Campylobacter jejuni infection is a leading cause of acute bacterial gastroenteritis worldwide 1. The widespread dissemination of C. jejuni is largely attributed to its transmission to humans by consumption of contaminated food sources frequently colonized by the bacteria. Clinical symptoms vary from mild to severe gastroenteritis to more complex and serious extraintestinal diseases, including the neuropathic disorders Guillain-Barré syndrome (GBS) or its variant Miller Fisher syndrome (MFS) 2.

GBS and MFS are acute neuropathies thought to result from a transient humoral immune response against host gangliosides in peripheral nerves 3. GBS is the most common form of acute flaccid paralysis with incidence rates of up to 4 per 100, 000 4. MFS, a variant of GBS, is characterized by acute ophthalmoplegia and ataxia. The onset of these syndromes is often preceded by infectious illness and C. jejuni is the most frequent infectious agent. 30% of GBS cases and 20% of MFS cases are preceded by an infection of C. jejuni within three weeks prior to the onset of neurological symptoms 567

A number of studies have sought to examine the population structure of GBS/MFS strains in the search for genetic commonalities that could account for a neuropathogenic phenotype. Initial surveys showing an over-representation of serotypes HS:19 and HS:41 among GBS strains 891011, raised the possibility that GBS strains comprise a clonal lineage with unique virulence factors associated with GBS. For example, a high proportion of HS:19 isolates was obtained from GBS patients in countries such as Japan, despite the low prevalence of this serotype among Japanese enteritis cases, suggesting that enteritis patients infected with a strain of HS:19 serotype have an increased risk of developing GBS 8. A similar association has been observed between Japanese MFS-related C. jejuni strains and the HS:2 serotype 10. Subsequent studies, however, have shown substantial genetic heterogeneity in other collections of neuropathogenic strains 121314.

Lipo-oligosaccharide (LOS) is one of the most important cell-surface structures expressed by C. jejuni, and strains associated with neuropathies express ganglioside-like LOS structures 1516. The development of GBS/MFS following C. jejuni infection is thought to be related to molecular mimicry between ganglioside-like moieties on certain Campylobacter LOS classes and ganglioside epitopes on neural tissue 17, with cross-reacting anti-LOS antibodies ultimately leading to nerve damage 18. The majority of patients with GBS subsequent to C. jejuni enteritis develop autoantibodies that react to GM1 or GD1a gangliosides 1920 whereas MFS patients develop anti-GQ1b antibodies 521. The LOS from GBS- and MFS-associated C. jejuni have been shown to induce anti-GM1 and anti-GQ1b antibodies in rabbits 22. Moreover, sensitization of rabbits with GM1-like LOS of C. jejuni isolated from a GBS patient has resulted in a disease model of GBS 23.

Among the three classes of C. jejuni LOS locus (A, B, and C) initially characterized by Gilbert et al. 24, the majority of HS:19 isolates harbour a Class A LOS locus, a gene cluster implicated in the expression of ganglioside mimics 2526. The A-class LOS carries the cst-II gene which, first isolated from the GBS-associated strain OH4384, encodes a bifunctional sialyltransferase capable of transferring sialic acid to either a terminal galactose residue or to a terminal sialic acid residue, resulting in linkages that lead to the production of ganglioside mimics 27. Taken together, both cst-II and the Class A locus are currently the strongest known determinants of GBS 2526282930 Recent studies using knockout mutants of C. jejuni and a mouse model have demonstrated the necessity of cst-II and of a related sialylation pathway gene (orf10 or neuA1, encoding a CMP-NeuAc synthetase) in the induction of anti-ganglioside antibodies 25.

Although anti-LOS cross-reactive antibodies are a major component of the development of C. jejuni induced GBS and MFS, the complete mechanism is not fully understood. Similarly, the extent to which microbial and/or host factors contribute to the development of an anti-glycolipid response and neurological symptoms remains a point of debate 31. Cases have been described in which GBS patients with C. jejuni infections fail to display ganglioside sero-reactivity, raising the possibility that other peripheral nerve antigens are the targets in these patients 32. Furthermore, ganglioside-like structures have been found in C. jejuni isolates from enteritis patients without GBS or MFS 33. Recent work aiming to characterize the LOS locus of GBS/MFS-associated C. jejuni strains also suggests that there is a strong, but incomplete, correlation between neuropathogenic strains and an A or B-class LOS locus 252629. Although neuropathogenic strains that carry a non-A/B locus may be less prevalent, their existence raises the possibility that other factors (pathogen and/or host-related) may contribute to the onset of neurological complications. We have performed a systematic search for additional pathogen-related GBS/MFS-associated factors by performing a detailed genetic comparison of strains isolated from GBS/MFS patients, and from uncomplicated cases of enteritis (i.e. "enteritis-control strains"), by means of microarray-based Comparative Genomic Hybridization (CGH). We present here the results of this comprehensive comparative genomic survey.

Current available evidence points to the direct involvement of genes that synthesize and transfer sialic acid to the LOS in the development of the cross-reactive anti-ganglioside antibodies thought to be the effectors in a large majority of GBS and MFS cases 2526. Although the association between neuropathogenic LOS genotypes and the GBS/MFS-associated phenotype is very strong, strains that bear the requisite genes for the synthesis of ganglioside mimics have been isolated from uncomplicated cases of enteritis. Similarly, a small number of GBS/MFS-associated strains do not appear to synthesize ganglioside mimics. Thus, the incomplete penetrance of the neuropathogenic LOS genotype has raised questions regarding the possible contribution of additional factors, whether bacterial or host-related, towards the development of these neuropathies. That the host contributes to such an auto-immune response should be considered likely.

A consistent finding among various genetic surveys of C. jejuni has been that GBS strains do not appear to represent a genetic lineage distinct from enteritis-control strains 1214383940. Studies attempting to correlate molecular typing results to the GBS/MFS phenotype did not reveal clusters of neuropathogenic strains distinct from enteritis-control strains, and results from our study suggest that many neuropathogenic strains are genomically related to enteritis-control strains. Some of the major clusters in the dataset include strains from more than one of the three major geographical regions represented in the study. For example, some HS:19 strains from The Netherlands and Curaçao cluster with, and are genetically similar to, strains from the Japanese HS:19 cluster. Similarly, HS:2 strains from The Netherlands cluster with strains from the two Japanese HS:2 lineages. It thus appears that the genomic stability previously suggested among clonal HS:19 strains of differing geographical source 3841 is also valid among HS:2 strains, as has been suggested in other studies 42.

While cluster analysis is prominent in this study, the fact that neuropathogenic strains do not form a coherent cluster, and the fact that genes related to neuropathogenesis are expected to represent only a small fraction of the data, expose the shortcomings of the use of cluster analysis to define potential neuropathogenic markers. Because of this, in contrast to previous CGH-based studies 3940, we opted to focus our search for neuropathogenic markers on the statistical testing of each individual gene present in the array for significant over-representation or under-representation among neuropathogenic strains. Using this novel approach, we were able to obtain unambiguous statistical signal for higher conservation rates in genes associated with LOS classes A and B among neuropathogenic strains 252943. These results not only agree with the current hypotheses regarding the mechanism of GBS/MFS but also concur with a recent study in which GBS/MFS strains and enteritis-control strains were screened for potential neuropathogenic markers using the high resolution comparative genomic method of high-throughput Amplified Fragment Length Polymorphism (htAFLP) 44. This study revealed 3 markers highly-associated with GBS which mapped to the LOS locus.

The strong association observed for LOS classes A and B and neuropathogenic strains in our data was found despite the extensive genomic heterogeneity in these strains. A recent study has provided direct evidence for horizontal transfer of genes, including putative neuropathogenic factors in an experimental setting 45. Our findings and the recent description of strain GB11, a GBS-associated HS:2 strain with high genetic similarity to the genome strain NCTC 11168 but found to have an A-class LOS locus instead of a C-class locus36 would suggest that horizontal transfer represents an important mechanism for the dissemination of neuropathogenic factors between otherwise unrelated GBS/MFS strains.

Among the known neuropathogenic genes identified in our screen cgtA and neuA1 are known to be involved in LOS biosynthesis 24. Another gene (orf11) has been found to be associated with class A/B LOS loci 25, and it was recently shown to encode a sialic acid acetyltransferase 46. Genes in the LOS locus needed for the synthesis of ganglioside mimics are strong GBS/MFS determinants 252629 but neuropathogenic strains with no known sialyltransferase genes are known to exist 25. This raises the possibility, among others, that additional bacterial factors are required to elicit neurological complications. Even though we examined every gene in our microarray for similar associations with GBS/MFS strains, after filtering the data to avoid over- or under-representation due to clonality effects, only two additional genes (Cj1421c, Cj1428c or fcl) had statistically significantly higher conservation rates among neuropathogenic strains compared to enteritis-control strains; and a single gene (HS:3 Cj1135) had a higher conservation rate among enteritis-control strains. Cj1135 is a one-domain glucosyltransferase involved in LOS biosynthesis47, while Cj1421c and Cj1428c are involved in capsule biosynthesis although their exact functions are still undefined 48. The potential involvement of these three genes in neuropathogenesis, if any, needs to be assessed. At the same time, the potential involvement of additional bacterially-encoded factors should not be discounted.

The microarray used to generate the bulk of the data in this study includes known GBS-associated factors and comprises greater gene diversity than that used in a previously published study of GBS strains 39. However additional genes important to neuropathogenicity could be missing from our array. This study represents the largest systematic screen for potential neuropathogenic factors in C. jejuni but anything less than a comparative approach involving full-genome sequences is only partially complete.

One of the key findings of this study is the close genetic relationship between some neuropathogenic strains and their enteritis-associated counterparts. This has been suggested by results from various molecular typing studies, and we have been able to observe these similarities with an extremely high level of resolution. Since several scenarios could help explain the high degree of genetic similarity observed between strains with different clinical outcomes, it is important to note that highly related strains can show major differences in virulence-associated phenotypes 49. The LOS locus also presents a unique challenge in that diversity in LOS structures can be obtained through genetic variation affecting the relevant genes 24. Enteritis-associated strains that carry neuropathogenic genes could have their neuropathogenic potential altered or silenced through mutation, and this mechanism is likely to play a significant role in the incomplete penetrance of the putative GBS/MFS genotype.

Our data suggest that in many cases GBS/MFS-related strains might not differ in their neuropathogenic potential with respect to highly genetically related enteritis-control strains. In some cases, differences in clinical outcome are likely to be attributable to differences in host-background. GB13 and GB14, two epidemiologically related strains isolated from a family outbreak in which only one of three individuals afflicted with enteritis went on to develop neurological complications 31, serve as a reminder of how host factors are likely to play a role in the development of neuropathic clinical outcomes. At the same time we have recently shown that GB11, a GBS strain with a close genetic relationship to the genome strain NCTC 11168, appears to have acquired potential GBS factors in the form of an A-class LOS locus 36. That study clearly underscores the valuable insight that could be gained by comparing closely related strains with differing clinical outcomes in the discovery of potential neuropathogenic factors. Future efforts should be aimed at comparative-genomic sequencing of strain pairs such as the ones described in this study in order to address whether differences in pathogen or host are responsible for differences in the clinical outcome of Campylobacteriosis cases.

GBS, Guillain-Barré syndrome;

MFS, Miller Fisher syndrome;

LOS, Lipo-oligosaccharide;

CGH, comparative-genomic hybridization.

The author(s) declares that there are no competing interests.

ENT designed M-CGH experiments, carried out downstream data analysis, and drafted the manuscript. RRA performed hybridizations, performed preliminary data analysis and assisted with downstream data analysis. MK characterized the Japanese strains. PCRG characterized the Dutch and Curaçao strains. JHEN, MG, and AvB conceived of the study, and participated in its design and coordination and helped to draft the manuscript. HPE, and NY participated in the conception and supervised the design of the study. All authors submitted comments on drafts and read and approved the final manuscript.