free capsular polysaccharides which are unable to adsorb to erythrocytes without coupling reagents such as chromium chloride, cyanogen bromide or stearoyl chloride ( Hämmerling & Westphal 1967; Baker et al. 1969 ; Ammann & Pelger 1972; Eriksen 1973; Richter & Kågedal 1974).

However, Chart et al. (1996) could not demonstrate the reaction of antisera against serostrains with homologous high-Mr LPS in immunoblotting and suggested that capsular polysaccharides were the HS antigens involved in serotyping. The inability to demonstrate a reaction with LPS in the former study may have been due to technical problems as Jones et al. (1984) inhibited PHA reactions of HS antigens with purified LPSs. Furthermore, these latter investigators showed, using eluates from electrophoresis gels of heat-extracted antigen preparations, that LPSs are involved in serotyping Camp. jejuni by PHA, whereas other bacterial cell surface components had no activity.

Neveretheless, to test the possibility that polysaccharide antigens are involved in the PHA system of serotyping Camp. jejuni, extracts were prepared by methods used to isolate lipid-containing capsular polysaccharides of Neisseria meningitidis ( Liu et al. 1971 ; Gotschlich et al. 1981 ). The presence of polysaccharides was demonstrated, but they were found to have the same antigenic specificities as the HS antigen and purified LPS, and to contain 2-keto-3-deoxyoctulosonic acid (Kdo), an 8-carbon acidic sugar unique to LPSs and LOSs ( Moran & Kosunen 1989). The authors concluded that the polysaccharides were derived from LPS as a result of the polysaccharide extraction process employed, and evidence for capsular components like those of N. meningitidis was not obtained.

3.2 Low-M▵ molecules that confer serotypic specificity

Further evidence that the HS antigens of Camp. jejuni are of the LPS or LOS type molecule was obtained in experiments in which both phenol-water extracts ( Westphal & Jann 1965) and proteinase K-treated whole-cell lysates ( Hitchcock & Brown 1983) of serostrains were examined by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) with LPS-specific silver staining ( Tsai & Frasch 1982; Mills et al. 1985 ; Preston & Penner 1987, 1989). With these procedures, only low-M▵ components, ranging in M▵ from 4500 to 5000, were detected for each of the 38 strains examined ( Preston & Penner 1987). Similar results were obtained by other investigators (Logan and Trust l984; Perez-Perez & Blaser 1985).

The presence of only low-M▵ components implied that the HS antigens are OSs that resemble either the LOSs that occur in Haemophilus and Neisseria ( Inzana et al. 1985 ; Mandrell et al. 1986 ; Moran et al. 1996b ), or the cores of rough enterobacterial mutant strains ( Rietschel et al. 1990 ). However, the observation that these short molecules display a wide diversity of antigenic specificities, a feature uncharacteristic of enterobacterial rough strains, indicated that they are more closely related to the LOSs.

Investigators studying the chemical composition of the HS antigens suggested that they were LOSs ( Adeyeye et al. 1989 ; Conrad & Galanos 1990). Further evidence supportive of this suggestion was the finding that three Camp. jejuni serostrains for serotypes HS:1, HS:2 and HS:4 possess sialic acid (N-acetylneuraminic acid, Neu5Ac)–containing oligosaccharides (OS) in the low-M▵ molecules ( Moran et al. 1991a ). As terminally linked Neu5Ac is a feature characteristic of the LOSs of Neisseria and Haemophilus ( Moran et al. 1996b ; Preston et al. 1996 ), the view that the HS antigens of Camp. jejuni are LOSs has gained support.

3.3 HS antigens with both high-M▵ and low-M▵ components

Unexpected, however, was the finding that not all of the Camp. jejuni strains have HS antigens that conform completely in structure to the LOS molecule. When electrophoresed extracts of HS antigens were immunoblotted with homologous antisera, 16 of 38 serostrains were shown to have, in addition to the low-M▵ components, a series of high-M▵ components in ladder-like patterns typical of LPS molecules with O chains ( Preston & Penner 1987, 1989). This new evidence indicated that the species consists of two classes of strains; one class with HS antigens consisting only of low-M▵ components, and the second class with HS antigens that have both low-M▵ and high-M▵ components. Examples of reactions of six serostrains are shown in Fig. 2. Although it has been reported that high-M▵ components could be visualized by silver staining under certain conditions ( Blake & Russell 1993), this has not been confirmed by other investigators. Instead, re-electrophoresis of high-M▵ material from LPS gels and LPS radiolabelling experiments have indicated that high-M▵ bands are aggregates of low-M▵ LPS ( Logan & Trust 1984; Mills et al. 1985 ).

Figure 2.

SDS-PAGE gel of proteinase K digests of six Campylobacter jejuni serostrains were (a) silver-stained and (b) immunoblotted with homologous antisera. Lanes: 1, serostrain HS:17; 2, serostrain HS:44; 3, serostrain HS:52; 4, serostrain HS:12; 5, serostrain HS:21; 6, serostrain HS:33. Note the ladder-like banding in lanes 4, 5 and 6 of panel (b)

Immunoblotting of electrophoresed extracts with both homologous and heterologous antisera also showed that both the low-M▵ and high-M▵ components are strain-specific antigens ( Preston & Penner 1987, 1989; Moran & Kosunen 1989). These unanticipated complexities clearly indicated that neither the typical LPS nor LOS molecule constitutes a model that accommodates the structure of the HS antigen of Camp. jejuni.

4. Molecular structure of the HS antigens

4.1 Summary of progress in brief

The HS antigens of Camp. jejuni and Camp. coli reflect, in some cases, the two component structure of LOSs composed of lipid A and core OS ( Moran et al. 1996b ; Preston et al. 1996 ) and, in others, the three component LPSs composed of lipid A, core OS and O-polysaccharide chain ( Rietschel et al. 1990 ; Moran 1995a) ( Fig. 1), except for the greater diversity in the structure of their core OS. It is not surprising therefore that much attention has focused on determining the structures of the HS antigens because of the interest in investigating the basis for the unique combination of components from these two distinctly different bacterial glycolipids.

Progress to date includes the complete determination of the structure of lipid A of one strain, Camp. jejuni serostrain HS:2 ( Moran et al. 1991b ). The molecular structures of the HS antigens have been determined for serostrains HS:1, HS:2 and HS:3 that show only low-M▵ components in immunoblots ( Aspinall et al. 1993a , b, 1995). The structures of the HS antigens have also been determined for serostrains HS:19, HS:23 and HS:36 that have both high-M▵ and low-M▵ components ( Aspinall et al. 1992a ; 1993b; 1994b,c). Only the low-M▵ component of serostrain HS:4 has been analysed structurally ( Aspinall et al. 1993b ).

In addition, the structures of both low-M▵ and high-M▵ molecules of five serotype HS:19 isolates from patients with the GBS have been determined ( Yuki et al. 1993 ; Aspinall et al. 1994b , c; Moran & O’Malley 1995). The structure of the low-M▵ molecules of a serotype HS:41 isolate from a GBS patient ( Prendergast et al. 1998 ), and of a serotype HS:10 isolate and a serotype HS:23 isolate from patients with MFS, have been determined ( Salloway et al. 1996 ; Aspinall et al. 1998 ). Although the structures of only a small number of HS antigens have been determined, the results have led to new insights into unique structural variations previously unknown in Gram-negative bacteria. The structures are described briefly in the following sections.

4.2 Lipid A

The lipid A of LPS and LOS molecules from diverse bacterial species adhere to a similar structural principle in which occur three essential elements, namely, a d-hexosamine disaccharide backbone, substituted or non-substituted phosphate groups linked to the backbone, and fatty acid chains bound to the disaccharide backbone. Structural variations from one bacterial species to another may occur in one or more of these three components, but the basic structural principle is essentially maintained ( Rietschel et al. 1990 ; Zähringer et al. 1994 ; Moran 1995a).

In Camp. jejuni, however, there is more variation than in most Gram-negative species, whereby three different lipid A backbone disaccharides occur. In Camp. jejuni serostrain HS:2, approximately 73% of the lipid A molecules have a disaccharide backbone of diaminoglucose (2,3-diamino-2,3-dideoxy- d-glucose) and d-glucosamine; 15% have backbones in which both sugars in the disaccharide are diaminoglucose, and 12% have disaccharides of two glucosamine residues, which is the composition most typical of disaccharide backbones in enterobacterial lipid A. All three disaccharide backbones are phosphorylated and acylated in the same manner ( Moran et al. 1991b ; Moran 1997a).

The occurrence of both glucosamine and diaminoglucose is a common property of all Camp. jejuni serostrains examined ( Moran et al. 1991a ; Moran 1995b). Preliminary structural analysis of lipid A from different Camp. jejuni serostrains has confirmed the presence of the same type of backbone structures identical to those encountered in lipid A of Camp. jejuni serostrain HS:2, but the relative proportions of the various backbones differ ( Moran 1993). Thus, minor interstrain variation in Camp. jejuni lipid A may occur ( Moran 1997a).

4.3 Structures of the OSs of low-M▵ molecules

In examining the structures of the core OSs of eight Camp. jejuni HS antigens of serostrains HS:1, HS:2, HS:3, HS:4, HS:10, HS:19, HS:23 and HS:36 ( Fig. 3), it is apparent that two distinctly different regions exist in each OS ( Aspinall et al. 1992b ; 1993a,b; 1994a,c, 1995; Nam Shin et al. 1998 ). Linked to lipid A is the inner OS region which includes a trisaccharide of Kdo and two heptoses. The heptose adjacent to Kdo is substituted by glucose in all OSs examined whereas the second heptose is substituted by glucose only in the OSs of reference strains for serotypes HS:1 and HS:2 ( Aspinall et al. 1993a , b). The only other variation in the inner OS region is the phosphate or phosphoethanolamine substituent at position 6 of the first heptose.

Figure 3.

Structures of the core oligosaccharides of Campylobacter jejuni serostrains HS:1 (a), HS:2 (b), HS:3 (c), HS:4 and HS:19 (d), HS:23 and HS:36 (e). Abbreviations: Gal, galactose; Glc, glucose; GalNAc, N-acetylgalactosamine; ldHep, l-glycero- d-manno-heptose; Neu5Ac, N-acetylneuraminic acid (sialic acid); Kdo, 3-deoxy- d-manno-octulosonic acid; PEA, phosphorylethanolamine; QuiNac, quinovosamine (3-amino-3,6-dideoxy- d-glucose)

The outer regions of the core OSs are much more variable in structure ( Fig. 3). Each consists of either two or three hexoses which are terminally or laterally substituted with hexoses, sialic acid (Neu5Ac) or, more rarely, with quinovosamine (QuiNAc). Differences among the OSs in sequences of the hexoses, variations in the anomeric configurations of the hexoses, and variations in the lateral and terminal substituents, provide further heterogeneity. These variations provide the serotypic diversity that enables differentiation of strains that have HS antigens with only low-M▵ components ( Aspinall et al. 1992b ).

It should be noted that the core OSs of serostrains HS:4 and HS:19 are structurally identical ( Aspinall et al. 1993b ; 1994c)(see Fig. 3). Similarly, the OSs of serostrains HS:23 and HS:36 are identical ( Aspinall et al. 1993b ) but different from those of serostrains HS:4 and HS:19 ( Aspinall et al. 1993b ; 1994a,c). From these results, it is evident that a wide variety of core OS structures are present in this bacterial species, but also that strains of different serotypes may share the same OS structure. Strains with identical OSs are differentiated on the basis of their OS-linked serotype-specific polysaccharide polymers contained in the high-M▵ components of serotypes HS:4, HS:19, HS:23 and HS:36 that are described in a subsequent section.

4.4 Molecular mimicry of human ganglioside saccharide moieties by OSs

Except for the OS of serotype 3 ( Moran et al. 1991a ; Aspinall et al. 1995 ; Moran 1995b), the OSs of all other serotypes for which the molecular structures have been determined contain one to three residues of sialic acid linked terminally and/or laterally to residues of galactose ( Fig. 3). The presence of the sialic acid residues makes the terminal structures of the OSs similar to the terminal regions of several human gangliosides ( Moran et al. 1996a , b). The OSs of serostrains HS:1, HS:23 and HS:36 ( Fig. 3a,e) have identical terminal structures that mimic the saccharide of the ganglioside GM2 ( Aspinall et al. 1993b ), whereas the OSs of serotypes HS:4 and HS:19 ( Fig. 3c) mimic the sacharride of ganglioside GD1a ( Aspinall et al. 1993b ; 1994c). Shorter molecules lacking the terminal sialic acid were also found in the low-M▵ components of both serotype HS:4 and HS:19. These monosialo OSs mimic the GM1 ganglioside ( Moran et al. 1996a , b). The terminal disaccharide (Neu5Acα2→3Gal) of the OS of the serostrain HS:2 ( Fig. 3b) is present as the terminal disaccharide of ganglioside GM4, and is also present as the terminal disaccharide in other gangliosides including GD1a, GT1b and GM3 ( Aspinall et al. 1993a ; Moran et al. 1996b ).

The molecular mimicry of human gangliosides and the implications for Camp. jejuni pathogenesis have been reviewed elsewhere ( Moran et al. 1996a , b). In brief, Camp. jejuni LPS can induce antiganglioside antibodies in animals ( Schwerer et al. 1995 ; Ritter et al. 1996 ), and GBS and MFS patients with antecedent Camp. jejuni enteritis frequently have autoantibodies to gangliosides during the acute phase of illness ( Yuki et al. 1990 ; Oomes et al. 1995 ; Schwerer et al. 1995 ; Neisser et al. 1997 ). Furthermore, anti-GM1 antibodies recognize epitopes at the Nodes of Ranvier and presynaptic end-plates ( Thomas et al. 1991 ; Illa et al. 1995 ) and can suppress voltage-sensitive Na+ currents and, thus, potentially interfere with the nerve impulse ( Waxman 1995).

4.5 Characteristics of the HS antigen of serotype HS:3

The OS of serostrain HS:3 ( Fig. 3c) is markedly different from those of the other serostrains in that it does not contain sialic acid. It also has an unusual sugar, quinovosamine (3-acetamido-3,6-dideoxy-β–d-glucose), linked laterally as a side chain. A Camp. jejuni strain of serotype HS:57 and serostrain HS:3 are the only ones in the species in which the quinovosamine residue has been detected to date ( Beer et al. 1986 ; Aspinall et al. 1995 ). Although no high-M▵ components were visualized after immunoblotting the electrophoresed phenol-water extract from serostrain HS:3 with homologous antiserum ( Preston & Penner 1987), on chemical analysis, the extract was found to contain a polysaccharide polymer. The polymer consists of repeating units of [→4)-α- d-galactose-(1→3)-{(3-hydroxypropanoyl)-L-α-ido-heptose}-(1→]n but has no detectable heptose units, which indicates a lack of linkage to the core OS ( Aspinall et al. 1995 ). Of all the serotypes for which the HS antigen structure has been determined, the core OS of this serotype is the only one that does not mimic known structures of human tissue components. Thus, extracts of HS antigen and LPS from this strain can serve as negative controls in experiments with antiganglioside antibodies ( Schwerer et al. 1995 ).

Like most polymeric polysaccharides, the polymer of the HS:3 serostrain is most likely to be thermostable, although tests to determine its susceptibility to heat have not been undertaken. Its serological reactivity and specificity have not been established, either through experiments in which electrophoresed extracts are immunoblotted with homologous antiserum, or by the PHA technique used in serotyping. The inability to visualize high-M▵ components of the polymer in immunoblotted gels of electrophoresed extract is attributable to the lack of a lipid moiety that is required to enable migration in SDS-PAGE gels that have been designed for analysis of LPS ( Tsai & Frasch 1982; Preston & Penner 1987). Failure to identify an antigenic specificity for the polysaccharide with the PHA technique is consistent with results of numerous investigators who have shown that polysaccharides are not natural haemosensitins, but require coupling reagents for attachment to erythrocyte membranes ( Hämmerling & Westphal 1967; Baker et al. 1969 ; Ammann & Pelger 1972; Richter & Kågedal 1974). Conversely, as the established HS serotyping scheme of Camp. jejuni uses the PHA technique, it can be deduced that this polysaccharide does not contribute to the serospecificity of the scheme.

It is conceivable that the serotype HS:3 polymer is a capsular constituent, but the inability to detect lipid in the extract prepared for structural analysis ( Aspinall et al. 1995 ), and the inability to extract the polysaccharide with procedures designed to isolate lipid-terminated capsular components from N. meningitidis ( Moran & Kosunen 1989), suggests, but does not prove, that the serotype HS:3 polymer lacks the terminal lipid moiety that is a characteristic feature of N. meningitidis capsules ( Gotschlich et al. 1981 ). Its lack of sialic acid, however, clearly indicates that it is not a capsular polysaccharide of the type present in N. meningitidis ( Liu et al. 1971 ; Jann & Jann 1987). The determination of the nature of the polymer and its location in the cell requires morphological studies at the ultrastructural level. The discovery of the polymer is significant, however, because it provides evidence that carbohydrate-based antigens, other than the HS antigens identified by PHA, are present in Camp. jejuni.

4.6 Structures of HS antigens that have O-polymers

Immunoblotting of electrophoresed phenol-water extracts from serostrains HS:4, HS:19, HS:23 and HS:36 that produce high-M▵ components migrating in ladder-like patterns typical of enterobacterial LPS, indicate that these strains have HS antigens with O-polymers, in addition to low-M▵ molecules composed of only lipid A and OS ( Mills et al. 1992 ; Preston & Penner 1987). Chemical analysis has confirmed the presence of such polymers in serostrains HS:19, HS:23 and HS:36 and the structures of these polymers have been elucidated ( Aspinall et al. 1992a ; 1994a,b). To date, the structure of the serotype HS:4 polymer has not been determined.

The impetus for determining the structure of the HS antigen of serotype HS:19 was the discovery in Japan that enteric infections with strains of this serotype were associated with subsequent development of GBS ( Kuroki et al. 1991, 1993 ). The serotype HS:19 polymer is linked to the core OS and consists of disaccharide repeating units of β–d-glucuronic acid amidated with 2-amino-2-deoxyglycerol linked to N-acetylglucosamine ( Fig. 4a); it reflects the structure of the disaccharide of hyaluronic acid ( d-glucuronic acid and N-acetylglucosamine) ( Aspinall et al. 1994a , b). A potential role of this polymer in pathogenesis is discussed in a following section.

Figure 4.

Figure 4.

Structures of the repeat units of oligosaccharide-linked polysaccharide polymers of serostrain HS:19 (a) and structures of the heptoses in the repeat units of serostrains HS:23 and HS:36 (b)

Figure 4.

Figure 4.

Structures of the repeat units of oligosaccharide-linked polysaccharide polymers of serostrain HS:19 (a) and structures of the heptoses in the repeat units of serostrains HS:23 and HS:36 (b)

In the scheme for serotyping Camp. jejuni, it was recognized that serostrains HS:23 and HS:36 cross-react, indicating identity or shared antigenic specificities ( Penner et al. 1983 ; Preston & Penner 1987). In the course of serotyping, some isolates were found to react only in antiserum 23, others only in antiserum 36, and some reacted in both antisera. To determine the basis of these reactions, the HS antigens of both reference strains were subjected to chemical analysis ( Aspinall et al. 1992a ; 1993b). The OSs of the two strains are chemically indistinguishable ( Fig. 3e), but the polymers of these serotypes were found to be singularly unusual in complexity. After cleaving the Kdo-lipid A bond, degradation of the high-M▵ fraction of serostrain HS:36 extract yields residues of N-acetylglucosamine, galactose, and four closely related heptoses. Compositional and structural data ( Aspinall et al. 1992a ) have established the existence of trisaccharide repeating units consisting of N-acetylglucosamine, galactose and one of four heptoses, [→3)-β- d-N-acetylglucosamine-(1→3)-α- d-galactose-(1→2)-heptose]n ( Fig. 4b). The results obtained do not preclude the unexpected, but possible, occurrence of polysaccharides with heterogeneous units ( Aspinall et al. 1992a ), but can also be interpreted to provide evidence for the occurrence of four O-polymers with repeating units characteristic of the O-polysaccharides of enterobacterial LPS. The synthesis of more than one O-polymer by a Gram-negative strain has been reported previously (Rivera et al. l988; Lam et al. 1992 ), but the occurrence of multiple structurally distinct units based on variations in structure of a single heptose is a novel phenomenon.

In the extract of serostrain HS:23, three repeating units identical to three from serostrain 36 were identified ( Aspinall et al. 1992a ). The absence from serotype HS:23 of a fourth repeating unit [GlcNAc→Gal→d-glycero-α- d-altro-Hep]n found in serotype HS:36 may be due to the inability of serotype HS:23 to synthesize the heptose, or to its presence at a concentration in the extract too low to be detected by the analytical procedure.

The cross-reactions between serostrains HS:23 and HS:36, and the reactions of isolates that type in both serotype HS:23 and HS:36 antisera, can be accounted for by antibodies against the common repeating units of the polymers ( Aspinall et al. 1992a ). Strains producing a polymer containing the repeating unit [GlcNAc→Gal→d-glycero- d-altro-Hep]n would be expected to react in antiserum 36 but not in antiserum 23 if, in fact, it can be confirmed that serotype HS:23 does not produce this repeating unit. However, the basis for the reactions of strains that react only in antiserum 23 is not clear and further investigations of such strains are required. Investigations should also focus on the possibility that controlled synthesis to produce polymers in varying proportions provides an advantage to these bacteria in resisting the host immune defences.

The linkage of the serotype HS:23 and HS:36 polysaccharide polymers to the terminal N-acetylgalactosamine of the core OSs has been suggested to be via a disaccharide consisting of N-acetylglucosamine and galactose ( Aspinall et al. 1992a , b). The establishment of the linkage site, the visualization of ladder-like banding patterns in electrophoresed and immunoblotted phenol-water extracts, and the proven correlations between serotyping and immunoblotting results, confirm the view that the HS antigens of Camp. jejuni are O-polysaccharide chains attached to core OS-lipid A moieties in LPS, and are not unassociated polysaccharides of the type identified in serotype HS:3.

5. HS antigens of isolates from patients with neurological diseases

Campylobacter jejuni strains of HS serotype 19 were discovered in 1991 to be associated with the development of GBS ( Kuroki et al. 1991 ). Recent reports have confirmed this finding and it has been noted that isolates of other serotypes are also identified with this syndrome ( Yuki et al. 1994 ; Lastovica et al. 1997 ; Nachamkin 1997; Penner & Aspinall 1997). However, serotype HS:19 was most frequently implicated, and the serostrain and two isolates (Camp. jejuni OH 4382 and Camp. jejuni OH 4384) of the same serotype from patients with GBS were selected for detailed study.

Comparison of immunoblots of electrophoresed phenol-water extracts of the serostrain with the two patient isolates showed that serotype HS:19 antiserum reacted with both the low-M▵ and high-M▵ components of each strain as expected ( Aspinall et al. 1994a ). Unexpected, however, was the observation that for one isolate, Camp. jejuni OH 4382, the banding patterns of the low-M▵ component indicated a core OS lower in molecular weight than the OS of the reference strain. The banding patterns of the low-M▵ components of the other isolate, Camp. jejuni OH 4384, and the serostrain were approximately the same. The high-M▵ components of the three strains produced ladder-like banding with similar spacings, but the bands of the isolates were displaced with respect to the bands of the serostrain. These results indicated that the isolates had core OSs of structure different from the OS of the serotype strain, but linked to identical O-polymers ( Aspinall et al. 1994a , b, c).

Prompted by the unanticipated results of the immunoblotting tests, chemical and structural analysis of the patient isolates were undertaken. They confirmed that the OSs of the two patient isolates were markedly different in structure from the OS of the serostrain ( Aspinall et al. 1994a , c). The OS of Camp. jejuni OH 4384 differs in that a disaccharide, neuraminobiose [Neu5Acα(2→8)Neu5Ac], instead of a single residue of sialic acid, is terminally linked to the galactose ( Fig. 5a). As predicted from the immunoblotting results, the core OS of Camp. jejuni OH 4382 is a shorter molecule ( Fig. 5b). It lacks the terminal galactose and N-acetylgalactosamine, but a neuraminobiose disaccharide is linked to an inner galactose ( Aspinall et al. 1994c ). Both patient isolates thus have a trisaccharide terminus of Neu5Ac→Neu5Ac→Gal which is the same structure that occurs in the terminal saccharide of the human ganglioside GD3. Furthermore, the outer OS region of Camp. jejuni OH 4384, which extends from the inner galactose to the terminal neuraminobiose, mimics the ganglioside GT1a ( Aspinall et al. 1994c ).

Figure 5.

Core oligosaccharide structures of Campylobacter jejuni HS:19 isolates from GBS patients, OH 4384 (a) and OH 4382 (b), and of a Camp. jejuni HS:10 isolates from a MFS patient, PG 836 (c). For abbreviations see Fig. 3.

In the immunoblot of the HS antigen of Camp. jejuni PG 836, a serotype HS:10 isolate from a patient with MFS, the migration patterns indicated the presence of a polymer linked to the core OS ( Salloway et al. 1996 ). The molecular structure of the OS has been determined ( Nam Shin et al. 1998 ) and was shown to have a terminal trisaccharide (Neu5Ac→Neu5Ac→Gal) identical in structure to those of the serotype HS:19 isolates from the GBS patients described above ( Fig. 5c). As the trisaccharide is present in the OSs of the three patient isolates but absent in the core OSs of seven serostrains, it suggests a special role for the trisaccharide in the pathogenesis of neurological disease ( Salloway et al. 1996 ). Moreover, preliminary analysis suggests that this same trisaccharide is present in the HS antigen of a serotype HS:23 MFS isolate which exhibits mimicry of various gangliosides, including that of GD3, in its core OSs ( Aspinall et al. 1998 ).

The synthesis of the trisaccharide requires an OS structure with a terminal β- d-galactose to accept the first Neu5Ac at the O-3 position and also an enzyme [α-(2, 3)-sialyltransferase] specific for this reaction. Another enzyme [α-(2,8)-sialyltransferase] is required to link the second Neu5Ac to the first Neu5Ac to complete the trisaccharide. The trisaccharide structure is identical to the terminal saccharide of ganglioside GD3, but the Neu5Acα(2→8)Neu5Ac disaccharide occurs either as a terminus or as a lateral side chain in several other gangliosides, including the ganglioside GQ1b which has been identified in an isolate of serotype HS:2 from a patient with MFS ( Yuki et al. 1994 ).

From these results it is evident that the α-(2, 8)-sialyltransferase is an important factor in the synthesis of structures that mimic the gangliosides of neuropathic strains. The results also suggest that the enzyme could serve as a biomarker to differentiate strains that have a potential to become associated with the subsequent development of neurological diseases from those that are more limited in their pathogenesis and cause only enteric infections. Greater understanding of the role of the α– (2,8)-sialyltransferase in the pathogenesis of neurological disease is expected to be forthcoming from future research on its activation, regulation, and on the basis for its acquisition by particular strains of Camp. jejuni.

Isolates of serotype HS:19 have been reported to occur in enteritis cases at a frequency of 1·6% ( Penner et al. 1983 ), but only a small fraction of these have been implicated in the development of GBS. To gain insights into differences between strains associated with the development of GBS and those that cause enteric infections only, isolates from both groups have been compared ( Moran & O’Malley 1995). An isolate (strain 98) from a patient who acquired only the enteric infection had a mixture of two OSs that mimic the structures of GM1 and GD1a identical to the core OSs of the serostrain HS:19. An isolate (strain 157) from a GBS patient was found to have a single core OS that mimics the structure of the monosialoganglioside GM1 and a second isolate (strain 101), also from a GBS patient, had one OS structure which mimicked the structure of the disialoganglioside GD1a. The absence of the neuraminobiose disaccharide (Neu5Ac2→8Neu5Ac) in the phenol-water extracts of these two patient isolates indicates the lack of the trisaccharide (Neu5Ac2→Neu5Ac2→Galβ). This is evidence that strains of serotype HS:19 lacking the trisaccharide may also be associated with the development the development of GBS. Supporting this, preliminary analysis indicates the occurrence of GM1 mimicry in the core OS of a serotype HS:41 isolate from a GBS patient ( Prendergast et al. 1998 ).

Analysis of the molecular structures of the O-polysaccharide polymers of the serotype HS:19 isolates from the GBS patients, and from a patient who suffered only an enteric infection ( Aspinall et al. 1994b ), showed that they are identical to the structure of the polymer from serostrain HS:19 ( Fig. 4a). The finding that Camp. jejuni strains of the same serotype may have antigenically identical polysaccharide polymers but structurally different core OSs, as seen in strains of serotype HS:19, has important implications in studying pathogenic mechanisms. Investigations based on the assumption that Camp. jejuni isolates of the same serotype will have identical core OSs mimicking the same gangliosides could be misleading. Structural analyses of HS antigens of isolates of other serotypes are necessary to determine the prevalence in Camp. jejuni of serotypes that include multiple OS structures. From the data currently available it cannot be ascertained if the HS antigen of strains of serotype HS:19 generally reflects the bacterial species in this respect, or if the multiple OSs are a unique feature of this particular serotype. Also important would be more information on structures of OSs of serotype HS:19 isolates that are infectious, but not associated with subsequent development of GBS.

It has been suspected that the serotype HS:19 O-polysaccharide polymer could also have a role in pathogenesis. The rationale for this hypothesis is twofold. From an epidemiological perspective it is curious that strains of serotype HS:4 are rarely isolated from GBS patients, although the core OSs of this serotype are the same as those of serotype HS:19 and can mimic the same gangliosides ( Aspinall et al. 1993b ). The major difference between the two serotypes is the polysaccharide polymer and thus, the polysaccharide of the serotype HS:19 has been implicated in pathogenesis. Secondly, the serotype HS:19 polymer resembles hyaluronic acid, suggesting the possibility of another case of molecular mimicry. This observation is of importance as hyaluronic acid occurs in human connective tissue and is also the major component in capsules of group A haemolytic streptococci, where it is believed to have an important role as an antiphagocytic factor. Although the serotype HS:19 polymer is not identical in structure to hyaluronic acid ( Aspinall et al. 1994a ), the possibility that it may serve as a contributing factor in pathogenesis merits consideration.

Experimental evidence in support of a role for the serotype O-polysaccharide polymer in virulence has been obtained ( Moran & O’Malley 1995). A significantly greater amount of the polysaccharide is present in strains from GBS patients than in strains from patients who develop enteritis alone. In the case of Camp. jejuni, complexity, not simplicity, may be the final outcome of investigations on the basis of its pathogenicity in various neuropathies.

6. Nomenclature and classification of the HS antigens

In serotyping, the typable strain is assigned the serotype of the reference strain against which the antiserum was produced. This provides a numerical designation that does not convey information on the nature of the HS antigen, but it does meet the practical requirements of the epidemiologist for whose discipline serotyping systems are produced. Authors have not agreed on the biochemical nature of the Camp. jejuni HS antigen and various designations have been used. As evidence for the presence of lipid A, core OS and O-polysaccharide (in some strains) became available, the practice of referring to the HS antigens as O-antigens has become popular. The controversy over the introduction of the LOS designation (see Preston et al. l996) for the glycolipids of the mucosal pathogens may also have been influential in favouring the O designation. However, from recent research on the HS antigens, it has become clear that the present nomenclature and classification schemes for LPSs and LOSs do not adequately embrace the unique features of Camp. jejuni HS antigens. Potential alternate forms of nomenclature and classification should recognize within the serotype designation the contributions to serological specificity of both the core OS and the O-polysaccharide components of the HS antigen.

7. Questions and issues

As in many other aspects of Camp. jejuni, its HS antigens have also provided a number of interesting surprises. There are now convincing lines of evidence that the HS antigen is a prominent bacterial glycolipid analogous to the LPS of enteric bacteria and the LOS of non-enteric mucosal pathogens, but with significant differences from each of them to be considered as a different, but related glycolipid. Through evolution the HS antigen in some strains has retained (or acquired) the polysaccharide polymer analogous to the O-antigen of LPS, but has outer core OS regions displaying structural diversity resembling those of LOSs.

The genetic basis for the production and variation of these molecules is only now becoming the focus of attention. In part aided by the availability of data on the LOS genetic locus of Campylobacter hyoilei for gene screening ( Korolik et al. 1997 ), a putative LPS biosynthetic locus of Camp. jejuni containing a range of genes and showing variation between strains has been identified ( Ketley 1997; Fry et al. 1998 ; Wood et al. 1998 ) and promises to yield new insights. In particular, genes potentially involved in biosynthesis of amino sugars including Neu5Ac, genes encoding sugar transferases, and genes controlling O-antigen-like polysaccharide synthesis, core OS synthesis and transport of core OS biosynthetic intermediates across the inner membrane have been identified ( Fry et al. 1998 ; Wood et al. 1998 ). However, the functions of these genes have been deduced on the basis of sequence homology to LPS/LOS genes of other bacteria. Confirmation of their putative roles await the construction of relevant mutants and subsequent chemical characterization of the HS antigens. In addition, the availability of a more detailed physical and genetic map of Camp. jejuni ( Karlyshev et al. 1998 ), as well as the future availability of the complete genome sequence of Camp. jejuni, should aid these biosynthetic investigations. Nevertheless, as experienced with other bacterial species whose complete genome sequence is available, in order to exploit the sequence data fully, a general mutagenesis system will be required for Camp. jejuni ( Moran 1997b). In general, the knowledge gained of the genes controlling HS antigen biosynthesis may aid the development of DNA based typing methodologies, particularly of Camp. jejuni strains expressing sialylated structures and mimicry of gangliosides in their HS antigen.

Interesting new insights into the nature of the HS antigen have been gained, but the results have also raised a number of questions and new issues have evolved. A critical issue in Camp. jejuni pathogenesis is the importance of molecular mimicry relative to other potentially important contributing factors, including host factors, in the development of post-infection neuropathies ( Moran et al. 1996a ). Much current research is directed towards determining the role of molecular mimicry in the pathogenesis of these neurological diseases and, for successful outcomes, more information on the HS antigens will be crucial. A wide range of human gangliosides are mimicked ( Moran et al. 1996a , b), but it is not clear which of these mimicries are significant and which have no role or only a minor role in establishing post-infectious disease. The numbers of strains studied at the molecular level is small in comparison with the number of strains of different serotypes that have been implicated in the development of neurological syndromes ( Nachamkin 1997; Penner & Aspinall 1997). The question arises as to whether too little attention is being directed towards strains of serotypes other than those of serotype 19. If molecular mimicry is the basis of neuropathogenesis by Camp. jejuni, there are likely to be more interesting surprises in store for the curious investigator of the unknown HS structures of other serotypes. The immunological mechanisms by which tolerance to host ganglioside structures are diminished by Camp. jejuni infection also need to be determined ( Moran & Prendergast 1998). Campylobacter jejuni is primarily an agent infective for the gastrointestinal tract and queries ought to be directed at the role of the unique HS antigen in the pathogenesis of infectious disease. At present, only limited data on the toxicity, immunological activity and biological properties of this antigen of Camp. jejuni are available ( Blaser et al. 1986 ; McSweegan & Walker 1986; O’Sullivan et al. 1988 ; Moran 1995b) and further investigations should prove fruitful.

Acknowledgements

The authors thank Dr G.O. Aspinall and M. Prendergast for discussions during the preparation of this review, and the many collaborators who contributed to the studies described in this review. The financial support of the Irish Health Research Board is gratefully acknowledged.

Ancillary